Elovl3: a model gene to dissect homeostatic links between the circadian clock and nutritional status.

The ELOVL3 protein is a very long-chain fatty acid elongase found in liver, skin, and brown adipose tissues. Circadian expression of the Elovl3 gene in the liver is perturbed in mutant CLOCK mice but persists in mice with severe hepatic dysfunction. A reliance on an intact clock, combined with the refractoriness to liver decompensation and the finding of a robust sexually dimorphic pattern of expression, evince a particularly complex mode of transcriptional control. The Elovl3 gene upstream region was repressed by RevErbα and activated by sterol-regulatory element binding protein-1 (SREBP1) transcription factors. We propose that the temporal coordination of RevErbα and SREBP1 activities integrates clock and nutrition signals to drive a subset of oscillatory transcripts in the liver. Proteolytic activation of SREBP1 is circadian in the liver, and because the cycle of SREBP1 activation was reversed after restricting meals to the inactive phase of the day, this factor could serve as an acute sensor of nutritional state. SREBP1 regulates many known lipogenic and cholesterogenic circadian genes; hence, our results could explain how feeding can override brain-derived entraining signals in the liver. This mechanism would permit a rapid adjustment in the sequence of key aspects of the absorptive and postabsorptive phases in the liver.

Supplementary key words fatty acid & elongase & sterol-regulatory element binding protein target genes & restricted feeding & daily rhythm & phase shift ELOVL3 is a 30 kDa membrane-bound protein expressed only in liver and brown adipose tissues (significantly lower levels can also be detected in skin) that is strongly induced during perinatal development, consumption of a high-fat diet, or exposure to cold stress (1). Elovl3 is a member of a recently described mammalian gene family involved in, according to current evidence, the elongation of fatty acids to produce very long-chain fatty acids (VLCFAs) and, secondarily, sphingolipids (2) in a tissue-specific manner (3). The dramatic induction observed in Elovl3 expression in brown adipose tissue during cold stimulation and the ensuing brown fat recruitment links this protein to the thermogenic process (3).
The regulation of the Elovl3 gene presents several interesting features. First, steady-state levels of Elovl3 mRNA follow a robust circadian profile in the liver, according to a recent transcriptome analysis (4). This rhythm, which appears to be perturbed in the CLOCK mutant mouse (4), displays a late-night acrophase, suggesting that if there is a link to the circadian BMAL/CLOCK pathway, it is most likely indirect, because BMAL/CLOCK activity peaks during the second half of the light phase (5). Therefore, analysis of the Elovl3 promoter could lead to the identification of downstream transcriptional programs linked to the core mechanism of the clock in the liver. Second, thermogenesis in the brown adipose tissue can be acutely modulated by photic signals acting via the suprachiasmatic nucleus (6,7). Because of its presumed role in thermogenesis, Elovl3 represents a potentially useful link between clock output from the suprachiasmatic nucleus and the regulation of physiological variables such as circadian changes in core body temperature. Finally, a day/night liver transcriptome analysis of the transgenic "fatless" mouse model (A-ZIP/F-1) showed that despite the known metabolic disruption in the liver, the daily rhythm in Elovl3 gene expression remained virtually unaffected. In fact, the nocturnal induction of the Elovl3 gene displayed the largest amplitude of day/night rhythmic expression between the two time points selected, among all the genes present in the microarray. This observation suggested that at least some rhythmic BMAL/CLOCK activity is present in A-ZIP/F-1 mice and that Elovl3 represents a potential downstream clock-controlled gene.
This hypothesis prompted us to look more closely at the regulation of the Elovl3 gene in normal mice. In this study, we found that Elovl3 is a tightly regulated sexspecific gene, with virtually undetectable expression in the female liver. In contrast, a uniform low level of Elovl3 gene expression was observed in both genders in the interscapular brown adipose tissue (iBAT). In addition, we found that the Elovl3 gene promoter was not activated by BMAL/ CLOCK in transfected cells, in spite of the presence of a perfect E-box. This result is consistent with the notion that its deregulation in the CLOCK mouse reflects an indirect connection to the core clock mechanism. In apparent contrast to a previous report, we found that expression of the Elovl3 gene was arrhythmic in the liver of CLOCK mutant mice.
Our results suggest the RevErba and the sterolregulatory element binding protein-1 (SREBP1) transcription factors as likely mediators of circadian Elovl3 transcriptional activity in the liver. Upon closer examination, we observed that the rhythmic Elovl3 transcript level in the liver was modulated by the feeding schedule, which was consistent with changes in the rhythm of proteolytic activation of SREBP1 observed under identical conditions. Therefore, the activity of the SREBP1 factor likely functions either to integrate or to dissociate the central circadian cues and the oscillating rhythms in the liver in response to the availability of food, a time cue significantly less predictable than the photoperiod.

Animal care and experimental protocols
Animals (C57BL/6 or CLOCK heterozygous and homozygous mice) (8) were entrained to a 12 h/12 h light/dark schedule [lights on at 6 AM; this time is referred to as Zeitgeber time (ZT) 0] for 2 weeks before experiments were carried out. The level of photo entrainment in mice was assessed by monitoring the locomotor activity using the Vital-View system software (Mini-Mitter Co.). Experimental procedures involving the A-ZIP/F-1 fatless strain of transgenic mice and all related supplemental data can be viewed at the following URL: http://neuroscience.nih. gov/Baler/CIG30ms/supplementaldata.pdf. (In accordance with the Mouse Nomenclature Committee, the assigning symbol for the mouse gene Cig30 is changed to Elovl3.) For restricted feeding experiments, C57BL/6 male mice were entrained to a 12 h/12 h photoperiod for 2 weeks. During the third week, half of the animals were allowed to feed only between ZT5 and ZT9. Animals were anesthetized by CO 2 inhalation and immediately decapitated. Liver samples were rapidly removed and quickfrozen in a 1.5 ml microtube in solid CO 2 . All experiments were performed in accordance with the Public Health Service Policy on Humane Care and Use of Laboratory Animals, the Guide for the Care and Use of Laboratory Animals, and all pertinent Animal Welfare Act regulations, according to experimental protocols that were approved by the Animal Care and Use Committee and met the guidelines of the National Institutes of Health. No procedures performed during the course of this study caused more than slight momentary pain or distress.

Northern blot analysis
Total cellular RNA was extracted from frozen pieces of tissue (in duplicate) using the Trizol reagent (Invitrogen-Life Technologies, Carlsbad, CA), fractionated (8-20 mg/lane) on a 1% formaldehyde-agarose gel, and examined by Northern blot analysis. Probes were labeled by random-priming with [a-32 P] dCTP (Amersham Pharmacia Biotech, Piscataway, NJ) using the MegaPrime kit (Amersham Pharmacia Biotech). A rat Elovl3 probe was obtained through a RT-PCR amplification procedure. Briefly, total rat liver RNA (1 mg) was reverse-transcribed using oligo-dT primers (Invitrogen-Life Technologies), and an 857 bp Elovl3 cDNA product was then amplified by PCR using the following primer pair (bracketing the fragment between positions 8,384 and 10,208 in GenBank AF054504) (1): forward (in exon 5), 59-GGTAAGCTCATTTCTCATAGTGGTCG-39; and reverse (in the 39 untranslated region), 59-GGGTGACATTTCTTGGTTGTCA-TAGCTTCC-39. Northern blot membranes were stripped (boiling in 0.1% SDS and 0.1% sodium chloride, sodium citrate-trisodium salt stock solution, twice for 15 min) and reprobed with a glyceraldehyde-3-phosphate dehydrogenase probe. Densitometric analysis of mRNA levels was performed using ImageQuant TM software (version 3.0; Amersham Pharmacia Biotech).

Cell lines, plasmid vectors, and recombinant DNA protocols
The mouse hepatocyte line AML12 (9) (American Tissue Culture Collection; CRL-2254) was cultured in a 1:1 mixture of DMEM and Ham's F12 medium containing insulin (5 mg/ml), transferrin (5 mg/ml), selenium (5 mg/ml), dexamethasone (80 ng/ml), and 10% fetal bovine serum. NIH-3T3 fibroblasts (CRL 1658) were cultured in DMEM (4 mM L-glutamine, 1.5 g/l sodium bicarbonate, and 4.5 g/l glucose) and 10% bovine calf serum (HyClone, Logan, UT). Cell cultures were maintained at 378C in a humidified atmosphere of 5% CO 2 . A mouse Elovl3 promoter reporter vector was constructed by splicing a 1.16 kb genomic region (located upstream and immediately adjacent to the mouse Elovl3 translation start site) between the NheI and BglII sites in the pGL3Basic firefly luciferase vector (Promega, Madison, WI). To generate a cytomegalovirus-driven RevErba expression vector, a fragment derived from the human RevErba sequence (10) was generated by RT-PCR using the following primers (59 to 39): TGAAGACATGACGACCCTGGACTCCAA-CAACAACACAGGT and GGTCACTGGGCGTCCACCCGGAAG-GACAGCAGCTTC. The resulting 1.8 kb coding sequence was subcloned into a NotI/XbaI-linearized pCDNA3.1 (Invitrogen) and fully sequenced. The CS10a, CS10c, and CS2 expression vectors for the N-terminal truncation derivatives of SREBP1a, -1c, and -2, respectively (11), were kindly provided by Dr. Tim F. Osborne (University of California). Human BMAL and mouse CLOCK expression vectors were a gift from Drs. Nick Gekakis and Charles Weitz (Harvard University). The vasopressin/LUC reporter used in this study has been described previously (12).

Transient transfection assays
Transfections were performed using Lipofectamine/Plus (Invitrogen)-DNA lipoplexes as described previously (12). Elovl3 transactivation patterns were analyzed in the nontransformed mouse hepatocyte cell line AML12 and in the NIH-3T3 mouse embryo fibroblast line. Results are representative of at least five independent experiments performed in triplicate. Equivalent transfection efficiency was assessed by cotransfection with a Renilla luciferase reporter gene driven by a cytomegalovirus promoter. Statistical analysis was performed using Student's t-test for unpaired samples. The intragroup variability remained consistently at or below the 10% level.

Western blot analysis
Duplicate liver samples were harvested from groups of four mice every 3 h during a 12 h/12 h light/dark cycle and placed immediately in dry ice. For whole cell protein extraction 10 ml of ice-cold buffer C (20 mM HEPES, pH 7.9, 1.5 mM MgCl 2 , 0.42 M NaCl, 0.2 mM EDTA, 1 mg/ml aprotinin, 1 mg/ml leupeptin, 1 mM sodium fluoride, 5 mM sodium orthovanadate, 0.5 mM phenylmethylsulfonyl fluoride, 0.5 mM dithiothreitol, and 25% glycerol) was added for every 1 mg of liver tissue. The tissue was then homogenized on ice with 20 strokes of a tight-fitting, 1.5 ml microtube pestle, vortexed, and refrozen on dry ice for 10 min. Lysates were then incubated in ice water for 15 min and centrifuged for 15 min at 12,000 rpm (48C). For Western blot analysis, z40 mg of soluble protein from whole liver lysates was electrophoresed through a 4-12% Tris-Glycine-SDS Nu-Page cassette (Invitrogen) and transferred onto an Immobilon P (Millipore, Bedford, MA) polyvinylidene difluoride membrane according to the manufacturer's recommendations. Membrane-bound proteins were investigated with anti-human SREBP1a 301-407 (13) (BD Biosciences, San Diego, CA) and finally with a mouse monoclonal antibody against the rat 14.3.3 e protein (14), which was kindly provided by Dr. David Klein (National Institute of Child Health and Human Development). After incubation with a HRPconjugated secondary antibody ( Jackson ImmunoResearch Laboratories, West Grove, PA), luminescence was quantified using an Image Station 440 CF (Kodak, Rochester, NY).

Rhythmicity of Elovl3 gene expression in the mouse liver persists in transgenic fatless mice
In a recent transcriptome analysis comparing day/night rhythms in gene expression between wild-type and A-ZIP/ F-1 mice (http://neuroscience.nih.gov/Baler/CIG30ms/ supplementaldata.pdf), the Elovl3 gene displayed the most robust amplitude among only 11 transcripts whose circadian oscillations remained largely unaffected by the severe liver decompensation characteristic of this transgenic model (Table 1) (15,16). The level of Elovl3 expression had been reported previously to increase by z4.8-fold in the CLOCK mutant mouse at ZT8 (4). This is interesting because the late-night peak in Elovl3 mRNA develops in antiphase with the diurnal peak in BMAL/CLOCK activity in the liver (5). Together with the refractory nature of the Elovl3 rhythm to the perturbation in hepatic physiology in A-ZIP/F-1 mice, these observations suggest that the rhythm in Elovl3 transcription could provide an interesting model for the study of the homeostatic integration between the core clock machinery and downstream gene oscillations in the liver.
Rhythmic, sex-specific, and tissue-specific expression of the mouse Elovl3 gene To analyze the pattern of Elovl3 gene expression in normal animals, groups of male and female C57BL/6 mice were entrained to a 12 h/12 h light/dark photoperiod for 2 weeks before being euthanized every 4 h for the collection of liver, iBAT, and muscle tissue samples. Northern blot analysis (Fig. 1A) confirmed the rhythmic expression in liver that had been observed previously in a liver transcriptome analysis performed on cDNA microarrays (4). Unexpectedly, the circadian expression of Elovl3 transcription in the liver was sex-specific, because no signal was detected in the livers of female mice (Fig. 1A, compare left and right panels). Interestingly, no such difference was observed in the iBAT, which showed low and virtually constant levels of Elovl3 mRNA in both sexes (see supplementary data). As expected, no Elovl3 expression was detected in skeletal muscle of either gender (data not shown). The rhythmic, tissue-specific, and sexually dimorphic pattern of Elovl3 gene expression suggests interesting links between the complex regulation of this gene and the well-established differences in the control of thermogenesis (17,18) and liver gene expression (19) between genders in the rodent species. Genes are listed from strongly nocturnal to strongly diurnal. This list of 11 genes represents the small group of overlapping genes with rhythmic expression that remained unchanged between the livers of wild-type and fatless (A-ZIP/F) mice (see Materials and Methods and supplementary data for more details).

Effect of the CLOCK mutation on rhythmic Elovl3 gene expression in the liver
It is reasonable to propose that daily oscillations in the level of Elovl3 mRNA in the liver are somehow linked to the circadian transcriptional activity of the BMAL/CLOCK heterodimeric complex, because Elovl3 mRNA levels have been reported to increase by z4.8-fold in the livers of CLOCK mutant mice at ZT8 (4,8). To independently corroborate that result at time points that are more relevant to Elovl3 gene rhythmicity, heterozygote and homozygote CLOCK mutant male mice were photo-entrained for 2 weeks and euthanized at ZT3, ZT8, and ZT23 to assess the level of liver Elovl3 mRNA by Northern blot analysis (Fig. 1B). We found that Elovl3 mRNA was marginally higher in the livers of Clock 2/2 mice at ZT8; more importantly, the daily rhythm in Elovl3 gene expression was essentially abolished in Clock 2/2 mice (Fig. 1B, right  panel). This result represents a significant refinement of previously available data (4). The present study suggests a more complex effect of the CLOCK mutation upon rhythmic Elovl3 gene expression in the rodent liver, which should be reinterpreted in light of the promoter and functional studies described below.
Isolation and functional characterization of the mouse Elovl3 gene promoter The promoter region of the mouse Elovl3 gene was isolated via high-fidelity genomic PCR with primers de-signed using available genomic sequence information (20). To identify putative DNA consensus regulatory sites, we scanned the z1.1 kb fragment of the Elovl3 region upstream of the translation start site [including a 171 bp long 59 untranslated region (boldface in Fig. 2A, B)] for significant matches to entries in the TRANSFAC database (21). This procedure revealed, most significantly, a potential RevErba binding element at position 2846 (22, 23), a putative signal transducer and activator of transcription (STAT) (24) recognition site at position 2726, two sterol response elements at 2776 and 2426, a perfect activator protein-1 site at position 2326, a likely TATA box z27 bp upstream of the reported transcription start point, and a perfect E-box, within the 59 untranslated region at position 121 relative to the transcription start point.
Next, we created a preliminary functional map of possible regulators of Elovl3 gene expression by testing the performance of an Elovl3/LUC promoter reporter construct during the transient coexpression of selected transcription factors in AML12 cells, a murine hepatocyte cell line (9). The factors tested were chosen after the identification of suitable potential cognate sites along the Elovl3 regulatory region (Fig. 2B) and because of their known involvement in circadian and lipogenesis-related transcriptional regulation.
To ascertain whether the links between the clock and Elovl3 gene expression operate directly or indirectly, we first evaluated the ability of the Elovl3 gene upstream region to respond to the action of the circadian heterodimer BMAL/CLOCK in transient cotransfection assays. The 1.1 kb Elovl3 promoter fragment was not transactivated after overexpression of the circadian BMAL/ CLOCK factor (Fig. 2C, upper panel), despite the presence of one perfect and three noncanonical E-boxes within the Elovl3 promoter sequence and the robust transactivation of the control vasopressin gene promoter (data not shown), a well-established BMAL/CLOCK-responsive region (12,25). In fact, we have consistently observed a modest repression of Elovl3 promoter-driven luciferase basal activity upon forced expression of BMAL/CLOCK. This behavior indicates that the E-box sites found in the mouse Elovl3 promoter region used are not likely to be direct positive targets of the clock. This result was expected based on the fact that circadian expression of Elovl3 in the male liver peaks at the end of the dark period, in relative antiphase to the reported peak activity time of BMAL/ CLOCK at ZT8-ZT12 in peripheral tissues (5,26). The Elovl3 promoter, however, must be strongly connected to the central clock machinery, because its expression is obviously affected by a mutation in the CLOCK gene (4), albeit in a complex manner (Fig. 1B).
RevErba, another of the likely candidates to provide clock-dependent information to the pathways controlling Elovl3 gene expression, is a nuclear hormone receptorrelated protein that has a potential binding site along the Elovl3 promoter (Fig. 2B). Additionally, RevErba plays a role in the positive limb of the interlocking loops that maintain oscillations in the central core of the mammalian clock (27), it is known to repress the transcription of target genes (28), and its early peak of activity in peripheral oscillators (27,29) is suitably in phase with the trough of Elovl3 mRNA expression in the liver. Interestingly, we found that basal Elovl3 promoter activity was significantly reduced after overexpression of RevErba (Fig. 2C, second  panel from top). This effect is consistent with a possible role for this factor in maintaining the low levels of Elovl3 mRNA in the male liver during the day and provides the first possible connection to the clock machinery.
In an attempt to identify the factor(s) responsible for the activation of the Elovl3 gene, we next tested the performance of the Elovl3 promoter fragment after overexpression of the mature forms of SREBP1a, -1c, and -2. These factors, which also possess putative binding sites along the Elovl3 promoter (Fig. 2B), have been shown to control the expression of a Putative cis-acting elements identified are highlighted and boxed (A) and also depicted schematically along the promoter (B); their positions, as indicated in the text, are relative to the transcription start point (11). AP-1, activator protein-1; SRE, sterol-regulatory element; STAT, signal transducer and activator of transcription. C: The performance of a mouse Elovl3/LUC promoter-reporter construct was assessed after its cotransfection into AML12 cells along with the indicated vectors designed to overexpress the BMAL/CLOCK (top), RevErba (second), and sterol-regulatory element binding protein-1a (SREBP1a; third) and SREBP1c (bottom) transcription factors. Error bars represent 6 SD of five independent experiments performed in triplicate. The asterisks indicate statistical significance of transcription factors overexpression versus basal Elovl3 promoter activity at P , 0.05 and P , 0.01, respectively. Elovl3 promoter-driven firefly luciferase activity was normalized to the activity generated by a cytomegalovirus promoter-driven Renilla luciferase control reporter. Similar results were observed when the experiments were performed in NIH-3T3 cells. large number of lipogenic and cholesterogenic genes (11,(30)(31)(32). We found that SREBP1a and -1c had robust and positive effects on Elovl3 promoter-driven firefly luciferase activity in AML12 cells (Fig. 2C, bottom panels). However, no significant effect was observed after SREBP2 expression (data not shown). The observation of different potencies between the two SREBP proteins is consistent with previous studies showing that the SREBP1 and -2 proteins display differential transactivation properties (33) and effects on gene regulation (31).
SREBPs are unusual basic helix-loop-helix proteins in that they can recognize not only a sterol regulatory element, such as the one found in the low density lipoprotein receptor gene promoter (32,34) and related sequences (11), but also the palindromic E-box motif (30). Mutation of the Elovl3 perfect E-box at position 121, however, had no negative effect on SREBP1-mediated stimulation of luciferase activity. On the contrary, E-box mutation resulted in a significantly more robust transactivation (higher stimulation index) of Elovl3 by SREBP1 (data not shown). Thus, a more detailed analysis of this regulatory region will be needed to identify which of the several potential SREBP binding sites along the Elovl3 promoter fragment used in this study is responsible for mediating this response.
Together, the responses of the mouse Elovl3 promoter fragment to BMAL/CLOCK, RevErba, and SREBP1 suggest that a complex network of factors can potentially cooperate to establish the observed temporal pattern of Elovl3 gene expression in the rodent liver. It is apparent that SREBP1 may play an important role in driving the robust induction of Elovl3 transcription in the liver toward the end of the night. The positive action of SREBP1 could reflect a constitutive default mechanism that becomes manifest when RevErba repression disappears. Alternatively, the SREBP1 rhythmic and antiphasic pathway (35) could have been designed to amplify the response during the dark (active) phase of the cycle.

Restricted feeding during the day shifts the phases of both SREBP1 activation and Elovl3 gene expression
To test the hypothesis that SREBP1 activation plays a role in the rhythmic activation of the Elovl3 gene in the liver in vivo, we subjected mice to a restricted feeding paradigm. Providing animals with restricted access to food only during the day (the inactive phase of the cycle for mice) has been found to reverse the phase of many cycling transcripts (36,37). We found that restricted feeding during the day had two noticeable effects on the pattern of SREBP1 activation. First, it was evident that the relatively shallow rhythm of activation seen in mice fed ad libitum had consolidated around the time of feeding; this was accompanied by a significant increase in the amplitude of the response at peak time. Second, and more importantly, the peak of SREBP1 activation was shifted by z9 h relative to the peak detected in mice fed ad libitum. An equivalent shift was also observed in the rhythmic expression of Elovl3 (Fig. 3). This result is consistent with a causal relationship between the nutritional activation of the SREBP1 pathway and the setting of the phase of the Elovl3, and, likely, many other gene cycles.

Elovl3: a model of complex gene regulation
Elovl3 has been identified several times during independent investigations of differentially expressed transcripts. First detected as a gene induced by cold stress in brown adipose tissue and termed cold-inducible glycoprotein of 30 kDa (Cig30) (38), Elovl3 later reemerged as a result of its strong predawn induction in the liver (4). Finally, our own analysis of the transcriptome in the liver of the fatless mouse (Table 1 and supplementary data) singled out Elovl3 as a gene under the control of particularly robust and complex transcriptional mechanisms in the rodent liver.
Previous studies have classified the Elovl3 gene product as a member of the fatty acid chain elongase family, likely involved in thermogenesis and/or brown fat recruitment (i.e., brown adipose tissue hypertrophy) (1). Many biological processes depend on VLCFAs containing .20 carbon atoms. The mouse ELOVL3 protein is very similar to yeast ELO2 and ELO3 gene products. In yeast, disruption of either ELO2 or ELO3 leads to a broad range of defects, encompassing altered expression of the plasma membrane ATPase gene and decreased glucose uptake capacity (39), modified phospholipid composition (40), and resistance to inhibitors of sterol synthesis (41), among others.
This study places the Elovl3 gene within a large family of sexually dimorphic genes in the rodent liver, a finding that raises many exciting questions that should be explored further. For example, many genes in this set (42,43) have been shown to be modulated by androgenic signaling working through sex-specific growth hormone secretion profiles (43,44). This cascade, beyond the growth hormone receptor, appears to recruit STAT5 transcriptional activity, which closely correlates with the sexually dimorphic patterns of growth hormone pulsatile secretion (44). It is possible that the putative STAT5 sites along the Elovl3 promoter contribute to the dramatic sex differences in Elovl3 expression we have discovered (Fig. 1A). In addition, and considering current evidence about the effect of a sexually dimorphic growth hormone secretory pattern on SREBP1c and its target genes in liver (45), it will be of interest to investigate the molecular basis for the failure of this activation strategy in the female liver and compare hypophysectomized, as well as ovariectomized, with control animals in future experiments.
In light of its sexually dimorphic pattern of expression and its robust circadian rhythm in the liver, it is tempting to propose the Elovl3 gene as a fertile model for studying the integration of major physiological systems, such as energy homeostasis, sex, and circadian timing at the level of transcriptional control. The sharp increase in the levels of Elovl3 mRNA in the livers of male mice at the end of the night already suggested that the transcription factor BMAL/CLOCK was unlikely to be involved directly in this transcriptional rhythm. In fact, such a late-night surge of a circadian gene in the liver, which is a strong peripheral oscillator, was more consistent with a RevErba-like activity. RevErba does not possess an Activation Function-2 domain and behaves as a powerful transcriptional repressor that interacts strongly with the Nuclear Receptor Corepressor (46). RevErba is active during the day, when it represses the activity of target promoters. BMAL itself is driven by a RevErba-sensitive promoter, and its release from RevErba repression constitutes the basis for its antiphasic circadian pattern of expression (27). Together with the presence of a putative RevErba binding site along the Elovl3 promoter, it was reasonable to hypothesize that a similar RevErba derepression could be at least partially responsible for the rhythmic increase in Elovl3 mRNA levels in the liver at night. Thus, the loss of RevErba rhythmicity in CLOCK null mice (27) could contribute to the observed disruption of Elovl3 rhythmicity in these mice (Fig. 1B). Our inability to transactivate the Elovl3 promoter after BMAL/CLOCK overexpression, coupled with the significant reduction in its basal level of expression after RevErba expression (Fig. 2C, top and middle panels), is consistent with this hypothesis. Furthermore, the slight repression of Elovl3 promoter-driven expression after BMAL/CLOCK overexpression suggests the induction of a suppressive pathway that might involve the RevErba protein. Therefore, we propose that Elovl3 rhythmicity might be at least partially maintained by the negative action of RevErba during the day, effectively linking the Elovl3 rhythm to the upstream status of the clock (Fig. 4). Moreover, an interruption of the repressive RevErba cycles in CLOCK mutant mice could explain the transiently higher levels of Elovl3 mRNA in their livers at ZT8 (4).
The responsiveness of the Elovl3 promoter to the action of SREBP1 is of interest for several reasons. There is overwhelming evidence that SREBPs play an essential role in both cholesterol and fatty acid metabolism (33,47). SREBPs might be differentially regulated at multiple levels by sterol depletion (48), but they appear to act in concert to maintain the right composition of lipids in cell membranes (31). Our results suggest that Elovl3 should be added to the growing list of genes targeted by SREBPs (33). This list includes FAS, low density lipoprotein receptor, HMG-CoA reductase, acetyl-coenzyme A carboxylase (ACC), and lipoprotein lipase. Interestingly, many of the genes in this list are also circadian in the liver, and some (e.g., ACC1, FAS, and HMG-CoA reductase) are in phase with the daily Elovl3 cycle (4,35). Together, these observations suggest that SREBP1 might play an important role in the orchestration of temporal gene expression, downstream of the circadian clock, in the liver. This hypothesis can now be tested in vivo in suitable genetically modified animals.
It is of interest that a perfect E-box (CACGTG) appears not to be the target of SREBP1 action in Elovl3 or in other target genes. This site, however, could play a role in preventing SREBP1-mediated activation at inappropriate phases of the cycle, because an E-box mutation resulted in a significantly more robust transactivation (higher stimulation index) of Elovl3 by SREBP1 (data not shown).

SREBP-CLOCK connections
The results of this study also give rise to the question of whether or not the BMAL/CLOCK pathway of transcriptional regulation can feed into the SREBP1 activation system in peripheral oscillators. The SREBP1 gene can be induced by differentiation (49), diets rich in PUFAs (48,50), or refeeding after fasting (51). SREBP1 mRNA (UniGene Cluster Mm.214958) also displays a circadian rhythm of expression in the rodent liver (4), as well as SREBP1 active protein, yet SREBP1 precursor remains at a constant level throughout the day (35) (Fig. 3C). The SREBP1 precursor is a cell membrane-anchored protein that becomes activated through proteolytic cleavage of the N terminus upon sterol depletion (52). Thus, the nucleusbound and transcriptionally active SREBP1 fragments can translate extremely low sterol levels (e.g., cholesterol) into specific patterns of gene expression (53).
It has also been suggested that SREBP1 in the liver could be involved in the transcriptional control of insulinresponsive genes (51) and that the regulation of gene expression by feeding could be exerted in part by insulin acting on the posttranscriptional activation of SREBP proteins (54). To our knowledge, ours is the first report of a detectable overt daily rhythm in the levels of active SREBP1 in the liver during ad libitum feeding (35) (Fig. 3C), which is consistent with an underlying rhythm in food intakemediated (and perhaps insulin-mediated) SREBP1 activation. Should this interpretation be valid, one would predict that the amplitude of the rhythmic SREBP1 cleavage will vary significantly depending on the extent of consolidation in the feeding behavior. Thus, deviations from our ad libitum results might be observed in other strains or under different conditions.
It is well established that the sterol depletion that ensues after fasting is a strong cue for SREBP activation. The mechanism involves the localized activity of an SREBP cleavage-activating protein (SCAP) whose shuttling back and forth between the endoplasmic reticulum and the Golgi is sterol level-dependent (55). Thus, it is likely that low sterol levels play a role in the consolidation of the phase-shifted pattern of SREBP1 activation we observed during the reverse restricted feeding paradigm (Fig. 3). It will be of interest to investigate whether the SREBP1 regulators, SCAP and CP332 (56), are themselves subject to circadian modulation in this context. More generally, we do not currently know to what extent the rhythmic cleavage of SREBP1 in the liver is also affected by the clock independent of the clear cyclic nature of the feeding behavior. On the other hand, as we and others have shown, key SREBP targets, such as ACC1, FAS, acyl-CoA synthase, glycerol-3-phosphate acyltransferase, ELOVL6, CIG30, and HMG-CoA reductase, display circadian expression patterns in several animal models (4,35,57), probably as a result of a dual regulatory mechanism involving nutritional (SREBP) and circadian (RevErba or BMAL1) signals, as proposed in the Elovl3 model.

A hypothetical model
We can now attempt to integrate these results into a coherent hypothetical model of Elovl3 transcriptional control that would incorporate the effect of the CLOCK mutation upon the levels of Elovl3 gene expression and the likely role of SREBP1 activation as a circadian sensor of nutritional status. We propose the existence of at least two input pathways controlling the Elovl3 promoter (Fig. 4). The first is clock-dependent, affected by RevErba, and active during the day. The second is food-dependent, mediated by SREBP1, and operates at the end of the active phase or an unscheduled fasting period (Fig. 3), when sterol levels are at a nadir and glucose levels begin to increase (58). SREBP1 mRNA has also been found to be under circadian control in the liver (4); this underlying rhythm could serve to replenish the pool of SREBP1 precursor being depleted during the dark phase.
The combination of the RevErba and SREBP1 antiphasic and opposing pathways could account for the high amplitude rhythm observed in the levels of Elovl3 mRNA. Restricted feeding during the day has already been shown to reverse the RevErba mRNA cycle (36). Because we have shown in this study, for the first time, that a similar paradigm shifts the SREBP1 activation cycle and Elovl3 mRNA acrophase by z9 h (Fig. 3), these observations are consistent with the hypothesis that the coordinated actions of RevErba and SREBP1 are ideally positioned to drive the observed rhythm in Elovl3 expression, and likely many other late-night-peaking genes, in the liver. We would have predicted that once the impact of the RevErba (clock) loop is abrogated, the influence of the second control mechanism, driven by the SREBP1 (nutrition) factor, would become the sole force driving transcription. Our finding that liver Elovl3 expression settles for a tonic level in CLOCK mutant mice, however, suggests that the two pathways might be interdependent. One possibility is that the expression of SREBP1 mRNA is dampened in the CLOCK mutant, which, in turn, would deplete the pool of activatable SREBP1 protein. We are currently analyzing the effect of clock mutations upon the rhythm of SREBP1 mRNA and SREBP1 proteolytic processing in the liver. Alternatively, the products of the peroxisome proliferator-activated receptor (PPAR) family of transcription factors could play a role in connecting the loops shown in Fig. 4. PPARg, for example, is a circadian (4) and SREBP1-driven (59) gene and has been reported to affect RevErba gene transcription. Similarly, mice that do not express the PPARa gene [a strongly circadian gene in the liver (4)] display a severe disruption of SREBP-dependent transcriptional pathways (60).

Potential physiological implications of Elovl3 circadian expression
In terms of how circadian regulation of Elovl3 could fit with the needs of the liver to synthesize long-chain fatty acids, we should mention that the main fraction of intracellular VLCFAs in the liver is esterified in various lipids, predominantly in sphingolipids. The bulk of sphingolipids are located in the plasma membrane, and a large body of evidence has been accumulated that implicates sphingolipids and their degradation products in signal transduction, cell proliferation, differentiation, and apoptosis (61) as well as in the formation of functional lipid microdomains (62). Given its circadian expression, as well as its role in synthesizing VLCFAs, it would be reasonable to propose that ELOVL3, along with other lipogenic enzymes, could contribute to establishing and/or sustaining a circadian pattern of membrane fluidity and, consequently, regulating the putative circadian rhythmicity of membranelinked processes in the liver, such as signal transduction involving sphingolipids, apoptosis, metabolite uptake, and turnover of phospholipids.
Regarding daily variation in lipid metabolism, studies using hamsters showed that the diurnal phase is characterized by inactivity and fuel storage (maximal enzymatic activity in fatty acid and triglyceride synthesis), whereas the night phase is one of exercise and fuel mobilization (fatty acid oxidation and sterol synthesis) (63). A circadian rhythmicity has also been reported in the expression level of lipogenic ACC1, FAS, ACS4, and G3PAT genes in mice (4,35). In this work, the mRNA level of Elovl3 shows the same pattern as those of FAS and G3PAT, peaking at the end of the night/beginning of the day. Elovl3 would, in this way, participate in the anabolic activity during the inactivity phase.
Additionally, the fact that Elovl3 was very sensitive to the acute reversal in the schedule of food consumption in mice is consistent with the significant phase shift in the circadian expression of Fas and g3pat genes observed under an identical feeding schedule (35). Such a mechanism could contribute to the increased energy storage at the end of the night/beginning of the day, preparing the body for or anticipating a period of food deprivation. The significant reversal in the expression of Elovl3 under diurnal feeding conditions is also consistent with the reported rapid switch of hepatic fatty acid metabolism from oxidation to synthesis and the concomitant hyperactivation of lipogenesis upon refeeding of diurnally fed rats (64).

Conclusions
The influence of the suprachiasmatic nucleus upon rhythmic gene expression in the liver is likely exerted through direct and indirect connections. It is well accepted that the autonomic nervous system and the establishment of activity/feeding schedules are likely the ultimate drivers of hepatic rhythmic gene expression, as pointed out in several previous studies (36,37,65).
The possible communication between the RevErba and SREBP1 pathways that emerges from our study of the Elovl3 gene promoter (Fig. 4) introduces a potentially attractive interphase for the homeostatic integration of clock and nutritional signals. The reversal of liver rhythms during restricted feeding occurs even if the photoperiod remains unchanged, and it does not affect the endogenous rhythms in the suprachiasmatic nucleus (36,37). Therefore, it has been concluded that restricted feeding can effectively uncouple the clock in the liver from the circadian humoral cues derived from the central oscillator in the brain (36). This phenomenon can be interpreted as evidence for the existence of a molecular sensor of nutritional status able to override other circadian entraining signals (37). A possible scenario for how this might be implemented emerges from the restricted feeding experiment (Fig. 3) involving the rapid modulation of the daily rhythm in SREBP1 precursor cleavage we have seen in the liver, which is predominantly linked to nutritional status.
In summary, we found that the expression of Elovl3 is sexually dimorphic: its circadian expression in the liver is detectable only in the male, whereas both genders express it in iBAT. We also propose for the first time a preliminary functional map of the Elovl3 upstream regulatory region and report that RevErba likely plays a connecting role between the core clock machinery and Elovl3 expression. In addition, our studies revealed that SREBP1 could play an important role in the regulation of the Elovl3 gene, activating it during the dark phase to consolidate and/or amplify the temporal pattern established by the diurnal action of the RevErba transcriptional repressor. Finally, we propose that the daily cycles of SREBP1 activation in the liver are loosely coupled to the clock machinery. We predict that this coupling is disengaged during diurnal restricted feeding, when a reversed pattern of SREBP1 activation overrides central circadian cues to reset the rhythms of many targets in the cholesterogenic and lipogenic pathways. Such a mechanism would be essential to accomplish an acute reversal in the sequence of absorptive and postabsorptive phases in the liver while avoiding a protracted clock-dependent reentrainment that would require several days to be completed.