The phosphorylation status of T522 modulates tissue‐specific functions of SIRT1 in energy metabolism in mice

Abstract

SIRT1, the most conserved mammalian NAD ⁺‐dependent protein deacetylase, is an important metabolic regulator. However, the mechanisms by which SIRT1 is regulated in vivo remain unclear. Here, we report that phosphorylation modification of T522 on SIRT1 is crucial for tissue‐specific regulation of SIRT1 activity in mice. Dephosphorylation of T522 is critical for repression of its activity during adipogenesis. The phospho‐T522 level is reduced during adipogenesis. Knocking‐in a constitutive T522 phosphorylation mimic activates the β‐catenin/GATA3 pathway, repressing PPARγ signaling, impairing differentiation of white adipocytes, and ameliorating high‐fat diet‐induced dyslipidemia in mice. In contrast, phosphorylation of T522 is crucial for activation of hepatic SIRT1 in response to over‐nutrition. Hepatic SIRT1 is hyperphosphorylated at T522 upon high‐fat diet feeding. Knocking‐in a SIRT1 mutant defective in T522 phosphorylation disrupts hepatic fatty acid oxidation, resulting in hepatic steatosis after high‐fat diet feeding. In addition, the T522 dephosphorylation mimic impairs systemic energy metabolism. Our findings unveil an important link between environmental cues, SIRT1 phosphorylation, and energy homeostasis and demonstrate that the phosphorylation of T522 is a critical element in tissue‐specific regulation of SIRT1 activity in vivo.

Introduction

Metabolic syndrome is defined as a cluster of metabolism‐related disorders, such as central obesity, type 2 diabetes, dyslipidemia, and high blood pressure, all of which are considered as major contributors of mortality in industrialized countries 1, 2, 3, 4. Both genetic factors and environmental influences contribute to the pathogenesis of metabolic syndrome. Among which, a class III histone deacetylase and a mammalian homologue of yeast silent information regulator (Sir2) protein, SIRT1, play a central role in the regulation of transcriptional networks in various critical metabolic processes in multiple tissues. For example, SIRT1 is a key modulator of both glucose and fatty acid metabolism in the liver 5, 6. Knocking‐down or deletion of hepatic SIRT1 impairs fatty acid oxidation, thereby increasing the susceptibility of mice to dyslipidemia and hepatic steatosis 7, 8, 9. Conversely, hepatic overexpression of SIRT1 attenuates hepatic steatosis and ER stress and restores glucose homeostasis in mice 10. SIRT1 is also an important regulator of maturation and remodeling of adipose tissues 6. It has been reported that SIRT1 represses a master regulator of adipogenesis in the white adipose tissue (WAT), PPARγ, thereby suppressing the expression of adipose tissue markers, such as a fatty acid binding protein, aP2, and inhibiting fat mobilization in response to fasting 11. Moreover, genetic ablation of SIRT1 in adipose tissues leads to increased adiposity and insulin resistance 12, whereas treatment of mice on a high‐fat diet with resveratrol, a polyphenol that activates SIRT1 in cells directly or indirectly 13, 14, 15, 16, 17, protects animals against high‐fat induced obesity and metabolic dysfunctions 18, 19, 20. Therefore, current studies point to the notion that SIRT1 functions as an adaptor that is “beneficial” to cellular and organismal metabolism. Consequently, dysfunction of this sirtuin contributes to the development a number of human metabolic diseases, particularly metabolic syndrome 5.

Although the role of SIRT1 in metabolic regulation of a variety of biological processes has been well studied, how the activity of SIRT1 is regulated in vivo in response to different biological/environmental cues remains elusive, and the functional/physiological consequences of disruption of its regulation are still unclear. As a NAD⁺‐dependent protein deacetylase, the activity of SIRT1 is tightly controlled at multiple levels, either by the cellular levels of NAD⁺, which are hypersensitive to a number of environmental cues including fasting, caloric restriction, exercise, or high‐fat diet feeding, and/or at the posttranslational level by small chemicals, protein–protein interactions, or through posttranslational modification 21. In particular, we have previously reported that SIRT1 can be phosphorylated and activated by two anti‐apoptotic members of the dual‐specificity tyrosine phosphorylation‐regulated kinase (DYRK), DYRK1A and DYRK3, in response to acute environmental stresses 22. We further showed that this modification activates its deacetylase activity independently of the cellular NAD⁺ level through preventing the formation of less‐active SIRT1 oligomers/aggregates 23. Our findings suggest that phosphorylation modification of SIRT1 might provide a molecular mechanism that fine‐tunes SIRT1 activity in vivo independently of the cellular NAD⁺ level.

To assess the physiological impacts of the phosphorylation of T522 on SIRT1, we generated SIRT1 T522 phosphorylation mimic (threonine to glutamic acid, or T522E mutation, or TE) and dephosphorylation mimic (threonine to alanine, or T522A mutation, or TA) knock‐in mouse models. In this report, we show for the first time that the phosphorylation of T522 renders tissue‐specific regulation of SIRT1 activity in response to developmental and nutritional signals in vivo.

Results

Generation of SIRT1 TEKI and TAKI mice

To investigate whether phosphorylation of T522 of SIRT1 is an important posttranslational modification that modulates SIRT1 activity in vivo, we used standard gene‐targeting technology to generate two different SIRT1 mutation knock‐in mouse lines, SIRT1 TAKI (TAKI) and SIRT1 TEKI (TEKI), with T522A (TA) to mimic the dephosphorylated SIRT1 and T522E (TE) to mimic the phosphorylated SIRT1, respectively (Figs 1 and EV1). Both KI strains were born at the expected Mendelian ratio with no gross phenotypes, indicating that the phosphorylation status of T522 does not affect normal animal development and survival. Immunoblotting analysis of total protein lysates from different tissues with anti‐SIRT1 antibody revealed that the protein levels of two mutant proteins in these two lines were comparable to those of wild‐type (WT) SIRT1 (Fig 1A). Further analyses with the anti‐p‐SIRT1(T522) antibody confirmed that knocking‐in the TA mutant abolished the phosphorylation signals on the endogenous SIRT1 protein (Fig 1B, Lane TAKI). Since our p‐SIRT1(T522) antibody only binds weakly to the phospho‐mimics 22, knocking‐in the TE mutation also yielded a mutant SIRT1 protein undetectable by this antibody (Fig 1B, Lane TEKI).

Figure 1. Generation of SIRT1 TEKI and TAKI mice.

1. The SIRT1 expression levels in wild type (WT), TEKI, and TAKI mice are comparable. Total protein lysates from indicated tissues were analyzed by immunoblotting with an anti‐SIRT1 antibody.
2. Endogenous SIRT1 proteins from indicated tissues from both TEKI and TAKI mice display decreased p‐SIRT1(T522) levels. Total protein lysates from indicated tissues were analyzed by immunoblotting with antibodies against SIRT1 and p‐SIRT1(T522). Please note that the p‐SIRT1(T522) antibody only displays weak affinity to the SIRT1TE protein.
3. SIRT1 TEKI MEFs have an increased deacetylase activity toward p53 in response to genotoxic stress. MEFs isolated from WT, TEKI, and TAKI mice were treated with adriamycin (0.2 μg/ml) for 8 h or with PBS (as control). Total cell lysates were immunoblotted with acetyl‐p53 (K382) or p53 antibodies.
4. SIRT1 TEKI MEFs have elevated deacetylase activity to the p65 subunit of NF‐κB in response to inflammatory signals. WT, TEKI, and TAKI MEFs were treated with 10 ng/ml of TNFα, or 1 μg/ml of Escherichia coli O111:B4 lipopolysaccharide (LPS) for 30 min with PBS (as control). Total cell lysates were immunoblotted with acetyl‐p65 and p65 antibodies.

Source data are available online for this figure.

Figure EV1. Generation of SIRT1 TEKI and TAKI mice.

1. Schematic representation of knock‐in strategies for generating SIRT1 TEKI and SIRT1 TAKI strains. Locations and sequences of genotyping primers are labeled.
2. Representation images of genotyping PCR results.
3. Primers and qPCR results for specific detection of WT, TEKI, and TAKI mRNA (n = 4 mice for each group).

Data Information: In (C), data are presented as mean ± SEM.

Our previous studies have demonstrated that in cultured cells, the phospho‐mimics of SIRT1 T522 are hyperactive toward acetyl‐p53 protein upon genotoxic stress, whereas the dephospho‐mimics of SIRT1 T522 are partially inactive 22, 23. To confirm the changes of SIRT1 deacetylase activity in the knock‐in mouse lines, we isolated mouse embryonic fibroblasts (MEFs) from WT, TAKI, and TEKI embryos and analyzed cellular SIRT1 activities. As shown in Fig 1C, acetyl‐p53 levels in WT MEFs were increased when treated with a DNA intercalating drug, adriamycin (Adria). This increase was blunted in TEKI MEFs (top panels), indicating that the TE mutant protein has an increased deacetylase activity toward acetyl‐p53. The knocked‐in SIRT1 TA protein, on the other hand, did not appear to have defects in deacetylation of p53 protein in response to adriamycin‐induced genotoxic stress (bottom panels), possibly due to stable knock‐in induced compensatory effects in regulating p53 acetylation in these cells. However, both TE and TA mutants displayed expected enhanced (TEKI, top panels) and reduced (TAKI, bottom panels) deacetylase activities toward acetyl‐p65 when MEFs were challenged with pro‐inflammatory stimuli, TNFα or LPS (Fig 1D). Taken together, our observations indicate that when stably knocked‐in mice, the SIRT1 TEKI allele exhibits the expected enhanced activity upon all tested environmental stress, while the TAKI allele displays the expected activity in response to specific environmental stimuli.

Phosphorylation of SIRT1 at T522 inhibits adipogenesis in vitro

SIRT1 is a negative regulator of adipogenesis 6, suggesting that the activity of this sirtuin must be repressed during the process of normal adipogenesis. To test whether phosphorylation of T522 plays a role in regulation of SIRT1 activity in this process, we analyzed the p‐SIRT1(T522) levels of endogenous SIRT1 protein during a 7‐day in vitro adipogenesis of WT primary MEFs. As shown in Fig 2A, p‐SIRT1 levels were gradually reduced while total SIRT1 protein remained constant during in vitro adipogenesis, indicating that SIRT1 phosphorylation (activity) but not SIRT1 expression is negatively correlated with adipogenesis. Moreover, this reduction in p‐SIRT1 levels was accompanied with decreased levels of DYRK3 but not DYRK1A (Fig 2A and B), suggesting that reduced expression of DYRK3 might be a reason for the diminished phosphorylation of SIRT1 in this process.

Figure 2. The SIRT1 TEKI allele inhibits adipogenesis in vitro .

1. The endogenous SIRT1 protein is dephosphorylated at T522 during in vitro differentiation of MEFs into adipocytes. Primary MEFs isolated from WT mice were treated and induced to differentiation into adipocytes in vitro as described in Materials and Methods. The levels of indicated proteins were analyzed by immunoblotting.
2. The expression levels of Dyrk3 but not Dyrk1a are decreased during in vitro adipogenesis (n = 3 independent experiments). The mRNA levels of indicated genes were analyzed by qPCR.
3. SIRT1 TEKI MEFs have reduced in vitro adipogenesis. In vitro differentiated adipocytes from primary MEFs isolated from WT, TAKI, and TEKI mice were stained by Oil Red O. Scale bars, 50 μm.
4. In vitro differentiated adipocytes from primary TEKI MEFs have reduced expression levels of adipocyte markers (n = 3 independent experiments).

Data Information: In (B and D), data are presented as mean ± SEM. *P < 0.05 (Mann–Whitney test). Source data are available online for this figure.

In line with above observations, primary MEFs isolated from dephosphorylation defective TEKI mice accumulated much less fat as determined by Oil Red O staining after 7 days of differentiation compared with WT MEFs, whereas MEFs from constitutively dephosphorylated TAKI mice displayed a comparable ability to be differentiated into adipocytes as WT MEFs (Fig 2C). Consistently, the mRNA levels of PPARγ2, a nutrition‐sensitive isoform of PPARγ, aP2, as well as Glut4, an adipose tissue‐specific glucose transporter, were significantly reduced in TEKI but not in TAKI cells during differentiation (Fig 2D). Taken together, our finding indicates that dephosphorylation of SIRT1, thereby reduction in SIRT1 activity, is required for normal adipogenesis in vitro.

The SIRT1 TEKI allele actives β‐catenin/GATA3 signaling, repressing PPARγ and impairing functions of WAT in vitro and in vivo

Although SIRT1 has been shown to indirectly repress or directly activate PPARγ in response to different environmental signals in WAT 11, 24, the reduced induction of PPARγ2 during adipogenesis of TEKI cells (Fig 2D) raises the possibility that the SIRT1 TE mutant protein may blunt the transcription of PPARγ2, thereby repressing adipocyte differentiation. To test this possibility, we analyzed the expression of several genes that have been previously shown to be important in transcriptional regulation of PPARγ, such as β‐catenin, GATA2, and GATA3, in WT and TEKI MEFs before and after differentiation into adipocytes. As shown in Fig 3A–C, both mRNA and protein levels of GATA3 but not its close homolog GATA2 nor β‐catenin were significantly elevated in TEKI MEFs before and at the early stages of differentiation.

Figure 3. The SIRT1 TEKI allele actives β‐catenin/GATA3 signaling.

1. TEKI MEFs have increased expression levels of GATA3 during in vitro adipogenesis (n = 3 independent experiments). The mRNA levels of indicated genes were analyzed by qPCR.
2. Primary TEKI MEFs have increased mRNA levels of GATA3 but not β‐catenin (n = 4 independent experiments). The mRNA levels of indicated genes were analyzed by qPCR.
3. Primary TEKI MEFs have increased protein levels of GATA3 but reduced β‐catenin protein. The levels of indicated protein were analyzed by immunoblotting.
4. The binding of β‐catenin to GATA3 promoter is enhanced in primary TEKI MEFs (n = 3 independent experiments). The relative enrichment of β‐catenin on the β‐catenin binding site of the GATA3 promoter was determined by the ChIP assay as described in Materials and Methods.
5. The binding of GATA3 to its target promoters is increased in primary TEKI MEFs (n = 3 independent experiments). The relative enrichment of GATA3 on the GATA3 binding sites of indicated promoters was determined by the ChIP assay as described in Materials and Methods.

Data Information: In (A, B, D, and E), data are presented as mean ± SEM. *P < 0.05 (Mann–Whitney test).Source data are available online for this figure.

GATA3 has been shown to be highly expressed in preadipocytes, inhibiting adipogenesis in part through repressing the transcription of PPARγ2 25, 26. In preadipocytes, the transcription of GATA3 is directly activated by itself, as well as by β‐catenin, a known SIRT1 deacetylation substrate 27, 28. The increased expression of GATA3 in TEKI cells suggests that the hyperactive SIRT1 TE protein may induce GATA3 expression through deacetylation/activation of β‐catenin. In support of this notion, the binding of β‐catenin to the GATA3 promoter was enhanced in these cells despite its reduced protein levels (Fig 3C and D). The association of GATA3 to its own promoter, as well as to PPARγ2 promoter, was also dramatically elevated in TEKI primary MEFs (Fig 3E). These observations suggest that the TE mutant enhances β‐catenin/GATA3 pathway thereby inhibiting the induction of PPARγ2 and adipogenesis.

In line with the finding in in vitro adipogenesis, β‐catenin was significantly hypoacetylated in WAT of TEKI mice compared to WT mice (Fig 4A), suggesting an activation of its transaction activity (Simic et al 27). Consistently, the expression of GATA3 was induced (Fig 4B), whereas the mRNA levels of PPARγ and several PPARγ target genes 29, including Pepck, Adiponectin, aP2, were significantly reduced in the WAT of TEKI mice (Fig 4C). The reduced PPARγ pathway in TEKI WAT was associated with reduced chromatin association of PPARγ to PPAR response element (PPRE) on one of its target promoters, aP2 (Fig 4D). As a control, the binding of PPARγ to a control region on this promoter, C/EBPα binding site (C/EBPα site), was not significantly reduced.

Figure 4. SIRT1 TEKI mice have enhanced β‐catenin/GATA3 signaling but impaired PPARγ pathway and WAT functions.

1. β‐Catenin is hypoacetylated in WAT of TEKI mice (n = 4 mice for each group). The acetylation levels of β‐catenin were analyzed by immunoprecipitation (IP) of β‐catenin followed by immunoblotting (IB) with an anti‐acetyl‐K antibody.
2. WAT of TEKI mice has increased mRNA levels of GATA3 but not β‐catenin (n = 6 mice for each group). The mRNA levels of indicated genes were analyzed by qPCR.
3. TEKI WAT has a reduced PPARγ‐signaling pathway (n = 6 mice for each group). The mRNA levels of indicated genes were analyzed by qPCR.
4. Reduced binding of PPARγ on the PPRE of the aP2 promoter in WAT of TEKI mice (n = 3 mice for each group). The relative enrichment of PPARγ on the aP2 promoter was determined by the ChIP assay as described in Materials and Methods.
5. SIRT1 TEKI mice have reduced expression levels of genes involved in lipogenesis in WAT (n = 6 mice for each group). The mRNA levels of indicated genes were analyzed by qPCR.
6. SIRT1 TEKI female mice have reduced body fat under the chow diet feeding (n = 9 WT and 6 TEKI mice). The percentage of fat mass and lean mass in 9‐ to 10‐month‐old mice were determined by Bruker LF90 minispec.
7. SIRT1 TEKI mice have reduced expression of genes involved in lipolysis in WAT (n = 6 mice for each group), and SIRT1 TEKI adipocytes have reduced lipolysis in vitro (n = 5 mice for each group). The mRNA levels of indicated genes were analyzed by qPCR, and the in vitro lipolysis assay was performed with isolated primary adipocytes as described in Materials and Methods.
8. SIRT1 TEKI mice have increased expression of genes in energy expenditure in WAT (n = 6 mice for each group). The mRNA levels of indicated genes were analyzed by qPCR.

Data Information: In all panels, data are presented as mean ± SEM. 0.05 < ^# P < 0.1, *P < 0.05 (Mann–Whitney test). Source data are available online for this figure.

Further analysis revealed that the expression levels of several lipogenic genes were reduced in the WAT of TEKI mice (Fig 4E), and SIRT1 TEKI female mice had significantly reduced fat composition but increased lean mass compared to WT mice (Fig 4F), indicating that TEKI mice have reduced adipogenesis in vivo. Intriguingly, the mRNA levels of a number of genes involved in lipolysis were also significantly reduced in WAT of TEKI mice after the overnight fasting, and isolated primary adipocytes from TEKI mice had reduced release of glycerol in response to the isoproterenol treatment in vitro compared to WT adipocytes (Fig 4G). Moreover, the expression of a couple of genes mediating energy catabolism was also significantly elevated in WAT of TEKI mice (Fig 4E), suggesting that knocking‐in the TEKI allele in WAT alters multiple functions of this tissue in addition to adipogenesis/lipogenesis. Together, our findings demonstrate that dephosphorylation of SIRT1 T522 is a critical step for proper differentiation of white adipocytes and that constitutive phosphorylation of SIRT1 at T522 represses white adipocyte differentiation and functions in vitro and in vivo.

The SIRT1 TEKI allele enhances systemic lipid oxidation and partially protects mice from high‐fat diet‐induced dyslipidemia

TEKI mice also had elevated serum β‐hydroxybutyrate levels after overnight fasting along with enhanced expression of a number of fatty acid oxidation genes in the liver compared to WT control mice (Fig 5A and B), indicating an elevation in hepatic fatty acid oxidation. Moreover, several fatty acid oxidation genes were also significantly increased in the BAT without significant alterations in levels of fatty acid synthesis genes (Fig 5C), further supporting the notion that the TEKI allele enhances systemic fatty acid catabolism. As an additional control, TAKI mice exhibited normal fatty acid oxidation in their liver and BAT (Fig EV2).

Figure 5. SIRT1 TEKI mice display enhanced systemic lipid oxidation and are partially protected from high‐fat diet‐induced dyslipidemia.

• A, B
  SIRT1 TEKI mice have increased β‐hydroxybutyrate levels in serum (A) and elevated expression levels of fatty acid oxidation genes in the liver (B) in response to fasting (n = 6 WT and 7 TEKI mice). The serum levels of β‐hydroxybutyrate was determined as described in Materials and Methods after 16‐h fasting, and the mRNA levels of indicated genes were analyzed in fasted liver samples by qPCR.
• C
  SIRT1 TEKI mice have elevated expression of fatty acid oxidation genes in basal condition in the BAT (n = 8 WT and 4 TEKI mice). The mRNA levels of indicated genes were analyzed in BAT samples under fed conditions by qPCR.
• D, E
  SIRT1 TEKI mice have reduced body fat (D), and decreased serum levels of cholesterol, free fatty acids, and glycerol (E) after 5 months of high‐fat diet feeding (n = 10 WT and 8 TEKI mice). The percentage of fat mass in indicated mice were determined by Bruker LF90 minispec, and the serum levels of indicated metabolites were measured as described in Materials and Methods.

Data Information: In (A–D), data are presented as mean ± SEM. *P < 0.05, **P < 0.01 (Mann–Whitney test).

Figure EV2. SIRT1 TAKI mice have normal fasting‐induced hepatic fatty acid oxidation and basal fatty acid oxidation in BAT .

• A, B
  SIRT1 TAKI mice have normal levels of serum β‐hydroxybutyrate levels (A) and expression of fatty acid oxidation genes in the liver (B) upon fasting (n = 6 mice for each group).
• C
  SIRT1 TAKI mice have normal expression of genes involved in fatty acid metabolism in basal condition in the BAT (n = 4 WT and 8 TAKI mice).

Data Information: In all panels, data are presented as mean ± SEM. 0.05 < ^# P < 0.1 (Mann–Whitney test).

To further assess the importance of SIRT1 phosphorylation in control of systemic lipid homeostasis, we challenged WT, TEKI, and TAKI mice with a Western style high‐fat diet containing 40% kcal fat and 0.21% cholesterol for 5 months. As shown in Fig 5D, after high‐fat diet feeding, TEKI mice displayed a mild but significant reduction in total fat percentage, despite that they gained similar weights (Fig EV3A) and had comparable food intakes (Fig EV3B) compared to WT mice. Moreover, TEKI mice showed decreased serum levels of free fatty acids, glycerol, and cholesterol after high‐fat diet feeding (Fig 5E), without detectable alterations in other serum lipids and hormones (Fig EV3E and F). Again, TAKI mice exhibited no significant changes in diet‐induced obesity and dyslipidemia (Fig EV3). Collectively, these data indicate that constitutive phosphorylation of SIRT1 on T522 enhances systemic lipid oxidation and partially protects animals from high‐fat diet‐induced dyslipidemia.

Figure EV3. SIRT1 TEKI and TAKI mice under high‐fat diet feeding.

• A, B
  Body weight (A) and food intake (B) of SIRT1 TEKI and TAKI mice under high‐fat diet feeding (n = 10 WT, 8 TAKI, and 8 TEKI mice).
• C
  SIRT1 TAKI mice have normal total fat percentage after 5 months of high‐fat diet feeding (n = 10 WT and 8 TAKI mice).
• D
  SIRT1 TAKI mice have normal serum levels of triglyceride, free fatty acids, glycerol, and cholesterol after 5 months of high‐fat diet feeding (n = 10 WT and 8 TAKI mice).
• E
  Serum levels of HDL and LDL after 5 months of high‐fat diet feeding (n = 10 WT and 8 TAKI mice).
• F
  Serum insulin, leptin, and adiponectin levels were measured after 5 months of high‐fat diet feeding (n = 10 WT and 8 TAKI mice).

Data Information: In all panels, data are presented as mean ± SEM.

SIRT1 TAKI mice display defective hepatic fatty acid oxidation and develop hepatic steatosis upon high‐fat diet feeding

In addition to adipose tissues, SIRT1 also has important roles in regulation of hepatic energy metabolism, particularly stimulation of hepatic fatty acid oxidation 6, 7, 8. But again, how the activity of hepatic SIRT1 is regulated in response to nutritional cues is still unclear. To explore the possible role of T522 phosphorylation in regulation of SIRT1 activity in liver, we analyzed the hepatic pSIRT1 (T522) levels in chow diet and high‐fat diet‐fed mice either under fed condition or after overnight fasting (Fig 6A and B). Surprisingly, compared to chow diet‐fed livers, p‐T522 levels were significantly induced in high‐fat diet‐fed livers but not upon fasting. Since cellular NAD⁺ levels have been reported to be reduced after high‐fat diet feeding 30, 31, this observation suggests that SIRT1 activity is increased upon over‐nutrition independently of cellular NAD⁺ levels and further suggests that TAKI mice, in which SIRT1 is defective in T522 phosphorylation, would be more sensitive to lipid loading than WT and TEKI mice in the liver. In agreement with this hypothesis, TAKI mice exhibited significant reduction in several fatty acid oxidation genes in the liver under the fed condition compared to WT and TEKI mice (Fig 6C). Moreover, TAKI primary hepatocytes showed a blunted ability to induce the expression of a number of fatty acid oxidation genes in response to lipid loading followed by the treatment with a PPARα agonist, WY14643 (WY; Fig 6D), indicating that the TA mutant decreases fatty acid oxidation cell autonomously in hepatocytes. On the other hand, although TEKI mice had enhanced hepatic fatty acid oxidation in response to fasting (Fig 5A and B), the isolated primary hepatocytes from these mice had a normal ability to induce the expression of fatty acid oxidation genes after lipid loading and WY treatment (Fig 6E), suggesting that the elevation in hepatic fatty acid oxidation in TEKI mice is non‐cell autonomous.

Figure 6. SIRT1 TAKI mice have impaired hepatic fatty acid oxidation under fed or lipid‐loading conditions.

• A
  T522 phosphorylation levels on endogenous SIRT1 protein are increased in high‐fat diet‐fed livers. WT mice were fed with a chow diet (Chow) or a high‐fat diet (HFD) for 5 months. Liver p‐SIRT1 (T522) and total SIRT1 levels were analyzed by immunoblotting under indicated conditions.
• B
  p‐SIRT1 (T522) levels in livers of mice under different conditions (n = 3 mice for each group). The relative levels of p‐SIRT1/total SIRT1 were determined by densitometry with ImageJ.
• C
  SIRT1 TAKI mice have reduced hepatic expression of genes involved in fatty acid oxidation under fed condition (n = 6 mice for each group). The mRNA levels of indicated genes were analyzed in liver samples under fed conditions by qPCR.
• D, E
  SIRT1 TAKI but not TEKI primary hepatocytes are defective in fatty acid oxidation upon lipid loading and treatment with a PPARα agonist, WY14643 (WY; n = 4 mice for each group). Primary hepatocytes isolated from WT, TEKI, or TAKI mice were pre‐incubated with 125 μM oleic acid/BSA for overnight in the presence or absence of 30 μM WY14643 in low glucose medium, followed by incubation with 125 μM oleic acid/BSA and 1 mM carnitine in glucose‐free DMEM for 4 h.

Data Information: In (B–E), data are presented as mean ± SEM. 0.1 < ^# P < 0.05, *P < 0.05, **P < 0.01 (Mann–Whitney test). Source data are available online for this figure.

Consistent with above observations, TAKI mice accumulated significantly higher amount of fat in their liver after 5 months of high‐fat diet feeding, as revealed by H&E staining of liver sections (Fig 7A) and enzymatic colorimetric quantification of extracted liver triglycerides, NEFA, and cholesterol (Fig 7B). Further analyses showed that after 5 months of high‐fat diet feeding, TAKI livers also had enhanced deposition of collagen (Fig 7C, arrows), hepatocyte nuclear invagination and enlarged nuclei (Fig 7D, arrows), and increased serum levels of AST (Fig 7E), indicating development of liver fibrosis and liver damage.

Figure 7. SIRT1 TAKI mice develop fatty liver after high‐fat diet feeding.

1. SIRT1 TAKI mice develop fatty liver after 5 months of high‐fat diet feeding. Liver sections were stained with hematoxylin and eosin. Scale bars, 100 μm.
2. TAKI mice have increased triglyceride and non‐esterified fatty acids (NEFAs) in the liver after high‐fat diet feeding (n = 7 mice for each group). The hepatic lipids metabolites were extracted and measured as described in Materials and Methods.
3. SIRT1 TAKI mice have increased liver fibrosis after high‐fat diet feeding. Liver sections were stained by Masson's trichrome staining (blue). Scale bars, 50 μm.
4. Hepatocyte nuclear invagination and enlarged nuclei (arrows) in TAKI liver after high‐fat diet feeding. Scale bar, 50 μm.
5. TAKI mice have enhanced liver damage after high‐fat diet feeding. Serum AST and ALT levels were measured after 5 months of high‐fat diet feeding (n = 7 mice for each group).
6. TAKI mice have a distinct hepatic gene expression profile compared to WT and TEKI mice. (Top) The numbers of differentially expressed gene probes between WT, TAKI, and TEKI livers. (Bottom) Venn‐diagram representation of significantly altered gene probes between TAKI vs. WT and TAKI vs. TEKI. The hepatic mRNA was analyzed by mouse whole‐genome microarray as described in Materials and Methods.
7. SIRT1 TAKI livers have differential expression patterns of a subset of genes in the liver toxicity and inflammation pathways compared to WT and TEKI livers (n = 5 WT mice, 5 TAKI mice, and 3 TEKI mice; cutoff P < 0.05). The hepatic mRNA was analyzed by mouse whole‐genome microarray as described in Materials and Methods.
8. SIRT1 TAKI livers have increased expression of several genes involved in liver toxicity and inflammation (n = 5 WT mice and 5 TAKI mice). The hepatic mRNA was analyzed by quantitative real‐time PCR.

Data Information: In (B, E and H), data are presented as mean ± SEM. 0.1 < ^# P < 0.05, *P < 0.05, **P < 0.01 (Mann–Whitney test).

Further microarray analysis of hepatic gene expression of high‐fat diet‐fed WT, TAKI, and TEKI mice revealed that consistent with above observations, TEKI and WT livers shared almost identical gene expression profiles under the high‐fat diet (Fig 7F). On the contrary, TAKI livers had alterations on the expression of 1,086 and 1,406 gene probes compared to WT and TEKI livers, respectively (Fig 7F, top, Tables EV1 and EV2), and 866 of these genes were common (Fig 7F, bottom), suggesting that TAKI livers were significantly different from WT and TEKI livers. In support of this notion, heat map analyses of total significantly changed 1,626 gene probes indicated that WT and TEKI liver were indistinguishable, whereas TAKI livers were clustered into a separate group (Fig EV4). Further Ingenuity Pathway Analysis (IPA) of these 1,626 genes revealed that TAKI livers had significantly impairments on pathways involved in hepatic fibrosis/hepatic stellate cell activation (Table EV3) as well as the liver toxicity and inflammation (Fig 7G and H) compared to WT and TEKI livers, confirming that SIRT1 phosphorylation on T522 is crucial to maintain hepatic functions in response to nutrient overloading. Collectively, our findings indicate that phosphorylation of the T522 residue on SIRT1 is vital for hepatic fatty acid oxidation upon high‐fat diet feeding.

Figure EV4. SIRT1 TAKI livers display distinct transcriptional response to high‐fat diet feeding, whereas SIRT1 TEKI livers have minimal effects compared to WT liver.

WT, TAKI, and TEKI mice were fed with a high‐fat diet for 5 months, the hepatic mRNA was analyzed by mouse whole‐genome microarray as described in Materials and Methods.

SIRT1 TAKI mice exhibit systemic alternations in energy metabolism

Our observation that phosphorylation modification of SIRT1 T522 was widespread in all tested tissues (Fig 1B) further suggests that knocking‐in a constitutive dephosphorylated SIRT1 allele may elicit systemic effects on SIRT1‐mediated metabolic processes in addition to lipid‐induced hepatic fatty acid oxidation. Indeed, SIRT1 TAKI mice were insulin resistant under normal feeding conditions (Fig 8A), indicating that they are less able to uptake glucose. On the contrary, SIRT1 TEKI mice have normal insulin sensitivity (Fig 8B). Paradoxically, TAKI mice also exhibited an enhanced respiration exchange ratio (RER) compared to WT and TEKI mice, especially at the night phase (Fig 8C and D), indicating a preference to use glucose over fatty acids of these mice. One plausible explanation for this paradox is that the blunted ability to utilize fat (Figs 6 and 7) increases the glucose dependence of TAKI mice in spite of the fact that they are less able to uptake glucose in response to insulin. Taken together, our studies demonstrate that the phosphorylation modification of the SIRT1 T522 residue is critically involved in normal function of SIRT1 in systemic energy metabolism in response to nutritional signals in vitro and in vivo.

Figure 8. SIRT1 TAKI mice display systemic alterations in energy metabolism.

• A
  SIRT1 TAKI mice are more insulin resistant than WT mice under normal feeding conditions. Six hours after fasting, WT, SIRT1 TAKI, and SIRT1 TEKI mice were i.p. injected with 0.75 μ/kg insulin, and blood glucose levels were monitored (n = 7 WT and 8 TAKI littermate mice).
• B
  SIRT1 TEKI mice display a normal sensitivity to insulin compared to WT controls under normal feeding conditions (n = 7 WT and 8 TEKI littermate mice).
• C, D
  SIRT1 TAKI mice have increased respiratory exchange ratio (RER) during the night phase, whereas SIRT1 TEKI mice display normal levels of RER. WT, TAKI, and TEKI mice were singly housed in the Labmaster calorimetry units, and the rates of O₂ consumption and CO₂ production were monitored and the RER = VCO₂/VO₂ (n = 7 WT and eight TAKI littermate mice or n = 7 WT and eight TEKI littermate mice).

Data Information: In (A and B), data are presented as mean ± SEM. *P < 0.05 (Mann–Whitney test). In (C and D), data are presented as mean.

Discussion

As a critical cellular metabolic sensor, SIRT1 has been well established as a master regulator of metabolism 6, 32, 33. However, the regulation of SIRT1 activity in vivo in response to environmental signals is still unclear, and the physiological consequences of disruption of its regulation are largely unknown. In our present study, we showed that the phosphorylation modification of SIRT1 on T522 is an important regulatory mechanism for modulation of the activity of SIRT1 in energy metabolism in different tissues. On the one hand, an allele that mimics the phosphorylation of T522, the TEKI allele, primarily affects maturation and function of WAT. One the other hand, a SIRT1 mutant allele that is defective in phosphorylation modification, the TAKI allele, disrupts systemic energy homeostasis, such as hepatic fatty acid metabolism, insulin sensitivity, and circadian switch of nutrient resources. Our observations suggest that maintaining a proper phosphorylation level of T522 is critical for whole body energy homeostasis, while dephosphorylation of T522 is an important regulatory mechanism to repress SIRT1 activity during adipogenesis. Our study highlights the importance of SIRT1 T522 phosphorylation in coupling environmental cues to energy metabolism in vivo.

Several lines of evidence support the notion that constitutive phosphorylation of T522 (the TE mutation) primarily impacts the function of SIRT1 in adipose tissues. For example, SIRT1 TEKI primary MEFs display defective adipogenesis in vitro (Fig 2). SIRT1 TEKI mice have a blunted fasting response in WAT, enhanced energy expenditure in BAT, and are relatively protected from high‐fat diet‐induced dyslipidemia (Figs 4 and 5). More importantly, the T522 residue of endogenous SIRT1 protein is gradually dephosphorylated during the process of in vitro adipogenesis (Fig 2A), indicating that the dephosphorylation of T522 is a critical mechanism in repression of SIRT1 activity in this process. Despite the strong influence of the TEKI allele on adipose tissues, however, the impact of this allele on functions of other tissues appears to be minimal. For instance, the TEKI allele only indirectly enhances hepatic fatty acid oxidation in response to fasting (Figs 5A and B, and 6E), and SIRT1 TEKI mice and WT controls have comparable hepatic responses to high‐fat diet feeding (Fig 7). SIRT1 TEKI mice also have normal expression levels of genes involved in lipid metabolism in the muscle (Fig EV5) and display normal insulin sensitivity (Fig 8B) and a normal nutrient preference (Fig 8D) under normal feeding conditions when compared with WT mice. In addition to inhibition of the PPARγ signaling (Figs 2, 3, 4 and 11), SIRT1 has been shown to directly enhance this signaling through deacetylation of PPARγ itself in response to cold exposure in the WAT, thereby promoting WAT browning 24. Therefore, the adipose tissue specificity of TEKI mice raises the possibility that they may have enhanced cold‐induced browning of WAT. It will be interesting to test this possibility in future studies.

Figure EV5. SIRT1 TEKI and TAKI mice have normal expression of genes involved in fatty acid metabolism in the muscle.

• A, B
  WT, TEKI, and TAKI mice were fed with the chow diet or a high‐fat diet for 5 months. The mRNA abundance of indicated genes involved in fatty acid oxidation (A) and lipid synthesis (B) was analyzed by qPCR (n = 6 mice for each group).

Data Information: In all panels, data are presented as mean ± SEM.

Despite the strong defects of TAKI mice in hepatic lipid metabolism upon high‐fat diet feeding (Figs 6 and 7), the molecular targets of the SIRT1 TAKI mutant protein in liver are still not clear. We have examined the acetylation status of a number of known SIRT1 targets, including PGC‐1a, FOXO1, NF‐kB (p65), p53, and H4K16, in livers of high‐fat diet‐fed animals, but none of these proteins displayed altered acetylation levels in the liver of TAKI mice compared to WT mice (Fig EV6). New assays are needed to identify possible targets of SIRT1 in liver of TAKI mice in future studies.

Figure EV6. The acetylation levels of indicated SIRT1 substrates in livers of high‐fat diet‐fed WT, TEKI, and TAKI mice.

Male TEKI, TAKI, and their WT littermate were fed with a high‐fat diet for 5 months, the acetylation levels of indicated SIRT1 deacetylation substrates were analyzed by immunoprecipitation (IP) followed by immunoblotting (IB) of acetylated protein or total protein.

The tissue‐specific impact of T522 phosphorylation on SIRT1's function is not altogether surprising. Although the SIRT1 T522 residue is phosphorylated in all tested tissues (Fig 1B, not shown), the expression of DYRK kinases is tissue‐specific 34, 35. In particular, the expression levels of several DYRKs, including DYRK3 and DYRK1B, are highly sensitive to environmental stress and nutritional cues 22, 36, 37. Additionally, the T530 of human SIRT1 (equivalent to T522 of mouse SIRT1) has been reported also as a target of cyclin B/cdk1 and JNK1 kinase 38, 39. Therefore, the phosphorylation level of SIRT1 is differentially regulated by different kinases in different tissues in response to different stimuli. Our observations that the TAKI allele behaves like the WT allele in tissues/conditions where the endogenous SIRT1 is hypophosphorylated (e.g., adipogenesis), while the TEKI allele behaves like the WT allele in tissues/conditions where the endogenous SIRT1 is hyperphosphorylated (e.g., HFD‐fed liver), are consistent with this notion. Despite the tissue specificity of T522 phosphorylation, however, our observations that TAKI mice have defects in multiple tissues support the idea that endogenous SIRT1 protein is generally phosphorylated at T522 in different tissues, and maintaining this basal phosphorylation level of T522 is critical for whole body energy homeostasis. Our observation that T522 was hyperphosphorylated in high‐fat diet‐fed livers in which cellular NAD⁺ levels are reduced 30, 31 further demonstrates that the activity of SIRT1 can be induced independently of cellular NAD⁺ level in vivo.

The link between DYRK and SIRT1 suggests that DYRK kinases may be also vital in regulation of energy metabolism in addition to neuronal development 40, stress response 41, and cell proliferation and apoptosis 34, 42, 43. Indeed, the essential role of DYRKs in energy metabolism has begun to emerge in recent years. For instance, minibrain/DYRK1A has recently been shown to regulate food intake in flies and mammals through the Sir2‐FOXO‐sNPF/NPY pathway 44, confirming our link between DYRK1A and SIRT1 in this process 22. DYRK1A has also been reported to induce pancreatic β‐cell mass expansion and improve glucose tolerance in mice 45, 46. Moreover, DYRK1B, a close member of DYRK1A, has been recently associated with a form of the metabolic syndrome in human 47. Our observation that the reduced expression of DYRK3 is coupled with decreased SIRT1 phosphorylation during adipogenesis (Fig 2A and B) further suggests that this kinase may be also involved in energy metabolism in mammals. Further studies with genetic modified mouse models will help to test this possibility in vivo.

It is important to note that although the phosphorylation defective TAKI allele displays a number of expected alterations in hepatic and systemic metabolic homeostasis, some of our current observations are inconsistent with our previous results or expected phenotypes. For instance, we have previously shown that in cultured cells and in vitro, the phosphorylation defective TA mutant protein is aggregation prone and less active, failing to protect cells from stress‐induced cell death 22, 23. Given the strong impact of the TA mutation on SIRT1 activity in vitro, as well as the importance of SIRT1 activity in animal development 48, 49, 50, 51, 52, it is unexpected that the TAKI allele does not significantly affect the deacetylase activity of SIRT1 in some tested substrates (Fig 1D) and the TAKI mice have minimal alterations on development (not shown). Since dephosphorylation of SIRT1 T522 primarily affects the stability of SIRT1 protein instead of its intrinsic deacetylase activity 22, 23, it is possible that compensatory effects induced by stable knocking‐in the TA mutant, including hyperphosphorylation of SIRT1 on additional sites, offset some effects of the TA mutation on SIRT1 oligomerization/aggregation. Nevertheless, our data indicate that maintaining a suitable phosphorylation level of T522 on SIRT1 protein is critical for energy homeostasis in multiple tissues, highlighting the importance of this modification in vivo.

In summary, we have shown that phosphorylation modification of T522 is an additional layer that mediates environmental modulation of SIRT1 activity in vivo, and plays a vital role in tissue‐specific regulation of energy metabolism in mice. The DYRK/SIRT1 signaling axis will likely provide new insights into gene–environment interactions that affect systemic energy metabolism, tissue homeostasis, and animal stress responses.

Materials and Methods

Animal experiments

SIRT1 TAKI and TEKI mice were generated with standard mouse knock‐in technology. Threonine 522 of SIRT1 was targeted and replaced with alanine (TAKI) or glutamate (TEKI) in C57BL/6 embryonic stem cells at Taconic Biosciences, Inc. The resulting TAKI, TEKI, and their WT littermates on the C57BL/6 background were housed in individualized ventilated cages (Tecniplast, Exton, PA) with a combination of autoclaved nesting material (Nestlet, Ancare Corp., Bellmore, NY and Crink‐l'Nest, The Andersons, Inc., Maumee, OH) and housed on hardwood bedding (Sani‐chips, PJ Murphy, Montville, NJ, USA). Mice were maintained on a 12:12‐h light: dark cycle at 22 ± 0.5°C and relative humidity of 40–60%. Mice were provided ad libitum autoclaved rodent diet (NIH31, Harlan Laboratories, Madison, WI, USA) and deionized water treated by reverse osmosis. Mice were negative for mouse hepatitis virus, Sendai virus, pneumonia virus of mice, mouse parvovirus 1 and 2, epizootic diarrhea of infant mice, mouse norovirus, Mycoplasma pulmonis, Helicobacter spp., and endo‐ and ectoparasites upon receipt, and no pathogens were detected in sentinel mice during this study.

All the experiments were carried out on TAKI, TEKI mice and their gender‐ and age‐matched WT littermates. Animal numbers in each group were chosen to achieve 1.5‐fold difference with 80% of power. No randomization method was used, and the investigators were not blinded to the group allocation.

Fat and lean body mass were determined by Bruker LF90 minispec at the age of 9–10 months. High‐fat Western diet (D12079B, Research Diets) was used to feed mice aged 2–3 months for 20 weeks. Tissues were harvested after 4‐h food withdrawal, starting at the beginning of the daytime cycle (fed), or after 16‐h food withdrawal, starting at the end of daytime cycle (fasted). Tissue lipids were extracted as described 53. Respiratory exchange ratio was measured using the Labmaster system (TSE systems).

All animal procedures were reviewed and approved by National Institute of Environmental Health Sciences Animal Care and Use Committee. All animals were housed, cared for, and used in compliance with the Guide for the Care and Use of Laboratory Animals and housed and used in an Association for the Assessment and Accreditation of Laboratory Animal Care, International (AAALAC) Program.

Histological analysis

Liver samples were embedded with paraffin and stained with hematoxylin and eosin for morphological study. Differentiated adipocytes were collected at different time of differentiation and stained with Oil Red O as described 54.

Biochemical analysis

Commercially available reagents or kits were used to measured serum lipids as listed: triglycerides and free glycerol reagent (Sigma), non‐esterified fatty acid (NEFA) and cholesterol (Wako Pure Chemical Industries). Serum β‐hydroxybutyrate was determined with Stanbio reagents (Stanbio). Liver lipids were extracted as described 53, and liver triglycerides, NEFA, and cholesterol were quantified by above commercially available kits.

Cell culture

Based on a previously described method 55, MEF cells were prepared from E14.5 embryos isolated from pregnant heterozygous TEKI (or TAKI) females mated with heterozygous TEKI (or TAKI) males. Isolated TEKI, TAKI, and WT MEF cells were further verified by PCR. MEF cells were maintained in Dulbecco's modified Eagle's medium (high glucose), supplemented with 10% fetal bovine serum (HyClone). To make immortalized MEF cells, 2 μg of SV40 T antigen vector was transfected into 1 × 10⁶ cells with Lipofectamine 2000 (Invitrogen). Transfected MEF cells were then split 1/10 at least five times to get immortalized cells.

Protein phosphorylation, acetylation, and Western blot analysis

To detect Thr522 phosphorylation level of SIRT1 in the mice, tissues were collected and homogenized in Nonidet P‐40 buffer (50 mM Tris–HCl, pH 8.0, 150 mM NaCl, 1% Nonidet P‐40) containing complete protease inhibitors (Roche Applied Science). Western blot analysis was used to detect Thr522‐phosphorylated SIRT1 and total SIRT1.

When p53 was used as the acetylation substrate of SIRT1, TEKI, TAKI, and their WT littermate MEF cells were treated with either 0.2 μg/ml of adriamycin (MP Biomedicals) or PBS (solvent control) for 8 h, with 5 μM of trichostatin A (TSA, Sigma) and 25 μM of MG‐132 (Calbiochem) treatment during the last 2 h. When NF‐κB was used as the acetylation substrate of SIRT1, TEKI, TAKI, and their WT littermate MEF cells were treated with either 10 ng/ml of TNFα (R&D), 1 μg/ml of Escherichia coli O111:B4 lipopolysaccharide (LPS, Sigma) or PBS (solvent control) for 30 min, also with 200 nM of TSA. Cell lysates were prepared with SDS lysis buffer and resolved by SDS–PAGE. Western blot analysis was used to detect acetylated substrates and total substrates.

The acetylation levels of β‐catenin were analyzed by immunoprecipitation of β‐catenin followed by immunoblotting with an anti‐acetyl‐K antibody.

All the Western blot images were captured with Odyssey Infrared Imaging System (Li‐Cor Biosciences). Antibodies used were listed as following: acetylated p53 (Millipore 06‐916), total p53 (Santa Cruz Biotechnology sc‐6243), acetylated NF‐κB (GeneTex GTX86963), total NF‐κB (Santa Cruz Biotechnology sc‐372), total SIRT1 (Sigma S5196), β‐catenin (GeneTex GTX101435), GATA3 (Santa Cruz Biotechnology, sc‐9009), acetyl‐lysine (Cell Signaling Technology #9441). Rabbit anti‐p‐SIRT1(T522) antiserum was generated and validated as described 22. Please also see source data files for validation of these antibodies.

Chromatin immunoprecipitation (ChIP) analysis

Chromatin immunoprecipitation (ChIP) analysis was performed essentially as described by the manufacturer, EMD Millipore (Upstate Biotechnology). Briefly, WT and TEKI primary MEFs, or white adipose tissues from WT and TEKI mice were fixed with 1% paraformaldehyde in PBS at room temperature for 15 min, minced, and dounced in SDS lysis buffer (50 mM Tris–HCl, pH 8.1, 10 mM EDTA, 1% SDS, and 2× Complete™ protease inhibitor mixture). Chromatin was then sonicated to an average length of 200–500 bp and subjected to immunoprecipitation with antibodies against β‐catenin (GeneTex), GATA3(Santa Cruz Biotechnology), or PPARγ (Santa Cruz biotechnology). DNA fragments were subjected to qPCR using primers flanking indicated promoter regions.

In vitro adipogenesis

Confluent primary MEFs (at the Passage #1, P1) were used for adipocyte differentiation as described 56. Confluent MEF cells were cultured in basal medium (high glucose DMEM, 10% FBS) for 2 days. Cell differentiation was then induced by differentiation medium I (basal medium with 0.5 mM IBMX, 0.25 μM dexamethasone, and 1 μg/ml insulin). After 48 h, the medium was changed to differentiation medium II (basal medium with 1 μg/ml insulin). Medium was changed every 2 days.

In vitro lipolysis

Lipolysis assay was performed as previously described 57. In brief, epididymal fat pads of mice were dissected, cut into 2‐ to 3‐mm pieces, and incubated with 1.5 mg/ml collagenase I solution (Worthington Biochemical). After shaking at 220 rpm at 37°C for 1 h, the mixture was filtered through a 250‐μm mesh. The floating adipocytes were washed three times and resuspended in 1 ml of adipocyte incubation solution. To stimulate lipolysis, 200 μl of resuspended adipocytes was incubated with 10 μM (−)‐isoproterenol (Sigma) at 37°C for 1 h. After incubation, the level of glycerol released into the buffer was determined using free glycerol reagent (Sigma).

Real‐time quantitative PCR (qPCR)

Total RNA was isolated and purified from tissues or cells using TriZOL (Invitrogen) and Qiagen RNeasy minikit (Qiagen). cDNA was synthesized with the ABI reverse transcriptase kit, and real‐time quantitative PCR was analyzed using SYBR Green Supermix (Applied Biosystems). All data were normalized to lamin A expression.

Microarray study and data analysis

Gene expression was analyzed using Agilent Whole Mouse Genome 4 × 44 multiplex format oligo arrays (014868; Agilent Technologies, Santa Clara, CA, USA) following the Agilent 1‐color microarray‐based gene expression analysis protocol. Starting with 500 ng of total RNA, Cy3‐labeled cRNA was produced according to manufacturer's protocol. For each sample, 1.65 μg of Cy3‐labeled cRNAs was fragmented and hybridized for 17 h in a rotating hybridization oven. Slides were washed and then scanned with an Agilent Scanner. Data were obtained using the Agilent Feature Extraction software (v9.5), using the 1‐color defaults for all parameters. The Agilent Feature Extraction Software performed error modeling, adjusting for additive and multiplicative noise. Microarray data are available from the Gene Expression Omnibus under accession GSE94904.

Resulting data were processed using Agi4x44PreProcess, limma in Bioconductor for the R software environment (www.r-project.org). A single log scale normalized expression measure for each probe set was obtained after background correction and normalization between samples. Box plots, density plots, MA plot, and spatial images of the raw and normalized data were examined in order to ensure data quality. Principal component analysis (PCA) was performed on all samples and all probes to reduce the data dimensionality while preserving the variation in the data set. This allowed us to assess the similarities and differences of samples within a treatment group and between treatment groups. The significance of the log ratio for each probe was determined by calculating one modified t statistic per probe using an empirical Bayesian approach. Probes with Benjamini–Hochberg multiple test corrected P‐values (q‐values) < 0.05 were considered to be differentially expressed between the two groups. Heat map was generated using BioConductor package HeatPlus. Dendrograms of samples (columns) and genes (row) were generated by hierarchical clustering. Color scale is from fourfold lower (log2‐fold = −2) than mean (green) to eightfold higher (log2‐fold = 3) than mean (red). Probe sets were further analyzed using the Ingenuity Pathway Analysis tool (version 16542223; Ingenuity Systems, Redwood City, CA, USA).

Statistical analysis

Values are expressed as mean ± standard error of mean (SEM). Significant differences between means were analyzed by the two‐tailed, unpaired, nonparametric Mann–Whitney test, and differences were considered significant at P < 0.05.

Author contributions

JL, QX, and MJ designed experiments, carried out experiments, and analyzed data. XG generated SIRT1 TAKI and TEKI mouse models. XX and DCF analyzed the microarray data. XL conceived the study, designed experiments, analyzed data, and prepared the manuscript.

Conflict of interest

The authors declare that they have no conflict of interest.

Supporting information

Expanded View Figures PDF

Table EV1

Table EV2

Table EV3

Review Process File

Source Data for Figure 1

Source Data for Figure 2

Source Data for Figure 3

Source Data for Figure 4

Source Data for Figure 6

EMBO Reports (2017) 18: 841–857