A consensus insulin response element is activated by an Ets-related transcription factor.

Insulin increases expression of somatostatin-chloramphenicol acetyltransferase (CAT) constructs 10-fold and thymidine kinase-CAT constructs 5-fold in GH4 cells. These responses are similar to our previously reported data on insulin-increased prolactin-CAT expression. They are also observed in HeLa cells and are thus not cell type specific. The evidence suggests that the insulin responsiveness of these genes is mediated by an Ets-related transcription factor. First, linker-scanning mutations and/or deletions of the prolactin, somatostatin, and thymidine kinase promoters suggest that their insulin responsiveness is mediated by the sequence CGGA. This sequence is identical with the response element of the Ets-related transcription factors. Second, CGGA-containing sequences placed at -88 in the delta MTV-CAT reporter plasmid conferred insulin responsiveness to the mammary tumor virus promoter. Third, expression of the DNA-binding domain of c-Ets-2, which acts by blocking effects mediated by Ets-related transcription factors, inhibits the response of these promoters to insulin. Finally, the Ets-related proteins Sap and Elk-1 bind to the prolactin, somatostatin, and thymidine kinase insulin-response elements. An Ets-like element was found in all insulin-sensitive promoters examined and may serve a similar function in those promoters.

The mechanisms involved in the regulation of gene expression by insulin are not well characterized. Insulin-induced alterations in the steady-state levels of numerous mRNAs have been documented (1). For several genes including phosphoenolpyruvate carboxykinase (2), glyceraldehyde 3-phosphate dehydrogenase (3), growth hormone (4), and prolactin (5), it has been established that these alterations are due to effects of insulin on the rate of transcription and not to effects on mRNA half-life or processing (1).
The effects of hormones on transcription are mediated through response elements in the promoters of genes. Several response elements may exist for a particular hormone that differs slightly in sequence or orientation, perhaps to allow for fine tuning of the hormonal response or for interaction with different tissue-specific factors. However, these elements are sufficiently similar to be recognized by the hormone-activated transcription factors and are thus said to form a consensus response element. The consensus hormone response element for cAMP-activated genes is the sequence TGACGTCA (6), while the thyroid-retinoid response element is a direct repeat of the sequence AGGTCA with varying numbers of intervening bases that determine hormone receptor specificity (7). A specific sequence comprising an insulin response element has been identified only for a small proportion of insulin-responsive genes. Comparison of these insulin response elements has not revealed sequence homologies that could constitute a consensus insulin response element (8).
The transcription factors that mediate responses to insulin have not been characterized. The putative insulin response element of the glyceraldehyde 3-phosphate dehydrogenase gene was identified by binding of an insulin-regulated protein to a specific sequence in the glyceraldehyde 3-phosphate dehydrogenase promoter. A protein that binds to this sequence has been cloned and is identical to the product of the testis determining gene, SRY (9). However, its role in the activation of glyceraldehyde 3-phosphate dehydrogenase gene expression by insulin was not further established.
We previously identified an insulin response element in the prolactin promoter (8). This insulin response element overlaps the cAMP response element of the prolactin gene TGACGGAA. However, mutagenesis and deletional analysis revealed that the insulin response element was separable from the cAMP response element and consisted of a direct repeat of the sequence CGGAAA. This sequence is identical to sequences that bind the Ets family of transcription factors. These studies report the identification of insulin response elements in the somatostatin and herpes simplex virus thymidine kinase genes that, together with the previously identified insulin response element of the prolactin gene, define a consensus insulin response element. The consensus sequence is identical to known binding sites for Ets-related transcription factors, and an Etsfactor inhibitor was found to inhibit insulin activation of all three promoters. The Ets-related proteins Sap-1 and Elk-1 specifically bind to these sequences.

EXPERIMENTAL PROCEDURES
Materials-[ 32 P]dCTP, 3000 Ci/mmol, and [ 14 C]chloramphenicol, 50 mCi/mmol, were obtained from ICN Biochemical Corp. Reagents and Taq polymerase for polymerase chain reaction were from Cetus. All other enzymes and linkers were obtained from either New England Biolabs or from Boehringer Mannheim and, unless otherwise indicated, were used under conditions recommended by the suppliers. Oligonucleotides were purchased from Operon. Duplex poly(dI⅐dC) was obtained from Pharmacia Biotech Inc. Reagents used for gel electrophoresis were purchased from Fisher. Acetyl-CoA and silica gel plates for thin layer chromatography were obtained from Sigma. Dulbecco's modified Eagle's medium containing 4.5 g/l glucose (DMEM) 1 was from Life Technologies, Inc., and iron-supplemented calf serum was obtained from Hyclone Laboratories. Triton X-100 and BCA reagent were from Pierce. All other reagents were of the highest purity available and were obtained from Sigma, Behring Diagnostics, Bio-Rad, Eastman, Fisher, or Boehringer Mannheim.
Plasmids-The construction of pPrl-CAT plasmids containing Ϫ173/ ϩ75 of prolactin 5Ј-flanking DNA was described (5). The vector pBLCAT2 (10) was obtained from G. Schutz (Institut fur Zell-und Tumorbiologie, Heidelberg, Federal Republic of Germany). A plasmid containing the 5Ј-region of the prolactin gene from the Sprague-Dawley rat was the generous gift of Dr. J. Gorski (University of Wisconsin) (11). The plasmid (AP1) 3 CAT was the gift of Dr. T. Curran (Roche Institute, Nutley, NJ) (12). This plasmid contains three concatamerized AP1 sites from the human metallothionine IIA gene fused to the SV40 early promoter. The ␣(Ϫ846/ϩ44)CAT was from Dr. J. L. Jameson (Case Western Reserve, Cleveland, OH) (13). This plasmid contains Ϫ846 bp of the human glycoprotein hormone ␣ gene that contains two cAMP response elements. The plasmid Pit-1(Ϫ738/ϩ11)CAT, containing two cAMP-responsive sequences (Ϫ210/Ϫ142), was described previously (14). Deletion and linker-scanning mutants of the thymidine kinase promoter were a generous gift of Dr. S. L. McKnight (Tularik, South San Francisco) (15). These were recloned into pUC 8 as described (16). The plasmid TK(Ϫ95/ϩ56⌬BamHI)CAT was made by digesting TK(Ϫ95/ϩ56)CAT with BamHI to linearize the plasmid at the 5Ј-end of the thymidine kinase promoter. This was then treated with mung bean nuclease to create flush ends, and the plasmid was recircularized by blunt end ligation. The plasmids TK(Ϫ385/ϩ56)CAT, TK(Ϫ218/ ϩ56)CAT, and TK(Ϫ128/ϩ56)CAT were constructed from TK(lsϩ5/ ϩ15⌬BamHI)CAT. First, TK(lsϩ5/ϩ15)CAT was digested with BamHI, blunt ended with mung bean nuclease, and religated to form TK(lsϩ5/ ϩ15⌬BamHI)CAT. HindIII-AccI (TK(Ϫ385/ϩ56)CAT), HindIII-PstI (TK(Ϫ218/ϩ56)CAT), and HindIII-NspI (TK(Ϫ128/ϩ56)CAT) deletions were made from this plasmid. The plasmids SS(Ϫ71/ϩ80)CAT and SS(Ϫ48/ϩ80)CAT were gifts of Dr. R. H. Goodman (Oregon Health Sciences University, Portland, OR) (6). SS(Ϫ71/Ϫ1)CAT and SS(Ϫ48/ Ϫ1)CAT were prepared by polymerase chain reaction using the SS(Ϫ71/ ϩ80)CAT as a template. The universal primer served as the sense primer. The antisense primer used was (5Ј-GCTCTCGAGCTCTC-CACGGTCTCC-3Ј). This adds an XhoI restriction site at Ϫ1 of the somatostatin promoter. The polymerase chain reaction product was digested with XbaI (there is an XbaI site in SS(Ϫ71/ϩ80)CAT in the polylinker) and XhoI and then ligated into XbaI-XhoI digested and dephosphorylated pBLCAT3. The plasmid SS(Ϫ71/ϩ80Etsmut)CAT was prepared by ligating an 82-bp duplex oligonucleotide with G3 C point mutations of each of the three potential Ets-response elements in the 5Ј-untranslated region of the somatostatin gene into the XhoI site of SS(Ϫ71/Ϫ1)CAT. These mutations converted three GGA motifs into GCA sequences that are not Ets binding sites. The plasmid SS(Ϫ71/ Ϫ1,ϩ23/ϩ80)CAT was made by deleting ϩ1/ϩ22 of SS(Ϫ71/ϩ80Etsmut)CAT with XhoI/PstI. This results in a plasmid in which the first two potential Ets-binding sites have been deleted and the third (ϩ43/ ϩ46) has a G3 C mutation at ϩ45. The plasmid SS(Ϫ71/Ϫ1,ϩ7/ ϩ47)CAT was prepared by ligating a duplex oligonucleotide to the sequence ϩ7/ϩ47 (containing the putative insulin response element) into the XhoI site of SS(Ϫ71/Ϫ1)CAT. All plasmid constructs were verified by sequencing. A human insulin receptor expression plasmid, pRT3HIR-2 was generously provided by Dr. J. Whittaker (State University of New York at Stony Brook) (17).
Transient Gene Expression Facilitated By Electroporation-Electroporation experiments and CAT assays were performed as described (18). GH4 cells were plated in DMEM, which contained 10% hormonedepleted, iron-supplemented calf serum (5), for 24 h and harvested with an EDTA solution, and 20 -40 ϫ 10 6 cells were used for each electroporation. All electroporations contained 5 g of the plasmid pRT3HIR-2, which expresses high levels of the human insulin receptor (18). This is necessary to achieve the high levels of insulin stimulation seen in these studies and is consistent with numerous other systems where cotransfection of receptors has been necessary to achieve physiological regulation of transfected genes (18). Trypan blue exclusion before electroporation ranged from 95 to 99%. The voltage of the electroporation was 1550 volts. This gives trypan blue exclusion of 70 -80% after electroporation. The transfected cells were then plated in multiwell dishes (Falcon Plastics) at 5 ϫ 10 6 cells per 9 cm 2 tissue culture well in DMEM with 10% hormone-depleted serum. Insulin was added at 1 g/ml to the appropriate flasks. Cells were refed with DMEM with 10% hormone-depleted serum Ϯ insulin at 24 h. After 48 h, the flasks were washed three times with normal saline and frozen. The cells were harvested in hypotonic lysis buffer using a rubber policeman. CAT enzyme activity was assayed as described (19).
Assay of DNA-Protein Binding by Gel Electrophoresis-Nuclear extracts used for gel mobility shift experiments were prepared from GH4 cells as described (20). Nuclei were disrupted with 400 mM KCl in a buffer containing 20% glycerol, 25 mM Tris, pH 8, 10 mM ␤-mercapto-ethanol, 0.5 mM EDTA, and 0.05% Triton X-100. It was then dialyzed and stored at Ϫ70 in the same buffer containing 100 mM KCl. Sap-1, Elk-1, and SRF proteins were produced using an in vitro transcription/ translation system (Promega). The plasmids T7 SAP1A (21), T7 ELK 1-428 (22), and T7 SRF ATG (23) were used in the reactions, which were performed as recommended by the manufacturer.

Insulin Increases Expression from the Somatostatin and the
Thymidine Kinase Promoters-The observation that the insulin and cAMP response elements of the prolactin gene were composed of overlapping sequences (8) suggested that other cAMP-responsive genes might be insulin responsive. A comparison of such sequences might suggest a consensus insulin response element. We therefore tested a number of cAMP-responsive promoter-CAT constructs in GH4 cells to determine if these were also insulin responsive (Fig. 1). The Pit-1(Ϫ738/ ϩ11)CAT, ␣(Ϫ846/ϩ44)CAT, and (AP1) 3 CAT are not significantly affected by insulin incubation (Fig. 1). However, somatostatin(Ϫ71/ϩ80)CAT expression increased 8-fold in response to insulin, and TK(Ϫ95/ϩ56)CAT expression increased 6-fold in response to insulin.

FIG. 1. Effect of insulin on cAMP-responsive promoters in GH4
cells. GH4 cells were cotransfected with 10 g of the CAT construct indicated in the figure and 5 g of pRT3HIR-2. Following a 24-h incubation without hormone, insulin was added at 1 g/ml. After an additional 24 h of incubation, the cells were harvested, and CAT enzyme was assayed. The average % acetylation in the control and insulinincubated cells was determined, and the results from the insulin-incubated cultures were compared with control levels to determine the fold stimulation (Fold-Control). The results are from three separate experiments done in duplicate. Basal CAT expression/g of protein was 0.052 Insulin also activates expression of Prl(Ϫ173/ϩ75)CAT, SS(Ϫ71/ϩ80)CAT, and TK(Ϫ95/ϩ56)CAT in HeLa cells (Fig.  2). Low levels of Prl(Ϫ173/ϩ75)CAT expression are increased 16-fold in HeLa cells in response to insulin. SS(Ϫ71/ϩ80)CAT is increased 10-fold and TK(Ϫ95/ϩ56)CAT is increased 4-fold in HeLa cells in response to insulin. As in GH4 cells, the expression of CAT is not increased by insulin using plasmids containing the Pit-1 promoter, glycoprotein hormone ␣-subunit promoter, or the repeated AP1 element. These data indicate that the effects of insulin to increase transcription of these three genes is not unique to the GH cells.
Identification of the Insulin-responsive Sequence in Thymidine Kinase Constructs-A panel of deletion and linker-scanning mutants was used in GH4 cells to identify the insulin response element of the thymidine kinase promoter. Deletion of 5Ј sequences to Ϫ46 (TK(Ϫ46/ϩ56)CAT, Fig. 3) reduced the insulin-increased expression of CAT to 3-fold but did not eliminate it. The 3Ј-deletion plasmid TK(Ϫ725/Ϫ7)CAT (Fig. 3) was stimulated 6-fold by insulin, but the plasmid TK(Ϫ725/ Ϫ16)CAT was not stimulated by insulin. Thus, it appeared that the insulin-responsive sequences were likely located between Ϫ46 and Ϫ7. However, linker-scanning mutants to this region (TK(lsϪ46/Ϫ36)CAT, TK(lsϪ42/Ϫ32)CAT, TK(lsϪ28/Ϫ18)CAT, TK(ls-21/Ϫ12)CAT, and TK(lsϪ16/Ϫ6)CAT) (Fig. 3) showed a loss of insulin-increased CAT expression only with the plasmid TK(lsϪ28/Ϫ18)CAT. This mutation deletes the TATA element and was reported to reduce basal expression from this promoter (15). Further, this deletion also eliminates the cAMP-mediated increase in thymidine kinase-CAT expression despite the absence of a cAMP response element. For example, the construct TK(lsϪ7/ϩ3)CAT is stimulated 9.9 Ϯ 2.3-fold by cAMP, while cAMP increases expression only 1.4 Ϯ 0.29-fold using TK(lsϪ28/ Ϫ18)CAT. Thus, this sequence does not likely represent the insulin-responsive element of the thymidine kinase gene. Comparison of the thymidine kinase gene with the insulin response element of the prolactin gene identified a sequence at Ϫ151/ Ϫ146 that contained the sequence CCGGAA. This sequence is present in all of the 3Ј-deletion mutants. The 5Ј-deletion mutants lack the putative wild type insulin response element, but they all have a BamHI restriction site containing the sequence CGGA at their 5Ј-ends. The linker used to generate the linkerscanning mutants is also a BamHI restriction sequence that has been shown to act as a response element for the Ets-related transcription factors (24). If the BamHI site confers insulin responsiveness to the 5Ј-thymidine kinase-CAT deletion plasmids, then mutation of the BamHI site upstream of Ϫ95 in the plasmid TK(Ϫ95/ϩ56)CAT should eliminate the insulin responsiveness of this plasmid. This BamHI site (CGGA) is removed in the plasmid TK(Ϫ95/ϩ56⌬BamHI)CAT, and CAT expression is not increased by insulin using this reporter (Fig.  3). Thus, the insulin response element of the 5Ј-deletion mutants is the BamHI linker.
To further confirm that the BamHI linker sequence used to make the linker-scanning plasmids was sufficient to mediate the effects of insulin on thymidine kinase-CAT expression, the linker sequence from the plasmid TK(lsϪ21/Ϫ12)CAT between Ϫ22 and Ϫ2, containing the CCGGAA motif, was cloned into ⌬MTV-CAT in both the normal and inverted orientation to create ⌬MTV(TKb)CAT and ⌬MTV(TKbi)CAT (Fig. 4, top). Insulin did not affect CAT expression from ⌬MTV(TKa)CAT and ⌬MTV(TKai)CAT that contain the Ϫ22/Ϫ2 sequence from the wild type thymidine kinase gene in both the normal and inverted orientation (Fig. 4, bottom). In contrast, insulin increases CAT expression 6-fold in ⌬MTV(TKb)CAT and ⌬MTV(TKbi)CAT that contain the CGGA sequence.
The deletion mutants shown in Fig. 5 address the possibility that the insulin response element of the native thymidine kinase promoter is the CCGGAA sequence located at Ϫ151/ Ϫ146. The linker-scanning plasmid TK(lsϩ5/ϩ15)CAT was used to make 5Ј-deletions of the thymidine kinase promoter since the 3Ј-deletion plasmids all have a 3Ј-CGGA sequence as a result of their construction. First, the BamHI site in the linker was removed with mung bean nuclease to create the plasmid TK(lsϩ5/ϩ15⌬BamHI)CAT. The remaining plasmid contains the thymidine kinase promoter between Ϫ725 and ϩ56. Insulin increases CAT expression 5-fold using this plasmid. Deletion to Ϫ385 (TK(Ϫ385/ϩ56)CAT, Fig. 5) did not reduce the increase in CAT expression due to insulin. A further deletion to Ϫ218 (TK(Ϫ218/ϩ15)CAT, Fig. 5) eliminates 5 potential Ets-binding sites including 2 CGGA sequences between Ϫ385 and Ϫ250, but this did not reduce the insulin-mediated increase in CAT production with this construct. Finally, a deletion to Ϫ128 (TK(Ϫ128/ϩ56)CAT, Fig. 5), which eliminates the CCGGAA at Ϫ151/Ϫ147, renders the thymidine kinase promoter insensitive to insulin.
Identification of the Insulin-responsive Sequence of the Somatostatin Promoter-Analysis of several deletion constructs suggests that the insulin response element of the somatostatin gene may be located in the 5Ј-untranslated region (Fig. 6A). Deletion of 5Ј sequences to Ϫ48 were previously shown to reduce the cAMP responsiveness of this promoter by eliminating part of the cAMP response element (6). However, this deletion had no effect on insulin-increased CAT activity (Fig.  6A, SS(Ϫ48/ϩ80)CAT). The 5Ј-untranslated region of the somatostatin gene is a GA-rich area that contains a CGGA sequence at ϩ43/ϩ46 that was similar to the insulin response element of the prolactin gene and of the insulin-responsive 5Ј-deletion of the thymidine kinase promoter. A somatostatin-CAT construct lacking this region SS(Ϫ71/Ϫ1)CAT was not stimulated by insulin (Fig. 6A). When the three potential Etsbinding sites were mutated in the plasmid SS(Ϫ71/ϩ80Etsmut)CAT, the increase in CAT expression mediated by insulin was reduced by 75% (Fig. 6B). Deletion of the first two potential Ets-binding sites in the plasmid SS(Ϫ71/Ϫ1,ϩ23/ϩ80)CAT did not further reduce the effect of insulin (Fig. 6B). This implies that the effect of insulin is predominantly mediated by the CGGA sequence at ϩ43/ϩ46. Finally, the putative insulin response region (ϩ7/ϩ47) of the somatostatin gene was added back to the non-insulin-responsive plasmid SS(Ϫ71/Ϫ1)CAT to produce the plasmid SS(Ϫ71/Ϫ1,ϩ7/ϩ47)CAT. This plasmid exhibits a 7-fold increase in response to insulin (Fig. 6B).

Insulin Stimulation of Gene Expression in GH Cells Is Ets Transcription Factor Mediated-
The DNA-binding domain of the Ets-related transcription factors is highly conserved, and the Ets-related transcription factors bind to each other's recognition sequences with only slightly different affinities. Thus, overexpression of the DNA-binding domain of any Ets-related protein will function as a dominant negative inhibitor of all related transcription factors. The overexpression of the DNA binding domain of c-Ets-2 has been adapted to this purpose (25).  (7) was used to construct a hybrid promoter with sequences from the thymidine kinase promoter or from a linker-scanning mutant of the thymidine kinase promoter. ⌬MTV-CAT contains 1200 bp of the mammary tumor virus long terminal repeat linked to the CAT structural gene terminated with an SV40 polyadenylation sequence. The glucocorticoid enhancer region, Ϫ190/ Ϫ88, was deleted, and a HindIII restriction sequence was inserted at Ϫ88. Oligonucleotides were synthesized to the sequence Ϫ22/Ϫ2 of the wild type thymidine kinase promoter and to the same location from the linker-scanning mutant TK(lsϪ21/Ϫ12)CAT. These oligonucleotides were then ligated into HindIII-digested ⌬MTV-CAT. The resulting plasmids were sequenced to confirm the presence of the proper sequence. The sequence of the final insert is given in the figure. Bottom, the response of the ⌬MTV(TK)-CAT plasmids to insulin was determined as above. GH4 cells were transfected with 10 g of the ⌬MTV(TK)CAT plasmid indicated and 5 g of pRT3HIR-2. They were incubated with 1 g/ml insulin as described in Fig. 1. The parental plasmid ⌬MTV-CAT was unresponsive to insulin treatment (data not shown). Basal CAT expression/g of protein was 0.14 Ϯ 0.065% for MTV(TKa)CAT, 0.065 Ϯ 0.011% for MTV(TKai)CAT, 0.1 Ϯ 0.027% for MTV(TKb)CAT, and 0.31 Ϯ 0.07% for MTV(TKbi)CAT.
FIG. 5.The effect of insulin on deletion mutants of the wild type thymidine kinase promoter in GH4 cells. GH4 cells were cotransfected with 10 g of the thymidine kinase-CAT construct indicated in the figure and 5 g of pRT3HIR-2. Following a 24-h incubation without hormone, insulin was added at 1 g/ml. After an additional 24 h of incubation, the cells were harvested, and CAT enzyme analysis was performed as in Fig. 1 teins could bind to putative insulin response elements of the prolactin, somatostatin, and thymidine kinase genes, the Etsrelated proteins Sap-1 and Elk-1 were used in gel mobility shift experiments with Ets-binding site-containing oligonucleotides from the insulin-responsive promoter constructs used in these experiments. The insulin response element of the prolactin gene is the Ets-binding sequence-related CGGAAA at Ϫ97/ Ϫ92. Fig. 8A shows that both Elk-1 (lane 2) and Sap-1 (lane 3) but not SRF (not an Ets-related protein) (lane 4), bind to the insulin response element of the prolactin promoter. Control lysates (lane 9) demonstrate no binding in the region of the shifted band seen with Sap-1 or Elk-1. An excess of non-radiolabeled prolactin Ϫ106/Ϫ87 inhibits the formation of this complex (data not shown). Nuclear extract (lane 1), run as a control, produced a characteristic pattern of specific binding (18). However, no bands equal in migration to the Sap-1 and Elk-1 retarded species were seen in nuclear extract even with longer exposures. Sap-1 appears to have higher affinity since it shifts more label to a more slowly migrating form. However, this might result from differential efficiencies of Sap-1 and Elk-1 production in the reticulocyte lysates. The shifts produced by Sap-1 and Elk-1 are indistinguishable from one another. SRF added with Elk-1 (lane 5) or Sap-1 (lane 6) slightly decreases the binding of the Ets-related member of the ternary complex. This may result from slight inhibition of binding by factors in the reticulocyte lysate. Addition of Sap-1 and Elk-1 together results in no enhancement of the shift produced by Sap-1 alone nor does it produce additional migrating forms.
Sap-1 incubation with an oligonucleotide to the somatostatin promoter (Fig. 8B) produces a more slowly migrating protein-DNA complex similar to that formed with PrlϪ106/Ϫ87 (Fig.  8B, lane 2). An excess of non-radioactive somatostatin ϩ7/ϩ47 inhibits the formation of this complex (Fig. 8B, lane 8). Nuclear extract proteins also bind to the somatostatin promoter (Fig.  8B, lanes 1 and 7), but no specific interactions of comparable migration with the Sap-1-DNA complex were seen even on longer exposure (Fig. 8B, lane 1 versus lane 2). No specific binding of Sap-1 was seen using an oligonucleotide whose Etsbinding sites were mutated by a G3 C conversion (compare lane 5, Sap-1, with lane 6, unprogrammed lysate) (Fig. 8B).
Sap-1 binds to the thymidine kinase promoter only when it has been mutated to contain an Ets-binding site (Fig. 8C). The wild type thymidine kinase promoter (Ϫ22/Ϫ2) shows no retarded bands when labeled DNA is incubated with Sap-1 (Fig.  8C, lane 2) that are not also present with unprogrammed lysate (Fig. 8C, lane 3). The oligonucleotide TKb corresponds to the this sequence (Ϫ22/Ϫ2) that is found in the insulin-sensitive TK(lsϪ21/Ϫ12)CAT and MTVTKb-CAT. Incubation of Sap-1 with 32 P-labeled TKb results in two retarded complexes (Fig.  8C, lane 5). These bands are not seen with unprogrammed lysate (Fig. 8C, lane 6), and they are inhibited by an excess of unlabeled TKb (Fig. 8C, lane 8). Again, no specific bands corresponding to the Sap-1 shifted DNA are seen using nuclear extract (Fig. 8C, lane 4). DISCUSSION Multiple lines of evidence presented here and previously (8) indicate that the sequence CGGA is a consensus response element for insulin effects in GH and HeLa cells. First, deletion and linker-scanning mutants of the prolactin promoter identi- fied the sequence CGGAAA as essential for the insulin effect on the prolactin promoter, and this sequence could confer insulin responsiveness to ⌬MTV-CAT (8). Second, the expression of CAT from several CGGA containing linker-scanning and deletion mutants of the thymidine kinase promoter is increased by insulin. When this linker sequence is inactivated, as in the plasmid TK(Ϫ95/ϩ56⌬BamHI)CAT, insulin responsiveness is lost. The CGGA-containing sequence from one of the linkerscanning mutants was shown to confer insulin sensitivity when inserted into ⌬MTV-CAT. Finally, CAT expression from a somatostatin promoter construct is also increased by insulin. Deletions that inactivate the cAMP response element of this gene have no effect on insulin regulation. However, deletion of sequences in the 5Ј-untranslated region of the gene, containing three Ets-binding motifs, eliminates the increased expression mediated by insulin. Point mutation of these motifs reduces the effect of insulin 75%, and a 24-bp deletion that removes the first 2 of these motifs completely eliminates the effect of insulin. When the three motifs are added back to the insulininsensitive plasmid SS(Ϫ71/Ϫ1)CAT, the effect of insulin is restored. These effects are seen both in GH cells and in HeLa cells. Thus, the presence of one or more CGGA sites in the proximal promoter region confers insulin responsiveness in these cell lines.
The effect of insulin to increase gene expression can be mediated by one copy of the insulin response element. The constructs ⌬MTV(PrlϪ106/Ϫ77)CAT and ⌬MTV(TKb)CAT are stimulated 4-and 6-fold by insulin, and they have only one copy of this sequence. The 5Ј-deletion mutants of the thymidine kinase promoter are also insulin responsive with only one copy of this sequence. However, the prolactin promoter has two Ets-related binding sequences, and the response of the prolactin promoter is approximately twice that of the thymidine kinase promoter constructs. Thus, multiple Ets motifs may mediate an increased response. This is not true of all cell lines. Chinese hamster ovary cells transfected with the prolactin-CAT constructs and an expression vector for Pit-1 show low levels of prolactin-CAT expression. However, this activity is not inducible either by insulin or cAMP. 2 Thus, it appears that Chinese hamster ovary cells lack transcription factors that are both important for high basal expression of this construct and its regulated expression. The sequence CGGA is able to confer insulin sensitivity only in cells with a necessary complement of transcription factors.
The involvement of Ets-related proteins in insulin-increased gene transcription is suggested by the experiment with the dominant negative Ets plasmid and by the gel shift experiments. Cotransfection of cells with a plasmid that expresses the DNA-binding domain of c-Ets-2 causes a 75% reduction in the insulin sensitivity of the prolactin, somatostatin, and thymidine kinase promoters. The Ets-related proteins Elk-1 and Sap-1 were shown to bind sequences from these promoters that 2 F. M. Stanley, unpublished observation. are 32 P-labeled somatostatin ϩ7/ϩ47, and lanes 4 -6 are 32 P-labeled somatostatin ϩ1/ϩ80EtsMut with point mutations converting the three GGA sequences in the somatostatin promoter to GCA. These oligonucleotides were incubated with nuclear extract (lanes 1, 4, and 7), Sap-1 (lanes 2, 5, and 8), or unprogrammed lysate (lanes 3, 6, and 9). Lanes 7-9 also had a 100-fold excess of unlabeled somatostatin ϩ7/ϩ47. C, lanes 1-3 are 32 P-labeled thymidine kinase Ϫ22/Ϫ2, and lanes 4 -9 are the 32 P-labeled Ϫ22/Ϫ2 sequence from the linker-scanning mutant TK(lsϪ21/Ϫ12)CAT (also found in MTV(TKb)CAT), in which the addition of a BamHI linker introduces a CGGA motif. Lanes 7-9 also had a 100-fold excess of unlabeled TK-22/Ϫ2 mutant. These oligonucleotides were incubated with nuclear extract (lanes 1, 4, and 7), Sap-1 (lanes 2, 5, and 8), or unprogrammed lysate (lanes 3, 6, and 9). are insulin sensitive but not sequences that are insulin insensitive.
The location of this sequence may also be important for its ability to mediate responses to insulin. The three insulin-sensitive promoters that we have described have the insulin response element inserted close to the transcription start site, the farthest away being the putative insulin response element of the wild type thymidine kinase gene at Ϫ150. Since this sequence is not uncommon in the genome, it is likely that this sequence is only effective within the first few hundred base pairs of the transcription start site.
These data allowed us to establish several criteria for screening insulin-responsive genes for potential response elements. First, the sequence GGA is key to the insulin response element. Second, preference was given to sequences containing CGGA as in the prolactin and thymidine kinase promoters, but (A/ T)GGA sequences were also considered (especially as a dimer with CGGA). Third, the limits Ϫ250/ϩ50 were established since the IRE in the somatostatin gene is in the region ϩ1/ϩ50, and the upstream limit was established in accord with other Ets-responsive promoters that contain Ets-binding sites in the Ϫ200/Ϫ300 region. Finally, inverse sequences were also considered since the insulin response is transferred to ⌬MTV-CAT by both the normal and inverse prolactin and thymidine kinase insulin response elements. These criteria allowed us to identify potential insulin response elements in 22 genes previously reported to be insulin sensitive.
The utility of this type of analysis is clear for promoters where extensive deletional analysis has defined a region that contains the insulin response sequence. For example, the insulin response element of gene 33 likely resides in the first 100 base pairs (26) of the promoter. Comparison of this region of the gene 33 sequence with the consensus insulin response element identifies the sequence Ϫ93 CCGGATTGGCTGCGCGGAGG Ϫ74 that contains a direct repeat of the insulin responsive sequence (underlined). The insulin response of the c-fos promoter was mapped to the serum response element (27). This sequence contains the sequence Ϫ225 GCGGAAGGTCTAG-GAGA Ϫ209 that binds Elk-1. Elk-1 is an Ets-related protein that was shown to be phosphorylated by insulin.
These studies do not rule out other insulin response elements, and it seems likely that there are other insulin-responsive sequences. Although all of the insulin-responsive promoters examined have sequences that are similar to the insulin response element of the prolactin promoter, the region of homology between our consensus insulin response element and the area of the gene known to be insulin sensitive does not correspond in all cases. For example, the negative insulin response element in the glucagon gene is apparently located at Ϫ268/Ϫ238, and this region does not contain a CGGAA sequence (28). Similarly, AGGA sequences reside outside of the insulin-responsive region of the amylase gene Ϫ167/Ϫ137 (29).
In summary, we have defined a consensus insulin response element, CGGA, that can act in several different promoter contexts and in different cell types. The activity of this response element is most likely dependent on the presence of the proper insulin response pathway and insulin-responsive transcription factors in the cells.