Chronic glucokinase activator treatment activates liver Carbohydrate response element binding protein and improves hepatocyte ATP homeostasis during substrate challenge

To test the hypothesis that glucokinase activators (GKAs) induce hepatic adaptations that alter intra‐hepatocyte metabolite homeostasis.


| INTRODUCTION
The recognition of the unique role of glucokinase in control of blood glucose homeostasis through its dual function in liver and pancreatic islets led to the development of glucokinase activators (GKAs) as candidate drugs for type 2 diabetes. 1,2 GKAs were identified that were very effective at lowering blood glucose in animal models and during short-term treatment in human diabetes. [3][4][5] However, in phase 2 clinical trials, efficacy declined during chronic therapy. [6][7][8] Three hypotheses have been proposed for this decline in efficacy. [3][4][5] The most widely held hypothesis is that liver glucokinase activation leads to increased hepatic production of triglycerides, aggravating fatty liver disease. [9][10][11][12] A second hypothesis is that liver glucokinase activation raises hepatic glucose 6-phosphate (G6P) and downstream metabolites with consequent activation of metabolite-responsive mechanisms, including the transcription factor Carbohydrate response element binding protein (ChREBP) 5 and induction of its target genes such as the Carbohydrate response element binding protein beta-isoform (ChREBP-β), Gckr which encodes the glucokinase inhibitor protein GKRP and G6pc which encodes glucose 6-phosphatase which degrades G6P. 13,14 A third hypothesis is that chronic overstimulation of pancreatic β-cell glucokinase causes β-cell failure by glucotoxicity. 4 The latter, together with the risk of hypoglycaemia by over-stimulating insulin secretion, could be mitigated by development of liver-selective GKAs. 2 The arguments supporting the first hypothesis are: first, that blood triglycerides were moderately raised in some phase 2 studies with GKAs 6,7 ; second, elevation in liver triglycerides occurred in some preclinical models [9][10][11] ; and third, that a common variant in the GCKR gene associates with raised blood lipids and fatty liver disease, possibly through raised glucokinase activity. 12 The arguments against are: first, that loss of GKA efficacy in humans also occurred in the absence of raised blood triglycerides 8 ; second, that several preclinical GKA models did not show raised blood or liver triglycerides [15][16][17][18][19][20][21] ; and third, that metabolites of some GKAs that caused hepatic steatosis were later found to have target-independent hepatotoxicity. 22 A key caveat to the second hypothesis is that hepatic changes result from induction of inherent adaptive mechanisms involving repression of liver Gck and/or induction of ChREBP target genes to safeguard intrahepatic metabolite homeostasis 23,24 rather than cytotoxicity by excess lipid production. The aim of this study was to test the second hypothesis in C57BL/6 mice treated for 8 weeks with AZD1656, a GKA with a good safety record. 25 We chose 8-week exposure because the decline in glycaemic efficacy with AZD1656 became evident within this interval 8 and we used a standard rodent diet because raised hepatic triglyceride levels by GKA therapy is more likely to manifest at low hepatic fat. 9,10 We show that hepatocytes from mice treated chronically with AZD1656 have raised basal mRNA levels of ChREBP-β and AMP-activated protein kinase catalytic subunit (AMPK-α) and improved ATP homeostasis during substrate challenge, together with attenuated induction of ChREBP target genes and the catalytic unit of AMPK (Prkaa1,2). The improved ATP homeostasis supports a beneficial role for ChREBP activation in the hepatic adaptive response to the GKA.

| Glucokinase activators
AZD1656 has been described 26 and was provided by AstraZeneca (Cambridge, United Kingdom). Synthesis of PF-04991532 27 (+) and (−) enantiomers for ex vivo studies is described in the Supplementary methods in Appendix S1 and PF-04991532 for in vivo studies was from Tocris Bioscience (C10363).

| Animals
Animal procedures conformed to Home Office Regulations and were approved by Newcastle University Animal Welfare Ethics Review Board.

| Experimental design
For the acute dose study (Figure 1), food was withdrawn 4 hours before glucose gavage. Blood was sampled after 2-hour fasting, mice were gavaged with AZD1656, blood was sampled after 2 hours, followed by glucose gavage (2 g/kg body weight) and blood sampling at the times indicated. For PF-04991532, mice were gavaged with the drug 60 minutes before glucose gavage. Mice were culled after the last blood sample and the livers snap-frozen in liquid N 2 for RNA extraction and mRNA analysis (quantitative RT-PCR). For the chronic studies (Figure 3) mice received a powdered diet (SDS, 801723, CRM) without or with GKA added to the diet at the doses indicated for the diet consumed (4 g/d). Free-feeding blood samples were collected from the tail vein for glucose, insulin and triglyceride analysis. For drug tolerance tests mice were fasted for 2 hours, a blood sample was collected followed by gavage with AZD1656 and blood was sampled after 2 hours. For the 4and 8-week studies mice were either culled (n = 6) for whole liver analysis or used for hepatocyte isolation (n = 3-5).

| Blood and liver analysis
Blood glucose was determined with a Roche Glucometer; plasma insulin by ELISA (Mercodia #10-1247-01); plasma triglycerides and liver triglycerides, extracted by the method of Bligh and Dyer, 28 were determined with a WAKO kit (290-63 701). Glucokinase activity (Figure 3D,L) was determined on liver 100 000-g supernatant by kinetic assay at 100 and 0.5 mM glucose as described 29 and GKA efficacy ( Figure 2A,B) using the same assay at 0.5 mM glucose. 29

| Hepatocyte studies
Hepatocytes were isolated by collagenase perfusion of the liver, seeded in 24-well plates and cultured overnight in serum-free minimum essential medium (MEM). 30 Table S1). Metabolites are expressed as nmol/mg cell protein and mRNA levels as percentage of respective 5 mM glucose control and also normalized to vehicle-treated controls. The effects of a GKA on the liver in vivo could be direct on the hepatocyte or indirect through altered blood insulin or glucagon 33,34 caused by the GKA targeting the pancreas. We compared the effects of AZD1656 with those of PF-04991532, a liver-selective GKA, 20 on liver gene expression after a single exposure to the GKA followed by an oral glucose tolerance test. AZD1656 (2-9 mg/kg), administered 2 hours before the oral glucose tolerance test, lowered blood glucose and glucose excursion ( Figure 1A,C,D) and raised insulin ( Figure 1E,G,H), whereas PF-04991532 (100 mg/kg) modestly decreased blood glucose ( Figure 1B,C) with no effect on insulin ( Figure 1F,G). Liver mRNA levels for various ChREBP target genes including ChREBP-β, G6pc, Pklr, Acly, Acac and Gpd2 were increased by PF-04991532 and AZD1656 ( Figure 1I).

| Induction of ChREBP target genes by (+)PF-04991532 but not (−)PF-04991532 in hepatocytes
To test whether liver gene regulation by the GKAs in vivo is consequent to glucokinase activation, we compared the two enantiomers of PF-04991532 on glucokinase activity and gene regulation in mouse F I G U R E 1 Effect of single exposure to AZD1656 or PF-04991532 on blood glucose and insulin and liver gene expression. Male C57BL/6 mice were treated with vehicle or AZD1656 (A,E) at the doses indicated 2 hours before an oral glucose tolerance test (GTT) or with PF-04991532 (B,F) 1 hour before the oral GTT. C,G, Changes in blood glucose (C) and insulin (G) after treatment with AZD1656 or PF-04991532 before glucose gavage. D, Blood glucose, area under the curve. H, Plasma insulin, area under the curve. I, Liver mRNA levels for the indicated genes normalized to TATA-Box Binding Protein (TBP) and to vehicle controls. Means ± SEM, n = 8, *P <0.05 relative to vehicle, except for C,G relative to time zero. https://doi.org/10.25405/data.ncl.12550508 hepatocytes ex vivo. In the glucokinase assay (−)PF-04991532 was inactive whereas (+)PF-04991532 (10 μM) was a potent GKA ( including ChREBP-β, whereas the (−) enantiomer had no effect ( Figure 2C), confirming that gene induction is consequent to glucokinase activation. AZD1656 caused similar gene induction but mostly with lower efficacy ( Figure 2D), consistent with the lower glucokinase activation ( Figure 2B).

| Comparison of GKAs with substrate challenge on cell metabolites and gene regulation in hepatocytes
The induction of ChREBP target genes is linked to raised metabolites, such as G6P and other phosphate esters. 23,31,32 We next compared the effects of the GKAs on cell metabolites [ATP,G6P, glycerol 3-P (G3P)] and ChREBP target gene induction with substrate challenges known to raise G6P (eg, high glucose) or more distal metabolites. For the latter we used xylitol (2 mM) which is metabolized independently of glucokinase and, similarly to fructose, lowers ATP because of excessive accumulation of phosphate esters, mainly as G3P. 36,37 We used an inhibitor of G6P hydrolysis (S4048, a chlorogenic acid derivative) 31,32 to further raise G6P with 25 mM glucose and an inhibitor of NADH shuttling to mitochondria [amino-oxyacetate (AOA)] to raise G3P with xylitol 30 ( Figure 2E). PF-04991532 at 5 mM glucose raised G6P more than AZD1656 or 25 mM glucose but less than 25 mM glucose with S4048 ( Figure 2F). However, xylitol, either alone or with AOA, had modest effects on G6P but markedly raised G3P ( Figure 2F), as expected. 30,36,37 Cell ATP was not affected by the GKAs but was F I G U R E 3 Chronic effects of AZD1656 on blood glucose, liver triglycerides and glucokinase activity. A-D, C57BL/6 mice were treated for 1 week without or with AZD1656 (at 0.3, 1, 3 and 9 mg/kg body weight): A, blood glucose; B, plasma insulin at the times indicated; C, liver mRNA levels; D, liver glucokinase activity. Means ± SEM for n = 6, *P <0.05 vs vehicle. E-L, C57BL/6 mice were treated for 4 or 8 weeks without (0) or with AZD1656 (1 or 3 mg/kg body weight): E, blood glucose; F, plasma insulin; G, acute (120 min) blood glucose-lowering by AZD1656 (drug tolerance test); H, blood triglycerides (TAGs); I, liver triglycerides (TAGs); J, liver mRNA after 4 weeks; K, liver mRNA after 8 weeks; L, liver glucokinase activity. Means ± SEM for n = (10-12, E-H) or n = 6 (I-L). https://doi.org/10.25405/data.ncl.12550574 lowered by xylitol and by high glucose with S4048 ( Figure 2G). ATP depletion by xylitol is analogous to fructose challenge and lowers mitochondrial ATP production by trapping phosphate in G3P. 37 ChREBP-α mRNA was modestly raised at the highest G6P ( Figure 2H), whereas ChREBP-β, a downstream target of ChREBP-α, 38 was induced two-to five-fold at raised G6P or G3P and was decreased with AOA alone (Figure 2I). Gck, which is not a ChREBP target gene, 31 was repressed by raised G6P but not G3P ( Figure 2J). Gckr, Pklr, G6pc, Txnip and FGF21, which are ChREBP target genes, were induced in association with raised G6P or G3P ( Figure 2K-O), with G6pc and FGF21 more strongly induced by G6P and G3P, respectively. Cumulatively, induction of ChREBP target genes is linked to raised hexose-P or triose-P, and GKAs raise predominantly hexose-P rather than triose-P, without lowering ATP.

| Effects of chronic exposure to AZD1656 on liver gene expression
We next performed a 1-week chronic study with four doses of AZD1656 (0.3, 1, 3 and 9 mg/kg) to select two doses for the 8-week study. Blood glucose-lowering by AZD1656 was dose-dependent, blood insulin was not significantly increased, ChREBP target gene mRNA levels were increased dose-dependently at or above 1 mg/kg and total liver glucokinase activity at or above 3 mg/kg ( Figure 3A-D).
Because the liver Gck gene is induced by raised insulin or by lowered glucagon, 24 we infer that the raised total glucokinase at 3 to 9 mg/kg suggests liver exposure to a raised insulin/glucagon ratio. Accordingly, we selected 1 and 3 mg/kg AZD1656 for further study.
During 8-week treatment with AZD1656 body weight gain was unchanged ( Figure S1), blood glucose-lowering was maintained and insulin was modestly but not significantly raised ( Figure 3E,F). The response to an intra-gastric load of AZD1656 (drug tolerance test [DTT], 120 minutes) showed maintained blood glucose lowering at 4 and 8 weeks ( Figure 3G) and stimulation of insulin secretion at 4 weeks but not 8 weeks ( Figure   S1). Blood triglycerides and liver triglycerides were unchanged ( Figure 3H, I). Liver mRNA levels of ChREBP target genes that were induced by single GKA exposure ( Figure 1I) or 1-week treatment ( Figure 3C) were mostly similar to vehicle after 4 to 8 weeks, except for ChREBP-β, which was raised by AZD1656 ( Figure 3J,K). Liver total glucokinase activity was raised by AZD1656 after 4 weeks and modestly decreased by 1 mg/kg AZD1656 (18%) at 8 weeks ( Figure 3L). Cumulatively, during 8-week treatment with AZD1656, blood glucose-lowering efficacy was maintained and liver glucokinase activity was modestly decreased by 1 mg/kg AZD1656. Protein levels for GKRP and mGPDH were modestly raised with the higher AZD1656 dose at 8 weeks ( Figure S2).

| Improved ATP homeostasis and blunted ChREBP-β induction in hepatocytes from AZD1656treated mice
To test for changes in metabolite homeostasis, hepatocytes were isolated from the three groups of mice after the 4-or 8-week treatments and cultured overnight at 5 mM glucose, followed by substrate challenge for cell metabolite and gene expression analysis. For the 4-week treatment ( Figure 4A-H) we used the same substrates as in Figure 2F,G.
Hepatocytes from AZD1656-treated mice had lower basal G6P and G3P at 5 mM glucose but similar G6P elevation with substrate challenge ( Figure 4A,B). Basal ATP at 5 mM glucose was similar across groups ( Figure 4C). However, the fractional lowering of ATP by xylitol was greater in the untreated mice than in the AZD1656 treatments (67%, 82% and 90%, 0, 1 and 3 mg/kg, respectively). Likewise, with 25 mM glucose + S4048, fractional ATP-lowering was attenuated by AZD1656 treatments (82%, 87% and 97%; Figure 4D), indicating resilience to ATP-lowering. Basal mRNA levels were mostly similar across treatment groups except for G6pc and Gck ( Figure 4E), which are known to be regulated oppositely by the direct GKA effect on liver versus indirect effect through pancreatic targeting. 24 To assess the gene To test for candidate mechanisms for the improved ATP homeostasis during substrate challenge, we determined mRNA levels of the catalytic subunits of AMPK (Prkaa1 and Prkaa2) because Prkaa2 was identified as a putative ChREBP target by chromatin immunoprecipitation sequencing. 39 We found raised basal Prkaa1,2 mRNA in hepatocytes from AZD1656 (3 mg/kg)-treated mice relative to vehicle ( Figure 5A,B) and 1.5-fold induction by ex vivo substrate challenge raising G6P or G3P in hepatocytes from vehicle-treated mice but not from AZD1656 (3 mg/kg)-treated mice ( Figure 5C,D). This attenuated Prkaa1,2 response to substrate challenge parallels the ChREBP-β response ( Figure 4F,N), supporting AMPK as a functional target gene of ChREBP in the adaptive response to AZD1656.

| DISCUSSION
In the present study we explored the hepatic adaptations in C57BL/6 mice during 4 to 8 weeks of treatment with AZD1656, a GKA with an established safety record 25 that has been studied preclinically 18,19 and clinically, and showed a decline in glycaemic efficacy at approximately 4 to 8 weeks. 8 Previous work with other GKAs in rat hepatocytes in vitro had shown induction of G6pc and Pklr, which are ChREBP target genes and repression of Gck, which is not a ChREBP target. 13,14,31 Here we tested the hypothesis that chronic exposure to AZD1656 causes hepatic adaptations linked to ChREBP activation and liver Gck repression. Various sets of evidence support adaptive changes consequent to ChREBP activation. However, we found very modest changes in liver total glucokinase activity. During 8-week exposure to AZD1656, blood glucose-lowering efficacy was maintained and there were no changes in blood or liver triglycerides.
F I G U R E 4 Improved ATP homeostasis and attenuated Carbohydrate response element binding protein (ChREBP) target gene induction in isolated hepatocytes after 4 and 8 weeks of treatment with AZD1656. C57BL/6 mice were treated for 4 weeks (A-H) or 8 weeks (I-P) without (0) or with AZD1656 (1 or 3 mg/kg body weight) and used for the studies in Figure 3 E-H before they were killed for hepatocyte isolation. After overnight culture of hepatocytes, parallel incubations were performed in minimum essential medium with the substrates indicated (see legend to We compared AZD1656 with PF-04991532, a liver-selective GKA, 20 to identify AZD1656 doses with optimal liver targeting and minimal targeting of pancreatic islet glucokinase, to minimize confounding effects from increased insulin exposure, which has converse effects on liver Gck and G6pc, to those expected from direct GKA effects on liver. 13 We show that AZD1656 and PF-04991532 induce several ChREBP target genes, including ChREBP-β, after acute exposure in vivo and in mouse hepatocytes ex vivo (Figures 1 and 2). These effects are unique to the active enantiomer of PF-04991532 that functions as a GKA, and associate with the raised G6P. At doses of AZD1656 and PF-0499153 causing comparable ChREBP activation, AZD1656 was far more effective at lowering blood glucose. This greater glycaemic efficacy of AZD1656 is best explained by stimulatory effects of the GKA on insulin secretion and possible inhibitory effects on glucagon secretion, 33,34 resulting in liver exposure to a raised insulin to glucagon ratio. Although we confirmed that Gck mRNA levels change inversely with raised G6P in mouse hepatocytes ( Figure 2) similar to rat hepatocytes, 13,31 we found very modest lowering of total glucokinase (18%) after 8 weeks with 1 mg/kg AZD1656 and no effect at 3 mg/kg AZD1656. Increased hepatic exposure to a raised insulin to glucagon ratio would promote induction of the Gck gene and counterbalance the repression through raised G6P and could explain the increase in total glucokinase activity in the 1-week study by higher doses of AZD1656 and possibly also the modest lowering of glucokinase by 1 mg but not 3 mg AZD1656 after 8 weeks. We infer that ChREBP activation has a more prominent role than Gck repression in the chronic hepatic adaptations to AZD1656.
ChREBP (encoded by Mlxipl) is a major transcriptional regulator in liver that is activated by high dietary carbohydrate and is frequently described as a "glucose sensor" because it is activated by high glucose. 38 However, the stimulus for its activation is the raised phosphate esters and not glucose itself, as shown by inhibitors of hexokinases and G6P hydrolysis which modulate cell G6P and ChREBP target induction. 23,31,32 ChREBP is expressed as two isoforms (α and β) by alternative splicing of the first exon. 38 ChREBPα protein accumulates in the cytoplasm at low glucose, and is regulated by an inhibitory domain which enables activation by metabolites causing translocation to the nucleus where it activates ChREBP target genes. 38 ChREBP-β lacks the inhibitory domain and is constitutively active in the nucleus and moreover the ChREBP-β promoter is itself a target of ChREBP-α. 38 Accordingly, conditions associated with raised phosphate esters, such as high-fructose diets, cause strong induction of ChREBP-β 38 , making ChREBP-β mRNA a convenient read-out of ChREBP activation. Although ChREBP is often functionally described as a "lipogenic" transcription factor, its wide array of target genes, 39 which include G6pc and Gckr, 14,32 implicate a more complex role in metabolite homeostasis. 23 This is supported by germ-line and liver-F I G U R E 5 Gene expression of Prkaa1 and Prkaa2 is increased by substrate challenge in hepatocytes from vehicle-treated mice ex vivo and by chronic treatment with AZD1656 (3 mg/kg) for 8 weeks. mRNA levels for Prkaa1 (AMPK-α1) and Prkaa2 (AMPK-α2) were determined in the experiments described in Figure 4 in hepatocytes isolated from mice treated for 8 weeks with vehicle or AZD1656 (1 or 3 mg/kg) and then challenged for 4 hours with the substrates indicated as in Figure 4I-P. A-B, mRNA levels for Prkaa1 and Prkaa2 normalized to hepatocytes from vehicle-treated mice in basal conditions (5 mM glucose). C-D. mRNA levels of Prkaa1 and Prkaa2 normalized to the respective controls at 5 mM glucose. Means ± SEM, n = 4, *P <0.05 vs respective control (5G); # P <0.05 vs respective untreated (0 mg/kg). https://doi.org/10.25405/ data.ncl.12550670 selective ChREBP deletion models which have markedly raised phosphate esters and compromised ATP homeostasis, particularly when challenged with dietary fructose. [40][41][42] The hypothesis that ChREBP activation is a component of the chronic effects of AZD1656, predicts changes in metabolite homeostasis in hepatocytes. To test this hypothesis we challenged hepatocytes isolated from mice treated chronically with AZD1656 with substrates that raise G6P or the more distal metabolite G3P. We used xylitol as a surrogate for fructose, because, like fructose, it compromises ATP homeostasis, but does so by raising G3P, 30,36,37 which can be monitored accurately.
Three key findings emerged from the hepatocyte studies on mice treated chronically with AZD165. First, there was lowering of basal G6P at 5 mM glucose but not maximal G6P levels in substrate-challenged conditions. Second, ATP depletion during substrate challenge was attenuated despite sustained elevation in G3P and G6P. Third, induction of ChREBP target genes (including ChREBP-β) by the substrate challenge ex vivo was attenuated despite lack of attenuation of the raised phosphate esters. The improved ATP homeostasis in the absence of attenuation of raised G6P or G3P was surprising. However, ATP homeostasis involves complex recovery mechanisms and moreover, the target genes of ChREBP include some of the seven subunits of AMPK 39 which has an established role in ATP homeostasis, as shown by the greater ATP depletion at raised G6P in hepatocytes from AMPK-KO mice. 43 The attenuated fractional induction of ChREBP-β in hepatocytes from AZD1656-treated mice is only in part explained by higher basal ChREBP-β mRNA. To our knowledge, this study is the first to demonstrate induction of the AMPK-α subunits (Prkaa1 and Prkaa2) at the mRNA level in hepatocytes with raised G6P or G3P. AMPK is an energy sensor but crucially also a negative regulator of lipogenesis. 44 For both Prkaa1 and Prkaa2 the substrate induction was abolished in hepatocytes from the high-dose AZD1656 treatment, in association with raised basal Prkaa1 and Prkaa2 mRNA.
The mechanism(s) by which metabolites of glucose or fructose activate ChREBP is far from understood and is highly unlikely to involve a single metabolite. 23 A tentative conjectural hypothesis to explain the attenuated induction of candidate ChREBP target genes in substratechallenged conditions despite sustained elevation in G6P or G3P, is that ChREBP activation may be a composite function of both raised phosphate esters (beyond the homeostatic range) and of compromised ATP homeostasis (adenine nucleotide phosphorylation potential), to varying degrees depending on the gene. The attenuated induction of some genes in association with improved ATP homeostasis may reflect their stronger regulatory links with ATP homeostasis as compared with raised phosphate esters. An analogous mechanism for composite control by a glucose metabolite and a signal from the electron transport chain, was recently reported for MondoA (encoded by