Identification of cis- and trans-acting factors regulating the expression of the human insulin receptor gene.

The functional organization of the human insulin receptor (hIR) promoter was analyzed by deletion mutagenesis and protein-DNA interaction studies. A series of deletion mutants was expressed transiently in two human hepatocytes, HepG2 and PLC. The results revealed that the promoter region between -692 and -345 is essential for efficient transcription of the hIR gene. Multiple trans-acting factors were identified by band shift and footprinting analyses. Sp1 binds to a cluster of GC boxes and two GGGAGG hexamers locating at -637 to -594. Adjacent to GC boxes, there are two regions, from -550 to -530 and from -522 to -503, which bind to two novel factors, IRNF-I and IRNF-II. These two factors are distributed differentially in different cell lines. Linker scanning mutations on GC, GA boxes, or the IRNF-I binding site significantly decreased the transcriptional activity, indicating that IRNF-I and Sp1 are important for hIR promoter activity. In addition, we demonstrated that glucocorticoid-dependent transcriptional induction of hIR mRNA in vivo is conferred by a glucocorticoid response element in the hIR promoter. Taken together, these results imply that transcription of the human insulin receptor gene is regulated by multiple protein-DNA interactions occurring within the defined promoter region.

The human insulin receptor gene has been cloned and characterized; it spans more than 120 kb' of genomic DNA and has 22 exons (11). By alternative splicing, two insulin receptor variants, differing by the presence or absence of 12 amino acids near the C terminus of the CY subunit, are generated in a tissue-specific manner (12). The promoter of the human insulin receptor gene has also been cloned and studied by several groups (11,13,14). It is extremely GC-rich and contains seven GC boxes (GGGCGG) which are putative binding sites for the mammalian transcription factor Spl (15). However, it contains neither a TATA box nor a CAAT box, reflecting the common features for the promoters of constitutively expressed genes (so-called housekeeping genes). Like other housekeeping promoters which lack TATA boxes, there are multiple transcription initiation sites within the first 300-bp GC-rich region (11,14).
Although the insulin receptor is ubiquitously expressed, albeit at low levels in almost all cells (16), there is growing evidence suggesting that the levels of insulin receptor can be regulated by a wide variety of factors (2) and under different environmental conditions (4). For example, glucocorticoids enhance transcription of the insulin receptor gene, while insulin down-regulates its own receptor through internalization (17)(18)(19). Also, the receptor mRNA levels were found to vary during differentiation, being much higher in differentiated mouse adipocytes as compared to undifferentiated mouse 3T3-Ll fibroblasts (20). In addition, developmental regulation of insulin receptor gene expression has been documented in Drosophila (21,22). However, the regulatory mechanisms controlling insulin receptor levels in the cells, especially at the level of expression, are poorly understood. Such information might be important for understanding the causes of impaired insulin action in various pathological states such as insulin resistance (23,24). Here, we have used mutagenesis and transfection studies to define the cis-acting elements in the 5"flanking sequence of the human insulin receptor gene that are important for high levels of expression and glucocorticoid responsiveness of this gene. In addition, we have identified multiple trans-acting factors, including Spl, glucocorticoid receptor, and two novel factors, IRNF-I (insulin receptor nuclear factor-I) and IRNF-11, binding to these cis-acting elements.

4639
-574 to -26 was generated by HindIII digestion. These promoter fragments and a wild type 1.8-kb promoter fragment, kindly provided by Dm. Graeme Bell and Susumu Seino (University of Chicago), were subcloned into the BglII site of the bacterial chloramphenicol acetyltransferase (CAT) reporter construct in pCAT3M (25) (Fig. 1, upper panel). A series of 3' deletion mutants, constructed in a similar way, was subcloned into pOVCAT-50 (26) where they were fused to a minimal promoter of the ovalbumin gene in a CAT reporter construct ( Fig. 1, lower p a n e l ) . Linker scanning mutations were constructed by recombining appropriate pairs of 5' and 3' fragments, each of which was generated by polymerase chain reaction. The resulting plasmids BglII linker (5'-CTAGATCTAT-3').
contain the replacement of 10 bp of natural sequences with a 10-bp CeU Culture and Transfection-Two human hepatoma cell lines (HepG2 and PLC) and one rat fibroblast line (Rat 208F) were obtained from the American Type Culture Collection (Rockville, MD) and maintained as monolayer cultures in minimum essential medium (GIBCO) supplemented with 10% fetal bovine serum, penicillin (100 pg/ml), and streptomycin (100 pglml). Transient DNA transfection was performed by the standard calcium phosphate precipitation method (27). Cells were harvested 2 days after transfection, and CAT assays were carried out as described by Gorman et al. (28). Conversion to acetylated [''C]chloramphenicol was quantified by liquid scintillation counting. The percent conversion was normalized for each experiment relative to the undeleted promoter. Each value shown is the mean of four to six separate experiments. For testing glucocorticoid receptor (GR)-dependent transcription, the GR expression vector (pRShGR) (29) was co-transfected with each of the 5' deletion constructs. In these experiments, medium was supplemented with lo-' M dexamethasone, if necessary.
Nuclear Extract-Nuclear extracts from HepG2 cells were prepared essentially as described by Dignam et al. (30). The protein concentration of each extract was determined by the Bradford assay (31). Aliquots of nuclear extract were stored at -80 "C.
DNase I Footprinting-Probes for footprinting were prepared as described in the previous section. DNA probe (20-40 fmol) was incubated in 20 pl of reaction mixture with 2 to 6 pg of nuclear extract or 0.3 to 0.6 pg of purified glucocorticoid receptor, 20 ng of poly(dI-dC), and the buffer used in the band shift assay. After a 15-min incubation at room temperature, 0.2 to 0.5 pg/ml DNase I was added at room temperature for 1 min. The reaction was stopped by the addition of 100 pl of stop solution containing 1 pg of pBR322,15 mM EDTA, 0.15% sodium dodecyl sulfate, and 1.5 pg of proteinase K. Samples were analyzed as described previously (32).
Methylation Interference-The 32P-labeled DNA fragment was partially methylated with dimethyl sulfate as described by Sakonju and Brown (33). Preparative band shift assays were performed with 10-20 ng of methylated probes (5-10 X 10' cpm) under the same conditions as described above. The protein-DNA complexes, as well as the free probe, were excised from the gel, extracted, and purified as described (34). The resultant DNA fragments having methylated purine residues were cleaved with 0.1 M NaOH in 20 mM NaPO, (pH 7.0) containing 1 mM EDTA. The samples were precipitated wlth ethanol and analyzed on a sequencing gel. Various deleted fragments were constructed as described under "Materials and Methods." Five 5' deletion mutants of the hIR promoter with a common 3' end point at position -26 were inserted into pCAT3M fused to a CAT reporter gene (upper panel). In a similar way, four 3' deletion mutants of the hIR promoter with a common 5' end point at position -1819 were inserted into pOVCAT-50 fused to the minimal promoter of the ovalbumin gene linked to a CAT reporter (lowerpanel). Each number in the name of the construct denotes the end point of the deletion.

Analysis
accordance with the convention used for genes having multiple transcription start sites, we designate the translation initiation site as +l.
A series of 5' deletion constructs of the hIR promoter was linked to the CAT gene and transfected into the human hepatoma cell line HepG2. As summarized in Table I, removal of the 1.1-kb promoter fragment spanning -1819 to -692 has little effect on the promoter efficiency. However, a greater than &fold decrease in promoter activity was observed when an additional 120 bp were deleted (phIRCAT-574), and further removal of 230 bp from -574 to -345 decreased transcription activity almost to the basal level. These results indicate that the proximal 692 bp upstream of the translation initiation site is important for transcription of the human insulin receptor gene and that the sequence spanning -692 to -574 contains a cis-acting element(s) which is essential for promoter/enhancer activity. Similar to the results obtained with HepG2 cells, transfection of these 5' deletion constructs into PLC cells, another human hepatoma cell line, showed that the deletion of sequences upstream of -692 has very little effect on the promoter activity (Table I). However, in contrast to the results from HepG2 cells, deletion of sequences between -574 and -345 had a more pronounced effect than deletion of the region between -692 and -574 in PLC cells. The relative activity of each construct in these two cell types in Table I represents the average of four to six separate experiments. As shown in Table I, all constructs except phIRCAT-574 showed comparable levels of CAT activity in both cell lines. Although the relative activity of each region is somewhat different in two cell lines, both sets of data suggest that the promoter region including the proximal 692 bp is required for the efficient transcription of the hIR gene.
In order to define the 3' border of the control region of the hIR gene, a series of 3' deletion mutants was generated and transiently transfected into HepG2 cells. Since these constructs lack a proximal promoter, they were inserted in front of a minimal ovalbumin promoter fused to the CAT reporter gene. The overall promoter activity is lower than that of the 5' deletion constructs, suggesting that the activity of the hIR promoter connected to the ovalbumin minimal promoter is lower than that of homologous promoter. Two constructs having deletion end points a t -772 and -692 produced no detectable CAT activity. However, the construct phIROVCAT-345 in which the most proximal region from -345 to -26 was deleted showed a similar level of activity as the wild type, and a slightly lower activity was found for the phIROVCAT-519. These results are consistent with those obtained from the 5' deletion analysis. Furthermore, they showed that the 345 bp proximal to the translation initiation site have no additional promoter function. Taken together, the results from both 5' and 3' deletion mutation analysis indicate that the promoter region spanning -692 to -345 contains sufficient information for efficient transcription of the human insulin receptor gene.

The tram-Acting Factor(s) Binding to the Region between -692 and -574
Is Related to Spl-Recently, the promoter region of the human insulin receptor gene was sequenced by Araki et al. (13) and Mamula et al. (14). Seven putative S p l binding sites, so-called GC boxes, are located in two clusters, -622 to -595 and -493 to -378. Both clusters are within the 350-bp functional promoter region defined above. T o determine whether they actually bind to specific protein(s), we used the 118-bp fragment from -692 to -574 as a probe and investigated the protein-DNA interaction(s) by band shift assays using HepG2 nuclear extract. Lane 2 of Fig The DNA probe used here is the 118-bp coding strand spanning -692 to -574. A, band shift assay. Each binding reaction in lanes 2-6 contained 0.3 ng of "P-labeled probe, 300 ng of pBR322/HinfI, and 3.3 pg of HepG2 nuclear extract and was incubated with reaction buffer as described under "Materials and Methods." For competition analysis, either unlabeled probe (lanes 3 and 4 ) or a promoter fragment of ovalbumin gene from -750 to -56 (lanes 5 and 6) was included in the reaction mixture prior to the addition of HepGP nuclear extract. The molar ratio of competitor used is indicated above each lane. Lane I shows protein-free probe, and complexes I, 11, 111, and IV are described in the text. B, DNase I footprinting analysis. As indicated, 0.5,1, or 2 pg of HepG2 nuclear extract was incubated with the same probe prior to DNase I digestion. The protected areas at the low and high protein concentrations are indicated by brackets and named FP1 and FP2, respectively. Arrowheads mark DNase I-hypersensitive sites observed at higher protein concentrations. Numbers on the right side refer to the nucleotide position upstream of the translation initiation site of the hIR gene. The molar ratio of competitor over the labeled probe is indicated. A G + A and G-only Maxam-Gilbert sequencing reaction of the same probe was used as a sequence reference (lanes I and 2). SV40, 42-mer oligonucleotide containing two 21-bp repeats of the SV40 early promoter; OV, ovalbumin promoter fragment spaning -750 to -56. Nucleotide sequence of the hIR promoter from position -638 to -594 was shown in the lower panel. Four GC boxes and two GA hexamers are underlined. Brackets indicate sequences protected by HepG2 nuclear extract at low (FP1) and high (FP2) concentrations.

Regulatory Elements
of hIR Gene 4641 111, and partially for the complex IV ( Fig. 2A, lunes 3 and 4).
In contrast, all four complexes could not be competed by a 75-fold molar excess of nonspecific competitor, ovalbumin promoter fragments ( Fig. 2A, lunes 5 and 6). These results show that protein factor(s) present in HepG2 nuclear extract generates specific complexes with the promoter fragment spanning -692 to -574. In order to map the boundary of the binding site of this factor(s), the same probe was subjected to DNase I footprinting analysis in the presence of increasing amounts of HepG2 nuclear extract. Binding activity was observed over a 40-bp region spanning -635 to -595 in which four overlapping GC boxes are located. The area from -620 to -602 encompassing three overlapping GC boxes was clearly protected at low levels of protein (Fig. 2B, lunes 5 and 6). With higher protein concentrations, the protected region was extended in both directions to encompass another GC box and two GGGAGG hexamers (lunes 7 and 8). In addition, hypersensitive sites were induced at -641 to -639 and also at -591 in the presence of higher levels of protein. The changes in footprint pattern seen with increasing amounts of nuclear extract suggest that multiple proteins with different affinities interact with the hIR promoter region between -692 and -574. To ascertain whether this binding activity is due to Spl, we performed competition experiments using a 42-mer oligonucleotide containing two 21-bp repeats of the SV40 early promoter as a competitor. Each 21-bp repeat contains two GC boxes, GC motifs 3 and 4 (35). Competition with 25-and 75-fold molar excesses of unlabeled competitor completely inhibited footprint formation on the GC boxes as well as on the two GA hexamers (lunes 9 and 10). No alterations in the protection pattern were observed even at a 75-fold molar excess of an ovalbumin promoter fragment (lanes 11 and 12). The detection of binding of Spl to the GGGAGG hexamers is not surprising, since the same hexamer located in the promoter of human low density lipoprotein receptor gene was previously shown to bind Spl (36). These results suggest that Spl binds to the four overlapping GC boxes and to the two GA hexamers. In contrast, the 105-bp fragment between -493 and -378 containing another cluster of GC boxes did not show any footprints or protein-DNA complex formation when tested (data not shown).

Factor(.$ IRNF-I and IRNF-11 Bind to the Promoter Region
Spanning -574 to -483-Since the region between -574 and -345 was also capable of promoting transcription of the insulin receptor gene, we also analyzed this region by protein binding assay. We initially performed a band shift assay with the 91-bp probe spanning -574 and -483 (Fig. 3A). Upon incubation with the HepG2 nuclear extract, two shifted bands, complex I and 11, were generated (Fig. 3A, lune 2 ) . The formation of complex I could be competed by inclusion of a 25-fold molar excess of unlabeled probe, but not by the addition of a 75-fold molar excess of ovalbumin promoter fragment (Fig. 3A, lunes 2-5). However, for complex 11, even a 75-fold molar excess of competitor only resulted in partial competition. These results suggest that at least two protein factors, present in HepGZ nuclear extract, generate two specific complexes with this hIR promoter region.
To determine the sites of protein-DNA interaction which might occur in this region, we performed DNase I footprinting assays. HepG2 nuclear extract protected two regions from DNase I digestion on the same probe (Fig. 3B). At relatively low levels of protein, binding activity was observed at the 3' end of the probe spanning -522 to -500 where two DNase I hypersensitive sites were detected at position -518 and -499 (Fig. 3B, lunes 2-5). An additional footprint became apparent at -548 to -530 when increasing amounts of nuclear extract were used (lunes 4 and 5 ) . Together the results from band shift assay and DNase I footprinting analysis imply that two distinct protein factors in HepG2 nuclear extract interact with two cis-acting elements located in the promoter region spanning -574 and -483.
Closer examination of the sequences encompassed by the two footprints indicated that these regions are highly conserved in human and mouse insulin receptor promoter (Fig.  40). This prompted us to investigate whether the two factors which bind to these regions are present in different cell types. Nuclear extracts from three other cell lines, HeLa, BHK (fibroblasts), and HIT (pancreatic p cell), were tested for the formation of specific protein-DNA complexes in the band shift assay (Fig. 3C). Interestingly, complex I was present in all cell lines except HeLa, while complex I1 was present in HepG2 and HeLa cells but not in BHK and HIT cells. This result suggests that two factors generating two complexes are differentially localized in different cell types. We designated the protein factor generating complex I as IRNF-I (insulin receptor nuclear factor-I) and the protein factor generating complex I1 as IRNF-11. In addition to complex I and 11, two slow migrating complexes were also detected in the other three nuclear extracts tested (Fig. 3C). Competition assays showed that both of them are specific complexes which might be generated by the interactions between protein factors in each extract and the hIR promoter fragment used as a probe (data not shown).
To elucidate the relationship between the two regions defined by footprinting analysis (Fig. 3A) and the two specific complexes generated in band shift assay (Fig. 3B), methylation interference experiments were performed to determine the exact protein-DNA contact points. Methylation at important purine contact points prevented complex formation. As shown in Fig. 3, 2 residues at position -514 and -513 of the lower strand are important for the formation of complex I1 (Fig. 4B), while 8 residues, 4 on the lower strand and 4 on the upper strand, are important for complex I formation (Fig.  4A). The location of these sequences are fully contained within the DNase I footprinting regions, indicating that the footprints observed over these sequences are a consequence of the binding of a specific protein factor(s). As summarized in Fig. 4C, contact points for complex I1 were located within the footprint 11, whereas all the contact points for complex I were exclusively localized in footprint I. Together, these results clearly illustrate the following relationship: IRNF-I binds to the promoter region from -550 to -530 and generates a slow migrating complex (I) in band shift assay, whereas IRNF-I1 binds to the region between -522 and -503 and generates a faster migrating complex (11). Fig. 4 0 shows the sequence conservation of the binding sites for IRNF-I and -11 of human and mouse insulin receptor promoters.

IRNF-I and Spl Binding Sites Are Functionally Important for Expression of the hIR
Gene-To substantiate that the protein-DNA binding sites we defined are important for hIR gene transcription, we constructed linker scanning mutations on the defined cis-acting elements and tested them for transcriptional activity (Fig. 5). Each linker scanning mutation is defined by two numbers which refer to the positions of the sequences immediately 5' and 3' to the BgZII linker. When the GA boxes or the GC boxes were mutated, in phIRLS-635/ -624 and phIRLS-618/607, respectively, the CAT activities were reduced at least %fold. More importantly, mutation of the IRNF-I binding site, in phIRLS-547/-536, drastically decreased the CAT activity to less than 10% of the original level. In contrast, the plasmid phIRLS-519/-508, with muta- tion of the IRNF-I1 binding site, is almost as active as the GR Binds to a Potential GRE in the hIR Promoter-It has wild type insulin receptor promoter. These results indicate been reported that glucocorticoids stimulate insulin receptor that IRNF-I and Spl binding sites are indeed important for biosynthesis in both hepatocytes and lymphocytes (18, 19, promoter activity and suggest that their binding proteins, 37). Recently, it was shown that the level of IR mRNA is IRNF-I and Spl, play important roles in the regulation of the increased by glucocorticoid treatment with no alteration in insulin receptor gene. mRNA stability or in transport activity from nucleus to A series of linker scanning constructs was transfected into HepG2 cells. Two numbers in each construct denoted the positions of the sequences immediately 5' and 3' to the BglII linker. They were analyzed for CAT activity as described by Gorman et al. (28), and the average activities were plotted.
cytoplasm (18). In order to test whether GR-dependent transcriptional induction occurs in cultured cells, transfection studies were performed using hIR 5' deletion constructs (Fig.  6A). A series of 5' deletion mutant constructs were co-transfected with or without GR expression vector pRShGR. The pPRE,TKCAT and pCAT3M reporters served as positive and negative controls. No difference in hIR promoter activity was detected in the presence or absence of dexamethasone in HepG2 cells (data not shown). Likewise, co-transfection of pRShGR into these cells had no effect on CAT activity. However, the hIR promoter activity was stimulated in the presence of M dexamethasone when transfected into the rat fibroblast line, Rat 208F. As shown in lanes 1 and 2 of Fig. 6A, undeleted construct phIRCAT-1819 showed a 3-to 5-fold induction. The other two mutant constructs containing promoter region downstream to the -772 or -692 also showed comparable levels of induction (lanes 3-6). These results imply that promoter region downstream to the -692 confers GR-dependent transcription.
While examining the sequence of the hIR promoter, we observed a sequence homologous to the consensus GRE at -359 to -345. Since the effects of glucocorticoids are known to be mediated by the interaction between glucocorticoid receptor (GR) and the GRE located in the target promoter (38), binding studies were performed. To test whether GR binds to this potential GRE, the 212-bp fragment from -483 to -271 was used in a DNase I footprinting analysis. We found that the region between -360 and -335 was selectively protected by the rat GR DNA binding domain expressed in Escherichia coli (Fig. 6B, lanes 5 and 6). To test the specificity of the protection, competition experiment was performed. An oligonucleotide containing the GRE sequence found in the TAT (tyrosine aminotransferase) gene was used as a specific competitor. DNase I protection was eliminated by 200-and 400-fold molar excesses of unlabeled probe (lanes 7-9), suggesting that the binding was specific for a GRE sequence. Similar results were obtained using the full length receptor (data not shown). Together, the results from transfection assay and DNase I footprinting analysis indicate that interaction between GR and the GRE located in the hIR promoter confers transcriptional responsiveness of the hIR gene.

DISCUSSION
To define the cis-acting elements controlling human insulin receptor gene expression, we have analyzed its promoter re- Transcriptional induction of the hIR gene by glucocorticoid. A, GR-dependent transcription of the hIR gene in rat 208F cells. A series of 5' deletion constructs described in Fig. 1 was transfected into Rat 208F cells together with GR expression vector (lanes 3-12) and analyzed for CAT activity. pSV,CAT was used as a positive control, and pCAT3M was used as a negative control. In addition, PREzTKCAT which contains two copies of a glucocorticoid thymidine kinase (TK) gene coupled to the CAT gene was used as a response element in front of the promoter of the herpes simplex virus reference for GR expression. + anddenote the presence or absence of dexamethasone M), respectively. B, DNase I footprinting analysis. The promoter fragment spanning -483 to -271 was endlabeled and incubated with either 0.3 or 0.6 pg of purified rat GR DNA binding domain. 31-mer oligonucleotides containing the GRE consensus sequence were used as a competitor in the amounts indicated above the lane. Footprinting reactions were performed essentially as described under "Materials and Methods." gion and identified cis-acting elements and the factors binding to them (Fig. 7). The promoter region spanning -692 to -345 was found to play a major role in transcription of the hIR gene in a human hepatoma cell line. At least four protein factors, Spl, glucocorticoid receptor, and two novel factors, IRNF-I and IRNF-11, interact with specific sequences within the defined promoter region, suggesting that these proteins play a major role in transcriptional regulation of the hIR gene.
In terms of functionally important promoter regions, our

4644
Regulatory Elements of hIR Gene The landmarks of the promoter region include four defined cis-acting elements: two GA hexamers, a cluster of GC boxes, two IRNF binding sites, and a putative GRE. Four conserved guanine residues in a putative GRE sequence are shown by asterisks.
data agree with the results of Seino et al. (11) in which the 600-bp 5' flanking region of the hIR gene was shown to be sufficient for transcription. However, our results are at variance with the results of Araki et al. (13) or McKeon et al. (39) who showed that the region between -574 and +1 or -532 and -392, respectively, is sufficient for promoter activity. Since HepG2 cells were used for transfection in these two cases as they were here, the reason behind this discrepancy is not clear. Interestingly, the relative activities of each region within the defined 350 bp seem to be very different in two human hepatoma cell lines. In PLC, the more proximal region harboring the two IRNF binding sites was much more important, whereas the more distal region containing four GC boxes was required for transcription in HepG2 cells. Although both of them are human hepatoma cell lines, PLC has been intensively studied with respect to the integration of the hepatitis B virus genomes in the cellular DNA. HepG2 line, however, retains a large number of liver-specific phenotypes and harbors neither hepatitis B surface antigen nor integrated viral DNA (40). In light of these differences in cellular physiology, it is conceivable that the concentration of the factors, Spl and IRNFs, which were shown to bind to each cis-acting element, might be different in these two cell lines. Alternatively, other positive or negative factors in these two cell lines might alter the transcription levels of the hIR gene through differential protein-protein interactions.
In this study, we found that Spl binds to one cluster of GC boxes in the hIR gene promoter. Three GC box-binding proteins have been characterized Spl ((15), MTF-1 (41), and LSF (42). Spl was initially isolated by Tjian and his colleagues (15,43) and characterized as a mammalian transcription factor that stimulates transcription by binding to GC box sequences in several viral and cellular promoters. Although the initial report postulated the hexanucleotide GGGCGG as a consensus binding sequence, several recent reports suggested that the recognition sequence is flexible. Two decanucleotide sequences which differ by two nucleotides from the consensus GC box were shown to bind Spl and to activate transcription of the human immunodeficiency virus (44). More recently, the hexanucleotide GGGAGG in the human low density lipoprotein receptor promoter, which differs by one nucleotide, was shown to bind Spl and act as a positive transcription element (36). This observation suggests that not only the four overlapping GC boxes but also the two adjacent hexanucleotides, GGGAGG, in the hIR promoter have a potential to bind Spl. A zinc-inducible factor, MTF-l, also binds to a GC box in the mouse metallothionein-I gene and activates its transcription (41). The third GC-box binding protein, LSF, has been purified from HeLa cells and shown to bind to the GC motifs 2 and 3 and LSF-280 site of the SV40 promoter (42). Although we have presented no direct evidence excluding the possible involvement of factors such as MTF-1 or LSF in complex formation, Araki et al. (45) reported that partially purified LacZ-Spl hybrid proteins can bind to the cluster of GC boxes in hIR promoter. Here we show that the linker scanning mutation of GA hexamers or GC boxes has similar effects on the activity of the hIR promoter. These results suggest that binding of Spl to GA as well as GC boxes must play an important role in transcription of the hIR gene. Surprisingly, another cluster of GC boxes, located at -437 to -401 of the promoter, showed no detectable protein binding, suggesting that the GC boxes in this region might not be functional. The factor(s) providing selective functional activity on the two clusters of GC boxes is not clear. However, it might involve the relationship between Spl and other tramacting factors occupying hIR promoter, such as IRNFs or GR (Fig. 7). It has been well documented that single DNA elements evoke differential regulatory effects depending upon their physiological contexts (46).
Using nuclear extract of HepG2 cells, we have also detected protein binding sites within a 91-bp promoter region, about 100 bp downstream of the functional GC boxes. Footprinting analysis revealed the protection of sequences AGATCCG-CGCCGCCTTTTCCCGCG and CTCCCGGGCGCAGAGT-CCCT. A computer search revealed that these sequences have no significant homology with other known cis-acting elements, suggesting that they might be binding sites for novel transcription factors. We named these new factors IRNF-I and IRNF-11. Sequence comparison of the only two available promoters of insulin receptor gene showed that the binding sites for IRNF-I and IRNF-I1 are highly conserved between the human and mouse (Fig. 40), supporting the importance of IRNFs in controlling insulin receptor gene expression. Therefore, it was surprising that only IRNF-I plays a crucial role in transcription in HepG2 cells. However, the differential distribution of IRNF-I and IRNF-I1 in different cell types (Fig. 3C) implies that these two transcription factors may play different roles in different cell types. In this regard, it is interesting to point out that Tewari et al. (47) reported that the hIR promoter between -852 and -570 is more important in HepG2 than in CV1 cells. We also found that phIR-CAT constructs had relatively higher activity in hepatocyte cell lines, such as HepG2 and PLC, than in other cell lines, such as fibroblasts (Rat 208F) and kidney (CV1) (data not shown). Additional experiments are being conducted in our laboratory to investigate the role of IRNF-I in the regulation of hIR gene expression in different cell types.
Glucocorticoid treatment has been observed to increase insulin receptor mRNA levels in uiuo. In this study, we have shown that purified GR binds to a potential GRE located in the proximal promoter region of the hIR gene. Although the sequence is not completely homologous to the consensus GRE (48), all 4 guanine residues previously shown to be essential for receptor binding (48,49) are conserved (Fig. 7). Upon binding to a GRE, GR induces target gene transcription by enhancing recognition of the promoter by other factors in the transcriptional machinery (50), or possibly by interactions with other protein factors (51,52). We have shown that glucocorticoids are capable of inducing transcription of the hIR gene in transiently transfected cells. Binding studies using purified GR support the idea that glucocorticoids indeed employ this induction on the hIR gene through the interaction with the GRE. The absence of detectable hormonal induction in HepG2 cells is possibly due to the strong basal activity of the hIR promoter in these cells which might mask the stimulation by glucocorticoid. In HepG2 and PLC cells, the CAT activity generated by the hIR promoter is comparable to that by pSV2CAT which has the SV40 promoter and enhancer. In fibroblasts, however, the hIR promoter activity is lower than that of pSVzCAT (Fig. 6B, lanes 1,2, and 8). Alternatively, it is also possible that the interactions between GR and other cellular factors are different in different cells, such as the interaction of GR with AP-1. In summary, (Fig. 7), we have demonstrated that the 350bp promoter region spanning -692 to -345 is sufficient for the regulation of transcription of hIR gene. Protein-DNA interactions occurring within the defined promoter region appear to be required for promoter function. In addition to Spl (or an Spl-like factor) and GR, we have identified two novel trans-acting factors IRNF-I and IRNF-I1 which interact with their respective cis-acting elements in the defined promoter region. Examination of the function of these cis-and trans-acting elements in vivo is currently underway.