Metformin Inhibits Growth Hormone–Mediated Hepatic PDK4 Gene Expression Through Induction of Orphan Nuclear Receptor Small Heterodimer Partner

Male and female wild-type C57BL/6 J mice and SHP-null mice were used in experiments. The study used null male mice for PDK4 (27,30), STAT5 (31), and SHP (16) (weight, 20–30 g; age, 12 weeks). Recombinant human GH (ProSpec, Rehovot, Israel) was administered intraperitoneally (i.p.) to fed mice at 2 μg/g body weight; metformin was given by oral gavage to fed mice at 200 mg/kg body weight. Protocols were approved by the Institutional Animal Care and Use Committee of the Korea Research Institute of Bioscience and Biotechnology, and experiments were performed in accordance with National Institutes of Health animal research standards.

Liver tissues were isolated, freeze-clamped, and extracted with 80% methanol, as described previously (32).

Western blot analysis from primary rat hepatocytes and mouse liver tissues were performed according to methods described previously (10). The membranes were probed with phospho-ATM, ATM, phospho-AMPK, AMPK phospho-STAT5 (Tyr694, Cell Signaling Technology), STAT5, SHP (H-160), GHR, β-actin (Santa Cruz Biotechnology), and phospho-PDH (Ser293, EMD Chemicals) and then developed using an enhanced chemiluminescence Western blot detection kit (Amersham Bioscience).

The small interfering (si)RNAs of human SHP (siSHP) and the plasmids encoding for the constitutively active form of AMPK (CA-AMPK) and the dominant-negative mutant AMPK (DN-AMPK) were described previously (10,33). The reporter plasmids STAT5-Luc and TEL-JAK2 were provided by Dr. Minho Shong (Chungnam National University School of Medicine), and the mouse PDK4-Luc was provided by Drs. Minho Shong and Robert A. Harris (Indiana University School of Medicine), respectively. STAT5 cDNA (34) was provided by Dr. Gregorio Gil, (Medical College of Virginia, USA). SHP cDNA, glutathione S-transferase (GST), and GST-fused SHP (GST-SHP) were described previously (10).

Adenoviruses (Ad) encoding full-length human SHP, siRNA for SHP (Ad-si SHP), c-myc–tagged DN-AMPK, and CA-AMPK have been previously described (11,33).

HepG2 and primary rat hepatocytes were cultured as previously described (11). Transient transfections were performed in accordance with the method described previously (11).

In vivo interaction assay between STAT5 and SHP was performed as described previously (10). For coimmunoprecipitation assay, total protein liver extract (700 μg) was immunoprecipitated with STAT5 antibody and then blotted with SHP.

RT-PCR was used to analyze mRNAs for SHP, PDK2, and PDK4, as described previously (10,12). Individual PCRs were performed in triplicate on liver samples using mouse β-actin (Mm00607939_s1) as a housekeeping gene and experimental probes for PDK4 (Mm01166879_m1), PDK2 (Mm00446681_m1), and suppressor of cytokine signaling 2 (SOCS2; Mm00850544_g1) genes (Applied Biosystems, Foster City, CA).

The chromatin immunoprecipitation (ChIP) assay was performed as described previously (10). The specific primers used for the analyses were proximal forward 5′-CTGTTCGCTAAGAAGAAAGT-3′ and proximal reverse 5′- CAATGTCACGCATTCCTAG-3′; and distal forward 5′-AGTGCATGTATTGTCCAAG-3′ and distal reverse 5′-TGCACACAGACACTGCCAGA-3′.

Results are expressed as means ± SEM. Analysis of variance was used to determine significant differences as detected by Student t tests and/or one-way ANOVA methods using the Prism program. Statistical significance was considered at P < 0.05.

Previous studies have shown GH upregulates the expression of PDK4 via activation of STAT5 in 3T3 L1 adipocytes (29). On the basis of this finding, we hypothesized that activation of STAT5 by GH would upregulate PDK4 expression and thereby alter the levels of key metabolites in the liver. Indeed, administration of GH to intact mice significantly increased STAT5 phosphorylation and PDK4 expression in a time-dependent manner (Fig. 1A and B). In contrast, GH had no effect on PDK2 expression (Fig. 1B). Pyruvate concentration is markedly reduced in the liver of PDK4-null mice (27) because of higher PDC activity; therefore, we anticipated finding the pyruvate concentration increased in GH-treated mice due to reduced PDC activity. As expected, GH increased hepatic levels of pyruvate, citrate, and glutamate, as well as the NAD⁺-to-NADH and NADP⁺-to-NADPH ratios. In contrast, acetyl-CoA concentrations were decreased, consistent with less PDC activity (Fig. 1C). Overall, these results suggest that the induction of PDK4 by GH-mediated STAT5 activation plays an important role in regulating metabolite concentrations in the liver, including conservation of three carbon compounds that can be diverted into glucose rather than oxidized by way of PDC and the citric acid cycle.

To determine the effect of metformin on the increase of GH-mediated STAT5 phosphorylation, we determined whether metformin treatment or overexpression of SHP with Ad-SHP transduction prevents activation of STAT5 or its transactivation of the PDK4 promoter in primary hepatocytes. As shown in Fig. 2A and B, GH-mediated STAT5 phosphorylation was not decreased by metformin or Ad-SHP, whereas GH-induced PDK4 expression was significantly reduced by metformin and Ad-SHP. Whether STAT5 transactivation is blocked by SHP was tested by knocking-down SHP with Ad-siSHP and oligonucleotide siSHP in primary hepatocytes. As expected, STAT5 activation by GH was not repressed under control and SHP-knockdown conditions, whereas SHP knockdown reversed metformin-mediated repression of GH-induced PDK4 expression compared with control siRNA (Ad-Scram; Fig. 2C). To test whether metformin or SHP regulate STAT5 and PDK4 promoter activities, transient transfection assays were performed with the STAT5 and PDK4 gene promoter. In HepG2 cells, GH-induced STAT5 and PDK4 promoter activities were significantly decreased by metformin or SHP in a dose-dependent manner (Fig. 2D). Metformin inhibited the increase of STAT5 and PDK4 promoter activities in the presence of GH, which is consistent with previous results, and this inhibition was significantly abolished by SHP knockdown (Fig. 2E and F). Overall, these results demonstrate that the metformin-SHP pathway represses hepatic PDK4 gene expression by inhibiting transactivation of the PDK4 promoter by STAT5.

To verify that AMPK activation is involved in the regulation of PDK4 expression by metformin, we examined the effects of the metformin-AMPK pathway on the regulation of PDK4 gene expression in primary hepatocytes. GH-stimulated PDK4 expression was markedly reduced by overexpression of the CA form of AMPK (Ad-CA-AMPK) in a dose-dependent manner (Fig. 3A and C) but not by overexpression of the DN form of AMPK (Ad-DN-AMPK; Fig. 3B and D). As expected, the level of PDK2 mRNA was not changed by Ad-CA-AMPK and Ad-DN-AMPK in the presence of GH or metformin (Fig. 3A and B). We confirmed regulation of PDK4 promoter activity by AMPK with transient transfection assays. In HepG2 cells, the increase of PDK4 promoter activity by GH and/or TEL-JAK2-STAT5 was effectively reduced by metformin and CA-AMPK in a dose-dependent manner but not that of DN-AMPK (Fig. 3E and F).

To further confirm whether GH-mediated induction of PDK4 expression is altered by the metformin-AMPK-SHP pathway, we determined the effect of SHP on GH-induced PDK4 gene expression in primary hepatocytes. As expected, GH induced PDK4, and this effect was blocked by metformin (Supplementary Fig. 1). Also as anticipated, metformin induced SHP, suggesting the inhibition of the effect of GH may be mediated by SHP. That this is the case was proven by the finding that knockdown of SHP abolished the inhibitory effect of metformin on the induction of PDK4 expression by GH. Taken together with the data presented in Fig. 2, these results strongly suggest that GH-mediated induction of hepatic PDK4 expression is repressed by the metformin-AMPK-SHP pathway.

To investigate whether SHP and STAT5 physically interact, endogenous coimmunoprecipitation assays were performed with mouse liver samples prepared with control, GH alone, and GH plus metformin. Endogenous SHP strongly interacted with STAT5 in the presence of GH and metformin relative to control and GH alone (Fig. 4A). To determine the in vivo interaction between SHP and STAT5, we cotransfected mammalian expression vectors encoding GST alone or GST-SHP together with STAT5 into HepG2 cells. As shown in Fig. 4B, STAT5 strongly interacted with GST-SHP but not with GST. Collectively, these results show that SHP directly interacts with STAT5 in vivo. A proximal STAT5 consensus motif (TTCNNNGAA, -389/-378 bp) in the PDK4 promoter has been reported previously to be essential for GH- and prolactin-induced PDK4 expression in adipocytes (29). PDK4 promoter activity was continuously retained with deletions up to -539 bp. Promoter activity was largely lost with the -301 bp construct in response to GH treatment as well as TEL-JAK2/STAT5 cotransfection (Fig. 4C). These observations demonstrate that the GH-responsive region lies between -539 to -301 bp on the PDK4 gene promoter, consistent with the previously established STAT5 consensus sequence. To further confirm the inhibition of DNA binding of endogenous STAT5 proteins on the PDK4 gene promoter by SHP, we performed ChIP assays using anti-STAT5 (nonphosphorylated STAT5) or anti-phosphorylated STAT5 (p-STAT5) antibodies in primary hepatocytes. Endogenous STAT5 directly binds to the proximal (-570/-351 bp) site with GH, and this activity was clearly removed by metformin. However, the binding activity of STAT5 on the PDK4 gene promoter was restored by SHP knockdown compared with controls (Fig. 4D, left). No binding was observed in the nonspecific distal region (-1970/-1751 bp) on the PDK4 gene promoter (Fig. 4D, right). Endogenous SHP and control IgG was not bound to this promoter in either site (Fig. 4D). Furthermore, evidence of p-STAT5 occupancy on the PDK4 gene promoter was demonstrated by ChIP assay (Supplementary Fig. 2). Overall, these results demonstrate that induction of SHP by metformin decreased GH-stimulated PDK4 transcriptional activity via the blockage of STAT5 occupancy on the PDK4 gene promoter.

To determine whether the expression of PDK4 by GH is mediated by STAT5, we evaluated the role of STAT5 on the regulation of PDK4 gene expression using Ad-CA-STAT5 or Ad-DN-STAT5. As shown in Fig. 5A, STAT5 overexpression with Ad-CA-STAT5 increased PDK4 gene expression in primary hepatocytes. Furthermore, the stimulatory effect of GH on PDK4 expression was decreased markedly by Ad-DN-STAT5. We next confirmed CA-STAT5 and GH regulate the PDK4 transcriptional activity with a PDK4 gene promoter-luciferase reporter construct in hepatocytes. The activity of PDK4 gene promoter was significantly increased by GH or CA-STAT5 but not in DN-STAT5 and AG490, a JAK2 inhibitor (Fig. 5B). Overall, these results indicate that GH-mediated induction of hepatic PDK4 expression and activation of the PDK4 gene promoter were mediated by STAT5 in primary hepatocytes. Next, we evaluated the physiologic effect of regulation of PDK4 expression by GH in wild-type and STAT5-null mice. GH increased PDK4 expression in the liver of wild-type but not in STAT5-null mice (Fig. 5C). Under the same conditions, the mRNA of the positive control SOCS2 increased in parallel with PDK4, while PDK2 mRNA levels were not increased (Fig. 5C). To elucidate whether PDK4 induction by GH is required for phosphorylation of PDC under these conditions, we determined whether upregulation of PDK4 expression correlated with phosphorylation of PDC in the livers of GH-treated wild-type and PDK4-null mice. As shown Fig. 5D, GH significantly increased PDC phosphorylation in the liver of wild-type but not in PDK4-null mice. Taken together, these findings suggest that the GH-STAT5-PDK4 network plays a key role in the regulation of the phosphorylation state and therefore the activity of PDC.

As reported previously, the pattern of GH secretion differs markedly between males and females, the basal GH level being higher in females than males (35,36). Initially, we examined basal expression of hepatic PDK2 and PDK4 in wild-type and SHP-null mice. Whereas PDK2 mRNA was similar in both genotypes, PDK4 mRNA level was significantly higher in the liver of male and female SHP-null versus the wild-type mice (data not shown). Also, endogenous PDK4 expression was higher in the liver of female relative to male mice (data not shown). Moreover, we evaluated the endogenous expression of GH receptor in wild-type, PDK4-null, and SHP-null mice. Our results show that endogenous growth hormone receptor expression was not impaired in the liver of wild-type, PDK4-null, and SHP-null mice (Supplementary Fig. 3). Metformin increased SHP expression via the activation of AMPK in both male and female wild-type mice (Fig. 6A). Induction of SHP by metformin significantly decreased GH-induced PDK4 expression in wild-type male and female mice but not in SHP-null mice (Fig. 6A). GH-induced STAT5 activation was not decreased by metformin in both mice, consistent with data presented in Fig. 2. To elucidate the correlation between PDC activation by GH-mediated PDK4 expression and metformin-induced SHP expression, GH and metformin were administered to wild-type and SHP-null mice. As expected, the increase of SHP by metformin-mediated activation of ATM and AMPK significantly decreased PDC phosphorylation via the reduction of PDK4 expression in the liver of wild-type but not in SHP-null mice (Fig. 6B). Collectively, these results demonstrate that GH induces hepatic PDK4 expression in a sex-dependent manner and metformin inhibits GH-stimulated hepatic PDK4 expression and PDC activation in wild-type but not in SHP-null mice, thereby indicating a requirement for SHP in metformin-mediated inhibition of PDK4 expression.

Finally, we measured the effect of metformin on liver metabolites in GH-treated wild-type and SHP-null mice. Metformin prevented the increase in pyruvate and lactate induced by GH in the livers of wild-type but not SHP-null mice (Fig. 6C). Metformin also prevented the decrease in acetyl-CoA concentration caused by GH in wild-type mice. This effect of metformin was not rescued in SHP-null mice because of low acetyl-CoA levels in all conditions (Fig. 6C). Overall, these results suggest that the induction of PDK4 by GH-mediated STAT5 activation plays an important role in the regulation of liver metabolite levels.

In the current study, we have demonstrated that changes induced in the phosphorylation state of PDC by the GH-STAT5-PDK4 pathway are prevented by the metformin-AMPK-SHP pathway under both in vitro and in vivo conditions. However, the inhibitory effects of metformin on this GH-dependent pathway were reversed by Ad-DN-AMPK and by Ad-si SHP in vitro conditions. The fundamental molecular mechanism of our current study demonstrated that metformin-induced SHP expression represses STAT5 occupancy on the PDK4 gene promoter through a protein–protein interaction. Moreover, the increase of PDK4 expression and PDC phosphorylation by GH was abolished in both STAT5-null and PDK4-null mice. Finally, these effects of metformin were lost in SHP-null mice.

As previously mentioned, the expression of the PDK4 gene is mediated by GH- and prolactin-stimulated STAT5 activation in adipocytes (29). However, the correlation between GH-induced PDK4 expression and the opposing effects of metformin in the liver had not been studied previously. We observed that GH significantly induced hepatic PDK4 expression through the JAK2-STAT5 signal pathway. Interestingly, we found that metformin-mediated inhibition of PDK4 expression was not related to STAT5 phosphorylation, whereas the inhibitory effects of metformin were altered by STAT5 transcriptional activity. Our results in the in vitro and in vivo experiments strongly suggest that GH acts as a positive regulator of PDK4 expression through the JAK2-STAT5 pathway and that metformin downregulated the stimulatory effects of GH by inhibition of STAT5 transactivation of the PDK4 promoter. Recent work has demonstrated that PDK4 expression is synergistically induced by AMPK and fatty acids in hepatoma cells and cardiomyocytes, whereas AMPK activation by hypoxia or AICAR (AMPK activator) had no significant effect on PDK4 expression by itself (37). However, our present findings demonstrate that metformin significantly repressed GH-induced PDK4 expression through AMPK activation in wild-type but not in SHP-null mice (Fig. 6). Furthermore, Ad-CA-AMPK inhibited GH-stimulated PDK4 expression but not that of Ad-DN-AMPK (Fig. 3), indicating that the metformin-AMPK-SHP cascade acts as a negative regulator of GH-induced PDK4 expression in both in vitro and in vivo conditions. Overall, these findings suggest that the metformin-AMPK-SHP pathway may provide with a key checkpoint for regulating PDK4 expression on hepatic metabolic disorders, such as diabetes, obesity, and insulin resistance, through the GH-dependent pathway.

Previous studies from our group have demonstrated that loss of PDK4 improves glucose homeostasis by lowering blood glucose and improving insulin sensitivity and glucose tolerance in diet-induced obese mice (27,30). As mentioned above, the concentration of three carbon substrates (pyruvate, lactate, and alanine) for gluconeogenesis is decreased in the liver of PDK4-null mice (27). Our results demonstrated that GH-mediated induction of PDK4 expression induced PDC phosphorylation in the liver of wild-type but not in PDK4- and STAT5-null mice. GH induced PDK4 expression and PDC activation elevated liver metabolites (pyruvate and lactate) associated with gluconeogenesis and reduced acetyl-CoA concentrations associated with pyruvate metabolism (Fig. 5 and Fig. 6). Furthermore, the change of metabolites caused by activation of the GH-PDK4–dependent pathway was significantly reduced by metformin in wild-type but not in SHP-null mice (Fig. 6). These results suggest that GH-mediated PDK4 expression and the resulting decrease in PDC activity may play an important role in glucose homeostasis during fasting.

Because upregulation of PDK4 and inactivation of the PDC promote gluconeogenesis in the diabetic state, our present findings further suggest that AMPK activators that mediate an increase of SHP expression may have a beneficial effect on hepatic metabolites and blood glucose in diabetes via inhibition of the GH-PDK4–signaling pathway. On the basis of the findings of this study, we suggest that metformin suppresses hepatic gluconeogenesis via inhibition of the expression of gluconeogenic genes and the reduction of gluconeogenic substrates. At this point, however, we are unable to evaluate the relative importance of these two pathways, which will surely vary in the short- and long-term.

We previously demonstrated that various pharmacologic agents, including metformin, HGF, sodium arsenite, and fenofibrate, induce SHP gene expression and repress hepatic gluconeogenesis and fibrosis through the AMPK-SHP cascade (11–13,16). We confirmed that metformin-induced SHP expression regulates GH-mediated target genes in both in vitro and in vivo conditions. Indeed, our result demonstrated that the increase of SHP by metformin markedly repressed the GH-activated PDK4 expression in both male and female wild-type mice, whereas this inhibitory effect of metformin was blunted effectively with the Ad-DN-AMPK and SHP-null mice (Fig. 3 and Fig. 6). Therefore, it is suggested that AMPK activators that increase SHP, including fenofibrate, HGF, and other natural products, may induce significant effects on hepatic PDK4 expression by the GH-STAT5 signaling pathway, whereas induction of SHP by different pathways, including bile acids, may provide a weak correlation between the AMPK-dependent pathway and GH-induced hepatic PDK4 expression.

As previously mentioned, GH is well-known to be associated with its target genes, such as IGF-1, SOCS2, and HNF-6, in hepatocytes via the activation and/or DNA-binding activity of STAT5 to these gene promoters (38-40). Thus, we determined whether the fundamental molecular mechanism of GH-induced PDK4 expression is mediated by the metformin-AMPK-SHP cascade in the liver. Our results demonstrate that endogenous STAT5 binds directly to a proximal site on the PDK4 gene promoter, that metformin prevented the association of STAT5 with the PDK4 gene promoter, and that this effect of metformin was reversed by SHP knockdown (Fig. 4D). We further observed endogenous interaction between SHP and STAT5 by coimmunoprecipitation and in vivo interaction assays (Fig. 4A and B). Therefore, our results strongly suggest that inhibition of the binding of STAT5 to DNA is the molecular mechanism responsible for SHP-mediated inhibition of GH-stimulated PDK4 expression.

In conclusion, we have shown that GH induces PDK4 gene expression via the JAK2-STAT5–dependent pathway and that the metformin-AMPK-SHP cascade represses GH-induced hepatic PDK4 expression and PDC phosphorylation, as described in Fig. 6D. Our findings provide evidence for a novel pathway for the regulation of hepatic PDK4 expression and metabolites by GH and a potential therapeutic approach for the treatment of hepatic metabolic disorders.

This work was supported by a National Creative Research Initiatives Center for Nuclear Receptor Signals Grant funded from the Korean Ministry of Education, Science and Technology (20110018305), Future-based Technology Development Program (BIO Fields) through the National Research Foundation of Korea (NRF) funded by the Ministry of Education, Science and Technology (20100019512 to H.-S.C.), and by the KRIBB Research Initiative Program of Korea (to C.-H.L.), and the World Class University program through the NRF funded by the Ministry of Education, Science and Technology (R32-10064), a Grant of the Korea Health Technology R&D Project, Ministry of Health & Welfare Grant A111345 (to I.-K.L.).

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

Y.D.K. performed the experiments, researched data, and wrote and reviewed the manuscript. Y.-H.K. performed the animal experiments and reviewed the manuscript. S.T., J.H.Y., and Y.-H.Y. researched data and contributed to discussion. N.H.J., M.S., and L.H. contributed to discussion and critical review of the manuscript. R.A.H. and I.-K.L. contributed to the analysis and interpretation of data and to the critical review and revision of the manuscript. C.-H.L. and H.-S.C. contributed to experimental design and critical review and revision of the manuscript. H.-S.C. is the guarantor of this work, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

The authors thank Dr. David D. Moore (Baylor College of Medicine, Houston, TX) for permitting them to use SHP-null mice and Dr. Seok-Yong Choi (Chonnam National University Medical School, Gwangju, Republic of Korea) for his critical suggestions and helpful discussions. Specific antibodies for PDK4 and PDC were kindly provided by Drs. In-Kyu Lee (Kyungpook National University School of Medicine) and Robert A. Harris (Indiana University School of Medicine).