Defect in peroxisome proliferator-activated receptor alpha-inducible fatty acid oxidation determines the severity of hepatic steatosis in response to fasting.

Fasting causes lipolysis in adipose tissue leading to the release of large quantities of free fatty acids into circulation that reach the liver where they are metabolized to generate ketone bodies to serve as fuels for other tissues. Since fatty acid-metabolizing enzymes in the liver are transcriptionally regulated by peroxisome proliferator-activated receptor alpha (PPARalpha), we investigated the role of PPARalpha in the induction of these enzymes in response to fasting and their relationship to the development of hepatic steatosis in mice deficient in PPARalpha (PPARalpha(-/-)), peroxisomal fatty acyl-CoA oxidase (AOX(-/-)), and in both PPARalpha and AOX (double knock-out (DKO)). Fasting for 48-72 h caused profound impairment of fatty acid oxidation in both PPARalpha(-/-) and DKO mice, and DKO mice revealed a greater degree of hepatic steatosis when compared with PPARalpha(-/-) mice. The absence of PPARalpha in both PPARalpha(-/-) and DKO mice impairs the induction of mitochondrial beta-oxidation in liver following fasting which contributes to hypoketonemia and hepatic steatosis. Pronounced steatosis in DKO mouse livers is due to the added deficiency of peroxisomal beta-oxidation system in these animals due to the absence of AOX. In mice deficient in AOX alone, the sustained hyperactivation of PPARalpha and up-regulation of mitochondrial beta-oxidation and microsomal omega-oxidation systems as well as the regenerative nature of a majority of hepatocytes containing numerous spontaneously proliferated peroxisomes, which appear refractory to store triglycerides, blunt the steatotic response to fasting. Starvation for 72 h caused a decrease in PPARalpha hepatic mRNA levels in wild type mice, with no perceptible compensatory increases in PPARgamma and PPARdelta mRNA levels. PPARgamma and PPARdelta hepatic mRNA levels were lower in fed PPARalpha(-/-) and DKO mice when compared with wild type mice, and fasting caused a slight increase only in PPARgamma levels and a decrease in PPARdelta levels. Fasting did not change the PPAR isoform levels in AOX(-/-) mouse liver. These observations point to the critical importance of PPARalpha in the transcriptional regulatory responses to fasting and in determining the severity of hepatic steatosis.

Higher animals, under fed conditions, preferentially burn carbohydrate to generate ATP, and surplus carbohydrate is converted into fatty acids, which are then stored as triacylglycerols (TG) 1 in adipose tissue. When glucose availability is low during periods of starvation, the TG stored in adipose tissue are hydrolyzed to free fatty acids (FFA) and mobilized into plasma to reach liver where they play a major role in energy production (1)(2)(3). In liver, the influxed fatty acids are oxidized predominantly by the mitochondrial ␤-oxidation system and to a lesser extent by the peroxisomal ␤-oxidation, as well as by CYP4A-catalyzed microsomal -oxidation pathways (4 -6). Partial oxidation of fatty acids by mitochondrial and peroxisomal ␤-oxidation systems in liver leads to the production of acetyl coenzyme A (acetyl CoA), which then condenses with itself to form ketone bodies. Ketone bodies generated in liver are exported out of the liver to serve as fuels for other tissues such as the skeletal and cardiac muscle and brain during starvation. Thus, alternate use of carbohydrate and fatty acids to produce ATP is well regulated, and this regulatory energy consumption is referred to as "glucose fatty acid cycle" that requires the maintenance of efficient hepatic fatty acid oxidation. It is estimated that in the adult, FFA and their ketone body derivatives provide ϳ80% of caloric requirements after 24 h of fasting (7). Fasting causes a more dramatic depletion of carbohydrate energy source in infants and children, and as a consequence they exhibit greater dependence on efficient FFAdependent ketogenesis during starvation, underscoring the importance of fatty acid oxidation in energy metabolism (8). At present, several genetically determined metabolic defects at the individual enzyme level in mitochondrial and peroxisomal fatty acid ␤-oxidation pathways have been identified, and individuals with these defects remain essentially asymptomatic as far as basal energy metabolism is concerned under normal feeding conditions (8). However, during conditions that lead to short term fasting, they manifest severe hypoketotic hypogly-cemia, increased plasma FFA, variable degree of hepatic steatosis, and sudden death in early life because of their inability to oxidize FFAs in liver due to enzymatic defects (8). These metabolic diseases have provided valuable insights pertaining to the role of individual enzymes in fatty acid oxidation and energy utilization.
It is now well recognized that fasting causes a rapid transcriptional activation of genes encoding mitochondrial, peroxisomal, and microsomal fatty acid oxidation in liver in healthy individuals (9,10). These observations point to the importance also of regulatory step(s) controlling the levels of inducible fatty acid oxidation enzymes. Any defect in the inducibility of these enzymes can also impact on the energy metabolism and degree of hepatic steatosis in response to fasting similar to those encountered with metabolic defects at the enzymatic level. Fatty acid oxidation occurs in mitochondria, peroxisomes, and microsomes, and some of the critical enzymes of these oxidation systems are transcriptionally controlled by peroxisome proliferator-activated receptor ␣ (PPAR␣), a member of the nuclear hormone receptor superfamily (11). PPARs, which derive the designation by virtue of their ability to mediate predictable pleiotropic effects in response to peroxisome proliferators (12), consist of three isotypes, namely PPAR␣, PPAR␦ (also called PPAR␤), and PPAR␥ which are products of separate genes (11,13,14). Peroxisome proliferators are structurally diverse agents which, when administered to rats and mice, induce not only a marked peroxisome proliferation and increase in the enzyme proteins of the peroxisomal fatty acid oxidation but also induce changes in carbohydrate and lipid metabolisms (4,12). The induction of mitochondrial, peroxisomal, and microsomal CYP4A genes involved in fatty acid oxidation requires the formation of PPAR␣ heterodimerization with retinoid X receptor, and this PPAR␣⅐retinoid X receptor complex binds to PPAR response element, a region consisting of a degenerate direct repeat of the canonical AGGTCA sequence separated by 1 base pair (DR1), present in the 5 Ј-flanking region of target genes (15). The generation of PPAR␣ Ϫ/Ϫ mice established that PPAR␣ is critical for peroxisome proliferation and the coordinate transcriptional activation of fatty acid oxidation enzymes in liver (16). Furthermore, PPAR␣ Ϫ/Ϫ mice have provided valuable information on the constitutive levels of expression of mitochondrial and peroxisomal fatty acid-metabolizing enzymes in liver (17) and the response of these mice to dietary overload as well as short term fasting (18 -20). Mice deficient in peroxisomal fatty acyl-CoA oxidase (AOX Ϫ/Ϫ ) exhibited sustained PPAR␣ hyperfunction presumably caused by accumulation of endogenous ligand(s) due to the impairment of the peroxisomal fatty acid oxidation pathway (21). Mice nullizygous for both PPAR␣ and AOX (PPAR␣ Ϫ/Ϫ AOX Ϫ/Ϫ double nulls (DKO)) have also served as valuable tools to explore the role of PPAR␣ and fatty acid oxidation in constitutive lipid metabolism and hepatic fatty liver phenotype under fed state (22). The availability of these genetically altered PPAR␣ Ϫ/Ϫ (16), PPAR␣ Ϫ/Ϫ AOX Ϫ/Ϫ (DKO) (22), and AOX Ϫ/Ϫ (21) mice provides an opportunity to examine the comparative responses to changes in energy metabolism imposed by fasting. We demonstrate the critical importance of PPAR␣-dependent induction of fatty acid oxidation in determining the degree of hepatic steatosis.

EXPERIMENTAL PROCEDURES
Animals-Wild type (C57BL/6J), AOX-null (AOX Ϫ/Ϫ ) (21), PPAR␣null (PPAR␣ Ϫ/Ϫ ) (16), and AOX Ϫ/Ϫ PPAR␣ Ϫ/Ϫ double knock-out (DKO) (22) mice were housed in a controlled environment with a 12-h light/ dark cycle with free access to water and standard laboratory chow as described (22). All experiments were performed using mice ranging in age from 16 to 20 weeks. Starvation was commenced by removing food at 8:00 a.m., and groups of mice were fasted up to 96 h. Control mice were fed ad libitum. After mice were anesthetized, blood was collected in heparinized tubes and centrifuged, and the plasma was frozen until use. Organs were removed and frozen in liquid nitrogen and stored at Ϫ80°C. All animal procedures used in this study were reviewed and preapproved by the Institutional Review Boards for Animal Research of the Northwestern University.
Morphological Studies-For light microscopy, pieces of liver were fixed in 10% neutral buffered formalin, embedded in paraffin, and 4-m-thick sections stained with hematoxylin and eosin. Frozen sections of formalin-fixed liver (5 m) were stained with Oil Red O and counterstained with Giemsa. For cell proliferation analysis, mice were given bromodeoxyuridine (0.5 mg/ml) in drinking water, and their livers were processed for immunohistochemical localization as described previously (23), using antibodies raised against bromodeoxyuridine (Becton Dickinson). Histological analysis and image processing were carried out using Leica DMRE microscope equipped with a Spot digital camera. Images were taken at ϫ 20 and 40 magnification and captured at 1315 ϫ 1033 pixels. Montages of images were prepared with the use of Photoshop 5.0 (Adobe, Mountain View, CA).
Determination of Metabolites-Plasma glucose (24), lactate (25), and 3-hydroxybutyrate (3-HB) (26) were determined by the cited procedures. Plasma FFA and TG were determined by the use of reagent kits (NEFA C-Test Wako and Triglyceride E-Test Wako, respectively, from Wako Pure Chemical Industries, Ltd. Osaka, Japan). Liver glycogen was determined by the use of coupling reactions of amyloglucosidase and glucose oxidase (27). Total carnitines were determined using carnitine acetyltransferase (CAT) (28).
Western Blot Analysis and Quantification of Proteins-Protein concentrations were determined using a protein assay kit (Bio-Rad) using bovine serum albumin as standard. Liver, kidney, and heart extracts were subjected to 10% SDS-polyacrylamide gel electrophoresis and transferred to nitrocellulose membranes. The membranes were incubated with the primary antibody (see Refs. 17 and 22 for the source of various primary antibodies used in this study) followed by alkaline phosphatase-conjugated goat anti-rabbit IgG. Antibodies against 3-hydroxy-3-methylglutaryl-CoA synthase (HS), 3-hydroxy-3-methylglutaryl-CoA lyase (29), and succinyl-CoA:oxoacid transferase (SCOT) (30) were provided by Dr. G. A. Mitchell and Dr. T. Fukao, respectively. The Western blot signals were quantified by scanning densitometry, and the values from mice fed control diets were assigned the number 1.0. Results are expressed as the means Ϯ S.D. of three determinations.
Northern Blot Analysis and RNase Protection Assay-Total RNA was isolated from liver using the acid guanidinium thiocyanate/phenol/chloroform extraction method. RNA was glyoxylated, electrophoresed, transferred to a nylon membrane, and then hybridized at 42 o in 50% formamide hybridization solution using 32 P-labeled cDNA probes as described previously (21,22). Equal loading was verified by the intensity of methylene blue-stained 18 S and 28 S RNA or by probing the blots with 18 S RNA probe. Changes in mRNA levels were estimated by densitometric scanning of autoradiograms.
RNase protection assay was performed using the following genespecific probes: PPAR␣, nucleotides 1186 -1565 (GenBank TM accession number X57638) (16); PPAR␥, nucleotides 1597-1914 (GenBank TM accession number U01841) (39); PPAR␦, nucleotides 1004 -1268 (Gen-Bank TM accession number U10375) (22); and CYP4A1, nucleotides 1421-1555 (referred to as CYP4A10, GenBank TM accession number ABO18421). Antisense RNA probes were transcribed in the presence of [ 32 P]UTP (20 mCi/ml, 800 Ci/mmol, Amersham Pharmacia Biotech) using the MAXIscript in vitro transcription kit (Ambion, Austin, TX). After transcription, the labeled riboprobes were purified in a 5% TBE/ urea polyacrylamide Ready Gel (Bio-Rad). Probes were eluted from the polyacrylamide gel fragments, and their activity was measured in a scintillation counter. Total RNA isolated from liver was hybridized with labeled probes overnight and then digested for 30 min with RNase A/RNase T1 mix at 37°C. Protected fragments were precipitated and resuspended in 3 l of gel loading buffer. The samples were loaded onto a 6% polyacrylamide sequencing gel 0.4 mm in thickness (Bio-Rad). After electrophoresis, the gel was dried and exposed to film or a Phos-phorImager plate (Molecular Dynamics, Amersham Pharmacia Biotech) overnight at room temperature without intensification. Quantitation was with a Molecular Dynamics Storm 860 PhosphorImager.
Statistical Analysis-Statistical comparisons were made by using Student's t test or two-way analysis of variance. A statistically significant difference was defined as p Ͻ 0.05.

RESULTS
Fatty Liver in Fasted Mice-Since fasting increases the capacity for fatty acid oxidation in liver under normal conditions, we subjected wild type, PPAR␣ Ϫ/Ϫ , DKO, and AOX Ϫ/Ϫ mice to fasting for up to 96 h for a comparative analysis of liver morphology. After 48 h starvation, the livers of PPAR␣ Ϫ/Ϫ and DKO mice were paler compared with those of fasted wild type mice, and this difference in pallor indicative of severe steatosis was grossly exaggerated with prolonged fasting. A representative example of a typical gross appearance of liver of a fed and 66-h fasted DKO mouse is illustrated in Fig. 1 A. Comparative histologic appearance of liver of fed and fasted wild type, PPAR␣ Ϫ/Ϫ , and DKO mice, as revealed by Oil Red O staining (to visualize neutral lipid) of frozen sections, is illustrated in Fig. 1. In fed wild type mice there is no detectable fatty change in hepatocytes other than the presence of Oil Red O-positive droplets in stellate cells (Fig. lB). When fasted for 48 -72 h, these wild type mice exhibited subtle steatosis in centrizonal hepatocytes (Fig. 1C, arrows). As reported elsewhere (22), under fed state, only a few centrilobular hepatocytes in PPAR␣ Ϫ/Ϫ mice and few scattered periportal hepatocytes in DKO mice revealed fatty change (Fig. 1, D and F The degree of hepatic steatosis appeared slightly more prominent in females during the first 48 h of starvation but with prolonged starvation (66 -96 h); the differences in fatty change between males and females were not apparent. We also subjected AOX Ϫ/Ϫ mice to 48-and 72-h starvation and found that regenerated hepatocytes with eosinophilic cytoplasm are resistant to lipid accumulation, whereas cells already steatotic appeared nearly the same or only slightly more steatotic (Fig. 2G). In order to demonstrate that the eosinophilic hepatocytes that do not exhibit steatosis in response to fasting are indeed cells that have regenerated and therefore resistant, we administered bromodeoxyuridine for 4 days in drinking water and assessed its incorporation in hepatocyte nuclei by immunoperoxidase staining (Fig. 2H). During the 4-day labeling period, several hepatocytes have incorporated this precursor indicating DNA synthesis and cell proliferation in cells that are at the interface between steatotic cells and cells with abundant eosinophilic cytoplasm that are resistant to fatty change. In contrast, an occasional cell showed bromodeoxyuridine incorporation in the livers of either fed or starved wild type, PPAR␣ Ϫ/Ϫ , and DKO mice (Fig. 2I) indicating minimal cell proliferation.
Changes in Major Metabolic Fuels upon Starvation-Plasma glucose levels were not much different among the four groups of mice under fed conditions, and these levels were decreased by about 50% at 48 h of starvation, and similar levels were maintained at 72 h in all groups (Fig. 3A). Plasma lactate levels in all groups were nearly the same under non-starved conditions, and on fasting gradual decreases in lactate concentrations were observed in all four groups of animals (data not shown). Lactate, together with alanine, serves as a major precursor for gluconeogenesis in liver under starvation. Glycogen in liver, a reservoir of glucose, was largely depleted within 48 h of starvation, and this reduction was sustained at 72 h of fasting in wild type, PPAR␣ Ϫ/Ϫ , and DKO mice (Fig. 3B). The hepatic glycogen content in AOX Ϫ/Ϫ mice was lower than that found in wild type mice even under the non-starved state. Similar decrease in hepatic glycogen content was noted in non-fasted wild type mice maintained on a diet containing a peroxisome proliferator such as ciprofibrate suggesting that PPAR␣ activation leads to reduction in hepatic glycogen content. 2 Hepatic glycogen content of PPAR␣ Ϫ/Ϫ and DKO mice maintained on control diet was similar to that of wild type mice, but the glycogen content in liver of these mice did not decrease following cipro-2 T. Hashimoto, unpublished data. fibrate, further suggesting that the basal glycogen content is affected by the PPAR␣ function. 2 In the wild type mice, fasting caused a nearly 2-fold increase in plasma FFA levels at 48 h, and by 72 h the FFA returned to near normal levels (Fig. 3C). Plasma FFA levels in other groups were nearly the same as those of wild type mice at 48 h fasting. Unlike in wild type mice, high levels of FFA were maintained at 72 h in PPAR␣ Ϫ/Ϫ , DKO, and AOX Ϫ/Ϫ mice. A large part of the product of hepatic fatty acid oxidation, acetyl CoA, is converted into ketone bodies, acetoacetate, and 3-HB. These are exported by the liver to other organs where they are utilized as energy substrates. The plasma 3-HB concentration was dramatically increased in fasted wild type and AOX Ϫ/Ϫ mice, suggesting enhanced hepatic mitochondrial fatty acid ␤-oxidation. In both PPAR␣ Ϫ/Ϫ and DKO mice, increases in plasma ketone body levels were not as marked (Fig. 3D), suggesting that fasting in these animals does not lead to increased mitochondrial oxidation.
Under fed conditions, plasma TG levels were slightly higher in both PPAR␣ Ϫ/Ϫ and DKO mice as compared with wild type mice (Fig. 3E). Plasma TG levels decreased following fasting in wild type, PPAR␣ Ϫ/Ϫ , and DKO mice (Fig. 3E). Interestingly, plasma TG levels in AOX Ϫ/Ϫ mice were lower even under nonstarved conditions, and there appeared to be a slight increase in TG levels as a result of starvation. As expected TG accumulated in liver when starved (Fig. 3F). This is in part to increased TG synthesis using an excess amount of FFA from adipose tissue, and this synthetic capacity exceeds that of he-patic TG secretion. As shown in Fig. 3F, the TG content in wild type mouse increased about 10-fold by starvation, reaching about 150 mg/g liver. TG accumulation in livers of mutant mice (i.e. in mice lacking PPAR␣) was more than 300 mg/g liver. In contrast, the TG content in AOX Ϫ/Ϫ mouse liver was high under fed conditions, but TG accumulation by starvation in this mouse was lower than that in the wild type mouse (Fig.  3F). This is in most part due to the predominance of regenerated hepatocytes in AOX Ϫ/Ϫ livers that are resistant to lipid accumulation (Fig. 2G). Dramatic increases in hepatic TG/ protein ratio are evident in 72-h fasted PPAR␣ Ϫ/Ϫ when compared with similarly fasted wild type mice, but the increase was more pronounced in fasted DKO livers (Fig. 4A). In contrast, the TG/protein ratio was increased only modestly in 72-h fasted AOX Ϫ/Ϫ mouse liver, which is consistent with the presence in liver of hepatocytes that are regenerated and resistant to steatosis (Fig. 2G). Hepatic carnitine levels were high in the livers of AOX Ϫ/Ϫ mice under fed state when compared with other groups (Fig. 4B). Fasting for 48 and 72 h produced dramatic increases in hepatic carnitine levels in the wild type mice, but mice deficient in PPAR␣ and those nullizygous for both PPAR␣ and AOX showed no changes in carnitine content. No significant increases in TG levels were observed in kidney and heart of starved wild type, PPAR␣ Ϫ/Ϫ , DKO, and AOX␣ Ϫ/Ϫ mice (data not shown). This may be due to controlled uptake of FFA by these extrahepatic organs and that uptake does not exceed the energy demand of these organs.
Changes in Quantities of Fatty Acid ␤-Oxidation Enzymes-Systematic quantification of fatty acid ␤-oxidation enzymes in liver was conducted by immunoblot analysis. As shown in Table I, hepatic very long chain acyl-CoA synthetase (VLACS) was higher in AOX Ϫ/Ϫ mice under non-starved conditions and increased by starvation in wild type but not in PPAR␣ Ϫ/Ϫ and DKO, suggesting that the increase of this enzyme is dependent on the presence of PPAR␣. In AOX Ϫ/Ϫ mice, VLCAS amount did not increase any further because of sustained PPAR␣ activation in these mice and attendant spontaneous peroxisome proliferation (21). The levels of VLACS do not relate to the mitochondrial ␤-oxidation activity, because of its presence only in microsomes and peroxisomes and not in mitochondria.
The mitochondrial fatty acid ␤-oxidation enzymes that increased by starvation in wild type mouse liver consisted of CAT, very long chain acyl-CoA dehydrogenase (VLCAD), and the mitochondrial trifunctional protein (TFP), an enzyme complex exhibiting the activities of enoyl-CoA hydratase/3-hydroxyacyl-CoA dehydrogenase/3-ketoacyl-CoA thiolase. These enzymes were higher in AOX Ϫ/Ϫ mice than in wild type mice. It is noteworthy that VLCAD and TFP are inner mitochondrial membrane-associated proteins and are responsible for the initial ␤-oxidation spiral of long chain fatty acids before these fatty acids are oxidized by the classical matrix enzymes. Although the levels of MCAD and SCAD were elevated in the liver of fed AOX Ϫ/Ϫ mouse, and these high levels were sustained following fasting, no increases in MCAD and SCAD protein content were found in the liver of wild type mouse following 48 and 72 h of fasting (Table I).
Mitochondrial acetoacetyl-CoA-specific thiolase (MTL2) and HS are the ketolytic enzymes. Ketogenesis is a final phase of fatty acid oxidation in liver, forms a single pathway utilizing acetyl CoA, and is regulated by the supply of acetyl CoA and by the activity of rate-limiting reaction catalyzed by HS. It was described that the half-life of the enzyme protein was shorter (31), and its expression was related to the PPAR␣ function (32) and that the enzyme catalytic activity was reduced by succinylation (33). The HS content in AOX␣ Ϫ/Ϫ mouse was higher than that in the wild type mouse, and the contents in both groups increased by starvation (Table I). Most of the peroxisomal enzymes listed here increased by starvation in the liver of the wild type mouse. The protein content of these enzymes was lower in PPAR␣ Ϫ/Ϫ and DKO mice and higher in AOX Ϫ/Ϫ mice.
Changes in the content of kidney and heart mitochondrial and peroxisomal enzyme proteins were less responsive to starvation (Table II). None of the mitochondrial enzymes, except for carnitine palmitoyltransferase II, changed in the kidney of wild type mice following fasting. The amounts of several enzymes in kidney under fed state were lower in PPAR␣ Ϫ/Ϫ mice (Table II). The contents of octanoyl-CoA synthetase, a mitochondrial matrix enzyme activating medium chain fatty acids, and carnitine palmitoyltransferase II were higher in AOX Ϫ/Ϫ mouse kidney. The changes in heart enzymes were much less remarkable (data not shown). No increase in the enzyme content by starvation was observed. The content of a few enzymes such as VLCAD, SCAD, and carnitine palmitoyltransferase were found to be lower in PPAR␣ Ϫ/Ϫ and DKO mice than those of the wild type mice under fed conditions. Ketolysis takes place in the mitochondria of extrahepatic cells. Acetoacetate is activated to the CoA ester by SCOT and is then converted to acetyl CoA by MTL2. Ketone body utilization is proportional to the circulating level (34), and the ketolytic capacity varies directly with the SCOT activity. Kidney and heart have high SCOT activity among the extrahepatic tissues. The SCOT content in these organs was not much different among the animal groups and was not changed by starvation.
Liver mRNA Levels-The expression levels of a select subset of genes encoding peroxisomal and microsomal fatty acid oxidation enzymes have been ascertained by Northern analysis (Fig. 5). In fasted wild type mouse liver, the mRNA levels of L-PBE, PTL, CY4A1, and CYP4A3 increased 5-20-fold, whereas no induction occurred in the livers of starved PPAR␣ Ϫ/Ϫ and DKO mice (Fig. 5). The mRNA levels of all these genes were already elevated in AOX Ϫ/Ϫ mouse liver, and fasting did not cause additional increases. Catalase mRNA levels remained unchanged in all groups. We determined the mRNA levels of PPAR␣, PPAR␦, and PPAR␥ in liver of wild type and mutant mice following 72 h of starvation by Northern analysis and RNase protection assay (Fig. 6). PPAR␣ mRNA content was slightly reduced in fasted wild type mouse liver, and as expected no PPAR␣ mRNA was detected in PPAR␣ Ϫ/Ϫ and DKO mouse livers (Fig. 6A). In AOX Ϫ/Ϫ livers, the PPAR␣ mRNA level was somewhat lower than that seen in fed wild type mouse (Fig. 6, A and B). PPAR␥ mRNA concentration was lower in fed PPAR␣ Ϫ/Ϫ and DKO mouse livers as compared with wild type. Fasting caused an up-regulation in PPAR␥ mRNA level in PPAR␣ Ϫ/Ϫ and DKO mouse liver, whereas the PPAR␥ mRNA level decreased in wild type mouse liver. The level of PPAR␥ mRNA in fasted PPAR␣ Ϫ/Ϫ and DKO mouse liver was similar to that seen in the liver of fed wild type mouse (Fig. 6B). PPAR␦ mRNA levels in the liver of all mutant mice were lower than that found in wild type mice, and fasting caused a reduction in all animals. We also examined the PPAR␣ mRNA level in wild type mice fasted for 24, 48, and 72 h and found that fasting caused a progressive decrease in mRNA concentrations at 48 and 72 h (data not shown).

DISCUSSION
Fatty acids are metabolized via the mitochondrial and peroxisomal ␤-oxidation enzyme systems with partly overlapping substrate spectra (4). Oxidation of the major portion of medium and long chain fatty acids occurs in mitochondria and that of the very long chain fatty acids takes place preferentially in peroxisomes (4). Long chain and very long chain fatty acids are also metabolized by the microsomal CYP4A1 and CYP4A3 fatty acid -oxidases, resulting in the formation of dicarboxylic acids that are further degraded by peroxisomal ␤-oxidation system (5,6). Under normal physiological conditions, mitochondrial ␤-oxidation is the dominant metabolic pathway, whereas the extramitochondrial fatty acid oxidation occurring within peroxisomes and endoplasmic reticulum plays a minor role (4). During starvation, fatty acids entering into the liver constitute the major source of energy, and they require efficient hepatic oxidation to generate ketone bodies to serve as fuels for other tissues (1,2,7). The availability of mice (i) deficient in peroxisomal AOX (21), (ii) deficient in PPAR␣ (16), and (iii) those nullizygous for both PPAR␣ and peroxisomal AOX (22) enabled us to explore the effect of genotype on energy utilization during fasting and on hepatic phenotype. In this paper, we show that fasted PPAR␣ Ϫ/Ϫ and DKO mice exhibit profound impairment of fatty acid oxidation and that DKO mice reveal a greater degree of hepatic steatosis when compared with PPAR␣ Ϫ/Ϫ mice ( Figs. 1 and 2). Following 48 and 72 h of fasting, hepatic TG/protein ratio was substantially higher in DKO mice than that observed in PPAR␣ Ϫ/Ϫ mice. Both DKO and PPAR␣ Ϫ/Ϫ mice manifested hypoglycemia, hypoketonemia, increased serum FFA, and increased serum TG, all indicative of impaired hepatic fatty acid oxidation (Fig. 3). The reductions in hepatic glycogen level and increase in hepatic TG concentration resulting from fasting further attest to disturbed fatty acid oxidation in these DKO and PPAR␣ Ϫ/Ϫ animals. On the other hand, as discussed below, in AOX Ϫ/Ϫ mice starvation did not substantially affect total hepatic TG levels suggesting that regenerated hepatocytes with massive spontaneous peroxisome proliferation are resistant to steatosis. Fasting did induce hypoglycemia and an increase in plasma FFA in AOX Ϫ/Ϫ mice. Nevertheless, they exhibited ketogenesis similar to that occurring in wild type mice as evidenced by increased levels of serum 3-HB reflecting increased mitochondrial and microsomal fatty acid FIG. 4. Changes in liver TG/protein ratio and carnitine content following fasting. Wild type, PPAR␣ Ϫ/Ϫ DKO, and AOX Ϫ/Ϫ mice fed (0) or fasted for 48 or 72 h were analyzed for hepatic TG, carnitine, and protein content. A represents TG/protein ratios, and B represents liver carnitine levels in various groups under fed and fasted conditions.

Quantification of mitochondrial and peroxisomal fatty acid oxidation enzymes in kidney
Total kidney proteins were subjected to immunoblot analysis, and the signals were quantified. The values represent the signal intensities obtained with the fed wild-type mice (ϭ 1.0). SCPx was undetectable. DKO, double knock-out mice nullizygous for both PPAR␣ and AOX.
Group starvation oxidation due to sustained activation of PPAR␣ by endogenous ligands (21).
In wild type mice, as well as in other groups of mice used in this study, starvation caused marked increases in plasma FFA due to excessive lipolysis in adipose tissue resulting from carbohydrate deficit. If mobilization of FFA exceeds the demand for lipid oxidation, re-esterification of surplus FFA to TG will occur in liver. There is sufficient evidence to indicate that certain crucial enzymes involved in fatty acid metabolism in mitochondria, peroxisomes, and endoplasmic reticulum are upregulated during starvation by PPAR␣ (9, 18 -20). In the present study, animals were starved for up to 72 h to determine whether hepatic mitochondrial and peroxisomal fatty acid-metabolizing enzymes are increased by starvation and if such increases depend on the presence of PPAR␣. In wild type mouse liver, starvation produced increases in certain enzymes such as VLACS, CAT, VLCAD, TFP, MTL2, HS, AOX, L-PBE, D-PBE, and sterol carrier protein X or 3-ketoacyl-CoA thiolase/sterol carrier protein 2 (Table I). Increases in hepatic MCAD and SCAD mRNA levels were reported in wild type mice following 24 h of starvation (19,20), whereas in the present study we observed no significant increases in the levels of these two proteins following 48 and 72 h of fasting. The observed differences may be due to the fact that mRNA levels and not protein levels were determined, and the duration of fasting was 24 h in previously reported studies (19,20) and not prolonged as in the present study. In AOX null mouse liver, the basal MCAD and SCAD protein levels were significantly higher than that of wild type mice, and these levels remained high following fasting (Table I). Our studies also confirm increases in mRNA levels of PPAR␣-regulated genes such as L-PBE, PTL, CYP4A1, and CY4A3 in livers of wild type mice following fasting. Thus, PPAR␣-dependent increases in fatty acid oxidation systems observed in fasted wild type mice appear to metabolize FFA entering the liver efficiently and minimize the development of hepatic steatosis. Indeed, in wild type mice fasted for 96 h, there was no morphologically discernible hepatic steatosis.
As reported elsewhere, the constitutive levels of expression of several mitochondrial enzymes involved in lipid metabolism are substantially lower in the livers of PPAR␣ Ϫ/Ϫ and DKO mice as compared with wild type mice indicating defective mitochondrial fatty acid catabolism (17,22). Absence of PPAR␣, in both PPAR␣ Ϫ/Ϫ and DKO mice, impairs the induction of mitochondrial ␤-oxidation in liver following fasting which contributes to hypoketonemia and hepatic steatosis. Fasting was associated with even greater increases in TG/ protein ratio and more pronounced steatosis in the liver of DKO mice than that observed in PPAR␣ Ϫ/Ϫ mice. We attribute this added severity of hepatic steatosis in DKO mice to the absence of peroxisomal ␤-oxidation system due to AOX deficiency. Furthermore, since peroxisomal ␤-oxidation system is required for the metabolism of toxic dicarboxylic acids generated by constitutive levels of microsomal fatty acid -oxidation enzymes, these metabolites can act synergistically to induce a greater degree of fatty change in fasted DKO mice than that encountered in PPAR␣ Ϫ/Ϫ mice. We determined the hepatic mRNA levels of PPAR␣, PPAR␥, and PPAR␦ isoforms in all mice following 72 h of fasting using RNase protection assay, and we found slight reductions in the levels of all three isoforms in wild type mice. In PPAR␣ Ϫ/Ϫ , and DKO mouse livers, fasting caused a less than 2-fold increase in PPAR␥ mRNA levels, whereas the levels of PPAR␦ were reduced as seen in wild type mice. No appreciable reductions in PPAR isoform levels occurred in AOX Ϫ/Ϫ mice following fasting. In a previous study, increases in PPAR␣ mRNA content were noted in the livers of wild type mice fasted for 24 h (19). Since we found reduction in PPAR␣ mRNA level in mice starved for 72 h, we determined PPAR␣ levels in mice fasted for 24 -72 h, and we noted a decrease in mRNA levels in animals fasted for 48 and 72 h, suggesting that prolonged fasting results in an adaptive state.
In this paper we also present data on the effect of fasting in mice deficient in peroxisomal AOX. Although the basal content of hepatic TG in AOX Ϫ/Ϫ mice was somewhat higher than that of other groups, reflecting the presence of preexisting microvesicular steatosis in some centrizonal hepatocytes (Fig. 2G), fasting for 48 and 72 h induced only a minimal increase in TG/protein ratios. There are two possible explanations for the absence of fasting-related amplification of hepatic steatosis in these animals. First, PPAR␣ is hyperactive in these animals leading to sustained spontaneous peroxisome proliferation and increases in the levels of mitochondrial and microsomal fatty acid oxidation systems in liver (21). PPAR␣-mediated spontaneous increase in hepatic mitochondrial ␤-oxidation system in these AOX Ϫ/Ϫ animals, as evidenced by increases in VLCAS, CAT, VLCAD, MCAD, and SCAD, appears highly effective in generating ketone bodies. Increases in plasma 3-HB levels occurred in AOX Ϫ/Ϫ mice similar to those observed in fasted wild type mice. In contrast, the plasma 3-HB levels in fasted PPAR␣ Ϫ/Ϫ and DKO mice indicate a significantly lower activity of fatty acid oxidation than that occurring in the liver of wild type and AOX Ϫ/Ϫ mice. Second, the reduced TG/protein ratio in 16 -20-week-old AOX Ϫ/Ϫ mice subjected to starvation in this study can be due to the fact that their livers consist of many regenerated hepatocytes that progressively replace steatotic hepatocytes, and this regenerative process extends toward centrizonal areas (Fig. 2G). We have presented evidence for hepatocyte proliferation at the interface between steatotic and non-steatotic hepatocytes. These regenerated hepatocytes with abundant cytoplasm and massive peroxisome proliferation are resistant to fatty change. This is analogous to the emergence of hepatocytes in chronic alcoholic liver disease that no longer become steatotic (35). No liver cell proliferation occurred in wild type, PPAR␣ Ϫ/Ϫ , and DKO mice either fed or fasted under the conditions used in this study. The livers of fed AOX Ϫ/Ϫ mice show a remarkable increase in microsomal fatty acid oxidation enzymes due to sustained activation of PPAR␣ by unmetabolized endogenous ligands (21), and fasting produced no further increase in the levels of CYP4A1 and CYP4A3 mRNA levels reflecting saturation of the receptor by ligands. Increased -oxidation generates dicarboxylic acids that cannot be metabolized FIG. 5. Northern blot analysis of total RNA extracted from the liver of wild type (wild), PPAR␣ ؊/؊ PPAR␣ ؊/؊ /AOX ؊/؊ (DKO), and AOX ؊/؊ mice. Lanes 0, 48, and 72 in each group represent fed (0), 48-h fasting (48), and 72-h fasting (72). Twenty g of total RNA was electrophoresed on a 0.8% agarose gel, blotted onto a nylon membrane, and hybridized with different random-primed 32 P-labeled L-PBE, peroxisomal 3-ketoacyl-CoA thiolase (PTL), CYP4A1 CYP4A3, and catalase (CTL) probed as shown. It should be noted that rat CYP4A1 and CYP4A3 probes used for hybridization may exhibit differences in the hybridization to mouse CYP4A mRNAs. The 18 S RNA is for loading control.
in the absence of peroxisomal ␤-oxidation. Although these toxic metabolites seem to exert severe fatty change in the livers of younger AOX Ϫ/Ϫ mice (21), the regenerated hepatocytes in older mice appear immune to fatty change.
It has been known by in vitro experiments that maximal uptake of FFA by liver is not changed by fasting (36), that the uptake rate is dependent on the FFA concentration, and that exogenously imported fatty acids are oxidized to ketone body and converted into TG and phospholipids in proportion to the uptake of FFA (37). However, a recent report (38) describes that uptake of FFA in vivo is saturated at the level of about 1 mM. Rate of FFA uptake was not determined in this study, but increases in the plasma ketone body and the hepatic TG content suggest that the FFA uptake did not limit the oxidation and esterification of the imported FFA. Studies on the relationship of PPAR␣ Ϫ/Ϫ expression to the proteins involved in fatty acid import (39,40) suggest that proteins such as fatty-acid translocase, fatty acid transport protein, and L-fatty acid-binding protein play a role in the uptake of fatty acids by liver. But marked increases in hepatic TG content in PPAR␣ Ϫ/Ϫ and DKO mice, irrespective of the PPAR␣ function, suggest that a major mechanism of FFA uptake by liver is not limited by these proteins under the starved conditions. Marked differences in fasting-related increases in the levels of TG between liver and these extrahepatic organs are evident. Differences in lipid metabolisms between liver and extrahepatic organs under starvation are attributed to rapid activation of PPAR␣-responsive genes in liver and PPAR␣-inducible fatty acid oxidation systems in liver play a vital role in energy metabolism and in the prevention of hepatic steatosis.
The mitochondrial ␤-oxidation is regulated by the carnitine content (4,41). The hepatic carnitine content under fed conditions was slightly lower in mice without PPAR␣ than that of wild type mice. Carnitine content in wild type mice increased about 3-fold upon starvation for 48 h, whereas in PPAR␣ Ϫ/Ϫ and DKO mice carnitine levels remained unchanged, suggesting that absence of PPAR␣ in these two genetically altered mice influences the carnitine metabolism and its response to fasting. It is of considerable interest to note that hepatic carnitine content in AOX Ϫ/Ϫ mice, which are under PPAR␣ hyperfunction (21), is higher than that found in wild type mice, and these high levels were maintained during starvation. We propose that hepatic carnitine level is regulated by PPAR␣ although the mechanism remains unknown.
Inborn errors in mitochondrial and peroxisomal ␤-oxidation enzymes have received major attention thus far as causes of hypoglycemia and hepatic dysfunction (1)(2)(3). In summary, our results with PPAR␣ Ϫ/Ϫ , DKO, and AOX Ϫ/Ϫ mice and data from other laboratories with PPAR␣ Ϫ/Ϫ and other genetically altered mice also focus on the importance of transcriptional regulation of genes involved in lipid metabolism in energy utilization (18 -22, 42). In humans, the PPAR␣ levels appear lower than that found in rats and mice (43), raising the issue of the effectiveness of PPAR␣-inducible fatty acid oxidation systems in different species in dealing with conditions of stress that lead to reduced energy intake. A PPAR␣ splice variant that may negatively interfere with wild type PPAR␣ has been described recently (44), and this finding also raises the question of countering the induction of PPAR␣-regulated genes leading to abnormal energy utilization.
FIG. 6. The levels of PPAR␣, PPAR␥, and PPAR␦ in mouse liver following fasting. A and B, wild type (WT), PPAR␣ Ϫ/Ϫ , DKO, and AOX Ϫ/Ϫ mice were either fasted (ϩ) or fed (Ϫ) for 72 h. Total RNA was isolated and hybridized with an equal amount of radioactivity of antisense riboprobes specific for each of the PPAR genes, PPAR␣, PPAR␥, and PPAR␦. Actin is used as a control for RNA. After RNase digestion, the reaction mixtures were separated on a 6% polyacrylamide gel. As expected, PPAR␣ mRNA is absent in PPAR␣ Ϫ/Ϫ and DKO mouse liver. B, the data are presented as average of three animals in each group and expressed as relative light units and normalized to actin mRNA content.