Mimetics of Caloric Restriction Include Agonists of Lipid-activated Nuclear Receptors* □ S

The obesity epidemic in industrialized countries is associated with increases in cardiovascular disease (CVD) and certain types of cancer. In animal models, TM , SAS Institute). The survival data was analyzed using the Kaplan-Meier test for significance.

The obesity epidemic in industrialized countries is associated with increases in cardiovascular disease (CVD) and certain types of cancer. In animal models, caloric restriction (CR) suppresses these diseases as well as chemical-induced tissue damage. These beneficial effects of CR overlap with those altered by agonists of nuclear receptors (NR) under control of the fastingresponsive transcriptional co-activator, peroxisome proliferator-activated co-activator 1␣ (PGC-1␣). In a screen for compounds that mimic CR effects in the liver, we found statistically significant overlaps between the CR transcript profile in wild-type mice and the profiles altered by agonists of lipid-activated NR, including peroxisome proliferator-activated receptor ␣ (PPAR␣), liver X receptor, and their obligate heterodimer partner, retinoid X receptor. The overlapping genes included those involved in CVD (lipid metabolism and inflammation) and cancer (cell fate). Based on this overlap, we hypothesized that some effects of CR are mediated by PPAR␣. As determined by transcript profiling, 19% of all gene expression changes in wild-type mice were dependent on PPAR␣, including Cyp4a10 and Cyp4a14, involved in fatty acid -oxidation, acute phase response genes, and epidermal growth factor receptor but not increases in PGC-1␣. CR protected the livers of wild-type mice from damage induced by thioacetamide, a liver toxicant and hepatocarcinogen. CR protection was lost in PPAR␣-null mice due to inadequate tissue repair. These results demonstrate that PPAR␣ mediates some of the effects of CR and indicate that a pharmacological approach to mimicking many of the beneficial effects of CR may be possible.
Obesity has reached epidemic proportions in most industrialized nations. More than 60% of U.S. adults are now overweight or obese, predisposing millions of Americans to chronic lifestyle diseases, including cardiovascular disease (CVD), 1 diabetes mellitus, and certain forms of cancer (1). Caloric restriction (CR) increases longevity in diverse species (2). In experimental models, CR decreases circulating levels of cholesterol, triglyceride, and glucose, increases insulin sensitivity in the liver and peripheral tissues, and decreases the incidence of diabetes and atherosclerosis as well as both spontaneous and environmentally induced cancer (3,4). Many of the transcriptional changes associated with aging can be reversed by CR as demonstrated by transcript profiling (5). However, the gene regulatory networks that underlie the beneficial effects of CR have not been identified.
A strategy for identifying drugs that mimic the beneficial effects of CR includes those that optimize glucose and lipid homeostasis and suppress inflammation (6). These responses are partly under control of the nuclear receptors PPAR␣, LXR␣/LXR␤, and RXR. PPAR␣ is activated by endogenous fatty acids (7), environmental chemicals, and clinically relevant hypolipidemic agents (8). PPAR␣ regulates genes involved in gluconeogenesis as well as fatty acid transport and ␤-oxidation (8). LXR␣ and LXR␤ are activated by naturally occurring oxysterols and regulate genes involved in reversed cholesterol transport, bile acid and triglyceride synthesis, glucose homeostasis, and inflammation (9,10). PPAR␣ and LXR heterodimerize with RXR family members, which themselves can be activated by natural and synthetic retinoid compounds. Activation of one or more of these receptors in mice and rats leads to responses that overlap with those of CR, including decreases in serum cholesterol and triglyceride levels, decreased incidence of atherosclerosis in models of heart disease, increased insulin sensitivity in diabetes models, and inhibition of cancer incidence (8,10,11). Given these phenotypic overlaps, beneficial effects of CR might be mediated in part by mechanisms that overlap with those regulated by these NRs.
In addition to a role in lipid metabolism, PPAR␣ also determines responses to diverse forms of physical-and chemicalinduced stress. When compared with PPAR␣-null mice, wildtype mice pre-exposed to PPAR␣ agonists exhibit decreased mortality, decreased cellular damage, and increased tissue repair after exposure to physical and chemical stressors in the liver (12)(13)(14)(15). The ability of PPAR␣ to regulate tissue repair genes may contribute to protection in the liver (12,16). Like activation of PPAR␣, CR protects tissues from chemical-induced toxicity (6). Protection is thought to occur partly through elimination of damaged cells by apoptosis (17) or by modulation of genes involved in xenobiotic metabolism (18). CR protected the livers of rats from permanent damage after exposure to the model hepatotoxicant thioacetamide through a timely and robust compensatory tissue repair response (19). The mechanism by which CR protects the liver from damage is not known.
Mammals have evolved highly regulated systems to maintain blood glucose levels within tight limits, even during times of food deprivation. Blood glucose levels are controlled by peripheral glucose uptake as well as the hormonal modulation of glucose production. Gluconeogenesis occurs during fasting in which insulin levels are low and in diabetic states in which the liver is insulin-resistant. PGC-1␣, a transcriptional co-activator is a central regulator of glucose levels and the response to fasting. PGC-1␣ coordinately regulates genes involved in gluconeogenesis and fatty acid ␤-oxidation in the heart and liver during fasting (20). The PGC-1␣ gene is induced at the transcriptional level by the cAMP response element-binding protein, CREB, a transcription factor activated by glucagon and cAMP under conditions of low glucose. PGC-1␣ mediates effects on metabolic pathways through both ligand-dependent and -independent activation of NR expressed in the liver, including hepatocyte nuclear factor-4␣, glucocorticoid receptor, and PPAR␣ (20). In the heart, PGC-1␣ and PPAR␣ were identified as downstream effectors of p38 kinase-dependent stress-activated signaling, indicating a link between extracellular stressors and alterations in energy metabolic gene expression (21). PGC-1␣ may also play a role in the regulation of other NRs, including LXR (22), farnesoid X receptor (23), and constitutive androstane receptor (24), all of which exhibit increased expression and/or activity during fasting.
In an effort to identify compounds that mimic the CR tran-scriptional response, we compared the CR transcript profiles with the transcript profiles altered by agonists of NR under control of PGC-1␣. In addition, we posed the hypothesis that many of the beneficial effects of CR are mediated by the NR under control of fasting-induced PGC-1␣. As a first step to test this hypothesis, we determined the role of PPAR␣ in CR transcriptional responses as well as CR protection from chemicalinduced damage in the liver.

EXPERIMENTAL PROCEDURES
Animals-Wild-type and PPAR␣-null mice on a SV129 background were used in these studies. Control and treated mice were provided with NIH-07 rodent chow (Zeigler Brothers, Gardeners, PA) and deionized filtered water ad libitum. Lighting was on a 12-h light/dark cycle. Mice were fed 90% of calculated AL levels for 1 week followed by 65% of AL for 4 weeks as described previously (19). Under these conditions, we see the typical effects of CR in wild-type mice, including decreases in mean body weights (wild-type AL, 32.3 Ϯ 1.4 g; wild-type CR, 21.6 Ϯ 1.9 g) and decreases in serum glucose and triglyceride levels. PPAR␣null mice exhibited similar changes in body weight (PPAR␣-null AL, 31.1 Ϯ 3 g; PPAR␣-null CR, 21.9 Ϯ 1.2 g) and decreases in glucose compared with wild-type mice (see below). In addition, our CR protocol increased apoptosis and decreased proliferation of hepatocytes in WT mice (data not shown), as observed earlier (17). In separate studies carried out at CIIT Centers for Health Research, mice were given by gavage each day for 3 days the RXR agonist AGN 194,204 (Allergan, Irvine, CA) at 3 mg/kg/day or the PPAR␣ agonist WY-14,643 (ChemSyn Science, Lenexa, KS) at 50 mg/kg/day. Mice were also fed a control diet or a diet containing WY in the diet (0.05%) for 1 week followed by a 40-min heat stress at 42°C or heat stress alone. Mice were sacrificed 4 h after the heat stress. Experiments with T0901317 (25), D3T (26), and DEHP and CR (27) have been described. In the thioacetamide studies, female mice were fed Harlen Teklad rat chow (No. 7029, Madison, WI) AL or 65% of AL for 21 days. On day 22, all four groups were treated with a lethal dose of TA (1000 mg/kg, intraperitoneally in saline) and observed for signs of toxicity and mortality twice a day for 14 days. For time-course experiments, separate groups of mice were treated with TA (1000 mg/kg, intraperitoneally) and killed at 0, 12, 24, 36, 48, and 72 h after TA administration. Animals were sacrificed under diethyl ether anesthesia. Blood was collected for analysis of plasma glucose and triglyceride levels. Portions of the livers were snap frozen in liquid nitrogen and stored at Ϫ70°C until analysis. Slices of liver were fixed in 10% neutral buffered formalin for 48 h, transferred to 70% ethanol, and embedded in paraffin; 5-m sections were cut and mounted on slides and stained with hematoxylin and eosin. Hematoxylin and eosin-stained liver sections were examined by light microscopy. All animal studies were conducted under the federal guidelines for the use and care of laboratory animals and were approved by Institutional Animal Care and Use Committees.
RNA Isolation and Transcript Profiling-For the analysis of livers, three mice were analyzed from each treatment group. Liver RNA was isolated using a modified guanidium isothiocyanate method (TRIzol®, Invitrogen) and was further purified using silica membrane spin columns (RNeasy®, Qiagen, Valencia, CA). RNA integrity was assessed by ethidium bromide staining followed by resolution on denaturing agarose gels and also by the RNA 6000 LabChip® kit using a 2100 Bioanalyzer (Agilent Technologies, Palo Alto, CA). For each sample, 15 g of biotin-labeled cRNA was generated from 10 g of total RNA and hybridized to GeneChip® Test3 arrays (Affymetrix, Inc., Santa Clara, CA) to determine the quality of the RNA. Subsequently, the same samples were hybridized to Murine GeneChip® U74Av2 GeneChips (Affymetrix). All procedures were carried out according to the manufacturer's recommendations using the antibody amplification technique. Images were processed using the Microarray Suite 5.0 software (Affymetrix). Hybridization quality was assessed by visual inspection of the image and from a report generated by Microarray Suite 5.0. Criteria for an acceptable hybridization were the following: background Ͻ 100, noise (RawQ) Ͻ 5, 3Ј/5Ј ratio for select housekeeping genes Ͻ 4. Hybridizations not meeting these criteria were repeated, beginning at the target preparation step. The data were analyzed and statistically filtered using Rosetta Resolver® version 3.0 software (Rosetta Inpharmatics, Kirkland, WA). The threshold for significance was set at p Յ 0.001 and Ն1.5-fold or Յ1.5-fold, and the relative increase or decrease in mRNA abundance for each gene was reported as a -fold change relative to control. Genes altered by CR in wild-type mice were further filtered using a Bonferroni correction of the p value (28). Because CR responses are dependent on the genetic background of the mouse strain (29), the significant genes were compared with those genes altered by CR in an independent study in C57Bl/6 mice (27) to identify common transcriptional targets of CR. Only those genes that were altered by CR in an identical direction in both studies were examined further. Genes were grouped into functional classes using KEGG (www.kegg.org) and Gene Ontology (www.geneontology.org). Identification of expressed sequence tags was facilitated by euGenes (iubio.bio.indiana.edu:89/ mouse/). The CR, heat stress, WY, and AGN profiles were compared with those genes reported as statistically changed (Ն1.5-fold or Յ1.5fold) from the studies of Stulnig et al. (25), Kwak et al. (26), and Wong and Gill (27) for T1317, D3T, and DEHP, respectively. The Spearman rank correlation test was performed using SAS, version 6.12 (SAS Institute, Research Triangle Park, NC). A complete list of the 87 genes altered by CR, WY, T1317, AGN, and D3T in wild-type and nullizygous mice, including unique identifiers, is provided in the Supplementary Materials as Table S1.
Reverse Transcriptase-PCR Analysis of Gene Expression-Real-time reverse transcriptase-PCR was performed as follows. After DNase treatment, total RNA was quantified (Ribogreen®, Molecular Probes, Inc., Eugene, OR) and diluted with water. 50 ng of RNA and PCR reagents was aliquoted into 96-well plates using an ABI Prism TM 6700 Automated Nucleic Acid Workstation (Applied Biosystems, Foster City, CA) and subjected to real-time quantitative PCR (TaqMan®, Applied Biosystems) using gene-specific primers (Table S2) and fluorescently labeled probes (Molecular Probes) designed by the Primer Express® software (Applied Biosystems). Amplification curves were generated using the ABI Prism TM 7900HT Sequence Detection System (Applied Biosystems). Expression relative to vehicle control animals was determined after normalizing to the ribosomal 18S gene. There were three animals per treatment group, and each sample was analyzed in duplicate. Variability is expressed as standard error of the mean.
Western Blot-Procedures for preparation of liver lysates, Western blotting, and description of antibodies used have been described before (30).
Estimation of Liver Injury-Plasma was separated by centrifugation and estimated for alanine aminotransferase activity (EC 2.6.1.2.), a marker of liver injury, by using Sigma kit 59UV (ALT) (Sigma). Liver injury assessed by ALT was corroborated by histopathology using liver sections stained with hematoxylin and eosin.
Estimation of Tissue Repair-Liver cell division was determined by proliferating cell nuclear antigen (PCNA) staining on paraffin-embedded, 5-m-thick liver sections as previously described (31). Cells in various phases of the cell cycle were identified as follows: G o , cells with no staining; G 1 , cells with light brown nuclear staining; S, cells with deep brown nuclear staining; G 2 , cells with diffuse speckled nuclear staining and brown cytoplasmic staining; and M, cells with diffuse cytoplasmic staining with deep blue chromosomal staining (32). Three slides per time point per group were observed under light microscopy. In all, 1000 cells were counted per animal per time point, and cells in S-phase of cell division were recorded. The percentage of cells in Sphase at each time point was calculated and used as an index of cell division.
Statistical Analysis of Data-Overall statistical significance of gene expression by TaqMan was determined by two-way analysis of covariance using Proc.MIXED with ␣ ϭ 0.05 (SAS Institute, Research Triangle Park, NC). Student's t test was used to determine if statistically significant differences existed between individual groups. The t-tests for the four individual comparisons per gene were not adjusted for multiplicity. Other statistical tests of significance were done by analysis of variance post-hoc testing performed using the Tukey-Kramer test with a p value of Ͻ0.05 (JMP TM , SAS Institute). The survival data was analyzed using the Kaplan-Meier test for significance.

Identification of PPAR␣-dependent Genes Regulated by CR
in Mouse Liver-To identify genes regulated by CR in wildtype mice and determine the role of PPAR␣ in their regulation, transcript profiles were generated in the livers of AL and CR wild-type and PPAR␣-null mice using Affymetrix mouse U74Av2 GeneChips® containing ϳ12,500 genes. We identified 87 genes altered by CR in wild-type mice (Ն1.5-fold or Յ-1.5-fold) as described under "Experimental Procedures." This number decreased to 78 as 9 of the genes were represented twice. In all cases, the expression behavior of the two expressed sequence tags was very similar (see below). The CR transcript profile from WT mice was dominated by downregulated genes (85%). Fifteen of the 78 CR genes (19%) were dependent on PPAR␣ for CR-mediated alteration, because the changes occurred only in wild-type mice (Fig. 1). To ensure the analysis resulted in a low false discovery rate, the expression of a number of genes was analyzed using TaqMan (Table I). We initially examined the expression of five genes that were altered in the transcript profiles in the same direction in the SV129 and C57Bl/6 strains. Changes in the expression of all of these genes were verified as to direction and magnitude in both the wild-type and PPAR␣-null mice indicating our methods of analysis were reliable. We next examined the expression of additional genes that fell into a number of functional categories of interest (described below). These genes were identified as altered in the wild-type SV129 mice by CR (p Յ 0.001), but they either did not pass the fairly strict statistical criterion determined by a p Յ 0.001 with a Bonferroni adjustment or they were not coordinately regulated in the two strains. Out of 22 comparisons (11 genes in wild-type and PPAR␣-null mice), 17 were confirmed by Taq-Man (Table I). These results indicate that 1) our criteria for selection of the 78 CR-regulated genes is stringent with a low false discovery rate (Ͻ0.08), and 2) the less stringent methods for gene selection that we used in our initial analysis and that are more commonly used in the field (i.e. p Յ 0.001) led to a higher false discovery rate indicating a greater need for confirmation of altered expression using other techniques.
Recent studies have identified genes either regulated by CR or by fasting in mouse livers. We compared the genes regulated in our studies with those recently identified as CR-responsive by Dhahbi et al. (33). Comparison was facilitated by the fact that identical DNA chips were used in the analyses. Out of the 18 overlapping genes that were CR-responsive in the two studies, 17 genes were regulated in an identical manner, despite the fact that there were differences in study design, including sex, genotype, and length of CR (male, SV129, and 5 weeks CR (our study) versus female, B6C3F1, and 2-8 weeks CR (33)). Only one gene was regulated in an opposite manner in the two studies (Orm1). Most of the common genes were changed early after the initiation of CR (within 4 weeks), although 7 of the genes were classified as "late genes" and altered only after 8 weeks of CR.
Bauer et al. (34) identified genes regulated by a 24-or 48-h fast in the livers of male SV129 mice. Out of the 17 overlapping genes that were responsive in the two studies, 16 genes were regulated in the same direction. Interestingly, many of these genes (7 out of 17) are involved in cholesterol synthesis and were also down-regulated by CR in the Dhahbi et al. study. The overlap in CR and fasting profiles is not surprising, given the fact that CR mice were sacrificed after an 18-h (our study) or 24-h (33) fast. These overlaps indicate that 1) we have identified genes regulated by CR independent of the mouse strain used and 2) many of the genes regulated by CR are also regulated by fasting.
Overlap in the Transcript Profiles of CR and Agonists of Lipid-regulated Nuclear Receptors-We examined overlaps between the transcript profiles in wild-type CR mice and mice treated with agonists that activate NR under control of PGC-1␣. Under conditions of CR used in this study, we observed 5-to 6-fold increases in PGC-1␣ mRNA in the livers of CR wild-type and PPAR␣-null mice by TaqMan (Table I). The CR profiles were compared with 21 individual transcript profiles generated from the livers of mice treated with experimental drugs, chemicals, or heat stress. Using hierarchical clustering (35), the CR profiles were most closely related to the profiles from wild-type mice treated with the PPAR␣ agonist WY in the feed for 7 days with or without a subsequent heat stress (Fig. 1). There was also a close relationship between the CR profiles and the profiles altered in wild-type mice treated in the feed for 7 days by a LXR agonist (T1317), 3 daily gavage doses of a RXR panagonist (AGN), or 3 daily gavage doses of WY. Spearman rank correlation, used as another measure of similarity between groups showed significant similarities between the genes regulated by CR and WY, AGN, or T1317 in wild-type mice (Table  II). Similarity between the CR and compound profiles in PPAR␣-null mice was insignificant for WY and AGN, indicating that PPAR␣ is required to generate the CR-like profiles through effects on either receptor in the PPAR␣-RXR heterodimer. Likewise, the similarity between CR and T1317 also became insignificant in LXR␣/LXR␤-null mice, indicating one or both LXRs are required for T1317 to generate the CR-like profile.
As CR is thought to activate genes in the liver that decrease oxidative stress (36), we also compared the CR profiles to that of D3T, an anti-oxidant compound that enhances nuclear accumulation of the transcription factor Nrf2 by triggering the release of Nrf2 from its cytoplasmic tether Keap1 (26). An overlap in the profiles was observed, but was independent of Keap1/Nrf2 (Table II).

Regulation of Lipid Homeostasis Genes by CR-CR
and agonists of PPAR␣, RXR, and LXR prevent or delay CVD in disease models of susceptibility as well as in wild-type animals (10,11). Decreases in CVD may be partly through modulation of circulating levels of lipids. From the initial 78 CR-responsive genes, we identified genes involved in lipid metabolism. We compared the regulation of these genes by CR, WY, T1317, AGN, and D3T in wild-type mice and mice nullizygous for the receptor. In this comparison we included lipid metabolism genes that were not part of the initial 78 genes but were examined by TaqMan (Table I). CR increased the expression of genes involved in fatty acid mobilization (lipoprotein lipase (Lpl) and phospholipase A2 gamma type XII (Pla2g12)), fatty acid transport (Abcd2 and Abcd4), and fatty acid ␤-(acyl-CoA dehydrogenase very long chain; Acadvl) and -(Cyp4a10, Cyp4a14) oxidation ( Fig. 2A) indicating increased dependence on fatty acids as an energy source. These results are consistent with metabolic effects of fasting on fatty acid metabolism regulated by PGC-1␣ (20). Assessment of mRNA and protein expression confirmed the increased expression of fatty acid metabolism genes during CR, including PPAR␣-dependent increases in Cyp4a14 gene expression (Table I) and Cyp4a protein levels using a pan-specific antibody (data not shown). Most of the fatty acid metabolism genes were regulated by CR in both wild-type and PPAR␣-null mice indicating PPAR␣-independent mechanisms involved in regulating fatty acid metabolism. In contrast, all fatty acid  a Genes in bold are those identified using the more stringent criteria described under "Experimental Procedures." Other genes were identified as altered by CR in wild-type SV129 mice (p Յ 0.001).
b Numbers are expressed as ratios of changes in CR versus ad libitum (AL) fed in the same strain as determined by TaqMan. The numbers in bold are significantly different (p Յ 0.05) from their respective controls in the same strain. Underlined numbers are those that are significantly different (p Յ 0.05) in either the AL or CR groups between wild-type and PPAR␣-null mice. metabolism genes regulated by WY except Cyp4a10 and Cyp4a14 were PPAR␣-dependent. T1317, AGN, and D3T exposure led to altered regulation of a subset of these genes consistent with less pronounced effects of T1317 and AGN on fatty acid catabolism compared with PPAR␣ agonists (25,37).
In contrast to the up-regulation of fatty acid utilization genes, CR decreased the expression of genes involved in fatty acid, triglyceride, and glycerolipid synthesis. These genes in-cluded fatty acid elongase 1 (Fae1), fatty acid synthase (Fasn), fatty acid coenzyme A ligase, long chain 2 (Facl2), glycerol-3phosphate acyltransferase (Gpam), glycerol kinase (Gk), and NAD(P)-dependent steroid dehydrogenase-like (Nsdhl) involved in a number of functions, including glycerolipid metabolism and bile acid synthesis ( Fig. 2A). All of these changes were PPAR␣-independent. Consistent with the changes in triglyceride synthesis genes, CR decreased serum triglyceride levels in wild-type mice (Fig. S1A). CR decreased the abnormally high serum levels of triglycerides in ad libitum PPAR␣-null mice as well as the inherent hepatocellular steatosis observed in this strain (Fig. S1B). Decreases in glucose levels by CR were PPAR␣-independent (Fig. S1A).
Exposure to WY and, to a lesser extent, T1317 and AGN resulted in modest increases in the expression of fatty acid, triglyceride, and cholesterol synthesis genes. LXR agonists increase circulating triglyceride levels through increases in sterol regulatory element-binding protein, SREBP1c, an impediment to the clinical use of these compounds (10). In contrast, D3T exposure led to a coordinated decrease in the expression of fatty acid, triglyceride, and cholesterol synthesis genes similar to CR. Most of these changes by D3T were observed in both wild-type and nullizygous mice or in nullizygous mice alone indicating Keap1/Nrf2-independent mechanisms are responsible for these unexpected, potentially beneficial effects.
CR altered expression of genes involved in bile acid synthesis, including increases in Cyp7a1 and decreases in Cyp7b1 and Cyp8b1 (Fig. 2A), a pattern identical to that altered by T1317 through an LXR-dependent mechanism. The spectrum of CR changes then included those associated with beneficial LXRmediated effects on bile acid synthesis (10) in the absence of adverse effects on triglyceride synthesis. Cyp7a1 is induced by fasting through a mechanism that may involve PGC-1␣ activation of HNF4␣ or COUP-TFII (38,39) demonstrating multiple mechanisms of induction by fasting in addition to possible involvement of LXR. These results demonstrate that CR regulates a battery of lipid metabolism genes associated with beneficial effects on circulating lipid levels.
CR Alters the Expression of Risk Factors for CVD-Acute or chronic inflammation contributes to atherosclerosis (40). CR uniformly decreased the expression of risk factors for CVD, including the expression of acute phase response genes (Fig.  2B). These acute phase proteins (APPs) included Grp78 (Hspa5), transthyretin or prealbumin (Ttr), orosomucoid 1 (Orm1), and serum amyloid A4 (Saa4). Expression of Saa4 is also positively associated with increases in very low density lipoprotein and triglyceride levels (40). Four components of the complement cascade that are also APP were down-regulated by CR, including mannose binding lectin (Mbl1), complement component 4 binding protein (C4bp), complement component C9 (C9), and 1700013L23Rik (71% homologous to the human complement component C8 gamma chain precursor). Decreases in APP expression during CR may be due in part to the decrease in expression of leukemia inhibitory factor receptor (Lifr), which partially controls interleukin-6-dependent expression of the APP (41). CR also decreased the expression of elastase 1 (Ela1). Four out of 10 genes were completely

FIG. 2. Overlap in the expression of genes with functions in lipid metabolism (A) and associated with CVD (B) by CR and lipid-activated nuclear receptors.
Each lane represents the genes differentially regulated by CR or the compound compared with AL control mice. In A, genes shown in bold are those regulated by CR in two independent studies (see text). The other genes are those regulated by CR in SV129 mice and confirmed by TaqMan (Table I). Primary sequence names are shown. Full gene information, including Unigene identifiers, can be found in the Supplementary Materials. Numbers on the scale are in -fold changes. or partially regulated by CR through a PPAR␣-dependent mechanism. Remarkably, WY and T1317 (8 out of 10) were effective at mimicking the CR profile of these risk factors. AGN and D3T also down-regulated many of these genes. These results are consistent with the effects of PPAR␣ and LXR agonists on suppression of inflammation which contributes to CVD (9,10).
Genes Involved in Cell Growth-CR suppresses spontaneous and chemically induced cancers in part through suppression of proliferation and augmentation of apoptosis (4). Critical components of two growth regulatory pathways were downregulated by CR (Fig. 3). Epidermal growth factor receptor (Egfr) has been implicated in the genesis of animal and human tumors and is a target for anti-cancer drugs (42). The down-regulation of Egfr by CR was dependent on PPAR␣ (Table I). CR also down-regulated growth hormone receptor (Ghr). Mice with a targeted mutation of the Ghr gene or that carry spontaneous mutations in genes that determine growth hormone secretion (Pit1 and Prop1) exhibit increased longevity (43), possibly through increased resistance to different forms of stress (44). CR and either the Ghr (28) or Prop1 (45) mutation regulate an overlapping set of genes in the mouse liver. Importantly, growth hormone signaling determines both spontaneous and carcinogen-induced liver tumor frequency (46). CR down-regulated deiodinase type I (Dio1), which catalyzes the deiodination of the thyroxine prohormone (T4) to T3. Decreases in Dio1, decreases in Ttr, a carrier of T3 as well as increases in a thyroid hormone sulfotransferase (Sultn) (Fig.  S2) may underlie decreases in circulating thyroid hormone (T3) levels and metabolic rate during CR (47). Decreases in T3 may also be important in CR suppression of cell proliferation, given that hepatocyte proliferation is regulated by T3 (48).
Additional genes with links to cancer were responsive to CR. Tubulin subunits are targets of taxol and vinca alkaloids that disrupt microtubulin polymerization (49). The tubulin subunits Tuba2, Tuba6, Tubb2, Tubb3, and Tubb5 were uniformly down-regulated. Folylpolyglutamyl synthetase (Fpgs), also down-regulated by CR, converts anti-folate drugs used in chemotherapy into polyglutamate derivatives (50). Genes down-regulated by CR also included Bcl-6, a transcriptional regulator overexpressed in B-cell non-Hodgkin's lymphoma (51), S100a10 (also known as calpactin) overexpressed in stomach carcinomas (52) and inhibin ␤C (Inhbc) involved in gonadal cancers (53). One gene up-regulated by CR was RNA binding motif 3 (Rbm3), underexpressed in melanomas (54). We also observed down-regulation of two genes that play roles in both cancer and inflammation. Cxcl1 (also known as melanoma growth stimulatory activity) is a chemokine regulated by NF-B with roles in tumor formation and angiogenesis (55). 0610025L15Rik is 92% homologous to human HSCO (hepatoma subtracted clone 1), which binds to NF-B, inhibits apoptosis and is overexpressed in hepatocellular carcinomas (56). CR may also modulate chemical carcinogenesis by altering genes involved in DNA repair (57). CR up-regulated Rad51-like 1 (Rad51l1) involved in DNA double-strand break repair (Fig. 3).
Although WY, AGN, and D3T were poor at mimicking the expression of the genes associated with cell proliferation and DNA repair, T1317 exposure led to almost half the number of changes as CR in these genes (7 out of 16) in an LXR-dependent manner. This result was surprising given that LXR agonists are not generally considered as potential anti-cancer drugs.
Miscellaneous Genes Regulated by CR-Additional genes in a number of categories are described in the Supplemental Materials.
CR Requires PPAR␣ for Protection from Chemical-induced Hepatotoxicity-PPAR␣ activation, like CR, protects the liver from chemical-induced toxicity (14). To determine if PPAR␣ is involved in CR-mediated protection of the liver, WT and PPAR␣-null mice were fed AL or CR diets and challenged with a prototypical hepatotoxicant and nongenotoxic carcinogen, thioacetamide (TA). CR significantly protected 50% of the wildtype mice from a single lethal dose of TA up to 14 days (p ϭ 0.0014), whereas the CR PPAR␣-null mice and AL mice from both strains were not protected (p Ͼ 0.05) (Fig. 4A). Liver damage at 24 h was significantly less in the CR WT mice compared with the CR PPAR␣-null mice (Fig. 4B). At 48 h, however, there were no differences between groups. Compensatory cell division showed a significant increase in the number of cells in S-phase in CR wild-type but not in CR PPAR␣-null mice when observed as early as 24 h after TA challenge (Fig. 4, C and D). The number of hepatocytes in S-phase further increased, peaking at 72 h in CR wild-type mice. In comparison, PPAR␣-null mice did not exhibit significant cell proliferation until 72 h. This extensive and timely compensatory cell division in the wild-type CR mice contributed to restoration of the structure and function of the liver and likely contributed to the survival of the mice. DISCUSSION PGC-1␣, a global regulator of responses to food deprivation, regulates both ligand-dependent and -independent activation of many nuclear receptors that have known or putative roles in energy homeostasis and that are targets of drugs used to treat diabetes and cardiovascular disease. PPAR␣, a target of PGC-1␣, regulates responses to fasting, including fatty acid ␤and -oxidation, gluconeogenesis, and ketogenesis (58,59). In this study, we posed the hypothesis that PPAR␣ is required for some of the physiological responses to CR.
To test this hypothesis, we used transcript profiling to identify putative target genes of CR that are dependent on PPAR␣ and that may underlie the beneficial effects of CR in the mouse liver. Using a fairly strict set of criteria, we identified a modest number of genes altered by CR involved in lipid metabolism, inflammation, and cell growth consistent with effects of CR on these processes. We showed that 19% of all gene expression changes after CR were dependent on PPAR␣. The genes included two fatty acid -hydroxylases (Cyp4a10 and Cyp4a14) activated by xenobiotics, fasting, and diabetes in wild-type but not PPAR␣-null mice (60). Despite the fact that PPAR␣ agonists regulate genes involved in fatty acid utilization, genes involved in release of fatty acids from peripheral stores, transport, and ␤-oxidation were increased by CR but in a PPAR␣independent manner. This is in contrast to the PPAR␣-dependent increases in fatty acid metabolism genes during fasting (58,59). PPAR␣ was partially or completely required for CR to down-regulate acute phase genes responsive to inflammatory cytokines (C4bp, C9, Mbl1, Orm1, and Saa4), possibly through the ability of PPAR␣ agonists to suppress inflammatory re-sponses by sequestration of NF-B and AP-1 components (8). Lastly, PPAR␣ was required for CR to down-regulate Egfr, also down-regulated by PPAR␣ agonists in rat liver (61). Our work indicates that PPAR␣ is required for regulation of a subset of CR-responsive genes in the liver involved in fatty acid metabolism, inflammation, and cell growth.
PPAR␣ regulates responses to diverse forms of stress. We found that PPAR␣ was required for CR to protect the liver from damage induced by the hepatotoxicant, thioacetamide. The mechanism of CR protection likely includes enhanced repair of damage as compensatory cell proliferation was increased earlier and to a greater extent in wild-type CR mice compared with PPAR␣-null CR mice. We did not observe major differences in the amount of tissue damage between the strains at 48 h, the peak of liver damage. The molecular basis for the PPAR␣-dependent increased cell proliferation in CR mice is not known, but it is possible that PPAR␣ is required for energy production needed for tissue repair or that PPAR␣ is required for the optimal expression of repair genes (12,16). Given that PGC-1␣ is increased in expression with CR (our study), we hypothesize that PGC-1␣ through PPAR␣ is required for CR to protect the liver from chemical-induced stress. This hypothesis is supported by a number of studies. The livers of diabetic mice in which PGC-1␣ is induced (20) are protected from toxicity induced by acetaminophen, carbon tetrachloride, and bromobenzene due to robust tissue repair (15,16). Like CR, the protection observed in diabetic mice treated with acetaminophen required PPAR␣ (62). Taken together, these data indicate the importance of PGC-1␣-mediated regulation of PPAR␣ in CRinduced hepatoprotection.
Drugs that mimic CR responses include those with effects on glucose and lipid homeostasis and inflammation (6). Given the importance of PGC-1␣ in regulating responses to fasting, we compared the CR profiles generated in wild-type and PPAR␣null mice to the profiles altered by agonists of NR that are targets of PGC-1␣. The wild-type CR transcript profile significantly overlapped with those of agonists of NR regulated by PGC-1␣. The overlap required an intact PPAR␣ for WY and the RXR agonist AGN and intact LXR␣/LXR␤ for T1317, indicating PPAR␣ and LXR mediate the CR-like effects of these compounds. The fact that the CR transcriptional program significantly overlaps with those regulated by PPAR␣, LXR and possibly RXR is not surprising given the close functional relationships between these receptors. PPAR␣ activation not only leads to increases in LXR␣ expression (63) but in the levels of endogenous LXR activators (64). Treatment with a RXR agonist leads to activation of PPAR␣-regulated genes (37) presumably through activation of PPAR␣-RXR heterodimers. In addition, we also compared the CR profiles to the profile of D3T, an activator of the PGC-1␣-regulated gene, Nrf2. Although there was a significant overlap in the profile of CR and D3T, the overlap was Keap1/Nrf2-independent indicating other mechanisms may be important for D3T to mimic the CR response. The overlaps between CR and these compounds were observed despite the fact that the profiles were generated in four laboratories using mice with different genetic backgrounds housed under different conditions. Our results indicate that additional comparisons, including those across array platforms and between species, as recently reported for genes involved in aging (65), may yield further insights as to drugs, physiological processes, or genetic changes that mimic the beneficial effects of CR.
In conclusion, our results not only identify novel transcriptional targets of CR but support the contention (6) that drugs can be developed to mimic the beneficial effects of CR. Drugs that activate one or multiple receptors of the PPAR␣-RXR-LXR axis are rational choices for CR mimetics.