Dynamic Combinatorial Networks in Nuclear Receptor-mediated Transcription*

The nuclear hormone receptor (NR)2 superfamily describes a diverse array of transcription activatorsworking in a ligand-dependentmanner and exerting action by regulating the expression of specific subsets of genes (1, 2). The receptors for steroid hormones, which include the estrogen (ER), androgen (AR), progesterone (PR), and glucocorticoid (GR) receptors, are held in the cytoplasm, in associationwith chaperone complexes. Ligand binding acts as an on-switch, inducing their release from the chaperone molecules, their dimerization, their entrance into the nucleus, and their binding to hormone response elements (HREs) within the regulatory regions of target genes. In contrast, the receptors for non-steroidal hormones such as the vitamin D (VDR), retinoic acid (RARs), and peroxisome proliferator-activated (PPAR) receptors are found primarily in the nucleus and bound at HREs as heterodimers with the retinoid X receptor (RXR). They are associated with histone deacetylase-containing complexes tethered through corepressors (N-CoR and SMRT), resulting in chromatin compaction and gene silencing. Upon ligand binding, the corepressor-binding interface of NRs is destabilized, leading to the dissociation of corepressors (3). The basic mechanism for switching on gene transcription by liganded NRs (either steroid or non-steroid) relies on a complex and ever growing network of interactionswith coregulatory proteins. FormostNRs, this network is directed by a specific domain, the activation function 2 (AF-2) domain, located in the C-terminal ligand-binding domain (LBD) (Fig. 1A). The LBD is composed of 12 helices organized as an antiparallel sandwich (Fig. 1B). Upon ligand binding, the AF-2 domain undergoes major structural rearrangements (4, 5), creating a new binding surface for coactivators (Fig. 1B). The AF-2 domain cooperates with a second activation domain residing in the N terminus, the AF-1 domain (4), to recruit multiple complexes to alter the chromatin structure surrounding the promoter of target genes (6–8). In the end these events pave the way for the recruitment of the transcription machinery, including RNA polymerase II (RNA PolII) and the general transcription factors (GTFs) (9). A concept that has developed over the last several years is that NRs and their coregulators are subjected to rapid modifications (phosphorylation, ubiquitination, acetylation, methylation) and proteasomal degradation. Therefore it emerged that they exert transcriptional control in a combinatorial, coordinated, and sequential manner. These findings challenged the conventional view of stable, NR-based, template-bound complexes and suggested a dynamic and multistep model involving rapid and carefully orchestrated series of exchanges between NRs, coregulators, and the promoter. In this review, we will focus on the dynamics of the interactions between NRs and their coregulators and on their fine tuning by a variety of post-translational modifications or degradation processes that occur in response to the ligand or to other cellular signalings, so that, in the end, the correct proteins are present with the right activity, at the right place, and at the right time.

The nuclear hormone receptor (NR) 2 superfamily describes a diverse array of transcription activators working in a ligand-dependent manner and exerting action by regulating the expression of specific subsets of genes (1,2). The receptors for steroid hormones, which include the estrogen (ER), androgen (AR), progesterone (PR), and glucocorticoid (GR) receptors, are held in the cytoplasm, in association with chaperone complexes. Ligand binding acts as an on-switch, inducing their release from the chaperone molecules, their dimerization, their entrance into the nucleus, and their binding to hormone response elements (HREs) within the regulatory regions of target genes. In contrast, the receptors for non-steroidal hormones such as the vitamin D (VDR), retinoic acid (RARs), and peroxisome proliferator-activated (PPAR) receptors are found primarily in the nucleus and bound at HREs as heterodimers with the retinoid X receptor (RXR). They are associated with histone deacetylase-containing complexes tethered through corepressors (N-CoR and SMRT), resulting in chromatin compaction and gene silencing. Upon ligand binding, the corepressor-binding interface of NRs is destabilized, leading to the dissociation of corepressors (3).
The basic mechanism for switching on gene transcription by liganded NRs (either steroid or non-steroid) relies on a complex and ever growing network of interactions with coregulatory proteins. For most NRs, this network is directed by a specific domain, the activation function 2 (AF-2) domain, located in the C-terminal ligand-binding domain (LBD) (Fig. 1A). The LBD is composed of 12 ␣ helices organized as an antiparallel sandwich (Fig. 1B). Upon ligand binding, the AF-2 domain undergoes major structural rearrangements (4,5), creating a new binding surface for coactivators (Fig. 1B). The AF-2 domain cooperates with a second activation domain residing in the N terminus, the AF-1 domain (4), to recruit multiple complexes to alter the chromatin structure surrounding the promoter of target genes (6 -8). In the end these events pave the way for the recruitment of the transcription machinery, including RNA polymerase II (RNA PolII) and the general transcription factors (GTFs) (9).
A concept that has developed over the last several years is that NRs and their coregulators are subjected to rapid modifications (phosphorylation, ubiquitination, acetylation, methylation) and proteasomal degradation. Therefore it emerged that they exert transcriptional control in a combinatorial, coordi-nated, and sequential manner. These findings challenged the conventional view of stable, NR-based, template-bound complexes and suggested a dynamic and multistep model involving rapid and carefully orchestrated series of exchanges between NRs, coregulators, and the promoter. In this review, we will focus on the dynamics of the interactions between NRs and their coregulators and on their fine tuning by a variety of post-translational modifications or degradation processes that occur in response to the ligand or to other cellular signalings, so that, in the end, the correct proteins are present with the right activity, at the right place, and at the right time.
The orderly and sequential recruitment of these coactivators with different enzymatic activities has been recognized only recently. In chromatin immunoprecipitation (ChIP) assays, Metivier et al. (16) defined the concept of a "transcriptional clock" directing this orderly "procession." ER and the SRC family of coactivators are the first to be recruited to the estrogen-responsive pS2 promoter, followed by the HATs and HMTs, in agreement with their essential role in the construction of the complexes involved in chromatin remodeling. Then the mediator is recruited, facilitating the arrival of RNA PolII and GTFs. However, the temporal order of recruitment of these complexes may differ from one target gene to the other (17), indicating that a given NR can employ multiple programs for gene activation depending on promoter context. As an example, a recent study demonstrated that both the mediator and RNA PolII can be recruited at the promoter of the RAR␤ gene even when it is in the inactive state. Transcription is initiated by a gene-specific cofactor PARP-1, which converts the mediator to an active form through triggering the release of the Cdk8 module and the recruitment of TFIIH (18).

The Ubiquitin-Proteasome Pathway: a Central Regulator of the Dynamics through Proteolysis
After the notion of clock, further work revealed the notion of dynamism controlling the chronological sequence of the association/dissociation of NRs and their coregulators at a promoter. Indeed, once recruited, each coactivator paves the way for the next one and then dissociates rapidly. Moreover, a new aspect that has been demonstrated in studies using ChIPs and fluorescence recovery after photobleaching is that steroid receptors, together with their coactivators, cycle on and off the promoter many times, interacting only briefly with response elements (16, 19 -21). This raised the idea that cyclic interaction with transcription complexes would be a common feature of promoters activated by NRs. It has been proposed that this cyclic recruitment of NRs and their cofactors on a target promoter might be directed by cyclic changes in chromatin structure and in the acetylation/methylation/phosphorylation status of histones, NRs, and their cofactors. It might also rely on allosteric changes induced within partner proteins of the transcription apparatus during their physical interaction (19,22,23). However, recent evidence argues that this process is likely driven by the ubiquitin-proteasome pathway.
Classically, the ubiquitin-proteasome pathway targets regulatory proteins for proteolysis (24,25) and is broken up in two parts: 1) the ubiquitination machinery, which consists of an E1 ubiquitin-activating enzyme, several E2 ubiquitin-conjugating enzymes, and multiple E3 ubiquitin ligases directing ubiquitin to target proteins and 2) the 26 S proteasome itself, which comprises a 20 S core complex and a 19 S complex that recognizes the ubiquitinated proteins and prepares them for entry into the 20 S core.
Several laboratories have demonstrated that NRs, as well as coactivators, corepressors, and components of the GTF machinery are ubiquitinated and degraded by the 26 S proteasome in response to the ligand (26 -29). A radical hypothesis first suggested that the functional role of this pathway is to provide an efficient suicide mechanism for attenuation of the transcriptional signal (30,31).
* This minireview will be reprinted in the 2005 Minireview Compendium, which will be available in January, 2006. The studies mentioned in text were supported by funds from CNRS, INSERM, the Hô pital Universitaire de Strasbourg, and the Association pour la Recherche sur le Cancer. 1 To whom correspondence should be addressed. However, in a perhaps more unexpected twist on the ubiquitin-proteasome theme, it emerged recently that proteasomal degradation is inextricably linked to transcription. Then the question was how? It has been proposed that degradation by the ubiquitin-proteasome pathway might provide an efficient mechanism for regulating the cyclic interaction of NRs with the promoter (19,23). However, the most plausible mechanism would be to clear out corepressors and/or coactivators so that other coregulators can subsequently bind (29). In line with this theme, E2 ubiquitin-conjugating enzymes (UbcH7) and E3 ubiquitin ligases (E6-AP) have been shown to interact with NRs and their coactivators and to be implicated in NR-dependent transcription as well as in the degradation of the transcriptional machinery (29). Moreover, Perissi et al. (32) demonstrated that TBL1 and TBLR1, which are components of E3 ubiquitin ligase complexes, function as adaptor molecules for the recruitment of the proteasome and the degradation of the N-CoR/SMRT corepressors. Such a process could mediate the exchange of corepressor for coactivator complexes. Proteasomal degradation of NRs and their coregulators could also promote disassembly of the initiation complex, facilitating the transition to a productive elongation complex (33,34). Finally, it has been demonstrated that the 26 S proteasome associates with elongating polymerase complexes and that its proteolytic activity is important for resolving stalled complexes at termination sites (35). Altogether, these models are consistent with the idea that transcription is a dynamic process with continual exchange and turnover of NRs, coactivators, and other components of the transcriptional machinery.

New Roles for Ubiquitin and Proteasomal Subunits without Proteolysis
Today, the interconnectivity of transcription and the ubiquitin-proteasome system is clearly established, but the emerging picture is increasing in com-plexity. Indeed, recent evidence points out that the ubiquitin-proteasome system can also function in transcriptional activation independently of proteolysis (36,37). In line with this, E3 ubiquitin ligases are integral components of the coactivator complexes and of the RNA PolII transcription machinery (37) and could participate in NR-mediated transcription without signaling degradation. In the same order of ideas, evidence suggests that ubiquitination may facilitate transcription elongation by controlling the recruitment of the elongation factor P-TEF-b by transcription activators (38). Another striking example is the proteasome itself. Indeed, SUG-1, which is one of the six ATPases of the 19 S regulatory complex of the proteasome (also known as APIS), interacts with NRs, the general transcription factor TFIIH, and RNA PolII and is recruited to transcriptionally active genes (39,40). It has been suggested that the 19 S complex could facilitate both transcription and elongation by remodeling protein complexes and/or chromatin through its ATP-dependent chaperone activity (41)(42)(43). This role of the 19 S complex would be to prevent the interactions from stalling and therefore to maintain the dynamics of the complexes. Finally, an alternative mechanism has also been suggested from the observation that histones are ubiquitinated (44). In agreement with the histone code hypothesis, ubiquitination could create new platforms for binding additional chromatin modifiers and thereby influence other modifications such as acetylation and methylation (45). It is also probable that histone ubiquitination plays a structural role by recruiting the 19 S proteasome (46). From all these examples, it is clear that components of the ubiquitin-proteasome pathway could function to orchestrate the dynamics of NR-mediated transcription in several diverse ways, ranging from the regulation of chromatin to the degradation of the transcriptional complexes. However, due to the growing number of proteins with double duties in the transcription and proteasome pathways, we can expect many new regulatory strategies in the future.

MAP Kinases Signaling to NRs and Their Coregulators: a Coordination of the Dynamics
Another emerging level of transcription regulation by NRs involves phosphorylation processes (47). A number of studies demonstrated that steroid and non-steroid hormones are able to cross-talk with several signaling pathways through the rapid activation of kinase cascades. As an example, estradiol and retinoic acid signaling leads to the rapid activation of MAPKs (Erks and p38 MAPK), which can enter the nucleus and then phosphorylate the N-terminal AF-1 domains of ER␣ and RAR␥ (40,47,48) (Fig. 1A).
It is becoming increasingly clear that MAPKs play a key role in the dynamics of NR-mediated transcription (49). In particular, phosphorylation of the AF-1 domain of NRs influences the recruitment of p160 coactivators, thereby helping the reorganization of chromatin (47). It also triggers the dissociation of repressor proteins. In line with this, Bour et al. (50) demonstrated recently that phosphorylation of the AF-1 domain of the RAR␥ isotype induces the dissociation of vinexin ␤, a repressor of RAR␥-mediated transcription. Whether such a dissociation process is followed by the recruitment of other complexes, in order to achieve maximal transcription efficiency, remains to be determined.
Phosphorylation by MAPKs also signals the ubiquitination and the degradation of some NRs such as RAR␥ (40) and PR (51). As the transcription and ubiquitin-proteasome systems are intimately linked (see above), it has been proposed that phosphorylation signals recruitment of ubiquitin ligase complexes, which on the one hand could control transcription by bridging NRs to the transcription machinery and, on the other hand, could trigger NR degradation via ubiquitin-dependent proteolysis (36,37,52). It remains to be determined, however, whether the phosphorylated AF-1 domains of NRs recruit complexes with ubiquitin ligase activity as suggested by this model.
Further complexity came with the finding that corepressors and coactivators are also phosphorylated by MAPKs in response to steroidal and nonsteroidal hormones (53)(54)(55). In agreement with the transcription clock hypothesis, it has been suggested that phosphorylation of the N-CoR/SMRT corepressors might represent a mark for their ubiquitylation and degradation to allow the recruitment of coactivators (32). The p160 coactivators are also phosphorylated. The best example is SRC-3/AIB1, which becomes phosphorylated at several residues by p38 MAPK in response to estrogens (55) and retinoic acid. 3 O'Malley and colleagues (55) proposed that phosphorylation would provide the basis for a combinatorial code determining the ability of SRCs to distinguish among various NRs. However, our group demonstrated that phosphorylation also represents a mark for proteasomal degradation.
Finally, histones were also found to be phosphorylated by MAPKs (49,56). According to the histone code hypothesis, a combinatorial and coordinated dynamic model has been suggested whereby phosphorylation participates with acetylation, methylation, and ubiquitination to provide motifs for the recruitment of other chromatin modifying or remodeling complexes.
Thus, it is becoming increasingly evident that MAPK-mediated phosphorylation processes also orchestrate the dynamics of NR-mediated transcription through influencing the rapid exchange and turnover of NRs and their coregulators within transcriptional complexes.

General Transcription Factor TFIIH: Another New Central Regulator of NR Phosphorylation and Transcriptional Activity
A fascinating new concept that has developed over the last several years is that NRs are phosphorylated by the Cdk7/cyclin H/MAT1 (CAK) subcomplex of TFIIH, which is a general transcription factor composed of 10 subunits (57). This phosphorylation process has been studied extensively by our group in the case of RARs (58,59), but other NRs including ER␣ (60), AR (61), and PPARs (62), have also been reported to be phosphorylated by Cdk7. The importance of TFIIH-mediated phosphorylation has been highlighted in recent studies using cells from patients bearing mutations in the XPD subunit of TFIIH. These mutations result in incorrect positioning of the Cdk7 kinase relative to its substrate. Indeed, in these cells, NRs are hypophosphorylated, and the liganddependent response is decreased (62)(63)(64). Similarly, expression of RAR␥ mutated at its phosphorylation sites in a RAR␥-null background affected dramatically the differentiation of mouse embryocarcinoma cells (F9 cells) and activation of target genes in response to retinoic acid (65).
TFIIH is recruited to the promoter in concomitance with the transcriptional machinery and thus after the coactivators and the chromatin modifying/remodeling complexes (16,18). Binding to the promoter makes TFIIH able to dock NRs bound at their cognate HRE, through specific subunits (59,60), thereby allowing the recognition by the Cdk7 kinase subunit of the phosphorylation site located in a proline-rich motif within the N-terminal AF-1 domain of NRs (Fig. 1A). In an unexpected twist on the molecular communication between TFIIH and NRs, Drané et al. (64) recently demonstrated that in the absence of a functional AF-1 domain, as observed for VDR, a second transcriptional activator can be phosphorylated by TFIIH, upon interaction with VDR, thereby filling up the role of an AF-1 domain. However, for other NRs such as RAR␥, the serine residue phosphorylated by Cdk7 is adjacent to the MAPK phosphorylation site (Fig. 1A) so that in the end, both Cdk7 and MAPKs cooperate for the activation of target genes, highlighting a new level of combinatorial regulation through coordinated phosphorylation processes (52).
It has been suggested that NR phosphorylation by TFIIH would define a "code" specifying the recruitment of partners participating in formation of the preinitiation complex. However, one cannot exclude the possibility that it promotes disassembly of the preinitiation complex, facilitating transition to elongation, according to a model very similar to that proposed for phosphorylation of the RNA PolII CTD by the same Cdk7 kinase within TFIIH (9). Such a code might also include cis-trans isomerization of the proline residues that follow the phosphorylated serines (66). In line with this the proline isomerase pin1/Ess1 has been recently shown to bind the phosphorylated form of RAR␣ (67). It would be interesting to determine whether phosphorylation changes the equilibria between the cis and trans forms of the (Ser(P)/Thr)-Pro motifs, as each state is potentially a specific recognition state for an interacting factor (66).

NR Phosphorylation in Response to Other Cellular Signaling Pathways: a Mechanism to Switch Off the Ligand Response
Today, new roles for phosphorylation in regulation of NR-mediated transcription are being discovered at an accelerated pace. Several extracellular signals such as growth factors, insulin, stress, cytokines, or other signals, activate cytosolic serine kinase cascade pathways, ending at different kinases, including MAPKs (Erks, JNKs, p38 MAPK), protein kinase C, or cyclic AMP-dependent protein kinase, which can enter the nucleus and phosphorylate NRs. In fact, phosphorylation in response to such signalings rather inactivates NR-mediated transcription. An 3 M. Gianni and C. Rochette-Egly, submitted for publication. In step 1, the hormone activates NRs and induces their phosphorylation by MAPKs. Phosphorylation cooperates with the ligand for dissociation and proteolysis of the histone deacetylase corepressor complexes. The p160 family of coactivators is also phosphorylated and recruited as well as HAT, HMT, and E3 ubiquitin ligase complexes. Histones are phosphorylated, methylated, acetylated, and ubiquitinated, and these modifications cooperate for the recruitment of additional transcriptional complexes such as ATP-dependent nucleosome remodeling complexes. In step 2, after modifying their targets, the coactivators are ubiquitinated by the E3 ubiquitin ligases to signal their destruction by the proteasome. This process continues many times with the purpose of recruiting all the coactivators with an essential function in chromatin remodeling, leading to the displacement of impeding nucleosomes within the proximal promoter region and thus facilitating the arrival of RNA PolII and GTFs. At this step, TFIIH becomes able to interact with NRs, and the Cdk7 kinase subunit phosphorylates their AF-1 domain. This phosphorylation process participates in the assembly of the transcriptional preinitiation complex (PIC). In step 3, the preinitiation complex disassembles and RNA PolII escapes from the promoter with its own set of E3 ubiquitin ligases and chromatin remodelers, facilitating transition to elongation. Then, in step 4, NRs are degraded by the ubiquitin-proteasome system. According to this model, on-going transcription would entail continuous reloading of newly synthesized NRs. However, upon cessation of the signal, histone deacetylase-containing complexes are recruited and chromatin is recompacted.
interesting example is that of PPAR␥, the transcriptional activity of which is switched off in response to insulin, subsequent to phosphorylation by MAPKs. Several scenarios have been proposed, such as the reduction of the ligand-binding affinity through interdomain communication between the phosphorylated AF-1 domain and the ligand-binding pocket (68) or the degradation of the receptor by the ubiquitin-proteasome pathway (69). In addition, phosphorylation of several NRs in their DNA-binding domain has been shown to switch off their activity by inducing them to dissociate from the promoter (47). Finally, the activity of RARs is negatively modulated in response to stress upon phosphorylation of the LBD of their RXR heterodimerization partner (70). It has been predicted that this phosphorylation sterically precludes the formation of the surface for the recruitment of coactivators. Thus, whatever the mechanism is, phosphorylation in response to exogenous cellular signals appears to provide a mechanism to "terminate" the response to the ligand, ensuring the escape of NRs from the transcription initiation complex. Such a process would switch the cell toward other transcription events.

Conclusion
Gene induction by NRs is controlled not only by simple on/off switches induced by the ligand but also by multiple players acting in a fine-tuned, spatially and temporally coordinated manner. The central theme of this review is the extremely dynamic and combinatorial nature of the signaling events that impinge on NRs and chromatin during gene induction (Fig. 2). The dynamics and the order of the different steps is orchestrated mainly by combinatorial phosphorylation, acetylation, methylation, ubiquitylation, remodeling, and degradation events that target either NRs or their coregulators, chromatin histones, and the transcription machinery. Considering the diversity of these events, it is likely that more novel types of coregulators will be identified. Moreover, structural approaches will further our understanding of how the different coregulators interact with each other and how the enzymes recognize their substrate. As genes are typically regulated by multiple transcription activators, each with its associated set of coactivators and corepressors, which respond to different signaling cascades, it is evident that major advances in our understanding of NR-mediated transcription will be derived from studying complex promoters. The challenge of the future will be to understand how the different signaling pathways are integrated through the different activators to give a gene-specific transcriptional response.