Hsp90 modulates PPARγ activity in a mouse model of nonalcoholic fatty liver disease.

Nonalcoholic fatty liver disease (NAFLD) is a highly prevalent complication of obesity, yet cellular mechanisms that lead to its development are not well defined. Previously, we have documented hepatic steatosis in mice carrying a mutation in the Sec61a1 gene. Here we examined the mechanism behind NAFLD in Sec61a1 mutant mice. Livers of mutant mice exhibited upregulation of Pparg and its target genes Cd36, Cidec, and Lpl, correlating with increased uptake of fatty acid. Interestingly, these mice also displayed activation of the heat shock response (HSR), with elevated levels of heat shock protein (Hsp) 70, Hsp90, and heat shock factor 1. In cell lines, inhibition of Hsp90 function reduced Pparγ signaling and protein levels. Conversely, overexpression of Hsp90 increased Pparγ signaling and protein levels by reducing degradation. This may occur via a physical interaction as Hsp90 and Pparγ coimmunoprecipitated in vivo. Furthermore, inhibition of Hsp90 in Sec61a1 mutant hepatocytes also reduced Pparγ protein levels and signaling. Finally, overexpression of Hsp90 in liver cell lines increased neutral lipid accumulation, and this accumulation was blocked by Hsp90 inhibition. Our results show that the HSR and Hsp90 play an important role in the development of NAFLD, opening new avenues for the prevention and treatment of this highly prevalent disease.

and triglyceride using a triglyceride assay kit (Cayman Chemicals, Ann Arbor, MI).

Quantitative PCR
All real-time PCR was performed on an ABI 7900HT using the Quantitect Probe RT-PCR Kit (Qiagen, Valencia, CA). Taqman primers and probes were from IDT (San Jose, CA). Data were analyzed using the comparative Ct method.

Lipid synthesis
Primary hepatocytes grown in 3.5 cm dishes were labeled with 150 nCi of 14 C-acetate (Perkin Elmer, Waltham, MA) for 16 h and then washed 3× with PBS, dissolved fi rst in 500 l of 0.1 M NaOH, then with 500 l of water, and combined. An aliquot was saved for protein determination. One milliliter of ethanol and 500 l of 75% KOH were added, and the sample was incubated at 85°C for 2 h followed by addition of 1 ml of ethanol. Cholesterol was extracted 3× using 1.5 ml of petroleum ether (Sigma). Afterward, 1 ml of concentrated HCl was added, and fatty acids were extracted 3× using 1.5 ml of petroleum ether. Counts associated with the different lipid fractions were then determined using a liquid scintillation counter and Ultima Gold XR cocktail (Perkin Elmer). Samples were normalized to protein content.

Fatty acid uptake
Primary hepatocytes were cultured in 24-well plates. Cells were washed 3× with PBS and incubated in 37°C assay buffer (PBS, 50 M fatty-acid-free BSA, 37.5 M sodium oleate, 500 nCi of 1-14 C oleic acid) for 5 min. Reactions were stopped with 1 ml of cold stop buffer (PBS, 0.1% fatty-acid-free BSA, and 200 M phloretin), washed 3×, and lysed with 300 l of 0.1 N NaOH. An aliquot was saved for protein determination, and the rest was added to 5 ml of Ultima Gold XR scintillation fl uid and read on a liquid scintillation counter. Nonspecifi c counts from cells incubated with ice-cold assay buffer mixed with stop buffer were subtracted from all values.

Hsp90 immunoprecipitation
Cells or tissues were lysed in lysis buffer (20 mM Tris pH 7.5, 125 mM NaCl, 1 mM EDTA, 20 mM Na 2 MoO 4 , 5 mM ATP, 0.5% NP-40). Lysates were precleared, and 100 g of protein for cells or 1 mg for tissue was incubated with 2 g of anti-FLAG antibody or anti-PPAR ␥ B-5 clone overnight at 4°C, followed by the addition of a 50% slurry of ␥ -bind for 1 h. Lysates were washed, eluted in SDS sample loading buffer, and fractionated by SDS-PAGE electrophoresis for Western blot.

Luciferase assay
F442A cells were incubated in 96-well plates and treated with the specifi ed compounds for 16 h before addition of Bright-glo protein that forms an aqueous pore in the ER membrane, which serves as the entry point for all newly synthesized proteins destined for the secretory pathway ( 11 ). The cause of apoptosis in these mice may be increased basal ER stress and reduced ER-associated degradation (ERAD), as has been demonstrated in yeast that carry this mutation, but others have documented disrupted calcium homeostasis and ER import of some substrates in cells carrying the Y344H mutation (12)(13)(14). These mice also exhibit hepatic steatosis. While ␤ -cell-specifi c expression of a wild-type Sec61a1 under control of the rat insulin promoter in these mice ( Sec61a1 Y344H/Y344H , RIP-Sec61) was suffi cient to rescue ␤ -cell apoptosis and diabetes, it did not rescue steatosis in the liver ( 10 ). Therefore, the Sec61a1 Y344H/Y344H , RIP-Sec61 mouse represents a novel, nondiabetic, nonobese model of hepatic steatosis without the requirement for dietary or chemical induction.

Animals
All mice were bred and housed at the Scripps Research Institute in accordance with institutional animal care and use committee and National Institutes of Health guidelines. All experiments used age-and sex-matched male littermates aged 12 weeks and maintained on a chow diet. The generation of the Sec61a1 Y344H/Y344H , RIP-Sec61 mouse has been described previously ( 10 ).
Cell culture and transfection. HEK293 and Huh7 cells were cultured in DMEM with 10% FBS (Hyclone) and penicillin and streptomycin. All transfections were performed with Lipofectamine 2000 (Invitrogen, Carlsbad, CA). F442A-Ap2Luc preadipocytes were maintained in DMEM-10% FBS and seeded in 96-well plates at 5 × 10 3 cells/well density. Two days after confl uency, medium was replaced with DMEM-10% FBS and differentiation cocktail (5 g/ml insulin, 2 M dexamethasone, and 0.5 mM isobutylxanthine). Two days later, medium was changed, and cells were maintained in DMEM-10% FBS and insulin 5 g/ml, changing every other day. Treatments were performed in DMEM-2% charcoal dextran-treated FBS overnight.

Liver lipid content
Lipids were extracted from liver homogenates using the method of Blighe and Dyer ( 15 ). Cholesterol levels were measured using the Amplex Red cholesterol assay kit (Invitrogen), widespread as indicated by staining of histological sections with Masson's trichrome ( Fig. 1C ) and was mainly localized to the perisinusoidal space (data not shown). This is in contrast to these mice on a high-fat diet, which show a much more signifi cant degree of fi brosis ( 10 ).
Because we have not previously seen differences between heterozygous and wild-type mice with respect to body mass, fat or lean mass distribution, fasting plasma lipid, and glucose or insulin levels, all subsequent experiments were performed with the unaffected heterozygous Sec61a1 +/Y344H , RIP-Sec61 mice as controls ( 10 ). Sec61a1 Y344H/Y344H , RIP-Sec61 mice weighed slightly less than heterozygotes (supplementary

AdipoRed staining
After transfection, HuH7 cells were incubated with 200 M oleate for 16 h, washed with PBS, and stained with AdipoRed (Lonza, Basel, Switzerland), diluted 1:200 in PBS, for 20 min. Cells were then washed and read on a Synergy Mx multimode plate reader (excitation: 485 nm; emission 572 nm).
Examining expression levels of genes that encode proteins involved in infl ammation and fi brosis indicated increased collagen synthesis and elevation of some markers of infl ammation ( Fig. 1B ), including elevation of serum alanine aminotransferase and aspartate aminotransferase (supplementary Fig. IA, B). However, fi brosis was not yet Increased expression of Pparg protein was confi rmed by Western blot, which showed an elevation in protein levels of both the Ppar ␥ 1 and Ppar ␥ 2 isoform in Sec61a1 Y344H/Y344H , RIP-Sec61 mice, although levels of Ppar ␥ 1 appeared to be the higher in absolute terms ( Fig. 2B ). It has been hypothesized that increased levels of Pparg mRNA and protein may be due to increased levels of C/EBP ␣ / ␤ proteins, which contribute to its expression in adipose tissue ( 16 ). However, we were unable to detect an increase in either C/EBP ␣ or ␤ levels by Western blot ( Fig. 2B ). Taken together, these results indicate that upregulation of Ppar ␥ in Sec61a1 Y344H/Y344H , RIP-Sec61 mice is through a C/EBP ␣ / ␤ -independent pathway.

Reduced synthesis and increased uptake of lipid in hepatocytes from Sec61a1 Y344H/Y344H
, RIP-Sec61 mice mRNA expression levels of lipid metabolic genes suggested that livers of Sec61a1 Y344H/Y344H , RIP-Sec61 mice should exhibit lower rates of de novo lipogenesis and increased uptake of free fatty acids. We tested both of these hypotheses at the biochemical level in primary hepatocytes. qPCR on isolated hepatocytes verifi ed that isolation did not alter gene expression with regard to Pparg and Cd36 mRNA levels ( Fig. 3A ). Using 14 C-acetate, we observed a decrease in synthesis of both cholesterol and fatty the media of primary hepatocytes from Sec61a1 Y344H/Y344H , RIP-Sec61 or heterozygous mice appeared the same (supplementary Fig. IIC).

Upregulation of Pparg and its target genes in Sec61a1 Y344H/Y344H
, RIP-Sec61 mice We next measured the expression of lipid metabolic genes in the livers of Sec61a1 Y344H/Y344H , RIP-Sec61 or heterozygous control mice, to reveal transcriptional pathways that might affect the development of steatosis. Transcription factors that affect lipid metabolism include Srebpf1 and Cepba , which are associated with lipid synthesis; Pparg , which promotes lipid storage; and Ppara , which directs fatty acid oxidation.
We found a reduction in Srebf1 mRNA levels in Sec61a1 Y344H/Y344H , RIP-Sec61 mice compared with heterozygous mice. This reduction coincided with decreased mRNA expression of Srebp-1c target genes Acacb , Scd1 , and Dgat2 ( Fig. 2A ). Furthermore, we did not detect a change in gene expression levels of Ppara or its targets Cpt1a and Acox1 ( Fig. 2A ). Most notable, however, was an 8-fold increase in Pparg1 and a 6-fold increase in Pparg2 mRNA in livers of Sec61a1 Y344H/Y344H , RIP-Sec61 mice and increased expression of their target genes: Cd36 , Lpl , and Cidec ( Fig. 2A ).

Fig. 2. Upregulation of Pparg and its targets in Sec61a1
Y344H/Y344H , RIP-Sec61 mice. A: qPCR analysis of metabolically relevant genes from RNA isolated from livers of mice of the indicated genotype (n = 4 mice per genotype). B: Western blot analysis of Ppar ␥ , C/EBP ␣ , and C/EBP ␤ proteins from liver tissue homogenates. Error bars represent SEM, and * indicates P р 0.05. Both our group and others have reported that the equivalent mutation in yeast leads to defective ERAD, which may lead to ER stress ( 13,14 ). Taken together, these observations suggest a defect in protein processing in Sec61a1 Y344H/Y344H cells, which may evoke a compensatory response that could contribute to the steatotic phenotype of mutant mice. We therefore surveyed expression of markers of the ER stress response and the heat shock response (HSR) in liver lysates from Sec61a1 Y344H/Y344H , RIP-Sec61 mice. We observed increases in markers of both the ER stress response and the HSR in Sec61a1 Y344H/Y344H , RIP-Sec61 mice compared with heterozygous controls. Consistent with our earlier fi nding of ER stress [i.e., dilated ER and upregulation of both Bip and Chop mRNA ( 10 )], we saw increased levels of spliced Xbp1 mRNA, Eif2 ␣ phosphorylated at serine 52, and Sec24d and Bip protein ( Fig. 4A, B ). In addition to elevation of ER stress markers, Sec61a1 Y344H/Y344H , RIP-Sec61 mice exhibited increased protein levels of key cytosolic chaperones, Hsp90 and Hsp70, and Hsf1, the transcription factor responsible for orchestrating the HSR, compared with controls ( Fig. 4C ).

Inhibition of Hsp90 activity reduces Ppar ␥ signaling and expression
Because Sec61a1 Y344H/Y344H , RIP-Sec61 mice exhibited signs of increased ER stress and an increased HSR along with elevations in Ppar ␥ , we asked if perturbation of either pathway could affect Ppar ␥ signaling. As an indicator of Ppar ␥ signaling, we used the F442A-Ap2 reporter cell line, which carries an integrated luciferase gene under control of the Ap2 promoter, which is a Ppar ␥ target gene ( 19 ). To induce the ER stress response, we used the small molecules thapsigargin and tunicamycin ( 20,21 ); to induce the HSR, we used the small molecules celastrol, which inhibits cytosolic chaperone activity, and 17-DMAG, which acid ( Fig. 3B ) in Sec61a1 Y344H/Y344H , RIP-Sec61 mice, in agreement with reduced levels of Srebf1 , Acc2 , Dgat2 , and Scd1 mRNA. Using 14 C-labeled oleic acid to monitor lipid uptake, we observed an ‫ف‬ 2-fold increase in fatty acid uptake ( Fig. 3C ), which agrees with elevated Cd36 mRNA. These data indicate that steatosis, despite reduced levels of de novo lipogenesis, in Sec61a1 Y344H/Y344H , RIP-Sec61 mice is largely a result of increased uptake and storage of fatty acid.

Sec61a1 Y344H/Y344H
, RIP-Sec61 mice have normal adipose tissue Hepatic steatosis as a result of elevated Pparg expression and increased uptake of fatty acid is reminiscent of the phenotype seen in lipoatrophic mice, in which defects in adipose tissue lead to an accumulation of lipid in the liver ( 17,18 ). Therefore, we examined the adipose tissue of Sec61a1 Y344H/Y344H , RIP-Sec61 mice for defects. In contrast to lipoatrophic mice, histology of white adipose tissue depots appeared normal (supplementary Fig. IIIA), as did fat mass generally (supplementary Fig. IIIB) and plasma adiponectin levels (supplementary Fig. IIIC). We even observed an increase in Cebpa and Pparg mRNA levels by qPCR in adipose tissue from Sec61a1 Y344H/Y344H , RIP-Sec61 mice (supplementary Fig. IIID). Furthermore, we observed no defect in adipocyte differentiation of mouse embryonic fi broblasts from Sec61a1 mutant mice when compared with wild-type or heterozygous mice (supplementary Fig. IIIE).

Activation of ER stress and the heat shock responses in Sec61a1 Y344H/Y344H , RIP-Sec61 mice
The Sec61 ␣ protein is involved in both translocation and the initial stages of protein processing in the ER ( 11 ). Schauble et al. ( 12 ) have reported excessive calcium leakage and a defective interaction between Bip and the Sec61 ␣ Y344H mutant, which may impede protein translocation.

Fig. 3. Sec61a1
Y344H/Y344H , RIP-Sec61 mice display increased uptake of free fatty acid. A: qPCR analysis on RNA isolated from hepatocytes from Sec61a1 +/Y344H , RIP-Sec61 or Sec61a1 Y344H/Y344H , RIP-Sec61 mice (n = 4 per genotype). Error bars represent SEM, and * indicates P р 0.05. B: Primary hepatocytes were isolated from age-and sex-matched, fasted mice. Rates of sterol and triglyceride synthesis were determined by labeling with 14 C-acetate for 16 h. Triplicate assays were performed from each of four ( Sec61a1 +/Y344H , RIP-Sec61) or three ( Sec61a1 , RIP-Sec61) mice. C: Primary hepatocytes were isolated as in B. Rate of fatty acid uptake was determined in triplicate after cells were incubated with 14 C-labeled oleate for 5 min, washed, and lysed. Cell-associated radioactivity was read by a scintillation counter (n = 2 per genotype).

Fig. 5.
Hsp90 inhibition reduces Ppar ␥ signaling. A: F442A cells stably transfected with luciferase under the control of the AP2 promoter (F442A-Ap2Luc cells) were treated with rosiglitazone, tunicamycin, or thapsigargin at the indicated concentrations for 16 h, followed by lysis and homogeneous assay for luciferase activity. B: Cells were treated and assayed as in A, except that the HSR inducers celastrol and 17-DMAG were used. Error bars represent SD, and * indicates P р 0.05 by Student's t -test.

Fig. 4. Increased ER stress and HSR in Sec61a1
Y344H/Y344H , RIP-Sec61 mice. A: Western blot analysis for markers of ER stress activation of liver extracts from mice of the indicated genotype. B: qPCR analysis of spliced Xbp1 message on RNA isolated from livers of the indicated genotype (n = 4 per genotype). Error bars represent SEM, and * indicates P р 0.05 by Student's t -test. C: Western blot analysis of Hsf1, Hsp70, and Hsp90 from liver homogenates prepared from mice of the indicated genotype.
inhibits Hsp90 ( 22,23 ). Treatment with either thapsigargin or tunicamycin resulted in no detectable alteration in Ppar ␥ signaling ( Fig. 5A ). Likewise, treatment with celastrol did not modify Ppar ␥ signaling via the Ap2 promoter. However, treatment of cells with 17-DMAG did lead to a reduction in Ap2 luciferase activity ( Fig. 5B ), indicating that Hsp90 activity is necessary for Ppar ␥ signaling.
These experiments suggest a mechanism whereby Hsp90 interacts with and stabilizes the Ppar ␥ protein. To test this hypothesis, we asked if Ppar ␥ and Hsp90 physically interact. Endogenous Ppar ␥ coimmunoprecipitated with endogenous Hsp90 from livers of chow-fed C57Bl/6 mice ( Fig. 6C ), as has been shown previously in 3T3-L1 adipocytes ( 25 ). To determine whether Ppar ␥ ligand modifi es this interaction, we transfected HEK293 cells with CMV:FLAG-Ppar ␥ . Here also anti-FLAG antibody coimmunoprecipitated Ppar ␥ and endogenous Hsp90 (supplementary Fig. IVB), and treatment of cells with rosiglitazone increased the amount of Hsp90 coimmunoprecipitated with Ppar ␥ . Finally, to see if Hsp90 had a functional effect on Ppar ␥ signaling, we used a PPRE-luciferase reporter assay. Hsp90 overexpression had a positive effect on basal Ppar ␥ signaling with a PPRE luciferase reporter in HEK293 cells, though this effect disappeared upon treatment with rosiglitazone ( Fig. 6D ).

Hsp90 inhibition blocks Ppar ␥ signaling in primary hepatocytes
The previous experiments indicate that Hsp90 enhances Ppar ␥ signaling. To place this in physiological context with respect to the development of NAFLD, we performed experiments in primary hepatocytes to see if Hsp90 inhibition could block ligand-induced expression of Ppar ␥ target genes. In hepatocytes from unaffected mice, 17-DMAG completely blocked the upregulation of Cd36 , Cidec , and Lpl mRNA in response to rosiglitazone, indicating that Hsp90 activity is required for the transcriptional activity of Ppar ␥ ( Fig. 7A ).
We also performed this experiment using hepatocytes from Sec61a1 Y344H/Y344H , RIP-Sec61 mice, to see if 17-DMAG treatment could lower the already elevated mRNA levels of these same Ppar ␥ target genes. In this case, treatment of primary hepatocytes lowered levels of Cd36 , Cidec , and Lpl ( Fig. 7B ). This was concomitant with reduced levels of Ppar ␥ protein ( Fig. 7C ) To determine whether Hsp90 overexpression has an effect on lipid storage in hepatocyte cell lines, we examined Huh7 cells treated with oleate, which promotes lipid accumulation ( 26 ), and stained with the fl uorescent dye AdipoRed. Oleate increased lipid accumulation in Ppar ␥transfected cells. Additional transfection of Hsp90 also increased lipid accumulation, with little additional effect of oleate ( Fig. 7D ). This effect required Hsp90 activity as treatment with 17-DMAG reversed lipid accumulation.

DISCUSSION
While NAFLD is a signifi cant public health prob lem, little is known of its etiology at the cellular level. Sec61a1 Y344H/Y344H , RIP-Sec61 mice develop steatosis in the absence of high-fat diet, obesity, hyperinsulinemia, or chemical induction, and we used them here to identify metabolic pathways relevant to the development of NAFLD. Fatty acid uptake was higher in primary hepatocytes from Sec61a1 Y344H/Y344H , RIP-Sec61 mice compared with nonsteatotic heterozygous mice. This increase in uptake correlated with an increase, at the mRNA and protein level, of Pparg1 and 2 and mRNA expression of their target genes Cd36 , Lpl , and Cidec , which encode proteins involved in the uptake and storage of fatty acids. Conversely, fatty acid , RIP-Sec61 mice compared with hepatocytes from heterozygous mice. This correlated with lower levels of Srebpf1 and its targets in livers of homozygous mutant mice compared with heterozygous controls.
Ppar ␥ plays a well-established role in liver lipid accumulation, as loss of Ppar ␥ function prevents steatosis in genetic or diet-induced obesity ( 27 ), and its activation leads to lipid accumulation in both adipocytes and liver ( 16,17 ). Indeed, overexpression of Cd36 or CideC, both Ppar ␥ targets, is suffi cient to increase hepatic lipid accretion ( 5,28 ). In this way, the Sec61a1 mutant mouse emulates steatosis as seen in both rodent models and in humans, which are marked by increased levels of Ppar ␥ and its targets ( 29,30 ).
To investigate activation of Ppar ␥ , we looked at stress pathways likely to be activated in Sec61a1 , RIP-Sec61 mice, ER stress and heat shock. We have previously documented elevated ER stress in livers of Sec61a1 Y344H/Y344H mice. Oyadomari et al. ( 31 ) have shown that ER stress in the liver can lead to greater translation of C/EBP ␣ / ␤ protein, which could in turn activate Pparg expression. However, we saw no increase in C/EBP ␣ / ␤ in the liver of Sec61a1 Y344H/Y344H , RIP-Sec61 mice and sought other explanations for Ppar ␥ activation. In addition to elevation of markers of the ER stress response, we also saw an increase in the HSR. Hsp70, Hsp90, and Hsf1 protein levels were elevated in the livers of Sec61a1 , RIP-Sec61 mice. Furthermore, inhibition of Hsp90 activity with 17-DMAG reduced Ppar ␥ signaling and protein levels. Hsp90 overexpression, on the other hand, increased both Ppar ␥ protein levels and signaling. 17-DMAG also had the effect of blocking the rosiglitazone-induced expression of Ppar ␥ target genes and lowering the levels of those same target genes in primary hepatocyte culture from heterozygous and Sec61a1 Y344H/Y344H , RIP-Sec61 mice, respectively. Finally, anti-Ppar ␥ immunoprecipitated Hsp90 in vivo, suggesting that a physical interaction underlies Hsp90's effects on Ppar ␥ protein abundance and signaling.
Recently, two different labs have found that Hsp90 blockade reduces Ppar ␥ signaling and inhibits adipocyte differentiation in vitro, which requires Ppar ␥ . Nguyen et al. ( 25 ) have shown that Hsp90 chaperones Ppar ␥ , while Desarzens et al. ( 32 ) have shown that Hsp90 inhibition blocks adipocyte lipid accumulation in vivo. Indeed, Hsp90 interacts with and chaperones many different transcription factors ( 33 ), most notably the nuclear hormone receptors, providing precedent for this observation ( 34 ). Notably, mice that lack Hsf1, which is the major transcription factor that orchestrates the HSR, display reduced levels of Hsp90 transcript and Ppar ␥ protein and are resistant to steatosis associated with HCC ( 35 ).
We are unaware of any human genetic variant of SEC61A1 associated with NAFLD, but this is not surprising given the essential role that SEC61 plays in protein translocation and processing, and the pleiotropic effects of the mouse mutation. Our investigations, however, uncover a role for the HSR in the development of NAFLD, specifi cally that an interaction between Hsp90 and Ppar ␥ positively regulates levels of Ppar ␥ and thus leads to increased lipid accretion in the liver. This link between NAFLD and Hsp90 is supported by proteomic analysis that indicates that both Hsp90 isoforms are elevated in patients with NAFLD ( 36 ).
In conclusion, our studies indicate a role for the evolutionarily conserved HSR in the development of NAFLD. By interacting with a key transcription factor involved in lipid storage, Ppar ␥ , Hsp90 facilitates Ppar ␥ signaling in the liver. This is not the only pathological state that implicates Hsp90, as the HSR and Hsp90 play a role in oncogenic transformation, possibly linking the initiation of steatosis with downstream effects such as the development of HCC.