The Nuclear Receptor Superfamily: A Personal Retrospect on the First Two Decades

Although lipophilic hormones (steroids, thyroid hormones, and vitamin D3) were isolated in the first part of the 20th century, their mediators remained elusive until the synthesis of radiolabeled estradiol by Jensen and Jacobsen in 1960, which allowed the identification of one or more estradiol-binding proteins. The finding that this protein was translocated from cytoplasm to nucleus upon hormone administration suggested that there was a link between control of transcription and the physiological action of the hormone (1). This concept was supported by the work of the Clever and Ashburner laboratories (2) showing that chromosomal puffing in insect polytene chromosomes was triggered by ecdysteroids. The subsequent identification of transcriptionally responsive genes in vertebrate steroid hormone target tissues, together with the finding that the glucocorticoid receptor (GR), could exhibit a modular structure with separable hormone- and DNA-binding sites (3), and also that a MMTV (mouse mammary tumor virus) GR response element (GRE) had the characteristic properties of a transcriptional enhancer element, led to a model of steroid hormone action by postulating that steroid hormone receptors are ligand-regulated enhancer-binding transcriptional factors (Refs. 4 and 5 and references therein). At this stage, the cloning of steroid receptors became an imperious necessity, because reverse genetic studies were obviously required to elucidate the molecular basis of this model. Twenty years have passed since the human GR cDNA was cloned by Ron Evans and his collaborators (6, 7); concomitantly, using antibodies prepared by Geoffrey Greene and Elwood Jensen against purified human estrogen receptor (ER), together with oligonucleotide probes derived from its amino acid sequence, we isolated and sequenced ER cDNA clones (8) and showed that one of them contained the entire open reading frame of the ER now known as ERα (9, 10). The GR and the ER, which incidentally were the first RNA polymerase II transcription factors to be cloned, exhibited a surprising homology with the v-erbA viral oncogene whose cellular counterpart, the c-erbA protein, was shown shortly afterward to be the thyroid hormone receptor (TR) (11, 12). This provided the first indication that steroid receptors could belong to a superfamily of structurally related receptors interacting with chemically unrelated ligands (13). A comparison of the above receptors with the chicken ER revealed six regions of homology (A to F, from the N-terminal to the C-terminal end), two of which (C and E) being highly conserved (14). In vitro studies showed that regions E and C were the ligand binding domain (LBD) and the DNA binding domain (DBD), respectively (15). This was unequivocally confirmed by the end of 1986, when Stephen Green demonstrated in our laboratory that a chimeric receptor, in which he had swapped region C of ER for that of GR, could bind estradiol but not activate a chimeric estradiol-responsive gene (vit-tk-CAT), whereas it activated a glucocorticoid-responsive gene (MMTV-CAT) in the presence of estradiol, but not in the presence of a glucocorticoid (16). Using a reciprocal construct, Vijay Kumar et al. (17) demonstrated that, in the presence of a glucocorticoid, the chimeric receptor could activate the expression of the vit-tk-CAT reporter gene, but not of MMTV-CAT. These experiments not only unequivocally demonstrated the modular structure of NRs, in which the DBD and the LBD can function independently (see also Ref. 18), but even more importantly they indicated how such chimeric receptors could be used to characterize the ligand requirement of a novel putative member of the NR superfamily: hooking the LBD of a novel receptor candidate to the DBD of either the ER or the GR would result, upon addition of a cognate agonistic ligand, in activation of an estrogen- or a glucocorticoid-responsive reporter gene, respectively, unless this ligand is endogenously produced by the transfected cultured cells, or the LBD of the novel receptor is constitutively transcriptionally active. The value of this approach was illustrated shortly afterward by the independent discovery of the first RAR [retinoic acid (RA) receptor] α by Martin Petkovitch in our laboratory (19) and Vincent Giguère in Ron Evans’ laboratory (20), which was followed by the subsequent finding of two additional RAR isotypes (subtypes), RARβ (21), and RARγ (22, 23). Clearly, in 1987, the stage was set to clone all members of the NR superfamily, based on sequence homology of their DBDs and LBDs. This was achieved in 1999: 48 human and 49 mouse NRs have been cloned (examination of the human and mouse genome sequences did not reveal any additional member of the superfamily), among which 24 are orphan receptors for which no ligand have yet been discovered (24, 25). Which ones will remain true orphans needs to be established (25, 26). The functional dissection of the ER demonstrated that it is a ligand-regulated transcription factor that modulates target gene transcription. Two distinct synergistic transcriptional activation functions (AFs) were unveiled: the ligand-independent AF-1 located in the N-terminal A/B region, and the ligand-dependent AF-2 encompassing region E (the LBD) (27, 28). Both ER AF-1 and AF-2 were found to act in promoter context- and cell-specific fashions (29). Similarly, the dissection of the activation functions of the six RAR and RXR isotypes (α, β, and γ) and, for each of them, of their isoforms that differ in their N-terminal A regions, revealed the existence of isotype-specific AF-2 and isoform-specific AF-1, the synergism of which was dependent on both the promoter context and the nature of the retinoic acid response element (RARE). Moreover, the AF-1 and AF-2 of a given RAR isoform (e.g. RARα) appeared to be significantly different from those of the two other RAR isotypes (RARβ and γ), as the AF-1 from a given RAR isoform synergized with the AF-2 of its cognate RAR isotype, but not with the AF-2 of the two other RAR isotypes (Refs. 30–33 and references therein). A very weak, autonomous and constitutively active, activation domain (AF-2 AD) was identified within the C-terminal end of the homology region E of RARs and RXRs (34), and shown to contain an amphipathic α-helix core motif (the AF-2 AD core), which was also present in ER and TR (Refs. 24 , 33 , and 34 and references therein). Its integrity was required for AF-2 activity, but in general not for ligand and DNA binding, which initially suggested that it may belong to an interactive surface for the binding of factors mediating AF-2 (24, 33–35). The observation that AFs acted in a promoter context- and/or cell-specific fashion suggested to us that NRs could cooperate with cell-specific promoter-bound transcription factors and/or interact with cell-specific factors mediating their activity. The latter assumption was supported by the results of transcriptional interference/squelching experiments in cultured cells, in which the activity of a given receptor was inhibited “off the DNA,” and in an agonist- and AF-2 integrity-dependent manner, by an excess (relative to its cotransfected cognate-responsive reporter gene) of the same receptor (autosquelching) or by addition of a different receptor (heterosquelching). These squelching data (36–38) were interpreted as resulting from the sequestration of transcriptional intermediary factors (TIFs) (also named mediators/coregulators) that, acting as coactivators, may mediate the AF-1 and AF-2 activities of NRs on the transcription machinery and chromatin template. To substantiate this concept, searches for proteins that would interact with the LBD in the presence of agonistic ligands, but not in the presence of antagonists, were undertaken using both yeast two hybrid-based and Far-Western blotting-based screens. The cDNAs of several such proteins were cloned using human and mouse expression libraries in 1995–1996 (33). Two of these proteins, steroid receptor coactivator (SRC)-1 (39) and TIF-2 (40), isolated by O’Malley and colleagues and our own group, respectively, had the expected properties of bona fide coactivators. For instance, TIF-2 was shown 1) to interact directly, in vitro and in cultured cells, with the LBDs of several NRs in an agonist- and AF-2 AD-core-integrity-dependent manner; 2) to harbor an autonomous activation function; 3) to relieve NR autosquelching; and 4) to enhance the activity of NR AF-2s when overexpressed in mammalian cells (40). Moreover, sequence similarities between and TIF-2 unequivocally supported the existence of a gene of NR In our for additional that would bind in a ligand-dependent fashion to LBDs of NRs, we isolated which did not exhibit all of the above bona fide of a NR but revealed the presence of a the NR motif which was also present in two putative and and references therein). we demonstrated that this NR motif was and for ligand-dependent between or and a cognate LBD surface that required the integrity of the AD core (see The presence of such NR in a of putative NR as as their in binding, was confirmed in two subsequent The as as those of its and the two other members of the was found to NR binding each the NR motif and references therein). TIF-2 known as or receptor coactivator and (also known as or were shown to contain an activation domain which with binding protein a motif which was distinct from the NR motif (Refs. and references therein). The observation that and to a the coactivators, an activity and the that an in the of the chromatin transcription provided the first indication that those of the could mediate the NR AF-2 transcription activity of the chromatin of target it to the transcription machinery (Refs. and references therein). our finding of a ligand-dependent between the ER LBD and of the it that additional chromatin interacting with LBDs could be in mediating their AF-2 activity and references therein). the other it had been found that NRs RARs and when in the (Refs. 24 , 33 , , and 48 and references therein). to those used for the of led to the identification of the receptor and of and that bind a to (e.g. RARs and or NRs (Refs. 24 , 33 , , 48 , and and references therein). of the AF-2 AD core the binding of or to TR and RAR and their agonistic were shown to in or molecular that have activities 48 and references therein). because are known to be with regions of the transcriptional by NRs appeared to be by the binding of to their which their target genes transcriptionally chromatin AF-2 could be related to the of RNA polymerase II was when its was in as the thyroid hormone or vitamin that activation of transcription by TR and vitamin receptor The that with the NR LBD was identified as to and shown to contain a functional NR motif 24 , 48 , and or between NRs and of the transcription machinery (e.g. of and were also revealed in studies in and to enhance the transcriptional activity of NRs, in experiments (Refs. 5 , 24 , and and references and been yet the that may mediate the of the ligand-independent AF-1 (see Ref. and references of the of ER AF-1 and AF-2 to and to with one and with AFs of other as as a comparison of their transcriptional interference/squelching properties suggested that they could interact with different it been shown that using distinct binding can bind to isolated ERα AF-1 and which in for the synergistic activation of transcription by the two AFs the functions of NRs and ligand binding, from an or to a transcriptionally be in of that mediate these The studies of the and the RARγ in our in with group, allowed us to in a structure for NR LBDs that a protein This was as a with to and one of LBD sequences of all NRs suggested that they all a and this been found in all NR LBDs whose have been as yet (Refs. 24 , 33 , , , and and references therein). These and additional and molecular studies revealed the of the and of the ligand binding of RARγ and as as the basis for the of β, and for the binding of which was shown to only in in their in the of and isotypes did not reveal any with the the the assumption shown to be that the structure of the would be to that of the RARγ the comparison of the and LBD suggested that a of NR LBDs was by binding of the ligand, the cognate surface required for with the transcriptional that mediate the AF-2 activation function to the chromatin and/or the transcription that this was achieved a mouse that a of the that to the core of the AF-2 activation domain The of our was supported by subsequent studies by with between and coactivator NR which showed that these latter with and binding to a on the receptor surface (Refs. 48 and and references therein). The that (AF-2 AD was required for of from and also that more to LBDs the AF-2 AD core (see initially suggested to us that the that to could the surface present in TR and RAR the binding of these This was supported by studies LBD that interact with and as as studies of between LBD and was shown that the bind in the same as the but with from the (Refs. 48 and and references therein). The observation that purified RARs did not bind to led us to and to clone the factor that the binding of RARs to in we discovered that this protein was et al. in a for proteins that could interact with also identified as the factor that the binding of RARs to Both led to the that, the of RXRs (α, β, and γ) not only the binding of RARs to but also of to thyroid hormone response and to vitamin response in Ref. these were confirmed by several that showed that cloned RXRs could for in binding of and receptors to their cognate response in Ref. the of between RXRs and receptor receptor and was also (see 24 , , and 33 for The finding that between RXRs and a of NRs were the functional the of their cognate ligands, the as to in such the RXR was a or a transcriptionally using the results of in vitro ligand binding and experiments were in Ref. studies to physiological with cultured cells and to of and in synergistic activation of expression of a of whose expression was known to both RARγ and (Refs. and references the expression of these genes was also activated by on their own but could not be activated by on their even This of RXR activity in the of a RAR ligand was to RXR or by (Refs. 24 , , and 33 and references therein). RXR were also for and but not for and RXR with other (see A molecular for RXR and in been In RXR in with can bind its ligand and in in the cellular not from the which coactivator binding, as of and are RXR can be in that bind or in cells with coactivator and The of et al. which showed how RXR could be a transcriptionally within also demonstrated that the between RXR ligands and RAR ligands results from of two NR present in a coactivator with both and of the transcriptionally RXRs can be to in been supported by our genetic which have shown that the ligand-dependent activation functions of and are for mouse (Refs. and and references and our we also found that the AF-2 is required in that control in cells that RXR AF-2 activity the binding of an agonistic ligand, the of the of this In RXRs are activated by which to RXR and RAR LBDs with 33 and references therein). been to in which suggested that it not be the ligand the ligand-dependent AF-2 of RXRs and references therein). In any that be the ligand RXRs in (e.g. would a of as it would in activation of the For instance, this would not in the same the of transcriptional activation by with transcriptional by In this we have found that the AF-2 activity of was required for the control of in cells, and that it was to be by a a acid that is in and RXR with a been found to activate RXR in These the that the RXR of could be to ligands to synergistic the transcriptional activity of the to in the of the cognate ligand of its For may and which are all in and 24 and references therein). the as to physiological RXR could be transcriptionally on their own binding to RXR response element even it been shown that RXR to response are to for and to in the This functional required the administration of of to the The of of the RAR and RXR isoforms vertebrate as as their expression in and (Refs. , 33 , and and references initially suggested to us that each isoform may for the highly of the of The results of subsequent studies in vitro led to the that these highly could highly molecular in which the of the (the and their to the expression of of target genes (33). A genetic dissection of the functions of RAR and RXRs in was obviously required as a first to the physiological of this first the as a model for in vitro that to physiological in Ref. upon and on the the cells either one of the distinct the additional presence of and in These are by a in of and in expression of of responsive genes (see Ref. and references therein). cells an model to the molecular by Moreover, as they to these cells also a to the molecular with other such as the protein A genetic of RARs and RXRs followed by of or receptors in and a using RAR isotype-specific and we showed and references that 1) were required for and whereas were required for in the presence of In all RAR and RXR could act but RXR activity was to that of its RAR 2) the expression of genes was by in which the RXR activity was also to that of its 3) AF-1 and AF-2 of and acted and to control the physiological and molecular of cells to 4) in gene functional between RAR or RXR which did not for expression of which is by a but not a RARβ in cells, was activated by a RARβ in cells, that gene can of of gene that not in upon with a general transcription factor and RARγ were constitutively in a ligand-independent in their AF-1 domain by the of (Refs. and and references therein). this was the first of activation of a binding to and by a Using the in and cells, we also demonstrated that of RARγ AF-1 was required for whereas of the same AF-1 domain in or was not for of the AF-1 domain of RARγ was required for of but in a fashion by a conserved within the LBD of and RARγ was required in but not in RARγ for the existence of between and for of its The observation that the AF-1 domain of and RARγ could also be by (Refs. and and references the of the also with the as first by us in the of the estrogen In any these studies supported the that binding of cognate ligands and of their activation receptors are to several our studies on cells also revealed that of by the was dependent on of the RARγ the that transcriptional activation and are and to a model in which this may in the of and of action (see references in Ref. The above studies supported our assumption that distinct could be in cellular and molecular by and also that their activation functions AF-1 and AF-2 could be required in these Moreover, they revealed that of the AF-1 and AF-2 could be for between and other that genetic in (the will be required to the and molecular the function of the receptors physiological from the to the of the in cells (see , , and and references we in the mouse for the RAR isotypes (α, β, and γ), the RXR isotypes (α, β, and γ), as as for the RAR isoforms of these RAR were and of these mouse revealed that RARs in as RAR (α, β, and γ) and isoforms and several of the vitamin A and to the that of been vertebrate to and functions RAR were to only a of the these receptors and we that RARs could be To this a of isotypes and or two or more isoforms to distinct were of these with the of in or because of that, the of to the by (Refs. , , and and references therein). This demonstrated that RARs distinct of and These also provided the first that was the of vitamin A which was confirmed by the that by the as an hormone (see references in Ref. and an of the that was in and in RAR which suggested that was in the of a also additional to the and also found in RAR all the exhibited by were also in RAR or , , and and references provided the first genetic of a between RAR and RXR and also the first that are the functional that in RXRs are transcriptionally within been a (see To the by AF-1 and AF-2 activities in we mouse that proteins either 1) of the A/B region that AF-1 or 2) the AF-2 core or both AF-1 and AF-2 activities in and and our Both and In genetic that functional between both activation functions were required for whereas they were for that their requirement on the nature of the (α, β, or γ), as as (α, β, or γ) in and exhibited a of that the of the showing that both RXR AF-1 and AF-2 can be to the activity of in in which a of a given RAR (α, β, or γ) was with either 1) a or 2) a or 3) a or 4) a also the exhibited by RAR (Refs. , , and and references therein). This synergism between RAR and RXR supported the that are the functional that the exhibited by these also identified the that the in a given our mouse genetic studies the that of the molecular the of the by as they have been from in vitro studies using cellular and are also to physiological the RAR to be by functional that are in given on the the RXR can be transcriptionally active, which the of the existence of physiological RXR ligands, and these ligands could be (see of this our that the highly of the which to the expression of of target genes that control the and of the and of our genetic of the physiological of RAR and RXR also revealed an functional within the members of each in all each of these members to one physiological functional may not be surprising within members a the of a given β, or isoform the this is or when a given RAR or RXR is as it been shown to be the in cells (see In is within members of the RAR or RXR a given being or in RAR or RXR it is or in whereas in is in the of given in which this is in a and references that are the functional the to these is to that can only when one of the two of the physiological is In other the activity of a given that is in the control of a given may be above a physiological when either one of the two or is but not when both are the of a given RAR or RXR be revealed only the is not which would for the observation that the of the AF-1 and AF-2 functions be revealed unless the activity of the is in (e.g. additional of the RAR additional of RXR of the of be in the in it is that the functions of the RAR and RXR isoforms would be revealed only of In any the results of our genetic studies the that the functional between receptors in the mouse are upon of one of the of physiological Clearly, the of have provided the functions of RARs and as as of other members of the NR superfamily an Ref. this the of a may be the of a the other the could be in (e.g. or of the functions of the gene or The could also the of a given an stage, of the gene functions Moreover, in it very to from functions of a gene to such as RARs and that are in highly In these may the of the function of a given gene in a a given of the This is obviously the for RARs and To these we have a for in the which is based on the of genes and expression of a a by with a LBD of which but not and references therein). This approach been used for in of the of and of both and whose are in This allowed us to the in the Using the same approach to the function of RXR in of the we have shown that are to and of is no that the of mouse the chimeric in (the together with mouse NRs will mouse to the that to the the et and the as as to the and the in for their to the et et for is not a to the and with the for their and to also colleagues the et et and for and our and for their activation function; binding DNA binding estrogen glucocorticoid ligand binding ligand binding mouse mammary tumor receptor protein retinoic retinoic acid response steroid receptor transcriptional intermediary thyroid hormone vitamin A vitamin