The transcription factor STAT3 participates in various critical cellular processes7
. In the liver, STAT3 is known to suppress expression of the transcriptional co-activator of gluconeogenesis PGC-1α and to suppress gluconeogenic genes. Ectopic expression of STAT3 in leptin receptor mutant (lepr-/-
) mice reduces PGC-1α
transcription and reverses diabetes. This effect of STAT3 is abolished when Tyr 705 is mutated to a phenylalanine (Y705F; ref. 5
). Other protein modifications of STAT3, such as acetylation, have recently been reported8-10
. However, the functional significance of STAT3 acetylation remains ill-defined, and its relationship with STAT3 tyrosine phosphorylation remains unclear.
We hypothesized that STAT3 acetylation regulates physiological processes by mediating changes in the STAT3 phosphorylation status. We found that STAT3 acetylation was decreased after a 24-h fast and increased after feeding in the livers of C57BL/6J mice. STAT3 tyrosine phosphorylation directly correlated with the level of STAT3 acetylation, indicating that the reversible acetyl-modifications are functionally relevant (). The observation that both acetylation and phosphorylation of STAT3 were evident in fed, but dramatically reduced in fasted, animals suggests that STAT3 acetylation and phosphorylation are actively downregulated on fasting. Overall, these observations signify that cellular metabolic status regulates STAT3 function in the liver, presumably owing to the sensitivity of the liver to the overall nutritional status of the organism11
Figure 1 SirT1 is involved in regulating STAT3 acetylation. (a) STAT3 acetylation and phosphorylation changed in mouse livers under different nutritional conditions. Male C57Bl/6J mice were fed (normal laboratory chow), fasted (24 h), re-fed (24 h, normal laboratory (more ...)
SirT1 can be induced to promote gluconeogenesis11,12
under conditions of fasting. Therefore, we asked whether SirT1 affects fed/fast-regulated STAT3 acetylation and phosphorylation. We injected the SirT1 inhibitor EX527, which has shown increased potency and specificity for SirT1 (ref. 13
), into C57Bl/6J mice. EX527 increased acetylation and phosphorylation of hepatic STAT3 (). These results are similar to those seen in animals treated with nicotinamide (Supplementary Information, Fig S1a
), a less specific SirT1 inhibitor. EX527 also increased the acetylation of p53, a known SirT1 substrate, suggesting that a reduction of SirT1 function was achieved with EX527 treatment (Supplementary Information, Fig. S4b
. In addition to EX527, we used an antisense oligionucleotide (ASO)14,15
to knockdown hepatic SirT1 on a chronic basis. SirT1 ASO induced significant STAT3 acetylation and tyrosine phosphorylation (). Together, these results support the idea that SirT1 is critically involved in hepatic STAT3 regulation.
Next, we studied the effect of nicotinamide and resveratrol (a SirT1 activator) on STAT3 acetylation and phosphorylation in an SV40-transformed mouse hepatic cell line, previously used in gluconeogenic studies16,17
. In cells treated with nicotinamide (0.3 mM and 3 mM), the levels of STAT3 acetylation and phosphorylation increased in a dose-dependent manner and were independent of the kinase JAK2 () and class I and II HDACs (other cells were treated with trichostatin A,
TSA; ). Conversely, resveratrol18
(0.2 μM) reduced STAT3 acetylation and phosphorylation in the cultured hepatic cells (). To determine whether deacetylation of STAT3 requires SirT1, we used SirT1 knockout (KO) and wild-type mouse embryonic fibroblasts19
(MEFs). First, we found that levels of STAT3 acetylation and phosophrylation were constitutively higher in the SirT1 KO MEFs than in the wild-type MEFs (). Second, treatment with EX527 increased levels of STAT3 acetylation and phosphorylation in wild-type MEFs, but not in SirT1 KO MEFs (). Moreover, resveratrol decreased STAT3 acetylation and phosphorylation in wild-type MEFs, but not SirT1 KO MEFs (). Together, these results further indicate that deacetylation of STAT3 is dependent on SirT1.
To investigate STAT3 as a SirT1 substrate, we studied the role of ectopically expressed SirT1. We found that the level of STAT3 acetylation was greatly reduced by cotransfection of human (h) SirT1, in HEK293T cells (). Moreover, the introduced hSirT1 was able to potently suppress p300/CBP-induced STAT3 acetylation, suggesting that both factors affected the same set of lysine residues of STAT3 (). The deacetylation of STAT3 by hSirT1 was more effective than that by HDAC1, HDAC 3 () or HDAC2 (data not shown), which were previously implicated in STAT3 deacetylation10,20
Figure 2 STAT3 phosphorylation and transactivation were downregulated by SirT1. (a) SirT1 deacetylates STAT3 in cultured cells. The effect of p300, SirT1 or HDACs on STAT3 acetylation was measured in transfected HEK293T cells (H1, HDCA1; H3, HDCA3). (b-e) SirT1 (more ...)
To test whether SirT1 and STAT3 formed a complex in cells, HEK293T cells were co-transfected with the wild-type genes of hSirT1 and hSTAT3. Exogenous () and endogenous () hSirT1 proteins were detected in immunoprecipitation products, suggesting that STAT3 and hSirT1 formed complexes. Next, endogenous STAT3 was co-precipitated by endogenous hSirT1 (). To determine whether this physical interaction between hSirT1 and STAT3 occurs in vivo, we examined their association in mouse liver. SirT1 was observed in the STAT3 immunoprecipitation products, suggesting an interaction between endogenous STAT3 and SirT1 in the liver of fasted animals (), and less so in the liver of fed animals ().
To determine the region(s) in STAT3 responsible for SirT1 binding, a set of five c-Myc tagged STAT3 deletion mutants (T1 to T5) were generated that were designed to test each of the domains in STAT3. The DNA binding and linker domain (from amino acid 330-590) of STAT3 (ref. 21
) was found to be the key region involved in the STAT3-SirT1 interaction (). The various truncated STAT3s that did not contain the DNA binding and linker domain failed to form complexes with SirT1, whereas the truncated STAT3s that contained this domain pulled down SirT1 (). Moreover, the latter had greatly reduced acetylation and phosphorylation on cotransfection with SirT1 (data not shown), suggesting that a direct interaction between the STAT3 DNA binding and linker domain and SirT1 is required for the enzymatic function of SiT1 on STAT3.
We next analysed whether the acetylation state of STAT3 affects STAT3 phosphorylation and transactivation function. After hSirT1 overexpression, the phosphorylation of exogenous () and endogenous () STAT3 was again downregulated with high efficiency. Low levels of exogenous hSirT1 DNA (0.05 μg) resulted in a reduction of exogenous and endogenous STAT3 acetylation and phosphorylation (), whereas transfection with HDAC1 and HDAC 3 had a limited effect (). Consistently, an assay of STAT3-dependent luciferase reporter10
(p4x IRF-Luc) revealed that SirT1, but not HDAC1 or HDAC3, suppressed the transactivation function of STAT3 (). However, at increased doses of plasmid DNA, HDAC3 appreciably reduced STAT3 activity (Supplementary Information, Fig. S2a
); this may contribute to downregulation of STAT3 acetylation and phosphorylation under certain conditions and in selected cells8,20
. From these results, we conclude that SirT1 specifically and effectively deacetylated STAT3 in cultured cells and in vivo,
and that this modification is coupled with a downregulation of STAT3 phosphorylation and transactivation.
The direct interaction between SirT1 and STAT3, and its effect on STAT3 phosphorylation and function, suggest that the state of acetylation of STAT3 may directly regulate its phosphorylation. Acetylation in a limited number of lysine residues in STAT3 was reported8-10
; however, the coupling of acetylation and phosphorylation through these sites was not established, suggesting more lysine acetylation sites are involved. To identify these, particularly in the carboxy-terminal region, which is crucial for STAT3 phosphorylation, we initially examined acetylation in the K685R-STAT3 mutant: Lys 685 was reported to be the only acetylation site at the C terminus of STAT3 (ref. 10
). We used an anti-acetyl-STAT3 antibody and, although it detected a minor reduction in acetylation, significant signal was still detected (). This signal was downregulated by SirT1, suggesting that other acetylation sites exist. Three new lysine-acetylation residues, K679, 707 and 709, were identified by tandem mass spectrometry analysis (Supplementary Information, Fig. S2
). These sites are evolutionally conserved among mammalian STATs, () and located either in the end β-sheet structure of the SH2 domain21
(βH), or in the disordered signal stretch immediately after the SH2 domain. SH2 is known to specifically bind to phospho-tyrosine peptides22,23
; therefore, it is critical to tyrosine signalling. Notably, the three new lysine residues are all in the vicinity of the Y705 of STAT3 (), indicating that these sites may be pertinent to STAT3 phosphorylation7,24
Figure 3 Critical novel acetylation sites regulate STAT3 phosphorylation and transactivation. (a) HEK293 cells were transfected with wild-type (WT) or K685R-STAT3 plasmids alone, or co-transfected with SirT1. STAT3 acetylation was determined. (b) A schematic representation (more ...)
Next, a systematic site-mutagenesis of these lysine sites of interest was performed. Mutation of all four lysine to arginine abolished the acetylation signals of STAT3 (). Single or double lysine-to-arginine mutations had a limited effect on STAT3 phosphorylation, whereas changes of all four lysines to arganine (4K/R) largely abolished STAT3 phosphorylation. This effect is specific to the C-terminal acetylation sites, as mutations of the amino-terminal acetylation sites (K49/87R) did not affect the phosphorylation of STAT3 (Supplementary Information, Fig. S3a, b
). These results imply that multiple lysine-acetylation sites adjacent to Y705 are vital for STAT3 phosphorylation. In contrast, the dominant-negative Y705F mutation did not affect the acetylation of STAT3 (Supplementary Information, Fig. S3c
). More importantly, the 4K/R-STAT3 mutant was no longer sensitive to stimulation by IL-6 (interleukin) and IFN-γ (interferon-γ) which otherwise acetylated and phosphorylated STAT3. This suggests that the newly identified lysine acetylation sites are critical for cytokine stimulation of STAT3 ().
To test whether the mutations of these acetylation sites affect STAT3 transcriptional function, STAT3-dependent luciferase reporter assays were performed in HEK293T cells and human ovarian cancer A2780 cells25
. Consistent with the phosphorylation results, the 4K/R mutation largely abolished IL-6-induced STAT3 activation and had a dominant-negative-like effect, comparable to that of the Y705F mutation (); whereas individual mutations had limited effects on STAT3 transactivation. Moreover, 4K/R-STAT3 was not capable of activating a known endogenous STAT3 target (hAGT; ref. 20
) in HepG2 cells (Supplementary Information, Fig. S3d
). Finally, ectopic expression of p300 (ref. 10
) and treatment with EX527, which increased STAT3 acetylation and phosphorylation in vivo13
, failed to activate 4K/R-STAT3, whereas the same treatments increased the transactivation of wild-type STAT3 (). These results suggest that the acetylation of the cluster of C-terminal lysine sites is up or downregulated by p300 or SirT1, respectively, and is crucial for STAT3 phosphorylation and transactivation. Next, we asked if the 4K/R mutations affect STAT3 localization, as phosphorylation is crucial for STAT3 nuclear translocation7
. We introduced wild-type-, 4K/R- or Y705F-STAT3 into liver-STAT3 knockout (STAT3 LKO) hepatoma cells using a retrovirus (pbabe-6xcMyc-STAT3 containing wild-type-, Y705F- and 4K/R- STAT3) to test STAT3 nuclear translocation by immunofluorescence microscopy staining. Similarly to Y705F-STAT3, the mutant 4K/R-STAT3 significantly disrupted the nuclear translocation of STAT3 (). This acetylation-related STAT3 localization was further supported by experiments in primary hepatocytes treated with EX527 (). In addition, we found that the efficiency of 4K/R-STAT3 dimerization was greatly reduced (Supplementary Information, Fig. S3e
If SirT1 promotes gluconeogenesis, in part through the suppression of STAT3, the knockout of STAT3 in liver should mimic the effect of SirT1 and increase gluconeogenesis independent of SirT1 activity. In normal chow fed STAT3 LKO mice, we observed a significant upregulation of pepck1
, but not of the control genes (cytochrome-c
). However, the further upregulation of pepck1
genes on fasting26
was limited (). The lack of repression of gluconeogenic gene expression after feeding, was due to the absence of STAT3 bound to the promoter region of pepck1,
as shown by chromatin immunopreciptiation (ChIP) assay (Supplementary Information, Fig. S3f
Figure 4 Liver-STAT3 deficiency disrupted fasting/SirT1 controlled gluconeogenesis. (a) Liver-STAT3 deficiency mimicked the effect of SirT1 in promoting gluconeogenic gene expression, independently of nutritional status and SirT1 activity. The mRNA levels of (more ...)
Next, we asked whether STAT3 is involved in the SirT1-mediated induction of hepatic gluconeogenesis. EX527 was used to inhibit SirT1 activity in STAT3 LKO mice and littermate controls after a 24-h fast. As reported previously6
, STAT3 LKO mice maintained normal glucose levels under such conditions, despite increasing gluconeogenic gene expression and plasma insulin levels (). EX527 (i.p. 10 mg per kg body weight) reduced plasma glucose levels in fasting wild-type mice, whereas the plasma glucose levels of STAT3 LKO mice were unchanged (), suggesting that STAT3 deficiency disrupted the ability of EX527 to lower fasting glucose levels independent of insulin concentration (). Consistent with the reduction of plasma glucose levels in wild-type animals after EX527 treatment, the expression of the gluconeogenic genes was reduced in the livers of wild-type mice but was blunted in STAT3 LKO mice (). Next, we studied the effect of SirT1 knockdown on glucose homeostasis in the STAT3 LKO model. The levels of hepatic STAT3 acetylation and phosphorylation were significantly increased with SirT1 ASO in wild-type mice (; Supplementary Information, Fig. S4a
). Moreover, SirT1 ASO decreased hepatic gluconeogenic gene transcription (pepck1, g6pase
) in wild-type, but not STAT3 LKO, mice (data not shown).
Finally, a glucose tolerance test (), a pyruvate tolerance test () and a glucagon-stimulation test () were conducted to evaluate various aspects of glucose homeostasis in wild-type and STAT3 LKO mice, with or without SirT1 knockdown. An overall reduction in glucose production was indicated in the SirT1 ASO-treated control animals. SirT1 ASO treatment of STAT3 LKO mice had little effect on the parameters measured by all three tests (). These data led us to conclude that SirT1 promotes gluconeogenesis, in part, by suppressing the inhibitory effect of STAT3 on the expression of gluconeogenic genes. To test the effect of hepatic insulin resistance on the interaction of SirT1 with STAT3, all groups of mice were fed with a high-fat diet (Supplementary Information, Fig. S5a
). Blood glucose levels on fasting were decreased in SirT1 ASO-treated wild-type mice, but unchanged in SirT1-ASO treated STAT3 LKO mice (; Fig. S5b
). This observation indicates that the loss of hepatic STAT3 is critical for SirT1 function. In addition, our results are consistent with the SirT1 LKO model12
, but differ from the SirT1 transgenic mouse model.3
. However, gluconeogenic gene expression was increased in isolated primary SirT1 transgenic hepatocytes treated with cAMP and deprived of insulin treatment, suggesting that insulin-mediated inhibition of gluconeogenesis may be independent of the SirT1 pathway3
If the C-terminal cluster of acetylation lysine sites are crucial for STAT3 transactivation, the C-terminal 4K/R mutant should disrupt the suppressive effect of STAT3 on hepatic glucose production. First, we found that the glucose production in STAT3 LKO primary hepatocytes was increased, compared with that in wild-type cells (). To further test the inhibitory effect of STAT3 on glucose production in hepatocytes, we promoted cellular glucose production in these cells, either by introducing PGC-1α or by treating cells with dexamethasone (50 nM) and 8-bromo-cAMP (2 μM; refs 2, 26
). Equivalent amounts of STAT3 total proteins were detected in STAT3-/-
cells in which wild-type STAT3, 4K/R- or Y705F-STAT3 had been re-introduced (). Wild-type STAT3 suppressed hepatic glucose production (), whereas, 4K/R- or Y705F-STAT3 had little effect. Consistent with the data on glucose production, we observed a strong reduction in expression of the gluconeogenic pepck1
transcripts by wild-type-STAT3, but not by 4K/R-STAT3, in STAT3 LKO hepatocytes ectopically expressing PGC-1α or treated with dexamethasone/cAMP ().
Figure 5 The 4K/R mutant STAT3 is defective in suppressing hepatic gluconeogenesis. (a) Basal-, PGC-1α- or dexa/cAMP-stimulated glucose production in primary hepatocytes was significantly increased in the absence of STAT3 (* P < 0.05; FF, STAT3 (more ...)
To determine whether the change in hepatic glucose production by re-introduced wild-type-STAT3 (but not 4K/R-STAT3) is mediated by SirT1, we co-infected STAT3 LKO primary hepatocytes with both adeno-SirT1 and adeno-STAT3s (wild type, 4K/R, Y705F and GFP). Ectopic SirT1 blunted the suppression of glucose production by wild-type-STAT3; however, this effect was limited in 4K/R-STAT3- or Y705F-STAT3-expressing cells (). These results further support the idea that hepatic glucose production is dependent on SirT1-mediated downregulation of STAT3.
Our findings reveal a new molecular mechanism whereby SirT1 suppresses the inhibitory effect of STAT3 on gluconeogenesis, while activating PGC-1α and Foxo1 (ref. 4
) to stimulate gluconeogenesis in the liver, in response to nutrient signals. This dual function of SirT1 has a central role in preventing concurrent activation of the two counter-mechanisms of gluconeogenesis regulation (). These findings have implications for defining the basic pathways of energy homeostasis, diabetes and lifespan.