The Cellular Transcription Factor CREB Corresponds To

MOLECULAR AND CELLULAR BIOLOGY, Dec. 1990, p. 6192-6203 Vol. 10, No. 12 0270-7306/90/126192-12$02.00/0 Copyright C 1990, American Society for Microbiology

The Cellular Transcription Factor CREB Corresponds to Activating Transcription Factor 47 (ATF-47) and Forms Complexes with a Group of Polypeptides Related to ATF-43 HELEN C. HURST,' NORMA MASSON,2 NICOLAS C. JONES,3 AND KEVIN A. W. LEE2* Gene Transcription Group, Imperial Cancer Research Fund, Hammersmith Hospital, Ducane Road, London W12 OHS,' Gene Regulation Group, Imperial Cancer Research Fund, London WC2A 3PX,3 and Imperial Cancer Research Fund, Clare Hall Laboratories, South Mimms, Hertfordshire EN6 3DL,2 England Received 7 May 1990/Accepted 30 August 1990

Promoter elements containing the sequence motif CGTCA are important for a variety of inducible responses at the transcriptional level. Multiple cellular factors specifically bind to these elements and are encoded by a multigene family. Among these factors, polypeptides termed activating transcription factor 43 (ATF-43) and ATF-47 have been purified from HeLa cells and a factor referred to as cyclic AMP response element-binding protein (CREB) has been isolated from PC12 cells and rat brain. We demonstrated that CREB and ATF-47 are identical and that CREB and ATF-43 form protein-protein complexes. We also found that the cis requirements for stable DNA binding by ATF-43 and CREB are different. Using antibodies to ATF-43 we have identified a group of polypeptides (ATF-43*) in the size range from 40 to 43 kDa. ATF-43* polypeptides are related by their reactivity with anti-ATF-43, DNA-binding specificity, complex formation with CREB, heat stability, and phosphorylation by protein kinase A. Certain cell types vary in their ATF-43 complement, suggesting that CREB activity is modulated in a cell-type-specific manner through interaction with ATF-43*. ATF-43* polypeptides do not appear simply to correspond to the gene products of the ATF multigene family, suggesting that the size of the ATF family at the protein level is even larger than predicted from cDNA-cloning studies.

Transcriptional activation of eucaryotic genes occurs, in and by other promoter elements (25, 49, 55). Second, mul- part, through the interaction of sequence-specific DNA- tiple cellular activator proteins interact with ATF-binding binding proteins with promoter and enhancer elements (8, sites, as described below. 24, 47, 58, 66, 77, 86). One poorly understood aspect of Gel mobility shift assays demonstrate that multiple factors transcriptional activation concerns the ability of some pro- bind directly to ATF sites (3, 15, 37, 40, 41, 43, 44, 57, 60, 62, moter elements to participate in disparate transcriptional 79, 88) and sequence-specific DNA affinity chromatography responses. Transcription factor families, the members of has enabled purification of some of these binding activities which have similar or identical DNA-binding specificities (1, 4, 15, 18, 34, 40, 44, 70, 75, 92). Of significance to the (20, 27, 35, 85, 91, 97), clearly play a major role in expanding studies described here is that a family of immunologically the activation potential of single promoter elements. How- related polypeptides of -43 and 47 kDa (referred to as ever, the mechanisms by which a variety of transcriptional ATF-43 and ATF-47, respectively [34, 40]) have been puri- responses are generated from groups of factors with appar- fied from HeLa cells and a factor of similar size, designated ently indiscriminate DNA-binding activity are not yet clear. CREB response has been Promoter elements containing the critical core motif (cAMP element-binding protein), from PC12 cells and rat brain The CGTCA (referred to here as activating transcription factor purified (70) (100). between CREB and ATF has not been estab- [ATF]-binding sites) are important for the activity of several relationship promoters that respond to different signals. ATF-binding lished, although in light of the above observations, CREB sites are implicated in positive control by the adenovirus Ela has been assumed to correspond to a member of the ATF protein (7, 9, 31, 40, 56, 57, 88, 102) and intracellular cyclic family. AMP (cAMP) (2, 14, 17, 19, 26, 38, 59, 71, 80, 82, 84, 87, 94) cDNA-cloning studies have demonstrated that an even and may also be involved in the function of enhancer more complex array of factors can bind to ATF-binding elements that respond to the human T-cell lymphotropic sites. First, it has been demonstrated that members of the virus type I (HTLV-1) transactivator p40tax (10, 28, 30, 45, Jun family can bind to ATF-binding sites when associated 69, 72) and the bovine leukemia virus transactivator p38tax with members of the ATF family as heterodimers (6, 42, 64). (51). In addition, the hydroxymethylglutaryl (HMG)-coen- Second, cDNAs corresponding to a multigene family (the zyme A reductase and 1-DNA polymerase promoters con- ATF family) encode several distinct proteins that bind to tain ATF-binding sites and are under negative control by ATF-binding sites (33, 35, 39, 50, 65). All of these proteins sterols (74) and p4Otax (46, 99), respectively. The functional bind to DNA as dimers through the action of a "leucine diversity of ATF-binding sites probably results from inter- zipper" that is required for dimerization and an adjacent action of multiple cis and trans determinants. First, the basic region that makes direct contacts with DNA (35). The ability of ATF sites to activate transcription is influenced by ATF family is therefore part of a broader class of activator sequences in close proximity to the CGTCA motif (19, 55) proteins that has been referred to as bZIP proteins (96) and includes the Jun family (11-13, 29, 36, 52, 83, 95). As is the case for other bZIP proteins, some members of the ATF * Corresponding author. family selectively form heterodimers (35), thus further in- 6192 VOL. 10, 1990 CREB CORRESPONDS TO AND FORMS COMPLEXES WITH ATFs 6193 creasing the repertoire of factors that might participate in embryonal carcinoma cells were maintained on standard specific transcriptional responses. tissue culture dishes coated with 0.1% gelatin (-300 bloom) In light of these observations it is of importance to identify from porcine skin (Sigma G-2500). For comparative experi- ATF (CREB)-related heterodimers that exist in vivo. Here, ments, cells were harvested at similar confluence one day we demonstrate that CREB is identical to ATF-47 and forms after addition of fresh media. For in vivo labeling with 32pi, complexes with ATF-43. We have identified a new group of HeLa cells in suspension were placed in phosphate-free E4 polypeptides (ATF-43 ) that share a range of properties with medium containing 5% dialyzed FBS (dialyzed against 50 ATF-43, including complex formation with CREB. ATF-43 mM NaCI) and 200 ,Ci of 32p; per ml and labeled for 12 h. polypeptides differ in their cell type distribution, suggesting Nuclear extracts were prepared by using a small-scale pro- that CREB activity may be modulated through interaction cedure as previously described (54). For in vivo labeling with with other members of the ATF family in a cell-type-specific [35S]methionine, monolayer cells were placed in methionine- manner. We also provide evidence that the cis requirements free E4 medium containing 5% dialyzed FBS (dialyzed for stable DNA binding of ATF-43 and CREB are different, against 50 mM NaCl) and 200 ,uCi/ml of [35S]methionine suggesting a mechanism that can contribute to the differen- (specific activity, 1,000 Ci/mmol, in vivo cell-labeling grade tial transcriptional activity of ATF-binding sites. from Amersham) and labeled for 12 h. Nuclear extracts from 35S-labeled cells were produced as described below. MATERIALS AND METHODS Small-scale nuclear extracts. Nuclear extracts were prepared by the same procedure for all monolayer cells used. Purification of ATF-43 from HeLa cells. ATF-43 was puri- Cells were washed twice with phosphate-buffered saline fied from suspension cultures of HeLa cells as previously (PBS), harvested, and resuspended in -300 ,ul of ice-cold described (40). In later preparations, the DNA-Sepharose lysis buffer (20 mM HEPES [pH 8.0], 20 mM NaCl, 0.5% column was omitted and instead the crude extract was Nonidet P-40, 1 mM DTT, protease inhibitors [0.5 mM heated for 10 min at 65°C and denatured proteins were PMSF, 2 p.g of leupeptin per ml, and 2 ,g of trasylol per ml]) removed by centrifugation (34). The supernatant was applied per 107 cells. Cells were left on ice for 5 min and centrifuged to a Biorex-70 column as previously described and step for 1 min at top speed in an Eppendorf microcentrifuge at eluted with 0.15, 0.3, 0.6, and 2 M KCl in dialysis buffer (20 room temperature. The crude nuclear pellet was resus- mM HEPES [N-2-hydroxyethylpiperazine-N'-2-ethane- pended in 60 pI of buffer (20 mM HEPES [pH 7.9], 25% sulfonic acid]-KOH [pH 8.0], 20o glycerol, 1 mM MgCI2, [vol/vol] glycerol; 420 mM NaCl, 1.5 mM MgCl2, 0.2 mM 0.5 mM EDTA, 0.1% LDAO; 0.5 mM phenylmethylsulfonyl EDTA, 1 mM DTT, protease inhibitors [0.5 mM PMSF, 2 ,ug fluoride [PMSF]; 1 mM dithiothreitol [DTT]). The 0.6 M of leupeptin per ml, and 2 ,g of trasylol per ml]) per 107 cells, fraction was dialyzed to 0.3 M and applied to a DNA affinity left on ice for 15 min, resuspended again, and left for a column prepared by using the ATF site from position -50 in further 15 min. Nuclear debris was removed by centrifuga- the adenovirus E4 promoter (40, 56). The column was tion for 1 min in a microcentrifuge at room temperature, and washed extensively with 0.3 M buffer and bound protein was the supernatant was diluted to a low salt concentration by eluted with 1.0 M buffer. Fractions containing ATF were the addition of 2 volumes of 20 mM HEPES (pH 7.4). Heat dialyzed back to 0.3 M KCI for another round of affinity denaturation at 60°C for 7 min (or variations as indicated) of chromatography. The above changes resulted in better this diluted extract produced the heat-denatured nuclear yields but produced a doublet band of -40 and 43 kDa, in extract used for most of the experiments described. Protein contrast to the previously described singlet of 43 kDa (40). concentrations were determined by using a dye assay (Bio- Heat denaturation of the nuclear extract appears to result in Rad). production of the lower-molecular-weight polypeptide (data Small-scale, sequence-specific DNA affinity purification.32P- not shown). or 35S-labeled heat-denatured nuclear extracts were incu- Production of antibodies to CREB and ATF-43. Rabbits bated batchwise with DNA affinity resin by resuspending the were immunized with a synthetic peptide corresponding to resin several times over a period of 30 min in a 1.5-ml the carboxy-terminal 10 amino acids of CREB (33, 39) microcentrifuge tube at room temperature. For a typical conjugated to thyroglobulin. The antibody produced is re- experiment, 200 ,u of nuclear extract (derived from -10' ferred to as anti-CREB throughout this paper. For produc- cells) was incubated with 20 RI of affinity resin. Incubations tion of anti-ATF-43 antibodies, 0.5 ,ug of purified ATF-43 in were performed in the presence of specific and nonspecific -300,ll of 0.1 M KCI dialysis buffer was precipitated with 90 competitors (5 ,uM oligonucleotides containing wild-type or ,ul of 10% potassium alum and 12 RI of 4 M NaOH. Each of mutated ATF-binding sites or 4 Rg of poly[dI/dC] per ml) three female Black 10 x BALB/c mice was injected subcu- which were added to the DNA affinity resin before it was taneously with this precipitate and boosted twice at fort- mixed with nuclear extract. After binding, the resin was nightly intervals. Tail bleeds were then tested for activity washed with five changes of affinity column wash buffer (20 against ATF compared with preimmune sera from the same mM HEPES [pH 7.4], 100 mM KCI), and the bound proteins mice. A week prior to subsequent tail bleeds, the mice were were eluted with 1 M KCI and added directly to sodium boosted as described above. The three antisera produced are dodecyl sulfate-polyacrylamide gel electrophoresis (SDS- referred to collectively as anti-ATF and individually as Cl, PAGE) sample buffer. C2, and C3. All experiments were performed with crude In vitro transcription and translation. 35S-labeled CREB anti-CREB or anti-ATF sera. protein was synthesized in vitro by transcription of CREB Cell culture and in vivo labeling. All cell lines were cDNA by using 17 RNA polymerase (Boehringer) and maintained as monolayers in Dulbecco modified Eagle me- subsequent translation in rabbit reticulocyte lysate (Amer- dium. HeLa, F9 (mouse embryonal carcinoma), JEG3 (hu- sham). Transcription reactions were as previously described man choriocarcinoma), CHO, and 293 cell cultures con- (35) and contained 500 ,uM m7GpppG and 100 ,uM GTP to tained 10% fetal bovine serum (FBS). PC12 cultures ensure efficient capping of transcripts. In vitro translation contained 10% FBS and 5% horse serum, and MEL28 was performed according to the manufacturers' instructions. cultures contained 10% calf serum. Undifferentiated F9 DNA probes and competitors. Oligonucleotides containing 6194 HURST ET AL. MOL. CELL. BIOL. consensus wild-type or mutated ATF-binding sites were as lulose, as previously described for Western blots. The filter follows. The ATF-binding site core motif TGACGTCA or was incubated at 4°C overnight in buffer containing 25 mM corresponding mutated sequence is underlined. (i) WT con- HEPES-KOH (pH 8.0), 10% glycerol, 150 mM KCl, 1 mM tains the sequence GGATCCATGACGTCATGGATC. (ii) EDTA, 0.1% LDAO, 5 mM DTT, and 2.5% Marvel. Probe Ml contains the sequence GGATCCATGAATTCATGGA was prepared by labeling 0.5 p.g of the double-stranded E4 TC. (iii) M2 contains the sequence GGATCCATGGCGCCA oligonucleotide described above with [y-32P]ATP and T4 TGGATC. (iv) f contains the following sequence from the polynucleotide kinase (BRL). The labeled oligonucleotide human ,B-globin pre-mRNA: GCCCTCTATTTTCCCACCC (with HindIl overhangs) was then concatemerized over- TTAGG. E2A contains the following duplicated ATF site night by incubation with T4 DNA ligase. Probing of filters from the E2A promoter: CTACGTCATCTGATGACGTAT was carried out in 5 to 10 ml of blocking buffer without TTCAGATC. The E2A oligonucleotide was concatemerized Marvel for 60 to 90 min at room temperature. The filter was and used to make the BSE2 affinity resin used for the off-rate then washed on ice for 20 min in large volumes (four experiment (see Fig. 10). The BSE4 affinity resin was made changes) of the same buffer (without glycerol or Marvel but from a concatemerized oligonucleotide with the following containing 200 mM KCI), dried, wrapped in Saran Wrap, and repeated sequence: CCCGGGATGACGTCAT. autoradiographed. In vitro labeling with protein kinase A and immunoprecipitations. Proteins in heat-treated nuclear extracts were labeled as follows. Extracts were adjusted to kinase-labeling RESULTS conditions by addition of 4 x kinase buffer (200 mM Antibodies that distinguish CREB and ATF-43. Sequence- KH2PO4-K2HPO4 [pH 7.2], 40 mM MgCI2, 10 ,uM ATP), 1 specific DNA affinity chromatography has been employed to p.Ci of [32P]ATP (5,000 Ci/mmol) per RI and 1 U of the purify multiple proteins that bind directly to ATF-binding protein kinase A catalytic subunit (Sigma P2645) per p.1. sites (1, 4, 15, 18, 34, 40, 44, 70, 92). Of these proteins, ATF Labeling reactions were performed at 30°C for 30 to 60 min. has been defined as two major polypeptides (34, 40), desig- For immunoprecipitations, crude antiserum (1 to 3 plI/100 pul) nated ATF-43 (43 kDa) and ATF-47 (47 kDa), from HeLa was added directly to the labeling reactions and left on ice cells and JY cells. Another factor, designated CREB (cyclic for 1 h. Immunoprecipitates were collected on protein AMP response element-binding protein), has been purified A-Sepharose beads (10 ,ul of packed beads/i ,ul of antiserum from PC12 cells (70) and rat brain (100) as an -43 kDa added) by mixing at room temperature for 20 min. Following polypeptide and subsequently cloned (33, 39). These obser- removal of supernatant, protein A beads were washed with vations suggested that CREB corresponds to ATF-43. five changes (700 ,ul each change) of wash buffer (25 mM To address this question we have used polyclonal antibod- HEPES [pH 7.0], 125 mM NaCl, and 0.1% Nonidet P-40) ies to ATF-43 purified from HeLa cells and an anti-peptide and finally with wash buffer in the absence of Nonidet P-40. antibody directed against the C terminus of CREB (33, 39) Washed beads were resuspended in SDS-PAGE sample and established the specificities of these anti-sera (Fig. 1 and buffer, boiled, and electrophoresed. Gels were fixed in 50% 2). [35S]methionine-labeled CREB (Fig. 1A, lane L) was methanol and dried for autoradiography. synthesized in vitro by transcription and translation of a Western immunoblot analysis. Proteins were resolved by full-length CREB cDNA. Anti-CREB (Fig. lA, lane C) 10% SDS-PAGE and transferred to nitrocellulose filters for specifically immunoprecipitated a protein of the expected 90 min at 100 V or 300 mA. Filters were incubated for 1 h in size of CREB (32, 33) and this protein was not precipitated 1% bovine serum albumin in PBS-T (PBS containing 0.05% by preimmune serum (Fig. 1A, lane P). CREB produced in Tween 20) to block nonspecific binding, incubated with vitro bound specifically to a DNA affinity column containing anti-ATF (crude C3 anti-serum diluted 1:200 in blocking ATF (CREB)-binding sites (Fig. 1A). Binding was prevented solution for 1 h) or anti-CREB (crude anti-serum diluted by an oligonucleotide containing a wild-type ATF-binding 1:200 for 3 h), and then washed six times in PBS-T over 20 site (Fig. 1A, lane WT) and not by an oligonucleotide min. The second antibody was either alkaline phosphatase- containing a mutated site (Fig. 1A, lane Ml). conjugated goat anti-mouse (Bio-Rad) antibody or peroxi- Three independent polyclonal mouse antisera (referred to dase-conjugated swine anti-rabbit (DAKO) antibody diluted as Cl, C-2, and C-3) were raised against purified HeLa cell 1:1,000 in blocking solution, incubated for 1 h at room ATF-43 containing no detectable ATF-47 (40). Western blot temperature and washed as described above. analysis and immunoprecipitation demonstrated that all Gel mobility shift assays. Appropriate concentrations of three antibodies recognized purified ATF-43 (Fig. 1B) (As- purified ATF, in vitro-made CREB, and 32P-labeled probe signment of -43 kDa as the apparent molecular mass for were determined empirically (40). Purified ATF-43 or in vitro ATF-43 is on the basis of previous determinations [34, 40] translation mixture was diluted in 20 pI of gel retention and our data [Fig. 2B and 3]. The size markers used here are buffer (25 mM HEPES-KOH [pH 8.0], 0.5% bovine serum the same as those used to describe ATF-43 and ATF-47 [34], albumin, 1 mM EDTA, 10% glycerol, 150 mM KCI, 5 mM and the migration of ATF-43 and ATF-47 in our experiments DTT, 0.5 mM PMSF) and incubated with crude antisera for is in agreement with previous data [34]. ATF-43 appears to at least 4 h or overnight at 4°C. Approximately 0.1 ng of migrate -35 to 40 kDa, according to a different set of size 32P-labeled, 24-bp oligonucleotide corresponding to the markers [Fig. 1B]. For the purpose of discussion, we have ATF-binding site at position -50 in the adenovirus E4 assigned apparent molecular weights according to the size promoter (56) was added and incubated for 20 min at room markers used in Fig. 2B and 3 and in other studies [34].) In temperature. The samples were then loaded directly onto an addition, these antibodies detected an additional, faster- 8% polyacrylamide gel (acryl-bis ratio, 44:0.8) and run in migrating species, the significance of which is addressed 0.5x TBE buffer for 90 min at 200 V. Gels were then fixed below. Gel mobility shift assays using purified ATF-43 also and dried for autoradiography. demonstrate that Cl and C2 antibodies recognized ATF-43 Southwestern (DNA-protein) blot analysis. Samples con- (Fig. 1B). Cl produced a DNA-protein complex with lower taining -85 ng of affinity-purified ATF were run on SDS- mobility than that of the complex generated by ATF-43 PAGE with 14C-labeled markers and transferred to nitrocel- alone, whereas C2 inhibited binding of ATF-43 to DNA. VOL. 10, 1990 CREB CORRESPONDS TO AND FORMS COMPLEXES WITH ATFs 6195

A Q.C. A AT F CREB B kDa u1m1-V a 1 -.-W- - .-.._. e ~~~..*

CREB -e -6' -so 75k _ CREB.- - u -§ N 50k _ -39 CREB 5 ]- ATF 39k -_ 27k - IMMUNOPRECIPITATI DNA AFFINITY

B M P C0 C2 C3 C N PC1C2 f _

92k -_ 92k -_ 69k- - _A 69k _ 46k- 46k N_ a ;. ATFt[ mm ~0K- 30k -- FIG. 2. (A) Effect of anti-CREB and anti-ATF-43 on DNA binding. Samples of affinity-purified ATF-43 and in vitro-made CREB were incubated with 2 p.1 of crude antiserum as indicated above the figure for 4 h at 4°C. Lanes: P, preimmune serum; aC, WESTERN IMMUNOPRECIPITATION GEL anti-CREB serum; mouse mouse MOBILITY Cl, Cl antiserum; C2, C2 antiserum. A standard gel mobility shift was then performed, using a FIG. 1. Antibodies to CREB and ATF-43. (A) Rabbit anti-pep- 32P-labeled oligonucleotide probe containing an ATF (CREB)-bind- tide antibody to CREB. Total 35S-labeled products from an RTL ing site. (B) Detection of ATF-43 (with C3 antibody) and CREB translation reaction programmed with CREB RNA (lanes L) were (with anti-CREB) in crude HeLa cell nuclear extracts by Western immunoprecipitated with anti-CREB (IMMUNOPRECIPITA- blotting. Size markers are indicated on the left. k, Kilodaltons. TION) or purified with a sequence-specific DNA affinity resin (DNA AFFINITY). Immunoprecipitation was with crude preimmune serum (lane P) or anti-CREB (lane C). DNA affinity purification was in recognized multiple species in the size range from -40 to 43 the presence of 5 ,uM mutant oligonucleotide competitor (lane Ml) kDa, whereas anti-CREB recognized a distinct -47-kDa or 5 puM oligonucleotide containing a wild-type ATF (CREB) polypeptide and an additional polypeptide of -65 kDa. The binding site (lane WT). (B) Mouse polyclonal antisera to HeLa cell significance of this latter polypeptide is currently unknown. ATF-43. Purified HeLa cell ATF-43 (60 ng) was Western blotted results (WESTERN) against crude mouse polyclonal antibody (unmarked Taken together, the above demonstrate that ATF-43 are was lane on right; C3) or 32P labeled and immunoprecipitated (IMMU- and CREB different factors. In addition, there no NOPRECIPITATION) with crude preimmune (lane P) or three cross-reactivity between anti-ATF-43 and CREB or anti- independent mouse antisera (lanes Cl, C2, and C3). Size markers CREB and ATF-43, either in solution (Fig. 2A) or on are indicated at the left (unmarked lane on left and lane M). k, denaturation (Fig. 2B and 3). Kilodaltons. (C) Recognition of DNA-protein complexes by anti- CREB corresponded to ATF-47 and formed complexes with ATF-43 (GEL MOBILITY). Affinity-purified ATF was incubated ATF-43. The above experiments indicated that, in contrast overnight at 4°C with no addition (lane N) or 2 [L1 of mouse serum to previous estimations (34, 40, 70), the mobility of CREB from preimmune (lane P), or C2 mice. An excess of 32P-labeled Cl, was more similar to that of ATF-47 than to that of ATF-43. oligonucleotide containing an ATF (CREB)-binding site was then To test the that to added, and a standard gel mobility assay was performed. possibility CREB corresponded ATF-47 we analyzed CREB and ATF-47 on SDS-polyacrylamide gels (Fig. 3). HeLa cell protein was labeled with 32p in vivo Thus, although these ATF antibodies are polyclonal, they or in vitro by using the catalytic subunit of protein kinase A, recognized some distinct epitopes on ATF-43. and 32P-labeled ATF was purified by using DNA affinity We used the gel mobility shift assay to determine whether chromatography. Sequence-specific binding was adjudged ATF-43 and CREB are the same factor. Assays were per- by competition with oligonucleotides containing wild-type formed with either purified ATF-43 or in vitro-translated (Fig. 3, lanes 2, 4, and 8) or mutated (Fig. 3, lanes 1, 3, and CREB and an oligonucleotide probe containing an ATF- 7) ATF-binding sites. The in vitro kinase assay has been binding site from the adenovirus E4 promoter (Fig. 2A). As used previously to define ATF-43 and ATF-47 (34) and does described above, DNA-protein complexes containing not result in detection of API-related species, since these are ATF-43 were shifted to a slower mobility (Cl) or inhibited not substrates for protein kinase A (34). Both in vivo (Fig. 3, (C2) by prior incubation with these antibodies. Similarly, lanes 1 and 2) and in vitro (Fig. 3, lanes 7 and 8) labeling anti-CREB retarded the DNA-protein complex produced by resulted in detection of sequence-specific binding of ATF-43 in vitro-generated CREB. In contrast, neither Cl nor C2 and ATF-47 (Fig. 3). As mentioned previously (Fig. 1B) recognized the CREB-DNA protein complex and anti-CREB there is more than one polypeptide with mobility similar to did not recognize the ATF-43-DNA complex. Western blot that of ATF-43. In vivo labeling (Fig. 3, lanes 1 and 2) analysis using the antibodies described above (Fig. 2B) enabled detection of additional polypeptides of -65 to 70 further indicated that ATF-43 and CREB are distinct pro- kDa that might correspond to the previously described factor teins. In crude HeLa cell nuclear extracts, anti-ATF-43 (C3) EivF (15). Western blot analysis (Fig. 3, lanes 3 to 6) of DNA 6196 HURST ET AL. MOL. CELL. BIOL.

'-l < ~ ANTI-CREB 2 3: co Xj 2 3 - :-_-oo _'.CREB MrD_ kDa7 DNA-AFFINITY

* -ATF47 ]-ATF43

50- -50 miX

SDS

-39 -IEF FIG. 4. Two-dimensional gel analysis of CREB and ATF47. HeLa cell nuclear extract was 32P labeled with the catalytic subunit of protein kinase A and immunoprecipitated (ANTI-CREB) or affinity purified (DNA-AFFINITY) and analyzed in two dimensions by isoelectric focusing (IEF) in the first dimension and SDS-PAGE 1 2 3 4 5 6 7 8 (SDS) in the second according to the method of O'Farrell (73). A mixture of the immunoprecipitated and DNA affinity-purified sam- FIG. 3. Mobility of CREB and ATF-43 on SDS-polyacrylamide ples was run on another gel (MIX). gels. The apparent molecular weights of ATF-43 and CREB were compared by sequence-specific DNA affinity chromatography of in vitro and in vivo 32P-labeled proteins and Western blotting of the corresponding fraction with anti-CREB. Lanes: 1 and 2, in vivo- that anti-CREB did not directly recognize ATF-43, we asked labeled polypeptides bound to the DNA affinity resin in the presence if anti-CREB could precipitate ATF-43 in the absence of of 5 ,uM oligonucleotide containing a mutated (Ml) or wild-type CREB (Fig. 5A). In vitro 32P-labeled extracts were depleted (WT) ATF-binding site; 3 through 6, Western blot (with anti-CREB) of CREB by using anti-CREB followed by immunoprecipi- of polypeptides that bind to the DNA affinity resin in the presence of tation with anti-ATF-43. Several cycles of immunoprecipi- 5 p.M oligonucleotide containing an ATF (CREB)-binding site (WT and E2A [an oligonucleotide containing the ATF-binding site from the adenovirus E2A promoter]) or 5 p.M nonspecific oligonucleotide (Ml and 1B [an oligonucleotide containing sequences from the human ,B-globin pre-mRNA]); 7 and 8, in vitro-labeled polypeptides bound 1 2 3 4 5 6 1 2 3 4 5 to the DNA affinity resin in the presence of 5 p.M oligonucleotide containing an Ml or WT ATF (CREB)-binding site. Molecular size markers are shown at the left and right. _M_D affinity-purified fractions using anti-CREB identified a single polypeptide that comigrated with ATF-47. This polypeptide bound specifically to ATF-binding sites. Binding was inhib- _- CREB -a. 9- ited by competition with oligonucleotides containing ATF- -e binding sites (Fig. 3, lanes 4 and 6) and not by mutated (Fig. ] -ATF43 -[ 3, lane 3) or nonspecific (Fig. 3, lane 5) oligonucleotides. These results suggested that CREB corresponds to ATF- A B 47. To show this unequivocally, we analyzed CREB and ATF-47 by two-dimensional gel electrophoresis (Fig. 4). In FIG. 5. (A) Anti-CREB antibody did not directly recognize ATF- vitro-labeled HeLa cell ATF (including both ATF-43 and 43. HeLa cell nuclear extract was 32p labeled with protein kinase A ATF-47) was purified by using the DNA affinity column, and and immunoprecipitated with anti-CREB (lane 1). Two further labeled CREB was purified by immunoprecipitation with cycles of anti-CREB immunoprecipitation of the supernatant (lanes 2 and addition of A beads and a final round anti-CREB. In the region of 47 kDa, a number of isoelectric 3), protein only (lane 4), of immunoprecipitation (lane 5) ensured depletion of CREB and variants were detected in the pI range from 5.2 to 6.8. Mixing CREB-ATF-43 complexes. The supernatant from above was then of the two samples (anti-CREB immunoprecipitate and DNA immunoprecipitated with anti-ATF (C2 mouse antiserum; lane 6). affinity-purified material) demonstrated that all of the CREB (B) The protein(s) that coimmunoprecipitates with CREB is ATF- and ATF-47 species comigrate. We conclude from these 43. Approximately 30 ,ug of HeLa cell nuclear extract was 32p experiments that CREB and ATF-47 are identical proteins. labeled with protein kinase A, diluted to 400 p1l, and immunoprecip- In addition, anti-CREB precipitated a protein(s) that be- itated with anti-CREB. The immunoprecipitate was collected on haved similarly to ATF-43 on two-dimensional gels. The protein A beads and washed, and 1/4 of the sample was removed as significance of this result is addressed below. the no-treatment control (lane 1). The remaining beads were resus- The possibility that ATF-43 would coprecipitate with pended in 300 ,u1 of wash buffer and heated for 10 min at 50°C. The supernatant was divided in three and incubated with anti-ATF (C2 CREB was in two (Fig. 4) unexpected; independent assays mouse antiserum; lane 3), pre-immune serum (lane 4) or 40 p.l of (gel mobility shift assays [Fig. 2A] and Western blotting [Fig. DNA affinity resin (lane 5). The DNA affinity resin was washed with 2B and 3]), anti-CREB did not recognize ATF-43. This 0.3 M KCI dialysis buffer and then suspended in SDS-PAGE sample suggested that coprecipitation of CREB and ATF-43 is buffer prior to electrophoresis. Lane 2 contained 1/3 of the heat- because of CREB-ATF-43 complexes. To show rigorously treated beads. WITH ATFs VOL. 10, 1990 CREB CORRESPONDS TO AND FORMS COMPLEXES 6197 tation with anti-CREB (Fig. SA, lanes 1 to 5) resulted in H P F depletion of CREB and putative CREB-ATF-43 complexes. H P F eo%N0mo%?WN F P P The first three cycles resulted in efficient depletion (Fig. 5A, lanes 1 to 3), with no detectable precipitation occurring in the last two cycles (Fig. SA, lanes 4 and 5). Following CREBEB- depletion of extracts by using anti-CREB, anti-ATF-43 was a ATF43[ able to precipitate the substantial amount of ATF-43 that [3 a -0-40I remained in the supernatant (Fig. 5A, lane 6). These results I m clearly demonstrate that anti-CREB does not directly recognize ATF-43. To verify that ATF-43 was the protein immunoprecipitated by anti-CREB, we disrupted the putative complex and B A c probed for the release of ATF-43 (Fig. 5B). Purified immune aCREB complexes (Fig. 5B, lane 1) were heated to 50°C (Fig. 5B, Ab: caCREB a ATF lane 2) to disrupt the putative CREB-ATF-43 complex. This resulted in partial release of ATF-43 (compare Fig. 5B, lanes e2 1 and 2). Confirmation that the polypeptides released were in 0 fact ATF-43 was provided by immunoprecipitation of the t released polypeptides by anti-ATF (Fig. 5B, lane 3) (and not by preimmune serum [Fig. 5B, lane 4]) and DNA affinity purification of the released polypeptides (Fig. 5B, lane 5). Taken together, the above results demonstrate that the polypeptide(s) coprecipitating with CREB is ATF-43 and that CREB and ATF-43 can associate in a protein-protein CREB-_-Uc a0 p complex. W foomosla,. Multiple ATF-43-related species (ATF-43*) were present in ATF43[ s different cell types. cDNA cloning and purification studies have led to the identification of multiple factors that bind directly to ATF-binding sites. However, as evidenced by the above results, sequence-specific DNA affinity purification from different cell types can result in the isolation of different Ab: C C P A C P A C C C proteins (34, 40, 70). We therefore decided to examine the FIG. 6. Detection of cell-type-specific ATF-43-related proteins. cell type distribution of ATF-43 and CREB (Fig. 6). Nuclear extracts were prepared from monolayers of HeLa (H), Initially we tested PC12 cells which apparently lack PC12 (P), and F9 embryonal carcinoma cells (F). Equal amounts of ATF-43 (CREB is the only polypeptide purified from PC12 nuclear protein from each cell line were 32p labeled with protein cells [69], and we show here that CREB is not ATF-43) and kinase A, immunoprecipitated as described in Materials and Meth- F9 embryonal carcinoma cells, which appear deficient in ods, and analyzed by SDS-PAGE. (A) Immunoprecipitation with CREB, as determined by immunofluorescence staining (32, anti-ATF (C2 antibody). For each cell line, immunoprecipitations Using anti-ATF-43, a different pattern of polypeptides were performed with C2 antibody (right-hand lanes) or preimmune 33). serum (left-hand lanes). (B) lmmunoprecipitation with anti-CREB. ranging in size from -40 to 43 kDa was immunoprecipitated (C) Effects of mixing F9 cells and PC12 cells prior to extract from HeLa, PC12, and F9 cells (Fig. 6A). We note that the preparation. Immunoprecipitations using anti-CREB were per- anti-ATF-43 did not recognize CREB-ATF-43 complexes. formed on nuclear extracts prepared from F9 cells or PC12 cells This apparent paradox is addressed in the Discussion. Anti- alone or from a mixture (F+P) of F9 and PC12 cells mixed at the CREB precipitated CREB from all three cell lines and in time of cell harvest. (D) Detection of CREB-ATF-43 complexes in a addition precipitated a profile of polypeptides that was variety of different cells. Nuclear proteins were prepared and similar or identical to that precipitated by anti-ATF-43 (Fig. labeled in vitro with protein kinase A as previously described. The 6B). The pattern of ATF-43-related species might be gener- cell lines (described in Materials and Methods) examined and the antibodies used are indicated above and below the figure, respec- ated from a common polypeptide in a cell-type-specific anti-ATF (C2); preimmune. manner during extract preparation. To test this possibility, tively. C, Anti-CREB; A, P. we made extracts from mixtures (F+P) of PC12 (P) and F9 (F) cells (Fig. 6C). Migration of the higher-molecular-weight F9-specific polypeptide(s) was not modified in the presence in HeLa, F9, and PC12 cells, we compared the profile of of PC12 cells, and similarly, the lower-molecular-weight polypeptides selected by DNA binding with that immuno- PC12-specific polypeptide(s) was not modified by F9 cells. precipitated by anti-CREB (Fig. 7). A very similar (or The same result was obtained for mixtures of PC12 cells and identical) profile was obtained for immunoprecipitation by HeLa cells (data not shown). Thus, the different ATF-43- anti-CREB (Fig. 7, lanes C) compared with DNA affinity related species observed were not produced during extract purification (Fig. 7, lanes DNA). This demonstrates that the preparation and were therefore cell type specific. These different species that make up ATF-43* were all able to bind results identify a group of cell-type-specific polypeptides to DNA. (ATF-43 ) present in nuclear extracts that are antigenically In vitro labeling of ATF with protein kinase A allowed related to ATF-43. Screening of additional cell lines (Fig. detection only of polypeptides that are not phosphorylated 6D) did not reveal any further ATF-43-related polypeptides on their protein kinase A phosphoacceptor site. The cell type and indicated that CREB-ATF-43 complexes are present in a differences described above might therefore reflect differen- wide variety of mammalian cells, including 293, JEG-3, tial phosphorylation in vivo of the same polypeptides in each MEL-28, and CHO. cell. To address this issue we performed immunoprecipita- To examine the DNA-binding activity of ATF-43* present tion assays following metabolic labeling of cells with 6198 HURST ET AL. MOL. CELL. BIOL.

DNA DNA A M1WT C M1 WT C Ml WT C DNA AF FINITY SOUTHWESTERN

75 - 69- CREB_-a- 410 4=10 I ATF43 U alo 50 - 46-

-U,o

3 9 - 30- I~~~~~~~~~~~~~,I

HeLa F9 PC12 2 7- FIG. 7. Comparison of ATF-43* by DNA binding and immunoprecipitation. In vitro 32P-labeled proteins from HeLa, F9, and PC12 cells were DNA affinity purified (DNA) in the presence of wild-type CIP: + -+ (WT) and mutant (Ml) oligonucleotides or immunoprecipitated with anti-CREB (C). FIG. 9. Effects of phosphatase treatment on ATF-43. Analysis of DNA affinity-purified ATF-43 (DNA AFFINITY). Heat-killed HeLa cell nuclear extract (30 ,ul) was incubated at 37°C for 45 min in the [35S]methionine (Fig. 8). The pattern of 35S-labeled polypep- presence or absence of 25 U of calf intestinal phosphatase (CIP; tides immunoprecipitated with anti-CREB (Fig. 8, lanes C) Boehringer), affinity purified, washed extensively, 32p labeled with closely resembles the pattern obtained for 32p in vitro- protein kinase A, and electrophoresed. For Southwestern blot labeled polypeptides (Fig. 6B), and the same result was analysis (SOUTHWESTERN), purified ATF-43 was treated as obtained with anti-ATF-43 antibody (data not shown). An described above, resolved on an SDS gel, transferred to nitrocellulose paper, and probed with a 32P-labeled oligonucleotide containing 35S-labeled polypeptide of -55 kDa was immuno- additional an ATF-CREB-binding site, as described in Materials and Methods. precipitated by anti-CREB. The significance of this polypeptide is not clear. We conclude that detection of ATF-43* by using the in vitro kinase assay does not simply reflect Differential phosphorylation produced alternative forms of differential cell-type-specific phosphorylation of the protein ATF-43. Many sequence-specific DNA-binding proteins are kinase A phosphoacceptor site in these polypeptides. The phosphorylated (5, 21, 53, 89, 93), and phosphorylation has observed differences in migration might therefore result from been suggested to play a role in the activities of factors that additional covalent modification (for example. because of bind to ATF-binding sites (1, 32, 33, 48, 68, 76, 78, 100). In other kinases; see below) of the same protein or from the light of this, we examined the effects of dephosphorylation of presence of distinct cell-type-specific proteins. HeLa cell ATF-43 (Fig. 9). Treatment of crude nuclear extracts (prepared from HeLa cells grown in suspension) with phosphatase, followed by DNA affinity purification and 32p labeling with protein kinase A, resulted in loss of 3 IV detectable ATF-43 and an increase in a faster-migrating __ species (Fig. 9). Similar results were obtained by using a Southwestern blot analysis (Fig. 9). Dephosphorylation of purified HeLa cell ATF-43 reduced levels of detectable ATF-43 (filters were probed with a 32P-labeled oligonucleotide probe containing an ATF-binding site) and resulted in detection of a faster-migrating polypeptide. A probe containing a simian virus 40 APl-binding site (40) failed to bind to CREB--- purified ATF, demonstrating the specificity of the South- western assay (data not shown). ATF43 These results suggest that dephosphorylation alters the mobility of ATF-43 without reducing DNA-binding activity. Because phosphatase treatment of ATF-43 produced a polypeptide with a mobility similar to that of the fastest-migrating polypeptide from HeLa cells observed in other assays (Fig. 3 and 5), it is possible that HeLa cell ATF-43* arises (at least Ab: P C P C P C partly) from differential phosphorylation of the same protein. Merino et al. (68) have reported that dephosphorylation of of CREB-ATF-43 FIG. 8. Detection [3S]methionine-labeled HeLa cell polypeptides in the size range of ATF-43 reduces complexes. Nuclear extracts were prepared from [355]methionine- labeled HeLa, PC12, and F9 embryonal carcinoma cells. Equal DNA-binding activity and does not affect migration on SDS amounts of nuclear protein from each cell line were immunoprecip- gels. These contrasting effects suggest that the factors de- itated with anti-CREB (C) or preimmune (P) antiserum and analyzed scribed by Merino et al. are distinct from ATF-43. by SDS-PAGE. Stable DNA binding of ATF-43 depends on flanking se- VOL. 10, 1990 CREB CORRESPONDS TO AND FORMS COMPLEXES WITH ATFs 6199

ATF-43 (43 kDa) and ATF-47 (47 kDa). The sequence- comp: m2 Wt M2 Wt specific DNA-binding activity and size of CREB (33, 39, 70) indicated that it was a member of the ATF family, although the relationship between CREB and ATF has not been established. We show here that CREB corresponds to ATF- CREB- 47. Moreover, the antibodies that we have used enabled us ATF 1 to distinguish CREB and ATF-43. Three anti-ATF sera that recognized at least two epitopes on ATF-43 did not cross-react with CREB, and the C-terminal epitope recognized by anti-CREB was not present in ATF-43. The ability to distinguish CREB and ATF-43 is in contrast to the previous observation that CREB (ATF-47) column: BSE2 BSE4 and ATF-43 are highly related (34). Their common properties include reactivity with a range of different antibodies (34), heat stability, DNA-binding specificity, presence of BSE4: cccgggatga CGTCA tcccgggatga CGTCA tcccggg phosphoacceptor sites for protein kinases A and C, and BSE2: cta CGTCA tctga TGACG tatttcagatc promoter-specific transcriptional activity in vitro. Our results demonstrate that CREB and ATF-43 have several FIG. 10. Stability of DNA binding by CREB and ATF-43. Heat- structural differences and therefore predict different func- denatured HeLa cell nuclear extract was 32P labeled with protein tional roles for these two proteins. kinase A, and proteins were allowed to bind to two different DNA We have identified a subgroup of factors (ATF-43*) within affinity resins (BSE2 and BSE4) in the presence of 4 jig of poly(dI- the ATF family that is highly related to ATF-43. These dC) per ml. The DNA sequences of concatemerized oligonucleo- are tides for the BSE2 and BSE4 affinity resins are indicated at the factors of similar size and might therefore correspond to bottom. The BSE2 sequence is derived from the ATF site present in previously identified factors that interact with CREs (4, 18, the adenovirus E2A promoter, and the BSE4 site is a consensus 44, 75) or the Tax-responsive element of the HTLV-1 long ATF site. DNA-protein complex stability was assessed by compe- terminal repeat (92). The relationship between ATF-43* and tition with 5 ,uM wild-type (wt) oligonucleotide (specific competitor) previously described ATF cDNA clones is not entirely clear, or 5 i1M M2 oligonucleotide (nonspecific competitor) for 30 min. although certain possibilities appear to have been ruled out. Following removal of supematant, the DNA affinity resins contain- We have shown that ATF-43 did not correspond to CREB. ing stably bound proteins were suspended in SDS-PAGE sample In addition ATF-43 is most unlikely to correspond to CRE- buffer and electrophoresed. BP1 or ATF-2 (because of a different size [65]), ATF-3 (ATF-3 does not form heterodimers with ATF-1 [35] and would therefore not form heterodimers quences or spacing between ATF sites. It has been suggested with CREB), or that the ability of ATF-binding sites to mediate different ATF-6 (which only binds to multimerized ATF-binding sites [35]). The observation transcriptional responses is influenced by sequences that that multiple proteins are related to immediately flank the CGTCA motif (19, 55, 72). This might ATF-43 but are not encoded by previously isolated cDNA be related to variations in the precise interaction of distinct clones indicates that the size of the ATF family at the protein ATF-related proteins with DNA (35) or due to interaction of level is even larger than that predicted from cDNA-cloning additional proteins with sequences that overlap ATF-binding studies (35). sites (15, 79, 98). To probe for differences in the Complex formation between CREB and ATF-43. The anti- DNA- bodies described protein complexes formed by CREB and ATF-43, we have above have enabled us to demonstrate that examined stable complex formation between these proteins CREB forms protein-protein complexes with ATF-43. Lack and two different immobilized ATF-binding sites (Fig. 10). of an in vitro assay for complex formation precludes a The two sequence-specific DNA affinity resins employed demonstration of direct interactions between CREB and or contained either a consensus ATF-binding site (BSE4) or the ATF-43 mapping of determinants involved in complex ATF-binding site from the E2A promoter (BSE2) and dif- formation. A cDNA clone that encodes ATF-43 will be fered in sequences flanking the CGTCA motif and in the required to analyze complex formation in detail. It is possi- spacing between ATF sites (Fig. 10). In vitro 32P-labeled ble that CREB and ATF-43 interact indirectly or that their CREB and ATF-43 were allowed to bind to BSE4 interaction is not mediated through leucine zippers. How- and BSE2 ever, followed by competition with a mutant (Fig. 10, lanes mt) or several considerations suggest that CREB and ATF-43 wild-type (Fig. 10, lanes wt) oligonucleotide. CREB formed most probably form heterodimers, with direct protein con- a stable complex with both BSE4 and BSE2, as indicated by tacts mediated through the CREB leucine zipper and the lack of competition by the wild-type oligonucleotide, putative leucine zipper of ATF-43. First, isolation of several whereas ATF-43 formed a stable complex with BSE4 but not cDNA clones encoding factors that directly interact with with BSE2. These results demonstrate that sequences flank- ATF-binding sites reveals that these factors belong to the ing the CGTCA core motif or the spacing between ATF sites bZIP class of transcriptional activators (35, 96). Second, can selectively affect stable complex formation between CREB binds independently to ATF-binding sites as a ho- ATF-43 and DNA. modimer (100), and ATF-43 can also bind independently (40), strongly suggesting that CREB-ATF heterodimers would probably bind to DNA in the same manner as CREB DISCUSSION homodimers. Relationship between CREB and ATF. ATF is a multigene Our results demonstrate that protein-protein interactions family of cellular transcription factors related by sequence- between members of the ATF family can be detected in cell specific DNA-binding activity (33, 35, 39, 50, 65). At the extracts, thus identifying specific complexes that almost protein level, ATF binding and transcriptional activities certainly occur in vivo. This extends previous results (35) from HeLa cells copurify with two polypeptides designated that have demonstrated heterodimer formation between 6200 HURST ET AL. MOL. CELL. BIOL. members of the ATF family in vitro. The specificities of the types examined (Fig. 6; data not shown). Second, a subset of antibodies that we have employed enable identification of ATF-43 probably arises by differential phosphorylation of a CREB-ATF-43* complexes and are therefore useful re- common polypeptide (ATF-43) in HeLa cells by a kinase(s) agents for further characterization of these complexes in other than protein kinase A. vitro and in vivo. In light of the recent demonstration that CREB is required Functional role of CREB and ATF-43 complexes. It is likely for cAMP-inducible transcription in F9 cells (32), the role of that the variety of transcriptional responses that are medi- ATF-43 in these cells is of particular significance. The ability ated by ATF-binding sites requires different combinations of of CREB to independently bind to DNA in vitro (100) CREB and ATF-43. Our results demonstrate that ATF-43 together with its ability to mediate a cAMP response in F9 has distinct homodimeric and heterodimeric conformations, cells (32), has led to the suggestion that CREB homodimers as indicated by lack of recognition of CREB-ATF-43 com- are responsible for cAMP-inducible transcription. Our re- plexes by three antisera that recognize at least two determi- sults raise the possibility that CREB-ATF complexes might nants on ATF-43. Thus, ATF-43 could differentially interact be the active components in the cAMP response. Alterna- with other parts of the transcriptional machinery, dependent tively, ATF-43 might negatively regulate CREB in F9 cells on heterodimerization with CREB. By analogy with oc- (which do not respond to cAMP [32, 81]) or expand the tamer-binding proteins, the ability of ATF-binding sites to activating potential of CREB for promoters that are not confer multiple transcriptional responses could occur, in regulated by cAMP. part, through the selective interaction of CREB and ATF-43 with other proteins that do not bind to DNA independently. ACKNOWLEDGMENT The study of viral systems suggests that the adenovirus Ela protein and the HTLV-1 Tax protein are likely candidates. We thank Jeff Williams for critical reading of the manuscript. In the case of Ela, it has been suggested that ATF might function by recruiting Ela into transcription complexes (61), LITERATURE CITED raising the possibility that Ela could discriminate between 1. Andrisani, O., and J. E. Dixon. 1990. Identification and purifi- different CREB-ATF-43 complexes. Similar interactions cation of a novel 120-kDa protein that recognizes the cAMP- responsive element. J. Biol. Chem. 265:3212-3218. might occur between CREB-ATF and the HTLV-1 Tax 2. Andrisani, 0. M., T. E. Hayes, B. Roos, and J. E. Dixon. 1987. protein that has been shown to bind indirectly to ATF- Identification of the promoter sequences involved in the cell binding sites found in Tax response elements (67). specific expression of the rat somatostatin gene. Nucleic Acids The potential of ATF-binding sites to mediate transcrip- Res. 15:5715-5728. tional responses to ElA, cAMP, or Tax appears to be 3. Andrisani, 0. M., D. A. Pot, Z. Zhu, and J. E. Dixon. 1988. influenced by cis-acting determinants other than the CGTCA Three sequence-specific DNA-protein complexes are formed core motif (19, 25, 49, 55, 72). Sequences immediately with the same promoter element essential for expression of the flanking the core CGTCA motif and additional promoter rat somatostatin gene. Mol. Cell. Biol. 8:1947-1956. elements can both alter the activating potential of particular 4. Andrisani, 0. M., Z. Zhu, D. A. Pot, and J. E. Dixon. 1989. In vitro transcription directed from the somatostatin promoter is ATF sites. We have provided evidence that CREB and dependent upon a purified 43-kDa DNA-binding protein. Proc. ATF-43 form complexes with DNAs of different stabilities in Natl. Acad. Sci. USA 86:2181-2185. a manner which is determined either by a flanking sequence 5. Bagchi, S., P. Raychaudhuri, and J. R. Nevins. 1989. Phosphor- or by spacing between multiple ATF sites. Our data are ylation-dependent activation of the adenovirus-inducible E2F insufficient to define determinants of stable complex forma- transcription factor in a cell free system. Proc. Natl. Acad. Sci. tion between ATF-43 and DNA. However, our results USA 86:4352-4356. suggest that differential stability of DNA binding is one 6. Benbrook, D. M., and N. C. Jones. 1990. Heterodimer forma- mechanism that can regulate activities of members of the tion between CREB and JUN proteins. Oncogene 5:295-302. ATF family and contribute to the differential 7. Berk, A. J. 1986. Adenovirus promoters and ElA transactiva- activities of tion. Annu. Rev. Genet. 20:45-79. ATF-binding sites. 8. Berk, A. J., and M. C. Schmidt. 1990. How do transcription Cell type distribution of ATF-43*. We have identified at factors work? Genes Dev. 4:151-155. least four polypeptides that share several properties with 9. Boeuf, H., D. A. Zajchowski, T. Tamura, C. Hauss, and C. ATF-43 but differ in cell type distribution. Given the number Kedinger. 1987. Specific cellular proteins bind to critical pro- of possible dimers that might exist between members of the moter sequences of the adenovirus early EIla promoter. Nu- ATF-CREB and Jun-Fos families, changes in abundance of a cleic Acids Res. 15:509-527. single factor could dramatically change the cellular comple- 10. Brady, J., K.-T. Jeang, J. Duvall, and G. Khoury. 1987. ment of functional dimers. In this regard it is of interest that Identification of p40x-responsive regulatory sequences within the ratios of free (homodimeric?) the human T-cell leukemia virus type 1 long terminal repeat. J. ATF-43 to ATF-43-CREB Virol. 61:2175-2181. complexes varied dramatically among different cell types 11. Busch, S. J., and P. Sassone-Corsi. 1990. Dimers, leucine (Fig. 6), as does the activity of several cAMP response zippers and DNA-binding domains. Trends Genet. 6:36-40. elements (2, 16, 17, 38, 49, 80, 90). 12. Chiu, R., W. J. Boyle, J. Meek, T. Smeal, T. Hunter, and M. We have not established any interrelationships among Karin. 1988. The c-Fos protein interacts with c-Jun/AP-1 to members of ATF-43* present in HeLa, F9, and PC12 cells. stimulate transcription of AP-1 responsive genes. Cell 54:541- Differential splicing, expression of different genes, or post- 552. translational modifications could all be contributary factors. 13. Cohen, D. R., C. P. Ferreira, R. Gentz, B. R. Franza, Jr., and Production of highly related protein products by differential T. Curran. 1989. The product of afos-related gene,fra-1, binds splicing has been described in the case of CREB cooperatively to the AP-1 site with Jun: transcription factor (101). Our AP-1 is comprised of multiple protein complexes. Genes Dev. data are consistent with the possibility that posttranslational 3:173-184. modifications might contribute to generation of ATF-43*. 14. Comb, M., S. E. Hyman, and H. M. Goodman. 1987. Mecha- First, the cell type distribution of ATF-43* is strikingly nisms of trans-synaptic regulation of gene expression. Trends different but is not absolute, as indicated by detection of Neurosci. 10:473-478. small amounts of most members of ATF-43 in most cell 15. Cortes, P., L. Buckbinder, M. A. Leza, N. Rak, P. Hearing, A. VOL. 10, 1990 CREB CORRESPONDS TO AND FORMS COMPLEXES WITH ATFs 6201

Merino, and D. Reinberg. 1988. EIVF, a factor required for 35. Hai, T.-Y., F. Liu., W. J. Coukos, and M. R. Green. 1989. transcription of the adenovirus E4 promoter binds to an Transcription factor ATF cDNA clones: an extensive family of element involved in ElA-dependent activation and cAMP leucine zipper proteins able to selectively form DNA binding induction. Genes Dev. 2:975-990. heterodimers. Genes Dev. 3:2083-2090. 16. Dean, D. C., M. S. Blakeley, R. F. Newby, P. Ghazal, L. 36. Halazonetis, T. D., K. Georgopoulos, M. E. Greenberg, and P. Hennighausen, and S. Bourgeois. 1989. Forskolin inducibility Leder. 1988. c-Jun dimerizes with itself and with c-Fos, form- and tissue-specific expression of the fibronectin promoter. ing complexes of different DNA binding affinities. Cell 55:917- Mol. Cell. Biol. 9:1498-1506. 924. 17. Delegeane, A. M., L. H. Ferland, and P. L. Mellon. 1987. 37. Hardy, S., and T. Shenk. 1988. Adenoviral control regions Tissue-specific enhancer of the human glycoprotein hormone activated by ElA and the cAMP-response element bind to the subunit gene: dependence on cyclic AMP-inducible elements. same factor. Proc. Natl. Acad. Sci. USA 85:4171-4175. Mol. Cell. Biol. 7:3994-4002. 38. Harrington, C. A., E. J. Lewis, D. Krzemien, and D. M. 18. Deutsch, P. J., J. P. Hoeffler, J. L. Jameson, and J. F. Habener. Chikaraishi. 1987. Identification and cell type specificity of the 1988. Cyclic AMP and phorbol ester-stimulated transcription tyrosine hydroxylase gene promoter. Nucleic Acids Res. 15: mediated by similar DNA elements that bind distinct proteins. 2363-2384. Proc. Natl. Acad. Sci. USA 85:7922-7926. 39. Hoeffler, J. P., T. E. Meyer, Y. Yun, J. L. Jameson, and J. F. 19. Deutsch, P. J., J. P. Hoeffler, L. J. Jameson, J. C. Lin, and Habener. 1988. Cyclic AMP-responsive DNA-binding protein: J. F. Habener. 1988. Structural determinants for transcrip- structure based on a cloned placental cDNA. Science 242: tional activation by cAMP-responsive DNA elements. J. Biol. 1430-1433. Chem. 263:18466-18472. 40. Hurst, H. C., and N. C. Jones. 1987. Identification of factors 20. Dorn, A., J. Bollekens, A. Staub, C. Benoist, and D. Mathis. that interact with the ElA-inducible adenovirus E3 promoter. 1987. A multiplicity of CCAAT box-binding proteins. Cell Genes Dev. 1:1132-1146. 50:863-872. 41. Hyman, S. E., M. Comb, Y.-S. Lin, J. Pearlberg, M. R. Green, 21. Driever, W., and C. Nusslein-Volhard. 1989. The bicoid protein and H. M. Goodman. 1988. A common trans-acting factor is is a positive regulator of hunchback transcription in the early involved in transcriptional regulation of neurotransmitter Drosophila embryo. Nature (London) 337:138-143. genes by cyclic AMP. Mol. Cell. Biol. 8:4225-4233. 22. Dwarki, V. J., M. Montminy, and I. M. Verma. Both the basic 42. Ivashkiv, L. B., H.-C. Liou, C. J. Kara, W. W. Lamph, I. M. region and the "leucine zipper" domain of the cyclic AMP- Verma, and L. H. Glimcher. 1990. mXBP/CRE-BP2 and c-Jun responsive element binding (CREB) protein are essential for form a complex which binds to the cyclic AMP, but not to the transcriptional activation. EMBO J. 9:225-232. 12-O-tetradecanoylphorbol-13-acetate, response element. Mol. 23. Dynan, W. S. 1989. Modularity in promoters and enhancers. Cell. Biol. 10:1609-1621. Cell 58:1-4. 43. Jalinot, P., B. Devaux, and C. Kedinger. 1987. The abundance 24. Dynan, W. S., and R. Tjian. 1985. Control of eukaryotic RNA and in vitro DNA binding of three cellular proteins interacting synthesis by sequence-specific DNA-binding proteins. Nature with the adenovirus Ella early promoter are not modified by (London) 316:774-778. the EIa gene products. Mol. Cell. Biol. 7:3806-3817. 25. Evans, M. J., and R. C. Scarpulla. 1989. Interaction of nuclear 44. Jalinot, P., M. Wintzerith, M. Gaire, C. Hauss, J. M. Egly, and factors with multiple sites in the somatic cytochrome c pro- C. Kedinger. 1988. Purification and functional characterization moter. J. Biol. Chem. 264:14361-14368. of a cellular transcription factor that binds to an enhancer 26. Fink, J. S., M. Verhave, S. Kasper, T. Tsukada, G. Mandel, element within the adenovirus early EllA promoter. Proc. and R. H. Goodman. 1988. The CGTCA sequence motif is Natl. Acad. Sci. USA 85:2484-2488. essential for biological activity of the vasoactive intestinal 45. Jeang, K.-T., I. Boros, J. Brady, M. Radonovich, and G. peptide gene cAMP-regulated enhancer. Proc. Natl. Acad. Sci. Khoury. 1988. Characterization of cellular factors that interact USA 85:6662-6666. with the human T-cell leukemia virus type I p4ox-responsive 27. Fletcher, C., N. Heintz, and R. G. Roeder. 1987. Purification 21-base-pair sequence. J. Virol. 62:4499-4509. and characterization of OTF-1, a transcription factor regulating 46. Jeang, K.-T., S. G. Widen, 0. J. Semmes IV, and S. H. Wilson. cell cycle expression of a human histone H2b gene. Cell 1990. HTLV-1 trans-activator, Tax, is a trans-repressor of the 51:773-781. human,B-polymerase gene. Science 247:1082-1084. 28. Fujisawa, J.-I., M. Toita, and M. Yoshida. 1989. A unique 47. Jones, N. C., P. W. J. Rigby, and E. B. Ziff. 1988. Trans-acting enhancer element for the trans activator (p40tax) of human T- protein factors and the regulation of eukaryotic transcription: cell leukemia virus type 1 that is distinct from cyclic AMP and lessons from studies on DNA tumor viruses. Genes Dev. 12-O-tetradecanoylphorbol-13-acetate-responsive elements. J. 2:267-281. Virol. 63:3234-3239. 48. Jones, R. H., and N. C. Jones. 1989. Mammalian cAMP- 29. Gentz, R., F. J. Rauscher III, C. Abate, and T. Curran. 1989. responsive element can activate transcription in yeast and Parallel association of Fos and Jun leucine zippers juxtaposes binds a yeast factor(s) that resembles the mammalian transcrip- DNA binding domains. Science 243:1695-1699. tion factor ATF. Proc. Natl. Acad. Sci. USA 86:2176-2180. 30. Giam, C.-Z., and Y.-L. Xu. 1989. HTLV-1 tax gene product 49. Kanei-Ishii, C., and S. Ishii. 1989. Dual enhancer activities of activates transcription via pre-existing cellular factors and the cyclic AMP-responsive element with cell type and pro- cAMP-responsive element. J. Biol. Chem. 264:15236-15241. moter specificity. Nucleic Acids Res. 17:1521-1537. 31. Gilardi, P., and M. Perricaudet. 1986. The E4 promoter of 50. Kara, C. J., H.-C. Liou, L. B. Ivashkiv, and L. H. Glimcher. adenovirus type 2 contains an ElA-dependent cis-acting ele- 1990. A cDNA for a human cyclic AMP response element- ment. Nucleic Acids Res. 14:9035-9049. binding protein which is distinct from CREB and expressed 32. Gonzalez, G. A., and M. R. Montminy. 1989. Cyclic AMP preferentially in brain. Mol. Cell. Biol. 10:1347-1357. stimulates somatostatin gene transcription by phosphorylation 51. Katoh, I., Y. Yoshinaka, and Y. Ikawa. 1989. Bovine leukemia of CREB at serine 133. Cell 59:675-680. virus trans-activator p38tax activates heterologous promoters 33. Gonzalez, G. A., K. K. Yamamoto, W. H. Fischer, D. Karr, P. with a common sequence known as the cAMP-responsive Manzel, W. Biggs III, W. W. Vale, and M. R. Montminy. 1989. element or the binding site of a cellular transcription factor A cluster of phosphorylation sites on the cAMP-regulated ATF. EMBO J. 8:497-503. nuclear factor CREB predicted by its sequence. Nature (Lon- 52. Kouzarides, T., and E. Ziff. 1988. The role of the leucine zipper don) 337:749-752. in the fos-jun interaction. Nature (London) 336:646-651. 34. Hai, T., F. Liu, E. A. Allegretto, M. Karin, and M. R. Green. 53. Krause, H. M., R. Klemenz, and W. J. Gehring. 1988. Expres- 1988. A family of immunologically related transcription factors sion, modification, and localisation of the fushi tarazo protein that includes multiple forms of ATF and AP-1. Genes Dev. in Drosophila embryos. Genes Dev. 2:1021-1036. 2:1216-1226. 54. Lee, K. A. W., A. Bindereif, and M. R. Green. 1988. A 6202 HURST ET AL. MOL. CELL. BIOL.

small-scale procedure for preparation of nuclear extracts that glutaryl coenzyme A reductase gene. Proc. Natl. Acad. Sci. support efficient transcription and pre-mRNA splicing. Gene USA 84:3614-3618. Anal. Tech. 5:22-31. 75. Philippe, J., and M. Missotten. 1990. Functional characteriza- 55. Lee, K. A. W., J. S. Fink, R. H. Goodman, and M. R. Green. tion of a cAMP-responsive element of the rat insulin 1 gene. J. 1989. Distinguishable promoter elements are involved in tran- Biol. Chem. 265:1465-1469. scriptional activation by Ela and cyclic AMP. Mol. Cell. Biol. 76. Poteat, H. T., F. Y. Chen, P. Kadison, J. G. Sodroski, and 9:4390-4397. W. A. Haseltine. 1990. Protein kinase A-dependent binding of a 56. Lee, K. A. W., and M. R. Green. 1987. A cellular transcription nuclear factor to the 21-base-pair repeat of the human T-cell factor E4F1 interacts with an ElA-inducible enhancer and leukemia virus type 1 long terminal repeat. J. Virol. 64:1264- mediates constitutive enhancer function in vitro. EMBO J. 1270. 6:1345-1353. 77. Ptashne, M. 1986. Gene regulation by proteins acting nearby 57. Lee, X. A. W., T. Y. Hai, L. SivaRaman, B. Thimmappaya, and at a distance. Nature (London) 322:697-701. H. C. Hurst, N. C. Jones, and M. R. Green. 1987. A cellular 78. Raychaudhuri, P., S. Bagchi, and J. R. Nevins. 1989. DNA- protein activating transcription factor activates transcription of binding activity of the adenovirus-induced E4F transcription multiple ElA-inducible adenovirus early promoters. Proc. factor is regulated by phosphorylation. Genes Dev. 3:620-627. Natl. Acad. Sci. USA 84:8355-8359. 79. Raychaudhuri, P., R. Rooney, and J. R. Nevins. 1987. Identi- 58. Levine, M., and J. L. Manley. 1989. Transcriptional repression fication of an ElA-inducible cellular factor that interacts with of eukaryotic promoters. Cell 59:405-408. the regulatory sequences within the adenovirus E4 promoter. 59. Lewis, E. J., C. A. Harrington, and D. M. Chikaraishi. 1987. EMBO J. 6:4073-4081. Transcriptional regulation of the tyrosine hydroxylase gene by 80. Rice, D. A., L. D. Aitken, G. R. Vandenbark, A. R. Mouw, A. glucocorticoid and cyclic AMP. Proc. Natl. Acad. Sci. USA Franklin, B. P. Schimmer, and K. L. Parker. 1989. A cAMP- 84:3550-3554. responsive element regulates expression of the mouse steroid 60. Leza, M. A., and P. Hearing. 1988. Cellular transcription factor 110-hydroxylase gene. J. Biol. Chem. 264:14011-14015. binds to adenovirus early region promoters and to a cyclic 81. Rickles, R. J., A. L. Darrow, and S. Strickland. 1989. Differ- AMP response element. J. Virol. 62:3003-3013. entiation-responsive elements in the 5' region of the mouse 61. Lillie, J. W., and M. R. Green. 1989. Transcription activation tissue plasminogen activator gene confer two-stage regulation by the adenovirus Ela protein. Nature (London) 338:39-44. by retinoic acid and cyclic AMP in teratocarcinoma cells. Mol. 62. Lin, Y. S., and M. R. Green. 1988. Interaction of a common Cell. Biol. 9:1691-1704. transcription factor, ATF, with regulatory elements in both 82. Sassone-Corsi, P. 1988. Cyclic-AMP induction of early adeno- Ela- and cyclic AMP-inducible promoters. Proc. Natl. Acad. virus promoters involves sequences required for ElA trans- Sci. USA 85:3396-3400. activation. Proc. Natl. Acad. Sci. USA 85:7192-71%. 63. Lin, Y. S., and M. R. Green. 1989. Identification and purifica- 83. Sassone-Corsi, P., L. J. Ransone, W. W. Lamph, and I. M. tion of a Saccharomyces cerevisiae protein with the DNA- Verma. 1988. Direct interaction between fos and jun nuclear binding specificity of mammalian activating transcription fac- oncoproteins: role of the 'leucine zipper' domain. Nature tor. Proc. Natl. Acad. Sci. USA 86:109-113. (London) 336:692-698. 64. Macgregor, P. F., C. Abate, and T. Curran. 1990. Direct 84. Sassone-Corsi, P., J. Visvader, L. Ferland, P. L. Mellon, and cloning of leucine zipper proteins. Jun binds cooperatively to I. M. Verma. 1988. Induction of proto-oncogene fos transcrip- the CRE with CRE-BP1. Oncogene 5:451-458. tion through the adenylate cyclase pathway: characterization 65. Maekawa, T., H. Sakura, C. Kanei-Ishii, T. Sudo, T. Yoshi- of a cAMP-responsive element. Genes Dev. 2:1529-1538. mura, J. Fujisawa, M. Yoshida, and S. Ishii. 1989. Leucine 85. Scheidereit, C., A. Heguy, and R. G. Roeder. 1987. Identifica- zipper structure of the protein CRE-BP1 binding to the cyclic tion and purification of a human lymphoid specific octamer- AMP response element in brain. EMBO J. 8:2023-2028. binding protein (OTF-2) that activates transcription of an 66. Maniatis, T., S. Goodbourn, and J. A. Fischer. 1987. Regula- immunoglobulin promoter in vitro. Cell 51:783-793. tion of inducible and tissue-specific gene expression. Science 86. Serfling, M., M. Jasin, and W. Schaffner. 1985. Enhancers and 236:1237-1245. eukaryotic gene transcription. Trends Genet. 1:224-230. 67. Marriott, S. J., I. Boros, J. F. Duvall, and J. N. Brady. 1989. 87. Silver, B. J., J. A. Bokar, J. B. Virgin, E. A. Vallen, A. Milsted, Indirect binding of human T-cell leukemia virus type 1 tax1 to and J. H. Nilson. 1987. Cyclic AMP regulation of the human a responsive element in the viral long terminal repeat. Mol. glycoprotein hormone subunit gene is mediated by an 18-bp Cell. Biol. 9:4152-4160. element. Proc. Natl. Acad. Sci. USA 84:2198-2202. 68. Merino, A., L. Buckbinder, F. H. Mermelstein, and D. Rein- 88. SivaRaman, L., S. Subramanian, and B. Thimmappaya. 1986. berg. 1989. Phosphorylation of cellular proteins regulates their Identification of a factor in HeLa cells specific for an upstream binding to the cAMP response element. J. Biol. Chem. 264: transcriptional control sequence of an ElA-inducible adenovi- 21266-21276. rus promoter and its relative abundance in infected and unin- 69. Montagne, J., C. Beraud, I. Grenon, G. Lombard-Platet, L. fected cells. Proc. Natl. Acad. Sci. USA 83:5914-5918. Gazzolo, A. Sergeant, and P. Jalinot. 1990. Taxl induction of 89. Sorger, P. K., and H. R. B. Pelham. 1988. Yeast heat shock the HTLV-I 21 bp enhancer requires co-operation between two factor is an essential DNA-binding protein that exhibits tem- cellular DNA-binding proteins. EMBO J. 9:957-964. perature-dependent phosphorylation. Cell 54:855-864. 70. Montminy, M. R., and L. M. Bilezikjian. 1987. Binding of a 90. Stamminger, T., H. Fickenschen, and B. Fleckenstein. 1990. nuclear protein to the cyclic-AMP response element of the Cell type-specific induction of the major immediate early somatostatin gene. Nature (London) 328:175-178. enhancer of human cytomegalovirus by cyclic AMP. J. Gen. 71. Montminy, M. R., K. A. Sevarino, J. A. Wagner, G. Mandel, Virol. 71:105-113. and R. H. Goodman. 1986. Identification of a cyclic AMP- 91. Staudt, L. M., H. Singh, R. Sen, T. Wirth, P. A. Sharp, and D. responsive elemnent within the rat somatostatin gene. Proc. Baltimore. 1986. A lymphoid-specific protein binds to the Natl. Acad. Sci. USA 86:6682-6686. octamer motif of immunoglobulin genes. Nature (London) 72. Nakamura, M., M. Niki, K. Ohtani, and K. Sugamura. 1989. 323:640-643. Differential activation of the 21-base-pair enhancer element of 92. Tan, T.-H., M. Horikoshi, and R. G. Roeder. 1989. Purification the human T-cell leukemia virus type 1 by its own trans- and characterization of multiple nuclear factors that bind to the activator and cyclic AMP. Nucleic Acids Res. 17:5207-5221. TAX-inducible enhancer within the human T-cell leukemia 73. O'Farrell, P. H. 1975. High resolution two-dimensional gel virus type 1 long terminal repeat. Mol. Cell. Biol. 9:1733-1745. electrophoresis of proteins. J. Biol. Chem. 250:4007-4021. 93. Tanaka, M., and W. Herr. 1990. Differential transcriptional 74. Osborne, T. F., G. Gil, M. S. Brown, R. C. Kowal, and J. L. activation by Oct-1 and Oct-2: interdependent activation do- Goldstein. 1987. Identification of promoter elements required mains induce Oct-2 phosphorylation. Cell 60:375-386. for in vitro transcription of the hampster 3-hydroxy-3-methyl- 94. Tsukada, T., J. S. Fink, G. Mandel, and R. H. Goodman. 1987. VOL. 10, 1990 CREB CORRESPONDS TO AND FORMS COMPLEXES WITH ATFs 6203

Identification of a region in the human vasoactive intestinal EMBO J. 9:841-847. polypeptide gene responsible for regulation by cyclic AMP. J. 99. Widen, S. G, P. Kedar, and S. H. Wilson. 1988. Human Biol. Chem. 262:8743-8747. 3-polymerase gene. J. Biol. Chem. 263:16992-16998. 95. Turner, R., and R. Tjian. 1989. Leucine repeats and an 100. Yamamoto, K. K., G. A. Gonzalez, W. H. Biggs, and M. R. adjacent DNA binding domain mediate the formation of func- Montminy. 1988. Phosphorylation-induced binding and tran- tional cFos-cJun heterodimers. Science 243:1689-1694. scriptional efficacy of nuclear factor CREB. Nature (London) 96. Vinson, C. R., P. B. Sigler, and S. L. McKnight. 1989. 334:494-498. Scissors-grip model for DNA recognition by a family of leucine zipper proteins. Science 246:911-916. 101. Yamamoto, K. K., G. A. Gonzalez, P. Menzei, J. Rivier, and M. R. 1990. Characterization of a bipartite activa- 97. von der Ahe, D., S. Janich, C. Scheidereit, R. Renkawitz, G. Montminy. Schuetz, and M. Beato. 1985. Glucocorticoid and progesterone tor domain in transcription factor CREB. Cell 60:611-617. receptors bind to the same sites in two hormonally regulated 102. Zajchowski, D. A., H. Boeuf, and C. Kedinger. 1987. Ela promoters. Nature (London) 313:706-709. inducibility of the adenoviral early E2a promoter is determined 98. Watanabe, H., T. Wada, and H. Handa. 1990. Transcription by specific combinations of sequence elements. Gene 58:243- factor E4TF1 contains two subunits with different functions. 256.