Transcriptional control by steroid hormones

J. Steroid Biochem. Molec. BioL Vol. 41, No. 3-8, pp. 241-248, 1992 Printed in Great Britain

TRANSCRIPTIONAL

CONTROL

0960-0760/92$5.00+ 0.00 PergamonPress plc

BY STEROID

HORMONES

MATHIASTRUSS,* GEORGESCHALEPAKIS,+BENJAMINPII~IA,~DOMINGOBARETTINO,§ ULF BRCrGGEMEIER,¶MARTHAKALFF,EMILYP. SLATER and MIGt~L B~ATO Institut fiir Molekularbiologie und Tumorforschung, E.-Mannkopff-Stra~ 2, D-3550 Marburg, Fed. Rep. Germany

Smmnary--Gene regulation by steroid hormones leads to induction or repression of particular sets of genes. These effects are mediated by intracellular hormone receptors that, in the unliganded state, are maintained in an inactive form by unknown mechanisms possibly involving association with other cellular proteins. Induction of the mouse mammary tumor virus (MMTV) requires binding of the hormone receptor to a complex hormone-responsive element (HRE) located between 75 and 190 bp upstream from the start of transcription. The interaction of several receptor molecules with the four receptor binding sites in the HRE is highly cooperative on circular DNA molecules and each individual site is needed for optimal induction. In chromatin the HRE is precisely organized in phased nucleosomes. Following hormone treatment and receptor binding, changes in chromatin structure are detected that correlate with binding of transcription factors, including nuclear factor I, to the MMTV promoter. However, though nuclear factor I acts as a basal transcription factor on the MMTV promoter it does not cooperate with the hormone receptors in terms of binding to free DNA, and mutation of the nuclear factor I binding site does not eliminate hormonal stimulation. This residual induction is mediated by octamer motifs, upstream of the TATA box, that bind the ubiquitous transcription factor OTF-1. Mutation of these octamer motifs does not influence basal transcription in vitro, but completely abolishes the stimulatory effect of progesterone receptor.

INTRODUCTION

protein components of chromatin. Conversely, Understanding the mechanisms by which cells binding of regulatory proteins to chromosomal are able to specify the unique fraction of their DNA may alter the structure and/or the funcgenetic information to be expressed in a particu- tion of chromatin. In this paper we will focus on the molecular lar context and how they control the extent of mechanism o f reversible gene regulation by sterits expression at a given time, are essential goals oid hormones taken as an example of signals of molecular biology. In many cases, this is acting through nuclear receptors. We will first mediated by an interaction of D N A regulatory review what is known about the structural and proteins with specific nucleotide sequences. As functional organization of the hormone recepeukaryotic D N A is organized into chromatin, tors and the D N A sequences to which they bind. detection of sequence information by regulatory To illustrate gene activation by steroid horproteins is not a trivial task and may be affected by the relationship between DNA and the mones we will use as an experimental model the hormone-responsive element (HRE) of mouse mammary tumor virus (MMTV). In this context we will discuss the role played by D N A topProceedings of the lOth International Symposium of the Journal of Steroid Biochemistry and Molecular Biology, ology in facilitating the synergistic interactions Recent Advances in Steroid Biochemistry and Molecular among receptor molecules and between recepBiology, Paris, France, 26--29 May 1991. tors and other transcription factors. In particu*To whom correspondence should be addressed. tPresent address: Max-Plank-Institut fiir Biophysikalisehe lar I will mention the function of nuclear factor Chemic, Postfach 2841, 3400 Gfttingen, Fed. Rep. I (NFI) and the octamer-binding factor OTF-1 Germany. :~Present address: M.I.T., Department of Biology, Cam- with respect to basal and induced activity of the bridge, MA 02139-4391, U.S.A. M M T V promoter. In discussing these topics, §Present address: EMBL, Meyershofstr. 1, 6900 Heidelberg, emphasis will be placed on the role of nucleoFed. Rep. Germany. ¶Present address: Bayer AG., 5090 Leverkussen, Fed. Pep. some positioning in determining the accessibility Germany. of the M M T V promoter elements to different 241

242

MATHIASTRUSS et al.

transcription factors, and on the use of cell-free transcription assays to dissect the different aspects of regulated transcription. During the past few years the receptors for steroid hormones have been cloned and sequenced. As a result of this work, it is now clear that the steroid hormone receptors belong to a large superfamily of nuclear receptors that comprises the receptors for retinoic acid, thyroid hormones, and several genes for which a physiological ligand is not yet known [1]. This later class of "orphan" receptors includes genes of known function, such as the knirps gene of Drosophila, and several genes of unknown function. Common to all members of the nuclear receptor family is a short DNA-binding domain composed of some 70 amino acid residues containing many conserved cysteines. Eight of these cysteines can be organized into two so-called zinc fingers, each containing four cysteine residues tetrahedrally coordinating a zinc ion. This structure was originally proposed for the transcription factor TFIIIA from Xenopus laevis, where instead of four cysteines, a pair of cysteines and a pair of histidines serve to coordinate each of the nine zinc ions that build up the basic repeated structure of the protein [2]. A comparison of the amino acid sequence in the DNA-binding domain of the different nuclear receptor genes allows the classification into two subfamilies (Fig. 1, top). The glucocorticoid

receptor (GR) is the prototype of the smaller subfamily that includes the progesterone receptor (PR), the androgen receptor and the mineralocorticoid receptor. The prototype of the larger subfamily is the estrogen receptor (ER) and this group includes the vitamin D3 receptor, the various thyroid hormone receptors, the receptors for retinoic acid and many of the "orphan" receptors. The main differences between the members of the two subgroups of nuclear receptors reside in the knuckles of the two zinc fingers, in regions that have been shown to be important for DNA sequence recognition and receptor dimerization (Fig. 1, top, see below). The amino acids following the first finger have been shown to be organized into an or-helix that serves as the reading head penetrating the major groove and recognizing the specific nucleotide sequence of the HRE. In the knuckle of the second zinc finger there is a five amino acids sequence, called region D, that participates in the dimerization between two receptor molecules, and determines the correct spacing of the dimer. This is important for the recognition of the proper spacing between the two halves of the palindromic HRE. Following the second zinc finger there is another s-helical region that serves to position the recognition helix in the correct orientation. Therefore, the general structure of the DNA-binding domain of the nuclear receptors is reminiscent of the structure found in many other prokaryotic and eukaryotic

Helix I

D

r

1

Loop

I

|1

Helix II II

I

KLCLVCSDEASGCHYGVLTC~]CKVFFKRAVEGQHN-YLCAGRNDC I I OK I RRKNCPACRYRKCLQAGM h GR --C--C ............. C--C ................ C..... C......... C--C .......... RYCAVCNDYASGYHYGVWSC~CK~FFKRS I -QGHNDYMCPATNQCT I DKNRRKSCQACRLRKCYEVGM h ER I I I It II I I

II

Finger I

I

l

Finger II

Intron

Position: G R E / PRE:

ERE:

1

2

~ T ~...~...~G ~ ~:~'"~ "'" ~l~ii ~:

3

4

5

6

T A

T C

T C

C A

C

N

G

Fig. 1. Top. The DNA binding domain of the nuclear receptors. The one letter amino acid code is used t o illustrate sequence conservation in the DNA-binding domain of nuclear receptors. Only those positions

that differ from the GR are indicated by letters. The position of the cysteine residues involved in zinc coordination are indicated at the center. All of the members of the B subgroup have the sequence EG at positions 21/22. Bottom. Nucleotidesequence of the half palindromes recognizedby the GR/PR and by ER.

Transcriptional control by steroid hormones

DNA-binding proteins, namely: helix, loop/ turn, helix. The DNA-binding domain has been postulated to be transcriptionally active in the absence of hormone, although to a much lesser extent than the intact receptor. In addition, this small region seems to also contain a nuclear localization signal that acts independently of ligand binding [3]. In addition to their characteristic DNAbinding domain, the nuclear receptors exhibit a large carboxy-terminal domain responsible for binding of the hormone ligand. In the GR and PR, some of the amino acid residues that are contacted by the hormone ligand have been identified in photo-crosslinking experiments [4, 5]. Genetic analysis of the ER has also been used to map the region of the hormonebinding domain responsible for direct binding of estrogens [6]. In these latter studies, it was possible to separate the region implicated in hormone binding from that responsible for ligand-induced receptor dimerization. The hormone binding domain also contains a so-called transcriptional activation function that is dependent for activity on binding of the ligand [7]. A refined analysis of this transcription activation function is complicated by the existence, in the same region, of a nuclear translocation signal that is also dependent upon hormone binding [3]. The amino terminal half of the nuclear receptors is their most variable region. The function of this region is also less well-defined. In GR, ER and PR a transactivation function, which seems to be independent of ligand binding, has been assigned to this region. Whether this region plays a role in the synergistic interactions among receptor dimers bound to DNA is not clear yet. In some cases, like for PR in human, rodents and chicken, several forms of the receptor are found that differ in the length of this amino terminal region [8, 9]. There are indications that these different forms may differ in their ability to induce different genes, and that these effects could also be dependent on the cell line. BINDING OF HORMONE RECEPTORS TO HORMONE-RESPONSIVE ELEMENTS

The nucleotide sequences recognized by the steroid hormone receptors and responsible for their effects on gene activity, the so-called hormone-responsive or regulatory elements (HRE),

243

have been studied in great detail [1]. In most cases, these sequences have a palindromic structure with two unequal halves separated by three non-conserved base pairs. In compliance with the division of the nuclear receptors into two subgroups according to the structure of their DNA-binding domain (see above), the HREs can also be divided into two subgroups: the GRE/PRE subgroup and the ERE subgroup. The half palindrome of the GRE/PRE has the general structure TGTYCT, whereas the prototype ERE half is TGACC (Fig. 1, bottom). We have recently shown that the last base of the half GRE/PRE, though highly conserved, is not essential for receptor binding nor for functional activity [10]. Therefore, the main differences between the two types of HREs reside in the third and fourth positions. As the fourth position of the GRE/PRE is a C in 40% of the cases, the most important distinction is the T in the third position of GRE/PRE as opposed to the A of the ERE. We know that GR and PR contact the 5'-methyl group of the T in the third position, but this base can be replaced by an A, provided that a T is located in the fourth position. Moreover, the ERE of the rabbit uteroglobin promoter contains a C in the third position [11], suggesting that ER does not directly contact this position of the ERE. It seems that the sequence context of the individual positions allows some flexibility in the way the DNA-binding domain of the receptors contacts the HRE. This flexibility is further documented by the observation that hybrid HREs composed of a half GRE/ PRE linked to a half ERE are recognized by GR and PR as well as by ER, and respond to glucocorticoids, progestins and estrogens in transfection experiments [12]. The amino acid side chains responsible for the distinction between GRE/PRE and ERE are apparently located at the knuckle of the first zinc finger, between cysteines 3 and 4, and immediately downstream in the recognition helix (Fig. 1, top) [13-16]. In addition, a few amino acid residues located between the cysteines 1 and 2 of the second finger, region D (Fig. 1, top), are also important in specifying HRE target recognition, probably by dictating the precise orientation of receptor dimers [15-17]. As the three-dimensional structure of the zinc finger region of GR has been elucidated in NMR-studies[17], it will soon be possible to postulate and to test direct contacts between individual amino acid side chains and individual base pairs within the HRE.

244

M^TmASTRUSSet al.

The initial identification of DNA elements able to mediate hormone induction in gene transfer experiments was accomplished with the MMTV system (for review see Ref. [1]). Later, this system provided the first demonstration for specific binding of a steroid hormone receptor to DNA. Consequently, the interaction of GR and PR with the HRE of the MMTV was studied in great detail[18]. In the G R mouse strain, there are four copies of the hexanucleotide motif TGTTCT between positions -190 and - 7 5 of the MMTV promoter and each of these motifs is contacted by GR or PR. Whereas, the upstream motif is part of an imperfect palindrome between - 190 and - 160, the other three motifs do not have a corresponding half with the correct spacing. Thus, a dimer of the receptor binds to the promoter-distal palindrome, but the exact stoichiometry of receptor binding to the promoter proximal sites is unknown[18]. The results of insertions between the distal and the proximal sites suggest that there is a strong functional cooperativity between receptor molecules bound to these two regions[18]. This cooperativity is strongly dependent upon the topology of the transfected plasmid[19] and is not accompanied by corresponding changes in the affinity of the receptors for linear DNA fragments[18]. We assume, therefore, that a functional interaction among receptor molecules requires bending or other deformations of the MMTV DNA that are favored by negative supercoiling [19]. A similar argument applies for the interaction between the hormone receptors and other transcription factors mediating induction of the adjacent promoter [19]. A role for DNA topology on receptor binding to the MMTV-HRE has been suggested by binding experiments with plasmids of various topologies [19], and confirmed using minicircles. We found that the affinity of the receptor is low for relaxed minicircles and increases with the degree of negative supercoiling, with an abrupt increase between topoisomers - 2 and - 3 [M. Truss and M.B., unpublished]. Interestingly, a similar topology-dependent transition is observed with respect to the sensitivity of the minicircles to the nuclease Bal31. This suggests that a conformational transition is induced by supercoiling, which favors binding of the hormone receptor to the HRE. The reverse effect should therefore be predicted and has been observed, namely, that binding of the receptor

favors a topological transition of plasmids carrying the MMTV-HRE [20].

MECHANISM OF TRANSACTIVATION: INVOLVEMENT OF TRANSCRIPTION FACTORS NFI A N D OTF-I

After proving that binding of several receptor molecules, and their functional interaction are essential for induction of MMTV transcription, the next question concerned the mechanism of the transactivation itself. More specifically, we wanted to know what factors mediate the effect of receptor binding on transcriptional efficiency. Several reports had been published describing the participation of the transcription factor NFI in hormonal induction of the MMTV promoter[21 and references therein]. However, the role of NFI as a transcription factor on the MMTV promoter had never been proven. In complementation experiments with NFIdeficient choriocarcinoma cell lines we demonstrated that NFI indeed acts as a transcription factor in vivo [21]. However, binding of NFI to its cognate sequence between - 7 5 and - 6 3 on the MMTV promoter is not favored by binding of the hormone receptors. On the contrary, both proteins compete for DNA binding, in agreement with the overlap of their respective binding sites [21]. Therefore, a simple mechanism involving binding cooperativity between hormone receptors and NFI is not in agreement with our findings. In cell-free transcription experiments we have shown that addition of the PR to templates driven by the MMTV promoter enhances their transcriptional efficiency about 10-fold [22]. In this system, deletion of the NFI binding site, or addition of an excess of oligonucleotide carrying the NFI consensus sequence, reduces the basic expression of the MMTV promoter dramatically, but does not influence the stimulatory effect of preincubation with the PR [22]. These in vitro data confirmed the in vivo results and in addition show that NFI acts as a basal transcription factor on the MMTV promoter, without any indication for synergism or cooperativity with the hormone receptors. We conclude that factors other than NFI must mediate the transactivation of the MMTV promoter by hormone receptors. A search for other possible factors involved in transcriptional activation of the MMTV promoter lead to the identification of two octamer motifs between the NFI binding sites and the

Transcriptional control by steroid hormones

TATA box. Mutations at these sites resulted in a significant reduction of the hormonal induction of the MMTV promoter in gene transfer experiments [23]. More interestingly, these mutated promoters were completely unresponsive to the addition of PR in vitro, suggesting that the effect of PR on cell-free transcription of the MMTV promoter is actually mediated by an octamer binding factor [23]. In fact, since OTF1 (Octl) is the main octamer binding factor in HeLa cells, from which the nuclear extracts are prepared, and purified OTF-1 binds to the two octamer motifs of the MMTV promoter, we assume that the effect of PR on transcription of MMTV-DNA is mediated by OTF-1 [23]. MECHANISM OF TRANSACTIVATION: ROLE OF CHROMATIN STRUCTURE

The lack of cooperativity between hormone receptors and NFI in terms of binding to DNA was unexpected, because in vivo hormone treatment does induce NFI binding to the MMTV promoter [24]. On the other hand, the MMTV-LTR region is precisely organized into nucleosomes, and following hormone administration, a DNaseI hypersensitive site appears over the HRE [21]. It seemed, therefore, possible that changes in chromatin structure could mediate the effect of hormone treatment on NFI binding. To directly test this hypothesis we have performed nucleosome reconstitution experiments with the relevant region of the MMTV promoter. We confirm the previous in vivo [25] and in vitro [26] observations that a nucleosome core is precisely positioned over the region containing the HRE [27]. In addition, we find that the binding site for NFI is included in this nucleosome that extends from -198 to -45. The orientation of the NFI binding sequences is such that their major groove is pointing

=

II

-~0 i

I

=1

-'=~I ~I

II I

=1

-1701 il

=

inwards toward the histone octamer, and is unaccessible for the protein (Fig. 2). This prediction was confirmed experimentally with purified pig liver NFI that binds very efficiently to naked MMTV-DNA, but is unable to recognize the MMTV promoter organized in nucleosomes [27]. On the contrary, the hormone receptors do bind efficiently to reconstituted nucleosomes [26], with an affinity only 4- to 5-fold lower than their affinity for naked MMTV-DNA [27]. This is in agreement with the prediction based on the structure of the reconstituted nucleosome, as only two of the four TGTTCT motifs have their major grooves exposed to the exterior and accessible for receptor binding (Fig. 2). Mutation of one of the two sites that are masked in nucleosome B has a dramatic influence on receptor binding to linear free DNA (over 10-fold). This suggests that the accessible hormone binding sites exhibit a higher affinity when organized in nucleosomes than they have in free DNA. A possible reason for this observation is that the major groove is widened when the DNA double helix is wrapped around the nucleosome, and may therefore enable a better contact between the recognition helix of the receptor and the relevant base pairs of the HRE. Thus, it seems that the precise positioning of the DNA double helix on the surface of the histone octamer could account for the lack of binding of NFI prior to hormone treatment. In the absence of hormone, the promoter is inaccessible to NFI and silent in vivo. If the model is correct, and promoter occlusion is due to nucleosome positioning, binding of receptor to the nucleosomally organized MMTV promoter should alter its structure and expose the recognition sequence of NFI. We have preliminary evidence supporting this concept. Upon binding of receptor, the nucleosome is not disassembled,

~

-160 i

245

=

-Iso i

-140 i

=

-130 i

T T A T T ¢,^ A A T/~":,C ^^.^^ TT^ T ~T~T T ( ~ ¢ A A{-"~^~A T T T T {.'~TTC~ T ^C,Ac T ~ T ~ T T C ^C'G A ^ A