Studies on the mechanism of functional cooperativity between progesterone and estrogen receptors.

Steroid response elements (SREs) cooperate with many different cis-acting elements including NF-1 sites, CACCC boxes, and other SREs to induce target gene expression (Schule, R., Muller, M., Otsuka-Murakami, H., and Renkawitz, R. (1988) Nature 332, 87-90; Strahle, U., Schmid, W., and Schutz, G. (1988) EMBO J. 7, 3389-3395). Induction of gene expression can be additive or synergistic with respect to the level of activation by either transactivators. Two mechanisms have been proposed for how synergism occurs: 1) cooperative binding of transcriptional activators to DNA or 2) simultaneous interaction of individually bound activators with a common target protein. We have shown previously that cooperative binding of receptors is important for synergism between two progesterone response elements (PREs). Here we showed that an estrogen response element (ERE) and a PRE can also functionally cooperate and this synergism between an ERE and a PRE is not contributed by cooperative DNA binding. Furthermore, we have demonstrated that the activation domains of the progesterone receptor (PR) (C1Act) are required for synergism between two PREs and sufficient for confirming cooperative binding. However these two activation domains of PR are not sufficient for synergism between an ERE and a PRE. Additional regions within the NH2-terminal and COOH-terminal domains are also required for synergistic interaction between two heterologous SREs.

can be additive or synergistic with respect to the level of activation by either transactivators. Two mechanisms have been proposed for how synergism occurs: 1) cooperative binding of transcriptional activators to DNA or 2) simultaneous interaction of individually bound activators with a common target protein. We have shown previously that cooperative binding of receptors is important for synergism between two progesterone response elements (PREs). Here we showed that an estrogen response element (ERE) and a PRE can also functionally cooperate and this synergism between an ERE and a PRE is not contributed by cooperative DNA binding. Furthermore, we have demonstrated that the activation domains of the progesterone receptor (PR) (ClAct) are required for synergism between two PREs and sufficient for confirming cooperative binding. However these two activation domains of PR are not sufficient for synergism between an ERE and a PRE. Additional regions within the NH2-terminal and COOH-terminal domains are also required for synergistic interaction between two heterologous SREs.
Steroid hormones control gene expression via specific interactions of their cognate receptors with steroid response elements (SREs)' in the 5"flanking DNA of steroid regulated genes (3-5). Once bound, the mechanism by which receptors activate transcription is only partially understood (6,7). Many eucaryotic genes are under the control of multiple hormones and steroid response elements are usually found in multiple * This work is supported by grants from the National Institutes of Health (to M. J. T. and B. W.O.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The abbreviations used are: SREs, steroid response elements; PREs, progesterone response elements; GRE, glucocorticoid response element; ER, estrogen receptor; ERE, estrogen response element; PR, progesterone receptor; aa, amino acids; PCR, polymerase chain reaction; CAT, chloramphenicol acetyltransferase. copies or tightly clustered with other cis-acting DNA elements or SREs (1, 3, 8-13). The tryptophan oxygenase gene (8), murine mammary tumor virus long terminal repeat (ll), and vitellogenin A2 and B1 genes (9) contain multiple SREs. In addition, the murine mammary tumor virus long terminal repeat (1 1) and the rat tryptophan oxygenase gene (1) contain other cis-acting elements in close conjunction with SREs. In all of these cases, when one of the SREs or the adjacent cisacting elements are mutated, steroid responsiveness is greatly diminished. Thus, SREs interact synergistically with each other and with other cis-acting elements to induce gene expression. The mechanism of synergism between SREs and various cis-acting elements is not known.
Ptashne (14) has proposed two models by which synergistic activation occurs. First, synergistic activation can occur when two factors bind cooperatively to adjacent cis-acting elements. Second, synergism can occur when two factors bind independently to DNA but simultaneously interact with a third transcription factor to initiate transcription.  (18) have demonstrated recently that synergism occurs between GAL4 binding sites and between GAL4 and ATF sites even when all sites are saturated with activators. This result is consistent with the hypothesis that activators cooperate by simultaneously touching a target protein. The observation that ER can squelch the activation of a progesterone-responsive target gene by the progesterone receptor supports this possibility (19,20).
In this paper, we demonstrate that the estrogen receptor cooperates functionally with the progesterone receptor, but that the synergism does not result from cooperative binding to their respective SREs. The two activation domains of the progesterone receptor are shown to be essential for synergism both between two PREs and between an ERE (estrogen response element) and a PRE. These two activation domains are sufficient for partial synergism of two PREs but not sufficient for synergism between an ERE and a PRE. Additional domains residing within the NH2-terminal and COOHterminal regions of the heterologous receptors are required. These data are consistent with the hypothesis that progesterone and estrogen receptors regulate target genes cooperatively via simultaneous protein-protein contacts with a target protein(s) in the transcriptional machinery.
Expression vectors for production of wild-t.ype steroid hormone receptors were: 1) pS9K (22,23) containing the chicken progesterone receptor cnNA and 2 ) AHER/p91023 (21) containing the human estrogen receptor cDNA. Oligonucleotide site-directed mutants of progesterone receptor (1'13) were constructed as descrihed in Dohson pNC,Act was created hy I T 1 3 to contain amino acid sequence from AUG, to amino arid 411 at the 3' end.
At least two individual isolates of each PCR-generated plasmids were used in all suhsequent transfection assays t o eliminate the possihilitv of mutation during PC13 reaction. I;schrrichin coli expression vector (pGEX-2T C,Act) for CST-ClAct fusion protein WAS constructed as the following: the Ecnlil fragment of ClAct (1459-1958) was inserted into the I:'coRl site of pGI<M"izf plasmid, it was then digested with Avnl enzvme and rendered hlunt-ended with Klenow enzvme. The ClAct expression vector. pCE?<-2'1"ClAct, was constructed hy inserting the s526-hase pair hlunt-ended Aonl fragment into the SmnI site of pGEX-27' (37) (I'harmacia I,KR Hintechnnlogy Inc.).
Trnns/rctions-The reporter const rurt ions and the receptor expression vectors were cotransfected into CV-1 cells t~v t h e p o l v t~r e n e method as desrrit)ed previously (22). The only exceptions were the anti-hormone experiments in which fetal calf serum was replaced hv the defined medium ITS+ (Collahorator Research) and hormone was rldded as descrihed in Fig 2. At 48-h post-transfection, extracts were assaved for CAT activity hv the method of Gorman rf 01. (27).
The end-Irlheled fragment was incul~ated with 11 ng of partially purified chicken progesterone receptor (288) and 0.4 ng of partially purified human estrogen receptor from MCF-7 cells (29). Reaction conditions were as desrril)ed previously (28). Unlaheled competitor PRE oligonucleotide was added hefore the addition o f receptor in the molar ratios indicated in Fig. 3 . The DNA-receptor complex and free DNA were separately excised and counted in a Heckman scintillation counter.

The Estrogpn Hrccptor Can Cooprrntc F'unctionnl!\. with t t w
Progesterone Hecppptor-Two PRES have been shown to art synergistically in the absence of a distal promoter element when located near the TATA box (21). To determine if an estrogen response element (ERE) can cooperate functionally with a PRE, we inserted a single ERE -5' to A single I'RF: upstream of a minimal promoter fused to the CAT reporter gene t,o creat,e pOVCAT-50 (ERE/PRE) (Fig. 1, hottompnnrl). ??
gether (PR + E R ) , or vector control ( V ) . All these experiments were carried out in the presence of their cognate ligands at 10" M. Fig. 1A shows that PR or ER alone induce very low levels of gene expression, hut the two receptors together act synergistically to induce expression. T h e level of CAT activity with both receptors is 4-%fold higher than expected for the additive effect of each individual receptor. These data demonstrate that the estrogen receptor can synergize functionally with the progesterone receptor. Hormone Is Hcyuired for Synergism of the Estrogen and Progesterone Reccptors-To determine if hormone is required for synergism between estrogen and progesterone receptors, we used anti-estrogens in t.he cotransfection experiments. Anti-estrogens have recently been shown to interact with the estrogen receptor in such a way that the receptor hinds to DNA hut is not activated (33). The ERE/PRE containing reporter was cotransfect.ed with both receptors in the presence or absence of various antiest,rogens . Fig 2 shows t,hat in the absence of any hormone no induction occurs. The addition of estrogen ( E ) and progest.erone ( P ) cooperatively induce gene expression. Replacing estrogen with either lo-' M tamoxifen (Tam) or lo-' M nafoxidine (Naf) blocked induction. Thus, DNA binding of a transcriptionally inactive form of estrogen receptor is not enough for Synergism between the estrogen and progesterone receptors. Activation by authent.ic ligand is necessary for the estrogen receptor to synergize with progesterone receptor.
Cooperative Binding Does Not Contrihute to Functional Synergism of Heterologous Receptors-We have demonstrated previously that progesterone receptor and estrogen receptor bind specifically t.o their target PRE and ERE (28, 34). In addition, we have shown that the progesterone receptor has higher affinity for a PRE when an adjacent PRE is already occupied (15). Using an in vitro transcription assay we have demonstrated that this cooperative binding contributes to functional synergism hetween two PRES (7). To determine if cooperative binding contributes also to functional synergism between heterologous receptors, we used a hand-shifting competition assay. A fragment of DNA containing both steroid response elements was labeled and incubated with either PR alone or P R a n d E R . Fig. 3, A and H , show that in the presence of unlabeled P R E oligonucleotide the rate of competition was the same for the complex containing either PR alone or both receptors. This result suggests that the specific binding of ER does not stabilize the binding of P R t o a nearby PRE. Thus, functional synergism between an ER and PR must occur by a mechanism other than cooperative binding to DNA.
Two Regions of the Progesteronr Rrcrptor Arr Rpyuirrd for Cooperatiuity-To determine what regions of P R a r e required for cooperativity, we utilized PR deletion mutants (24, 2 5 ) .
PRClH contains only the amino terminus and the DNA binding domain of P R . P R C l C 2 cont.ains only the DNA binding domain and the carboxyl terminus. Each of these constructions was cotransfected with a reporter containing an ERE/PRE. Neither mutant alone, or with ER, was able to induce gene expression (Fig. 4). Thus, synergistic induction is lost upon deletion of the NH2-terminal (aa 1-280) or COOH-terminal (aa 369-659) regions of the receptor.
T o further define the important regions of PR, we used oligonucleotide site directed mutants of P R (Fig. 4). Each of these mutants has been shown to bind DNA and hormone with wild-type affinity (24). Fig. 4 shows that mutants PR14, PR16, and PR17 have a t least wild-type ability to cooperate with ER. However PR18 and PR5 have much reduced ability t o cooperate with ER to induce target gene expression. Similarly, Dobson et al. (24) have shown that these same two mutants have greatly reduced ability to induce transcription from a target gene containing two PRES. Thus, the two regions flanking the DNA binding domain, aa 262-287 (PR18) and aa 383-411 (PR5), are important for functional synergism of both heterologous and homologous receptors. These regions contain the transcriptional activation domains of the progesterone receptor.
We used Western analysis, to demonstrate that the receptor mutants described in Fig. 4 are expressed to similar levels in transient transfections. Fig. 5 shows that wild-type receptor ( 7 9 K ) and mutant receptors PR18 and PR5 are expressed to similar levels in transfected cells. These recombinant receptors migrated on gels to a position approximate to that of partially purified progesterone receptor (form A) from chicken oviduct (lane 2). The molecular sizes of the mutants are not expected to be significantly smaller than the wild-type receptor, because the deletions are too small to he det,ected by this analysis. We conclude that the reduced CAT expression driven by PR5 and PR18 is not due to instabilitv of these mutants in transfected cells.

Different Regions of PR Are Requirpd for K~gulation of Trrw PRES and ERI;/PRR--If
the two subregions of the receptor characterized in the previous experiment are indeed the activation domains of PK, we would expect that these two regions. in conjunction with the DNA binding domain, would activate transcription. T o test this hypothesis, we fused these reEions individually or together to the DNA binding domain of PR. Fig. 6 shows the constructions in which the DNA binding domain is fused either to the COOH-terminal activation domain, pC1-5 (aa 368-421), or the NH,.-terminal activation domain, pC1-18 (aa 199-287). Neither of these constructions was capable of inducing expression of reporters containing either two PRES or an ERE/PRE (Fig. 6 ) . pClAct (aa 2.50 To investigate the regions of PR, in addition to the two activation domains, that are required for synergism between an ERE and a PRE, we subcloned a fragment containing aa 250-659 (pClActC2) and a fragment containing aa 1-411 (pNC1Act). The construction pClActC2 contains both activation domains and the carboxyl terminus. It differs from deletion mutant PRClC2 in that the latter does not contain the 5' activation domain. When pClActC2 or pNClAct was cotransfected with the reporter containing two PREs, CAT activity was induced to 50% of wild-type PR in both cases (data not shown). In contrast, when the reporter contains an ERE and a PRE, the recombinant receptor forms did not induce expression in the presence of ER (Fig. 7). ClActC2 is identical to ClAct except that ClActC2 also contains the entire COOH terminus. Yet, ClActC2 appears to have lower levels of expression in the presence of ER. This is due to a difference in transfection efficiencies in the experiments in Fig. 7. The lower transfection efficiency of these experiments indicated by the lower level of wild-type PR + ER expression.
Therefore, the carboxyl-or amino-terminal region alone is not sufficient to induce expression of an ERE and a PRE, suggesting that multiple regions in both NH2-and COOHterminal domains are required for functional synergistic interaction with an estrogen receptor.
Cooperative Binding of ClAct to Two PRES-As shown in Fig. 6, ClAct construct has 38% of the wild-type PR activity on the PRE2 reporter and has very little activity above the basal level on the ERE/PRE reporter (compare +ER or +PR with ER or PR alone, 9.8 versus 3.0 or 7.3, respectively. These results indicate that ClAct of PR can synergize with one another but not able to do so with ER. Since synergistic induction of PRE, reporter is through cooperative binding (7, 15) and the fact that ClAct can not synergize with ER, it suggests that ClAct synergizes with one another through the same mechanism. In order to confirm this, we synthesized ClAct protein in an E. coli expression system. ClAct was synthesized as a fusion protein by attaching it to glutathione  7. Effect of amino or carboxyl terminus on cooperativity. Cotransfections were as described in Fig. 1A. The reporter contained an ERE and PRE (Fig. 1B). Various  S-transferase so that it can be easily purified through glutathione-Sepharose 4B affinity column. The fusion protein was then cleaved by thrombin to generate ClAct. The protein purified in such a manner has a major band at 18 kDa and a small amount of partially cleaved products, Band shift experiments were then carried out to examine whether the ClAct cooperatively binds to probe containing two PREs. As shown in Fig. 8A, two protein-DNA complexes (11 and I V ) were observed. The identity of two complexes were confirmed by binding of ClAct to a probe containing a mutation in one of the PRE sites (15). As shown in Fig. 8A, the formation of complex IV was drastically decreased when one of the PRE site is mutated. In contrast complex I1 is not effected by this mutation. These data together with those obtained in our earlier paper (15) indicate that complex I1 represent the binding of ClAct dimer to one of the PRE site and complex IV represent binding of two dimers to both PREs.
To examine cooperative binding of ClAct, we next carried out binding experiments with increasing concentrations of ClAct. Fig. 8, B and C, shows that formation of complex I1 is transient and never reach more than 20% of the total probe before complex IV taking over. This suggested that ClAct binds to PRE, probe in a cooperative manner. Using the equation derived in our earlier publication ( E ) , one can calculate the relative Kd for the binding of ClAct to first PRE and the subsequent binding to the adjacent PRE. We estimated that ClAct binds to the PRE with an affinity of 20 times higher when the adjacent PRE site is already occupied as compared to that when the adjacent site is unoccupied. Thus, cooperative binding of ClAct to PREs contributes to the synergistic induction of PRE, template.

DISCUSSION
We have shown that the estrogen receptor can cooperate functionally with the progesterone receptor. Induction is 4-8fold higher than expected for the additive effect of each individual receptor. Ankenbauer et al. (35) have previously demonstrated synergism between an ERE and GRE in the chicken vitellogenin I1 gene. Using a weak GRE, they demonstrated a 2.5-&fold synergism with an ERE which is analogous to the 4-8-folds of synergism observed by us. In the present study, we have extended these studies to demonstrate a likely mechanism of synergism between an ERE and a PRE. Ptashne (14) proposed that transcriptional synergism can occur by two mechanisms (Fig. 8): A, cooperative binding of factors to adjacent cis-acting elements and B, cooperative transactivation of a target protein by two independently bound activators. We have reported previously that progesterone receptors bind cooperatively to two PREs from the TAT gene (15). In contrast, no cooperative binding of cognate receptors occcurs between an ERE and a PRE. This deduction is further supported by the anti-estrogen data since these drugs allow the receptor to bind but not activate, yet antiestrogens block synergism of the estrogen receptor with the progesterone receptor (23). Therefore, binding of estrogen receptor to DNA is not sufficient for synergism with progesterone receptor, suggesting that cooperative transactivation may occur.
Recently, Carey et al. (17) and Lin et al. (18) have shown that multiple GAL4 sites or GAL4 and ATF sites can act synergistically in the absence of cooperative binding. This was demonstrated by adding an excess of activators in vitro and in transfections to saturate available binding sites SO that cooperative binding could not affect the interpretation of their results, yet synergism was still observed. Their results suggest that synergism can occur when activators simultaneously showed that addition of progesterone receptor to an in uibo transcription system stimulates transcription of test genes harboring one and two copies of PREs 5 -and 27-fold, respectively. When similar experiments were carried out in a vast excess of progesterone receptor, the levels of transcription of a test gene containing two PREs were only 2-4-fold higher than a test gene containing one copy of a €'RE.' Taken together with our earlier published data (15) these results reveal t h a t cooperat.ive binding of progesterone receptors occurs to two PREs and that binding affinity is the primary reason for synergism in this system. This is different. however, from interactions between receptors at an ERE and a PRE.
In this case, cooperative transactivation may occur even when the binding sites are saturated.
The cooperative transactivation model predicts that in many instances the activation domains of receptor should he important for such interactions with a target protein. We have used progesterone recept,or mutants to delineate regions of the receptor important for induction of transcription. Two regions, aa 262-287 (PR18) and aa 383-41 1 (I'RS), located on either side of the DNA binding domain, are important for activation of templates containing either two PRES or a n ERE and a PRE. These regions contain the activation domains of the progesterone receptor. In conjunction with the DNA binding domain, these two activation domains are sufficient for induction of transcription from two PRES. Neither region alone is active, suggesting that both activation domains are required for cooperativity.
Since the transactivation model does not play much role in the synergistical induction of two PREs, we predicted that this synergistic activity is derived from cooperative binding. DNA binding data suggested that this indeed occurs with ClAct. Therefore the sequence important for cooperative binding must reside within the expressed sequence. The lack of synergistic induction on PREr template when individual activation domain was used suggests that both activation domains may be required for the cooperative binding. We are current,ly attempting to identify the sequence(s) important for the cooperativity. It is interesting to point out that in our DNA binding assay, we did not observe monomer binding of ClAct to the PRE half-site, suggesting that either ClAct already dimerizes in solution or the monomeric C1 Act has very low affinity for t,he half-site as compared with ClAct dimer. This observation differs significantly from our previously published results when DNA binding domain of mouse glucocorticoid receptor was used in similar studies (28). In t.hat case the DNA binding domain readily hinds to PRE half site as a monomer (28). I t is likely that ClAct may contain the dimerization domain.
There are two forms of chicken progesterone receptor derived by translation from two AUGs in the same reading frame (19). Cato and Ponta (36) identified the amino terminus of the chicken progesterone receptor (R-specific region) to be essential for synergism with the estrogen receptor. However.  (19,20) have shown that the NH2terminal 128 amino acids of the chicken progesterone receptor constitute a promoter specific activation domain. Since the reporter gene used in these studies (PRE/ERE OVCAT-50) differs from that used by Cat0 and Ponta (PRE/ERE TKCAT), this difference may account for the discrepancy between the two sets of data. Furthermore, these data would lead us to speculate that different activation domains of the progesterone receptor contribute to cooperative transactivation in a promoter-specific manner. The PR expression vector, ClAct, contains the two activating domains flanking the DNA binding domain. This construction partially induces transcription from a promoter containing two PRES but is insufficient for induction from an ERE/PRE construct. Maximal activation requires additional sequences in both the carboxyl and amino termini of PR. We conclude that multiple regions throughout the progesterone receptor are required for interaction with a heterologous receptor to produce cooperative transactivation.