Transcriptional regulation of the rat insulin-like growth factor-I gene involves metabolism-dependent binding of nuclear proteins to a downstream region.

Insulin-like growth factor-I (IGF-I) gene transcription is mediated largely via exon 1. In an initial search for regulatory regions, rat hepatocytes were transfected with IGF-I constructs. Since omission of downstream sequences led to reduced expression, we then used in vitro transcription to evaluate potential metabolic regulation via downstream regions. With templates including 219 base pairs of downstream sequence, transcriptional activity was reduced 70-90% with hepatic nuclear extracts from diabetic versus normal rats. However, activity was comparable with templates lacking downstream sequences. The downstream region contained six DNase I footprints, and templates with deletion of either region III or V no longer provided reduced transcriptional activity with nuclear extracts from diabetic rats. Nuclear protein binding to regions III and V appeared to be metabolically regulated, as shown by reduced DNase I protection and activity in gel mobility shift assays with nuclear extracts from diabetic rats. Southwestern blotting probes corresponding to regions III and V recognized a approximately 65-kDa nuclear factor present at reduced levels in diabetic rats. These findings indicate that a downstream region in exon 1 may be important for both IGF-I expression and metabolic regulation. Altered concentration or activity of a transcription factor(s) binding to this region may contribute to reduced IGF-I gene transcription associated with diabetes mellitus.

Insulin-like growth factor-I (IGF-I) gene transcription is mediated largely via exon 1. In an initial search for regulatory regions, rat hepatocytes were transfected with IGF-I constructs. Since omission of downstream sequences led to reduced expression, we then used in vitro transcription to evaluate potential metabolic regulation via downstream regions. With templates including 219 base pairs of downstream sequence, transcriptional activity was reduced 70 -90% with hepatic nuclear extracts from diabetic versus normal rats. However, activity was comparable with templates lacking downstream sequences. The downstream region contained six DNase I footprints, and templates with deletion of either region III or V no longer provided reduced transcriptional activity with nuclear extracts from diabetic rats. Nuclear protein binding to regions III and V appeared to be metabolically regulated, as shown by reduced DNase I protection and activity in gel mobility shift assays with nuclear extracts from diabetic rats. Southwestern blotting probes corresponding to regions III and V recognized a ϳ65-kDa nuclear factor present at reduced levels in diabetic rats. These findings indicate that a downstream region in exon 1 may be important for both IGF-I expression and metabolic regulation. Altered concentration or activity of a transcription factor(s) binding to this region may contribute to reduced IGF-I gene transcription associated with diabetes mellitus.
The insulin-like growth factors (IGFs) 1 are polypeptides with sequence, structure, and biological actions similar to those of insulin (1). Since circulating levels of IGF-I are more responsive to changes in metabolic status than are levels of IGF-II (2, 3), IGF-I is thought to be a more important regulatory factor during postnatal life. While IGF-I is expressed in many organs and tissues, consistent with paracrine regulation and a role as a local growth factor, its expression is 50 -100 times higher in the liver than in other tissues, consistent with hepatic origin of circulating IGF-I, and a role as an endocrine regulator of growth (1,3,4). In the liver, IGF-I expression appears to be regulated pretranslationally (5-7). Modulation at the level of gene transcription is indicated by findings such as decreased IGF-I gene transcription in streptozotocin-diabetic animals (7) and the ability of insulin to stimulate IGF-I gene transcription in hepatocyte primary culture (8). However, underlying mechanisms are poorly understood.
The single rIGF-I gene gives rise to a complex family of mRNAs with both size and coding sequence heterogeneity (9 -11) and polypeptides which are products of multiple translational initiation sites. Multiple in-frame initiator codons within 5Ј sequences specify different amino-terminal signal peptides, and preprolGFs with signal peptides containing 22, 32, or 48 amino acids are synthesized depending on utilization of different AUGs (9, 12). Initiation of transcription is also complex, as several laboratories have identified multiple transcription initiation sites in exons 1 and 2 of the rat, sheep, and human genes (12-16). In the rIGF-I gene, initiation sites extend over 140 bp in exon 1, and 770 bp in exon 2 (13). However, Adamo et al. (13) found that two initiation sites in exon 1 could account for 70 -80% of IGF-I gene transcription in adult rat liver. Thus, although rat IGF-I gene transcription is regulated by two distinct promoters, the exon 1 promoter appears to be dominant.
Since relatively little is known about molecular regulation of IGF-I gene transcription, we have focused on the liver; while several laboratories have begun to study the basis of IGF-I gene transcription in different immortal cell lines (14,(17)(18)(19), there has been little evaluation of underlying mechanisms in the dominant source of IGF-I production (1, 3, 4). Analysis of findings from several laboratories suggests that downstream regions may play a role in IGF-I gene expression (17-20), but such an hypothesis has not been tested in the liver. In the present study, we demonstrate that sequences downstream from the exon 1 major transcription initiation site are important both for hepatic IGF-I expression and metabolic regulation, we characterize nuclear protein binding to downstream sequences, and we identify two regions that may be involved in the decreased IGF-I gene transcription associated with diabetes mellitus.

MATERIALS AND METHODS
Chemicals-Restriction endonucleases and DNA modifying enzymes were obtained from New England BioLabs (Beverly, MA); streptozotocin from Pfanstiehl (Waukegan, IL); [-32 P]ATP (6000 Ci/mmol), [␣-32 P]dATP, and [␣-32 P]dGTP (800 Ci/mmol) were from Amersham Corp. Oligonucleotides were from Operon Technology (Alameda, CA); luciferase and luciferin were from Boehringer Mannheim. Superscript RNase H Ϫ reverse transcriptase was from Life Technologies, Inc., and DNase I was from Pharmacia Biotech Inc. All other chemicals (of molecular biology grade) were purchased from Sigma.
Animals-Male Sprague-Dawley rats (Charles River, Lexington, MA), weighing approximately 120 -160 g, were fed ad libitum. Chronic diabetes was produced through tail vein injection of streptozotocin 140 mg/kg. Animals were sacrificed by cervical dislocation 5-7 days later, and livers were used for nuclear extract preparation immediately. Streptozotocin at 250 mg/kg was used to produce acute diabetes, with sacrifice of animals two days after injection.
Transfection-Constructs with IGF-I sequences cloned into a luciferase reporter (p0Luc) were generously provided by Dr. Peter Rotwein from Washington University, as reported previously (14). Relative to the rIGF-I exon 1 major transcription initiation site (181 bp upstream from the leader sequence (7, 13)), constructs used for transfection extended from Ϫ4398 to Ϫ32 bp; Ϫ1859 to Ϫ32 bp; Ϫ1262 to Ϫ32 bp; Ϫ1859 to ϩ55 bp; and Ϫ1859 to ϩ180 bp. Hepatocytes were isolated from 150 -200-g male Sprague-Dawley rats using a modification of the collagenase perfusion method of Seglen (21), as reported previously (5,8). Cells were transfected with 6 g of supercoiled DNA calcium-phosphate precipitate in a volume of 0.3 ml/plate 5 h after plating (22). Following overnight exposure to the DNA, the cells were rinsed twice with 4 ml of phosphate-buffered saline and placed in 3 ml of serum-free defined medium containing 10 Ϫ7 M insulin and 500 ng/ml human growth hormone, since both insulin and growth hormone increase IGF-I secretion in primary cultures of hepatocytes (23). After 24 h, the cells were rinsed twice with 4 ml of phosphate-buffered saline and scraped into 0.7 ml of 50 mM Tris-MES, pH 7.8, containing 1 mM DTT and 1% Triton X-100. Following centrifugation, 50 l of the supernatant was assayed for luciferase activity in a 200-l reaction containing 50 mM Tris-MES, pH 7.8, 10 mM magnesium acetate, 2 mM ATP, and 0.5 mM luciferin. Standard curves were performed with purified luciferase to ensure linearity. In all experiments, transfections were performed in triplicate plates for each condition.
Rat Liver Nuclear Extract Preparation-Liver nuclear extracts were prepared as described by Gorski et al. (24) and Triezenberg et al. (25). Briefly, 15 g of tissue were homogenized in 120 ml of buffer containing 10 mM Hepes, pH 7.6, 25 mM KCl, 0.15 mM spermine, 0.5 mM spermidine, 1 mM EDTA, 10% glycerol, and 2.0 M sucrose. The homogenate was layered onto a 2 M sucrose cushion and centrifuged at 27,000 rpm in an SW28 rotor at 4°C for 1 h. The nuclear pellet was resuspended in lysing buffer containing 10 mM Hepes, pH 7.6, 100 mM KCl, 3 mM MgCl 2 , 0.1 mM EDTA, and 10% glycerol at a concentration of 10 A 260 /ml. Nuclei were lysed by adding one-tenth of a volume of 4 M (NH 4 ) 2 SO 4 , and chromatin was removed by centrifugation at 39,000 rpm in an SW40 rotor for 2 h. Nuclear proteins were concentrated by (NH 4 )2SO 4 precipitation (0.33 g/ml) and dialyzed against buffer containing 25 mM Hepes, pH 7.6, 100 mM KCl, 0.1 mM EDTA, and 10% glycerol for 4 h. Proteinase inhibitors (0.1 mM phenylmethylsulfonyl fluoride, 0.1 mM benzamidine, 1 g/ml pepstatin A, 1 g/ml leupeptin, and 1 g/ml aprotinin) were added to buffers just before use. Extracts prepared from animals with either acute or chronic diabetes had comparable activity with in vitro transcription assays.
Construction of 3Ј Deletion Mutants-For mutants with deletion of ϩ84 to ϩ152 bp, the wild type template was linearized with BstXI, treated with Bal 31 nuclease, and then religated. Other mutants were constructed by polymerase chain reaction, similar to methods described by Higuchi et al. (26). Primers used are in Table I. In each case, a Ϫ309/ϩ373-bp polymerase chain reaction product with specific internal deletions was digested with BanII and BglII, gel purified, and then subcloned into a wild type template. All mutants were sequenced.
In Vitro Transcription Assay-Transcription reactions (30 l) contained 1.0 g of linear DNA template, 50 ng of pAML(C 2 AT) 190 , 60 g of liver extract, 50 mM KC1, 6 mM MgCl 2 , 0.5 mM ATP and CTP, 35 M UTP, 10 Ci of [␣-32 P]UTP (400 Ci/mmol), 0.1 mM 3Ј-O-methyl GTP, 10% glycerol, I u/l RNasin, 0.05 mM EDTA, and 1 mM DTT. The DNA template and extract were incubated on ice for 20 min. Transcription was then initiated by the addition of nucleotides and carried out for at 30°C for 45 min. The transcripts were purified by phenol/chloroform extraction, ethanol precipitated, separated on an 8 M urea, 6% polyacrylamide gel, and visualized by autoradiography. When transcription assays were performed in the presence of 0.5 mM of ATP, UTP, GTP, and CTP, the DNA template was degraded by 50 units of RNase-free DNase I as described previously (7), and the RNA synthesized in vitro was quantitated by primer extension.
Primer Extension-Oligonucleotides complementary to the sequence from ϩ42 to ϩ61 or ϩ79 to ϩ101 were end-labeled by polynucleotide kinase to a specific activity of 10 8 cpm/g. Probe (50 ϫ 10 4 cpm) was annealed to RNA synthesized in vitro, to 10 g of yeast tRNA, or to 10 -30 g of liver RNA at 42°C for 3 h. The cDNA was synthesized as described by Boorstein and Craig (27) at 42°C for 1 h, ethanol precipitated, and separated on an 8 M urea, 6% polyacrylamide gel.
Gel Mobility Shift Assay-End-labeled 272-bp AccI/BglII fragments were incubated with 0.5-15 g of nuclear extract in 25 l of binding buffer containing 10 mM Tris, pH 7.5, 50 mM KCl, 1 mM EDTA, 0.5 mM DTT, 0.2% Nonidet P-40, 20 g of bovine serum albumin, 4 g of poly(dI-dC), and 10% glycerol at 25°C for 25 min. Protein-DNA complexes were separated from free probe at 4°C on a 5% polyacrylamide gel in 0.5 ϫ TBE (45 mM Tris, pH 8.0, 45 mM boric acid, 1 mM EDTA) at 11 V/cm for 2-3 h, and visualized by autoradiography. For doublestranded oligonucleotides, the probe was incubated with 5-15 g of extract in same buffer, except that 200 g/ml of salmon sperm DNA and 24 g/ml of pBR322 were included to reduce nonspecific binding.
DNase I Protection Assay-End-labeled 272 bp AccI/BglII fragments were incubated with 4 -24 g of nuclear extract in 25 l of binding buffer containing 10 mM Hepes, pH 7.9, 50 mM KCl, 1 mM EDTA, 0.5 mM DTT, 1 g of poly(dI-dC), and 10% glycerol at 25°C for 25 min. An equal volume of binding buffer containing 10 mM MgCl 2 , 2 mM CaCl 2 , and 10 units/ml of DNase I was added, and the sample was incubated at 25°C for 2 min. The reaction was stopped with buffer containing 40 mM FIG. 1. Luciferase activity in cultured normal rat hepatocytes transfected with constructs containing rIGF-I sequences in a reporter vector (see "Materials and Methods"). Sequences are shown relative to the exon 1 major transcription initiation site. Promoter activity was expressed relative to activity with p0Luc. Results shown are representative of three experiments. Separate studies indicated that transfection efficiency (measured according to expression of CMV-␤-galactosidase) was unaffected by the presence of IGF-I elements. Mean Ϯ S.E., n ϭ 3.

TABLE I
Oligonucleotides used to construct mutants with deletions in regions III, IV, and V Oligonucleotides mF corresponding to Ϫ309/Ϫ285, and mR corresponding to ϩ349/ϩ373 were used as upstream and downstream primers, respectively. The sequences corresponding to message strand are designated as "F," to anti message strand as "R." For each construct, PCR involves primers mR and d-F along with mF and d-R, followed by PCR of the products with mF and mR. Oligos Downstream Regulation of rIGF-I Transcription EDTA and 10 g/ml tRNA and deproteinized with plenol. DNA was precipitated with ethanol, electrophoresed on an 8 M urea, 6% polyacrylamide gel, and visualized by autoradiography. Southwestern Blotting-20 g of extract were mixed with an equal volume of loading buffer containing 5 mM Tris, pH 6.8, 200 mM DTT, 5% SDS, 20% glycerol, and 0.05% pyromin Y. Proteins were denatured at 100°C for 3 min, separated on a 10% SDS-polyacrylamide gel, and then blotted onto nitrocellulose paper (29). The blots were incubated with buffer containing 10 mM Hepes, and 5% nonfat dry milk at 25°C for 1 h, and then incubated with binding buffer containing 10 mM Hepes, pH 8.0, 50 mM NaCl, 1 mM EDTA, 1 mM DTT, 0.25% nonfat dry milk, 5 g/ml salmon sperm DNA, and 1 ϫ 10 6 cpm/ml probe at 25°C for 1 h. After washing with 2 changes of binding buffer containing 300 mM NaCl at 25°C for 2 h, the filter paper was air dried and subjected to autoradiography.

Sequences Downstream from Exon 1 Transcription Initiation Sites Are Important for IGF-I Gene Expression and Metabolic
Regulation-Constructs with IGF-I 5Ј-flanking sequences were transfected into rat hepatocytes in primary culture, as summarized in Fig. 1. Relative to expression of a promoterless luciferase reporter vector (p0LUC), expression was decreased 30% by the presence of IGF-I sequence extending from ϳϪ4 kilobase pairs to Ϫ32 bp relative to the exon 1 major transcription initiation site (7,13,14). Expression was increased 40 and 113% with constructs containing 1.86 and 1.26 kilobase pairs of 5Ј sequence, respectively, and the same 3Ј terminus. Compared with expression of p(Ϫ1859/Ϫ32)LUC (40% above that of p0LUC), increased expression was obtained with constructs containing additional downstream sequences. Expression was increased 126% above p0LUC with a construct containing 1.86 kilobase pairs of 5Ј sequence and 3Ј sequence terminating at ϩ55 bp, and maximum expression, 230% above p0LUC, was obtained with a construct containing the same 5Ј sequence and 3Ј sequence extending to ϩ180 bp; expression was significantly greater than that of the construct with similar 5Ј sequence but lacking downstream sequence (p Ͻ 0.005). Thus, these findings suggested that downstream sequences enhanced IGF-I gene expression.
Because of relatively low expression in transient transfection studies (possibly due to the difficulty in maintaining IGF-I gene expression in hepatocytes (30) as well as the difficulty of transfecting cells in primary culture (31)), a different model was used to test the hypothesis that downstream sequences contribute to metabolic regulation of IGF-I gene transcription. A genomic IGF-I template containing 471 bp of upstream sequence and 219 bp of downstream sequence was incubated with nuclear extracts from the livers of normal and diabetic rats, and in vitro transcriptional activity was evaluated by primer extension. As shown in Fig. 2A, the dominant transcription initiation site in vitro was identical to that used in vivo. The signal originated from RNA polymerase II transcripts, since it was sensitive to ␣-amanitin (not shown). Moreover, no signal was detected when extracts were incubated in the absence of a DNA template, indicating that signals originated from tran-  5 and 7) rats, and the in vitro transcripts were quantitated by primer extension; 30 g of total liver RNA from normal (lane 2) and diabetic (lane 3) rats were used as positive controls, and yeast tRNA (lane 8) was used as a negative control. f174 DNA digested by HinfI was used as a size marker (lane 1). Panel B, a genomic IGF-I fragment from Ϫ471 to ϩ3 bp was subcloned to pUC13(C 2 AT) and used as a template in in vitro transcription assays, and a template for the adenovirus major late promoter (pML(C 2 AT) 190 , AMLP) was used as an internal control. Panel C, relative transcriptional activities of normal and diabetic liver nuclear extracts. Data were obtained from three different pairs of normal and diabetic extracts; for each pair of extracts, IGF-I transcriptional activity of normal liver extracts (determined by densitometric scanning and expressed relative to adenovirus major late promoter activity) was designated as 100%. Mean Ϯ S.E., * indicates p Ͻ 0.05.

FIG. 3. Binding of nuclear proteins to the 272-bp AccI/BglII
(؊54/؉219) fragment. End-labeled probe (5000 cpm) was incubated with liver extract at 25°C for 25 min. Protein-DNA complexes I and II were visualized on a 5% polyacrylamide gel as described under "Materials and Methods." scripts generated in vitro (not shown). Since in vitro transcriptional activity for the adenovirus major late promoter template driven by nuclear extracts from the livers of diabetic rats was comparable or greater to that with extracts from normal rats (Fig. 2B), we concluded that nuclear extracts from the livers of diabetic rats contained adequate transcriptional machinery and that changes in IGF-I gene transcription were likely to be specific. Using templates containing downstream sequence, in vitro transcriptional activity of nuclear extracts from the livers of diabetic rats was reduced 90%, compared with nuclear extracts from the livers of normal rats, (p Ͻ 0.05), but transcriptional activity with an IGF-I template lacking downstream sequence was not significantly decreased with extracts from diabetic versus normal rats (p Ͼ 0.1, Fig. 2C). Since this observation is consistent with our previous finding (7), that IGF-I gene transcription rates were reduced ϳ97% in the livers of diabetic rats as compared with normal rats, downstream sequences may be important for both IGF-I expression and metabolic regulation.
Nuclear Protein(s) Binding to Downstream Sequences-To search for regions that might be involved in gene regulation, the binding of hepatic nuclear factors to the 272-bp AccI/BglII fragment (Ϫ54/ϩ219 bp) was first studied by gel mobility shift analysis, as shown in Fig. 3. Densitometric scanning revealed that the intensity of shifted protein-DNA complexes was reduced 30 -60% with extracts from streptozotocin-diabetic as compared with normal rats. Binding was specific, since formation of DNA-protein complexes could be competed with an unlabeled 272-bp fragment but not with pBR322 DNA (not shown). All protein-DNA binding studies were repeated with at least three different batches of extracts, and extract activity was monitored by in vitro transcription assays.
Protein Binding Sites Assessed by DNase I Footprinting-Protein binding sites were determined on both coding and noncoding strands, and results are shown in Fig. 4. Region I corresponded to the major transcription initiation site in exon 1. Five additional protected regions were observed consistently, with footprints at ϩ17/ϩ25 (region II), ϩ42/ϩ68 (region III), ϩ79/ϩ101 (region IV), ϩ129/ϩ152 (region V), and ϩ155/ϩ169 (region VI). Binding of factors in nuclear extracts from the livers of diabetic as compared with normal rats was reduced in both region III (especially with the noncoding strand) and region V (especially with the coding strand) (lanes 4, 5, and 6  versus lanes 1, 2, and 3 in panels A and B; differences in binding were not consistent in other regions. Nucleotide sequences and protected regions are summarized in Fig. 5. Regions III and V Are Necessary for the Diabetes-associated Reduction in IGF-I Gene Transcription-The importance of FIG. 4. DNase I protection assay. An end-labeled 272-bp AccI/BglII (Ϫ53/ϩ219 bp) fragment was incubated with normal or diabetic liver extract at 25°C for 20 min and then digested with DNase I as described. Protected regions determined according to Maxam-Gilbert sequencing are shown as boxes, with naked DNA digested with DNase I as a negative control (C). pBR322 plasmid DNA digested with HpaII was used as a size marker (M). Panel A, probe labeled on coding strand; Panel B, probe labeled on noncoding strand. downstream regions in metabolic regulation was evaluated with in vitro transcription assays using deletion mutants as templates, as shown in Fig. 6A. A template lacking regions IV and V (ϩ84/ϩ152) no longer provided reduced transcriptional activity with nuclear extracts from diabetic versus normal rats (panel B). Similar results were also obtained with templates lacking regions III (ϩ42/ϩ68) or V (ϩ129/ϩ152) (panel C). In contrast, a template with deletion of region IV continued to exhibit decreased transcriptional activity with nuclear extracts from diabetic versus normal rats (panel D). While the potential involvement of other regions is still being investigated in our laboratory, these data were reproducible with different batches of extracts, and the transcriptional activities of normal and diabetic extracts remained comparable as determined with the adenovirus major late promoter as template.
Nuclear Factors Associated with Regions III and V Are Reduced by Streptozotocin-induced Diabetes-Double-stranded oligonucleotides corresponding to regions III (ϩ42/ϩ68 bp) and V (ϩ129/ϩ152 bp) were used in gel mobility shift analyses to examine DNA-protein interactions, as shown in Fig. 7. Addition of pBR322 DNA was necessary to decrease nonspecific binding (lanes 2 and 10 versus lanes 4 and 12). The association of nuclear factors with region V was stronger than that with region III, especially in formation of complex II (lanes 3 and 4 versus lanes 11 and 12). While formation of both complexes I and II could be competed with unlabeled oligonucleotides (lane 3 versus lanes 5 and 6 and lane 11 versus lanes 13 and 14), cross-competition was incomplete (lane 3 versus lanes 7 and 8  and lane 11 versus lanes 15 and 16). Thus, binding of nuclear factors to regions III and V was relatively specific, particularly for complex II. Activities of nuclear extracts from the livers of normal and diabetic rats are shown in Fig. 8. Nuclear proteins associated with region IV showed similar affinity with normal and diabetic rat liver extracts. In contrast, nuclear protein binding to regions III and V was reduced with extracts from diabetic animals, typically 30 -50% of that of extracts from normal rats.
Identification of Protein Factors Associated with Regions III and V-To characterize the size and relative abundance of proteins associated with regions III and V, hepatic nuclear extracts from normal and diabetic rats were subjected to polyacrylamide gel electrophoresis and then blotted to nitrocellulose and probed with corresponding oligonucleotides, as shown in Fig. 9. Proteins with apparent molecular weight of ϳ65 kDa were associated with both regions III and V and were present in extracts from both normal and diabetic animals. However, apparent abundance of the ϳ65-kDa protein in extracts from diabetic animals was ϳ75% of normal with region III and FIG. 6. In vitro transcription assay with deletion mutants as templates. Panel A, summary of deletion mutant constructs. Regions of DNase I protection are indicated, and deleted regions are shown by a thin line. The right side of the panel shows the relative transcriptional activities of diabetic rat liver nuclear extracts as compared with those from normal animals for each template. Panel B, relative transcriptional activities of normal (lanes 3 and 5) and diabetic (lanes 4 and 6) nuclear extracts with both wild type DNA (lanes 3 and 4) and a deletion mutant (lanes 5 and 6) as template. Lane 1 contained 20 pg of liver RNA; lane 2 contained tRNA. In vitro transcripts were quantitated by primer extension as described under "Materials and Methods." Sequencing reactions utilized an oligonucleotide complementary to ϩ141 to 161 bp electrophoresed along with primer extension products on a 6% polyacrylamide-8 M urea gel. Panel C, transcriptional activities of normal (lanes 1, 3, and 5) and diabetic (lanes 2, 4, and 6) nuclear extracts with both wild type DNA (lanes 1 and 2) and mutants lacking region III (lanes 3 and 4) or region V (lanes 5 and 6) as templates. Lane 7 contained 10 g of liver RNA. pBR322 DNA digested with HpaII was used as a size marker (M). A 23-bp oligonucleotide complementary to ϩ79/ϩ101 (region IV) was used as a primer to quantitate in vitro transcripts. Sequencing reactions were performed using the same primer, but with a construct lacking region III d(42-68) as a template. Both sequencing reactions and primer extension products were electrophoresed on a 10% polyacrylamide-8 M urea gel. IGF-I cDNA is indicated with an arrow. Panel D, transcriptional activities of normal (lanes 3 and 5) and diabetic (lanes 4 and 6) nuclear extracts with both wild type DNA (lanes 3 and 4) and a mutant lacking region IV (lanes 5 and 6) as templates. Liver RNA (10 g, lane 2) and yeast tRNA (lane 7) were used as positive and negative controls for primer extension. pBR322 plasmid DNA digested with HpaII was used as a size marker (lane 1). A 20-bp oligonucleotide complementary to ϩ42/ϩ61 was used as a primer to quantitate in vitro transcripts. IGF-I cDNA is indicated with an arrow. ϳ50% of normal with region V, as determined by densitometric scanning.

DISCUSSION
The IGF-I promoters analyzed to date have several common features, such as lack of a "TATA" box, presence of transcription "initiator" sequences (13, 32), and binding sites for well recognized transcription factors such as Sp1, C/EBP, and HNF-1 located upstream from the major transcription initiation sites (32). The present studies demonstrate that sequences downstream from the major transcription initiation site in exon 1 are important for both IGF-I gene expression and metabolic regulation. Within the Ϫ54/ϩ219 bp region of exon 1, we found six loci of binding with hepatic nuclear factors; protected regions were similar to those described by Thomas et al. (33). With our model, DNase I footprinting and gel mobility shift assays revealed that nuclear factors in the livers of diabetic rats have reduced interactions with region III (ϩ42/ϩ68) and region V (ϩ129/ϩ152). Transfection studies revealed a 230% increase in expression with a construct containing 180 bp of downstream sequence (including both regions III and V). Our findings are consistent with those of Hall et al. (14), who observed that the presence of downstream sequence increased IGF-I gene expression when the same constructs were transfected into SK-N-MC cells and Lowe et al. (18) and Adamo and co-workers (19) who found that downstream sequence increased IGF expression when constructs were transfected into C6 glioma cells. Further evidence of biological significance was provided by in vitro studies; specific differences in IGF-I transcriptional activity between normal and diabetic rat liver extracts could be detected only in the presence of downstream sequences.
The two protein-DNA complexes observed in gel mobility shift assays with oligonucleotides III and V likely result from binding of multiple nuclear factors rather than formation of a dimer, since only complex I could be cross-competed with both oligonucleotides III and V. A putative common factor could interact with motifs such as CCTGC(G/C)CA found within both regions III and V. In both gel mobility shift assays and Southwestern blotting studies, the formation of protein-DNA complexes could be competed with unlabeled oligonucleotides but was not blocked with a great excess of pBR322 DNA, indicating that binding was specific.
The DNA-binding protein(s) identified by Southwestern blotting appears to be metabolically regulated, as reduced binding was provided by hepatic nuclear extracts from diabetic as compared with normal rats. While gel mobility shift assays point to the presence of at least two DNA-binding proteins, we do not yet know if other putative factors are metabolically regulated as well. Lack of identification of a second DNA-binding factor by Southwestern blotting may also be attributed to the denaturing conditions used in this procedure, which could interfere with protein-protein interactions that may be stabilized by the caging effect in gel mobility shift analysis.
A number of viral and cellular transcriptional units contain essential sequences which are downstream from transcription initiation sites (34 -37). Such downstream elements may influence RNA elongation, processing, and translation, in addition to transcription initiation. Promoters commonly associated with housekeeping and growth control genes often require downstream elements to achieve full gene expression (38). Moreover, intragenic enhancers or activators have been described for numerous extrahepatic genes such as immunoglobulins (39), adenosine deaminase (40), and muscle creatine kinase (41). Thus, the requirement for both upstream and downstream elements to achieve full gene expression is not unique to the IGF-I gene.
There has been relatively little characterization of tissuespecific and hormone response elements of the IGF-I gene.
Since this manuscript was submitted, Nolten et al. (42) recently found that C/EBP and HNF-1 can stimulate hIGF-I gene ex- FIG. 7. Binding of nuclear proteins to oligonucleotides III and V. Each binding reaction contained 6 g of nuclear protein from normal rat liver, 5000 cpm probe, pBR322 plasmid DNA or/and cold double stranded oligonucleotides as indicated. Protein/DNA complexes were separated on a 6% polyacrylamide gel.