Diabetes Alters the Occupancy of the Hepatic 3-Hydroxy-3-methylglutaryl-CoA Reductase Promoter*

Hepatic 3-hydroxy-3-methylglutaryl-CoA reductase (HMGR) protein and mRNA are substantially decreased in diabetic animals and rapidly restored by the administration of insulin. To begin to examine the underlying molecular mechanisms, measurements of transcription by nuclear run-on assays and an investigation of occupancy of the promoter were performed. The rate of transcription was substantially reduced in the diabetic rats and fully restored within 2 h after insulin treatment. In vivo footprinting revealed several areas of protein binding as shown by dimethyl sulfate protection or enhancement. The cAMP-response element was heavily protected in all conditions, including diabetes, feeding of dietary cholesterol, or statin treatment. Striking enhancements in footprints from diabetic animals were visible at –142 and at –161 (in the sterol-response element). Protections at a newly identified NF-Y site at –70/–71 were observed in normal animals and not in diabetics. This NF-Y site was found to be required for efficient HMGR transcription in luciferase assays. CREB-1 was able to bind the HMGR cAMP-response element in vitro and the promoter in vivo. This evidence supports an essential role for cAMP-response element-binding protein in transcription of hepatic HMGR and identifies at least two sites where in vivo occupancy is regulated by insulin.

Type I diabetes is associated with lower rates of cholesterol synthesis and increased absorption of dietary cholesterol in humans (1). These individuals are at high risk for the development of cardiovascular disease (2) and have higher total serum cholesterol levels. In rats, streptozotocin-induced diabetes also renders animals particularly susceptible to a dietary cholesterol insult (3). For reasons that are still unclear, this sensitivity correlates well with decreased expression of hepatic HMG 2 -CoA reductase (4), the enzyme that catalyzes the rate-limiting reaction in cholesterol biosynthesis.
Hepatic HMG-CoA reductase is responsible for the majority of the regulatable cholesterol synthesis in the body. The expression of this enzyme is affected by cholesterol, insulin, thyroid hormone, bile acids, fasting, and refeeding and also varies diurnally (5). HMG-CoA reductase (HMGR) protein and mRNA levels are both decreased in diabetic animals and can be rapidly restored with insulin treatment (6), suggesting regulation at the transcriptional level. Previous experiments in H4IIE cells (rat hepatoma) showed that the proximal reductase promoter could be activated by insulin (7), at levels greater than or equal to those seen in live animals. Questions remain as to whether this mode of insulin activation mirrors the physiological regulation of the gene.
The hamster HMG-CoA reductase gene requires about 300 bp of sequence upstream of the transcription start site for high level expression (8). This proximal promoter was found to contain sequences sufficient for sterol regulation in cultured cells (9) and shares about 90% sequence identity with the rat promoter (GenBank TM accession number S78687 (10)). The HMGR promoter contains a sterol-response element (SRE) that can be activated by SREBP-1 and SREBP-2 in cultured cells and in transgenic mice overexpressing these proteins (11)(12)(13). Although SREBP-1c appears to be insulin-responsive at the mRNA level, recent evidence suggests this factor is more closely tied to lipogenesis than cholesterol biosynthesis (14 -16). SREBP-2 is a potent activator of the HMG-CoA reductase gene, but insulin regulation of this protein has not been reported. There are several other important elements in the HMG-CoA reductase promoter. These include possible binding sites for Sp1 and NF-Y, as well as a functional cyclic AMP-response element (CRE) (10). The CRE in particular was shown to be required for insulin activation of the HMGR promoter in rat hepatoma cells (7).
Because of the problems inherent in a cell culture model, especially for a gene that is sterol-sensitive, we decided to perform in vivo footprinting in rat liver. This approach allows for a complete unbiased survey of the HMGR promoter. Performing this technique in animals ensures that the footprint reflects physiological regulation of the gene, in the context of the many nutritional and hormonal stimuli that the liver receives. In vivo footprinting has been used successfully to map where transcription factors are bound to DNA in vivo (17,18). Previous in vitro footprinting studies of the HMGR promoter were successful in identifying sterol-responsive elements (19). The only previous in vivo footprints of this promoter failed to detect changes in occupancy in response to insulin in HepG2 cells, although a 1.5-fold increase in mRNA was observed (20). In this report we show that diabetes alters the occupancy of the HMG-CoA reductase promoter in live animals.

MATERIALS AND METHODS
Animal Care and Treatment-Male Sprague-Dawley rats, 125-150 g (Harlan), were allowed free access to Harlan Teklad 22/5 rodent chow and water. Animals were kept on a reverse cycle lighting system and were sacrificed at 9:00 -10:00 a.m., when HMG-CoA reductase expression is at its diurnal high. Animals were rendered diabetic by a single subcutaneous injection of streptozotocin (Sigma), 65 mg/kg. Diabetes was verified by the presence of urinary glucose using Clinistix from Bayer. Where indicated, animals were injected subcutaneously with 3.0 units/100 g of recombinant human insulin (Novolin 70/30, Novo Nordisk) 2 h prior to sacrifice.
Nuclei Isolation-Nuclei were prepared as described previously (21) by centrifugation through dense sucrose.
Nuclear Run-on Assay-Nuclear run-on assays were carried out essentially as described previously (22). After centrifugation, nuclei from 2 g of liver were resuspended in 100 l of PBS with 3 mM MgCl 2 . Next, 100 l of 2ϫ run-on buffer (160 mM Tris, pH 7.5, 20 mM MgCl 2 , 2 mg/ml heparin, 1% Sarkosyl, 0.7 M ammonium sulfate, 0.8 mM each of ATP, GTP, and UTP) and 250 Ci of [␣-32 P]CTP were added to each reaction. Samples were incubated at 37°C to continue extension of RNA transcripts. Samples were then treated with 50 units of DNase I, 40 g of proteinase K and lysed with the addition of 50 l of 10% SDS and 25 l of 0.2 M EDTA. After extraction and precipitation, equal counts of 32 P-labeled RNA were added to each membrane at ϳ5 ϫ 10 6 dpm. These membranes were spotted previously with 5 g of cDNA encoding HMG-CoA reductase, catalase, or the pBluescript vector. Hybridizations were performed overnight at 57°C. The next day membranes were washed with 2ϫ SSC, 0.1% SDS at room temperature for 1 min, then 0.2ϫ SSC, 0.1% SDS at 60°C for 30 min, followed by 2ϫ SSC containing 250 g of RNase A at 37°C for 30 min. Membranes were given a quick final rinse in 2ϫ SSC and dried, followed by exposure to autoradiography film with an enhancing screen at Ϫ70°C for 1-5 days.
Sequencing-The rat HMG-CoA reductase promoter was obtained by PCR of rat liver genomic DNA, using primers to the sequence published previously (10). PCR products were sequenced by Retrogen (San Diego).
In Vivo Footprinting of Rat Liver-Rat liver (2.2g) was minced in 8 ml of ice-cold PBS. Liver pieces were homogenized 4 -5 times on a drill press with a Teflon pestle in a glass vessel. A 5-ml portion of filtered homogenate was placed in a 50-ml polypropylene centrifuge tube. Each filtered homogenate was treated with 5 l of dimethyl sulfate for 2 min at room temperature. The DMS reaction was slowed by rapid dilution with 40 ml of ice-cold PBS. Tubes were centrifuged at 1000 ϫ g for 5 min at 4°C. Pellets were resuspended in 20 ml of PBS and washed again. The pellet was then resuspended in 15 ml of lysis buffer (60 mM Tris, pH 7.5, 100 mM EDTA, 0.5% SDS, and 100 g/ml proteinase K). Samples were rocked gently at room temperature for 3 h to completely lyse nuclei. Genomic DNA isolation and piperidine treatment were performed as described previously (17). Roughly 200 -300 g of DNA was obtained per sample.
Ligation-mediated PCR-Ligation-mediated PCR was performed according to the original method (17). The primers corresponding to the coding (top) strand of the HMGR promoter starting at Ϫ185 were as follows: primer 1, 5Ј-CAA TAG GAA GGC CGC GAT GC-3Ј; primer 2, 5Ј-ATG CTG GGA CCC GAC TAG CCA TTG-3Ј; and primer 3, 5Ј-ATG CTG GGA CCC GAC TAG CCA TTG GTT G-3Ј. The primers to reveal the template (bottom) strand starting at Ϫ58 were as follows: primer 1, 5Ј-CGG AAG GAA CTG CGC TTA CG-3Ј; primer 2, 5Ј-AAC CGG CCG CCA ATA AGG AAG GAT C-3Ј; and primer 3, 5Ј-CGG CCG CCA ATA AGG AAG GAT CGT CCG ATC-3Ј. The following annealing temperatures were used for both primer sets: 58, 63, and 68°C. All primers were ordered PAGE-purified from Integrated DNA Technologies. Products were resolved on a 6% polyacrylamide wedge gel with 7.75 M urea. Each PCR used 6 -9 g of DNA.
Nuclear Extract-Nuclei isolated from 2 g of rat liver were resuspended in 1 ml of PBS containing 3 mM MgCl 2 and were centrifuged at 3000 ϫ g for 5 min at 4°C. Nuclear pellets were resuspended in 0.5-1.0 ml of high salt buffer (420 mM NaCl, 20 mM HEPES, pH 7.9, 1 mM EDTA, 1 mM EGTA, 20% glycerol, 20 mM NaF, 1 mM Na 3 VO 4 , 1 mM Na 4 P 2 O 7 , 1 mM dithiothreitol, 0.5 mM phenylmethylsulfonyl fluoride, 1ϫ protease inhibitor mixture; Sigma). Nuclei were lysed by rotating slowly at 4°C for 30 min. The lysates were then centrifuged at 15,700 ϫ g for 15 min to pellet nuclear debris. The supernatant (nuclear extract) was collected and stored at Ϫ70°C until needed. Protein concentrations were determined using the BCA assay (Pierce).
Chromatin Preparation from Rat Liver-Rat liver (2.2 g) was placed in a beaker containing 10 ml of ice-cold PBS. The liver was minced into small pieces and diluted with an equal volume of 2% formaldehyde in PBS followed by a 10-min incubation at room temperature. Formaldehyde cross-linking was stopped by the addition of 2 ml of 1.25 M glycine. Liver pieces were washed three times with 10 ml of ice-cold PBS. Samples were then homogenized in 12 ml of nuclei isolation buffer ϩ Triton X-100 using a Teflon glass homogenizer in a drill press. Nuclei were isolated by centrifugation through dense sucrose. Nuclei were resuspended in 1-2 ml of PBS containing 3 mM MgCl 2 and centrifuged at 3,000 ϫ g for 5 min in 1.5-ml tubes. The nuclear pellet occupied an ϳ50-l volume and was white in color. Nuclei were then resuspended in 1 ml of nuclei lysis buffer (50 mM Tris, pH 8.1, 10 mM EDTA, 1% SDS, 1 g/ml leupeptin, 1 g/ml aprotinin, and 0.1 M phenylmethylsulfonyl fluoride). Sonication was accomplished with a Heat Systems-Ultrasonics W375 sonicater with a microtip. Ten 10-s bursts at a power setting of 4, 40% duty cycle, were sufficient to shear chromatin to an average size of 200 -600 bp. Shearing of chromatin was checked by electrophoresis on a 3% agarose gel. Chromatin concentrations were equalized to 0.3 mg/ml by A 260 and stored at Ϫ70°C until needed.
Chromatin Immunoprecipitation Assays-Chromatin immunoprecipitations were performed essentially as described previously (24). Briefly, chromatin suspensions were diluted 1:10 and pre-cleared with single-stranded DNA/protein A-agarose beads (Upstate). Each immunoprecipitation reaction received 15 g of chromatin and 5 g of the appropriate antibody: USF-2 (sc-862), CREB-1 (sc-187x), and phospho-CREB (sc7978r). All antibodies were polyclonal rabbit IgG from Santa Cruz Biotechnology. PCR was performed using the following primers Plasmid Construction-A PCR product from rat genomic DNA containing the HMG-CoA reductase promoter from Ϫ325 to ϩ70 was cloned into the PGL3 basic vector (Promega) using standard molecular biology techniques. Briefly, the 5Ј primer was designed to possess an overhang to introduce an MluI site by PCR, and the 3Ј overhang added an XhoI site. The resulting PCR product was digested and cloned into the pGL3 basic backbone. Mutant HMGR promoter-luciferase plasmids were generated using the QuikChange kit from Stratagene. Both mutants were verified by sequencing at the core sequencing facility at the Moffitt Cancer Center, University of South Florida, Tampa, FL. pRL-TK, a vector containing the Renilla luciferase gene driven by the thymidine kinase promoter was from Promega. (Primer sequences are available upon request.) Cell Culture-H4IIE cells (rat hepatoma) were purchased from the American Type Culture Collection. Cells were grown in Eagle's modified essential media supplemented with 10% fetal bovine serum, 100 units of penicillin/streptomycin per ml, and 1 mM sodium pyruvate. Cells were kept at 37°C and 5% CO 2 in a humidified incubator.
Transient Transfections-H4IIE cells were plated to an initial density of 100,000 cells per well in 24-well plates the day before the experiment. The following day, the media were removed, and the cells were washed one time with PBS. Cells were transfected with 1 g of DNA/well using Transfast reagent (Promega) in the recommended 2:1 ratio. Cells were co-transfected with reporter construct and pRL-TK in a 4:1 ratio. One h after transfection, the 200 l of transfection mix in each well was diluted with 800 l of growth media. 12-16 h later, cells were harvested in 100 l of passive lysis buffer and assayed for luciferase activity using the dual luciferase assay kit (Promega). Data are shown as the average ratio of firefly to Renilla luciferase counts Ϯ S.D. At least six independent yet identical transfections were performed per condition. All plasmid concentrations were checked by A 260 prior to transfection.

RESULTS
Nuclear run-on assays were performed to determine whether insulin acts to increase transcription of the HMG-CoA reductase gene. Inbred male Sprague-Dawley rats were injected with streptozotocin (65 mg/kg) to induce diabetes. Animals were sacrificed during the 3rd h of the dark cycle, at the diurnal high for hepatic HMG-CoA reductase expression. Nuclei were isolated from the livers of these animals, and nuclear run-on assays were performed as described under "Materials and Methods." HMG-CoA reductase transcription was greatly diminished in the diabetic animals (Fig. 1A). It was also found that administration of insulin to diabetic animals restored HMG-CoA reductase transcription to normal in just 2 h (Fig. 1B).
We next carried out in vivo footprinting to examine the occupancy of the hepatic HMGR promoter. Livers from normal and diabetic animals were treated with dimethyl sulfate and subjected to footprinting by ligation-mediated PCR. A primer set designed to reveal the top strand begins by reading cytosines at Ϫ185. Another primer set reveals the bottom strand by reading guanines beginning at Ϫ58. It should be noted that these primers were designed against the rat HMGR promoter sequence, which varies slightly (about 10 bp) from the hamster (10). Our primers were designed against sequence from the inbred Sprague-Dawley rats used in our experiments. Minor differences from the published rat sequence included an extra G at Ϫ15 and a reversal of the CG at Ϫ3 and Ϫ4. Numbering is therefore Ϫ1 bp relative to the previously published sequence, based on the transcription start site (Fig. 2).
On the top strand, several areas of DMS protection or enhancement were detected (Fig. 3). The Naked lane refers to DNA that was first extracted and purified and then treated with DMS in vitro. This lane is a control that shows all the reactive G residues in the sequence. Bands that are absent or reduced in intensity in the in vivo samples (Fig. 3, lanes  N and D) represent protections where protein binding shields the DNA from dimethyl sulfate attack. Bands that are significantly darker in the in vivo samples, or new bands that appear in these lanes, are known as enhancements. Enhanced DMS reactivity is indicative of protein binding in the nearby area, although generally not on that particular residue. In Fig. 3, the protections are noted with a filled triangle, and enhancements are marked with an open triangle. In both normal and diabetic samples, the CRE was completely protected at Ϫ101 and showed enhanced DMS reactivity at Ϫ104, Ϫ99, and Ϫ95/Ϫ94 (Fig. 3, right). This pattern was seen in all animals regardless of treatment. A significant protection seen only in normal animals occurred at Ϫ71, as shown in Fig. 3, right. Protections at Ϫ137 and Ϫ147 were not consistently observed.
On the bottom strand, the CRE is heavily protected at Ϫ100, Ϫ105, Ϫ103, and Ϫ109 (Fig. 4, bottom right) in both normal and diabetic animals. The A at Ϫ102 showed up as an enhancement in both cases. A key difference in the footprints is a very obvious enhancement at Ϫ142 seen only in the diabetic samples (Fig. 4, middle right). This particularly dark band, indicating enhanced DMS reactivity, was seen in 4/5 diabetic footprints and 0/5 normal footprints. Conversely, the nearby enhancement at Ϫ138 of the normal lane was not seen in diabetic footprints, suggesting possible competition for a binding site in this region. Another obvious difference is an enhancement at Ϫ161 in the diabetic lane (Fig. 4, top right). This residue is in the middle of the SRE located between Ϫ164 and Ϫ155. The SRE appears unoccupied under normal conditions and enhanced in diabetic samples (4/5 animals). Protections at Ϫ189/Ϫ190 were seen in all groups, whereas those at Ϫ70 were not observed in the diabetics. Both of these areas contain potential NF-Ybinding sites, with the sequence ATTGG.
Because there was an enhancement at Ϫ161 of the SRE in four of the five diabetic animals, we wondered if insulin activation could be a result of sterol regulation through the SRE. To investigate this possibility, we examined livers of rats fed lovastatin or cholesterol to alter liver cholesterol levels. Animals were fed 0.02% lovastatin or 1% cholesterol for 5 days. Previous research in our lab has shown that a similar dose of lovastatin elevates HMGR transcription 4 -6-fold (22), whereas dietary cholesterol reduces HMGR protein levels to about 1% of control. It should be noted that dietary cholesterol has only a minor repressive effect on the rate of HMGR transcription in these animals (26). We predicted that the lovastatin-fed animals would show strong protections at the SRE, because of elevated cleavage of SREBPs, induced by sterol deprivation. Curiously, no definitive protections or enhancements at the SRE were visible when animals were fed either lovastatin or cholesterol (Fig. 5, right). The footprints were the same for the two animals in each group. The CRE was also heavily protected in these animals but unchanged by either treatment. The enhancement at Ϫ138 seen in normal animals was also seen in both lovastatin and cholesterolfed rats. The enhancement at Ϫ142 seen in diabetics was noticeably absent from these animals. In addition, the NF-Y sites at Ϫ189/Ϫ190 and Ϫ70 are readily visible in this footprint and protected in both cases. A summary of the major differences between normal and diabetic footprints is presented in Fig. 6.
To identify some of the major factors bound to footprinted regions, we performed EMSAs with short oligonucleotide probes for these elements. Nuclear extracts prepared from normal and diabetic rats were used in this assay. By using the probe from Ϫ59 to Ϫ82, we observed strong binding that could be supershifted with two different NF-Y antisera but not with antisera to NF-1, an unrelated protein (Fig. 7A). This confirms the presence of another NF-Y-binding site in the HMGR promoter, in addition to the one previously described upstream of the SRE (9). NF-Y binding did not change in liver nuclear extracts, despite a marked difference in occupancy of this site in vivo (Fig. 6). A probe corresponding to the CRE was able to bind CREB-1 from normal and diabetic nuclear extracts (Fig. 7B). Overall binding to the CRE was not different between normal and diabetic extracts, as expected from the footprinting. Probes for the Ϫ119/Ϫ142 and Ϫ129/Ϫ152 regions showed strong binding of an unknown factor, whose in vitro binding ability was unaffected by diabetes (Fig. 7C).
In an effort to identify the factor binding to the Ϫ119/Ϫ142 region, we performed additional EMSAs. Fig. 8 shows an experiment in which a probe to the Ϫ119/Ϫ142 region was incubated in a binding reaction with normal rat liver nuclear extract (Fig. 8, 2nd lane). An unlabeled double-stranded oligonucleotide was added in 10-fold molar excess relative to the hot probe in the remaining lanes. The sequence at the top of Fig. 8 is the wild type promoter sequence, with an arrow highlighting the point mutation in each lane. The upper band in Fig. 8 can be efficiently competed away with the wild type competitor (3rd lane), and most of the mutant oligonucleotides. Noteworthy exceptions are "C_AG" from Ϫ137 to Ϫ134 and the entire sequence "GGGCGGTT" between Ϫ129 and Ϫ122. The presence of two binding sites explains why this band is seen as well in the overlapping Ϫ129/Ϫ152 probe, albeit with weaker signal. The sequence GGGCGGTT seems like a good match for Sp1, whose consensus sequence is "GGGGCGGGGC" with a strong requirement for the core GGGCGG (27). Curiously, neither of our Sp1 antisera were able to supershift this band, nor was antibody to Egr-1, another GC-box binding protein (data not shown). The lower band in the 1st lane of Fig. 8 was found to be nonspecific, as it was present regardless of probe used.   Given the strong protection of the CRE seen in all the in vivo footprints, and the ability of CREB to bind in vitro, we wanted to find out if CREB was in fact bound in vivo to the hepatic HMGR promoter. To accomplish this, liver sections from normal and diabetic rats were crosslinked with formaldehyde and subjected to chromatin immunoprecipitation analysis. CREB-1 antibody was able to pull down the HMGR promoter from both normal and diabetic chromatin, whereas an isotype-matched antibody to an irrelevant nuclear protein was not (Fig.  9A). The minor difference in intensity of this band between normal and diabetic samples was not reproducible, although CREB was clearly bound in both cases. This immunoprecipitation was not able to pull down exon 12 of the HMGR gene, confirming that DNA was sheared to an appropriate size. The inability of phospho-CREB to pull down the HMGR promoter could be a result of poor antibody-antigen interaction. We would expect that at least some of the endogenous CREB is phosphorylated. The data confirm reports that CREB is bound to the HMGR promoter in vivo (25) and validates this observation in the context of the live animal.
Another point of interest in the in vivo footprints was the consistent protection of the NF-Y site identified at Ϫ70. Because this site was strongly protected in 4/5 of the normal footprints, and none of the diabetics, we hypothesized that this element might play a critical role in activation of transcription. In order to investigate a possible functional role for this site, we constructed luciferase reporter plasmids containing the full-length HMG-CoA reductase promoter starting at Ϫ325 and  The number of times each protection or enhancement was observed is expressed as a fraction of the total animals in the group. In the sequence column, the sequence is given, showing the reactive residue that was detected. Possible factors that may bind are listed in the last column. Major differences are in A, and common elements are in B.  . EMSA analysis of Sp1-like factor binding to the ؊119/؊142 region. A short DNA probe corresponding to the footprinted region between Ϫ119 and Ϫ142 was used to detect in vitro binding ability in rat liver nuclear extracts. Competitor oligonucleotides were used in a 10-fold molar excess to the probe where indicated. The 1st lane (Ϫ) is a binding reaction without nuclear extract. The 2nd lane (none) is a binding reaction with rat liver nuclear extract but no competitor. The 3rd lane (w.t.) is a binding reaction with the wild type sequence used as a competitor. The remaining lanes are labeled with the wild type promoter sequence on top, and an arrow pointing to the point mutation in the competitor oligonucleotide.
ending at ϩ70 of the 5Ј-untranslated region. Two identical plasmids harboring mutations in the NF-Y site were also made. These plasmids were transfected into H4IIE cells, a rat hepatoma line. The cells were harvested and assayed for luciferase activity (Renilla luciferase was cotransfected for normalization purposes). As seen in Fig. 10, the wild type promoter shows a high level of activity relative to the vector backbone. Both of the mutants significantly inhibited luciferase production, indicating that this NF-Y site is required for efficient HMGR transcription.

DISCUSSION
These studies address the regulation of the HMG-CoA reductase promoter by insulin in live animals. Here we present the first evidence that insulin acts to directly increase transcription of the HMG-CoA reductase gene in rat liver. Diabetic rats have lower rates of HMG-CoA reductase transcription than normal rats. With only 2 h of insulin treatment, transcription was restored to normal. Previous work from our lab showed that the corresponding increase in mRNA could be accomplished even in the presence of cycloheximide (6). Taken together, these results suggest that insulin acts rather directly to stimulate the HMG-CoA reductase promoter and does not require protein synthesis.
In vivo footprinting revealed numerous protections and enhancements throughout the HMGR promoter. The most pronounced of these was at the CRE, which was occupied under all conditions tested. EMSA analysis confirmed that CREB-1 present in nuclear extracts from normal or diabetic rat livers could bind to this element in vitro, in agreement with observations in FRTL-5 cells (10). Chromatin immunoprecipitation analysis of rat liver confirms the previous finding that CREB is bound to the HMGR promoter in vivo (25). Given the overwhelming and invariant occupancy of the CRE in vivo, it seems unlikely that CREB binding is the regulated event in insulin activation. This may differ from sterol regulation in which SREBP binding has been shown to selectively recruit CREB to the promoter in CHO cells (28). Although we previously showed that the CRE was required for insulin activation of this promoter in H4IIE cells (7), in vivo occupancy of this site did not vary in rat liver. It is possible that insulin regulation in cultured rat hepatoma cells differs from the physiological regulation of this gene seen in whole animals. It is also possible that the CRE is necessary but not sufficient for insulin activation, inasmuch as it is required for maintaining an appropriate level of basal transcriptional activity.
The enhancement at Ϫ138 in normal footprints was not seen in diabetic samples. Given the GC-rich content of the nearby sequence, it is likely that this may be due to binding of an Sp1-related factor. In fact, strong in vitro binding activity was observed with the probe from Ϫ119 to Ϫ142. The sequence "gggcggctt" is a close match to the consensus binding sequence for Sp1. Although generally regarded as a more basal transcription factor, Sp1 has been invoked in the insulin regulation of several genes, including SREBP-1a (29).
Four of the five diabetic animals showed a particularly striking enhancement at Ϫ142. This enhancement was never seen under the other conditions examined, including cholesterol and lovastatin treatment. In addition, only diabetic samples showed a change in the DMS reactivity of the SRE. This enhancement at Ϫ161 of the SRE coincided with the enhancement at Ϫ142 and may be a result of binding of a repressive factor in the Ϫ161/Ϫ142 region. Binding of a factor in this region could distort the DNA in such a way that both Ϫ142 and Ϫ161 are more susceptible to dimethyl sulfate attack. In addition, this factor may preclude binding to the Sp1 site downstream of Ϫ138. This competition would explain why enhancements at Ϫ138 were only seen in the normal animals (due to binding at Sp1 sites) and the enhancement at Ϫ142 only in the diabetic animals.
When animals were fed either lovastatin or cholesterol, the footprint FIGURE 9. CREB is bound to the HMGR promoter in live animals. Liver pieces from normal (N) and diabetic (D) rats were cross-linked in 1% formaldehyde. Nuclei were isolated and lysed. Chromatin was sonicated to an average size of 200 -600 bp. 15 g of chromatin was immunoprecipitated (IP) with no antibody (beads), an irrelevant antibody (Irr. Ab) (USF-2), CREB-1 antibody, or phospho-CREB antibody. Forty cycles of PCR were performed, and products were run on a 2.2% agarose gel with ethidium bromide. A, PCR to detect the HMG-CoA reductase promoter. B, PCR to detect HMG-CoA reductase Exon 12. looked essentially identical to that seen in normal animals. This is peculiar because a similar dose of lovastatin was shown to elevate HMGR transcription 4 -6-fold (22). Three possible reasons for the lack of occupancy at the SRE come to mind. 1) SREBP binding has been proposed to be a rapid and transient event. It is known that SREBPs are by themselves weak binders of DNA, suggesting that nearby factors are needed to stabilize them (25). Therefore, SREBPs may not occupy their binding site in a large enough percentage of cells to show up in the footprint. 2) Lovastatin activation of this gene could occur through another site elsewhere in the promoter. This seems reasonable because HMGR is known to have a variant SRE, with lower affinity for SREBPs than those found in the low density lipoprotein receptor and HMG-CoA synthase. It has also been demonstrated that SREBP-1 can bind to the HMGR promoter at sites other than the SRE (11). 3) Perhaps not all of the cells in the liver respond to lovastatin. It has been shown previously by immunostaining of livers from rats fed mevinolin (lovastatin) that HMGR expression is clustered around the blood supply (30). Uneven distribution of the drug could mean that some cells show a drastic up-regulation of HMGR message, whereas a larger percentage remains unaffected. Because in vivo footprinting examines the net promoter occupancy of a population of cells, the effect could be muted even though mRNA is markedly higher. In either case, the data suggest that insulin activation of this promoter occurs by a mechanism distinct from sterol regulation. The protections at Ϫ71/Ϫ70 were not observed in footprints from diabetic animals. It was found that NF-Y from nuclear extracts could bind to this element. Curiously, in vitro binding of NF-Y from nuclear extracts did not vary with diabetes, suggesting that other factors may be necessary to stabilize its interaction with the promoter in vivo. This is the first identification of this NF-Y site in the HMGR promoter. It is also clear that this site is of particular functional importance. Both mutations to this site substantially decreased overall transcription in H4IIE cells. Given its proximity to the transcription start site, recruitment of this factor is probably a key event in insulin regulation of HMGR.
By using the information obtained from these experiments, we can construct a basic model of promoter occupancy in normal and diabetic rats (Fig. 11). CREB is bound to the promoter at the CRE in both normal and diabetic rats. The previously described NF-Y site around Ϫ189/ Ϫ190 is occupied in both situations. Sp1 or a related factor binds in the Ϫ119/Ϫ142 region, possibly accounting for the enhancement at Ϫ138 in normal rats. This binding is likely prevented by the presence of a repressive factor that occupies the region between Ϫ142 and Ϫ161 in the diabetic animals. The SRE is not protected to a detectable degree in the liver. Although it is likely that SREBP-2 participates to some extent in basal transcription, it is not known what percentage of endogenous SREBPs are bound to the hepatic HMGR promoter under normal con-ditions. The diagram also shows a newly identified NF-Y site at Ϫ71/ Ϫ70 that is preferentially occupied in the normal animals.
In summary, the areas of protection or enhancement identified in this study generally correspond with the large protected regions seen previously in DNase I footprinting studies of the hamster promoter (19). Previous in vivo footprints of the human promoter in HepG2 cells did not find any differences with insulin treatment, despite a 1.5-fold increase in mRNA (20). These previous in vivo studies also identified the SRE as a protected region, something that was not observed in rat liver. These reports were useful in both helping us design the experiments and in allowing us to compare results from cultured tumor cells to rat liver. Our work represents the first examination of the in vivo occupancy of the hepatic HMGR promoter in live animals. This is also the first demonstration that the CRE is occupied in vivo. We report a novel NF-Y site that is more often protected in normal animals and is required for efficient HMGR transcription. Most importantly, we have identified a few key areas where occupancy varies with diabetes, particularly Ϫ138, Ϫ142, and Ϫ161. This work will help focus future studies of insulin activation of this promoter and further our understanding of the transcriptional regulation of genes in the liver. Additional studies are needed to address the functional roles of these footprinted elements in the context of the live animal. Such investigations into insulin regulation of cholesterol biosynthesis are of paramount importance if we are to understand the links between diabetes and cardiovascular disease.