The factor binding to the glucocorticoid modulatory element of the tyrosine aminotransferase gene is a novel and ubiquitous heteromeric complex.

Glucocorticoid induction of the tyrosine aminotransferase gene deviates from that of many glucocorticoid-responsive genes by having a lower EC and displaying more agonist activity with a given antiglucocorticoid. A cis-acting element, located 3646 base pairs upstream of the start of tyrosine aminotransferase gene transcription, has been found to be sufficient to reproduce these variations with heterologous genes and promoters (Oshima, H., and Simons, S. S., Jr.(1992) Mol. Endocrinol. 6, 416-428). This element has been called a glucocorticoid modulatory element, or GME. Others have called this sequence a cyclic AMP-responsive element (CRE) due to the binding of the cyclic AMP response element binding protein (CREB). We now report the partial purification and characterization of two new proteins (GMEB1 and −2) of 88 and 67 kDa that bind to the GME/CRE as a heteromeric complex. This purification was followed by the formation of a previously characterized, biologically relevant band in gel shift assays. By several biochemical criteria, the GMEBs differed from many of the previously described CREB/CREM/ATF family members. Partial peptide sequencing revealed that the sequences of these two proteins have not yet been described. Size exclusion chromatography and molecular weight measurements of the gel-shifted band demonstrated that the GMEBs bound to the GME as a macromolecular complex of about 550 kDa that could be dissociated by deoxycholate. Similar experiments showed that CREB bound to the GME as heteromeric complexes of about 310 and 360 kDa. As determined from gel shift assays, GMEB1 and −2 are not restricted to rat liver cells but appear to be ubiquitous. Thus, these novel GMEBs may participate in a similar modulation of other glucocorticoid-inducible genes in a variety of cells.

Glucocorticoid induction of the tyrosine aminotransferase gene deviates from that of many glucocorticoidresponsive genes by having a lower EC 50 and displaying more agonist activity with a given antiglucocorticoid. A cis-acting element, located 3646 base pairs upstream of the start of tyrosine aminotransferase gene transcription, has been found to be sufficient to reproduce these variations with heterologous genes and promoters (Oshima, H., and Simons, S. S., Jr. (1992) Mol. Endocrinol. 6, 416 -428). This element has been called a glucocorticoid modulatory element, or GME. Others have called this sequence a cyclic AMP-responsive element (CRE) due to the binding of the cyclic AMP response element binding protein (CREB). We now report the partial purification and characterization of two new proteins (GMEB1 and -2) of 88 and 67 kDa that bind to the GME/ CRE as a heteromeric complex. This purification was followed by the formation of a previously characterized, biologically relevant band in gel shift assays. By several biochemical criteria, the GMEBs differed from many of the previously described CREB/CREM/ATF family members. Partial peptide sequencing revealed that the sequences of these two proteins have not yet been described. Size exclusion chromatography and molecular weight measurements of the gel-shifted band demonstrated that the GMEBs bound to the GME as a macromolecular complex of about 550 kDa that could be dissociated by deoxycholate. Similar experiments showed that CREB bound to the GME as heteromeric complexes of about 310 and 360 kDa. As determined from gel shift assays, GMEB1 and -2 are not restricted to rat liver cells but appear to be ubiquitous. Thus, these novel GMEBs may participate in a similar modulation of other glucocorticoid-inducible genes in a variety of cells.
For many years, the accepted model of steroid hormone action predicted that the responses of all regulated genes were a property of the steroid used. Thus, a gene is induced, or repressed, by agonists, and the action of agonists is prevented by antisteroids. Furthermore, the concentration of steroid required for half-maximal induction by an agonist and the amount of agonist activity possessed by a given antisteroid should be constant for each steroid and independent of the gene examined (reviewed in Refs. 1 and 2).
Recently, this model has had to be modified as exceptions were defined. Thus, jun⅐fos heterodimers (AP-1), and AP-1 inducers such as phorbol esters block steroid induction (3,4) in what can be a cell-specific manner (5,6), while jun⅐jun homodimers augment glucocorticoid induction (4). Cyclic AMP, via protein kinase A, can often (but not always (7)(8)(9)) cause greater induction by agonists (10,11) and increased percentages of agonist activity for antisteroids (11,12). Heat shock, or chemical shock, afforded a synergistic increase in glucocorticoid inducibility (13), while the immunosuppressive agent FK506 augmented the activity of subsaturating concentrations of glucocorticoids (14). Finally, dopamine can cause ligandindependent gene activation of some receptors (15). None of the above agents effected any shift in the dose-response curve for agonists except for FK506, which was postulated to increase the nuclear binding of activated complexes (14).
Other observations that did not appear to fit with the conventional model of steroid hormone action originated from studies on glucocorticoid induction of the tyrosine aminotransferase (TAT) 1 gene in rat hepatoma tissue culture (HTC) cells, which had become a paradigm for steroid-inducible genes. We found that the dose-response curve for dexamethasone induction of TAT gene expression in the related Fu5-5 rat hepatoma cell line was left shifted compared to that in HTC cells (16). Similarly, TAT enzyme activity was induced at lower cAMP concentrations in Fu5-5 cells than in HTC cells (7). Furthermore, all antiglucocorticoids examined displayed a higher percentage of agonist activity for TAT gene expression in Fu5-5 than in HTC cells (16 -18). This left shift in the TAT doseresponse curve, and increased agonist activity with antisteroids, was not a general response of all glucocorticoid-inducible genes in Fu5-5 cells (19) and occurred at the level of correctly initiated transcripts (7,19). Surprisingly, the magnitudes both of the left shift in the dose-response curve and of the increased amount of agonist activity were not constant but varied slowly over time (17,20) in a manner that was eventually found to be related to the density of the cells in culture (21). Therefore, it appeared that some event downstream of steroid binding to the glucocorticoid receptor selectively modulated the properties of TAT gene induction by glucocorticoid agonists and antagonists.
We previously proposed that this modulation of TAT gene induction in rat hepatoma cells occurred via the binding of a trans-acting factor to a cis-acting element of the TAT gene (1). Stable (22) and transient (23) transfection assays succeeded in identifying such a cis-acting element, at about Ϫ3646 bp of the rat TAT gene, that conveyed all of the glucocorticoid induction properties of the endogenous TAT gene to heterologous genes and promoters. This cis-acting element was called a glucocorticoid modulatory element (GME) and was found to bind a trans-acting factor(s) (23). The mechanism of action of the GME, unlike that of the commonly discussed transcription factor binding sites, does not involve synergism with the glucocorticoid response element, or GRE (24). This suggests that the GME-bound factor(s) (GMEB) might be a novel protein.
The binding site of the GMEB has also been identified as a cyclic AMP-responsive element (CRE) (25)(26)(27), but several lines of evidence indicate that two different sets of proteins are responsible for GME and CRE activity. First, the biological activities mediated by GMEB and the CRE binding protein (CREB) are quite dissimilar (Refs. 21-23 versus 25, 28). Second, no additional element is needed for GME biological activity, while a functional CRE requires a second TAT gene sequence, initially called BIII (25) and more recently found to bind HNF-4 (27). Third, both the GME (23) and the CRE (28) give a closely spaced, three-band pattern in gel shift assays. However, the GMEB is responsible for slowest migrating of the three bands (23), which has been shown not to contain CREB (26).
The GME contains the sequence CGTCA, which is a common CRE element that binds homo-and heterodimers of CREB/ATF along with other family members or unrelated proteins (29) such as AP-1 (30). Thus, many known and unknown transacting factors could bind to the GME/CRE site at Ϫ3646 bp of the TAT gene. The purpose of this paper, therefore, was to characterize and purify the binding protein(s) proposed to be responsible for the GME activity of modulating glucocorticoid receptor function.
Fractionation Procedures for GMEB-Nuclear extracts (50 l) in the absence or presence of 1 or 2.7 M guanidine hydrochloride were loaded onto a Microcon 100 (molecular cut-off, 100 kDa) (Amicon) and centrifuged at 2500 ϫ g for 20 -25 min until 50% of the initial volume had passed through the membrane. The retentates and pass-through fractions containing guanidine hydrochloride were dialyzed against buffer 20T at 4°C for 16 h and stored at Ϫ70°C.
For size exclusion chromatography, 200 l of nuclear extract or cytosol was loaded onto a Superose 6 HR 10/30 column (Pharmacia), which had been equilibrated with buffer 150T at 4°C. The column was run at 0.5 ml/min in a FPLC system (Pharmacia). Every 500-l fraction was collected and stored at Ϫ70°C.
For ion exchange chromatography, HTC cytosol (about 30 ml) was loaded with a Superloop in a FPLC system (Pharmacia) onto Mono Q HR 10/10 column that had been equilibrated with buffer 20T at 1 ml/min. After washing the column with the same buffer, the column was eluted with a linear gradient up to 40% of buffer 1000T. Every 2-ml fraction was collected and stored at Ϫ70°C.
For size fractionation by electrophoresis, partially purified nuclear extract or cytosol was separated on 6 or 8% SDS-PAGE gels (1.5 mm thick) (32) and the desired M r ranges, based on the migration of prestained molecular weight markers (Life Technologies, Inc., Bio-Rad) in adjacent lanes, were cut out of the gel and electroeluted in 1 ϫ SDS running buffer (25 mM Tris, 0.19 M glycine, 0.1% SDS) at 125 V in a model 1750 electroeluter (ISCO, Lincoln, NE) for 60 min at room temperature. After elution, the protein was precipitated with 4 volumes of acetone (Ϫ60°C), and the pellets were washed with 80% acetone, 20% buffer 20T (4°C) as described (33). The pellets were redissolved in buffer 20T and reprecipitated with acetone as before. After the second precipitation, the recovered samples were denatured and renatured by redissolving in 60 l of buffer 20T containing 6 M guanidine-hydrochloride and dialyzing against buffer 20T at 4°C for 16 h. After dialysis, the renatured samples were concentrated to about 20 l with a Microcon 10 concentrator (12,000 rpm for 15 min at 4°C) (Amicon) and stored at Ϫ70°C.
Methylation Interference Assay-The sequence of the doublestranded oligonucleotide used was as follows (capitalized letters correspond to the rat TAT sequence from Ϫ3654 to Ϫ3634 bp from the start of transcription): 5Ј-tcgaCTTCTGCGTCAGCGCCAGTATg-3Ј, 3Ј-gctGAAGACGCAGTCGCGGTCATAcg-5Ј. The probe was filled-in by Klenow fragment with [ 32 P]dCTP or [ 32 P]dATP for labeling specifically the upper or lower strand, respectively. The labeled probes were par-tially methylated with dimethyl sulfate as described by Maniatis et al. (34). The binding reaction and gel electrophoresis were performed as for the gel shift assay, except that the binding reaction was scaled up 8-fold and wider sample wells (5 cm width) were used. After electrophoresis, the gel was electroblotted onto DEAE filters (NA45; Schleicher & Schuell) in 0.5 ϫ TBE buffer at 30 V at 4°C for 16 h. The filter was exposed to Kodak X-Omat XAR-5 film at 4°C for 120 min. The retarded and free bands were identified from the x-ray film and cut out from the filter. The DNA was extracted from the excised filter in elution buffer (1 M NaCl, 0.1 mM EDTA, 20 mM Tris (pH 8.0 at room temperature)) for 30 min at 60°C, precipitated by ethanol with 5 g of native, sheered herring sperm DNA, and subjected to piperidine treatments (35). Equivalent amounts of radioactivity obtained from the retarded and free gel shift bands were loaded onto 15% sequencing gels. After electrophoresis, the gel was dried and exposed to Kodak X-Omat XAR-5 film at Ϫ70°C for 3-5 days.
DNA Affinity Chromatography-Oligonucleotides for the M3 sequence (5Ј-tcgaCTTCTGCGTCAGTATGCGTATg-3Ј, 3Ј-GAAGACGCA-GTCATACGCATAcagct-3Ј) or M2 sequence (5Ј-tcgaCTTCTGTATGA-GCGCCAGTATg-3Ј, 3Ј-GAAGACATACTCGCGGTCATAcagct-3Ј) (capitalized letters correspond to the rat TAT sequence from Ϫ3654 to Ϫ3634 bp from the start of transcription, bold letters indicate mutations of native TAT sequence, and lower case letters are for added nucleotides to make SalI cohesive ends) were synthesized and purified by HPLC. Sequence-specific DNA affinity columns were prepared as described (36) except that 500 g of oligonucleotides were coupled to 1 ml (bed volume) of CNBr-activated Sepharose CL-4B (Pharmacia). GMEB activity throughout the chromatography was monitored by gel shift assays with 32 P-labeled GME. Mono Q fractions containing GMEB activity (42.5 ml) were pooled, diluted with an equal volume of buffer TN to reduce the NaCl concentration to about 100 mM, mixed with sheered herring sperm DNA (final concentration of 5 g/ml) and spermidine (final concentration of 2 mM; U. S. Biochemical Corp.), and incubated on ice for 10 min before being spun at 12,000 ϫ g for 10 min at 4°C. The supernatant (20 ml) was applied with gravity onto the M2 DNA affinity column (400-l bed volume) equilibrated with buffer TN plus 100 mM NaCl. The column was then washed with 4 ϫ 2 ml of buffer TN containing 100 mM NaCl followed by 550 l of buffer TN with 200, 300, 400, or 800 mM NaCl in stepwise fashion. The flow-through fraction that contained GMEB activity was loaded onto a 400-l bed volume M3 DNA affinity column that was washed as for the M2 DNA affinity column and eluted with buffer TN containing 200, 300, 400, 500, 600, or 800 mM NaCl. The fractions (300 -500 mM NaCl) containing GMEB activity were pooled and diluted with buffer TN to a final NaCl concentration of 100 mM and reloaded on the M3 column. A total of three sequential M3 DNA affinity column purifications were performed. The final M3 DNA affinity column had a 250-l bed volume, and the GMEB activity was eluted in the 300 -500 mM NaCl fraction. All samples were aliquoted, quickly frozen in a dry ice-methanol bath, and stored at Ϫ70°C.

GMEB Is Present and Active in Non-hepatic Cells-
The GME was initially defined in the context of the cloned rat hepatoma tissue culture cells Fu5-5 and HTC. There, the percentage of maximal agonist activity obtained from transiently transfected GREtkCAT reporter constructs with either subsaturating concentrations of agonists or saturating concentrations of antiglucocorticoids was increased by the presence of the GME (23). As shown in Table I with the agonist dexamethasone and the antagonist dexamethasone 21-mesylate, GME activity was not restricted to the originally investigated cloned hepatoma cells. The GME caused increased percentages of agonist activity in uncloned Fu5-5 cells as well as transformed mouse fibroblasts (L cells) and human cervical adenocarcinoma cells (HeLa cells) in a manner that required the crucial CGTC sequence of the GME, which was not present in the inactive M2 mutation of the GME (23). Thus, the capacity to express GME activity, which presumably reflects the existence of the requisite GMEB factor(s), is not restricted to hepatic cells but crosses species lines. The inactivity of the GME in rat adrenal pheochromocytoma cells (PC-12) suggests, though, that there is some cellular selectivity for GME action.
The presence of the GMEB in the above cell lines was examined in gel shift assays. Normally, a three-band pattern was obtained, of which the slowest migrating band corresponds to the biologically relevant interaction of GMEB with the GME (23). Nuclear extracts from each cell line afforded somewhat different patterns; but in each case, a complex of the same low mobility was observed that was blocked only by excess nonlabeled GME oligonucleotide (Fig. 1A). In all cases, the biologically inactive M2 oligonucleotide (23) was unable to competitively inhibit the formation of this band (Fig. 1A). Thus, both liver and non-liver cells appear to contain GMEB.
Properties of GMEB-The GMEB-DNA complex observed in gel shift assays migrated only slightly slower than several other complexes (Fig. 1A). As expected, cytosol solutions prepared by lysis of cells with hypotonic buffer or by the freezethaw techniques used to obtain crude glucocorticoid receptors (31) contained very few DNA binding species. However, each cytosol solution still appeared to include the same GME binding species that was observed in nuclear extracts (Fig. 1B). This cytosolic binding protein (Fig. 1B) was further identified as the GMEB because the only DNAs that competitively inhibited the formation of the gel-shifted complex were the known biologically active oligonucleotides of GME, M1, M3, and M4 (23). This procedure thus provided a simple method for separating the GMEB from most of the other GME binding proteins. The presence of the GMEB in PC-12 cells, where the GME is inactive (Table I), suggests either that other factors besides the GMEB and glucocorticoid receptors are required for activity or that the GMEB is present in PC-12 cells in a biologically inactive form.
When cytosols were prepared from HTC cells treated with saturating concentrations of dexamethasone at 37°C, conditions that cause the nuclear translocation of most of the cyto-TABLE I Activity of GME in hepatoma and non-hepatic cells Duplicate dishes of the indicated cells were transiently transfected with the specified reporter plasmids. After incubation with steroids, the CAT activity was determined and then expressed as percent of the maximal induction with 1 M dexamethasone as described under "Materials and Methods." Similar results for each combination were obtained in at least one additional experiment except for M2GREtkCAT transfection of L cells. ND, not determined.  (37), there was no appreciable decrease in the intensity of the gel-shifted band with 32 P GME (data not shown). Similarly, cytosolic preparations of receptors treated with saturating concentrations of dexamethasone and then heated to activate the complexes to the DNA binding form did not display an increased amount of the GMEB complex in gel shift assays (data not shown). Therefore, we conclude that this GMEB-containing band does not involve any interaction of the glucocorticoid receptor with the GME. GMEB bound to several polycationic columns. GMEB was eluted from heparin-agarose at 0.3-0.8 M NaCl, from DEAE-Sepharose CL-6B by 0.1-0.2 M NaCl, and from Mono Q columns with 0.2 M NaCl. Thus, GMEB would be predicted to contain at least one surface containing a net abundance of anionic charges, as would be expected for a factor involved in the modulation of transcription. GMEB also bound to phenyl-Sepharose and a Mono S polyanionic column, but the recovery was very low (data not shown).
Several results indicate that GMEB and CREB are different proteins. First, heating the nuclear extracts to 65°C (but not 37°C) for 10 min eliminated the formation of the GMEB-bound complex (data not shown). In contrast, CREB is stable under these conditions (data not shown, Ref. 29). Second, an anti-CREB antibody supershifted some of the lower two bands in gel shift assays but none of the more slowly migrating GMEBcontaining bands (data not shown). This confirms the earlier report that only the lower two bands contain CREB (26). Third, a 0.5-h treatment with 10 M of forskolin, which increases the cellular cAMP (and activates CREB), had no effect on the amount of GMEB in the gel shift assay (data not shown) or on the percent agonist activity seen for 1 M dexamethasone 21mesylate with the GREtkCAT reporter (30 versus 34%) and only slightly increased the percent agonist for dexamethasone 21-mesylate with the GMEGREtkCAT reporter (67 versus 57%). Fourth, a 32 P oligonucleotide containing the consensus CRE (38) of the somatostatin gene (SOM/CRE) (25,26) differs from the GME in 9 out of 19 nucleotides (see below; identical nucleotides are underlined, and lower case letters indicate flanking DNA in the reporter plasmids) and did not afford the slower migrating, GMEB-containing band in gel shift assays (GME, 5Ј-tcgaCTTCTGCGTCAGCGCCAGTATtcga-3Ј; SOM/ CRE, 5Ј-gatccCTCTCTGACGTCAGCCAAGGAgatc-3Ј). Also, non-labeled SOM/CRE oligonucleotides only weakly competed for the formation of this band with the 32 P GME oligonucleotide (data not shown).
Two factors that are closely related to CREB, which heterodimerize with CREB, and can replace CREB in vivo are CREM and ATF-1 (39). However, neither a broad spectrum antibody (anti-CREM) nor a variety of mono-specific antibodies could supershift the GMEB-containing band in gel shift assays. Similarly, immunodepletion of the nuclear extracts with these antibodies did not prevent the formation of the GMEB-containing band (data not shown). These results argue that the GMEB is not ATF-1, -2, or -3, CREB-1 or -2, CBP, or CREM. Similar experiments with anti-jun and anti-fos antibodies ruled out jun and/or fos as being the GMEB (data not shown), even though AP-1 is active with a highly homologous DNA sequence (40).
GMEB Is a Multimeric Protein-Methylation interference experiments revealed that all of the guanines in a 10-bp region of the GME are important for binding in the gel shift assay (Fig. 2). An even larger region of 26 bp is protected from DNase I digestion (data not shown). Under most circumstances, more than one protein would be required to cover such a large stretch of DNA. When compared to molecular weight markers in the gel shift assay (41), the size of the GMEB was calculated to be ϳ550 kDa; the sizes of the CREB-containing bands were about 310 and 360 kDa (data not shown). A similar very large size of 600 kDa for the GMEB was observed by gel shift assays of the peak binding activity after fractionation by size exclusion chromatography on Superose 6 HR (Fig. 3, A and B) and Sepharose S-3000. To determine whether this 550 -600-kDa species was a monomeric or oligomeric protein, we made use of the report that deoxycholate dissociates protein-protein complexes, but not protein-DNA complexes, in a manner that can be reversed by added Nonidet P-40 detergent (42,43). As shown in Fig. 3C, the appearance of the GMEB-containing band (and the lower CREB-containing bands) was blocked by deoxycholate and restored by added Nonidet P-40. However, these GMEB and CREB protein complexes are relatively resistant to dissociation by salt. All three bands were seen in gel shift assays of material retained by a Microcon 100 filter in the presence of 1 M guanidinium hydrochloride while no band was observed in the flow through, which would contain species Յ100 kDa (Fig. 3D). Raising the guanidinium hydrochloride concentration to 2.7 M still did not allow any GMEB to appear in the flow-through of the Microcon 100 filter, as evidenced by the formation of the appropriate gel shift bands (data not shown). Collectively, these data suggest that GMEB and the ϳ42-kDa CREB either can exist as a multimeric complex that resists dissociation in salt or requires other proteins larger than 100 kDa to form the observed DNA complexes on gels.
GMEB Is a Heterooligomer-Southwestern blotting of proteins that had been separated on SDS-polyacrylamide gels and then renatured prior to being probed with 32 P GME oligonucleotide was performed to determine the size of the monomeric GMEB. No signal was seen with either crude or partially purified (by FPLC on a Mono Q column) GMEB under conditions where CREB was readily visualized at ϳ42 kDa (data not shown). This implies that the GMEB may be comprised of proteins of different molecular weights. Direct evidence for this conclusion came from gel shift assays with Mono Q purified material that was separated on, and then extracted from, SDSpolyacrylamide gels. No one size of fractionated proteins yielded the original band in gel shift assays, even though samples containing a broad size range of proteins retained the ability to form the gel-shifted band both after exposure to SDS (Fig. 4, lane 2) and after extraction from an SDS gel (Fig. 4,  lane 3). However, when two fractions encompassing species of 101-80 kDa and 80 -62 kDa were mixed, a strong gel shift band was obtained (Fig. 4, lanes 5, 13, and 20). Each fraction by itself bound DNA weakly and afforded different complexes. The reassociation of both components occurred more efficiently when they were mixed before, as opposed to after, renaturation (Fig. 4, lanes 20 versus 22). When the components were mixed after renaturation, increased incubation time favored complex formation (Fig. 4, lanes 20 -22).
Purification and Peptide Sequencing of GMEB1 and -2-The above data indicated that the GMEB is composed of two proteins. To confirm this conclusion, the GMEBs were purified from about 130 g of HTC cells (Table II). Each stage of the purification scheme was monitored for its ability to give the appropriate gel-shifted band. Thus, the GMEBs in crude cytosol devoid of CREB were fractionated first on a preparative Mono Q HR 10/10 column and then on a column containing tandem repeats (Ͼ10) of the biologically inactive GME mutant oligonucleotide M2 (23) to remove nonspecific binding proteins. The flow-through from this column was loaded onto a column with tandem repeats (Ͼ10) of the GME mutant oligonucleotide M3 to bind GMEB. The mutant oligonucleotide M3 was used because M3 had a higher apparent affinity for GMEB than did the native GME, as determined from competitive gel shift experiments (Fig. 1B) and affinity chromatography with multiple repeats of DNA oligonucleotides (data not shown). After three rounds of DNA affinity chromatography, the GMEBs had been purified about 5000-fold (Table II). As with crude GMEB, the gel-shifted band obtained with purified GMEB was inhibited only by added biologically active oligonucleotides (i.e. GME, M1, M3, M4 but not M2, AP-1, or CRE) (Ref. 23 and data not shown). Analysis of this purified material on denaturing SDS-polyacrylamide gels followed by staining with copper, Coomassie Blue, or silver (Fig. 5A) revealed three major species. However, only the combination of the highest and lowest molecular mass proteins at 88 and 67 kDa could reconstitute the GMEB band in gel shift assays after the elution of each band from SDS-polyacrylamide gels (Fig. 5B). Confirmation that this was the authentic gel-shifted band came from the ability of only the biologically active GME and not the mutant M2 oligonucleotide (23) to competitively inhibit the formation of this band (Fig. 5C).
About 130 g of the two proteins GMEB1 (88 kDa) and GMEB2 (67 kDa) was isolated from denaturing SDS-polyacrylamide gels and sent to the Kreck Foundation (Yale University) for peptide sequencing. The procedure involved in-gel trypsin digestion of each protein, HPLC separation of the fragments, and laser desorption mass spectroscopy to determine which fragments would be most amenable to micro-sequencing. The partial sequences of the three peptides from each protein digest that were sequenced are given in Table III. The yield of most sequenced peptides was 5-23%, which is the normal range. For peptide 2-73, a very high yield of 78% was obtained, which also proves that this protein was not a mixture of two or more proteins when isolated from the SDS-polyacrylamide gel. The high concentration of acidic residues in some of these peptides (i.e. 1-109 and 2-145) is compatible with the retention of the proteins on polycationic columns (see above). A TBLASTN search of numerous DNA and protein data bases using the NCBI BLAST E-mail server (44) revealed no significant homology with any of the peptides of Table III. Therefore, the GMEB may be a heterooligomeric complex of two novel proteins.
FIG. 2. Methylation interference assay for GMEB binding to GME. Nuclear extracts from Fu5-5 clone 27 cells were incubated with partially methylated 32 P GME probe and fractionated on a 5% nondenaturing PAGE gel. The top and bottom strands of free and GMEBbound probe DNA were processed for sequencing and autoradiographed as described under "Materials and Methods." The DNA sequence of the top and bottom strands is shown at the left and right, respectively, of the autoradiograph. B, bound probe; F, free probe; arrows mark those guanosine residues that must remain unmethylated for complex formation with GMEB to occur. DISCUSSION We have called the element at Ϫ3.6 kilobases of the rat TAT gene a glucocorticoid modulatory element, or GME, because it modulates the induction properties of both subsaturating concentrations of agonists and saturating concentrations of antagonists (23). The binding of a protein(s) to the GME was observed that was directly related to the biological activity of the GME oligonucleotide (23,24), which was different from synergism (24). We now report that the GMEB appears to be a heterooligomer of two previously unsequenced proteins, GMEB1 and GMEB2, of apparent molecular masses of 88 and 67 kDa, respectively. However, conclusive identification must await the cloning of both proteins and a demonstration of biological activity with the cloned proteins in cells lacking GMEBs.
Several properties of the GMEBs emerged during their purification that pertain to the mechanism of GME action. First, although the DNA sequence to which the GMEBs bind is very similar to that for the CREB/CREM/ATF and the Jun/Fos/AP-1 superfamilies, and CREB even binds to a non-consensus CRE at the same position as the GME at Ϫ3646 bp of the TAT gene (25)(26)(27), there is little similarity between the GMEBs and these other proteins. The GMEBs are not related to CREB by the criteria of size, biological activity (23, 24, 38), antibody reactivity, methylation interference patterns for protein binding ( Fig.   2A versus Fig. 7 of Ref. 28), or amino acid sequence (Table III). Some AP-1 sites contain the CGTC of the GME, and AP-1 may bind to the GME/CRE, as it has recently been reported that 12-O-tetradecanoylphorbol-13-acetate both inhibited glucocorticoid (and cAMP) induction of TAT and caused a decreased protein occupancy of the CRE at Ϫ3646 bp (45). However, again there was no similarity between the peptide sequences of the GMEBs and AP-1, an anti-AP-1 antibody did not cause a supershift of GMEB-GME complexes, and 12-O-tetradecanoylphorbol-13-acetate alone did not elicit any response from GMEcontaining constructs (data not shown). Thus, there is little physical or biological similarity between the GMEBs and the other factors binding to the same DNA sequence. This, then, is an additional example of different proteins that bind to the same DNA region (46,47).
Second, we do not know if GMEB and CREB can both bind to the GME/CRE at the same time. However, it seems that CREB is unable to block GMEB action. The low levels of the protein kinase A regulatory subunit (Tse-1) in liver cells are thought to result in high amounts of active CREB that would bind to the GME/CRE (48). Nevertheless, reporter constructs containing either a single GME (GMEGREtkCAT) or other elements that are needed for CRE activity, such as multiple tandem repeats of the GME or the GME plus the downstream BIII sequence (25), show full GME activity in Fu5-5 rat hepatoma cells (23).  35 kDa). B, gel shift properties of size-fractionated HTC cytosol. Unfractionated cytosol (lanes 1 and 24) and aliquots (3 l) of the indicated column fractions were assayed in the gel shift assay as in Fig. 1 and visualized by phosphorimagizer. The arrow marks the position of the GMEB-containing band. C, effects of deoxycholate and Nonidet P-40 on the binding of GMEB and CREB to GME. Fu5-5 nuclear extract was incubated with 32 P GME oligonucleotide in the presence of the indicated percentage of deoxycholate Ϯ 1% Nonidet P-40. Complexes were separated on 5% non-denaturing polyacrylamide gel as described under "Materials and Methods" and visualized by autoradiography. The filled arrow indicates the position of the GMEB-containing band, the two CREB-containing bands are just below, and the open arrow is at the position of nonspecifically bound species. D, size fractionation of GMEB in 1 M guanidine hydrochloride. Fu5-5 nuclear extracts Ϯ 1 M guanidine hydrochloride were centrifuged through a membrane (Microcon 100) whose molecular cutoff was 100 kDa. After centrifugation, the filtrate (Filt.) and retentate (Ret.) fractions were dialyzed and analyzed individually or in combination in gel shift assays with 32 P GME probe as described under "Materials and Methods." The filled arrow indicates the position of the GMEBcontaining band.
Furthermore, conditions that elevate protein kinase A activity, such as forskolin treatment, did not inhibit GME activity. Thus, while CREB binds to the same DNA sequence as GMEB, CREB does not appear to competitively inhibit GMEB binding in intact cells.
Third, GME activity is not limited to rat liver cells (Table I), and the GMEBs are not tissue-specific proteins (Fig. 1A). Furthermore, the fact that the GME was active with synthetic GREs and a variety of promoters, including a minimum thymidine kinase promoter (23), suggests that no tissue-specific DNA binding factors are required for GME activity.
Fourth, the GMEBs are clearly of nuclear origin but can be readily extracted from nuclei under conditions where other factors, such as CREB, stay in the nucleus. This is reminiscent of several other nuclear proteins (49), including the progesterone and estrogen receptors, which are predominantly nuclear but appear in most cytosolic preparations. The cytosolic appearance of the GMEBs could be indicative of a dynamic equilibrium between the two cellular compartments, as established for the progesterone receptors (49 -51), or may simply reflect a repartitioning of the GMEBs in the lysis buffer.
Fifth, the mass and stability of the GMEB complex are notable. The 550 -600-kDa size of the protein complex seen in both gel shift assays and size exclusion chromatography (Fig. 3,  A and B) argue against a nonspecific aggregate. Involvement of the 265-kDa protein CBP that binds phosphorylated CREB (52) was discounted by the observed sizes of GMEB1 and -2 and their lack of immunoreactivity with anti-CBP antibody. The most purified preparation of GMEB appeared to contain about equal amounts of GMEB1 and -2 (Fig. 5A), which would require three or four molecules of each protein in the final complex to achieve a molecular mass of 550 -600 kDa. Such a massive complex is probably not too large to be extracted intact from HTC cell nuclei because identically prepared nuclei were found to be permeable to molecules as large as the 240-kDa protein complex of phycoerythrin (53). However, the GMEB complex must be quite stable to retain specific binding to the GME after extraction from the nucleus (Fig. 1), even in the presence of up to 2.7 M guanidinium hydrochloride and after various degrees of purification (Table II). Despite the stability of the GMEB complex with regard to dissociation, the rate of reassociation of the separated components was relatively slow (Fig. 4).
Finally, from the yield of purified GMEB1 and -2 in Table III, it can be calculated that there are about 40,000 molecules of each GMEB per HTC cell. This is similar to the approximately 80,000 molecules of glucocorticoid receptor that are present in an HTC cell (16). Considering the fact that most glucocorticoidresponsive genes contain two GREs, each of which binds a dimer of the receptor, the ratio of GME-bound GMEB complexes to GRE-bound receptors is about 1:2. Given the facts that the GMEBs are not limited to rat liver cells and that GME-like modulation has been observed with several other FIG. 4. Reconstitution of GMEB activity requires two separable species. HTC cell cytosol (about 37 g of protein) that had been partially purified by Mono Q column chromatography was size fractionated on a 6% SDS-PAGE gel. Molecular weight ranges of HTC cell proteins were isolated by cutting the sample lanes, as indicated in the unshaded gel lane strip, relative to the migrations of prestained markers (Life Technologies, Inc., Bio-Rad) in adjacent lanes (note that gel was cut below the 71-kDa marker). Proteins in the excised pieces of the gel were electroeluted, kept separate, or mixed as indicated by the shaded gel lane strips, and then precipitated, denatured, and renatured as described under "Materials and Methods." About 1-2 l of the 20 l of renatured proteins was used in gel shift assays with 32 P-GME probe. Lane 1, mono Q-purified cytosol; lane 2, mono Q-purified HTC cytosol mixed with the SDS loading buffer followed by precipitation, denaturation, and renaturation; lane 3, proteins eluted from the entire molecular weight range of the SDS-PAGE gel; lanes 7-10, 14 -17, and 20, protein fractions were mixed before denaturation; lanes 21 and 22, proteins were renatured separately and then mixed for indicated time on ice just before the gel shift assay. The position of the GMEB-containing band is indicated by the arrow.

TABLE II
Purification of GMEBs Cytosol from 130 g of HTC cells were sequentially purified by Mono Q DNA affinity column chromatography as described in the text (see "Materials and Methods" for details). The total protein at various stages was quantitated by the Bio-Rad protein assay with bovine serum albumin as a standard or by comparison with the silver staining with known amounts of molecular weight markers for the DNA affinity column fractions. The GMEB specific activity was estimated from the least amount of sample to give a signal for the GMEB-containing band after overnight exposure of the gel shift assay gels to Kodak XAR film at room temperature. ND, not determined. glucocorticoid regulated genes (2), it will be interesting to pursue the possible role of a GME and its heteromeric binding complex in the transcription of genes other than TAT. While CREB has been identified as a component that also binds to the DNA sequence of the GME/CRE (26), it is not known if it is the only protein in the complex. Several lines of evidence argue that the CREB-containing complexes bound to the GME are also multimeric. Most obvious is the size of the CREB-containing complexes, which were ϳ310 -360 kDa in gel shift assays and 400 kDa on size exclusion columns ( Fig. 3A and data not shown). Deoxycholate blocked the formation of the CREB-containing bands in gel shift assays, just as was observed for GMEB (Fig. 3B). Given the fact that CREB is relatively small (ϳ42 kDa), it would seem that the CREB complexes must contain either multiple copies of CREB or other proteins. CREB will bind to the GME-containing oligonucleotide in Southwestern blots (see "Results") and will afford gelshifted bands with a CRE-containing oligonucleotide (26) so that homooligomeric complexes of CREB may form. However, the gel-shifted band that the somatostatin CRE formed with purified CREB exhibited a faster migration than that with crude nuclear extracts (26). Thus, the CREB-containing complex from nuclear extracts probably is not the same as that formed just from CREB and would contain some other protein(s). Further experiments are required to determine whether the suspected additional proteins are CBP (52), other members of the CREB/CREM/ATF superfamily that can heterodimerize with CREB (39), or even GMEB1 or -2.
In summary, we have found that a heteromeric complex of two potentially new proteins binds to a cis-acting element of the TAT gene. These two proteins, GMEB1 and GMEB2, are associated with changes in the transcriptional activity of antiglucocorticoids and low concentrations of glucocorticoids. These are phenomena that have not been previously described for steroid receptors and thus are of considerable mechanistic interest. It remains to be seen whether the GMEBs interact with glucocorticoid receptors and the transcriptional machinery in the manner that we have proposed (2,23). The cloning of GMEB1 and GMEB2 and the production of specific antibodies will be of major assistance in understanding the mechanistic details of this interesting system.  6 and 7). B, reconstitution of GMEB activity in gel shift assays. The three major proteins in the eluant from the third M3 DNA affinity column (proteins 1-3 in A) were individually cut out from a silver-stained gel such as in A, electroeluted, denatured, and renatured as described under "Materials and Methods." Aliquots (0.4 l out of 17 l) from the renatured sample were analyzed in the gel shift assay. lanes 1 and 9, purified GMEB from the third M3 DNA affinity column; lanes 2-7, indicated individual proteins or mixtures of proteins that were mixed, precipitated, denatured, and renatured. C, specificity of binding activity of purified and reconstituted GMEBs. Protein bands 1 and 2 from the SDS gel of A were isolated and reconstituted as in B and then analyzed in the standard gel shift assay without (lane 2) or with a 100-fold molar excess of non-labeled specifically (GME) or nonspecifically (M2) binding oligonucleotides. The complex formed with GMEB from the third M3 affinity column, but not fractionated on SDS gels, is shown in lane 1.

TABLE III
Partial sequence of tryptic peptides from GMEB1 and GMEB2 The purified GMEBs from the third M3 DNA affinity column were fractionated on an 8% SDS-PAGE gel and stained with Coomassie Blue. Bands corresponding to GMEB1 (88 kDa ϭ protein 1 of Fig. 5A) and GMEB2 (67 kDa ϭ protein 2 of Fig. 5A) were cut out and sequenced by the Keck Foundation (Yale University) as described under "Materials and Methods." The single letter amino acid sequence of three tryptic fragments from each protein that were purified and sequenced is given, along with the yield (in parentheses) of the sequenced peptide from the original protein sample.