Glucose Regulation of Mouse S14Gene Expression in Hepatocytes

Transcription of genes encoding enzymes required for lipogenesis is induced in hepatocytes in response to elevated glucose metabolism. We have previously mapped the carbohydrate-response elements (ChoREs) of the rat liver-type pyruvate kinase (L-PK) and S14 genes and found them to share significant sequence similarity. However, progress in unraveling this signaling pathway has been hampered due to the difficulty in identifying the key factor(s) that bind to these ChoREs. To gain further insight into the nature of the carbohydrate-responsive transcription factor, the glucose regulatory sequences from the mouse S14 gene were examined in primary hepatocytes. Three elements were found to be essential for supporting the glucose response: a thyroid hormone-response element between −1522 and −1494, an accessory factor site between −1421 and −1392, and the ChoRE between −1450 and −1425. Of these, only the accessory factor site was conserved between the rat and mouse S14 genes. Investigation of the ChoRE sequence indicated that two half E box motifs are critical for the response to glucose. Electrophoretic mobility shift assays revealed a complex formed between the mouse S14 ChoRE and liver nuclear proteins. This complex was also formed by ChoREs from the rat S14 and L-PK genes but not by mutants of these sites that are inactive in supporting the glucose response. These results suggest the presence of a novel transcription factor complex that mediates the glucose-regulated transcription of S14 and L-PK genes.

The mammalian liver is primarily responsible for the conversion of excess dietary carbohydrate to triglycerides, a process known as lipogenesis. The first hepatic response to carbohydrate intake is the activation of key rate-limiting enyzmes that convert carbohydrate to triglycerides. In the second phase, a longer term response is generated by the induction of a variety of enzymes involved in this process. These include enzymes of glycolysis, fatty acid biosynthesis, and triglyceride synthesis and maturation. Enzyme induction results from the increased levels of the respective mRNAs, and in a number of cases, this regulation occurs at the level of transcription (for review see Refs. [1][2][3][4]. The effects of dietary carbohydrate in the animal can be mimicked in primary hepatocytes by increasing media glucose concentrations and consequently rates of glucose metabolism (5,6). The glucose response in primary hepatocytes depends upon the presence of insulin, whose role is largely permissive through the activation of glucokinase gene expression (1,7,8). The intracellular mediator of the carbohydrate response has not conclusively been identified, although both glucose-6-phosphate and xylulose-5-phosphate have been proposed as candidates (9,10).
Other effectors that coordinately regulate most lipogenic enzyme genes include glucagon and polyunsaturated fatty acids, which repress gene expression, and thyroid hormones, which induce expression (for review see Refs. 2 and 11). In several cases, thyroid hormones support the synergistic activation of lipogenic enzyme expression with carbohydrate both in whole animals and in cultured primary hepatocytes. The molecular mechanism of the functional synergism by thyroid hormones and carbohydrate is unknown but does occur at the pretranslational level (5,6,12,13).
The rat S 14 gene was first investigated because of its rapid response to thyroid hormones in rat liver and primary hepatocytes (14 -16). Subsequent studies showed that S 14 mRNA levels are controlled by stimuli that modulate fatty acid formation, including dietary carbohydrate (6,12,17). Together with its restricted distribution in tissues active in lipogenesis, these data suggested a role for S 14 in lipogenesis (17). This suggestion was supported by experiments using antisense S 14 oligonucletides, which blocked the normal lipogenic response of hepatocytes (18,19). The rapid induction of S 14 mRNA levels by carbohydrate occurs mostly at the transcriptional level (13). Transfection analysis in primary hepatocytes led to the identification of regulatory sequences responsible for the carbohydrate response (20 -22). These sequences from Ϫ1467 to Ϫ1422 consist of two distinct sites. One site, designated the ChoRE 1 or carbohydrate-response element, is found from Ϫ1448 to Ϫ1422 in the rat S 14 gene and consists of two motifs related to the E box sequence CACGTG separated by 5 bp. When coupled in tandem copies, the ChoRE confers a response to glucose by itself, suggesting it is the binding site for a factor directly regulated by glucose. However, in the context of the natural promoter, an adjacent site from Ϫ1467 to Ϫ1449 is required for maximal induction. This accessory site binds an unknown hepatic transcription factor to augment the glucose response (22). The ChoRE of the rat S 14 gene bears striking similarity to a glucose regulatory sequence mapped in the L-PK gene, another gene regulated by carbohydrate metabolism (23,24). Based on the sequence of the ChoRE, it has been proposed that a member of the b/HLH/LZ transcription factor family binds to the ChoRE and activates transcription in response to glucose. The nature of the carbohydrate-responsive factor, however, is controversial. Whereas the b/HLH/LZ factor USF can bind to the ChoRE in vitro (21, 24 -26) and has been proposed to be required for the glucose response (27)(28)(29), other data argue against its direct participation (30). Clearly, identification of the carbohydrate-responsive factor is critical to further studies on the mechanism governing carbohydrate regulation of lipogenic gene expression.
Recently, the mouse homologue of the S 14 gene was cloned (31). Like its rat counterpart, the mouse S 14 gene is regulated by dietary carbohydrate in vivo (28). In this report, we map the mouse S 14 gene sequences critical for its control by carbohydrate. We demonstrate that the mouse S 14 gene employs three adjacent regulatory sequences for supporting a glucose response in primary hepatocytes. One of these sequences functioned analogously with the rat S 14 and L-PK ChoREs. Examination of this site in the mouse S 14 gene led to the refinement of the ChoRE consensus sequence and identification of a novel liver nuclear factor that binds to the ChoRE.

EXPERIMENTAL PROCEDURES
Primary Hepatocyte Culture and Transfection-Primary hepatocytes were isolated from male Harlan Sprague-Dawley rats (180 -260 g) using the collagenase perfusion method as described previously (30). After a 3-6-h attachment period, cells were transfected using either Lipofectin reagent (Life Technologies, Inc.) or F1 reagent (Targeting Systems) in modified Williams' E medium with 23 mM HEPES, 0.01 M dexamethasone, 0.1 unit/ml insulin, 1 unit/ml penicillin, 1 g/ml streptomycin, and mM glucose for 12-14 h. Subsequently, cells were cultured in medium containing either 5.5 or 27.5 mM glucose, with or without 500 nM T 3 . For studies using T 3 , Matrigel (Collaborative Biomedical Products) was added to the plates at a concentration of 0.33 mg/ml after transfection. This treatment has been shown to increase hormonal response of transfected genes (32). After a 48-h treatment, cells were harvested for CAT assay. Results are expressed as percentage conversion of chloramphenicol to its acetylated forms as determined by thin layer chromatography followed by phosphor screen autoradiography (Molecular Dynamics, Inc.) Plasmid Constructs-Mouse S 14 genomic DNA was obtained from an SV129 strain mouse genomic library and was provided by Dr. C. Mariash, University of Minnesota. To prepare the mS 14 (Ϫ5643/ϩ18)CAT construct, a XhoI site was first introduced at position ϩ18 in the 5Ј-untranslated region of mouse S 14 genomic DNA using polymerase chain reaction mutagenesis. A fragment from the unique XbaI site at Ϫ5643 to the engineered XhoI site at ϩ18 was inserted into the pCA-T(An) vector (33). The 5Ј-deletion series of mouse S 14 CAT constructs was generated using natural restriction endonuclease sites. To prepare the mS 14 (Ϫ1540/Ϫ1368)(Ϫ279/ϩ18)CAT construct, the mouse S 14 sequence from Ϫ1540 to Ϫ1368 was excised from mS 14 (Ϫ1540/ϩ18)CAT and inserted into the pTZ18R vector (Promega). The sequence was subsequently excised with BamHI and HindIII and inserted into mS 14 Clustered point mutations were generated by inverse polymerase chain reaction as described previously (34). Briefly, each oligonucleotide creating either a unique BglII site (TRE mut) or NsiI site (mut1-8) was synthesized and used to amplify the mS 14 (Ϫ1540/Ϫ1368)/pTZ18R plasmid. Polymerase chain reaction products were digested at the introduced restriction enzyme site, purified by gel electrophoresis, ligated, and transformed into Escherichia coli. Mutant constructs were isolated, and mouse S 14 sequences were excised with BamHI and HindIII to clone into the mS 14 (Ϫ279/ϩ18) CAT construct.
Oligonucleotides containing sequences of the mouse S 14 ChoRE (Ϫ1450/Ϫ1425), point mutants of the ChoRE (3-1 to 3-6), and the mouse S 14 accessory factor site (Ϫ1421/Ϫ1392) were synthesized with additional BamHI and BglII sites at the 5Ј-and 3Ј-ends, respectively. Each oligonucleotide was ligated and treated with BamHI and BglII to isolate a DNA fragment with three copies in a head-to-tail orientation. Fragments containing three copies were then inserted into the BamHI site of PK(Ϫ96)CAT construct (35). All plasmid constructs described above were confirmed by DNA sequencing.
Preparation of Nuclear Extracts-Liver nuclear extracts were prepared from male Harlan Sprague-Dawley rats that had been fasted overnight and then fed a high carbohydrate diet for 16 h as described previously (22). The crude liver nuclei were then fractionated by PEG (PEG8000, Hampton Research) precipitation. For the binding assays described, the PEG fraction from 0 to 6.7% was exclusively used.
Electrophoretic Mobility Shift Assay-EMSA was performed as described previously (20). A typical reaction contained 100,000 cpm (10 -30 fmol) of 32 P-labeled oligonucleotide with 10 -20 g of nuclear protein. Nonspecific competitors were 0.1 g of poly(dI⅐dC) and 1.9 g of poly(dA⅐dT). Following incubation at room temperature for 30 min, samples were subjected to electrophoresis on a 4.5% nondenaturing polyacrylamide gel and subjected to phosphorimager analysis. Antibodies to USF1 (C-20) and USF2 (C-20) were from Santa Cruz Biotechnology and were added together to nuclear extract for 20 min at 4°C prior to the addition of probe.

RESULTS
The Sequence of the Mouse S 14 Gene Corresponding to the Rat S 14 ChoRE Does Not Support a Carbohydrate Response-The rat S 14 gene sequence from Ϫ1467 to Ϫ1422 contains two sites, the accessory factor site and ChoRE, which are required to support a transcriptional response to glucose (22). The mouse S 14 gene sequence from Ϫ1411 to Ϫ1366 aligned to that region of the rat gene with an identity of 85%. Despite this high degree of similarity, the mouse sequence corresponding to the rat ChoRE contains a number of mismatches in the E box motifs at positions previously shown (30) to be essential for the glucose response (Fig. 1A). This observation raised the question of whether the rat ChoRE is functionally conserved in the mouse gene. To investigate this question, an oligonucleotide containing the mouse sequence from Ϫ1411 to Ϫ1366 was cloned into a glucose-unresponsive PK(Ϫ96/ϩ12)CAT reporter construct, which we have previously used as a basal promoter construct (22,30). As shown previously (22) and in Fig. 1B, a reporter  (30)). B, rat primary hepatocytes were transfected with CAT reporter constructs containing either the glucose-responsive sequence of the rat S 14 gene or the corresponding sequence of the mouse S 14 gene linked to the PK(Ϫ96) promoter. As a control, a CAT reporter construct containing the PK(Ϫ96) promoter alone was also transfected. Cells were cultured in 5.5 (solid bars) or 27.5 (hatched bars) mM glucose for 48 h. CAT activity is shown as relative percentage conversion of chloramphenicol to its acetylated forms with the value for the rS 14  construct containing the rat S 14 sequence from Ϫ1467 to Ϫ1422 conferred a glucose response when transfected into primary hepatocytes. The partial stimulation observed in low glucose conditions with this construct likely reflects the fact that 5.5 mM glucose is not the basal level in terms of glucose signaling. In contrast, a reporter construct containing the mouse S 14 sequence did not show any increase when compared with the basal promoter construct alone and did not respond to glucose. These data suggest that the mouse S 14 sequence corresponding to the rat S 14 ChoRE does not support a carbohydrate response in primary rat hepatocytes. Two possibilities may explain this observation. First, the glucose-responsive transcription factor in rat hepatocytes may not recognize the corresponding mouse ChoRE. Second, the mouse ChoRE may be located in a different position from that of the rat gene. To distinguish between these possibilities, a larger segment of the mouse S 14 gene was examined for its ability to respond to glucose in rat hepatocytes.
Thyroid Hormone Works as a Permissive Signal for Supporting a Glucose Response of the Mouse S 14 Gene-The mouse S 14 genomic sequence from Ϫ5643 to ϩ18 was cloned into a CAT reporter construct and transfected into primary rat hepatocytes. Activity of this construct was only slightly increased by either high glucose or T 3 alone compared with cells maintained in low glucose without T 3 ( Fig. 2A). However, when transfected cells were cultured in medium containing both high glucose and T 3 , the mouse sequence conferred a strong induction, indicating a synergism between glucose and thyroid hormone in transactivating the gene. Thus, the mouse S 14 gene is capable of being regulated by factors in the rat hepatocyte.
To localize the regulatory sequences required for the effects of glucose and T 3 , deletions of the 5Ј-end of the mouse S 14 genomic sequences were generated and tested in primary hepatocytes. Deletions from the 5Ј-end to Ϫ1540 still supported the synergistic effects of glucose and thyroid hormone, although the magnitude of the response diminished somewhat ( Fig. 2A). Further deletion to Ϫ1147 completely abolished the response. These results suggest that the sequence from Ϫ1540 to Ϫ1148 contains a regulatory element or elements that supports the synergism between glucose and thyroid hormone. Further deletion analysis within this region demonstrated that the sequence from Ϫ1540 to Ϫ1368 is required and sufficient to confer the response to glucose and T 3 in primary hepatocytes (Fig. 2B). Promoter activity directed by sequences downstream from Ϫ1368 are unaffected by these treatments. Thus, regulatory sequences in the mouse S 14 gene are contained within the same DNA region as the rat gene but differ in requiring T 3 for their activity.
Sequence analysis revealed a putative thyroid hormone receptor binding site between Ϫ1522 to Ϫ1494 in the mouse S 14 gene (see Fig. 3A). This sequence contains two motifs related to the consensus TRE half-site (A/G)GGTCA in an inverted orientation with 7-bp spacing. Indeed, an oligonucleotide containing this sequence can bind to in vitro translated thyroid hormone receptor/retinoid X receptor heterodimers by EMSA. 2  point mutation, which disrupts the TRE, was generated and tested in the context of the mS 14 (Ϫ1540/Ϫ1368)(Ϫ279/ ϩ18)CAT reporter construct. As expected, the construct bearing a mutation in its TRE showed a much diminished response to glucose and thyroid hormone compared with the wild type sequence (Fig. 2B). However, a weak response to glucose is retained in the construct with the TRE mutation. These data suggest that the ligand-bound thyroid hormone receptor works as an accessory factor to augment the glucose induction of the mouse S 14 gene, and an independent carbohydrate-response element is also present within this region.
Two Separate Sequences, in Addition to the TRE, Are Essential for the Carbohydrate Induction-To identify additional regulatory sequences involved in the glucose response, further clustered point mutations were introduced into the mouse S 14 gene. For this purpose, we focused on the sequence from Ϫ1469 to Ϫ1368, as this segment was capable of supporting a glucose response independent of thyroid hormone when linked to the heterologous PK(Ϫ96/ϩ12) promoter (see Fig. 4). Eight clustered point mutants were designed spanning this segment (Fig.  3A). Because the TRE present in Ϫ1522 to Ϫ1494 can augment the glucose-generated signal, the mutations were generated in the context of the mouse S 14 gene from Ϫ1540 to Ϫ1368 and linked to the mouse S 14 (Ϫ279/ϩ18) promoter. Each mutant construct was transfected into primary hepatocytes and cultured in the presence of 5.5 mM glucose or 27.5 mM glucose plus 500 nM T 3 for 48 h. The ability of three of these mutant constructs (mut3, mut5, and mut6) to respond was largely or completely abolished (Fig. 3B). All other mutant constructs displayed responses that were comparable to or, in the case of mut7, greater than the wild type construct. It is worth noting that the mut8 mutation alters sequences corresponding to the rat S 14 upstream E box, further confirming that this region does not function as a ChoRE in the mouse gene. Comparable results were obtained when these mutations were tested in the context of the mS 14 (Ϫ1540/ϩ18)CAT construct. 2 Interestingly, the construct containing mut4, which is localized between the inactivating mut3 and mut5/6 mutations, gives a glucose response as high as the wild type. These data indicate that the glucose-responsive region of the mouse S 14 gene contains two regulatory sites. Based on previous characterization of the glucose-responsive rat S 14 and L-PK genes, we postulated that these two sites might function as a ChoRE and an accessory factor site (23,24,32,36).
To investigate the role of the two regulatory sites predicted by mutagenesis in the mouse S 14 gene, two oligonucleotides containing these sequences were generated. Each oligonucleotide was ligated in three copies in a head-to-tail orientation and linked to a CAT construct containing the PK(Ϫ96) basal promoter. These constructs were tested in primary hepatocytes at low or high glucose concentrations. The construct containing the sequence from Ϫ1421 to Ϫ1392 (corresponding to sequences mutated in mut5 and mut6) did not respond to glucose but showed higher basal activity in low glucose, suggesting that a regulatory sequence for an accessory factor is present in this region (Fig. 4). Interestingly, the corresponding region of the rat S 14 gene also contains an accessory factor site. The construct containing the sequence from Ϫ1450 to Ϫ1425 of the mouse S 14 gene (corresponding to mut3) conferred a glucose response even in the absence of other regulatory sequences. This result indicate that this site functions as a ChoRE in the hepatocyte. Surprisingly, the mouse S 14 ChoRE did not contain two E box motifs separated by 5 bp as had been found in previously identified ChoREs from the rat S 14 and the L-PK genes (22).
Two E Box "Half Sites" Are Important for the Glucose Response-The lack of clearly distinguishable E box motifs in the mouse S 14 ChoRE led us to investigate which sequences within this region are critical for the glucose response. Consequently, five point mutations of 2 bp were introduced across the ChoRE sequence (Fig. 5A). In particular, we targeted CA (or TG) dinucleotides that might represent portions of degenerate E box sequences. Oligonucleotides containing these mutations were linked in three copies to the PK(Ϫ96)CAT construct to test for glucose responsiveness. Only two of the mutant constructs, 3-2 and 3-4, showed significant reductions in glucosestimulated promoter activity (Fig. 5B). Examination of the sequences surrounding these inactivating mutations revealed that each was part of a potential E box half site: CACG (for 3-2) and CAGC (for 3-4). Further scrutiny of both the rat S 14 and PK ChoREs reveals that each also contains two potential E box half sites related to CACG (see "Discussion"). To test if this E box 'half site' might represent an essential element for the glucose response, an additional mutation was introduced into the mouse S 14 ChoRE to substitute GC in the second putative E box half site with CG, generating a CACG sequence. This mutated construct, 3-6, retained its ability to confer a strong response to glucose. Thus, the presence of two E box half sites related to the sequence CACG may be the critical determinants for mediating the glucose response.
A Novel Factor That Binds to ChoREs Is Present in Liver Nuclear Extract-To search for the specific nuclear factor that is responsible for the glucose response of the mouse S 14 ChoRE, rat liver nuclear extracts were prepared, fractionated by PEG precipitation, and used for EMSA with a mouse S 14 ChoRE probe. Using the fraction precipitated between 0 and 6.7% PEG, three complexes, designated bands x, y, and z, were observed (Fig. 6). The same three bands were detected with unfractionated nuclear extract, although the intensities of bands x and y were much lower relative to band z prior to fractionation. To determine whether any of these bands might be a candidate for the carbohydrate-responsive complex, the six mutant mouse S 14 ChoREs tested above were used for EMSA. Among the three bands detected with the wild type mouse S 14 ChoRE, only the slowest migrating band (band x) was retained in lanes with mutant ChoREs that were glucose-responsive but not with glucose-unresponsive oligonucleotides. Note that the weak band observed with the 3-4 oligonucleotide that migrates similarly to band x on this gel showed distinctively slower mobility with longer electrophoretic conditions and 3-4 could not compete for band x formation with the wild type mouse S 14 ChoRE. In contrast, band y was barely detectable with the glucose-responsive 3-6 oligonucleotide, whereas band z was not observed with 3-3. Thus, band x contains a factor(s) that is a candidate for the carbohydrate-responsive factor, and we have designated this complex as ChoRF.
To further evaluate this possibility, we asked whether this same complex could be observed with other ChoREs. EMSA was carried out using four different glucose-responsive elements together with four control elements. In addition to the mouse S 14 ChoRE, we tested the rat S 14 and L-PK ChoREs, as well as a mutant rat S 14 ChoRE designated mut3/5. This latter element contains a palindromic arrangement of the degenerate E box sequence CAtGcG and was previously shown to support a glucose response (30). Other probes tested included the adenovirus major late promoter USF binding site, the L-PK accessory site (hepatic nuclear factor-4) binding site, a mutant of the rat S 14 ChoRE with the E box sequence CAaGTG that did not respond to glucose (30), and a consensus binding site for the factor SREBP. As seen in Fig. 7, a band comigrating with ChoRF was observed with all of the oligonucleotides capable of conferring a glucose response, whereas none of the control oligonucleotides formed this complex. These results support the possibility that this complex contains the carbohydrateresponsive factor responsible for the glucose induction of S 14 and L-PK genes.
As an additional means of assessing involvement of the ChoRF complex in the glucose response, mutant ChoREs of the rat S 14 and L-PK genes were used for EMSA. These mutations were originally designed by either adding or subtracting one bp to change the spacing between the two E box motifs. We reported previously that a spacing of 6 bp (N6) gave only a partial glucose induction in primary hepatocytes compared with the wild type spacing of 5 bp (N5), whereas a 4-bp separation (N4) did not respond to glucose (22). If the ChoRF complex contains the carbohydrate-responsive factor, we expected to observe a correlation between binding of the factor to three probes and functional activity. Indeed, the N5 oligonucleotide of either the rat S 14 or L-PK ChoREs showed the strongest binding, the N6 FIG. 6. Detection of a novel complex formed by mouse S 14 ChoRE and ChoRE mutants that are glucose-responsive. EMSA was performed as described under "Experimental Procedures" with rat liver nuclear proteins and wild type (WT) or the six mutated mouse S 14 ChoREs shown in Fig. 5A. The arrows indicate the positions of three bands detected with the wild type mouse S 14 ChoRE oligonucleotide. Note that the band labeled "x" is detected with all oligonucleotides that support a glucose response. Band "y" is not observed with the glucoseresponsive mutant 3-6, whereas band "z" (USF) is not observed with the glucose-responsive mutant 3-3. All lanes contained 20 g of nuclear protein except lane 3-6, which contained 10 g. oligonucleotide bound weaker, and N4 did not detectably bind (Fig. 8). The faster migrating band observed in these gels represents USF, and there was no difference in intensity of USF binding to probes with different spacing. These results strongly suggest that the factor(s) forming the ChoRF complex is capable of binding to only glucose-responsive sequence elements and likely represents the carbohydrate-responsive factor.
Although ChoRF migrates more slowly than USF, it is still possible that USF may be a component of the ChoRF band observed on EMSA. To investigate this possibility, we tested whether anti-USF antibody could disrupt the formation of ChoRF. As a control, the adenovirus major late promoter USF binding site was used. USF binding of the adenovirus USF binding site was effectively inhibited by adding increasing amounts of anti-USF antibody (Fig. 9). However, the intensity of the ChoRF complex on two different ChoRE probes was unchanged in the presence of anti-USF antibody. Thus, USF is not likely to be a component of the ChoRF complex.

DISCUSSION
The ability of the hepatocyte to respond to elevated metabolism of glucose and other glycolytic substrates by increasing transcription of genes involved in fatty acid production has provided a valuable model for exploring nutrient control of mammalian gene expression. In particular, the rat L-PK and S 14 genes have been intensively analyzed by transfection studies to define the regulatory sites essential for supporting the glucose response (22)(23)(24). However, progress in dissecting this novel signaling pathway has been thwarted by the difficulty in identifying the key factor(s) that bind to the transcriptional regulatory sites. To gain possible insights into the nature of the carbohydrate-responsive factor, we have examined the sequences of L-PK and S 14 genes from other species to determine whether the regulatory elements are conserved or whether sequence variants might be identified. For example, a comparison of human and rat L-PK promoters revealed complete identity in the 36-bp region encompassing the ChoRE and hepatic nuclear factor-4 accessory factor sites. Other sequences within the first 200 bp of the 5Ј-flanking regions were less conserved (60% identity), including sites in the rat gene for hepatic nuclear factor-1 and nuclear factor-1 binding. This observation supports the importance of the glucose regulatory sites mapped in the rat L-PK gene and suggests that the ability to respond to carbohydrates is conserved in humans.
Consequently, it was unexpected to find substantial differences in the organization of the glucose-responsive regions between the rat and mouse S 14 genes (Fig. 10A). First, the mouse S 14 gene requires the presence of T 3 and a TRE adjacent to the ChoRE to confer a glucose response. Thyroid hormones can also transcriptionally activate the rat S 14 gene, and this effect is synergistic with carbohydrate (6,13). Thyroid hormones exert their action through multiple TREs present in the rat S 14 gene between Ϫ2790 and Ϫ2494 (34), which are largely conserved in the corresponding sequence of the mouse gene. However, in the rat S 14 gene, the glucose-responsive region does not include an additional TRE, and glucose can confer a response independent of the upstream TREs (20 -22). The second difference between the glucose-responsive regions of the rat and mouse S 14 genes is the location of the ChoREs relative to their accessory factor sites. Because the mouse and rat S 14 genes are so closely related with an overall identity of 84% in the 5Ј-flanking region, it is surprising that these genes utilize regulatory sequences localized at different sites to respond to the same signal. Based on both functional and binding studies, the mouse ChoRE appears to be a weaker response element than the rat ChoRE. Thus, it might be expected that the mouse element requires the cooperation of two accessory factor sites, rather than just one, to effectively support a glucose response. In this case, the positioning of the mouse ChoRE between the TRE and an accessory factor site may allow it to effectively interact with factors binding at both of these sites. How this change occurred evolutionarily following the divergence of these two species is a matter for speculation. It would be FIG. 8. Spacing mutations that modify glucose responsiveness of the ChoRE also alter formation of the ChoRF complex. EMSA was performed with 10 g of rat liver nuclear protein and oligonucleotides derived from either the consensus CACGTG ChoRE (30) or L-PK ChoRE. Mutations that change the spacing between E box motifs from the natural spacing of 5 bp between the two E box motifs (N5) to either 4 bp (N4) or 6 bp (N6) were used (22). The thick arrow indicates the novel ChoRF complex, whereas the thin arrow indicates the position of the USF complex. interesting to compare glucose regulatory sequences in other related species to explore which represents the primordial organization.
The organization of regulatory elements functioning coordinately with accessory factor sites is commonly found in genes encoding enzymes of metabolic importance. For example, genes encoding enzymes required for cholesterol synthesis and uptake are regulated by SREBP transcription factors in response to cellular levels of cholesterol (for review see Refs. 37 and 38). SREBPs bind to a common site, the SRE-1, found in those genes. However, in all cases tested, SREBPs are unable to transactivate without the presence of other transcription factors binding at sites adjacent or near to the SRE-1. Several different accessory factors, including Sp-1 and nuclear factor-Y, have been found to function with SREBPs to increase the transcriptional signal mediated in response to cholesterol depletion (39 -41). In the case of Sp1 and nuclear factor-Y, the synergistic effect was due to cooperative binding to DNA through direct interaction of the two factors. In other cases, functional synergism between two factors does not involve direct interaction and presumably arises from interactions with additional components (36). The mechanism by which the accessory factor for the mouse and rat S 14 genes functions to augment the glucose response is currently unknown.
Previous work on the rat S 14 and L-PK genes suggested that two CACGTG-type E boxes are critical for the glucose response of these genes (22)(23)(24). Each E box was proposed to bind to a b/HLH/LZ dimer and both dimers would presumably be required for glucose signaling. Based on the analysis of mouse S 14 ChoRE, we postulate a new model for the ChoRE. In this model, the ChoRE consists of two E box half sites related to the CACG motif. Such a regulatory site could act by binding to a single dimer of the b/HLH family. In the mouse S 14 ChoRE, mutations that disrupted the CACG or CAGC sequences inhibited the glucose response, but mutations of bases immediately downstream of these sequences did not. Furthermore, formation of the ChoRF complex in the EMSA was not affected by the latter mutations. Because we detected the same complex in the EMSA with the rat S 14 and L-PK ChoREs, it is possible that these ChoREs also function through two CACG-type E box half sites rather than the 6-bp E box motifs. Indeed, the rat S 14 and L-PK ChoREs fit the new model (Fig. 10B). Mutations of the rat S 14 or L-PK ChoREs that affect the first 4 bp of the CACGTG motif abolished the response to glucose, whereas mutations of the fifth and sixth positions did not disrupt the glucose response, suggesting that only the first 4 bp of the E box are critical (30). Similarly, mutations of the rat or mouse S 14 ChoREs that give a better fit to the consensus CACG motif generally behave as up-mutations. Despite these similarities, there is one distinct difference in the structure of the previously characterized ChoREs and that of the mouse S 14 gene. In the ChoREs of the rat S 14 and L-PK genes, the two CACG motifs are separated by 9 bp in an inverted orientation, whereas the CACG motifs in the mouse S 14 ChoRE are situated in a direct orientation separated by 7 bp. The ability of a single transcription factor to recognize different orientations of half site motifs is not unprecedented. For example, the thyroid hormone receptor and other members of the nuclear receptor superfamily recognize two AGGTCA motifs that can be organized with different arrangements. These include direct, inverted, and everted repeats with different spacing between the motifs (42). Similarly, SREBP is an example of an E box-binding protein than can recognize both direct repeats (SRE-1) and inverted repeats (CACGTG) of a half site motif (43). The carbohydrateresponsive factor also appears to display this flexibility in its binding requirements for different ChoREs.
USF first emerged as a candidate for the carbohydrate-responsive factor in hepatocytes. This possibility was based on the observation that USF forms a major complex with the rat S 14 and L-PK ChoREs using liver nuclear extract in the EMSA (21, 24 -26). Lefrancois-Martinez et al. (27) reported that the overexpression of USF can activate the transfected L-PK promoter in hepatocytes and hepatoma cell lines. In mice deleted for the USF2 gene, the induction of L-PK or S 14 mRNAs in liver is delayed after feeding of high carbohydrate diet (28,29). However, several lines of evidence argue against USF as a carbohydrate-responsive transcription factor. In the EMSA, USF lacks the ability to discern altered ChoREs that are glucose-responsive from glucose-unresponsive oligonucleotides either in binding experiments or transfected cells (30). Similarly, for the mouse S 14 ChoRE mutant 3-3, USF binding (band z) is greatly reduced with no apparent reduction in functional activity. Furthermore, overexpression of a dominant negative form of USF in hepatocytes failed to block the glucose response of the rat S 14 or L-PK promoters (30). Finally, the ChoRF activity detected by EMSA was found predominantly in a distinct PEG fraction than USF. ChoRF binding was not inhibited by adding USF antibody, indicating that USF is not present in this complex. Overall, these data indicate that USF is not likely to be directly involved in the transcriptional response of S 14 and L-PK genes through the ChoRE. The abnormal carbohydrate response observed in USF2 knockout mice or cells treated with dominant negative forms of USF for several days, however, indicates that USF may play an indirect role in the process. FIG. 10. Two E box half sites related to the sequence CACG are found in all active ChoREs. A, the organization of regulatory sequences required for supporting a response to glucose in the mouse and rat S 14 genes is compared. AF, accessory factor site. B, comparison of wild type and mutant ChoRE sequences that are capable of supporting a glucose response. Boxes indicate previously proposed E box motifs in the rat S 14 and L-PK ChoREs. Arrows represent CACG-like E box half site motifs found in all active ChoREs. Element are grouped based on the spacing and orientation of the E box half site motifs. rat S 14 (ϩ) is an oligonucleotide with a 1-bp mutation of the downstream E box site (CCTGTG to CCCGTG) that has been shown to increase glucose responsiveness (22). rat S 14 (MluI) is an oligonucleotide with three bp changes in the downstream E box site (CCTGTG to CACGCG) that also increases the functional activity (22). 2XCACGTG is an oligonucleotide with the consensus E box sequence (30). SREBP-1c is another b/HLH/LZ factor that has been proposed as a potential carbohydrate-responsive factor. SREBP-1c has the ability to bind either the SRE-1 or E box motifs and is highly expressed in the mammalian liver (44). SREBP-1c expression in liver is up-regulated by feeding a carbohydrate diet to rodents and by insulin treatment of hepatocytes (45,46). Overexpression of a constitutively active form of SREBP-1c in either cultured cells or transgenic mice can activate expression of several lipogenic enzyme genes, including fatty acid synthase (47)(48)(49). Finally, expression of a dominant negative form of SREBP-1c in primary hepatocytes resulted in a reduction of glucose-induced mRNA levels of the fatty acid synthase, S 14 , and L-PK genes (46). However, there are several inconsistencies with the binding and activation properties of SREBP-1c and the expected properties of ChoRF. For instance, nuclear SREBP-1 did not form a complex with several active ChoREs, including the mouse S 14 and L-PK ChoREs, and a consensus SRE-1 oligonucleotide did not compete with ChoRF binding. 2 In addition, cotransfected SREBP-1c showed little or no activation of L-PK ChoRE-containing constructs (50). Although SREBP appears to play a role in regulating some aspects of lipogenesis and perhaps coordinating cholesterol biosynthesis and lipogenesis, it does not appear to be the nuclear factor directly binding to the ChoRE of the S 14 and L-PK genes.
The demonstration of the ChoRF complex by EMSA suggests the possibility of a novel carbohydrate-responsive factor(s) in liver. The ability of various oligonucleotides to form this complex correlates with their competence to support a glucose response in hepatocytes. Three natural ChoREs and seven mutants of these ChoREs with functional activity all yielded the ChoRF complex on EMSA. In contrast, ChoRE mutants without functional activity and unrelated oligonucleotides showed very little or no ChoRF complex formation. Of particular note is the spacing mutant N4, which differs by only 1 bp from the N5 oligonucleotide, and yet has no functional activity or ability to form the ChoRF complex. Hence, we suggest that the ChoRF complex contains the nuclear factor(s) that receives the glucose signal. The failure to detect the ChoRF complex in our previous work stemmed from two factors. First, the ChoRF complex is apparently of low abundance and not easily detected in crude nuclear extracts. Using the PEG fractionation, we were able to enrich this factor sufficiently to allow its routine detection. Second, the formation of the ChoRF complex is largely refractory to poly(dI⅐dC), which was used in much of our earlier work as a nonspecific competitor. The ChoRF complex migrates more slowly than either the USF dimer or the SREBP (1a, 1c, or 2) dimers. However, we cannot rule out the possibility that ChoRF might contain USF or SREBP as a component of a larger complex at this time. In addition, two recent reports have described distinct nuclear factors that bind to the L-PK ChoRE (51,52). The relative migration of those bands was significantly faster than the ChoRF complex that we detected. Whereas we have not observed these faster migrating complexes in our experiments, these factors could also be components of the ChoRF complex. In preliminary experiments, we have not found a difference in the intensity of the ChoRF complex in extracts prepared from rats of varying dietary status. Hence, it would appear that glucose signaling controls the activity of the ChoRF complex rather than its binding to DNA. Further exploration of this pathway will require purification and characterization of the ChoRF complex.