Hormonal Regulation Of Hepatic Glucose Production In Health And Disease

Abstract

We review mechanisms that regulate production of glucose by the liver, focusing on areas of budding consensus, and endeavoring to provide a candid assessment of lingering controversies. We also attempt to reconcile data from tracer studies in humans and large animals with the growing compilation of mouse knockouts that display changes in glucose production. A clinical hallmark of diabetes, excessive glucose production remains key to its treatment. Hence, we attempt to integrate emerging pathways into the broader goal to rejuvenate the staid anti-diabetic pharmacopeia.

Introduction

Hepatic glucose production (HGP) is a key physiological process that becomes altered in diabetic patients (Bogardus et al., 1984) and represents the main target of the anti-hyperglycemic effect of biguanides (Stumvoll et al., 1995). The combination of tracer and spectroscopy methods in humans, selective catheterization of portal and hepatic veins in dogs during tracer studies, conditional knockouts and metabolic phenotyping in mice, and NMR-based flux analysis in perfused liver and primary hepatocytes has yielded a wealth of information on mechanisms controlling HGP. However, these investigations appear to occur in tangential, if not parallel universes, leading to often-conflicting claims. Purpose of this review is to integrate this knowledge by critically analyzing the extent to which different pathways meet exacting genetic, cellular biological, and integrated physiological criteria across different experimental models.

Features of HGP in vivo, in the perfused liver, and in cultured cells

Classic work by Exton and Park established that key features of HGP can be reproduced in isolated liver (Exton and Park, 1967). This work demonstrated that hormonal regulation of HGP is: (i) rapid, occurring within seconds of exposing the organ to glucagon or insulin; (ii) sensitive, as either hormone is effective at sub-picomolar concentrations; and (iii) independent of ongoing protein synthesis. The forgotten implications of these landmark observations are that putative physiologic regulators of HGP ought to be regulated at low hormone concentrations by post-translational modifications of existing signaling complexes and by substrate flux, rather than by de novo gene transcription.

In the 1980’s, work in rodents and in cultured hepatocytes conclusively demonstrated that, in addition to substrate flux, insulin and glucagon also regulate expression of the rate-limiting enzymes of glycogenolysis (the catalytic subunit of glucose-6-phosphatase, encoded by G6pc) and gluconeogenesis (the cytosolic phosphenolpyruvate carboxykinase, encoded by Pck1). This work emphasized the role of cAMP- and insulin-responsive transcription factors (O'Brien and Granner, 1996). The unintended consequence of this landmark research was that regulation of gene transcription and HGP are all too often conflated into a single process, with recurring interpretive errors.

Tracer studies in dogs have defined hormonal regulation of HGP in detail. As in the isolated rodent liver, HGP is exquisitely sensitive to glucagon and insulin. Glucagon sets the basal tone, but insulin trumps glucagon at any concentration–just as it does in vitro. Both hormones affect primarily glycogenolysis by reciprocal changes of glycogen synthase and glycogen phosphorylase, and by modulating glycolysis through glucokinase, fructose-bisphosphatase and pyruvate kinase (see below) (Cherrington, 1999). Hormonal regulation of gluconeogenesis has proven difficult to demonstrate. Acute elevations of insulin in the physiologic range have a transient effect on gluconeogenesis (Ramnanan et al., 2010), whereas high insulin concentrations are required to bring about persistent changes (Edgerton et al., 2009). Inhibition of gluconeogenesis by insulin parallels its effect to lower FFA and lactate, consistent with a bimodal mechanism: direct stimulation of glycolysis and glycogen synthesis, and indirect inhibition through decreased gluconeogenic precurors (Bergman and Ader, 2000). The Cherrington group also made a commendable attempt to link flux data with signaling events, as will be seen below (Ramnanan et al., 2010).

Physical exercise, by virtue of its combined actions on stress and glucoregulatory hormones, as well as tissue glycogen levels, is a potent regulator of HGP (Holloszy and Kohrt, 1996).

HGP and diabetes

Unlike muscle and adipose insulin resistance, which antedate hyperglycemia by years and remain relatively stable throughout the course of the disease (Weyer et al., 1999), the rise of HGP occurs ‘late’ in the natural history of diabetes, but appears to worsen progressively, and to become refractory to treatment (Monnier et al., 2007). In type 2 diabetes, HGP is higher in the post-absorptive state, and fails to be properly suppressed by insulin, resulting primarily from excessive gluconeogenesis, rather than glycogenolysis (Rizza, 2010). As HGP is inversely correlated to insulin levels (Bogardus et al., 1984), its increase likely reflects as much the plight of the β-cell as it does a deterioration of hepatic insulin action.

Several factors contribute to elevated gluconeogenesis in diabetes: (i) increased supply of glucogenic precursors to the liver (glycerol, amino acids, free fatty acids); (ii) increased liver lipid content; (iii) cytokines and adipokines; (iv) altered glucagon-to-insulin ratios; (v) in rodents, vagal control originating in the hypothalamus; and (vi) decreased insulin receptor signaling in hepatocytes (Figure 1).

Figure 1. Direct and indirect control of HGP.

The liver integrates cell-autonomous and cell-nonautonomous mechanisms to control glucose release into the bloodstream. The role of altered glucagon-to-insulin ratios in HGP of type 2 diabetics remains unsettled and of potential therapeutic import. Classic mechanisms of indirect control of HGP include release of gluconeogenic precursors from adipose tissue and muscle (FFA, glycerol, amino acids), adipocytokines (leptin, adiponectin, resistin), neuronal control–possibly mediated through the vagus nerve. In addition, intrahepatic fat plays an important role in promoting HGP, possibly through accumulation of complex phospholipids. Finally, intra-hepatic accumulation of resident macrophages has in recent years taken center stage as a potential mechanism of insulin resistance, leading to lipid accumulation and increased HGP.

Inhibition of gluconeogenesis by insulin in humans remains disputed. The balance of the evidence is consistent with a small effect that requires high insulin concentrations, and is possibly secondary to decreased FFA (Gastaldelli et al., 2001; Petersen et al., 1998). Equally controversial is the role of glucagon in the increased HGP of diabetes. Strong cases can be made for and against a dominant role of glucagon in driving HGP (Raju and Cryer, 2005). We would like to suggest that the effect of glucagon on HGP is important, but indirect. In fact, insulin trumps glucagon at any concentration to inhibit HGP, making it unlikely that hyperglucagonemia itself would be sufficient to raise HGP, in the absence of insulin resistance. This is illustrated by the biphasic effect of somatostatin–which initially inhibits, then stimulates HGP (Sherwin et al., 1976), indicating that glucagon may not be necessary for the development of hyperglycemia, and hence assigning a primary role to insulin. But glucagon trumps insulin in the regulation of hepatic de novo lipogenesis. Thus, even at concentrations that may not override the effect of insulin on HGP, glucagon might drive excessive liver fat deposition, providing a mechanism to explain the association between impaired hepatic glucose and fat metabolism, the so-called ‘selective’ insulin resistance (Kim-Muller and Accili, 2011).

The roles of adipokines, liver fat, and neural control in driving HGP are reviewed below.

Regulation of HGP through substrate flux

Glycogen Synthase and Phosphorylase

The rapid onset of hormone action on HGP is likely independent of gene transcription. The opposing effects of insulin and glucagon on HGP pivot around their actions on glycogen synthase and phosphorylase. By activating the former, insulin favors glycogen deposition; glucagon activates the latter, resulting in glycogenolysis. It should be emphasized that increased glycogenolysis doesn’t necessarily increase HGP, owing to glucose cycling (Petersen et al., 1998). Synthase activation by insulin involves phosphorylation at multiple sites and allosteric binding of glucose-6-phosphate (Roach, 2002). Phosphorylase is rate-limiting for glycogenolysis, and is oppositely regulated by covalent modifications, through PKA and phosphorylase kinase, and allosterically by its product glucose-6-phosphate. Demonstration that insulin inhibits phosphorylase by covalent modifications is elusive. In euglycemia, insulin paradoxically increases phosphorylase activity, shunting glucose to glycogen cycling. Phosphorylase inhibition by insulin is observed under hyperglycemic conditions, suggesting a role for elevated glucose-6-phosphate levels in this process (Petersen et al., 1998). Efforts to leverage phosphorylase inhibition for diabetes treatment have floundered in recent years.

Fructose-1,6-bisphosphatase

FBP catalyzes the penultimate step in gluconeogenesis, converting fructose-1,6-bisphosphate (F1,6BP) to fructose-6-phosphate (Figure 2). This step is necessary for the incorporation of three-carbon substrates into glucose (Pilkis and Claus, 1991), and makes FBP an attractive target for drug development, given that glycerol gluconeogenesis is increased in diabetes. Indeed, genetic mutations and pharmacological inhibition of FBP in rats and humans (Gumbiner et al., 2009) demonstrate its regulatory role in HGP. Feedback inhibition by fructose-2,6-bisphosphate (F2,6BP) and AMP reduces flux through FBP, while glucagon stimulation of cAMP/PKA rapidly reduces intracellular F2,6BP by favoring the phosphatase activity of 6-phosphofructo-2-kinase/fructose-2,6-bisphosphatase, and relieves FBP inhibition to drive gluconeogenesis (Pilkis and Claus, 1991). Accordingly, raising hepatic F2,6BP levels improves insulin sensitivity and lowers glycemia in mice, while reducing F2,6BP levels has the opposite effect (Wu et al., 2006). In addition, FBP activation depletes its substrate F1,6BP–an allosteric activator of the glycolytic enzyme liver-type pyruvate kinase (PK). PK–the opposing enzyme to PCK1 in the phosphoenolpyruvate/pyruvate cycle–is also inhibited by glucagon and ATP (Pilkis and Claus, 1991). Thus, glucagon simultaneously activates FBP and inhibits PK by phosphorylation of both enzymes and by a coordinate increase in F1,6BP and decrease in F2,6BP, driving glucose formation. Moreover, AMP and ATP levels directly regulate gluconeogenic and glycolytic enzymes, thus affecting the balance between glucose storage/oxidation and production.

Figure 2. Hormonal and nutrient pathways of HGP regulation in hepatocytes.

Some of the main signaling pathways reviewed in this article are summarized. Insulin, glucagon, and glucocorticoids remain the central regulators of HGP. Their transcriptional effects are mediated through the Akt/Foxo and possibly Crtc2 pathways. AMPK, by virtue of its energy-sensing role, is likely to participate in several of these processes, but its upstream regulators and downstream targets are elusive. Bile acids have profound effects on HGP through complex and redundant mechanisms. Cytokines released from a variety of sources (fat cells, circulating as well as resident macrophages) also affect HGP and are likely to impair insulin sensitivity. A second level of transcriptional integration is provided by cofactors such as Sirt1, Clock genes, Src1–3. Their hormonal regulation is unknown at present. But transcriptional effects account only for part of HGP regulation. Direct control over enzymatic flux through glycolysis and gluconeogenesis, summarized on the right, is likely to be the critical mechanism for rapid regulation of this process.

Pyruvate dehydrogenase kinase (PDK)

The pyruvate dehydrogenase complex (PDC) converts pyruvate to acetyl-CoA, committing the substrate to cellular respiration. PDC is phosphorylated and inhibited by PDK and dephosphorylated and activated by pyruvate dehydrogenase phosphatase. PDK is in turn activated by the products of PDC, acetyl-CoA and NADH (Figure 2). Among the four PDK isoforms, PDK2 and PDK4 are induced by fasting and inhibited by insulin. PDK inhibits PDC during fasting, sparing pyruvate for gluconeogenesis. Accordingly, Pdk4 knockout mice show fasting hypoglycemia, secondary to increased systemic glucose oxidation and decreased gluconeogenic substrate delivery to the liver (Jeoung et al., 2006). Pharmacological inhibitors of PDK reduce glycemia in diabetic rodents (Mayers et al., 2005), but the underlying mechanism of increasing glucose oxidation at the expense of fatty acid oxidation suggests a potential risk for hepatosteatosis.

Transcriptional regulation of HGP by insulin, cAMP, and somatostatin

Forkhead-O transcription factors

The identification of the forkhead transcription factor daf-16 as effector of insulin receptor signaling in C. elegans led to the testable hypothesis that its mammalian ortholog, FoxO1, was the long-sought hormone-regulated transcription factor that integrated cell surface receptor signaling with HGP (Accili and Arden, 2004). FoxO proteins meet the “Exton & Park” criteria for putative mediators of hormonal HGP. cAMP promotes FoxO nuclear retention and dephosphorylation, whereas insulin at low concentrations promptly inactivates FoxO by driving its nuclear exclusion via Akt-dependent phosphorylation (Nakae et al., 2001). The main FoxO target, G6pc is key to physiologic control of HGP. In dogs subjected to physiologic hyperinsulinemia, FoxO1 phosphorylation parallels inhibition of G6pc mRNA and suppression of glycogenolysis (Ramnanan et al., 2010). FoxO1 loss-of-function in liver reduces HGP by half by decreasing glycogenolysis, gluconeogenesis, and glucose cycling, and results in neonatal as well as starvation-induced hypoglycemia (Matsumoto et al., 2007), whereas its gain-of-function prevents hormonal regulation of HGP (Nakae et al., 2002). The extent to which residual HGP in FoxO1 knockout mice reflects non-transcriptional effects of hormones on glucose flux, indirect regulation of HGP (e.g., through the CNS), or additional hepatic transcription factors (e.g., CRTC2) is at present unknown. When all three Foxo genes (1, 3a, and 4) are ablated from mouse liver, fasting glycemia is extremely low (~30mg/dl), but decreases in G6pc are moderate, indicating that additional target genes and mechanisms contribute to FoxO regulation of HGP (Haeusler et al., 2010) (Figure 2).

Insulin signaling pathways to HGP

The Irs/PI-3K/Akt/FoxO pathway is critical for insulin regulation of hepatic glucose metabolism (Dong et al., 2008). Additional serine/threonine kinases, such as atypical PKCs and salt-inducible kinases (SIK1 and 2), have been proposed to regulate HGP. But atypical PKC knockout mice have normal HGP (Matsumoto et al., 2003), amidst lingering controversy on PKC activation by insulin. The involvement of SIK1 and 2 is indirect: SIK1 is regulated transcriptionally (Koo et al., 2005), whereas SIK2 is regulated via Akt-dependent phosphorylation (Dentin et al., 2007), and can thus likely be subsumed under the Akt pathway. Signal transducer and activator of transcription 3 (STAT3) is phosphorylated in response to insulin, and its ablation increases HGP (Inoue et al., 2004a). However, this appears to be an indirect effect of insulin, as will be discussed in the section on liver/brain interactions.

Attempts to identify Akt modulators that affect HGP have been marred by confusing results. Tribbles homolog 3 (Trb3) was identified as an inhibitory pseudo-substrate of Akt, whose gain-of-function induced insulin resistance and raised glycemia in mice (Du et al., 2003). But loss-of-function data are generally inconsistent with a role of Trb3 in insulin action or HGP: whereas one report showed improved glucose tolerance following siRNA-mediated Trb3 knockdown (Koo et al., 2004), Akt signaling and hepatic glucose metabolism were normal in different models of Trb3 loss-of-function in rodents (Okamoto et al., 2007). A similar fate befell adaptor protein APPL1: originally shown to prevent the interaction of Trb3 with Akt and increase HGP inhibition by insulin, its knockout had no effect on glucose tolerance and insulin sensitivity (Tan et al., 2010). In summary, it appears that the rise of HGP in diabetes cannot be explained by post-receptor mechanisms of insulin resistance, and requires decreased insulin receptor number and/or activity as the sine qua non of reduced insulin signaling (Lauro et al., 1998).

Peroxisome proliferator-activated receptor γ coactivator-1α

PGC-1α is induced by fasting and promotes transcription of glucogenic enzymes and mitochondrial proteins (Yoon et al., 2001). FoxO1 is the obligate partner for the former, but not for the latter effect (Matsumoto et al., 2007). The FoxO1 requirement explains why knockout of Pgc1α has modest effects on HGP (Lin et al., 2004). Substrate flux analysis by MR spectroscopy shows that PGC-1α affects HGP by promoting mitochondrial biogenesis and oxidative metabolism (Burgess et al., 2006). The extent and relative role of direct regulation of PGC-1α by insulin is disputed: one report indicates that insulin decreases Pgc1α mRNA (Herzig et al., 2001), while another shows that it promotes PGC-1α phosphorylation (Li et al., 2007). The metabolic phenotype of liver-specific PGC-1α knockouts has not been reported to date.

cAMP response element-binding protein (CREB)

Transcription factor CREB belongs to a family of stress-activated DNA binding proteins with multifaceted functions. Hepatic CREB is phosphorylated in response to glucagon, catecholamines, and insulin (Koo et al., 2005), and binds to cis-acting cAMP response elements on target promoters. Phosphorylated CREB acts as a scaffold for coregulators such as CBP, p300, and CREB-regulated transcription coactivator 2 (CRTC2) that activate gluconeogenic genes (Koo et al., 2005). A dominant-negative CREB mutant causes fasting hypoglycemia that can be rescued by PGC-1α overexpression (Herzig et al., 2001). Acute CREB knockdown in liver reduced glycemia and improved insulin sensitivity in diabetic mice and rats (Erion et al., 2009a), but to date we lack a model of liver-specific CREB knockout that would allow us to parse its complex effects. The mechanism by which CREB affects HGP remains unclear, given that the competing hormones insulin and glucagon have similar effects on CREB phosphorylation.

CREB Regulated Transcriptional Coactivator 2

CRTC2 (also known as TORC2, not to be confused with the Target of Rapamycin Complex 2) is a CREB coactivator that confers hormone regulation on HGP (Koo et al., 2005). During fasting, CRTC2 is dephosphorylated, allowing its nuclear translocation. Nuclear CRTC2 binds to CREB and recruits CBP and p300 to activate Pck1, G6pc, and Pgc1α transcription (Koo et al., 2005). Insulin triggers CRTC2 phosphorylation via SIK2, promoting its nuclear exclusion and proteosomal degradation (Dentin et al., 2007). Two CRTC2 knockout alleles have been made to test its role in vivo: a liver-specific knockout has no effect on HGP or glycemia (Le Lay et al., 2009), while a ubiquitous knockout lowers fasting glycemia by ~25% as well as Pck1 and G6pc levels, together with a twofold reduction of glucose output from isolated hepatocytes, but HGP was not measured in vivo (Wang et al., 2010). The reasons for this discrepancy are unclear, and will have to be addressed if a consensus on the role of CRTC2 in HGP is to be reached.

Glucocorticoids (GC) drive HGP directly by transcriptional induction of Pck1 and tyrosine aminotransferase (Tat), and indirectly through their actions in muscle and adipose tissue to promote amino acid and glycerol flux to the liver, and by decreasing insulin production in pancreatic β-cells. Pharmacological inhibition of the enzyme required to generate bioactive corticosterone, 11β-hydroxysteroid dehydrogenase type 1 improves glucose control in type 2 diabetic patients (Rosenstock et al., 2010). However, regulation of glycogenolysis in primary hepatocytes doesn’t require GC (Matsumoto et al., 2007), indicating that these hormones participate primarily in the regulation of HGP by stress, rather than by fasting and feeding.

GC action on Pck1 and Tat transcription is mediated through ligand-induced binding of the nuclear glucocorticoid receptor (GR, encoded by Nr3c1) to GC response elements. Liver-specific GR knockout mice are euglycemic in physiologic conditions, but prone to hypoglycemia during starvation (Opherk et al., 2004). Knockdown of liver and adipose GR reduced HGP and glycemia in diabetic rodents, as did a liver-selective synthetic GR antagonist (Watts et al., 2005). These studies indicate that hepatic GC/GR signaling is essential for the stress-induced glycemic response. Efforts to leverage the eminent pharmacological tractability of GC to modulate HGP are limited by potential side effects of GC blockade on the hypothalamic-pituitary-adrenal axis, and by lingering uncertainty on the contribution of GC excess to the pathogenesis of human diabetes.

NR4As

Transcription of the three orphan nuclear receptors of the NR4A family (Nurr77, Nurr1, and NOR-1, encoded by Nr4a1, 2, and 3) is rapidly induced by glucagon or fasting in liver (Pei et al., 2006). Overexpression of each NR4A isoform in cultured hepatocytes activates G6pc, fructose-1,6-bisphosphatase (encoded by Fbp), and enolase. Accordingly, a pan-NR4A dominant negative mutant lowered HGP, while Nurr77 overexpression in liver increased it (Pei et al., 2006). But whole-body Nurr77 knockout also increased HGP (Chao et al., 2009). It should be noted that glucagon promotes HGP without stimulating new protein synthesis (Exton and Park, 1967), making it unlikely that NR4As–whose induction requires de novo synthesis–be physiologic mediators of glucagon’s effects on HGP.

Adenosine 5'-monophosphate-activated protein kinase

AMPK is a cellular sensor of energy levels. Activated by AMP- or Ca²⁺-dependent phosphorylation through LKB1 and CaMKKβ, respectively, AMPK generates ATP by increasing fatty acid oxidation and reducing ATP hydrolysis through decreased lipogenesis and glucogenesis (Zhang et al., 2009). Glucagon can also activate AMPK, but fails to do so in liver-specific Pck1 knockout mice (Berglund et al., 2009), suggesting that its effects on AMPK are mediated by cellular ATP depletion through gluconeogenesis. Whole-body knockout of AMPKα2–the predominant catalytic subunit in liver–did not affect HGP, whereas liver-specific AMPKα2 knockout caused a modest increase in basal HGP (Andreelli et al., 2006). The multiplicity of AMPK functions complicates the dissection of its upstream regulators and downstream targets.

Upstream

AMPK activators, 5-aminoimidazole-4-carboxamide riboside (AICAR), metformin, A-769662, and adiponectin suppress HGP (Zhang et al., 2009). However, these effects are preserved in mice lacking AMPK (Foretz et al., 2010), indicating that they are mediated through different mechanisms, e.g., cellular respiration. Presently, it’s unknown whether the effect of the adipokine resistin on HGP is mediated through its ability to regulate AMPK. Inactivation of the AMPK kinase LKB1 increases glycemia and causes nuclear accumulation of CRTC2 (Shaw et al., 2005), but HGP was not measured in this study, and individual contributions of LKB1 substrates–AMPK, SIK1–2, and MAP/microtubule affinity-regulating kinase 2 (MARK2)–to this phenotype haven’t been determined.

Downstream

Despite the important role of FoxO in HGP, its regulation by AMPK has not been examined extensively; in vitro data suggest that AMPK promotes FoxO3a activity (Greer et al., 2007). More efforts have gone into demonstrating the role of AMPK in CRTC2 phosphorylation. The latter can be blocked by AMPK and related kinases (SIK1–2, MARK2) (Koo et al., 2005) and becomes impaired when AMPKα1 and α2 are ablated in liver. The meaning of this observation is mysterious, as these mice have normal HGP (Foretz et al., 2010). The contribution of additional AMPK substrates to HGP is thus far based on in vitro data and awaits experimental confirmation in vivo. Collectively, these data suggest that AMPK and related kinases can restrain gluconeogenesis during extreme energy deficit or diabetes, but not under physiological conditions.

Regulation of HGP by bile acid signaling

Farnesoid X receptor (FXR, NH1R4)

Bile acids (BA) inhibit HGP, and BA sequestrants decrease glycemia and improve dyslipidemia in type 2 diabetics. BA bind nuclear receptor FXR and lower glucose by hepatic and extra-hepatic mechanisms. The BA cholic acid inhibits Pck1 and G6pc via FXR-dependent induction of small heterodimer partner (SHP) (Ma et al., 2006). Phenotyping of FXR knockouts has yielded different but not irreconcilable results, with one study showing transient fasting hypoglycemia due to defective glycogen turnover (Cariou et al., 2005), and another one showing hyperglycemia secondary to hepatic and muscle insulin resistance (Ma et al., 2006). These observations likely reflect multiple glucoregulatory pathways affected by FXR deficiency, such as BA biosynthesis, which can potentially affect incretin secretion (Thomas et al., 2009), and FGF15/19 regulation as discussed below.

The corepressor SHP

The FXR target gene Shp (Nr0b2) encodes an atypical orphan nuclear receptor that lacks a DNA binding domain and represses expression of genes implicated in HGP, including G6pc, Pck1, and Pdk4 (Kim et al., 2008). Shp knockout mice show modest fed and fasted hyperglycemia and resistance to BA inhibition of Pck1 and G6pc (Ma et al., 2006). Conversely, SHP overexpression in liver reduced Pck1 and G6pc expression and lowered glycemia in diabetic mice (Kim et al., 2008).

Fibroblast Growth Factor 15/19

Post-prandial release of BA activates expression of FGF15 (19 in rodents) in the small intestine through FXR (Inagaki et al., 2005). FGF15 inhibits BA synthesis and promotes hepatic glycogen synthesis. Interestingly, these actions appear to be mediated by ERK rather than Akt (Kir et al., 2011). However, FGF15 can also activate Akt and inhibit Pck1 through the canonical FoxO1 pathway (Shin and Osborne, 2009). The insulin-like actions of FGF15/19 provide a potential alternative pathway to control HGP.

Orphan regulators of HGP

CCAAT/enhancer-binding protein (C/EBP) α and β

Ablation of transcription factor C/EBPα causes lethal neonatal hypoglycemia due to delayed onset of G6pc and Pck1 expression (Wang et al., 1995). But in the adult animal, data are discordant, with some loss-of-function studies showing significant hypoglycemia and liver damage (Lee et al., 1997), while others showed no effect on HGP (Inoue et al., 2004b). Regardless of this discrepancy, there is no evidence of C/EBPα regulation by hormone signaling. It's worth noting that C/EBPα coordinately controls glycogen synthase, Pck1, and G6pc, providing a potential mechanism for hepatic autoregulation. C/EBPβ deletion in mice also causes hypoglycemia in neonates and in fasted adults. This phenotype illustrates the difficulties of extrapolating from effects on gene expression to effects on HGP. In fact, while C/EBPβ gain-of-function activates Pck1 and G6pc in hepatocytes, its loss-of-function lowers HGP by decreasing cAMP production (Croniger et al., 2001), and not by affecting gene expression (Figure 2).

Steroid receptor coactivator (SRC) family

Histone acetyltransferases SRC-1, 2, and 3 coactivate several nuclear receptors and transcription factors. Hepatic expression of SRC-1 and SRC-3, but not SRC-2, is induced by fasting. Despite apparent functional promiscuity in cell culture experiments, different SRC knockout mice have distinct phenotypes. Ubiquitous and liver-specific Src1 knockout mice are hypoglycemic due to reduced HGP, associated with decreased expression of Pck1, Fbp, and pyruvate carboxylase, but not G6pc (Louet et al., 2010). Mechanistic studies revealed that SRC-1 coactivates C/EBPα and PGC-1α, and is required for Cebpα and Foxo1 expression in liver (Louet et al., 2010), consistent with a role in controlling gluconeogenesis during the fed-to-fasting transition. SRC-2 regulates G6pc expression by coactivating retinoid-related orphan receptor α, and hepatic SRC-2 loss-of-function led to fasting hypoglycemia, owing to low G6pc expression, and potential secondary effects of hepatic BA accumulation due to defective secretion into the gut (Chopra et al., 2011). The role of SRC-3 in HGP has not been determined. To establish that these interesting observations have physiological relevance, it will be necessary to study whether hormones and nutrients regulate the activity of these coactivators.

Sirtuin-1

The NAD⁺-dependent protein deacetylase Sirt1 can activate HGP by deacetylating its substrates PGC-1α, FoxO1, STAT3, and FXR (Schwer and Verdin, 2008). Changes of hepatic glucose metabolism in liver-specific Sirt1 knockouts are unremarkable (Chen et al., 2008), but this might reflect opposing effects of other Sirt1-dependent processes, such as CRTC2 degradation (Liu et al., 2008), SHP induction (Wei et al., 2011), and AMPK activation (Hou et al., 2008). In fact, Sirt1 ablation lowered HGP in insulin-resistant mice lacking both Irs1 and Irs2, restoring the post-prandial suppression of Pgc1α and Pck1. And acute knockdown of Sirt1 in liver of rodents (Erion et al., 2009b) reduced fasting glycemia, decreased Pck1, G6pc, and Fbp, and increased glucokinase. Conversely, Sirt1 overexpression in liver increased gluconeogenic gene expression during fasting (Rodgers and Puigserver, 2007), but ubiquitous overexpression reduced fasting glycemia and improved glucose tolerance in obese mice, despite constitutive deacetylation of PGC-1α and FoxO1 in liver (Banks et al., 2008), owing possibly to decreased hepatosteatosis and increased adiponectin levels. Under conditions of nutrient excess, these indirect pathways might trump Sirt1's direct gluconeogenic effects.

Circadian pacemakers

Circadian oscillations in glucose and lipid metabolism are well documented and disruptions of such temporal regulation are associated with metabolic diseases. Core clock components regulate rhythmicity and amplitude of HGP, including Clock, Bmal1, Period2, Cryptochrome (Cry) 1 and 2, and Rev-Erbα (Bass and Takahashi, 2010). The liver clock is regulated by cellular energy sensors such as AMPK, which modulates Cry1 degradation (Lamia et al., 2009), and NAD⁺, which regulates Sirt1- and poly [ADP-ribose] polymerase-1-dependent posttranslational modifications of clock components (Nakahata et al., 2008). PGC-1α and GR (So et al., 2009) can also modulate components of the liver clock. Additional work is necessary to dissect the significance of these pathways in diabetes pathogenesis and treatment.

Direct vs. indirect mechanisms

Central nervous system effects on HGP in rodents

In rodents, the direct effects of insulin in liver are necessary, but not sufficient to inhibit HGP (Okamoto et al., 2005). Insulin receptor signaling in hypothalamic neurons affects HGP by activating K_ATP channels and suppressing flux through G6PC, but not PCK1 (Obici and Rossetti, 2003). The site of insulin action includes orexigenic (appetite-promoting) NPY/AgRP neurons (Konner et al., 2007) as well as anorexigenic POMC neurons (Lin et al., 2010). The mechanism may also involve activation of STAT3 via IL-6 signaling (Inoue et al., 2006), providing a potential link with inflammatory changes that are associated with insulin resistance.

Whether neuronal control of HGP is unique to rodents remains disputed. HGP rates in rodents are ~10-fold higher than humans, indicating that potential CNS effects might go undetected in human studies. Direct delivery of insulin to the CNS of dogs doesn’t affect HGP, raising the possibility that the “CNS effect” on HGP reflect a failure to properly replace basal portal insulin levels during glucose clamps (Edgerton et al., 2006). However, it should be noted that insulin fails to suppress HGP in mice lacking peripheral, but not hepatic insulin receptors. In these mice, the porto-hepatic insulin gradient is likely ablated by the lack of receptor-mediated insulin clearance in tissues exposed to the systemic circulation, and portal insulin levels should suffice to inhibit HGP–but don't (Okamoto et al., 2005).

Free Fatty Acids, adipokines, and HGP

FFA uptake into the liver is thought to impair HGP regulation by insulin (Lewis et al., 1997). However, the rise of FFA in insulin resistance antedates fasting hyperglycemia–the hallmark of increased HGP. And the onset of hyperglycemia is not associated with worsening FFA profiles. Furthermore, manipulations of circulating FFA elicit similar changes in gluconeogenesis in type 2 diabetics and non-diabetic subjects (Boden et al., 2001). In mice lacking hepatocyte insulin receptors, acute lowering of FFA fails to suppress HGP (Fisher and Kahn, 2003). Therefore, FFA likely contribute to render HGP refractory to insulin, but are unlikely to be the inciting factor. Adding to the complexity of this signaling mechanism, hypothalamic sensing of circulating FFA regulates HGP, and may counteract the direct effect of FFA on hepatocytes (Lam et al., 2005). Whether FFA are the bellwether of other adipocyte secretory products that impair insulin action on HGP remains controversial. In vivo regulation of leptin and adiponectin by insulin requires prolonged incubations. Thus, neither is likely to contribute to the rapid hormonal regulation of HGP, but might contribute to setting a basal HGP tone.

Renal and intestinal gluconeogenesis

Renal epithelial cells produce glucose through a process regulated by both insulin and pH. Cultured renal epithelial cells lose the ability to respond to insulin, which can be restored by reconstituting FoxO1 (Nakae et al., 2001). The small intestine also expresses G6pc and makes a contribution to systemic glucose production that may become more significant in starvation and diabetes. Intestinal gluconeogenesis has been implicated in nutrient sensing in the portal vein (Troy et al., 2008). These observations need to be interpreted with caution because of the technical challenges in accurately measuring glucose production in tissues with highly active glucose utilization such as kidney and intestine.

Controversial areas

Flux control vs. gene transcription

As indicated, a creative tension exists between the physiologists, who never fail to remind us that flux control trumps transcriptional control of HGP, and the cell/molecular biologists, who have been tenaciously chasing hormone-regulated transcription factors as the holy grail of HGP. Based on the data reviewed here, the truth appears to lie somewhere in between: in vivo, it takes ~30 min to detect effects on glycogenolysis and gluconeogenesis, and while the latter are unrelated to changes in Pck1 mRNA–let alone protein–the former show a striking correlation with G6pc mRNA and FoxO1 phosphorylation (Ramnanan et al., 2010). Hence, we propose that flux control is important in the first 30 min, but gene transcription kicks in earlier than formerly surmised.

How many genes mediate hormonal effects on HGP?

HGP is a complex and genetically heterogeneous process that cannot be subsumed under a single mechanism. Nonetheless, as illustrated by the litany of “buts”, “unclears”, and “unknowns” that graces this overview, most of the genes proposed to play a role in this process fail one or more tests of physiologic relevance. Among the reasons for this discrepancy are: (i) Obliviousness to–and for younger researchers unawareness of the “Exton & Park” criteria. (ii) Variations of mRNAs encoding glucogenic enzymes, let alone changes in reporter gene activities in hepatoma cells, do not portend effects on HGP–the latter should be tested directly. (iii) Genetic ablation experiments resulting in fasting hypoglycemia don't necessarily indicate a physiologic role in hormonal regulation of HGP, given the latter’s redundancy. (iv) Gain-of-function experiments with transcription factors and coactivators should be interpreted cautiously, as they are especially prone to artifacts. (v) Physiologically relevant conclusions on the role of any given gene product should be supported by both acute and chronic manipulations in vivo. Acute transduction of the liver with DNA- or RNA-based reagents by direct delivery, while expeditious, often results in hepatocyte damage, itself a regulator of HGP. As a result, interpretation of such data is problematic. (vi) Conversely, compensatory mechanisms (e.g., glucose cycling) may obfuscate the interpretation of gene knockouts. A useful approach in this regard is to study induction of HGP at birth, as it first acquires hormone responsiveness (Girard et al., 1992). (vii) Many genes affect HGP indirectly, through effects on hepatocyte function or inter-cellular communication that are not involved in physiologic hormonal or nutrient regulation; to determine their relevance to physiologic conditions and disease states, it’s helpful to study their post-translational regulation in response to insulin and glucagon.

Gluconeogenesis and Pck1

Recent studies have led to a reassessment of the role of PCK1 as rate-limiting enzyme in gluconeogenesis. Liver-specific Pck1 knockouts show a blockade of lactate and amino acid-derived gluconeogenesis but maintain normal fasting glycemia, possibly owing to increased gluconeogenesis from glycerol and reduced glucose utilization (She et al., 2003). Ex vivo studies in mouse livers with various levels of Pck1 ablation show a tight correlation between PCK1 activity and TCA cycle flux, but weak control by PCK1 over gluconeogenic capacity (Burgess et al., 2007). The level of HGP regulation exerted through Pck1 transcription in vivo is also modest, as gluconeogenesis can be suppressed without changes in Pck1 expression in dogs (Ramnanan et al., 2010), and liver biopsies from patients with type 2 diabetes fail to demonstrate changes in Pck1 (Samuel et al., 2009). In vitro, PCK1 can be inhibited by acetylation independent of changes in protein levels (Lin et al., 2009), but the physiological significance of this finding in liver metabolism hasn't been determined.

Therapeutic implications

How does metformin work?

Several explanations for the effect of metformin to decrease HGP have been set forth. The original hypothesis, that it does so through activation of AMPK has not been borne out by genetic ablation of AMPK or its kinase LKB1 that, if anything, sensitizes to metformin action on plasma glucose levels (Foretz et al., 2010). Another study found that metformin was unable to lower glycemia in liver LKB1-deficient mice, but this study didn’t measure HGP (Shaw et al., 2005). Metformin reduces cellular respiration in hepatocytes by inhibiting mitochondrial respiratory chain complex I (El-Mir et al., 2000), and might thus blunt gluconeogenesis by reducing intracellular ATP. Interestingly, thiazolidinediones and berberine are also mild inhibitors of respiratory chain complex I (Turner et al., 2008). Other potential mediators include elevations in AMP/ATP or NAD⁺/NADH ratios, or reactive nitrogen species (Fujita et al., 2010). The alternative explanation that metformin decreases HGP via CBP phosphorylation by PKCλ, leading to decreased CREB/CRTC2 complex formation (He et al., 2009), is unlikely considering the lack of effect on HGP of CRTC2 (Le Lay et al., 2009) or PKCλ knockouts (Matsumoto et al., 2003). Metformin has been shown to decrease mRNA levels of transcription factor KLF-15, whose targets include G6pc and Pck1 (Takashima et al., 2010), but the signaling pathways by which metformin regulates gene expression are unknown. In sum, it’s unlikely that metformin has an obligate target. It’s more likely to act by lowering ATP levels, with pleiotropic consequences. The role of metformin transporters and their complex genetic variation in determining metformin sensitivity and failure remains underscrutinized.

Future prospects

Is HGP inhibition the right approach to diabetes treatment? Arguably, increased HGP is a “late” defect in disease progression, and it’s conceivable that preventive treatments (for example, increasing glucose disposal or enhancing β-cell function) will offset the need to intervene on this aspect of liver dysfunction in diabetes. On the other hand, other pathogenetic mechanisms (increased liver fat content and secretion of VLDL-rich lipoproteins) will continue to require targeting the liver in diabetes (Kim-Muller and Accili, 2011). Thus, we are unlikely to write off the liver as a site of action of antidiabetic medications any time soon. Among the mechanisms that have been explored to sensitize the liver to insulin, the key obstacle has been the potential for hypoglycemia, an inevitable concern with drugs that concurrently inhibit HGP and enhance glucose disposal.

As glucagon levels are inappropriately elevated in diabetes, inhibition of the glucagon receptor pathway is an attractive therapeutic path. Acute inactivation of the glucagon receptor (GcgR) lowers glucose in diabetic mice, while GcgR knockout mice are hypoglycemic and resistant to STZ-induced diabetes (Conarello et al., 2007). Inhibitors of glucagon/GcgR action have shown promise in diabetes treatment, but their long-term impact on lipid and amino acid metabolism is yet unknown.

Sirt1 activators lower glycemia in diabetic rodents, but are tainted by controversies over their bona fide Sirt1 activation and in vivo efficacy (Pacholec et al., 2010). Activators of the xenobiotic receptor, constitutive androstane receptor (CAR) lower hyperglycemia and body weight in mice (Dong et al., 2009), but the role of CAR in drug metabolism and thyroid hormone metabolism may hamper their utility as anti-diabetic compounds. Inhibition of PDK4 or FBPase has a long and checkered history dating back to dichloroacetate. It remains challenging to fine-tune glucose flux without causing irreversible hypoglycemia or contributing to lactic acidosis. Buoyed by new devices and a broader acceptance of parenteral treatments, peptide mimetics of insulin action, or “selective insulin sensitizers” deserve scrutiny (Kim-Muller and Accili, 2011).

Conclusions

Key biochemical, cellular, and integrated physiological mechanisms by which hormones regulate HGP have been clarified. But few players thus identified are tractable drug targets. Thus, given its efficacy, tolerability and inexpensiveness, metformin is likely to remain the mainstay of therapy, despite its limitations. New anti-diabetic agents should therefore either reduce HGP by a mechanism distinct from metformin’s, or target different aspects of diabetes pathophysiology (impaired glucose uptake, excessive lipolysis, increased tissue fat content, β-cell dysfunction, elevated atherogenic lipoproteins), or improve safety vis-à-vis lactic acidosis or use in renal failure, or outperform metformin’s durability. The studies reviewed here will help chart the way ahead.

Highlights.

• Hormones control glucose production by modulating enzyme flux and gene transcription

• The Akt–Foxo pathway accounts for a large component of insulin-dependent HGP

• Additional pathways (Crct2, Fxr, Src) show promise as mediators of Foxo-independent effects

• Alternative approaches to reducing excessive glucose production in diabetes are needed