Estrogen-related Receptor α (ERRα) Is a Transcriptional Regulator of Apolipoprotein A-IV and Controls Lipid Handling in the Intestine*

The estrogen-related receptor α (ERRα) is an orphan member of the superfamily of nuclear receptors involved in the control of energy metabolism. In particular, ERRα induces a high energy expenditure in the presence of the coactivator PGC-1α. However, ERRα knockout mice have reduced fat mass and are resistant to diet-induced obesity. ERRα is expressed in epithelial cells of the small intestine, and because the intestine is the first step in the energy chain, we investigated whether ERRα plays a function in dietary energy handling. Gene expression profiling in the intestine identified a subset of genes involved in oxidative phosphorylation that were down-regulated in the absence of ERRα. In support of the physiological role of ERRα in this pathway, isolated enterocytes from ERRα knockout mice display lower capacity for β-oxidation. Microarray results also show altered expression of genes involved in dietary lipid digestion and absorption, such as pancreatic lipase-related protein 2 (PLRP2), fatty acid-binding protein 1 and 2 (L-FABP and I-FABP), and apolipoprotein A-IV (apoA-IV). In agreement, we found that ERRα–/– pups exhibit significant lipid malabsorption. We further show that the apoA-IV promoter is a direct target of ERRα and that its presence is required to maintain basal level but not feeding-induced regulation of the apoA-IV gene in mice. ERRα, in cooperation with PGC-1α, activates the apoA-IV promoter via interaction with the apoC-III enhancer in both human and mouse. Our results demonstrate that apoA-IV is a direct ERRα target gene and suggest a function for ERRα in intestinal fat transport, a crucial step in energy balance.

Nuclear receptors are ligand-inducible transcription factors that control important metabolic pathways needed for development and homeostasis. This superfamily includes classic receptors for ligands such as steroid hormones, vitamin D, and thyroid hormones as well as orphan nuclear receptors for which there are no physiological ligands associated at the time of their discovery (1). Estrogen-related receptor ␣ (ERR␣) 1 is an orphan nuclear receptor originally identified on the basis of its homology with the estrogen receptor ␣ (2). The two receptors display transcriptional cross-talk and share some target genes and synthetic ligands (reviewed in Ref. 3). Furthermore, ERR␣ may have a function in bone remodeling (4,5) and as a prognostic marker of breast cancer (6,7), two classical estrogenresponsive tissues.
Apart from its proposed role as a modulator of estrogen receptor-dependent pathways, several lines of evidence indicate that ERR␣ acts primarily as a regulator of energy metabolism. First, ERR␣ is known to be expressed in tissues with high ␤-oxidation activity such as the brown fat, kidney, heart, and intestine (2,8,9). Furthermore, ERR␣ expression is induced during adipocyte differentiation (10) and in response to stimuli that increase energy demand such as in the liver under fasting conditions (11) and in skeletal muscle and brown fat from mice exposed to cold (12). Second, ERR␣ regulates the gene encoding medium chain acyl-CoA dehydrogenase (MCAD), which catalyzes the initial step in mitochondrial fatty acid ␤-oxidation (9,13,14). Third, the peroxisome proliferatoractivated receptor (PPAR)␥ coactivator 1␣ (PGC-1␣), a coactivator central to the control of energy expenditure (15), has recently been described as a key partner of ERR␣, assuring both its expression and transactivation potential (10 -12, 14, 16). In fact, ERR␣ appears to be an essential transducer of PGC-1␣ action in mediating mitochondrial biogenesis (17,18).
We recently observed that the absence of ERR␣ in knockout mice leads to reduced fat mass and resistance to high fat diet-induced obesity (19). In addition, gene expression profiling experiments in adipose tissue from ERR␣ knockout mice revealed alterations in the expression of genes implicated in the regulation of adipogenesis and energy metabolism. However, we could not detect significant changes in dietary intake or energy expenditure between the wild-type and ERR␣-deficient mice. Knowing that ERR␣ is expressed in the intestine, defective uptake or utilization of dietary nutrients in ERR␣-deficient mice could also contribute to the observed lean phenotype. In fact, the intestine is the site essential for the transport of alimentary fat. After the digestive phase, the lipolytic products are absorbed by the enterocytes, in which a complex series of sequential events result in their packaging as chylomicrons. In addition, the intestine contributes to total body ␤-oxidation because of its large surface and specific needs (20).
In the current study, we have addressed the functional role of ERR␣ in intestinal tissues, using the ERR␣ knockout mice model and the power of gene expression profiling. Our findings show that ERR␣ is involved in mitochondrial electron transport and intestinal lipid handling. In particular, ERR␣ interacts with the apolipoprotein (apo)C-III enhancer and specifically regulates the adjacent apoA-IV gene.

MATERIALS AND METHODS
Animals-ERR␣ Ϫ/Ϫ mice were generated as described (19). Mice used were either into 129/SvJ or C57BL/6J genetic backgrounds. Animals were housed under specific pathogen-free conditions in the animal facility of the McGill University Health Centre and were fed standard laboratory chow.
Lipid Load-ERR␣ Ϫ/Ϫ and wild-type littermate adult mice were fasted overnight (18 h) followed by the administration of either normal saline or 200 l of olive oil by oro-gastric gavage. Mice were sacrificed after 2 h, RNA was extracted from jejunal tissues and submitted to reverse transcription. cDNA samples were used as templates to measure mouse apoA-IV expression by quantitative RT-PCR. ApoA-IV gene expression levels were compared between mice fed saline versus olive oil as well as between ERR␣ Ϫ/Ϫ and wild-type mice.
RNA in Situ Hybridization-Intestinal tissues from C57Bl/6J mice were fixed in 4% paraformaldehyde, dehydrated, and embedded in paraffin. Sections (7-m thick) were processed for in situ hybridization according to published protocols with a 33 P-labeled riboprobe obtained from the 3Ј-untranslated region of mouse ERR␣ (bp 1730 -2027) (AS). Negative control was hybridized with a sense riboprobe (S).
Western Immunoblots-Nuclear proteins were isolated from the digestive epithelium of mice by an adaptation of a previously described method (21). Briefly, mice were sacrificed, and the digestive tract was separated into sections of stomach, duodenum, jejunum, ileum, and proximal colon. Each segment was opened longitudinally and rinsed with cold phosphate-buffered saline. The segments were further cut in 5-mm long pieces and incubated in 5 ml of cold MatriSperse (BD Biosciences) at 4°C for 18 h. The epithelial layer was dissociated by gentle manual shaking. The epithelial suspension was collected, centrifuged, and washed with cold phosphate-buffered saline. Nuclear proteins were then isolated from the epithelial cell pellet as described previously (9). Protein samples were separated on SDS-PAGE, transferred to a hydrophobic polyvinylidene difluoride membrane (Amersham Biosciences), and immunoblotted with ERR␣ polyclonal antiserum raised against the N terminus of mouse ERR␣ (19) and antilaminin B antibody (M20; Santa Cruz Biotechnology). The proteins were visualized with an ECL Western blotting detection system according to instructions of the manufacturer (Amersham Biosciences).
RNA Preparation and Microarray-Total RNA from mouse intestinal tissues (ileum) were prepared with the RNeasy kit with DNase treatment on column (Qiagen). RNA from 3-4 mice of each genotype was pooled to compare gene expression in three experiments: in female of the C57Bl/6J background, in male of the C57Bl/6J background and in both male and female of the 129/SvJ genetic background. The mouse 6.5K microarray was used for gene expression profiling experiments. This chip is based on the Mouse Genome Oligo Set (70 mer) version 1.1 from Operon, printed by the Samuel Lunenfeld Research Institute Microarray Laboratory. Briefly, 2.5 g of total RNA from ERR␣ Ϫ/Ϫ and ERR␣ ϩ/ϩ mice intestinal tissues were indirectly labeled with either Cy3-or Cy5-reactive dyes followed by co-hybridization on Operon Slides (protocol at the following Web site: www.queensu.ca/microarray/). Each RNA pair was hybridized on 2 arrays employing a fluor reversal strategy such as the ERR␣ Ϫ/Ϫ sample was labeled with Cy3 for one hybridization and with Cy5 for the other hybridization. Chips were scanned using the ScanArray4000, and quantitation was done using the Quant-Array software (Version 3.0). After pixel intensity determination, background subtraction, and normalization, the relative expression levels for every gene were depicted as the logarithm of the ratio of intensity in ERR␣ Ϫ/Ϫ over ERR␣ ϩ/ϩ samples. It was observed that logarithms of the ratio were distributed according to a Gaussian distribution, with mean and S.D. Genes having normalized ratio values outside of 2 S.D. were considered differentially expressed. Pertinent genes were defined as genes found to be differentially expressed in 4 of 6 microarray chip experiments.
Quantitative Reverse Transcription-PCR-2.5 g of total RNA from each of the 6 pooled RNA samples used for microarray were used to synthesize cDNA in 20 l using Superscript II (Invitrogen) reverse transcriptase and d(T)12-18 random hexamer primers (Amersham Biosciences). cDNA samples were used as templates, and real-time quantitative PCR was performed using LightCycler and Fast Start DNA Master SYBR Green (Roche Applied Science). Intron-spanning specific primers were designed based on sequences from the Celera data base. The specificity of the PCR product was documented by LightCycler melting curve analysis and migration on ethidium bromide-stained agarose gel. Efficiency of the PCR reaction was calculated using the formula E ϭ 10 (Ϫ1/slope) ; the slopes obtained by plotting cycle crossing point values (CP) as a function of cDNA template (22). The relative quantitation for any given gene was calculated after determination of the difference between CP of the target gene and that of the calibrator gene hypoxanthine ribosyl transferase (HPRT) in ERR␣ Ϫ/Ϫ and ERR␣ ϩ/ϩ mice, and correction for efficiency, using the formula: Ratio ϭ (E target ) ⌬CP target(WT-KO) /(E HPRT ) ⌬CP HPRT(WT-KO) (23). CP values are the mean of triplicate measurement, and experiments were done in duplicate.
␤-Oxidation Activity in Intestinal Epithelium-Intestinal epithelial cells from adult mice fasted for 18 h were isolated as described in the Western blot section and homogenized on ice with 5 vol of 250 mM sucrose containing 1 mM EDTA and 10 mM HEPES (pH 7.2). The homogenate was centrifuged at 600 ϫ g for 5 min, and the resulting supernatant was used for ␤-oxidation assay as reported previously (20). Briefly, the ␤-oxidation reaction mixture contained 50 mM Tris-HCl (pH 8.0), 40 mM NaCl, 2 mM KCl, 2 nM MgCl 2 , 1 nM dithiothreitol, 5 mM ATP, 0.2 mM L-carnitine, 0.2 mM NAD ϩ , L-malate 0.5 mM, 0.05 mM flavin adenine dinucleotide (FAD), 0.12 mM CoA, 0.1 Ci of [ 14 C]palmitic acid, 100 g of protein of extract in a final volume of 200 l. The reaction was started by adding the protein extract and incubating the preparation at 37°C for 25 min. The reaction was terminated by adding 200 l of 0.6 N perchloric acid. Negative controls where the reaction was stopped at zero time were carried out in parallel. The reaction mixture was washed three times with 800 l of hexane to remove residual radiolabeled palmitic acid. The radioactivity retained in the aqueous phase was measured. The assay was done in duplicate. Protein concentrations were determined in duplicate using Bio-Rad protein reagents.
Fecal Fat Analysis-As previously described for stool lipid analysis (24), 100-mg aliquots of droppings collected from ERR␣-null and wildtype mice of the indicated ages were mixed with a small amount of [carboxyl-14 C]triolein (112 mCi/mmol) and dried for 1 h in a vacuum oven at 70°C. The solid matter was extracted with 2 ml of chloroform/ methanol (2:1) for 30 min at 60°C, passed through a Whatman No. 1 filter, and brought to a final volume of 4 ml with chloroform/methanol (2:1). The material was back-extracted with 1 ml of H 2 O, and the organic phase was evaporated to dryness. The pellet was resuspended in 2 ml of chloroform/methanol (2:1) and transferred to preweighed vials. The solvent was evaporated, and the vial was taken to a constant weight by drying in a vacuum oven at 70°C. The difference in weight between the starting empty vial and the vial containing the dried lipid was the fecal lipid amount, which was expressed as a percentage of the weight of the starting fecal sample. To obtain stool from suckling animals, 10-day-old animals were sacrificed, and the entire colon was removed. The feces was removed by extrusion, pooled from several animals, and processed as for adult animals. The percent recovery of radiolabeled triolein (85-90%) was determined by subjecting the vial to scintillation counting.
ApoA-IV Promoter Cloning-The mouse apoC-III/apoA-IV intergenic region was amplified by PCR using mouse genomic DNA as template and subcloned into the luciferase reporter plasmid pLuc. Deletion mutants were produced with subcloning into pLuc or ptkLuc luciferase reporter plasmids. The integrity of each construct was confirmed by sequencing.
Plasmids and Cell Transfections-Plasmid expressing the VP16-ERR␣ fusion protein was constructed by subcloning PCR-amplified ERR␣ cDNA into pCMX-VP16 downstream of the transcriptional activation domain of the VP16 protein. The pCDNA3.1HA-hPGC-1␣ vector was obtained from A. Kralli (La Jolla, CA). Caco-2/15 cells were a gift from Dr. J. F. Beaulieu (Université de Sherbrooke, Canada), and COS-1 cells were obtained from the American Type Culture Collection. Cells were cultured in phenol red-free Dulbecco's minimal essential medium containing penicillin (25 units/ml), streptomycin (25 units/ml), and 10% heat-inactivated charcoal-treated fetal bovine serum. 18 h prior to transfection, the cells were seeded in 12-well plates. A total of 1-1.2 g of DNA/well was transfected including 0.5 g of reporter plasmid and 0.25 g of internal control pCMV␤Gal, 0.05-0.2 g of pCMX-ERR␣, 0.1 g of pCMX-VP16-ERR␣, 0.1 g of pCDNA3.1HA-hPGC-1␣ as specified in the figure legends. The DNA was introduced into cells using Fu-GENE 6 (Roche Applied Science). The cells were harvested in a potassium phosphate lysis buffer containing 1% Triton X-100. Luciferase assays were performed as previously described (25). The transfections were normalized to the ␤-galactosidase activity of each sample. All results represent experiment conducted in duplicate at least three times.

ERR␣ Is Expressed in Epithelial Cells throughout the Small
Intestine-In situ hybridization was first used to evaluate the expression of ERR␣ throughout the intestinal mucosa of adult mice. As shown in Fig. 1A (top panel, AS), ERR␣ expression is restricted to epithelial cells with very little stromal hybridization. Western blot analysis was then performed with nuclear extracts from epithelial cells isolated from mice throughout the digestive tract and at the postnatal development stage to examine the presence of the ERR␣ protein. Prominent ERR␣ expression was detected in the small intestine from postnatal days 0 through 21, with lower levels in adult mice (Fig. 1B). In adult mice, ERR␣ was mainly expressed in epithelial cells involved in nutrient absorption such as duodenal, jejunal, and ileal segments of the digestive tract (Fig. 1C). Lower levels of ERR␣ were detected in the colon and stomach.
Mice Deficient for ERR␣ Display Unique Intestinal Gene Expression Profiles Associated with Defects in ␤-Oxidation Capacity-To examine the in vivo function of ERR␣ in the intestine, we took advantage of the previously described ERR␣-null mice (19). Because intestinal tissues from ERR␣-deficient mice were histologically normal (data not shown), we first performed microarray studies to determine intestinal gene expression associated with ERR␣ deficiency (Table I). Because interindividual variation is a concern in gene expression profiles, intestinal tissues from 3-4 mice were pooled before measurements. Furthermore, to identify differentially expressed genes rele-vant to both sex and in various genetic backgrounds, we performed microarrays in female mice of C57/BL6 genetic background, male in the same background and pooled male and female in the 129SvJ background. Last, genes found to be consistently differentially expressed were confirmed by quantitative reverse transcription-PCR. For internal validation, our studies confirmed the absence of ERR␣ expression in null mice (Table I). Interestingly, a group of genes involved in mitochondrial electron transport were consistently down-regulated in the intestine from ERR␣-null mice. This includes NADH dehydrogenase and the cytochrome c oxidase (subunits 5a and 7c; both needed to couple electron transport to proton pumping in order to generate an electrochemical gradient), as well as ATP synthase functioning when the proton gradient across the mitochondrial membrane is effective for oxidative phosphorylation. PCG-1␣ is known to be involved in mitochondrial biogenesis and respiration. We found PGC-1␣ transcript levels to be decreased by 21% in the intestine from ERR␣-null compared with wild-type mice (p Ͻ 0.05), although PGC-1␣ was not consistently identified as differentially expressed in microarray experiments (data not shown). This order of magnitude has been previously reported and could be physiologically important (27). Electron transport chain is linked to the mitochondrial ␤-oxidation and inhibition of the former results in suppression of the latter (28). To investigate the physiological effect of ERR␣ deficiency on intestinal lipid catabolism, we measure the ␤-oxidation activity in intestinal epithelial cells isolated from ERR␣ Ϫ/Ϫ and ERR␣ ϩ/ϩ mice. As shown in Fig. 2, deficiency for ERR␣ is associated with a significant decrease in intestinal palmitic acid oxidation capacity.

ERR␣-null Mice Display Alterations in the Expression of Genes Involved in Intestinal Lipid Handling as Well as Fat
Malabsorption in Suckling Mice-Indeed, microarray results also revealed differential expression in a group of genes involved in dietary lipid handling (Table I). Among them, the expression of pancreatic lipase-related protein 2 (PLRP2) was found to be expressed at lower levels in the intestine from ERR␣-null versus wild-type mice. The expression of fatty acidbinding proteins 1 and 2 (L-FABP and I-FABP, respectively) were also reduced in ERR␣-null versus wild-type mice. Of note are the high affinity and binding of these cytoplasmic FABPs for long chain fatty acids (29). L-FABP is expressed in the liver and small intestine but I-FABP is restricted to the intestinal FIG. 1. Expression of ERR␣ in mice intestinal epithelium. A, ERR␣ expression was studied by in situ hybridization with a 300-bp antisense riboprobe hybridizing to the ERR␣ 3Ј-untranslated region (AS). A sense riboprobe (S) was hybridized as control. B, nuclear extracts were prepared from isolated epithelial cells of the ileum at postnatal days 0, 10, and 21 and in adult mice and analyzed by Western blot using a mouse ERR␣-specific antibody. The blot was stripped and incubated with a lamin B-specific antibody to monitor equal loading and protein integrity. C, nuclear extracts were prepared from epithelial cells of the stomach, duodenum, jejunum, ileum, and proximal colon from adult mice and analyzed by Western blot as above. epithelium. Last, apoA-IV was also intensely down-regulated in the microarray analysis relative to ERR␣ KO mice. ApoA-IV is a component of triglyceride-rich lipoproteins such as chylomicrons, very low density lipoproteins and high density lipoproteins secreted by the enterocytes. Intestinal apoA-IV synthesis is stimulated by fat absorption and the assembly and/or transport of chylomicrons (30). To confirm the effects of ERR␣ deficiency on the in vivo regulation of apoA-IV gene expression, overnight-fasted mice were subjected to a lipid load by gavages to up-regulate intestinal apoA-IV gene expression. The results shown in Fig. 3A demonstrate that ERR␣ Ϫ/Ϫ mice display a significantly lower basal and lipid-induced jejunal apoA-IV expression, compared with wild-type mice. Interestingly, the magnitude of apoA-IV induction by lipid was comparable between wild-type and ERR␣ Ϫ/Ϫ mice (2.6-and 2.39-fold, respectively).
Neonatal and suckling mice thrived on high fat diet from maternal milk and are known to have immature gut with physiological fat malabsorption and poor adaptation. Although we did not observe dietary fat malabsorption in adult mice, ERR␣-deficient pups at 10-days postnatal exhibited a significantly higher magnitude of steatorrhea, indicative of fat mal-absorption, than wild-type pups (Fig. 3B).
Apolipoprotein A-IV Is a Direct Target of ERR␣-The human and mouse apoA-I, apoC-III, and apoA-IV genes are tandemly organized within a 15-kb DNA segment. Within the gene cluster, the apoA-I and -A-IV genes are transcribed in the same direction, whereas the apoC-III gene is transcribed in the opposite direction. Therefore, the apoC-III/A-IV intergenic region constitutes a common 6.6-kb 5Ј-flanking sequence for these two genes. In human, the expression of the apoA-IV gene in the intestine is under the control of the Ϫ700/Ϫ310 promoter of the apoA-IV gene combined with the Ϫ780/Ϫ580 enhancer region of the apoC-III gene (31). This enhancer region has also been shown to direct the intestinal expression of the three genes of the cluster (32). Although there is some degree of conservation between the human and mouse genomes, the functional significance of regulatory sequences in the apoA-IV promoter are not well described in mice. Sequence analysis of the mouse apoC-III/A-IV intergenic region reveals the existence of several potential ERR responsive elements containing the core AGGTCA sequence (or its reverse complement). As a first step in evaluating the significance of these sites, the mouse apoC-III/A-IV intergenic region was cloned in front of a luciferase reporter gene. The binding capacity of ERR␣ to this region was characterized by transient cotransfection using the ERR␣-VP16 expression vector in Caco-2/15 cells. As shown in Fig. 4, the constitutively active ERR␣-VP16 chimera induced a strong apoA-IV promoter-dependent luciferase induction (11.4-fold). This result confirms that ERR␣ can interact directly with the apoA-IV promoter/enhancer regulatory unit. Furthermore, coexpression of ERR␣ with the coactivator PGC-1␣ results in synergistic activation of the apoC-III/A-IV intergenic region. To determine the region required for ERR␣ activation, 5Ј-nested deletion constructs of the apoC-III/A-IV intergenic region were fused to luciferase cDNA and cotransfected with ERR␣-VP16, ERR␣, or both ERR␣ and PGC-1␣. As shown in Fig. 4, truncation of the apoA-IV promoter/enhancer region to Ϫ4513 abolishes responsiveness to ERR␣-VP16 and to ERR␣ in the presence of PGC-1␣. This domain maps to the well defined apoC-III promoter/enhancer region. The proximal region of the apoA-IV promoter (-558) was unresponsive to ERR␣. Since putative ERR␣ response elements can be found in both the apoC-III promoter and enhancer region, we next sought to more precisely map the region required for ERR␣ action on apoA-IV expression. Reporter constructs were designed in which vari-  ous segments of the apoA-IV promoter were cloned in front of the minimal thymidine kinase promoter fused to the luciferase reporter gene. Cotransfection experiments with ERR␣-VP16 as well as ERR␣ in the presence of PGC-1␣ indicate that the apoC-III promoter/enhancer domain (Ϫ5846 to Ϫ4487) is indeed responsible for the ERR␣ regulation of the apoA-IV gene expression. Furthermore, we observed that a 229-bp segment corresponding to the apoC-III enhancer (Ϫ5355 to Ϫ5116) was specifically responsive to ERR␣.
Finally, because of the imperfect homology between the human and mice apoC-III/A-IV intergenic region, we tested whether ERR␣ also interacts with the human apoC-III enhancer, in the native chromatin environment. A ChIP assay was therefore performed in Caco-2/15 cells containing ectopic human ERR␣, a human colon cancer cell line frequently used in the investigation of intestinal apolipoprotein metabolism. As shown in the Fig. 5, ChIP with a human ERR␣-specific antibody resulted in an enrichment for the genomic fragment corresponding to the human apoC-III enhancer region (ϳ4-fold). A lower degree of enrichment was also observed for the apoC-III promoter. However, the two regions are very close and detection of the apoC-III promoter in long chromatin segment enriched for the immunoprecipitated apoC-III enhancer is plausible. Taken together, these data support that ERR␣ interacts with the apoC-III enhancer region and modulates the expression of intestinal apoA-IV via this regulatory region.

DISCUSSION
In this report, we used genetic deletion of the ERR␣ gene in mice in combination with the power of microarray genomic technology to explore the function of ERR␣ in the small intestine. By using this strategy, we characterized in vivo the role of ERR␣ in the regulation of intestinal energy metabolism and dietary lipid absorption. Furthermore, this approach allowed us to identify apoA-IV as a novel ERR␣ target gene involved in the absorption and transport of triglycerides.
We previously reported that ERR␣ was expressed early in intestinal mucosa of the developing mice (9). To further explore FIG. 3. ApoA-IV gene expression and dietary fat absorption in mice deficient for ERR␣. A, wild-type and ERR␣ Ϫ/Ϫ mice were fasted overnight and fed by gavages of 200 l of olive oil or normal saline. Mice were sacrificed after 2 h, RNA was extracted from jejunal tissues and submitted to reverse transcription. cDNA was used as a template to measure mouse apoA-IV expression by quantitative reverse transcription-PCR. Mice groups were: wild-type fasting (n ϭ 10), ERR␣ Ϫ/Ϫ fasting (n ϭ 9), wild-type lipid (n ϭ 13), and ERR␣ Ϫ/Ϫ lipid (n ϭ 10). Results are mean fold change over wild-type mice treated with saline Ϯ S.E., * p Ͻ 0.05. B, ten-day-old pups and adult mice on regular diet were sacrificed and lipids were extracted from feces as described under "Materials and Methods." Sample number were n ϭ 10/group for adult mice, 12/group for wild-type pups, and 11/group for ERR␣ Ϫ/Ϫ pups (feces from 3 pups were pooled together). The results are expressed as the percentage of fat comprising the total stool dry weight Ϯ S.E., * p Ͻ 0.05.
FIG. 4. ERR␣ activates the apoA-IV gene through the apoC-III enhancer. Caco-2/15 cells were plated in 6-well plates and cotransfected with the apoA-IV promoter luciferase reporter constructs (0.4 g per well) and the mouse VP16-ERR␣ chimeric receptor (0.1 g), ERR␣ (0.1 g) with or without PCG-1␣ (0.1 g) expression vectors. pCMX␤gal (0.1 g) was used as an internal control. Cells were harvested 48 h after transfection and assayed for luciferase and ␤-galactosidase activity. ApoA-IV promoter-dependent luciferase activities were determined as fold activation over the individual luciferase activity with vector alone. Subsequent cotransfections were assayed with truncated apoA-IV promoter reporters and segments of the apoA-IV promoter cloned in front of the minimal thymidine kinase promoter. the role of ERR␣ in the intestine, we examined the distribution of ERR␣ throughout the digestive tract. We noted that ERR␣ gene expression was restricted to epithelial cells of intestinal mucosa and that the small intestine displayed higher levels of ERR␣ compared with the stomach and colon. These data suggest a role for ERR␣ in dietary nutrient absorption. Furthermore, the high levels of ERR␣ in neonatal and weaning mice, where most of the energy derives from fat, support a function for ERR␣ in the regulation of intestinal lipid metabolism. This is in agreement with a previous report showing that ERR␣ regulates the medium chain acyl CoA dehydrogenase gene (MCAD) involved in fatty acid ␤-oxidation (9,13,14).
A role for ERR␣ in metabolic control was also supported by the observation that ERR␣ Ϫ/Ϫ mice display a lean phenotype associated with reduced lipogenesis and high fat diet-induced obesity (19). However, the fact that ERR␣-deficient mice did not display lower food consumption or higher energy expenditure over a 4-day period pointed out to a possible defect in intestinal processing of dietary nutrients, taking into account that the intestine is the first organ in the chain of energy metabolism. We then evaluated the gut morphology in ERR␣null mice and did not find any gross or histological abnormality. However, gene expression profiling revealed a subset of genes consistently down-regulated in the intestine from ERR␣deficient mice, without regard to gender or genetic background. In particular, ERR␣ modulated the expression of genes involved in oxidative phosphorylation such as cytochrome c oxidase 5a and 7c, NADH dehydrogenase and ATP synthase. Furthermore, functional evaluation indicated that epithelial cells from mice lacking ERR␣ present a lower capacity for ␤-oxidation than cells expressing ERR␣. Since ␤-oxidation is directly coupled to oxidative phosphorylation, our findings are in favor of the in vivo role of ERR␣ in optimal oxidative phosphorylation and energy production. This group of genes has also been found to be induced by PGC-1␣ in SAOS2 cells (17) and C2C12 mouse myoblasts (18). Interestingly, ERR␣ was required for PGC-1␣ induction of oxidative phosphorylation genes since treatment of SAOS2 cells with siRNA specific for ERR␣ reduced the ability of PGC-1␣ to induce these genes (17). Furthermore, treatment with an ERR␣ synthetic inhibitor abolished the PGC-1␣-mediated regulation of ATP synthase and cytochrome c oxidase, as well as cellular respiration (18). This interrelation between PGC-1␣ and ERR␣ is consistent with our previous study showing that PGC-1␣ increases ERR␣ levels via a polymorphic regulatory site (16). However, lower energy catabolism in the intestine cannot explain in itself why the ERR␣-null mice are leaner. In fact, the mice could be fatter as the ERR␣-PCG-1␣ tandem is shut down, in reverse to the athletic phenotype of the muscle specific and global transgenic PGC-1␣ and ␤ mice (10, 33). One possible explanation is that energy is not available to the ERR␣-null mice.
Given the fact that the gut produces energy from food available to the rest of the body, it was of physiological importance that ERR␣-deficient pups malabsorb dietary fats, as evidenced by steatorrhea. This was not observed in adult ERR␣ Ϫ/Ϫ mice, possibly because of compensatory mechanisms or to the large absorptive reserves of the adult small intestine. Nevertheless, fat malabsorption could explain the lower gain weight of ERR␣deficient pups compared with wild-type littermates. In addition, subtle defect in dietary lipid absorption, including faulty apoB synthesis, reduced microsomal triglyceride transfer protein activity or alterations in the Sar1 GTPase of COPII (34,35) could also contribute to resistance to high fat diet-induced obesity in adult ERR␣-deficient mice.
Indeed, it is striking that the second group of genes downregulated in the intestine from adult ERR␣ Ϫ/Ϫ mice play a role in dietary fat digestion and absorption. PLRP2 is part of the lipase gene family derived from common ancestral gene and is the major colipase-dependent pancreatic lipase in suckling mice (36). In addition to the pancreas, PLRP2 is well expressed in the small intestine and its role in triglyceride, galactolipids and phospholipids digestion is reflected by the dietary fat malabsorption in suckling PLRP2-deficient mice (24). Furthermore, FABP-L (liver) and FABP-I (intestine) are intracellular lipid-binding proteins found to be expressed at lower levels in the intestine from ERR␣ Ϫ/Ϫ mice compared with wild-type mice. It is generally accepted that FABPs facilitates the intracellular transport and metabolism of fatty acids. In addition, FABPs have been postulated to protect intracellular polyunsaturated fatty acids against peroxidation (37), as well as to regulate gene expression by delivering fatty acids to the nucleus, where they act as ligands for PPARs (38). In turn, L-FABP gene expression appears to be regulated by some fatty acids and PPAR agonists (39,40). The role of I-FABP in fatty acid uptake and metabolism is unclear as I-FABP overexpression in Caco-2/15 results in inhibition of fatty acid incorporation (41) and I-FABP knockout mice display gender-specific alterations in body weight (42). Finally, apoA-IV, which is also down-expressed in ERR␣-null mice, is a major protein component of intestinal triacylglycerol-rich lipoproteins such as chylomicrons and very low density lipoproteins. Although its precise function remains unclear, apoA-IV has been proposed to play a role in various processes: lipid absorption, transport, and metabolism; food consumption, gastric acid secretion and motility; and protection against lipoprotein oxidation and atherosclerosis (43). In accord with our data, apoA-IV acted as a post-prancial satiety signal, which caused anorectic effect mediated via the central nervous system (44). Furthermore, overexpression of apoA-IV enhanced lipid transport in intestinal cells from newborn swine (45). The physiological relevance of this suppression of food intake has yet to be confirmed since apoA-IV knockout mice had normal lipid absorption, weight gain, and food consumption (46). In the ERR␣ Ϫ/Ϫ mice, the observed down-regulation of intestinal apoA-IV could contribute to impaired fat absorption or be the consequence of lipid malabsorption. Indeed, the formation and secretion of chylomicrons stimulate the synthesis of apoA-IV and human pathological states of fat malabsorption are associated with lower levels of apoA-IV (reviewed in 43). In the current investigation, we observed that mice deficient for ERR␣ showed lower intestinal apoA-IV gene expression in the fasting condition, but normal apoA-IV induction by fat feeding. This suggests that ERR␣ does not mediate apoA-IV response to feeding but is essential for maintenance of basal apoA-IV intestinal expression.
ApoA-IV expression is limited to intestinal and hepatic cells. In the intestine, the transcriptional regulation of the apoA-IV FIG. 5. ERR␣ binds to the human apoC-III enhancer region in vivo. Caco-2 cells were transfected with human ERR␣ followed by cross-linking and immunoprecipitation of DNA/protein complexes using either preimmune serum or human ERR␣ antibody (Ab). Quantitative PCR amplifications of enriched chromatin were performed using primer pairs specific to either the apoC-III promoter, apoC-III enhancer, the apoC-III/A-IV intergenic region or an unrelated region located Ϫ3 kb away from the ERR␣ autoregulation site (16). Input lane i shows DNA that was PCR-amplified from extracts before immunoprecipitation. gene involves synergism between a proximal promoter region and a distant enhancer located in the upstream promoter region of the apoC-III gene. Our experiments showed that ERR␣ interacts with the mouse apoC-III enhancer when fused with VP16 or in the presence of the PGC-1␣ coactivator. Furthermore, we used a ChIP assay in the Caco-2/15 enterocytic cells to demonstrate that ERR␣ binds in vivo to the corresponding human enhancer region. Molecular interspecies preservation implies a physiological importance of this transcriptional control pathway in the intestine. Many transcription factors (Sp1, ATF-2, NFB) and nuclear receptors (COUPTF-I, COUPTF-II, RAR␣, T 3 R, HNF-4␣) have been shown in vitro to interact with the human apoC-III enhancer region (47). Furthermore, the nuclear farnesoid X receptor (FXR) has been found to be a negative regulator of both human and mice hepatic apoC-III expression, via a FXR response element located in the apoC-III enhancer (48). In addition, a recent study has demonstrate that liver X receptors (LXRs) modulate both human and mouse apoA-IV gene expression in the liver but not in the intestine (49). Interestingly, the direct regulation by LXRs was mediated by different interaction sites in humans and mice, respectively in the apoA-IV promoter and in the apoC-III enhancer.
In conclusion, by using the ERR␣ knockout mice in combination with intestinal microarray analysis, we identified ERR␣-modulated genes involved in two main pathways: oxidative phosphorylation and dietary lipid digestion and absorption. The first is in agreement with the known function of ERR␣ in energy expenditure when coactivated by PGC-1␣, and the second allowed us to reveal a function for ERR␣ in energy uptake from fat, a crucial first step in energy balance. Furthermore, we characterized the apoA-IV gene as a novel direct ERR␣ target gene. At the molecular level, ERR␣ interacted with the apoC-III enhancer region in both mice and human. The role of ERR␣ in dietary lipid handling should therefore be taken into consideration during the development of ERR␣ pharmacologic ligands (50).