TAZ inhibits GR and coordinates hepatic glucose homeostasis in normal physiologic states

The elucidation of the mechanisms whereby the liver maintains glucose homeostasis is crucial for the understanding of physiologic and pathologic states. Here, we show a novel role of hepatic transcriptional co-activator with PDZ-binding motif (TAZ) in the inhibition of glucocorticoid receptor (GR). TAZ interacts via its WW domain with the ligand-binding domain of GR to limit the binding of GR to gluconeogenic gene promoters. Therefore, liver-specific TAZ knockout mice show increases in glucose production and blood glucose concentration. Conversely, the overexpression of TAZ in mouse liver reduces the binding of GR to gluconeogenic gene promoters and glucose production. Thus, our findings demonstrate distinct roles of the hippo pathway effector proteins yes-associated protein 1 (YAP) and TAZ in liver physiology: while deletion of hepatic YAP has little effect on glucose homeostasis, hepatic TAZ protein expression decreases upon fasting and coordinates gluconeogenesis in response to physiologic fasting and feeding.


Introduction 1
The liver plays a critical role in organismal energy homeostasis by regulating diverse biologic processes 2 in response to nutrient availability 1 . During fasting, the activation of hepatic gluconeogenesis is required for the 3 supply of glucose to tissues with a high glucose demand, such as the brain, and for the maintenance of 4 glucose homeostasis, whereas in the fed state gluconeogenesis is suppressed 2, 3 . Precise control of hepatic 5 gluconeogenesis is crucial for normal physiology, and a failure to suppress hepatic gluconeogenesis post-6 prandially contributes to hyperglycemia in insulin resistance and diabetes 4 . 7 Glucagon and glucocorticoids (GCs; in humans, cortisol; in mice, corticosterone; synthesized in the 8 adrenal cortex) are hormones that are secreted during fasting and promote hepatic gluconeogenesis at 9 multiple levels, including via gene transcription. Glucocorticoid receptor (GR, encoded by the NR3C1 gene) is 10 a member of the nuclear receptor super-family 5 that is a key transcriptional regulator of gluconeogenic gene 11 expression in the fasting state 6 . GR not only directly responds to increases in GC concentration by activating 12 gluconeogenic gene transcription, but also plays a permissive role in the glucagon-mediated transcriptional 13 control of these genes. Therefore, the deletion of hepatic GR leads to fasting hypoglycemia 7 , while 14 adrenalectomy abrogates the induction of gluconeogenesis by fasting, glucagon, cAMP, or epinephrine in 15 rodents 8,9 . Similarly, a single dose of a GR antagonist is sufficient to reduce hepatic glucose output in healthy 16 humans 10 . 17 GR transactivation of gluconeogenic genes involves a series of molecular events 6 . Upon the binding of 18 a ligand (a synthetic agonist or an endogenous GC) 11 , GR undergoes conformational changes, translocates to 19 the nucleus, dimerizes, and binds to glucocorticoid response elements (GREs) in the promoters of key 20 gluconeogenic genes, such as phosphoenolpyruvate carboxykinase 1 (PCK1) and glucose-6-phosphatase 21 catalytic subunit (G6PC) 11,12 , which encode the rate-limiting and final enzymes of gluconeogenesis, 22 respectively 13, 14 . This activation mechanism differs from that of the transrepression of inflammatory genes by 23 GR, which involves the tethering of monomeric GR to DNA-bound proinflammatory transcription factors 15 . 24 Yes-associated protein 1 (YAP) and transcriptional co-activator with PDZ-binding motif (TAZ) are 25 downstream effectors of the Hippo pathway 16 . Inhibition of the Hippo pathway activates YAP and TAZ, and 26 they co-activate TEA domain (TEAD) transcription factors within the nucleus, which induce the expression of 27 genes involved in cellular proliferation 16 . However, YAP and TAZ also play other roles, mediated by 1 interactions with diverse transcription factors 17 . Although YAP and TAZ share nearly 50% amino acid sequence 2 similarity, the two proteins have distinct functions that are exerted via interactions with different transcription 3 factors 17,18 . For example, YAP activates epidermal growth factor receptor 4 (ErbB4) and tumor protein p73 19, 20 , 4 whereas TAZ specifically interacts with peroxisome proliferator-activated receptor gamma (PPARγ) and T-box 5 transcription factor (TBX5) 21,22 . In addition, although TAZ co-activates TEADs, it also acts as a repressor: the 6 binding of TAZ to PPARγ inhibits the PPARγ-induced differentiation of mesenchymal stem cells into 7 adipocytes 21 . Thus, YAP and TAZ have distinct roles in gene promoter-specific transcriptional regulation. 8 We previously reported that YAP integrates gluconeogenic gene expression and cell proliferation, 9 which contributes to its tumorigenic effects. However, YAP does not regulate normal glucose homeostasis, 10 because hepatocyte-specific deletion of YAP in normal mice has little effect on gluconeogenic gene expression 11 or blood glucose concentration 23 . Whereas YAP is primarily expressed in cholangiocytes in normal mouse 12 liver 24 , we show here that TAZ is abundantly expressed in pericentral hepatocytes and that its expression is 13 markedly reduced by fasting. In accordance with these data, we show that TAZ, but not YAP, interacts with GR 14 to inhibit the GR-transactivation of gluconeogenic genes, thereby coordinating hepatic glucose production with 15 physiologic fasting and feeding in normal mouse liver. 16 17

Fasting and feeding alters hepatic TAZ protein level 19
To determine whether TAZ plays a role in hepatic metabolic regulation under normal physiologic 20 conditions, we measured hepatic TAZ expression in mice that had been fed ad libitum or fasted for 24 h. As 21 expected, fasting increased the hepatic mRNA expression of genes encoding the key gluconeogenic enzymes 22 Pck1 and G6pc, but not those of canonical TAZ and YAP target genes involved in cell proliferation 16 , including 23 connective tissue growth factor (Ctgf) and cysteine-rich angiogenic inducer 61 (Cyr61) (Figure 1-Figure  24 supplement 1A-B). Hepatic TAZ protein level was reduced by >50% after 24 h of fasting, relative to ad libitum 25 feeding ( Figure. 1A-B). The phosphorylation of TAZ at serine 89, which is mediated by the Hippo pathway core 26 kinase large tumor suppressor kinase (LATS)1/2 25 , was commensurate with the reduction in total TAZ ( Figure  1 1A-B). Both nuclear and cytoplasmic TAZ proteins shared this regulation ( Figure 1C-D). 2 Immunohistochemistry confirmed that TAZ was abundantly expressed in both the nuclear and 3 cytoplasmic compartments of mature hepatocytes and that its expression was reduced by fasting ( Figure 1E). 4 The antibody used was validated using liver sections from liver-specific TAZ knockout (L-TAZ KO) mice (which 5 were generated by crossing TAZ floxed mice with Albumin-cre mice) (Figure 1-Figure supplement 2). 6 Interestingly, TAZ showed zonal expression, with the highest protein levels being found in the pericentral (or 7 perivenous) hepatocytes ( Figure 1E). This zonal expression negatively correlated with the expression and 8 activity of gluconeogenic genes in the liver lobule 26 . Immunoblotting of isolated pericentral and periportal 9 hepatocytes confirmed that TAZ was primarily expressed in glutamine synthetase (GLUL)-expressing 10 pericentral hepatocytes; whereas YAP is primarily expressed in periportal hepatocytes ( Figure 1F). These 11 results are consistent with the higher TAZ mRNA expression in pericentral than periportal mouse hepatocytes, 12 revealed by single-cell RNA analysis 27 . In addition, TAZ protein was less abundant in hepatocytes isolated 13 from fasting than ad libitum-fed mice ( Figure 1G), suggesting that fasting reduces TAZ protein in hepatocytes. 14 In contrast to TAZ protein levels, TAZ mRNA was not affected by fasting or feeding ( Figure 1H), 15 indicating that TAZ is post-transcriptionally regulated by physiologic fasting and feeding. These data are 16 consistent with previous findings that TAZ is subject to ubiquitin-mediated degradation 28 . Consistent with this, 17 TAZ protein, but not its mRNA, was induced in mouse primary hepatocyte cultures by supplementation of the 18 medium with a high glucose concentration (25 mM) and 10% fetal bovine serum (FBS), which mimics the fed 19 condition, in a time-dependent manner ( To define the role of TAZ in glucose homeostasis in mouse liver, we acutely knocked down hepatic TAZ 24 using adenoviruses expressing AdshTAZ or control AdshCon. Compared with the control virus, administration 25 of AdshTAZ to C57BL/6J mice reduced hepatic TAZ mRNA and protein levels (Figure 2A-B). TAZ knockdown 26 did not alter mouse body or epididymal white adipose tissue mass, but it slightly reduced liver mass, without 27 affecting liver histology (Figure 2-Figure supplement 1A-D). However, TAZ knockdown significantly increased 1 the mRNA expression of hepatic Pck1 and G6pc by 6-and 2-fold ( Figure 2C), respectively, and PCK1 and 2 G6PC protein levels 2-fold ( Figure 2D-E). Consistent with this, knockdown of hepatic TAZ significantly 3 increased the ad libitum-fed and fasting blood glucose concentrations ( Figure 2F). Mice with hepatic TAZ 4 knockdown also showed larger blood glucose excursions than control mice when challenged with pyruvate, a 5 gluconeogenic substrate ( Figure 2G). By contrast, the knockdown of TAZ did not alter the levels of YAP or key 6 gluconeogenic factors (e.g., hepatic nuclear receptor alpha (HNF4α), forkhead box O1 (FoxO1), and GR) 29 , 7 suggesting that the inhibition of gluconeogenic gene by TAZ is unlikely to occur via regulation of the protein  To determine whether overexpression of TAZ is sufficient to inhibit the expression of gluconeogenic 2 genes in the fasting state, we constructed an adenovirus expressing FLAG epitope-tagged human TAZ 3 (AdTAZ), which expresses a slightly larger protein than endogenous mouse TAZ ( Figure 3A-B). 4 Immunohistochemistry confirmed the expression of AdTAZ in both pericentral and periportal hepatocytes 5   Consistent with these data, AdTAZ infection also reduced glucose production after glucagon stimulation 26 ( Figure 4H). In addition, knockdown or overexpression of TAZ did not alter the protein levels of YAP or 27 gluconeogenic transcription factors (Figure 4-Figure supplement 1). Given that insulin and glucagon are major 1 regulators of gluconeogenic genes, we determined whether TAZ regulates their action in hepatocytes, and 2 found that it had no effects on either the basal or stimulated phosphorylation of CREB or AKT in hepatocytes 3 treated with glucagon or insulin, respectively (Figure 4-Figure supplement 2), suggesting that TAZ is unlikely 4 to directly affect gluconeogenic gene expression by altering the cellular sensitivity to these hormones in 5 hepatocytes. 6 7

TAZ inhibits the GR-transactivation of gluconeogenic genes 8
We next determined the molecular mechanisms by which TAZ inhibits gluconeogenic gene expression. 9 Because TAZ is not known to bind directly to gluconeogenic gene promoters, we determined whether the indicating that TAZ suppresses transcription of these genes. To determine whether the Hippo pathway and 16 TEADs are required for these effects of TAZ, we utilized TAZS89A 25 and TAZS51A 33 mutants, which abolish 17 TAZ phosphorylation by LATS1/2 and cannot interact with TEADs, respectively. Both mutants inhibited G6PC-18 Luc to a similar extent, compared to wild-type TAZ ( Figure 5A  To identify the molecular target of TAZ, we constructed a luciferase reporter containing three repeats of 23 a canonical GR response element (inverted hexameric half-site motifs, separated by a three-base-pair spacer) 24 (GRE-Luc) 34 . Similar to our findings using G6PC-Luc and PCK1-Luc, TAZ suppressed dexamethasone-25 stimulated GRE-Luc activity by >50%, whereas knockdown of TAZ increased promoter activity by 40% ( Figure  26 5C-D), suggesting that TAZ inhibits GR transactivation. 27

The interaction of TAZ with GR is dependent on its WW domain 1
The WW domain of TAZ has been reported to interact with proteins containing an I/L/PPxY (I, 2 isoleucine; L, leucine; P, proline; x, any amino acid; and Y, tyrosine) motif 35 21 . Moreover, an I/LPxY motif found 3 in the ligand binding domain of GR is highly conversed across species, including in humans, mice, and rats 4 ( Figure 5-Figure supplement 2). Co-immunoprecipitation (co-IP) assay revealed that TAZ, but not YAP, was 5 able to interact with GR ( Figure 5E). These data are consistent with our previous findings that YAP does not 6 interact with PGC1α 23 , a co-activator of GR. A TAZ mutant protein lacking the WW domain (TAZΔWW), but not 7 one lacking the coiled-coil (CC) domain (TAZΔCC), showed much weaker interaction with GR ( Figure 5F) and 8 was unable to inhibit GRE-Luc ( Figure 5G). Consistent with this, compared with the TAZS89A mutant that  Conversely, a GR mutant in which the conserved IPKY motif was mutated to alanines (A) (GR4A 16 mutant) interacted with TAZ much more weakly, despite being expressed at a level similar to that of the wild-17 type GR ( Figure 5M). In addition, compared with wild-type GR, the GR4A mutant was not subject to TAZ-18 mediated inhibition of GRE-Luc and displayed a greater ability to induce the activities of the GRE-Luc and 19 G6PC-Luc reporters ( Figure 5N-O). 20

TAZ inhibits the binding of GR to GREs 22
To understand how the interaction between TAZ and GR inhibits GR transactivation, we determined 23 how the binding of TAZ to GR inhibits GR nuclear localization, dimerization, and binding to promoter GREs. fact that the GR4A mutant can be activated by Dex ( Figure 5N-O) also suggests that the TAZ-GR interaction 2 does not impair GR nuclear localization, ligand activation, or dimerization. 3 To evaluate GR binding to the G6pc and Pck1 promoters, we performed chromatin immunoprecipitation 4 (ChIP) assays in primary mouse hepatocytes using control IgG and anti-GR antibodies. Glucagon treatment 5 strongly promoted the binding of GR to the Pck1 and G6pc promoter regions containing GREs ('Pck1 (GRE)', 6 −376 to −280 and 'G6pc (GRE)', −215 to −111), but not to a distal region of the Pck1 promoter lacking GREs 7 ('Pck1 (Con)', −3,678 to −3,564). However, the glucagon-induced binding of GR to the Pck1 or G6pc 8 promoters was significantly lower in cells overexpressing TAZ ( Figure 5P). Taken together, these data suggest 9 that the TAZ-GR interaction limits gluconeogenic gene expression by reducing GR binding to GREs. 10 To determine whether TAZ might have significant GR-independent effects to reduce hepatic 11 gluconeogenic gene expression, we treated primary mouse hepatocytes with RU486, a well-characterized and 12 potent GR antagonist 36 . Upon binding to GR, RU486 induces a conformational change that favors the 13 interaction of GR with co-repressors, but stabilizes the binding of dimeric GR to GREs 36 . Interestingly, we 14 found that RU486 blocked the interaction of GR with TAZ ( Figure 5Q), probably due to the change in GR G6pc expression in hepatocytes ( Figure 5S). Thus, not only is the TAZ-GR interaction necessary for the 20 reduction in gluconeogenic gene expression, but it is likely the sole mechanism for such a reduction. 21

TAZ inhibits GR transactivation in mice 23
Consistent with the interaction between GR and TAZ identified in cultured cells, GR in the nuclear 24 extracts of wild-type mouse liver could be immunoprecipitated using an antibody targeting endogenous TAZ 25 ( Figure 6A). To verify that TAZ regulates the association of GR with promoter GREs in mouse liver, we 26 conducted ChIP assays of endogenous gluconeogenic gene promoters. GR bound to regions of the G6pc and 27 Pck1 promoters containing functional GREs ( Figure 6B). The binding of GR to GRE-containing regions of the 1 G6pc and Pck1 promoters was significantly increased by TAZ knockdown ( Figure 6B). Moreover, the histone 2 acetylation of G6pc and Pck1 promoter regions containing or near to GREs was increased 50-80% by TAZ 3 knockdown ( Figure 6C), implying active transcription from the G6pc and Pck1 genes. Conversely, ChIP assays 4 revealed that TAZ overexpression caused a 25-45% reduction in the binding of GR to GREs, histone 5 acetylation, and the binding of the polymerase II subunit (Pol II) to the promoter regions of G6pc and Pck1 6 ( Figure 6D-F). However, TAZ bound normally to the promoter of Ctgf (a well-described TAZ target gene that is 7 not known to be regulated by GR) in AdTAZ-infected mice, but not to the promoters of G6pc or Pck1 ( GR and suppressing GR binding to GREs, rather than a model in which TAZ binding near or within GREs 10 inhibits the binding of GR to gluconeogenic gene promoters. 11 We next determined whether the inhibition of gluconeogenic gene expression by TAZ requires GR. 12 Because RU486 reduced the TAZ-GR interaction in cultured cells ( Figure 5Q) and mouse liver nuclear extracts 13 ( Figure 6G), we determined whether RU486 could abrogate the effects of TAZ on glucose homeostasis in mice. 14 As expected, TAZ knockdown increased mouse blood glucose concentration prior to RU486 injection. However, 15 this increase was completely abolished by RU486 treatment ( Figure 6H). Similarly, hepatic TAZ knockout 16 increased fasting blood glucose and glucose production after pyruvate administration to mice; and RU486 17 treatment abolished the effects of TAZ on blood glucose and glucose production ( Figure 6I-J). Consistent with 18 these loss-of-function studies, RU486 also entirely abolished the ability of overexpressed hepatic TAZ to 19 reduce gluconeogenic gene expression, to improve pyruvate tolerance, and to inhibit gluconeogenic gene 20 expression ( Figure 6K-M). Moreover, RU486 abolished the ability of TAZ to reduce GR binding to GREs in the 21 Pck1 or G6pc promoters and the amounts of acetylated histones on these promoters ( Figure 6N). These data 22 confirm that the interaction between GR and TAZ is required for the inhibitory effects of TAZ on hepatic 23 gluconeogenic gene expression in mice. 24 The binding of dimeric GR to GREs is required for the GR transactivation of gluconeogenic genes, but 25 not GR transrepression of anti-inflammatory genes 15 . Therefore, we next determined the effects of TAZ on the 26 regulation of gluconeogenic and inflammatory genes by Dex in mouse liver. TAZ overexpression inhibited the 27 Dex-induced increases in blood glucose and gluconeogenic gene expression, but the effects of TAZ on the 1 inhibition of inflammatory gene expression by Dex, including the expression of the tumor necrosis factor alpha 2 (Tnfα) and interleukin 1 (Il1) genes, were inconsistent (  together, these data suggest that the regulation of gluconeogenic genes by TAZ does not require insulin 21

signaling. 22
Fasting increases GR binding to gluconeogenic gene promoters in mouse liver and feeding reduces 23 this binding 38 . Consistent with this, we found that the binding of GR to GREs in the G6pc and Pck1 promoters 24 was reduced by 40% in the livers of mice that were fasted, versus those that were fed ( Figure 7H). Collectively, 25 our data suggest a role for hepatic TAZ in glucose homeostasis in normal mouse liver. In the fed state, high 26 hepatic TAZ expression inhibits GR transactivation of gluconeogenic genes by interacting with GR and 27 reducing the binding of GR to the promoters of these genes; whereas in the fasted state, lower TAZ expression 1 enables GR to bind to gluconeogenic genes and activate their transcription, which increases glucose 2 production ( Figure 7I). Distinct cellular functions of YAP and TAZ have been suggested by several studies 17,18,39,40 . In the liver, 7 YAP is expressed at low levels in normal hepatocytes, which is crucial to the maintenance of their 8 differentiated state 24 . Consistent with this, the knockout of hepatic YAP in adult mouse liver has no effect on 9 gluconeogenic gene expression or glucose homeostasis 23 . However, hepatic YAP expression negatively 10 correlates with the expression of gluconeogenic genes in human hepatocellular carcinoma, which suggests a 11 role for YAP in the integration of glucose metabolic regulation and cell growth 23 . By contrast, TAZ is expressed 12 in normal pericentral hepatocytes and regulates normal glucose homeostasis. The WW domain of TAZ is 13 required for its interaction with GR, because a TAZ mutant lacking this domain cannot interact with GR and 14 does not inhibit the transactivation of gluconeogenic genes by GR. Interestingly, although YAP also contains 15 WW domains, it does not interact with GR ( Figure 5E) or a GR co-activator, PGC1α 23 . Therefore, these results 16 demonstrate the distinct roles of TAZ and YAP in the modulation of GR activity and hepatic physiology. 17 The essential roles of GR and GCs in the promotion of hepatic gluconeogenic gene transcription are 18 well established. Hepatocyte-specific GR ablation leads to the death of half of newborn mice and severe 19 hypoglycemia in adult mice because of a defect in gluconeogenesis 7 . In addition, GR is required for the full 20 induction of glucose production, because fasting, glucagon, cyclic AMP (cAMP), and epinephrine-activated 21 glucose production are substantially blunted in adrenalectomized rodents 8,9 . The binding of GR to the 22 promoters of gluconeogenic genes is higher in the fasting state than in the fed state 38 , and GCs are believed to 23 play a role in this dynamic regulation. However, little is known about the GC-independent factor(s) or 24 mechanism(s) mediating this regulation. Our data reveal hepatic TAZ to be a novel regulator of GR that 25 modulates the binding of GR to gluconeogenic gene promoters. First, TAZ inhibits GR transactivation. The 26 overexpression of TAZ inhibits, while the knockdown of TAZ increases, the promoter activity of GRE-Luc, 27 G6PC-Luc, and PCK1-Luc reporter genes. In addition, TAZ inhibits the induction of gluconeogenic gene 1 expression by Dex. Second, TAZ acts as a repressor of GR. However, unlike a classical co-repressor of GR, 2 such as nuclear receptor compressor 1 (NCoR1), which does not alter DNA-binding by GR 41 , the binding of 3 TAZ to GR results in the dissociation of GR itself from the promoters of GR target genes. Consistent with this, 4 the overexpression of TAZ reduces GR binding to GREs in the promoters of gluconeogenic genes, whereas 5 the knockdown of TAZ increases GR binding to these GREs. Third, the effects of TAZ overexpression or 6 knockdown on gluconeogenic gene expression and the binding of GR to the Pck1 and G6pc promoters are 7 completely abolished in hepatocytes in the presence of RU486, which prevents the interaction of GR with TAZ. 8 RU486, a GR antagonist, binds to GR but elicits conformational changes that do not favor the 9 recruitment of histone acetyltransferase without an inhibition of GR nuclear localization and DNA-binding 36 . 10 Thus, one explanation for the reduction in the interaction between GR and TAZ following RU486 treatment is 11 that RU486-bound GR possesses a conformation that does not permit GR to interact with TAZ. The ligand 12 binding domain of GR is crucial for the interaction of GR with co-regulators and ligand-binding, but also for the 13 formation of homodimers, which permit GR to bind to GREs in gluconeogenic gene promoters. Our results 14 suggest that the binding of TAZ to the LBD of GR does not lead to a reduction in the nuclear localization of GR 15 or in its ability to respond to an agonist; however, exactly how the binding of TAZ to GR causes GR to 16 dissociate from GRE is unknown. GRβ, which is produced by alternative splicing, differs only at its C-terminus 17 from GRα, where the I/LPxY motif resides. Interestingly, unlike GRα, GRβ is not able to bind to conventional 18 GREs in gluconeogenic gene promoters and activate gluconeogenic gene transcription 42 , which implies that 19 these C-terminal amino acids are indispensable for the binding of GR to GREs. 20 The inhibition of gluconeogenic gene expression by TAZ is hepatocyte-autonomous and is unlikely to 21 be an indirect effect of glucagon or insulin signaling. TAZ does not affect the phosphorylation of CREB in 22 hepatocytes, nor are plasma glucagon concentrations altered when TAZ is overexpressed or knocked down. 23 Previous studies show that the overexpression of TAZ increases insulin signaling in liver cancer and that TAZ 24 deletion in white adipose and muscle increases insulin sensitivity 25, 30 . However, we found little effect of either 25 TAZ overexpression or knockdown on the phosphorylation of AKT or FoxO1 (Ser256), or on the plasma insulin 26 concentration and we found little effects of hepatic TAZ knockout on whole-body insulin sensitivity. In addition, 27 the overexpression of hepatic TAZ is sufficient to reduce blood glucose, pyruvate tolerance, and the hepatic 1 expression of Pck1 and G6pc in liver-specific IRS 1 and 2 double-knockout mice 43 , which strongly suggests 2 that the inhibition of the GR transactivation of gluconeogenic genes by TAZ is independent of hepatic insulin 3 signaling. 4 TAZ protein, but not mRNA expression, was altered by fasting and feeding, which suggests that TAZ is 5 subject to post-transcriptional regulation. It is also noteworthy that TAZ is zonally distributed, with the highest 6 expression being near the central veins and the lowest near the hepatic veins. This pattern of distribution is 7 negatively related to the expression and activity of gluconeogenic genes 26 , which is consistent with an 8 inhibitory effect of TAZ on hepatic gluconeogenesis under normal physiologic conditions. However, the 9 molecular mechanism(s) that mediate the post-transcriptional effects and zonal distribution of TAZ are unclear 10 and are currently under investigation in our laboratory. 11 Gluconeogenesis is substantially upregulated in diabetic patients and contributes significantly to the 12 lack of control of blood glucose concentration in these patients 4, 44 . Long-term increases in the concentration 13 and/or actions of endogenous GCs in both humans and rodents manifest as metabolic syndrome, which is 14 characterized by higher glucose production and insulin resistance, and resembles Cushing's syndrome 45 . GCs 15 are among the most widely prescribed anti-inflammatory and immunosuppressive drugs 46 . However, GC 16 therapy is associated with substantial increases in glucose production and insulin resistance; thus, therapies 17 that would have the anti-inflammatory effects of GCs, but not affect gluconeogenesis, would be of great 18 interest 47 . Similarly, strategies aimed at specifically inhibiting the ability of GR to induce gluconeogenesis have 19 been explored for the treatment of diabetes 48 . Therefore, whether the inhibition of GR by TAZ plays a role in 20 the regulation of hepatic gluconeogenesis in these abnormal states, and whether a non-tumorigenic TAZ 21 mutant(s), or small peptide(s) or molecule(s), which would mimic or enhance the TAZ-GR interaction could 22 normalize the hyperglycemia associated with obesity, insulin resistance, or the chronic use of GCs, should be 23 determined in future investigations. 24 In summary, the factors and mechanisms that regulate cell proliferation substantially, such as the 25 mTOR complex 49 , overlap with those that control metabolic homeostasis, and have emerged as an important 26 area of biology in recent years. We have identified hepatic TAZ, a downstream effector of the Hippo pathway, 27 as a novel regulator of glucose metabolism in the normal liver. TAZ acts as a GR co-repressor and inhibits the 1 binding of GR to GREs in the promoters of gluconeogenic genes, whereby it coordinates GR transactivation 2 and hepatic glucose production in response to fasting and feeding, to maintain energy balance ( Figure 7I).

Animals and treatments 3
All mice were fed a standard chow diet and maintained on a 12 h light/dark cycle, with free access to 4 food and water, unless otherwise indicated, and were sacrificed in the ad libitum-fed, fasted, or re-fed states, 5 as indicated in the main text. At that time, the livers were removed and then fixed in 4% formaldehyde or frozen 6 in liquid nitrogen and stored at −80°C, until RNA, protein, and immunoprecipitation analyses were performed. 7 All animal experiments were performed with the approval of the Institutional Animal Care and Research 8 Advisory Committee at Boston Children's Hospital. Both male and female cohorts of the same age were 9 studied and we found consistent results in each. 10 To knockdown hepatic TAZ, eight-to twelve-week-old C57BL/6J mice were administered adenoviruses 11 expressing shTAZ (AdshTAZ) or control shRNA (targeting human lamin) by retro-orbital injection at a dose of 12 1-2 × 10 9 pfu/mouse. Six days after virus injection, mice were fasted for 14 hours and subjected to a pyruvate 13 tolerance test (PTT), as previously described 50 . Briefly, mice were injected intraperitoneally with 2 mg/kg 14 pyruvate at time zero, and subsequently blood glucose concentrations were measured using a glucometer 15 (Contour). Alternatively, eight days after virus infection, the mice were sacrificed in the ad libitum state at 16 zeitgeber time (ZT) 13 for gene expression and chromatin immunoprecipitation analysis. To overexpress TAZ 17 in the liver, mice were injected with adenovirus expressing human FLAG-TAZ (AdTAZ) or control GFP (AdGFP) 18 retro-orbitally at a dose of 0.5-1 × 10 9 pfu/mouse. Five days later, the mice were subjected to PTT or glucagon 19 challenge after an overnight fast or sacrificed at ZT 13 after a 24 h fast. For glucagon challenge experiment, 20 eight-to twelve-week-old C57BL/6J mice were injected with glucagon (250 μg/kg, Sigma) intraperitoneally at 21 time zero, and subsequently blood glucose concentrations were measured using a glucometer. To treat RU486 22 (Sigma), mice were intraperitoneally injected with RU486 (50 mg/kg), fasted for 14 h, subjected to a PTT, and 23 then sacrificed 1 day later. To treat dexamethasone (Sigma), mice were injected intraperitoneally with 24 dexamethasone (10 μg/kg) daily for 3 days. 25 L-TAZ KO mice were generated by crossing TAZ flox/flox mice (Jackson Laboratory) with Albumin-Cre 1 mice (Jackson Laboratory). Age and sex-matched knockout and littermate control mice were studied. Mice 2 underwent PTT and RU496 treatment as described in C57BL/6J mice. 3 4

Histology, immunohistochemistry, and plasma insulin and glucagon measurements 5
Portions of liver were fixed overnight in 4% paraformaldehyde, embedded in paraffin, and sectioned, 6 and the sections generated were hematoxylin and eosin-stained and their histology was assessed. For TAZ 7 immunohistochemistry, liver sections underwent antigen retrieval in boiling 10 mM sodium citrate buffer (pH = 8 6) for 10 min and then were blocked with 5% goat serum in PBS for 1 h, treated with 3% hydrogen peroxide for 9 30 min at room temperature, and incubated with primary antibodies overnight at 4°C, followed by horseradish 10 peroxidase (HRP)-conjugated goat anti-rabbit IgG secondary antibodies (Vector Labs) for 1 h at room 11 temperature. Immunoreactivity was visualized by incubating slices with VIP peroxidase substrate (Vector Labs), 12 with cell nuclei being stained blue using methyl green (Sigma). All images were obtained using an EVOS2 13 microscope (Life Technologies). Plasma insulin and glucagon were measured using commercial 14 chemiluminescent ELISA kits (Alpco). 15 16

Plasmid and adenoviral vector constructs 17
Human cDNAs of wild-type TAZ and mutants containing deletions of the WW or CC domains (gifts from 18 Dr. Jeff Wrana: Addgene #24809, #24811, and #24816) were cloned into a pcDNA3-FLAG vector by PCR, 19 restriction enzyme digestion, and ligation. A cDNA expressing the human GR was purchased from ATCC and 20 was cloned into a pcDNA3 vector for overexpression in mammalian cells 51 . TAZS89A, TAZS51A, and GR4A 21 mutants were constructed using site-directed mutagenesis (Agilent). The GRE-Luc construct was constructed 22 by annealing two complementary DNAs containing three repeats of a canonical GRE (two inverted hexameric 23 half-site motifs separated by a three-base-pair spacer), followed by ligation into a pGL3 basic vector (Agilent). 24 The human PCK1 and G6PC luciferase constructs (PCK1-Luc and G6PC-Luc) and the plasmids encoding 25 HNF4α and PGC1α were gifts from Dr. Pere Puigserver 52 . The GFP-GR construct was a gift from Dr. A. Wong 26 (Addgene #47504), the pcDNA3-FLAG-human YAP1 vector was a gift from Dr. Yosef Shaul (Addgene 27 #18881), the TEADE-Luc (8xGTIIC-Luc) construct containing eight repeats of the TEAD response element was 1 a gift from Dr. Stefano Piccolo (Addgene #34615) 53 , and the TEAD1 expression vector was a gift from Dr.  Table 1. Adenoviruses overexpressing FLAG-tagged human TAZ or control GFP 7 were constructed using the AdTrack system 54 . All viruses were amplified in 293A cells (Life Technologies), 8 purified using cesium chloride gradient centrifugation, and titered using an end-point dilution method. The 293A 9 cells were maintained in high glucose (4.5 g/L) Dulbecco's modified Eagle's medium (DMEM) supplemented 10 with 100 units/ml penicillin, 100 units/ml streptomycin, and 10% FBS at 37°C and in a 5% CO 2 -containing 11 atmosphere in a humidified incubator, and they were free of mycoplasma contamination. Cell culture reagents 12 were purchased from Life Technologies unless otherwise indicated. 13 14

Primary hepatocyte studies 15
Primary mouse hepatocytes were isolated from eight-to twelve-week-old male C57BL/6J mice, as 16 described previously, using a two-step collagenase perfusion method 50 , and they were maintained at 37°C 17 and in a 5% CO 2 -containing atmosphere in a humidified incubator. Hepatocytes were seeded in collagen-18 coated 6-well dishes at a density of 5 × 10 5 cells/well in M199 medium supplemented with 2 mM glutamine, 100 19 units/ml penicillin, 100 units/ml streptomycin, and 10% FBS. After 4 h, the cells were washed with PBS and 20 incubated in fresh medium, in which various treatments were administered. 21 To measure glucose production, hepatocytes were fasted overnight in low glucose (1 g/L) DMEM 4 media and then incubated in glucose production medium (2 mM L-carnitine and 2 mM pyruvate, without 5 glucose and phenol red) in the presence of glucagon (20 nM). Small aliquots of medium were removed at 6 various time points, and the concentrations of glucose were measured using an Amplex glucose assay kit (Life 7 Technologies), as per the manufacturer's instructions. 8 Peripcentral and periportal hepatocytes were isolated as previously described 55 . Briefly, primary mouse 9 hepatocytes isolated via the two-step collagenase perfusion method were suspended in 1 ml DMEM media 10 and subjected to Percoll (GE Healthcare) gradient centrifugation at 1,000g for 30 minutes. The gradient 11 consists of 1 ml 70% Percoll followed by 3 ml 52%, 4 ml 42%, and 5 ml 30% Percoll. The periportal and 12 pericentral cell layers were removed and washed with PBS prior to immunoblotting. Renilla plasmid as an internal control for transfection efficiency, and 10-300 ng/well of expression vector. 21 Empty vector was added to ensure that the same amount of total DNA was used in each transfection reaction. 22 On day 2, the cells were either harvested or placed in serum-free media, with or without 100 nM Dex, for an 23 additional 16-20 h, and were harvested on day 3 by scraping into passive cell lysis buffer (Promega). 24 Luciferase activity was then measured using a commercial dual luciferase assay kit (Promega), after which 25 firefly luciferase activity was normalized to Renilla activity and the values for the control groups were set to 26 one. 27

RNA isolation and analysis 1
RNA was isolated using Trizol reagent (Life Technologies), according to the manufacturer's instructions. Whole cell lysates were prepared by collecting cells in lysis buffer (50 mM Tris, pH 7.5, 150 mM NaCl, 1 12 mM EDTA, 1% NP-40, 0.5% sodium deoxycholate, 1.0% SDS, 2 mM NaF, and 2 mM Na 3 VO 4 , supplemented 13 with protease and phosphatase inhibitors [Sigma]), sonicating, and then centrifuging at 13,000 × g for 10 min 50 . 14 Some cells were treated with 1 µM DSP (Life Technologies), a cross-linker, for 30 min on ice prior to 15 harvesting. Liver homogenates were prepared by homogenizing liver tissue in lysis buffer 50 . Nuclear extracts 16 were prepared using an NE-PER extraction kit (Thermo Scientific), according to the manufacturer's instructions, 17 or as previously described 56 . Briefly, livers were homogenized in a Dounce homogenizer in hypotonic buffer 18 (15 mM HEPES, pH 7.9, 1.5 mM MgCl 2 , 10 mM KCl, 0.2% NP-40, 1 mM EDTA, 5% sucrose, and 1 mM 19 dithiothreitol, supplemented with protease and phosphatase inhibitors). The homogenate was layered onto a 20 sucrose cushion buffer (300 mM sucrose, 60 mM KCl, 10 mM Tris-HCl, pH 7.5, 1 mM EDTA, 0.15 mM 21 spermine, 0.5 mM spermidine, and 1 mM dithiothreitol) and centrifuged at 2,000 × g for 2 min. 22 Lysates were subjected to SDS-PAGE and transferred onto a PVDF membrane (Thermo Scientific). 23 After blocking, blots were incubated overnight with primary antibody (1:1,000 to 5,000 dilution), and then 24 secondary antibody conjugated to horseradish peroxidase (Thermo Scientific) and chemiluminescent ECL 25 reagents (Thermo Scientific) were used to identify specific bands. Band intensities were determined using 26 ImageJ and normalized to the intensity of loading control bands and the values of controls were set to 1. The 1 antibodies used in this study were obtained from commercial sources and are listed in Supplemental Table 3. 2 3

Co-immunoprecipitation 4
Because of poor transfection efficiency in HepG2 and primary mouse hepatocytes, co-IP assays were 5 performed in 293A cells after the co-transfection of expression vectors, as indicated in the Figures, for 36-48 h. 6 Cells were lysed in RIPA buffer (50 mM Tris, pH 7.5, 150 mM NaCl, 1 mM EDTA, 0.5% NP40, and 0.5% 7 sodium deoxycholate) supplemented with protease and phosphatase inhibitors, sonicated, and centrifuged at 8 13,000 × g for 10 min. The resulting cell lysates were incubated with anti-FLAG antibody overnight at 4°C. 9 Immunoprecipitates were obtained by adding Protein A/G agarose (Santa Cruz Biotechnologies), washed three 10 times with RIPA buffer containing 0.025-0.05% SDS, and subjected to immunoblotting. Alternatively, IP was 11 performed using nuclear extracts from mouse livers. The antibodies used for IP are listed in Supplemental 12 Table 3. 13 14

Chromatin immunoprecipitation 15
ChIP assays using primary hepatocytes and liver tissues were performed as previously described, with minor 16 modifications 23, 56 . Cells were cross-linked with 1% formaldehyde at room temperature for 15 min, and then the 17 reaction was stopped by the addition of 125 mM glycine for 5 min at room temperature. Cells were washed 18 with PBS, collected in harvest buffer (100 mM Tris-HCl, pH 9.4, and 10 mM DTT), incubated on ice for 10 min, 19 and then centrifuged at 2,000 × g for 5 min. The cells were then sequentially washed with ice-cold PBS, buffer 20 I (0.25% Triton X-100, 10 mM EDTA, 0.5 mM EGTA, and 10 mM HEPES, pH 6.5), and buffer II (200 mM NaCl, 21 1 mM EDTA, 0.5 mM EGTA, and 10 mM HEPES, pH 6.5). Livers were prepared in a similar manner: livers 22 from two to four mice were pooled, minced into small pieces, cross-linked with 1% formaldehyde, and 23 homogenized in hypotonic buffer (15 mM HEPES, pH 7.9, 1.5 mM MgCl 2 , 10 mM KCl, 0.2% NP-40, 1 mM 24 EDTA, 5% sucrose, and 1 mM dithiothreitol) supplemented with protease and phosphatase inhibitor cocktails. 25 The nuclei were isolated by centrifugation of the resulting homogenates laid over a cushion buffer (300 mM 26 sucrose, 60 mM KCl, 10 mM Tris-HCl, pH 7.5, and 1 mM EDTA). The pellets were resuspended in lysis buffer 27 (1% SDS, 10 mM EDTA, and 50 mM Tris-HCl, pH 8.0) supplemented with protease and phosphatase inhibitor 1 cocktails; sonicated to reduce the DNA length to 0.3-1.5 kb; and then centrifuged at 8,000 × g for 1 min at 4°C. 2 The soluble chromatin was diluted 10-fold in dilution buffer (1% Triton X-100, 2 mM EDTA, 150 mM NaCl, and 3 20 mM Tris-HCl, pH 8.0) and then pre-cleared with protein A/G sepharose beads (Santa Cruz Biotechnology) 4 at 4°C for 1 h (protein A/G sepharose was incubated with sheared salmon sperm DNA and washed three times 5 in dilution buffer prior to use). Equal amounts of pre-cleared chromatin were then added to 2-3 µg of antibody 6 overnight. The next day, 25 μl protein A/G sepharose beads were added and the incubation was continued for 7 another 2 h. The beads were collected and then washed in TSE I (0.1% SDS, 1% Triton X-100, 2 mM EDTA, 8 20 mM Tris-HCl, pH 8.0, and 150 mM NaCl), TSE II (0.1% SDS, 1% Triton X-100, 2 mM EDTA, 20 mM Tris-9 HCl, pH 8.1, and 500 mM NaCl), and TE buffer (10 mM Tris, pH 8.0, and 1 mM EDTA), and were finally eluted 10 in 1% SDS/0.1 M NaHCO 3 . Eluates were heated to 65°C for at least 6 h to reverse the formaldehyde cross-11 linking and then treated with proteinase K (Life Technologies). DNA fragments were isolated using a DNA 12 purification kit (Qiagen). The immunoprecipitated DNA and 10% of the pre-cleared chromatin (input DNA) were 13 then subjected to real-time PCR using Power SYBR Green PCR Master Mix (Biolines), in triplicate. For each 14 immunoprecipitate, we calculated the relative enrichment as 2 −ΔCt , where ΔCt was calculated as the mean Ct           Eight-to twelve-week-old C57BL/6J mice were administered AdshTAZ or AdshCon, then sacrificed 8 days later in the ad libitum-fed state. Hepatic proteins were measured by immunoblotting whole cell lysates. Eight-to twelve-week-old C57BL/6J mice were administered AdshTAZ or AdshCon, then sacrificed 8 days later in the ad libitum-fed state. Plasma insulin (F) and glucagon (G) were measured. Data are means and SEMs; n = 7-9. Data were analyzed by unpaired Student's t-test.         Data were analyzed by one-way ANOVA. **p < 0.01 and ***p < 0.001 denotes comparisons with wells transfected with TEAD4 alone.  Eight-to twelve-week-old male C57BL/6J mice were administered adenoviruses expressing TAZ (flag-tagged) or GFP for 5 days. ChIP assays were performed from liver extracts using indicated antibodies. Data are means and SEMs of 3-4 immunoprecipitates. Data were analyzed by two-way ANOVA; * p < 0.05 and ** p < 0.01; # p < 0.05, ## p < 0.01, and ### p < 0.001; # denotes a comparison with IgG.  Eight-to twelve-week-old C57BL/6J mice were administered adenoviruses expressing TAZ mutants (flagtagged) or GFP for 5 days. Body weight (A) and liver weight (B) were measured and hepatic protein levels were measured by immunoblotting whole cell lysates. Data are means and SEMs; n = 7-8. Data were analyzed by one-way ANOVA; **p < 0.01 and ### p < 0.001. Eight-to twelve-week-old female L-DKO and flox controls were administered adenoviruses expressing TAZ or GFP for 5 days. Body (A) and liver (B) weigh were measured. Data are means and SEMs; n = 7-8. Data were analyzed by unpaired Student's t-test *p < 0.05 and ***p < 0.001.