Hepatitis C virus induces a prediabetic state by directly impairing hepatic glucose metabolism in mice

Virus-related type 2 diabetes is commonly observed in individuals infected with the hepatitis C virus (HCV); however, the underlying molecular mechanisms remain unknown. Our aim was to unravel these mechanisms using FL-N/35 transgenic mice expressing the full HCV ORF. We observed that these mice displayed glucose intolerance and insulin resistance. We also found that Glut-2 membrane expression was reduced in FL-N/35 mice and that hepatocyte glucose uptake was perturbed, partly accounting for the HCV-induced glucose intolerance in these mice. Early steps of the hepatic insulin signaling pathway, from IRS2 to PDK1 phosphorylation, were constitutively impaired in FL-N/35 primary hepatocytes via deregulation of TNFα/SOCS3. Higher hepatic glucose production was observed in the HCV mice, despite higher fasting insulinemia, concomitant with decreased expression of hepatic gluconeogenic genes. Akt kinase activity was higher in HCV mice than in WT mice, but Akt-dependent phosphorylation of the forkhead transcription factor FoxO1 at serine 256, which triggers its nuclear exclusion, was lower in HCV mouse livers. These findings indicate an uncoupling of the canonical Akt/FoxO1 pathway in HCV protein-expressing hepatocytes. Thus, the expression of HCV proteins in the liver is sufficient to induce insulin resistance by impairing insulin signaling and glucose uptake. In conclusion, we observed a complete set of events leading to a prediabetic state in HCV-transgenic mice, providing a valuable mechanistic explanation for HCV-induced diabetes in humans.


Virus-related type 2 diabetes is commonly observed in individuals infected with the hepatitis C virus (HCV)
; however, the underlying molecular mechanisms remain unknown. Our aim was to unravel these mechanisms using FL-N/35 transgenic mice expressing the full HCV ORF. We observed that these mice displayed glucose intolerance and insulin resistance. We also found that Glut-2 membrane expression was reduced in FL-N/35 mice and that hepatocyte glucose uptake was perturbed, partly accounting for the HCV-induced glucose intolerance in these mice. Early steps of the hepatic insulin signaling pathway, from IRS2 to PDK1 phosphorylation, were constitutively impaired in FL-N/35 primary hepatocytes via deregulation of TNF␣/SOCS3. Higher hepatic glucose production was observed in the HCV mice, despite higher fasting insulinemia, concomitant with decreased expression of hepatic gluconeogenic genes. Akt kinase activity was higher in HCV mice than in WT mice, but Akt-dependent phosphorylation of the forkhead transcription factor FoxO1 at serine 256, which triggers its nuclear exclusion, was lower in HCV mouse livers. These findings indicate an uncoupling of the canonical Akt/FoxO1 pathway in HCV protein-expressing hepatocytes. Thus, the expression of HCV proteins in the liver is sufficient to induce insulin resistance by impairing insulin signaling and glucose uptake. In conclusion, we observed a complete set of events leading to a prediabetic state in HCV-transgenic mice, providing a valuable mechanistic explanation for HCV-induced diabetes in humans.
Hepatitis C virus (HCV) 4 infects over 170 million individuals worldwide. Although newly developed treatment combinations based on direct-acting antiviral drugs cure a high proportion of infections (1,2), the vast majority of HCV-infected patients do not have access to these therapies because they are unaware of their infection, do not have access to a proper healthcare system, or cannot afford the high cost of treatment.
Chronic HCV infection is not only a leading cause of chronic liver disease, including cirrhosis and hepatocellular carcinoma, but also induces systemic disorders. In particular, HCV infection was reported to be an independent risk factor for type 2 diabetes (T2D), regardless of the presence of cirrhosis (3)(4)(5). Indeed, a meta-analysis showed a 1.8-fold excess risk of T2D among HCV-positive patients compared with HBV-positive/ HCV-negative ones (6). In a longitudinal study, the risk of developing diabetes was up to 12 times higher in HCV-infected patients than in the general population (7). In patients with chronic hepatitis C, T2D is an independent predictor of a more rapid progression of liver fibrosis (8) and impairs the response to antiviral treatments based on PEGylated interferon ␣ and ribavirin (9,10). In addition, patients with cirrhosis and T2D have an increased risk of developing hepatocellular carcinoma (11,12). In patients with compensated HCV-related cirrhosis, diabetes and marked insulin resistance are independently associated with higher liver morbidity and mortality (13).
T2D results from a combination of mechanisms, including insulin resistance (IR), resulting in increased hepatic glucose production and altered insulin secretion. Several arguments suggest that HCV is pro-diabetogenic per se (for a review, see Ref. 14,15). Indeed, IR persists in patients who do not respond to antiviral treatment despite a decrease in body mass index (16). Although still controversial, a correlation between IR and the HCV RNA level, a surrogate marker of viral replication, has been reported (17), and the incidence of IR is higher in patients infected with genotypes 1 and 4 than in those infected with genotypes 2 or 3 (18,19). This genotype specificity of HCVinduced IR could result from amino acid sequence differences across genotypes in the core sequence (20,21) and HCV core-induced hypoadiponectinemia (22)(23)(24). The cure of infection is often associated with IR reduction, in particular in patients infected with genotype 1 (25)(26)(27)(28).
The direct involvement of HCV in T2D has also been suggested in experimental models. HCV core-mediated down-regulation of insulin receptor substrates 1 and 2 (IRS-1/2) and protein phosphatase 2A (PP2A)-dependent Akt dephosphorylation mediated by the nonstructural NS5A protein have been suggested to play a role in HCV-induced IR (29 -33). Other studies showed that HCV modulates the activity of transcription factors such as peroxisome proliferator-activated receptor ␥ coactivator 1-␣ (PGC1␣) and FoxO1 and FoxA2, both implicated in metabolic enzyme expression (34 -37).
Thus, clinical observations and experimental data strongly suggest that HCV directly induces T2D in infected patients. However, the molecular mechanisms underlying HCV-induced T2D, the associated IR, and glucose metabolism abnormalities remain unknown. In this study, we show that FL-N/35 HCVtransgenic mice that express the full-length HCV genotype 1 open reading frame (38) are at a prediabetic stage, exhibiting glucose intolerance and IR, mirroring the situation in HCVinfected patients. Mechanistically, we demonstrate that the HCV proteins directly impair Glut2-mediated hepatic glucose intake and insulin-driven shutdown of gluconeogenesis by down-regulating IRS-2 and altering FoxO1 phosphorylation and nuclear exclusion.

Mice expressing the full-length HCV open reading frame are insulin-resistant and glucose-intolerant
We studied the impact of the HCV proteins on the prevalence of insulin resistance in our murine model. Transgenic and WT mice did not display different fasting and fed glycemias (Fig. 1A). In contrast, fasting and fed insulinemias were significantly higher in HCV-transgenic mice than in their WT littermates (Fig. 1B). As a result, the homeostatic model assessment of insulin resistance index was significantly higher in HCVtransgenic than in WT mice (13.0 Ϯ 3.4 versus 7.9 Ϯ 0.7 microunits/mg, p Ͻ 0.05), suggesting that HCV mice are resistant to insulin.
Because orally administered glucose can induce secretion of glucagon-like peptide 1 (GLP-1) by the gut, which increases insulin secretion by pancreatic ␤ cells, mice were subjected to intraperitoneal glucose tolerance tests (IPGTT), which bypass the effect of GLP-1 and are associated with lower glucose-stimulated insulin secretion than oral glucose tolerance tests. As shown in Fig. 1, C and D, IPGTT analyses confirmed that HCVtransgenic mice are glucose-intolerant.
Body weight and fat mass did not differ between HCV-transgenic mice and their WT littermates. Weight gain was similar in both groups until 13 months of age, followed by a steady state until 20 months of age (supplemental Fig. 1A). Moreover, body fat assessments by tomography scanner (39) showed similar body mass distributions in both groups, with no evidence of excessive fat storage in HCV-transgenic mice, ruling out an effect of accumulated lipids in the IR of HCV-transgenic mice (supplemental Fig. 1B), i.e. excluding a possible role of a meta-bolic syndrome in our findings. Together, these results indicate that the sole expression of HCV proteins in the livers of transgenic mice is associated with insulin resistance and glucose intolerance, i.e. a prediabetic stage unrelated to a metabolic syndrome.

Insulin resistance in HCV-transgenic mice is of both hepatic and muscular origin
To complement these data, glucose turnover rate and metabolic flux were measured in HCV-transgenic and non-transgenic animals by means of hyperinsulinemic-euglycemic clamps. The basal glucose turnover rate was similar in WT and HCV-transgenic mice (Fig. 1E). The glucose infusion rate (GIR), i.e. the amount of glucose needed to maintain euglycemia, and the whole-body glucose utilization rate (GUR) were significantly lower in HCV-transgenic than in WT mice, indicating a defect in glucose uptake from the bloodstream despite hyperinsulinemia (Fig. 1E). Furthermore, endogenous glucose production was significantly higher in HCV-transgenic animals ( Fig. 1E), indicating a defect in insulin-driven shutdown of gluconeogenesis and/or glycogenolysis in the liver. As shown in Fig. 1F, glucose uptake in the soleus oxidative skeletal muscle (S, soleus) and, to a lesser extent, the visceral white adipose tissue (WATv) were significantly lower in HCV-transgenic mice than in WT animals, whereas no difference was observed in glucose uptake in subcutaneous white adipose tissue (WATs). Together with our previous analyses, these results confirm that mice expressing the full-length HCV open reading frame are insulin-resistant and indicate that insulin resistance is of both hepatic and muscular origin.

Impaired liver glucose intake in HCV-transgenic mice is at least partly due to Glut2 down-regulation by HCV proteins
To examine the underlying mechanisms surrounding the IR observed in HCV-transgenic mice, we next examined the membrane expression of glucose and hexose transporter 2 (Glut2), which has been implicated as an important regulator of hepatic glucose intake (40,41). Western blot experiments showed significantly lower amounts of Glut2 in membrane fractions from HCV-transgenic mouse livers compared with WT controls (Figs. 2, A and B), without significant differences in its transcript levels (0.85 Ϯ 0.33-fold-change in HCV-transgenic compared with wild-type animals). Correspondingly, glucose uptake in cultured mouse primary hepatocytes was significantly lower in HCV-transgenic hepatocytes than in non-transgenic ones (Fig. 2C). These observations may, at least in part, account for the insulin resistance observed in the presence of the HCV proteins.

Glycogen storage is not altered in HCV-transgenic mice
Next we examined whether the HCV proteins affected glycogen synthesis by examining the expression and activation of glycogen synthase kinase 3, the enzyme that inhibits glycogen synthase. Western blotting showed no difference in the expression or phosphorylation of GSK3 on Ser 21 between transgenic and non-transgenic animals (Figs. 3, A and B). In addition, the amounts of liver-stored glycogen did not differ between fasting HCV-transgenic and WT mice, respectively (Fig. 3C). Therefore, HCV does not impact hepatic glycogen storage.

Insulin-driven shutdown of hepatic gluconeogenesis is impaired in HCV-transgenic mice
Given our previous data, we postulated that the regulators of glucose production, which operate in the switch from gluconeogenesis to glycolysis in response to insulin, may be affected by the HCV proteins. Therefore, we examined the transcription levels of key regulators of hepatic gluconeogenesis by means of quantitative RT-PCR in liver RNA extracts from fasting HCV-transgenic and non-transgenic mice. They included glucokinase (GK), fructose 1,6-bisphosphatase (PFKFB3), phosphoenolpyruvate carboxykinase (PEPCK, PCK1 gene), and glucose-6-phosphatase (Glc-6-PC). As shown in Fig. 4A, the expressions of GK and PFKFB3 were reduced by ϳ5-fold and 1.5-fold, respectively, in the livers of HCV-transgenic mice compared with WT animals. The amount of PCK1 transcripts was 2-fold higher in transgenic than in WT animals, whereas Glc-6-PC mRNA levels were identical in both groups. Correspondingly, Western blotting of crude liver extracts revealed that GK and PEPCK protein expression was lower and higher, respectively, in HCV-transgenic mice than in non-transgenic animals, as shown in Fig. 4B. Furthermore, although insulin injection increased GK mRNA expression in WT mice, GK mRNA levels were reduced in HCV-transgenic mice in response to insulin (Fig. 4C), demonstrating that regulation of gluconeogenesis is perturbed by the HCV proteins. Together, these data indicate that HCV transgenic mice develop hepatic IR leading to higher gluconeogenesis and lower glycolysis.

IRS2 expression is down-regulated in HCV-transgenic mice
To further understand the mechanisms for the IR observed in HCV-transgenic mice, we next examined the hepatic insulin receptor pathway. Similar basal amounts of insulin receptor were found in liver extracts from HCV-transgenic and nontransgenic mice by Western blot analysis (data not shown). In contrast, indirect immunofluorescence revealed lower basal levels of IRS2 in primary hepatocytes from HCV-transgenic mice than in those from WT animals (Fig. 5A). Moreover, both basal levels of IRS2 and those induced following intraperitoneal insulin injection were decreased in liver extracts from HCVtransgenic mice compared with the WT (Fig. 5, B and C). However, IRS2 mRNA levels were identical in transgenic and nontransgenic mouse livers by quantitative RT-PCR assay (Fig. 5D), ruling out modulation of IRS2 expression by HCV proteins at the transcriptional level. These results suggest that IRS2 expression is down-regulated at the posttranscriptional level in HCV-transgenic mice and demonstrates that the HCV proteins perturb the hepatic insulin receptor pathway.   21 GSK3␣ phosphorylation status in liver extracts from HCV-transgenic (HCV-tg) and control WT mice by Western blot analysis, normalized to total GSK3 expression (representative animals). B, average Ser(P) 21 GSK3␣ protein levels in liver extracts from HCV-tg and control WT mice by Western blot analysis, normalized to total GSK3 expression. NS, not significant. C, total liver glycogen content in HCV-tg and control WT mice, normalized to total liver protein weight.

Down-regulation of IRS2 expression in HCV-transgenic mice occurs via a SOCS3-dependent mechanism
To elucidate the underlying mechanisms for these observations, we examined whether the degradation of IRS2 was affected by the viral proteins. In mouse primary hepatocytes isolated from WT and HCV-transgenic mice, treatment with the proteasome inhibitor MG132 restored IRS2 levels in HCVtransgenic animals to those observed in WT animals (Fig. 5E), demonstrating that HCV down-regulated IRS2 levels by inducing proteasomal degradation.
Several kinases have been implicated in regulating the phospho-dependent degradation of IRS2, including PKC⑀, which phosphorylates IRS2 at serine 612 and induces its degradation by the proteasome (42); p70 S6 kinase (p70S6K), activated through a negative feedback loop resulting from the prolonged action of insulin, which phosphorylates IRS2 at multiple serine residues and induces its degradation (43); and suppressor of cytokine signaling 3 (SOCS3), which induces IRS1/2 degradation and inhibits IRS2 tyrosine phosphorylation, therefore blocking insulin signal transduction (44,45). We therefore examined the levels and activation of these kinases in our murine model of HCV-induced IR.
Western blot analyses of liver extracts showed similar levels of PKC⑀ phosphorylation at serine 728 in HCV-transgenic and WT mice (Fig. 5, F and G). In agreement, fasted HCV-transgenic mice did not show significantly lower levels of p70S6K phosphorylation at serine 389 than WT animals (Fig. 5, F and H). In contrast, the expression of SOCS3 protein (Fig. 5, F and I) and mRNA (Fig.  5J) was significantly higher in the livers of fasted HCV-transgenic mice than in those of their WT littermates. In addition, TNF␣ transcript levels were 2.5 Ϯ 0.5 times higher in HCV-transgenic than in WT mice. Because TNF␣ directly regulates SOCS3 expression, these data provide a potential explanation for the higher levels of SOCS3 expression in HCV-transgenic mice (data not shown). Together, these results indicate that the hepatic insulin signaling is impaired in mice expressing the full-length HCV open reading frame, through the proteasomal degradation of IRS2, potentially via a SOCS3-dependent mechanism.

Down-regulation of IRS2 expression down-regulates the hepatic insulin signaling pathway
Because our data strongly suggest that HCV proteins induce the degradation of hepatic IRS2, we examined the downstream effects on hepatic insulin receptor signaling. In normal hepatic tissues, phosphoinositide-dependent protein kinase 1 (PDK1) is activated downstream of IRS2 activation through phosphorylation at serine 241. In keeping with our previous data, Western blot analysis of liver extracts showed significantly lower amounts of Ser(P) 241 PDK1 in HCV-transgenic animals subjected to vehicle injections than in their WT littermates. Moreover, insulin treatment induced PDK1 hyperphosphorylation in the livers of WT mice but not in HCV-transgenic animals (Fig. 6, A  and B). Furthermore, phosphorylation of the downstream target of PDK1, Akt, at threonine 308 was also significantly lower in HCVtransgenic than in WT animals, both in the presence or absence of insulintreatment (Fig.6,CandD).However,despitethishypophosphorylation at threonine 308, phosphorylation of AKT at serine 473 and Akt kinase activity, assessed using recombinant GSK3 protein, were constitutively enhanced in the livers of older (14 -18 months old) HCV-transgenic mice (Fig. 6, E and F), confirming our previous observation in younger HCV-transgenic mice (46). These data suggest that, in our murine model, although decreased IRS2 expression has a knock-on effect on PDK1 activation, global Akt activity is elevated, probably because of an effect of the viral proteins on kinases regulating Akt Ser 473 phosphorylation.

FoxO1 phosphorylation at serine 256 by Akt is altered in HCV-transgenic mice
To examine whether the increased Akt activity observed in the presence of the HCV proteins elevated insulin signaling, we next examined the expression and phosphorylation of the forkhead in rhabdomyosarcoma transcription factor (FKHR or FoxO1) (47,48). Akt regulates glucose homeostasis by phosphorylating FoxO1 at serine 256. In keeping with a role for HCV in perturbing hepatic insulin signaling, significantly lower amounts of Ser(P) 256 FoxO1 were observed in liver extracts from vehicle-injected and insulin-treated HCV-transgenic mice compared with their WT littermates (Fig. 7, A and B). A similar result was found in cultured mouse primary hepatocytes using indirect immunofluorescence to detect Ser(P) 256 FoxO1 (Fig. 7, C, left panels, no insulin, and D). These findings were in keeping with the higher PCK1 and the lower GK and PFKFB3 transcript levels reflecting higher endogenous glucose production in the livers of HCV-transgenic mice. However, the reduced phosphorylation of FoxO1 in the presence of enhanced Akt kinase activity in the livers of HCV-transgenic mice suggests that HCV protein expression uncouples the activities of Akt/FoxO1 and that another IRS2-independent mechanism is involved.

Impairment of FoxO1 phosphorylation at serine 256 alters its nuclear exclusion in HCV-transgenic mice
The activity of FoxO1 is regulated by nuclear/cytoplasmic shuttling. Indeed, serine 256 phosphorylation of FoxO1 triggers its nuclear exclusion, thus removing it from gene promoters (49). We demonstrated that HCV proteins inhibit FoxO1-P.
We therefore examined the nuclear/cytoplasmic localization of FoxO1 in hepatocytes isolated from both HCV-transgenic and WT mice. In the absence of insulin, Ser(P) 256 -FoxO1 was mainly located in the nuclei of primary hepatocytes from both HCV-transgenic and WT animals. As expected, after insulin injection, Ser(P) 256 -FoxO1 relocated in the cytoplasm of WT hepatocytes. However, it remained in the nuclei of hepatocytes from HCV-transgenic mice (Fig. 7, C and D). These results show that, despite elevated Akt activity, HCV inhibits the phosphorylation of FoxO1 at serine 256, thereby inducing its nuclear retention and decreasing its FoxO1 transcriptional activities, leading to impaired insulin signaling.

Discussion
In patients with HCV, chronic infection has been shown to be frequently associated with insulin resistance and type 2 diabetes, accounting for significant virus-related morbidity and mortality, independent of the severity of liver disease (Refs. 18, 50, 51; for a review, see Refs. 4, 52-54). In this study, we showed that, like infected patients, transgenic mice expressing the fulllength HCV open reading frame are glucose-intolerant and insulin-resistant and that this is due to the sole expression of HCV proteins in the liver without the need for viral replication, local inflammation, or advanced liver disease such as fibrosis or cirrhosis. HCV core protein-expressing mice were reported  Figure 6. A, Ser(P) 241 -PDK1 assessed by Western blotting in liver extracts from WT and HCV-tg mice injected with insulin or vehicle. The results were normalized using total PDK1 expression (representative animals). B, quantitative results for Ser(P) 241 -PDK1 protein levels. C, Thr(P) 308 -Akt assessed by Western blotting in liver extracts from WT and HCV-tg mice injected with insulin or vehicle. The results were normalized using total Akt expression (representative animals). D, quantitative results for Thr(P) 308 -Akt protein levels. E, quantitative results for Ser(P) 473 -Akt protein levels. F, analysis of Akt activity by kinase assay in hepatic lysates from WT and HCV-tg animals. Akt activity was normalized to the relative expression of Akt in the lysates and is expressed as the percentage of WT Akt activity.
previously to be insulin-resistant, but, in contrast with our HCV-transgenic mice, they did not display glucose intolerance or altered endogenous glucose production (29). These differences could be due to the very young age of the HCV core transgenic mice (2 months) compared with the animals used in our study and due to the expression of only one HCV protein in the model, whereas ours express the full spectrum of HCV proteins at nearly physiological levels, better mim-icking HCV protein expression in infected humans (29). Our results also showed that IR is of both peripheral (striated muscle) and hepatic origin in HCV-transgenic mice, a result in keeping with clinical data showing little or no involvement of adipose tissue in this process (55). Although HCV protein expression in our model is hepatocyte-restricted, this peripheral effect may result from the endocrine effects of inflammatory cytokines (56), such as IL-8, whose overexpres-  sion is induced by the HCV NS5A protein, as we showed recently (57). Our results suggest that glucose intake is perturbed in HCVtransgenic mice, probably as a result of reduced amounts of Glut2 in the plasma membrane. Such a defect in Glut2 expression at the plasma membrane was also reported in Huh7.5 cells replicating the genotype 2 culture-adapted JFH1 HCV clone (58,59) as well as in the livers of HCV-infected patients (59). Our results demonstrate that this effect is solely related to the expression of HCV proteins in vivo. In addition, we found that insulin-driven shutdown of gluconeogenesis was defective in the livers of HCV-transgenic mice, which explains the higher endogenous glucose production observed in these mice during the clamp experiments. Previously, an increase in PCK1 and Glc-6-PC mRNAs and glucose production in Huh-7.5 cells harboring HCV genotype 1b RNA replicons or infected with the JFH1 strain has been reported (37). Other groups observed an increase in gluconeogenic activities in Huh-7 cells or in murine hepatocytes expressing the HCV NS5A protein as well as in livers from HCV-infected patients and engineered mice infected with HCV (60 -62). Nevertheless, our results extend these observations in an in vivo model and demonstrate that insulin resistance is directly due to the expression of the HCV proteins.
In HCV-infected patients, hepatic IRS1 and IRS2 levels are diminished compared with non-infected patients (25,31). In addition, HCV clearance is accompanied by normalization of IRS1 and IRS2 levels, proving the direct role of HCV in this perturbation (25). In our experiments, levels of IRS2 protein were reduced in HCV-transgenic mice, and this reduction was reversed by proteasome inhibitors. This in vivo result confirms previous findings in HCV core transgenic mouse livers and HCV core-transfected human hepatoma cells, showing proteasomal degradation of IRS1 and IRS2, possibly resulting from an increased production of proinflammatory cytokines, such as TNF␣, involving the activation of the JNK pathway by the HCV core protein (29,31,63). In keeping with these data, our HCVtransgenic mice displayed higher levels of TNF␣ in their livers, suggesting a role for pro-inflammatory cytokines in this process and providing a potential mechanistic explanation for our observations. Serine phosphorylation of IRS-1 was also described in HCV NS5A-expressing hepatoma cells (60). Furthermore, it was suggested that IRS1 down-regulation, rescued by MG132 treatment, is dependent upon mTORC1 activation in JFH1-infected hepatocytes (64). Accordingly, we observed that HCV protein expression activated Akt, an upstream activator of mTORC1, in our HCV-transgenic mice. It has been shown that SREBP1 suppresses IRS2-mediated insulin signaling in hepatocytes (65,66) via a negative feedback loop involving transcriptional regulation of the IRS2 gene. We reported previously that HCV triggers the activation of SREBP1 in the livers of HCV-transgenic mice (67). However, IRS2 mRNA levels were unchanged in our study, suggesting that this mechanism is not involved in our observations.
Our results suggest that HCV-mediated SOCS3 overexpression is responsible for IRS2 degradation in HCV-transgenic mice. Accordingly, hepatic SOCS3 levels were reported to be higher in HCV-infected patients than in healthy individuals (68,69). Furthermore, HCV core protein expression was reported to induce SOCS3 overexpression and IRS degradation, leading to IFN-␣ signaling impairment (31,70). Interestingly, SOCS3 expression was shown to be an independent predictor of IFN-␣ treatment response, especially in genotype 1-infected patients (71,72). It was also demonstrated, in a non-infected mouse model, that overexpression of SOCS induces IRS1 and IRS2 degradation, subsequently inducing insulin resistance (44). HCV core substitutions at positions Arg 70 and Leu 91 have been suggested to enhance SOCS3 expression in an IL-6/UPR-dependent mechanism (73). Interestingly, the HCV clone used to generate the FL-N/35 mouse model used in this study contains the core L91M substitution. However, we observed no UPR stimulation in the livers of these mice (67). Whether SOCS3 overexpression and the resulting degradation of IRS2 is sufficient to trigger insulin resistance and type 2 diabetes is questionable. A recent study suggested that HBV can also induce SOCS3 expression and IRS1 degradation in the liver, thereby inhibiting insulin signaling (74). Nevertheless, no association between HBV infection and type 2 diabetes or insulin resistance has been reported (75)(76)(77). Thus, other perturbations are probably necessary, justifying our assessment of the integrity of the IR pathway downstream of IRS2 in the livers of HCV-transgenic mice. Impaired gluconeogenesis inhibition by insulin generally reflects reduced Akt activation. However, we found that Akt kinase activity was constitutively enhanced in the livers of HCV-transgenic mice, in agreement with our previous findings in younger animals (46). In addition, Akt activity was significantly augmented in HCV-transgenic mice injected with insulin. This result might appear as paradoxical in the context of impaired insulin-driven gluconeogenesis shutdown and IRS2/ PDK1 down-regulation described above. Akt regulates glucose homeostasis gene expression through the phosphorylation of FoxO1 at serine 256, which, in turn, triggers its nuclear exclusion, thus removing it from gene promoters (for a review, see Ref. 78). Thus, the lower amounts of Ser(P) 256 -FoxO1 associated with the impairment of its insulin-driven nuclear exclusion in the hepatocytes of HCV-transgenic mice were also unexpected in the context of increased Akt activity in the same hepatocytes but in keeping with the elevated levels of expression of gluconeogenic genes. Akt phosphorylation at threonine 308 has been reported to be essential for glucose uptake whereas phosphorylation at serine 473 is not (79). The hypophosphorylation of Akt at threonine 308 we observed in the livers of HCVtransgenic mice could therefore explain the impairment of glucose uptake. However, others reported that the modulation of Akt phosphorylations had no effect on FoxO1 phosphorylation (35). Another study, reporting similar observations in Huh7.5 cell lines harboring an HCV genotype 1b replicon or JFH1 infection, suggested an Akt-independent mechanism involving JNK pathway activation through mitochondrial reactive oxygen species production (37). A correlation was indeed reported between oxidative stress and insulin resistance, through the analyses of hepatic thioredoxin levels, in HCV-infected patients (80). In addition, we and others have shown an accumulation of reactive oxygen species in the liver of HCV-transgenic mice (Refs. 46, 81-83; for a review, see Ref. 84), which could be at least partly responsible for the uncoupling of the Akt/FoxO1 signaling we observed.
In summary, we used the FL-N/35 transgenic mouse model, which expresses the full-length HCV open reading frame at nearly physiological levels in the liver of these animals, to demonstrate that, like in HCV-infected patients, HCV-transgenic mice are insulin-resistant and glucose-intolerant and that these effects are solely due to the expression of the viral proteins in hepatocytes. This model also allowed us to unravel the molecular mechanisms by which HCV protein expression alters glucose metabolism. Our data suggest that the impaired glucose intake in the liver of transgenic animals is partly due to a lower expression of Glut2 in the presence of HCV proteins. We also observed that the insulin-driven hepatic neoglucogenesis switch to glycolysis was impaired in HCV-transgenic mice. This resulted, at least in part, from an alteration of the hepatic insulin receptor cascade because of the down-regulation of IRS2 expression, potentially via a SOCS3-dependent mechanism, as a result of increased TNF␣ expression. Finally, the Akt/FoxO1 uncoupling we observed in HCV-transgenic animals coincided with an alteration of FoxO1 nuclear exclusion induced by HCV protein expression, which subsequently generated alterations in glucose homeostasis despite enhanced Akt activation in HCV-transgenic animals. Our results suggest that HCV-infected patients should be assessed for insulin resistance and prediabetes in their routine evaluation to avoid the double burden of diabetes mellitus and chronic liver disease. They also suggest that insulin resistance, glucose intolerance, and diabetes mellitus should be considered extrahepatic manifestations of chronic HCV infection and imply that patients presenting signs of severity should be prioritized for treatment with the new safe and efficacious IFN-free, all-oral combination regimens.

Animals
Animals were housed under specific pathogen-free conditions. 2-to 20-month-old C57BL6 male mice transgenic for the full-length HCV open reading frame (FL-N/35 lineage) (38) were used for body weight monitoring. 14-to 18-month-old HCV-transgenic mice were used for all of the other experiments in this study. Age-matched WT male littermates were used as controls. Animals harboring liver tumors at necropsy were removed from the study. The animals were housed in a temperature-controlled environment with a 12-h light/dark cycle and ad libitum access to water and diet (D04 from SAFE, Augy, France; 6.1% carbohydrate, 3.1% fat, and 15.8% protein).
When indicated, mice were injected with 0.25 units/g insulin lispro (Humalog KwikPen, 100 IU/ml, Eli Lilly, Indianapolis, IN) diluted in PBS, either directly in the portal vein in anesthetized animals sacrificed 2-5 min after injection or intraperitoneally in animals sacrificed 1 h after injection. After sacrifice by means of CO 2 intoxication, liver tissue fragments were immediately snap-frozen in liquid nitrogen and stored at Ϫ80°C before analysis. In the "fasting animal" group, animals were sacrificed after a 24-h fasting period. In the "fed animal" group, 24 h-fasting animals were fed using regular chow and glucosesupplemented water and sacrificed 3 h later.

Cell cultures
Hepatoma-derived Huh7 cells, purchased from the ATCC (Manassas, VA), were maintained in DMEM supplemented with 10% fetal calf serum, 100 units/ml penicillin, and 100 g/ml streptomycin. Murine hepatocytes from both HCVtransgenic and non-transgenic animals were isolated by portal vein perfusion of collagenase. Freshly isolated hepatocytes were cultured in DMEM supplemented with 10% fetal calf serum, 10 units/ml penicillin, 10 g/ml streptomycin, 10 g/ml insulin, 5.5 g/ml transferrin, and 5 ng/ml sodium selenite. Four hours after perfusion, the medium was removed, and fresh medium supplemented with 0.1 mol/liter dexamethasone (Sigma-Aldrich, St. Louis, MO) and 50 ng/ml epidermal growth factor was added. When indicated, cells were starved overnight by withdrawal of fetal calf serum and cultured in low-glucose (1g/liter) DMEM and then incubated with 5 g/ml insulin lispro for 20 min.

Total RNA isolation and RT-qPCR
Total RNA was isolated from frozen mouse livers using the Ambion PARIS RNA and protein isolation kit (Thermo Fisher Scientific, Waltham, MA), according to the instructions of the manufacturer. Then RNA quality and quantity were assessed by means of a 2100 Bioanalyzer and RNA Nano Chips (Agilent Technologies, Santa Clara, CA). RNA integrity number was calculated using Agilent software, and samples displaying an RNA integrity number below 6 were discarded. 1 g of RNA was reverse-transcribed using the high-capacity cDNA reverse transcription kit (Thermo Fisher Scientific). Real-time quantitative PCR was performed using TaqMan reagents and an Applied Biosystems 7300 thermal cycler (Applied Biosystems, Carlsbad, CA). Specific primers and probes were purchased from Applied Biosystems (supplemental Table 1). The expression level of each gene studied was normalized to that of 18S ribosomal RNA by using the comparative Ct method and expressed as -fold-change using the wild-type mRNA expression level for standardization.

Preparation of protein extracts and immunoblot analysis
Crude protein extracts were prepared by homogenization of frozen mouse livers (50 -100 g) in tissue lysis buffer from the Ambion PARIS RNA and protein isolation kit, supplemented with protease inhibitors (cOmplete TM EDTA-free protease inhibitor mixture, Sigma-Aldrich) and phosphatase inhibitors (PhosSTOP TM , Sigma-Aldrich), using a tissue homogenizer (MP Fast Prep24, MP Biomedicals, Santa Ana, CA) and MP Lysing Matrix A tubes. Proteins were quantified using the BCA assay (Thermo Fisher Scientific). For protein analysis, polypeptides from crude extracts were separated on precast 4 -15% SDS-polyacrylamide gels (Criterion XT precast gels, Bio-Rad) and transferred onto polyvinylidene fluoride membranes (GE Healthcare), and proteins were detected by immunoblotting. The primary antibodies used are listed in supplemental Table 2. The membranes were incubated with the corresponding secondary antibodies coupled to HRP-conjugated donkey anti-rabbit IgG or sheep anti-mouse IgG (GE Healthcare). Labeled antibodies were detected with ECL Prime detection kits (GE Healthcare). Chemiluminescent signals were detected and quantified using an Image Quant Las 4000 minicamera and Image Quant software (GE Healthcare). A polyclonal mouse ␥-tubulin antibody (supplemental Table 2) was used as a loading control for total lysates.
Akt kinase assay was performed as described previously (46) using an Akt non-radioactive kinase kit (Cell Signaling Technology, Danvers, MA). Briefly, hepatic samples (0.5 mg) were lysed and active Akt was immunoprecipitated, and its ability to phosphorylate GSK-3 in vitro was analyzed by Western blotting. Relative Akt activities are expressed as the levels of phospho-GSK3 normalized to Akt expression in the unprecipitated input.

Body mass composition
Body mass composition (lean tissue mass, fat mass) was analyzed prior to the clamp studies using a whole-body composition analyzer (EchoMRI 100, EchoMRI, Houston, TX) according to the instructions of the manufacturer. Briefly, awake animals were weighed before they were placed in a mouse holder and inserted into the MRI analyzer. Readings of body composition were obtained within 1 min. Body composition was expressed as a percentage of body weight.

Glucose tolerance tests and glucose and insulin assays
IPGTT was performed in overnight fasted mice. Briefly, glucose (2 mg/g) was injected intraperitoneally. Glycemia was assessed by means of a glucometer (Accu-Check Go and the corresponding test strips, Roche) from 2 l of blood collected from the tip of the tail vein 0, 10, 20, 40, 60, 90 and 120 min after glucose injection. Insulinemia was measured before glucose injection on blood samples collected from the mouse tail vein using heparinized capillaries. Plasma was separated by centrifugation, and the insulin concentration was assessed using an ultrasensitive mouse insulin ELISA kit (Mercodia, Uppsala, Sweden), according to the instructions of the manufacturer.

Glucose turnover rate at basal state and during hyperinsulinemic-euglycemic clamps
One week before the experiment, an indwelling catheter (BD Biosciences) was inserted into the right jugular veins of mice anesthetized with isoflurane. To measure the glucose turnover rate under basal conditions, a 5-Ci bolus of [3-3 H]glucose was injected through the jugular vein, followed by a continuous infusion of [3-3 H]glucose (15 Ci) at a constant rate of 1 l/min for 90 min. Blood samples were collected at the end of the experiment for radioactivity measurement (see below).
The experiment aimed at measuring the glucose turnover rate during hyperinsulinemic-euglycemic clamping was conducted in conscious mice fasting for 5 h as described previously (85). Briefly, a 5-Ci bolus of [3-3 H]glucose and a priming dose of insulin (83 milliunits/kg; Actrapid, Novo Nordisk, Bagsvaerd, Denmark) dissolved in isotonic saline was injected through the jugular vein, followed by a continuous infusion of [3-3 H]glucose (15 Ci) and insulin (2 milliunits/kg/min) at a constant rate of 1 l/min to maintain blood glucose levels at 100 mg/dl. During the clamping, blood was sampled from the cut tail every 10 min to determine glucose levels and to adjust the rate of unlabeled glucose infusion to maintain euglycemia. The euglycemic conditions were reached within 30 -40 min and subsequently maintained for 60 min. Steady-state specific glucose radioactivity and plasma glucose were determined during the last 20 min of the clamping. During the glucose clamping (50 -70 min after the onset of insulin infusion), the glucose disposal rate, which reflects the GUR, is equal to the rate of glucose appearance, which results from endogenous glucose production, added to the amount of infused glucose necessary to maintain euglycemia (GIR, expressed in milligrams/minute/ kilogram). To measure [3-3 H]glucose radioactivity, blood samples were deproteinized with Ba(OH) 2 and ZnSO 4 , and the supernatant was evaporated to dryness at 50°C to remove tritiated water. The dry residue was dissolved in 0.5 ml of water, to which 10 ml of scintillation solution was added (Aqualuma Plus, Lumac), and radioactivity was determined in a Packard Tri-Carb 460C liquid scintillation system.

Indirect immunofluorescence
For indirect immunofluorescence studies and proximity-dependent DNA ligation assays, Huh7.5 cells transiently expressing HCV core protein (genotype 1b (87)) or primary mouse hepatocytes were fixed with 2% paraformaldehyde in PBS for 10 min at room temperature and permeabilized with ice-cold methanol at Ϫ20°C for 10 min, washed with PBS, and incubated for 1 h at room temperature with blocking buffer (Duolink assay, Sigma-Aldrich). The primary antibody was diluted in antibody diluent reagent (Duolink assay, Sigma-Aldrich) and incubated for 1 h at room temperature. For indirect immunofluorescence, bound antibodies were detected with an Alexa Fluor-conjugated secondary antibody (Life Technologies), whereas cellular DNA was stained with DAPI (Life Technologies). Slides were mounted using Prolong Gold antifading solution (Life Technologies), and images were captured using a Zeiss Axioskop 40 microscope in conjunction with the Zeiss MRm Axiocam and Axiovision V40 4.7 software.

Statistical analysis
Results are expressed as means Ϯ S.E. Fisher-Yates-Terry or Mann-Whitney non-parametric rank tests were used when appropriate. p values of less than 0.05 were considered statistically significant.

Ethics statement
Animal housing and procedures were conducted in accordance with the recommendations from the Direction des Ser-