Multiple elements within the glucocorticoid regulatory unit of the rat alpha 1-acid glycoprotein gene are recognition sites for C/EBP.

Sequences between -106 and -42, located immediately downstream of the glucocorticoid response element, are essential for efficient glucocorticoid-stimulated expression of the alpha 1-acid glycoprotein (AGP) gene. We have used mobility shift assays with oligonucleotides bearing wild type and mutated sequences from segmented portions of this region to characterize the specific interaction of similar binding factors from rat liver and HTC rat hepatoma cell nuclear extracts. One of these factors, AGP nuclear factor 2 (ANF-2), appears capable of dual interaction with the homologous recognition sites, HA (-133 to -104) and HB (-81 to -72), which overlap and are located downstream of the glucocorticoid response element, respectively. Using an affinity matrix containing the HB sequence we have isolated ANF-2 from rat liver nuclear extracts. On the basis of immunological evidence rat liver ANF-2 was confirmed to be highly related and probably identical to CCAAT/enhancer-binding protein (C/EBP). Methylation protection analyses with partially purified, rat liver ANF-2 confirmed HA and HB as recognition sites for C/EBP-related factors and are consistent with the location of a third interaction site for these transactivating proteins at HX (-102 to -93). We propose that the sequences HA, HX, and HB, spanning residues -113 to -72 of the AGP promoter, might serve as recognition sites for a family of C/EBP-like nuclear factors that coordinate the glucocorticoid-mediated induction of the AGP gene.

Induction of AGP mRNA by dexamethasone is markedly reduced in cycloheximide-treated cells suggesting that efficient transcription of the AGP gene requires a pre-existing, labile protein in addition to the glucocorticoid receptor (3,13,14). The dexamethasone responsive region has been delineated between positions -121 and -42 of the AGP promoter (15) and contains a binding site for the glucocorticoid receptor centered approximately 110 base pairs upstream of the AGP transcription initiation site (15). Sequences between -106 and -42, located downstream of the glucocorticoid regulatory element (GRE), are essential for efficient glucocorticoid-stimulated expression from the AGP promoter (15). Additional results suggest that elements responsible for cycloheximide inhibition of this glucocorticoid-mediated induction reside within the same region (15).
In DNase I protection assays with nuclear extracts from HTC (JZ.l) rat hepatoma cells, workers from this laboratory have previously demonstrated a footprinted region spanning -110 to -68 of the AGP gene (15). The result suggested the presence of multiple elements within a control region known to be important for transcriptional activity of the AGP promoter in response to glucocorticoids (15). It was intriguing that these binding elements might represent interaction sites for positive factors that cooperate synergistically with the glucocorticoid receptor leading to induction of transcription by the AGP promoter. As an approach to further characterizing the structural and functional properties of the glucocorticoid regulatory unit of the rat AGP gene, we have used mobility shift assays with oligonucleotides bearing sequences from segmented portions of the DNase I footprinted region to demonstrate the specific interaction of similar factors from rat liver and rat hepatoma cell nuclear extracts. In this study we describe the isolation of one of these factors, designated AGP nuclear factor 2 (ANF-2), from rat liver nuclear extracts using sequence-specific DNA-affinity chromatography. We show that rat liver ANF-2 is immunologically indistinguishable from the CCAAT/enhancer-binding protein (C/EBP), a liver-specific transcription factor (16), and propose that three closely spaced DNA sequences, the most upstream one of which overlaps the GRE, might serve as recognition sites for a family of C/EBP-related nuclear factors that coordinate the regulation of AGP by glucocorticoids.
Preparation of DNA Affinity Resins-Two affinity gels were prepared for ANF-2 isolation. The first was synthesized by coupling multimers of the wild type oligonucleotide (see Fig. 1) containing the base pair -90 to -64 sequence of the rat AGP promoter to CNBractivated Sepharose 4B (17). A coupling of 25 pg of DNA/ml of resin was achieved. Our initial use of this gel for ANF-2 purification appeared to be complicated by nonspecific protein binding. A second affinity resin was prepared based on the Y mutant sequence (see Fig.  1). Coupling of the oligonucleotide to carbonyldiimidazole-activated Sepharose was facilitated by the introduction, during oligonucleotide synthesis, of a primary amine linked to the 5' end of the sense strand via an alkyl (C2H6) spacer arm. N-MMT-C6 amino modifier (Clontech) was used for this purpose in conjunction with an Applied Biosystems DNA synthesizer. Briefly, the amino-linked oligonucleotide (30 nmol) and its complimentary strand (34 nmol) were dissolved in 0.5 ml of coupling buffer (0.5 M NaHC03, pH 9.5) and annealed. T o monitor coupling efficiency, a small quantity of 32P-5' end-labeled complementary strand was added prior to the annealing reaction. Carbonyldiimidazole-activated Sepharose (18) was prepared for use by quickly washing 5 ml of the resin with 25 ml of ice-cold water and suspending the gel in 9.5 ml of ice-cold coupling buffer. The annealed amino-linked oligonucleotide product was added and reaction allowed to continue at 4 "C for 48 h with gentle agitation. Unreacted oligonucleotide was removed by filtration and alternate washing with 0.1 M NaHCOa, pH 8.5, 0.5 M NaC1; 0.1 M NaOAc, pH 5.5, 0.5 M NaC1; and finally water. The gel contained 9 pg of DNA/ml and was stored at 4 "C in BlOO buffer (50 mM Hepes, pH 7.8, 100 mM KCl, 0.1 mM EDTA, 1 mM dithiothreitol, 10% glycerol) with 0.02% azide.
Preparation of Nuclear Extracts-The method of Ohlsson and Edlund (19) was adapted for the preparation of crude rat liver nuclear extracts. All procedures were carried out at 4 "C. Fresh rat (male Wistar) liver tissue was washed free of blood with ice-cold phosphatebuffered saline followed by two washes of buffer A (10 mM Hepes, pH 7.8, 15 mM KCI, 2 mM MgC12,O.l mM EDTA, with freshly added 1 mM dithiothreitol, 1 mM phenylmethylsulfonyl fluoride, 0.3 pg/ml leupeptin, and 0.5 pg/ml antipain). The tissue (30 g, approximately 2 livers) was cut into small pieces (scissors) and homogenized on ice with 30 ml of buffer A in a 50-ml Elvejhem Potter homogenizer using 10 strokes of a motor driven Teflon pestle. The homogenate was divided equally among four 50-ml polypropylene centrifuge tubes. Each tube was filled with buffer A, mixed by inversion, and centrifuged at 2600 rpm for 10 min at 4 "C (Sorval RC3C centrifuge). After discarding the supernatants, the crude pellets were resuspended in 30 ml of buffer A and the homogenization/centrifugation procedure was repeated. The nuclear pellets were resuspended in 16 ml of buffer A and transferred to a Beckman Ti 45 rotor tube, and 4 M (NH4)2S04, pH 7.9, was added to 0.3 M final concentration. After lysis of the nuclei by frequent mixing over 30 min at 4 "C, the DNA was sedimented at 35,000 rpm in a Beckman Ti 45 rotor for 60 min at 4 "C. The supernatant was collected, and proteins were precipitated by the addition of 0.2 g of solid (NH4)2S04/ml of supernatant. After 30 min at 4 "C, the precipitated proteins were pelleted by centrifugation at 37,000 rpm in a Beckman Ti 45 rotor for 15 min. The protein pellets were suspended in 6 ml of buffer B50 (50 mM Hepes, pH 7.8,50 mM KC], 0.1 mM EDTA, 10% glycerol containing freshly added 1 mM dithiothreitol and 1 mM phenylmethylsulfonyl fluoride). The protein suspension was dialyzed overnight at 4 "C against 100 volumes of buffer B50 with one change of buffer. Extracts were centrifuged at 10,000 rpm for 10 min at 4 "C (Beckman SW 28.1 rotor), and the supernatant was used directly for transcription factor isolation studies or frozen and stored in aliquots at -70 "C. Nuclear proteins isolated by this procedure varied from 35 to 40 mg in 5 ml of extract from 30 g of tissue. Proteins were determined by the method of Bradford.
Purification of ANF-2-Batchwise heparin-Sepharose chromatography was used to obtain partially purified extracts of ANF-2 prior to sequence-specific DNA affinity chromatography. Typically, nuclear extract (25 ml) from 150 g of tissue (approximately 10 livers) was diluted to 50 ml with buffer B50 and added to a beaker containing 50 ml of heparin-Sepharose previously equilibrated with the same buffer. The mixture was stirred at 4 "C for 1 h after which the gel was filtered, washed three times with 100 ml of buffer B50, and packed in a 25 X 125-mm column. The column was washed successively with 100 ml of 0.2,0.3,0.45, and 0.6 M KC1 in buffer B (buffer BlOO without KC1). The peak protein fractions from each salt wash were combined (30 ml) and analyzed for ANF-2 by mobility shift assay (see below). The partially purified ANF-2 preparation (30 ml, 15-25 mg) recovered with 0.45 M KC1 was dialyzed against buffer B100, incubated for 10 min at 4 "C with poly(dI.dC) (3 pg/ml) and added to 1 ml of buffer B100-preequilibrated DNA affinity resin containing the multimerized wild type oligonucleotide. The mixture was shaken gently at 4 "C for 1 h after which the gel was washed three times with 10 ml of buffer BlOO by repeated centrifugation (3500 rpm/2 min/4 "C), careful aspiration of the supernatant and mixing with buffer. The gel was packed in a 8.5 X 90-mm column and washed three times with 2 column volumes of 0.35 M KC1 in buffer B. ANF-2 was recovered by elution with 2 ml of buffer B containing 0.6 M KCl. In preparation for protein analysis by denaturing SDS-PAGE (20) recovered fractions were lyophilized after dialysis against 50 mM NH,HCO,, pH 8.0, 0.05% SDS. Proteins were detected by silver staining (21).
We observed during this study that the ability of ANF-2 to interact with sequence-specific DNA was maintained after exposure of rat nuclear extracts to heat at 85 "C for 10 min. Denatured protein was cleared from the extract by centrifugation giving an ANF-2 preparation of 8-10-fold increased purity and suitable for direct DNA affinity chromatography. However, in contrast to the above-described affinity chromatography procedure for heparin-Sepharose-purified ANF-2, conditions for optimal retention of ANF-2 from heated extracts, by both affinity resins, required a 2-fold increase in the relative amount of gel used together with a decrease in poly(d1. dC) concentrations to 1 rglml.
Denaturation-Renaturation of ANF-2-ANF-2 purified by sequential heparin-Sepharose and affinity chromatography was electrophoresed, eluted, and renatured by the protocol of Hager and Burgess (22). After denaturing SDS-PAGE on a 12.5% gel, 10-mm slices, cut throughout the length of the gel, were eluted overnight at 23 "C with buffer containing 100 pg of carrier BSA. Protein in gel eluates was precipitated with acetone, pelleted by centrifugation and washed once with ice-cold 80% acetone. The dried pellets were dissolved in 6 M guanidine hydrochloride, and the protein was renatured by overnight dialysis at 23 "C against buffer BlOO containing 0.1% Nonidet P-40. Slice fractions were assessed for ANF-2 by mobility shift assay.
For ANF-2 the labeled wild type oligonucleotide was generally used as the binding template. In a typical mobility shift reaction (total volume of 25 pl) used for analysis of nuclear extracts and fractions from heparin-Sepharose chromatography, sample aliquots were added to an incubation mix of 1 pg of poly(dI. dC) and variable amounts of competitor DNA (where applicable) in buffer B (with and without KCl). After 10 min at 23 "C, 4 pl of radiolabeled probe (approximately 0.2 ng, 15,000 cpm) was added and the incubation continued for 15 min at the same temperature. The final KC1 concentration was normally 80 mM, although this was varied from 50 to 150 mM during assay of chromatography fractions. For affinity-purified ANF-2 preparations the initial reaction included 40 ng of poly(dI.dC) with 10 pg of BSA and 0.1% Nonidet P-40. Protein-DNA complexes were analyzed by electrophoresis through nondenaturing 4% polyacrylamide gels in TBE buffer (89 mM Tris-borate, pH 8.0, 2 mM EDTA).
Electrophoresis was performed at room temperature at 100 V (10 V/ cm) for 30 min, then at 125 V (12.5 V/cm) for 1.5 h. The gels were dried and analyzed by autoradiography.
DNase I Footprint and Dimethyl Sulfate Protection Assays-The fragment used in DNase I footprint and dimethyl sulfate (DMS) protection assays corresponded to the 5'-flanking region (nucleotides -260 to +1) of the AGP gene. For the noncoding strand, Hind111 digestion of the plasmid pAGPCATA20.1 (14) gave a 764-base pair fragment which was end-labeled with T4 polynucleotide kinase and [Y-~'P]ATP as described above. Upon digestion with KPNI the appropriate fragment was isolated by electrophoresis through a 2% agarose gel. A similar procedure was used to isolate coding strand of the al-Acid Glycoprotein Gene 11113 template DNA after labeling with the Klenow fragment of DNA polymerase and [ ( u -~~P I~C T P . Footprint reactions were performed as described by Ohlsson and Edlund (19). Varying amounts (3.5-15 pg) of heat-treated rat liver extract were added to BlOO buffer containing 1 pg of poly(dI.dC) in 90-pl final volume. After 10 min at room temperature, the end-labeled DNA fragment (approximately 1 ng in 5 pl) was added and incubation was continued for 20 min on ice. Reactions were placed a t room temperature for 2 min, and 5 p1 of a 50 mM CaCI2, 100 mM MgC12 solution was added followed immediately by 5 pl of DNase I (10 ng/ pl). Reactions were terminated after 60 s by addition of 100 pl of stop solution (100 mM Tris-HCI, pH 8.0, 100 mM NaCI, 10 mM EDTA, 100 pg/ml proteinase K, 50 pg/ml tRNA, and 1% SDS). After incubation at 37 "C for 15 min, nucleic acids were recovered by phenol/ chloroform extraction and ethanol precipitation. Pelleted DNA was dissolved in 95% (v/v) formamide in 10 mM NaOH and 1 mM EDTA with tracking dyes, heated a t 90 "C for 3 min, and electrophoresed through a 6% polyacrylamide, 8 M urea gel. After drying, gels were autoradiographed a t -70 "C with Kodak XAR film. DMS protection assays were performed as described by Johnson et al. (24). Binding reactions (50 pl) were treated with 50 pl of DMS diluted 1:200 into standard DMS buffer (23). Incubation continued a t 25 "C for 2 min, and the reactions were quenched with 25 pl of DMS stop solution (23). Template DNA was extracted with phenol/chloroform and precipitated with ethanol. Samples were treated as outlined for guanine cleavage in chemical DNA sequencing (25).
Western Analysis-Western blot analysis was performed as described by Towbin et al. (26) using l2'1-protein A to visualize immunoreactive material. Samples of a purified, bacterially expressed C/ E B P fusion protein and affinity purified ANF-2 from heated and nonheated nuclear extracts were submitted to SDS-PAGE on a 15% gel. After electrotransfer to Immobilon P, C/EBP-related proteins were detected with a C/EBP-specific antibody directed against an internal 14-amino acid peptide (16).

RESULTS
Separate DNA binding studies with partially purified rat glucocorticoid receptor and nuclear extracts from the HTC (JZ.1) rat hepatoma cell line have previously revealed a combined region of protection from DNase I digestion extending between residues -121 and -68 (15). The footprinted region afforded by rat hepatoma HTC (JZ.l) cell nuclear extracts is shown divided into X, Y, and 2 domains (Fig. lA). Both Y and 2 regions are included in the wild type oligonucleotide (-90 to -64) (Fig. lA) used in mobility shift analyses with rat hepatoma cell nuclear extracts. The major retarded complex observed in these experiments was attributed to the specific binding protein, AGP nuclear factor 2 (ANF-2) (  1B). Close examination of the protected region within the rat AGP promoter revealed two regions of shared homology: HA, sequence -113 to -104, and HB, sequence -81 to -72 (Fig. lA). It was of particular interest that the upstream homologous region overlaps the recognition sequence for the glucocorticoid receptor (Fig. IA). The oligonucleotide, HA (-120 to -90)) which includes both the glucocorticoid response element and the HA sequence ( Fig. lA), was shown to compete efficiently with the wild type oligonucleotide probe for binding to ANF-2 (Fig. 1B). In separate experiments, a similar sized oligonucleotide containing only the HA sequence flanked by a random selection of nucleotides displayed identical competition properties for ANF-2 (not shown). An extension of the same study showed that mutation of the bases e, within the HA sequence, effectively negated this competition confirming the importance of at least one of these bases for ANF-2 interaction (27). Within the HA sequence, this critical region was directly comparable in positional location to the triplet of nucleotides, ACA, found in the homologous HB sequence to be required for binding to ANF-2 ( Fig. 1).  the GRE (-121 to -107) and the two homologous sequences HA (-113 to -104) and HB (-81 to -72). Bases important for ANF-2 interaction with these homologous sites and for ANF-1 recognition in the Y region are underlined. Also illustrated are oligonucleotides (wild type, Y and 2 mutants, and HA (-120 to -90)) used in this study in competition mobility shift analyses. B, DNA binding properties of the AGP promoter. Nuclear extracts from HTC (JZ.l) cells were incubated in DNA binding reactions with labeled wild type oligonucleotide and 1 pg of poly(dI.dC) with or without a 100-fold molar excess of competing DNA.

G G A A C A r m G t G C A A G A C A r h C C C A A G T~T~T G A~~G T G C I L T G c mal
With nuclear preparations from the HTC (JZ.l) cell line we occasionally observed, in mobility shift assays, the interaction of a second binding factor with the wild type (-90 to -64) probe (not shown). This binding protein, designated AGP nuclear factor 1 (ANF-l), was seen as a faster migrating complex than that observed for ANF-2 and appeared  (Fig. 1B). Throughout this study the wild type oligonucleotide (-90 to -64) and the corresponding Y and Z mutant oligonucleotides (Fig. 1) were used routinely in competition mobility shift assays to assess the DNA binding specificity of ANF-2 in crude and purified extracts.
Purification of ANF-2 from Rat Liver Nuclear Extracts-Preliminary examination of rat liver nuclear extracts by competition mobility shift analysis with the wild type oligonucleotide as the labeled probe confirmed the presence of an ANF-2-like factor with the appropriate DNA binding specificity (data not shown). Rat liver was then chosen as a convenient tissue source for the isolation of ANF-2 by sequence-specific DNA affinity chromatography. Two different affinity resins were employed. Based on the Kadonaga/Tjian methodology (17) we prepared a DNA affinity resin using the wild type oligonucleotide encompassing the Y and 2 domains from -90 to -64 of the AGP promoter (Fig. 1). Our initial experience with this gel indicated considerable interference from nonspecifically adsorbed proteins, probably due to the inherent charge effects resulting from the CNBr-activated mode of coupling (18). To overcome this problem and to eliminate the possible interaction of ANF-1, a factor with binding specificity for the Y domain, we prepared a novel gel from an aminolinked oligonucleotide containing a 3-base substitution in the Y region (see Y mutant, Fig. 1). Coupling of the amino-linked DNA was achieved via carbonyldiimidazole-activated Sepharose. The advantage of this coupling method is that, with extended incubation under the coupling conditions, remaining activated groups are hydrolyzed to the native Sepharose matrix leaving only the coupled oligonucleotide (18). Although the amount of coupled DNA in the conventional gel is 2-3 times higher (25 uersus 9 pg/ml) than in the amino-linked resin, both gels appear to perform equally well after adequate washing with salt buffers. Fig. 2 shows the results of a two-step purification for ANF-2 from rat liver nuclear extracts involving sequential heparin-Sepharose and DNA affinity chromatography. Mobility shift assays with the labeled, wild type oligonucleotide (Fig. 1) were used to monitor DNA binding activity in recovered fractions. A 0.45 M KC1 eluate from batchwise heparin-Sepharose chromatography contained most of the ANF-2 activity which was shown to display the correct binding specificity by competition analysis with both wild type and mutated oligonucleotides (Fig. 2, A and B). This step afforded a 12-15-fold increase in ANF-2 purity, and after dialysis against BlOO buffer the partially purified extract was equilibrated with nonspecific competitor poly(d1. dC) and submitted to sequence-specific DNA affinity chromatography. ANF-2 was recovered by elution with 0.6 M KC1 buffer, and competition analysis confirmed the DNA binding specificity of the isolated fraction ( Fig. 2, C and D). Denaturing SDS-PAGE of the 0.6 M KC1 fraction, followed by silver staining, indicated the selective elution of three proteins corresponding to a major band migrating at 29 kDa and two minor bands at 43 and 20 kDa (data not shown).
The affinity-purified ANF-2 preparation was characterized further after denaturation-renaturation of proteins eluted from SDS-PAGE gel slices (22). The recovered proteins were precipitated with acetone, dissolved in 6 M guanidine hydrochloride and allowed to renature by dialysis against guanidinefree buffer. Mobility shift analysis of the renatured fractions confirmed the presence of DNA binding activity in slices 6,7, 8, and 10 corresponding to molecular mass windows of 50-41, 41-34, 34-29, and less than 20 kDa, respectively (Fig. 3A). Although only the result of competition analysis for slice 10 activity is shown in Fig. 3B, specific DNA binding was demonstrated for both major activities observed in slices 6 and 10.
A survey of the literature indicated that the molecular sizes of the observed DNA binding fragments matched closely those reported for the CCAAT/enhancer-binding protein (C/EBP) (16) previously isolated from rat liver by Johnson et al. (24). Native C/EBP exists as a 42-kDa protein, but can be proteolyzed to smaller 30-, 20-, and 14-kDa fragments (16) (Fig. 4). A comparison of the C/EBP binding site and the homologous regions (HA, HB) on the AGP promoter indicated considerable similarity. The optimal C/EBP binding site consists of a dyad-symmetric recognition sequence containing the 5'-GCAAT-3' half-site (28). On alignment the homologous regions, HA and HB, respectively, show a 7 and 6 out of 9 direct correspondence to the bases in the C/EBP site (Fig. 4).

SDS-PAGE.
A, ANF-2, purified by se-l4J"ER quential heparin-Sepharose and affinity chromatography, was electrophoresed through a 12.5% SDS-polyacrylamide gel. Proteins eluted from slices taken throughout the length of the separating gel were renatured and analyzed by mobility shift assay. B, competition mobility shift analysis for renatured slice 10 activity. Aliquots containing the renatured protein were incubated in binding reactions with the labeled wild type oligonucleotide in the presence and absence of a 100-fold molar excess of competing DNA.

FIG. 4. Comparisons between ANF-2 and C/EBP. The con-
sensus site for C/EBP has been compiled from several viral enhancers by Ryden and Beemon (29), while the optimal recognition sequence has been described by Vinson et al. (28).
However, by including the single bases adjacent 5' to both sequences in the AGP promoter, both regions display a 7 out of 10 identity to the C/EBP recognition sequence (Fig. 4).
The upstream sequence (HA) conforms better with the consensus C/EBP site compiled from recognition sequences present in several genes ( Fig. 4) (29).
McKnight and collaborators (24) have reported that C/ EBP DNA binding activity is substantially unaffected when heated to 80 "C for 5-10 min. Heat treatment of rat liver nuclear extracts indicated a similar stability for ANF-2 (not shown). The chromatographic properties of ANF-2 activity from heated extracts were also shown to parallel those observed for unheated preparations (Fig. 5). Heated ANF-2 could be recovered from heparin-Sepharose with buffers containing 0.4-0.5 M KCl, similar to salt concentrations (0.45 M KCl) used for elution of ANF-2 during chromatography of nonheated nuclear extracts (see Figs. 2A and 5A). On heating, >SO% of the protein in the nuclear extract is denatured and precipitated resulting in a preparation enriched 8-10-fold for ANF-2. This facilitated further purification by direct sequence-specific DNA affinity chromatography. As seen for unheated preparations, ANF-2 from heated extracts was eluted from the affinity column with 0.6 M KC1 (see Figs. 2B and 5B). The appearance of ANF-2 activity in the flowthrough fraction (Fig. 5B) was attributed to an overloading of the affinity column and could be prevented by increased gel volume and lower poly(dI .dC) concentration. Binding sequence specificity for the recovered ANF-2 was again confirmed by competition analysis (Fig. 5C).
Rat Liver ANF-2 Is Closely Identical to CIEBP-ANF-2 from rat liver was confirmed to be highly related and probably identical to C/EBP after Western blot analysis with a C/ EBP-specific antibody raised against a 14-amino acid peptide situated N-terminal of the C/EBP DNA binding domain (16) (Fig. 6A). Major antibody-reacting proteins of approximately 40 and 30 kDa were observed for both heated and nonheated affinity-purified extracts, consistent with molecular sizes for native and proteolytically degraded C/EBP, respectively (Fig.  6A). In a separate experiment an affinity-purified extract from a heated nuclear preparation was submitted to denaturing SDS-PAGE followed by silver staining. Fig. 6B shows that most protein contaminants were removed by successive washing of the affinity resin with 0.35 M KC1 buffer. On elution with 0.6 M KCl, five major bands were observed corresponding to proteins of approximately 40 (doublet), 30 (doublet), and 20 kDa in size (Fig. 6B, lane 4 ) . We assumed initially that all of these proteins were C/EBP-related. However, after recovery of the individual proteins in these bands by elution, SDS-PAGE followed by Western analysis with the C/EBP peptide-specific antibody confirmed only the two 30-kDa proteins and the 20-kDa species as C/EBP-related fragments (not shown). The identity of the remaining 40-kDa proteins was not pursued. One possibility is that these proteins were also C/EBP-related, but that a low elution recovery from the SDS-PAGE gel may have prejudiced the signal intensity in Western analysis. Alternatively, the original affinity-purified extract may have contained only degraded C/ EBP-related, fragments accounting for the observed antibody reactive proteins.
The Glucocorticoid Regulatory Unit of the AGP Promoter Contains Multiple Interaction Sites for C/EBP-like Factors-Using mobility shift analyses with the labeled wild type and HA (-120 to -90) oligonucleotides (see Fig. 1) we were able to confirm the dual binding site specificity of an authentic, bacterially expressed C/EBP fusion protein for the homologous HA and HB regions (data not shown). To further assess the interaction of C/EBP-like factors with sequences within the glucocorticoid regulatory unit of the AGP promoter we performed DNase I footprint assays, on DNA templates spanning this region, using an ANF-2 preparation derived from heated rat liver nuclear extract. For both coding and noncoding strands the observed area of protection ranged from

Regulatory Elements
of the al-Acid Glycoprotein Gene FIG. 5. Chromatographic properties of ANF-2 from heated rat liver nuclear extracts. After heat treatment (85 T / 1 0 min), precipitated protein was pelleted by centrifugation and the soluble extract was submitted separately to heparin-Sepharose and DNA affinity chromatography. The salt elution properties of ANF-2 from both adsorbants were monitored by mobility shift analysis. A, heparin-Sepharose chromatography. Load, heat-treated nuclear extract;  A , Western blot analysis of affinity-purified ANF-2 from heated and nonheated rat liver nuclear extracts. Purified proteins were separated by gel electrophoresis and analyzed by Western blotting using C/ E B P peptide specific antibody (16). Antibody reaction with an 11-kDa, bacterially expressed C/EBP fusion protein (16) is shown in the left-hand lane. B, analysis of affinity-purified ANF-2 from heated rat liver nuclear extracts by SDS-PAGE. Heated nuclear extract was cleared by centrifugation and added directly to amino-linked affinity resin (1 ml) for batchwise incubation. After extensive washing with buffer BlOO the gel was packed in a column and eluted three times with 2 ml of 0.35 M KCI. ANF-2 was recovered with a final 2-ml wash of 0.6 M KCl. All 0.35 and 0.6 M KC1 fractions were dialyzed against 50 mM NH,HCOa buffer containing 0.05% SDS and lyophilized. After electrophoresis on a 15% (w/v) SDS-polyacrylamide gel, the separated proteins were visualized by silver staining. Lanes 1-3,  residues -115 to -65 (Fig. 7A) and was identical to that previously reported by our group for crude extracts from rat hepatoma HTC (JZ.l) cells (15).
Rat liver ANF-2 interaction with recognition sequences  7B). There was no evidence of altered methylation on the noncoding strand for the HA recognition site (Fig. 723). However, marked changes in methylation pattern were apparent for residues located immediately downstream of the homologous HA sequence (Fig. 7B). Bound protein strongly enhanced the methylation of the guanine at -97 (Fig. 7B). From an autoradiography obtained by shorter film exposure (not shown) it was also clearly evident that methylation of the adjacent guanine at -96 had been inhibited (Fig. 7B). Methylation protection analysis of the coding strand revealed a weak protection, in the presence of rat liver ANF-2, of the guanine at -79 within the HB sequence (Fig. 7B). Immediately upstream of this binding element, methylation of the guanine residue at -83 was seen to be augmented (Fig.  7B). ANF-2 interaction with the HA recognition site resulted in less effective methylation of the guanine at -109 and was accompanied by increased alkylation of the guanine and adenine residues at -105 and -106, respectively (Fig. 7B). In between the homologous regions HA and HB, the pattern of methylation on the coding strand was unaltered by ANF-2 interaction except for enhanced methylation of the guanine at -93 and the adenine residue at -94 (Fig. 7 B ) . A summary of the observed pattern of protection and enhancement on both strands is presented in Fig. 7C. Our results confirm HA and HB as recognition sites for C/EBP-like factors and are consistent with the presence of a third interaction site, designated HX, for the protein located between residues -102 and -93 (Fig. 7C). This putative binding sequence, ATTTCCCAAG, is closely comparable with the consensus recognition sequence for C/EBP (see Figs. 4 and 7C).

DISCUSSION
We have used affinity chromatography with a matrix containing the HB sequence, derived from the glucocorticoid regulatory unit of the AGP promoter, to isolate the nuclear sequences. Uniquely labeled AGP fragments spanning residues -260 to +1 were incubated with 20 pg of BSA ( l a n e I ) or increasing quantities of heat-treated rat liver nuclear extract ( l a n e 2, 3.5 pg; lane 3, 7 pg; lane 4, 15 pg) and then subjected to partial DNase I digestion. Reaction products were electrophoresed on a 6% polyacrylamide, 8 M urea sequencing gel. B, DMS protection analysis. DNA fragments from coding and noncoding AGP strands were incubated with BSA ( l a n e 1 ) or 15 pg of heated rat liver extract under conditions identical to those for DNase I footprinting. After methylation of exposed guanine residues with DMS, DNA fragments were cleaved with piperidine and analyzed on a 6% sequencing gel as described above. 0, guanine residues protected from DMS upon ANF-2 binding; 0, hypermethylated guanine or adenine residues. C, summary of DMS protection analysis showing residues with altered methylation a t sites HA, HX, and HB. Symbols indicating protected and hypermethylated residues are as described in B.
factor ANF-2 from rat liver. Several points of evidence indicate that the isolated protein is highly related to C/EBP, a heat-stable DNA-binding protein initially identified in rat liver nuclear extracts (16, 24). ANF-2 is similarly stable to thermal denaturation, and both proteins display a dual binding site specificity by interacting with both homologous regions, HA and HB, on the AGP promoter. ANF-2 was confirmed by immunological analysis to be very similar and probably identical to C/EBP. Our evidence from methylation protection analysis suggests that C/EBP also recognizes a third binding sequence, HX, located between the homologous HA and HB regions, Chang et al. (30) have described three analogous motifs for C/EBP-like factors in the corresponding region of the mouse AGP gene. These authors have recently identified the murine C/EBP-related factor, AGP/EBP, which was shown to interact with an oligonucleotide incorporating sequences equivalent to HA and HX on the mouse AGP promoter (30). Whether this interaction involves one or other of these recognition sites, or both, was not clearly demonstrated.
While C/EBP appears capable of interacting with DNA sequences that significantly diverge from its consensus binding site (TGNNGTAAG) (29) the homologous sequences, HA and HB within the AGP promoter, match very closely the optimal recognition site (ATTGCGCAAT) (28) for C/EBP. C/EBP is a major regulatory protein with a basic DNA binding region related in amino acid sequence to the transforming proteins Myc, Fos, and Jun (16). A common structural motif, termed the "leucine zipper," facilitates the sequence-specific interaction of these proteins with DNA (28,31). In early studies, Southern blot analysis, using the conserved DNA binding/leucine zipper domain of C/EBP, sug-T C T gested the existence of additional C/EBP-related proteins? Indeed cDNA clones encoding a number of these factors have recently been isolated and characterized (32)(33)(34)(35). DBP, a recently described liver-specific transcription factor (341, shares the conserved basic domain, but lacks a leucine zipper structure, and in this respect differs from other members of the C/EBP family (34). However, NF-IL6, an interleukin-linducible factor (32); Ig/EBP-1, a ubiquitously expressed immunoglobulin enhancer-binding protein (35); and LAP (liver activating protein) (the AGP/EBP rat homolog), a factor found to be highly enriched in rat liver nuclei (33), all contain potential leucine zippers and are highly homologous to C/ EBP in the C-terminal DNA binding region. Both Ig/EBP-1 and LAP have a demonstrated ability to form heterodimeric complexes with C/EBP (33,35). Additionally, all of these proteins, including DBP (D binding protein), share common recognition sequences (32)(33)(34)(35). Of considerable interest is the observation that NF-IL6 binds to interleukin-6 responsive elements in the regulatory regions of several acute-phase protein genes implying a role for NF-IL6 in gene regulation during the acute-phase reaction (32). The recognition sequence (GTTGTGCAAT) for NF-IL6, pertaining to the interleukin-6 responsive element on the AGP promoter, lies approximately 5 kilobases upstream of the transcriptional initiation site (11) and is closely comparable with the homologous binding sites, HA and HB, located within the glucocorticoid regulatory unit (see Fig. 4). This report presents another example of the proximal location of recognition sequences for C/EBP-related nuclear factors and a glucocorticoid responsive element. Glucocorticoid-induced transcription of the rat angiotensinogen gene is S. L. McKnight, personal communication. of the al-Acid Glycoprotein Gene mediated by a multimodular enhancer consisting of an acutephase response element flanked on both sides by two functionally distinct GREs (36). The acute-phase response element has been shown to bind a cytokine-inducible NFKB-like factor and constitutive C/EBP-like proteins in a mutually exclusive manner (37). Complete glucocorticoid responsiveness of the angiotensinogen gene appears to be dependent on the association of either one of these DNA-binding proteins with the acute-phase response element leading to the formation of a synergistic enhancer complex together with activated glucocorticoid receptor (36). It is of note that the angiotensinogen acute-phase response element contains the sequence ATTTCCCAAC which is homologous to region HX on the AGP promoter. Our results suggest that HX may be an additional interaction site for C/EBP-related proteins (see Fig. 7C). Similar sequences have been previously reported to be associated with glucocorticoid regulatable elements of the a2u-globulin and tryptophan oxygenase genes (38,39).
Within the glucocorticoid regulatory unit of the AGP promoter we have identified three regions of interaction for C/ EBP-like factors. Two of these recognition sites, HX and HB, are located downstream of the GRE. The third C/EBP site, HA, overlaps the hormone response element suggesting a mechanism in which the enhancer-binding protein may attenuate the glucocorticoid inducibility of AGP by competing with glucocorticoid receptor for a common binding site. An alternative scenario, in which C/EBP and the glucocorticoid receptor are accommodated at the same binding locus by contacting opposite sites of the DNA helix, may result in positive proteinprotein interactions between both transcriptional activators leading to efficient glucocorticoid-mediated induction from the AGP promoter. The mutual influences of glucocorticoid receptor and C/EBP on their individual interactions with their respective recognition sequences in the GRE/HA region can be tested by analyzing nucleotide-protein contacts in DNA binding studies using protection assays and interference techniques (40-42). Such studies should provide a clearer definition of how these transcription factors are accommodated together on the AGP promoter. A similar study has recently been reported for glucocorticoid receptor and nuclear factor 1 on the mouse mammary tumor virus promoter (43). An additional consideration regarding occupancy of the HA interaction site by C/EBP-like factors is the possible restrictive influence (mediated through steric effects) of a related nuclear factor located on the adjacent recognition sequence at HX. For efficient binding the spacing (center to center) between tandemly linked C/EBP recognition sites should be at least 16 base pairs? On the AGP promoter HA is separated from HX by 12 nucleotides and the HB site resides an additional 20 base pairs downstream (Fig. 7C). It is conceivable that after sequential recruitment of C/EBP by the HB and HX interaction sites, binding to the HA site might not be favored.
The initial observation of ANF-2-like interaction with the AGP promoter was made using soluble nuclear extracts from HTC (JZ.l) rat hepatoma cells. with recent evidence that C/ EBP expression might be restricted to terminally differentiated cells (44) it was distinctly possible that the ANF-2 activity present in the proliferating cell line was not C/EBP, but an additional member of the C/EBP nuclear family. We have now further characterized ANF-2 from HTC cells, and our evidence suggests that this factor is not reactive to the C/ EBP peptide antibody either in mobility shift experiments or in Western analysis. These results and those relating to a functional analysis of the C/EBP binding sites on the AGP C. R. Vinson, personal communication.
promoter, are presented in a separate report (45).