Pharmacological TNKS inhibition enhances glucose tolerance and insulin sensitivity
To investigate the role of TNKS catalytic activity in metabolism, we treated wild-type male mice with G007-LK, a PARsylation inhibitor specific for TNKS and the closely related TNKS2 . To deliver this compound orally at a daily dose of 10 mg/kg to a typical 40 g mouse that consumed 2.5 g of HFD each day, we impregnated the diet with 0.015% (wt/wt) of G007-LK. This formulation effectively attenuated TNKS activity in vivo as assessed by upregulation of axin1 and downregulation of active (unphosphorylated) β-catenin in adipose tissue (Fig. 1a). Compared with HFD alone, treatment with G007-LK-containing HFD for 6 months did not affect body weight (Fig. 1b), food intake (ESM Fig. 1a), plasma albumin levels (an index of liver synthetic function, ESM Fig. 1b) or intestinal mucosal histology (ESM Fig. 1c), all arguing against overt toxicity. G007-LK treatment did not affect overall adiposity as determined by body composition analysis (Fig. 1c) despite there being a modest reduction in individual adipocyte size (Fig. 1e). Importantly, G007-LK treatment lowered fasting glucose levels (Fig. 1g), improved glucose tolerance (Fig. 1h) and potentiated the hypoglycaemic effect of insulin (Fig. 1i).
Since glucose homeostasis is modulated by adipokines, we investigated the effect of G007-LK on plasma RBP4 (a pro-diabetic adipokine ) and adiponectin. Immunoblotting of plasma showed that during G007-LK treatment RBP4 levels became progressively lower than in controls (Fig. 1j) whereas adiponectin levels persistently exceeded those in controls (Fig. 1k). Since G007-LK did not increase adipose tissue stores of adiponectin mRNA (ESM Fig. 1d) or protein (Fig. 1a), its effect on plasma adiponectin was likely post-transcriptional. Also supporting a post-transcriptional mechanism was the finding that overnight G007-LK treatment at a higher dose (100 mg/kg per day) robustly increased plasma adiponectin (Fig. 1l).
Adipose-selective deletion of the TNKS catalytic domain
The effect of G007-LK on glucose metabolism (Fig. 1g–i) could reflect inhibition of PARsylation in any of the many tissues that express TNKS or TNKS2. However, the rapidity of the effect on plasma adiponectin (Fig. 1l) pointed to adipose tissue as a key target. To confirm this, we set out to genetically inactivate TNKS selectively in adipocytes. First, we used homologous recombination to flox the TNKS gene as depicted in Fig. 2a. Cre-mediated recombination of the floxed allele TNKS
f, by shifting the open-reading frame, was predicted to delete the entire PARP domain (aa 1179–1320 encoded by exons 25–27) and shorten the protein product by 6 kDa. This was confirmed using mice homozygous for the recombinant allele ΔPARP (Fig. 2b). As expected, the recombination deleted the C-terminal epitopes of TNKS (Fig. 2b). This deletion did not disrupt the scaffolding function of native TNKS since the resulting fragment TNKSΔPARP was pulled down by GST-IRAPaa78-101 containing the TNKS-binding motif RxxPDG (Fig. 2b).
To selectively truncate TNKS in adipocytes, we expressed Cre recombinase in TNKS
f/f mice using either the aP2
Salk or the adiponectin promoter [23, 24]. Fig. 2c shows that only aP2-Cre induced efficient TNKS truncation in adipose tissue. Since no truncation was detected in the pancreas, heart, liver or peritoneal macrophages (Fig. 2d), aP2-Cre TNKS
f/f mice were referred to as CITA (catalytically inactive TNKS in adipocytes) in subsequent phenotyping. To corroborate the tissue specificity of this conditional model, we showed (Fig. 2e, f) that the CITA adipose tissue, but not liver, exhibited upregulation of axin1 and downregulation of active (unphosphorylated) β-catenin, the expected consequences of TNKS inactivation.
Female CITA mice have greater glucose tolerance and insulin sensitivity
Female CITA mice fed a chow diet exhibited lower overall adiposity (Fig. 3b) and smaller adipocyte size (Fig. 3d) than wild-type littermates despite upregulation of adipogenic markers including lipogenic genes (Fig. 3e). Plasma resistin and leptin levels did not differ between the CITA and wild-type mice (Fig. 3f, g) whereas plasma adiponectin was significantly increased in CITA mice (Fig. 3h). This increase was a post-transcriptional effect since neither adiponectin mRNA (Fig. 3e) nor protein (Fig. 3h) in the CITA adipose tissue was increased. Importantly, female CITA mice exhibited greater glucose tolerance (Fig. 3i) and insulin sensitivity than their wild-type littermates (Fig. 3j). This glucose phenotype and increased adiponectin in plasma was not discernible in male CITA mice (ESM Fig. 2) or ovariectomised female CITA mice (ESM Fig. 3), a sexual dimorphism that prompted us to further phenotype only female mice.
Liver of female CITA mice on a chow diet is more insulin sensitive
To specify the tissue(s) that mediated the glucose phenotype of female CITA mice on a chow diet, we performed hyperinsulinaemic–euglycaemic clamp studies. We found comparable rates of hepatic glucose production between CITA mice and controls in the basal state (23.1 ± 0.8 vs 23.6 ± 0.8 mg kg−1 min−1, p = 0.72). During insulin infusion, CITA mice tolerated a higher glucose infusion rate than wild-type mice (Fig. 4a) and showed a greater suppression of hepatic glucose production (Fig. 4b) and plasma NEFA levels (Fig. 4c), indicative of increased insulin sensitivity in liver and fat, respectively. This was in contrast to the muscle, where the insulin-stimulated glucose disposal rate indicated comparable insulin sensitivity between the two genotypes (Fig. 4d). To corroborate the above tissue specificity, we harvested various tissues before and after in vivo insulin stimulation to compare insulin signalling. Figure 4e shows that insulin-stimulated Akt phosphorylation in the liver and fat, but not muscle, of the CITA mice was more robust than that in control mice.
The CITA liver exhibited downregulation of the gluconeogenic marker Pdk4 and the inflammatory markers F4/80 (also known as Adgre1) and Mcp-1 (Fig. 4f). It also contained less triacylglycerol than control liver (Fig. 4g), which presumably reflected decreased lipogenesis and increased lipolysis, since the CITA liver showed downregulation of lipogenic genes (Acc, Acs and Acl [Acly]), upregulation of the lipolytic gene Atgl (Pnpla2), and no changes in markers of triacylglycerol export (Apob and Mtp) (Fig. 4f). Collectively, these data indicate that the CITA liver, despite being free from TNKS inactivation (Fig. 2d, f), was indirectly affected by TNKS inactivation in adipose tissue.
Liver of female CITA mice on an HFD is more insulin sensitive
To find out if the metabolic phenotype of lean CITA mice is conserved in the obese state, we fed female mice an HFD. This diet did not alter TNKS abundance (ESM Fig. 4a) or prevent the CITA mutation from blocking Wnt signalling in fat (Fig. 5a) and raising plasma adiponectin levels (Fig. 5e). The liver of HFD-fed CITA mice, like its counterpart from lean mice, showed downregulation of lipogenic genes, inflammatory markers and Pdk4 (ESM Fig. 5a). Importantly, HFD-fed CITA mice also exhibited greater glucose tolerance (Fig. 5f) and insulin sensitivity (Fig. 5g). During hyperinsulinaemic–euglycaemic clamps, the hepatic glucose production in CITA mice tended to be lower than that in wild-type mice in the basal state (20.3 ± 1.5 vs 25.2 ± 1.3 mg kg−1 min−1, p = 0.13) and was significantly more sensitive to suppression by insulin (Fig. 5i). This was in contrast to the findings in fat and muscle, where the two genotypes exhibited comparable insulin sensitivity based on suppression of plasma NEFA levels (Fig. 5j) and stimulation of glucose disposal rate (Fig. 5k). Collectively these data indicate that in diet-induced obese female mice, liver insulin sensitivity was enhanced by inhibition of TNKS-mediated PARsylation in adipose tissue. This phenotype was not attributable to differences in the severity of obesity since body weight and overall adiposity were comparable between the two genotypes (Fig. 5b, c).
Adipose TNKS activity modulates secreted factors to regulate hepatocyte gluconeogenesis
Given the indirect hepatic effect of adipose TNKS inactivation, we hypothesised that this effect was mediated by adipocyte-derived factors whose secretion was regulated by TNKS-mediated PARsylation. To test this hypothesis, we compared ex vivo gluconeogenesis in hepatocytes pre-treated with conditioned media (CM) generated by CITA or wild-type (TNKS
) adipocytes. Figure 6a (bar 3 vs bar 4) shows decreased gluconeogenesis in hepatocytes pre-treated with CM from female CITA vs wild-type mouse adipocytes. This anti-gluconeogenic effect was not observed in CM from male CITA mouse adipocytes (Fig. 6a, bar 1 vs bar 2), a sexual dimorphism that mirrored the female specificity of CITA phenotype in vivo. To corroborate the effect of genetic TNKS inactivation, we also treated female wild-type mouse adipocytes with G007-LK or vehicle (DMSO), and applied the resulting CM to hepatocytes. Fig. 6b (bar 1 vs bar 2) shows that CM from G007-LK-treated adipocytes imparted an anti-gluconeogenic effect on hepatocytes. This effect was not due to G007-LK carryover from adipocytes to hepatocytes, since direct G007-LK treatment of hepatocytes did not affect gluconeogenesis (Fig. 6c). Furthermore, we treated 3T3-L1 adipocytes with IWR-1, a structurally distinct TNKS inhibitor , and rinsed off the inhibitor before CM collection. Figure 6d shows that CM from adipocytes pre-treated with IWR-1 also exhibited an anti-gluconeogenic effect. Collectively, these ex vivo studies indicate that upon genetic or pharmacological TNKS inactivation, adipocytes used secreted factors to communicate with hepatocytes to reduce gluconeogenesis.
A candidate mediator of this TNKS-modulated adipocyte–hepatocyte crosstalk is adiponectin since TNKS inactivation increased plasma adiponectin and concomitantly decreased hepatic glucose output (Figs 3, 4 and 5). We therefore investigated whether adiponectin knockout could abrogate the impact of G007-LK on adipocyte–hepatocyte crosstalk. In line with the known anti-gluconeogenic effect of adiponectin , we found increased gluconeogenesis in hepatocytes exposed to CM from adiponectin-knockout vs wild-type female mouse adipocytes (Fig. 6b, bar 3 vs bar 1). However, G007-LK treatment of adipocytes, regardless of adiponectin genotype, conferred a comparable anti-gluconeogenic effect to the CM (bars 2 and 4). This ex vivo assay therefore suggested that upon short-term G007-LK treatment, adipocytes alter the secretion of factors apart from adiponectin to reduce gluconeogenesis in hepatocytes.