Emerging roles of BET proteins in transcription and co-transcriptional RNA processing

Transcription by RNA polymerase II (Pol II) gives rise to all nuclear protein-coding and a large set of non-coding RNAs, and is strictly regulated and coordinated with RNA processing. Bromodomain and extraterminal (BET) family proteins including BRD2, BRD3, and BRD4 have been implicated in the regulation of Pol II transcription in mammalian cells. However, only recent technological advances have allowed the analysis of direct functions of individual BET proteins with high precision in cells. These studies shed new light on the molecular mechanisms of transcription control by BET proteins challenging previous longstanding views. The most studied BET protein, BRD4, emerges as a master regulator of transcription elongation with roles also in coupling nascent transcription with RNA processing. In contrast, BRD2 is globally required for the formation of transcriptional boundaries to restrict enhancer activity to nearby genes. Although these recent findings suggest non-redundant functions of BRD4 and BRD2 in Pol II transcription, more research is needed for further clarification. Little is known about the roles of BRD3. Here, we illuminate experimental work that has initially linked BET proteins to Pol II transcription in mammalian cells, outline main meth-odological breakthroughs that have strongly advanced the understanding of BET protein functions, and discuss emerging roles of individual BET proteins in transcription and transcription-coupled RNA processing. Finally, we propose an updated model for the function of BRD4 in transcription and co-transcriptional RNA maturation

BET proteins, and particularly BRD4 and BRD2, have been implicated in RNA polymerase II (Pol II) transcription S. C. Hsu & Blobel, 2017). Interestingly, despite their strong sequence and structural similarities, recent studies suggest non-redundant functions of BRD4 and BRD2 in transcription control (Arnold et al., 2021;K. L. Cheung et al., 2017;M. T. Werner et al., 2020;Zheng et al., 2021). Recent work also indicates an implication of BRD4 in transcription-coupled RNA maturation (Arnold et al., 2021;Uppal et al., 2019). This review focuses on the functions of BET proteins in the regulation of Pol II transcription and co-transcriptional RNA processing in mammalian cells. We describe the pioneer studies that have initially linked BET proteins to Pol II transcription and transcription-coupled RNA maturation, and outline the experimental breakthroughs that uncovered the underlying molecular mechanisms indicating key functions of BET proteins in transcription and RNA processing. For roles of BET proteins in other cellular processes and their implication in disease mechanisms, we direct the reader to the following excellent reviews (Cribbs et al., 2021;García-Gutiérrez & García-Domínguez, 2021;S. C. Hsu & Blobel, 2017;Nicholas et al., 2017;Shi & Vakoc, 2014;Taniguchi, 2016;Valent & Zuber, 2014;C.-Y. Wang & Filippakopoulos, 2015 Transcription by Pol II is generally subdivided into three distinct phases: initiation, elongation, and termination ( Figure  3; Muniz et al., 2021;Svejstrup, 2004;Wissink et al., 2019). At the transcription start site (TSS) Pol II begins with the production of the nascent RNA. During the elongation phase the nascent RNA is synthesized by a productive transcription elongation complex (Figure 3; A. C. M. Cheung & Cramer, 2012;Kireeva et al., 2000). This active elongation complex includes elongation factors that directly bind to Pol II including the DRB Sensitivity-Inducing Factor (DSIF, composed of SPT4 and SPT5), the Polymerase Associated Factor 1 (PAF1) complex, and SPT6 (Vos et al., 2018). Transcription elongation is coupled with co-transcriptional RNA maturation including splicing and 3 0 -RNA processing (Cust odio & Carmo-Fonseca, 2016;Giono & Kornblihtt, 2020;Lenasi & Barboric, 2013;Neugebauer, 2019;Peck et al., 2019;Saldi et al., 2016;Tellier et al., 2020). Thereby, the phosphorylated C-terminal repeat domain (CTD) of transcribing Pol II can serve as a landing pad for RNA processing factors (Figure 3; Buratowski, 2009;Eick & Geyer, 2013;Harlen & Churchman, 2017;Saldi et al., 2016). RNA cleavage at the 3 0 -end of genes creates an entry site for the torpedo termination factor XRN2 (Eaton et al., 2018), leading to transcription termination at the termination zone (Fong et al., 2015;Schwalb et al., 2016).
The established view that transcription initiation was the major regulatory step in RNA synthesis was challenged by the discovery of widespread transcriptional pausing in the promoter-proximal region of genes (Margaritis & Holstege, 2008;Mayer et al., 2017) and throughout the gene-body (Gajos et al., 2021). Several factors have been implicated in promoter-proximal pausing (Core & Adelman, 2019;Dollinger & Gilmour, 2021;Gaertner & Zeitlinger, 2014;Noe Gonzalez et al., 2021). Whereas the Negative Elongation Factor (NELF) and DSIF are involved in the establishment of pausing, P-TEFb (CDK9 and cyclinT1) is required for pause release (Figure 3). The understanding of the causes and consequences of post-initiation transcription control remains incomplete.

| Uncovering first links between BRD4 and Pol II transcription
In this section, we outline the first studies that have connected BRD4 to Pol II transcription in mammalian cells. We describe the model systems applied and the type of experiments performed that led to initial models of how BRD4 functions in transcription control.
First evidence that BRD4 is implicated in the regulation of Pol II transcription came from the initial purification of the mammalian Mediator complex (Jiang et al., 1998). The Mediator is a coactivator providing an interface between Pol II and activators or repressors (Harper & Taatjes, 2018;Jeronimo & Robert, 2017;Soutourina, 2018). Using mouse Sp2/0 hybridoma cells for the purification trials a protein of unknown function, called the RING3-like protein, consistently co-purified with Mediator subunits (Jiang et al., 1998). The authors of this work indeed considered the RING3-like protein as a Mediator subunit (Jiang et al., 1998). Later, it could be shown that the RING3-like protein was most likely BRD4 (Houzelstein et al., 2002).
An important line of evidence for a role of BRD4 in Pol II transcription came from two studies that uncovered a link between BRD4 and P-TEFb in vitro and in human cells Yang et al., 2005). Both studies used epitopetagged versions of BRD4 ectopically expressed from vectors in HeLa cells to show that the P-TEFb subunits CDK9 and cyclinT1 co-purify with BRD4, indicative of an interaction. In the study by Jang et al. (2005), tagged BRD4 was purified from extracts obtained from the nucleus and the cytoplasm. A co-purification of P-TEFb with BRD4 was mainly detected for the nuclear extracts suggesting that BRD4 interacts with P-TEFb in the nucleus of mammalian cells .
Both studies provided first evidence that BRD4 is actively involved in Pol II transcription. Jang et al. used a luciferase reporter assay with the HIV-1 long terminal repeat (LTR) promoter known to be stimulated by P-TEFb and the cellular promoters of the c-MYC and c-JUN genes to assess the role of BRD4 in transcription. Transfection of cells with BRD4 increased reporter gene expression in a dose-dependent manner suggesting that BRD4 positively regulates transcription from these promoters . Yang et al. performed in vitro transcription assays with nuclear extracts obtained from HeLa cells that were simultaneously depleted of BRD4 and P-TEFb. Only the addition of both factors could fully restore transcription providing evidence for a role of BRD4 in transcription that is dependent on P-F I G U R E 3 Schematic view of the main stages of the Pol II transcription cycle. Selected key proteins of individual stages are shown. Co-transcriptional RNA processing-5 0 -capping (green circle), splicing, RNA cleavage, and polyadenylation-is indicated. Components of the spliceosome that assembles on the nascent RNA are symbolized by silver bullets. Pol II CTD phosphorylation of serine residues at positions 2 and 5 within the heptapeptide repeat are shown as purple and orange circles, respectively. Although RNA processing factors also bind to the Pol II CTD, main co-transcriptional RNA processing steps are shown below the DNA line for clarity. ADP, adenosine diphosphate; ATP, adenosine triphosphate; pA, poly-adenylation site; TSS, transcription start site TEFb (Yang et al., 2005). Furthermore, chromatin immunoprecipitation (ChIP) of P-TEFb coupled with real-time PCR upon a >48 h RNAi-mediated BRD4 knockdown at the HIV1-LTR reporter gene revealed a reduction in the P-TEFb occupancy at the reporter locus . Given the long depletion times of BRD4 it remained unclear whether the reduction of P-TEFb at the chromatin was a direct consequence of BRD4 depletion or an indirect effect.
Based on this work Yang et al., 2005), a model was proposed according to which BRD4 recruits P-TEFb to the gene-body region to positively regulate transcription elongation ( Figure 4; Peterlin & Price, 2006). However, key aspects of this model remained speculative. These included (i) the proposed general requirement of BRD4 in Pol II transcription regulation, (ii) the role of endogenous BRD4 in the native chromatin environment under the control of the natural regulatory elements, and (iii) the role of BRD4 in transcription elongation.
A giant leap in our understanding of BET proteins in transcription regulation came with the development of BET bromodomain-specific inhibitors. The advantages of BET bromodomain inhibition over BET RNA depletion (RNA interference [RNAi]-based methods) are the higher kinetic resolution of the functional disruption and that BET inhibitors act on the protein level. RNAi-mediated BET RNA depletion usually suffers from long treatment times (>48 h) and incomplete BET protein depletion ( Figure 5a, compare two leftmost panels). The first selective small molecule inhibitors for BET bromodomains were JQ1 and I-BET (Figure 5a,b; (Filippakopoulos et al., 2010;Nicodeme et al., 2010). Both inhibitors mimic acetylated lysine residues such as of histones and bind with high affinity and selectivity to the hydrophobic binding pocket of BET bromodomains (Filippakopoulos et al., 2010;Nicodeme et al., 2010). Both inhibitors displaced BET bromodomains with high efficacy from the chromatin (Filippakopoulos et al., 2010;Nicodeme et al., 2010).
Originally developed for therapy discovery of hematologic malignancies (Xu & Vakoc, 2017), both inhibitors have proven as powerful tools for the analysis of the mechanisms of transcription control by BET proteins. Using multiple myeloma cells (MM1.S) as a model, treatment with 500 nM JQ1 for 6 h led to a genome-wide reduction in the occupancy of BRD4, the Mediator complex, and P-TEFb particularly at super-enhancers (Lovén et al., 2013). This study also found that 50% of active genes showed a reduction in the density of elongating Pol II suggesting an elongation defect (Lovén et al., 2013). Additional studies quantified the impact of BET bromodomain inhibition on Pol II transcription in murine heart tissue, human MOLT4 cells, and for selected genes in mouse G1E-ER4 cells using the pausing index metric (Anand et al., 2013;Bartman et al., 2019;Winter et al., 2017). This metric compares the Pol II density in the promoter and the gene-body region (Wade & Struhl, 2008;Zeitlinger et al., 2007). Notably, JQ1 treatment led to a significant and global increase in the pausing index suggesting an elongation defect (Anand et al., 2013;Bartman et al., 2019;Winter et al., 2017). BET bromodomain inhibition with the inhibitor I-BET151 impaired the recruitment of BRD4, P-TEFb (CDK9), and PAF1 to the TSS of selected genes in human HL60 leukemia cells and uncovered F I G U R E 4 Classic view of BRD4 function in Pol II transcription. BRD4 recruits P-TEFb to target genes to allow phosphorylation of key components of the Pol II transcription machinery and elongation activation. Gray arrows point to main targets of P-TEFb-mediated phosphorylation. The model was adapted from Peterlin and Price (2006) interactions between BRD4 and the super elongation complex (SEC; Dawson et al., 2011). The inhibitor displaced BET proteins primarily downstream of the TSS correlating with increased Pol II promoter-proximal pausing in human K562 and MV4;11 cells (Khoueiry et al., 2019). These studies together provided evidence for a general role of BET proteins in transcription elongation.

| Link between BRD4 and P-TEFb
Several studies shed new light on the link between BRD4 and P-TEFb. One study could narrow down the region of BRD4 required for the interaction with P-TEFb to the C-terminal 34 amino acids of BRD4 (Bisgrove et al., 2007). Another study uncovered a new interaction between BRD4 and P-TEFb (Schröder et al., 2012). BD2 of BRD4 directly binds to tri-acetylated cyclin T1. Evidence was also provided that the CTD of BRD4 liberates and relieves P-TEFb from the inhibitory HEXIM1 bound complex and that BRD4 stimulates the kinase activity of P-TEFb to phosphorylate the CTD of Pol II (Itzen et al., 2014;Schröder et al., 2012). Two studies provided evidence that BRD4 is involved in the recruitment of P-TEFb to selected stimulus-responsive genes upon stimuli application (Hargreaves et al., 2009;Patel et al., 2013). 2.3.2 | Links between BRD4 and NSD3, JMJD6, and SWI/SNF Evidence was provided that BRD4 directly interacts with several chromatin regulators. It was shown that BRD4 binds to JMJD6 and NSD3 through the ET domain (Figure 2b; Crowe et al., 2016;Konuma et al., 2017;Liu et al., 2013;Rahman et al., 2011;Q. Zhang et al., 2016). ChIP experiments upon 72 h siRNA-mediated depletion of BRD4 ( Figure  5a, leftmost panel) in human C33A cells reduced the recruitment of NSD3 and JMJD6 to selected target genes (Rahman et al., 2011). NSD3 is a histone methyltransferase known to catalyze H3K36 methylation (Jeong et al., 2021;Li et al., 2009;Rayasam et al., 2003), a mark of active transcription elongation (Wagner & Carpenter, 2012). Knockdown of BRD4 led to a decrease of H3K36me3 particularly at the gene-body region of selected target genes (Rahman et al., 2011). JMJD6 is a nuclear protein that can demethylate histones (Oh et al., 2019). The interaction of BRD4 with JMJD6 and P-TEFb leads to the activation of P-TEFb and release of paused Pol II into productive elongation in HEK293T cells (Liu et al., 2013). Furthermore, it was shown that the ET domain and bromodomains of BRD4 directly bind to BRG1, a catalytic subunit of the SWI/SNF chromatin remodeling complex (Conrad et al., 2017). BRD4 colocalizes with the SWI/SNF complex at the genome and is implicated in the recruitment of SWI/SNF to the chromatin (Conrad et al., 2017;Flynn et al., 2016;Liu et al., 2014;Shi et al., 2013).

| Link between BRD4 and Mediator
Several studies revealed an interaction between BRD4 and the Mediator complex (Arnold et al., 2021;Donner et al., 2010;Jang et al., 2005;Jiang et al., 1998;Winter et al., 2017). ChIP-seq studies of BRD4 and Mediator subunits showed a genome-wide co-localization at enhancers and super-enhancers (SE; Bhagwat et al., 2016;Di Micco et al., 2014;Lovén et al., 2013). BET inhibition with JQ1 perturbed the occupancy of BRD4 and the Mediator at these sites and led to a reduction in target gene expression (Bhagwat et al., 2016;Di Micco et al., 2014;Lovén et al., 2013). Despite the positive role of BRD4 and Mediator in gene expression of SE-mediated genes a study also provided evidence for kinase subunits of the Mediator (CDK8 and CDK19) to function against BRD4 and to repress SE-associated genes in AML cells (Pelish et al., 2015). Further studies are required to clarify the functional interactions between BRD4 and Mediator.

| Additional links between BRD4 and Pol II
Two recent studies provided evidence that BRD4 directly binds the Pol II CTD in vitro and may act as an atypical kinase that phosphorylates serine 2 residues of the Pol II CTD (Devaiah et al., 2012;Weissman et al., 2021). BRD4-dependent Pol II CTD phosphorylation stimulated the activity of the DNA topoisomerase 1 (TOP1) facilitating pause release of Pol II and elongation by enhancing DNA relaxation especially in a zone 1.5 kb downstream of the Pol II pausing site (Baranello et al., 2016).

| Limitations of BET bromodomain inhibition
Although the wide application of BET bromodomain-selective inhibitors led to an advance in the understanding of BET protein functions in Pol II-mediated RNA synthesis key aspects remained unclear due to inherent limitations of BET bromodomain inhibition. First, inhibitors target bromodomains of all BET proteins in the cell, leaving the function of individual BET proteins unclear. Second, BET inhibitors exclusively target the bromodomains of BET proteins whereas the other domains such as the ET domain remain intact. Third, BET bromodomain inhibition leads only to a partial dissociation of BET proteins from the chromatin (Arnold et al., 2021). The latter two limitations cause incomplete phenotypes leaving cellular BET functions unclear. In the next section, we describe approaches that have been recently developed to overcome some of these limitations (Figure 5a) providing new insights into the roles of BET proteins in RNA synthesis and processing.

| New insights from BET proteolysis targeted chimeras (PROTACs)
In 2015, chemical biology methods became available that allow ligand-induced BET degradation (Lu et al., 2015;Winter et al., 2015;Zengerle et al., 2015). These methods rely on PROTACs that specifically target and degrade BET proteins in cells. BET PROTACs bind with high affinity and selectively to both BET bromodomains and an ubiquitously expressed E3 ubiquitin ligase (Figure 5a, third panel). This PROTAC-induced physical proximity leads to rapid BET degradation via the cellular proteasomal degradation pathway (Figure 5a; Lu et al., 2015;Winter et al., 2015;Zengerle et al., 2015). A main advantage of BET degradation over BET bromodomain inhibition is its holistic nature, rapidly eliminating all BET functions.
In the meantime, an improved BET PROTAC, called dBET6 ( Figure 5a, third panel, and 5c), was created with improved potency and efficacy in degrading BET proteins as compared to previous BET PROTACs (Winter et al., 2017). Treatment of human MOLT4 cells with dBET6 resulted in an almost complete elimination of BET proteins within less than 2 h (Winter et al., 2017). Direct consequences of acute BET protein loss on nascent Pol II transcription were revealed by native elongating transcript sequencing (NET-seq; (Winter et al., 2017)), an approach that captures transcriptionally engaged Pol II with single-nucleotide precision genome-wide (Churchman & Weissman, 2011;Mayer et al., 2015;Nojima et al., 2015). The combined application of rapid target protein degradation with readouts that have a similarly high kinetic resolution allows the analysis of the immediate consequences before the phenotype is complicated by indirect effects (Jaeger & Winter, 2021). Following this rationale, the study found that acute loss of BET proteins provoked a global collapse of productive elongation (Winter et al., 2017). BET degradation phenocopied the impact of P-TEFb (CDK9) inhibition but unexpectedly was inconsequential for the genome-wide localization of P-TEFb (Winter et al., 2017). This finding suggests BET protein-independent P-TEFb recruitment mechanisms questioning the classic model of BRD4 function (Figure 4).

| Rapid and selective degradation of BRD4 in cells
Although BET PROTACs provided new insights into the mechanisms of Pol II transcription regulation by BET proteins, BRD4-selective functions remained unclear. Several recent studies have addressed this limitation and shed new light on the BRD4-specific roles (Arnold et al., 2021;Muhar et al., 2018;Zheng et al., 2021). These studies used CRISPR-Cas9 gene editing to generate homozygous human knock-in cell lines that express degron-tagged forms of BRD4 from its endogenous locus. These systems allowed rapid and selective BRD4 degradation in cells upon ligand treatment. In two studies the auxin-inducible degron (AID) tag (Nishimura et al., 2009) was inserted into the BRD4 locus of human K562 and DLD-1 cells (Muhar et al., 2018;Zheng et al., 2021). Treatment with IAA (indole-3-acetic acid) led to an acute loss of BRD4. ChIP-seq experiments of total Pol II and different Pol II phospho-isoforms revealed an increase of Pol II density in the 5 0 -gene region and to a decrease over the gene-body of genes suggesting a defect in Pol II pause-release (Muhar et al., 2018;Zheng et al., 2021). This observation was consistent with the prior finding that pan-BET degradation prompted a collapse of elongation (Winter et al., 2017).
A recent study used the degradation tag (dTAG) system (Nabet et al., 2018) and CRISPR-Cas9 to generate human homozygous K562 knock-in cells that express a dTAGged form of BRD4 from its endogenous locus and under the control of the natural promoter (Figure 5a, fourth panel, and 5d; Arnold et al., 2021). Treatment with the degrader (dTAG-7) led to an almost complete elimination of long and short BRD4 protein isoforms (BRD4-L and BRD4-S, Figure 1) that are expressed in human K562 cells within less than 2 h (Arnold et al., 2021). Application of spike-in NET-seq upon acute loss of BRD4 revealed an immediate accumulation of transcriptionally engaged Pol II in the promoter-proximal region accompanied by a uniform decrease of transcribing Pol II over the gene-body at a large set of genes indicative of a global defect in pause release of Pol II (Arnold et al., 2021). This functional multiomics study also provided additional insights into the underlying molecular mechanism. The study found that acute BRD4 loss globally impaired the recruitment of the PAF complex at the promoter-proximal region of genes (Arnold et al., 2021).
Notably, all three genome-wide studies consistently found that the localization of P-TEFb (CDK9) was not impaired upon BRD4-selective elimination (Arnold et al., 2021;Muhar et al., 2018;Zheng et al., 2021) which was in line with prior observations upon pan-BET degradation (Winter et al., 2017). Together, these findings converge on a model of Pol II transcription control in which BRD4 underlies a general 5 0 -elongation checkpoint to help assemble a functional Pol II elongation complex that is capable of nascent RNA synthesis ( Figure 6).

| Roles of BRD4 in enhancer function and transcriptional condensate formation
Genes are not only under the control of promoter and promoter-proximal elements but also of distal regulatory sites such as enhancers (Andersson & Sandelin, 2020;Zabidi & Stark, 2016). Enhancers were defined as control elements that can increase transcription of their target genes irrespective of their orientation and their physical distance (Banerji et al., 1981;Benoist & Chambon, 1981;Gruss et al., 1981). Due to the sometimes large distance between enhancers and their target genes mechanisms are required that bring enhancers into spatial proximity to their targets (Robson et al., 2019). According to a conventional model sequence-specific TFs bind to enhancers and recruit co-activators such as the Mediator complex which in turn interact with Pol II at gene promoters through DNA looping (Furlong & Levine, 2018;Schoenfelder & Fraser, 2019). Initially, it was thought that TFs persist at enhancers for long periods due to stable DNA binding (Steger & Workman, 1997;M. H. Werner & Burley, 1997). The emerging view is that TFs bind rather dynamically to their genomic target sites and that their occupancy is sustained by condensates of TFs (Lim & Levine, 2021). The activation domain of TFs often possesses intrinsically disordered regions (IDRs) that were shown to F I G U R E 6 Emerging view of BRD4 function in Pol II transcription and co-transcriptional RNA processing. The model includes the concepts of transcriptional condensates and super-enhancers. Sequence-specific transcription factors that bind to enhancers are not shown. Although elongation and RNA processing factors can bind the Pol II CTD, the CTD is not shown for clarity. For the same reason, main cotranscriptional RNA processing steps are shown below the DNA line (as in Figure 3) although elongation and 3 0 -RNA processing factors can bind the nascent RNA. Components of the spliceosome that assemble on the nascent RNA are symbolized by silver bullets. Although the binding of NELF and PAF to Pol II is mutually exclusive in vitro (Vos et al., 2018), we show both factors to indicate that NELF and PAF are present in the promoter-proximal region of genes in cells and interact with BRD4 (Arnold et al., 2021;Chen et al., 2015;Dawson et al., 2011;Hou et al., 2019;Lambert et al., 2019;Rahl et al., 2010;Yu et al., 2015). CPSF, cleavage and polyadenylation specificity factor; CstF, cleavage stimulation factor; GTFs: general transcription factors; MED, Mediator complex induce liquid-liquid phase separation in vitro and in cells Chong et al., 2018). These enhancer-bound TF condensates associate with Mediator and Pol II at promoters to form transcriptional hubs that control target gene expression ( Figure 6; Hnisz et al., 2017;Lim & Levine, 2021). Evidence is accumulating that BRD4 plays an integral role in enhancer regulation and the formation of functional Pol II transcriptional condensates.

| Role of BRD4 in enhancer function
BRD4 binds to enhancers through direct interactions of its tandem BDs with acetylated histones (E. Hsu et al., 2021) and with enhancer RNAs (eRNAs; Rahnamoun et al., 2018). Enhancer binding by BRD4 is required for eRNA synthesis (Kanno et al., 2014;J.-E. Lee et al., 2017;Nagarajan et al., 2014) and can control target gene transcription in mature and differentiating cells also at the level of transcription elongation (Brown et al., 2018;E. Hsu et al., 2021;J.-E. Lee et al., 2017;Liu et al., 2013;W. Zhang et al., 2012). BRD4 also associates with super-enhancers (SEs; Figure 6). SEs are clusters of enhancers that are occupied by higher levels of components of the Pol II transcriptional apparatus including BRD4 and MED1 as compared to conventional enhancers, and predominantly regulate target genes with a role in cell identity Whyte et al., 2013). Transcription at SE-associated genes is more dependent on BRD4 as compared to genes that are not under the control of SEs (Di Micco et al., 2014). Individual enhancers of SEs show a higher interaction frequency with one another as with their target genes indicating spatial proximity (Dowen et al., 2014).

| Role of BRD4 in transcriptional condensate formation
A recent study showed that BRD4 can form puncta in live murine embryonic stem cells . Image analyses detected 1034 ± 130 BRD4 puncta per nucleus . These BRD4 puncta possess properties of phase-separated condensates and can associate with SEs . BRD4 contains large IDRs that can form phase-separated droplets in vitro and in cells Shin et al., 2018). BRD4 can co-condensate with other IDR-containing proteins including MED1, the histone-acetyltransferase p300, sequence-specific TFs, eRNA, and Pol II in vitro and in cells (J.-H. Ma et al., 2021;Sabari et al., 2018). Loss of BRD4 condensates adversely affects expression of target genes . Live-cell imaging analyses showed that Mediator and Pol II can form clusters at enhancers and promoters with condensate-like properties in live murine embryonic stem cells (Cho et al., 2018). Both clusters exhibited rapid dynamics and turnover of components, and co-localized to form transcriptional condensates (Cho et al., 2018). Interestingly, treatment with the BET inhibitor JQ1 dissolved Mediator and Pol II clusters indicating a fundamental role of BET proteins in condensate formation (Cho et al., 2018).
Several recent studies provided additional insights into the molecular mechanisms of Pol II transcriptional condensate formation by BRD4. One study found that BRD4 binding to acetylated chromatin can induce liquid-liquid phase separation in vitro creating two phases one of which is enriched with acetylated chromatin and the other with nonacetylated chromatin (Gibson et al., 2019). Furthermore, it was shown that the short BRD4 isoform (BRD4-S(a)) can also form puncta with condensate-like properties in human cancer cell lines . BRD4-S(a) puncta incorporated BRD4-L, MED1, P-TEFb (CDK9) and Pol II from transcription-competent nuclear extracts in vitro . Condensate formation was driven by the IDRs of BRD4-S and through binding of the bromodomains to acetylated chromatin .

| Role of BRD4 in higher-order genome organization
Recently, evidence has emerged that BRD4 is involved in higher-order chromatin folding (Linares-Saldana et al., 2021;R. Wang et al., 2012) with implications for transcription control. Mechanistically, genome topology regulation can be mediated through direct interactions of BRD4 with NIPBL (Linares-Saldana et al., 2021;Luna-Pel aez et al., 2019;Olley et al., 2018), which is a positive effector of cohesin that facilitates binding of cohesin to the chromatin (D. Gao et al., 2019). Another study found that the short isoform of BRD4 (BRD4-S(b)) can recruit the condensin II chromatin remodeling complex to acetylated chromatin to induce chromatin compaction in human U2OS cells (Floyd et al., 2013). Together, these findings suggest a role of BRD4 in enhancer-promoter interactions through liquid-liquid phase separation and changes in the 3D genome organization although a recent study provided evidence that BET proteins are dispensable for the maintenance of enhancer-promoter contacts (Crump et al., 2021). To clarify the role of BRD4 in enhancer-target gene communication is an interesting subject for future investigations.

| Implications of BRD4 in co-transcriptional RNA processing and other transcription-coupled processes
Pol II transcription is coupled and timely coordinated with co-transcriptional RNA maturation including splicing and 3'-RNA processing (RNA cleavage and polyadenylation;Cust odio & Carmo-Fonseca, 2016;Giono & Kornblihtt, 2020;Lenasi & Barboric, 2013;Mayer et al., 2017;Neugebauer, 2019;Saldi et al., 2016;Tellier et al., 2020). Evidence has emerged that BET proteins and particularly BRD4 are implicated in the coupling of nascent Pol II transcription and RNA processing. A recent study used a functional multi-omics approach to provide several lines of evidence for a direct role of BRD4 in coupling nascent transcription and 3 0 -RNA processing (Arnold et al., 2021). First, BRD4 interacted with subunits of the CPSF and CstF complexes (Arnold et al., 2021), two integral components of the general 3 0 -RNA processing machinery (Kumar et al., 2019;Sun et al., 2020). Second, BRD4 co-localized with 3 0 -RNA processing factors at the promoter-proximal regions of genes genome-wide suggesting an interaction in a native chromatin environment. This finding is in line with previous observations that 3 0 -RNA processing factors are already present at the 5 0 -end of genes (Dantonel et al., 1997;Davidson et al., 2014;Glover-Cutter et al., 2008;Kamieniarz-Gdula et al., 2019). Third, acute BRD4-selective degradation led to an immediate dissociation of 3 0 -RNA processing factors from the chromatin as revealed by chromatin mass-spectrometry. Fourth, acute loss of BRD4 provoked a genome-wide reduction in the occupancy of CPSF and CstF. Notably, the reduction in the occupancy levels was stronger for subunits of the CPSF and CstF complexes in the promoter-proximal region of genes as compared to the overall reduction of the Pol II occupancy. BRD4 degradation also led to a reduction of CPSF and CstF across the body of genes. This finding together with ChIP data, showing that CPSF and CstF crosslink throughout the gene-body region, indicate that the 3 0 -RNA processing machinery is present during elongation ( Figure 6; Arnold et al., 2021;Davidson et al., 2014;Glover-Cutter et al., 2008). Finally, acute BRD4 loss prompted RNA cleavage defects and massive readthrough transcription at a large set of genes (Arnold et al., 2021) which is consistent with the view that 3 0 -RNA cleavage is a requirement for transcription termination (Eaton & West, 2020). These data suggest that BRD4 helps to recruit 3 0 -RNA processing factors during a general 5 0elongation checkpoint, which may prime the Pol II elongation complex for 3 0 -RNA processing and transcription termination at the 3 0 -end of genes ( Figure 6).
Recent work also suggested an implication of BRD4 in RNA splicing. Two studies found that BRD4 can interact with splicing factors in human K562 and T-cell acute lymphoblastic leukemia (T-ALL, MOLT4) cells (Arnold et al., 2021;Uppal et al., 2019). Another study found that BET bromodomain inhibition in castration-resistant cancer cells (CRPCs) impacted alternative splicing of the androgen-receptor (AR) leading to decreased expression of AR and reduced growth of patient-derived CRPCs (Welti et al., 2018). Finally, it was found that BRD4 depletion provoked splicing inhibition during heat shock in human lung fibroblasts (WI-38 cells) leading to a reduction of mRNA levels of the affected transcripts (Hussong et al., 2017). Although these observations suggest a potential role of BRD4 in splicing, more work is required to clarify its direct implication.
Recently, evidence was provided that BRD4 prevents the transcription-coupled accumulation of R-loops caused during Pol II transcription by re-annealing of the nascent RNA with the DNA template strand (Edwards et al., 2020;Kim et al., 2019;Lam et al., 2020). There is also emerging evidence that BRD4 can couple transcription with metabolic processes (M. Gao et al., 2021;Olp et al., 2017;Sdelci et al., 2019).

| EMERGING ROLES OF BRD2 AND BRD3 IN POL II-MEDIATED RNA SYNTHESIS
Initial work provided evidence that BRD2 and BRD3 facilitate transcription of Pol II through acetylated nucleosomes in vitro (LeRoy et al., 2008). Genome-wide occupancy profiling studies using ChIP-seq showed that BRD2 and BRD3 were enriched at the TSS region of genes in murine and human cells (K. L. Cheung et al., 2017;Daneshvar et al., 2020;Lamonica et al., 2011;Zheng et al., 2021) and localized to enhancers genome-wide (K. L. Cheung et al., 2017;Daneshvar et al., 2020;Zheng et al., 2021). Comparative occupancy profiling revealed that the genomic binding patterns of BRD2 and BRD4 strongly differ in murine Th17 cells suggesting non-redundant functions in RNA synthesis (K. L. Cheung et al., 2017). However, a recent study found that the genomic occupancy profiles of BRD2, BRD3, and BRD4 were similar in human DLD-1 cells (Zheng et al., 2021). More work is needed to clarify the reasons for the observed differences between studies.
A comparison of acute BRD4-selective and pan-BET degradation revealed interesting differences on the PoI ll density in the promoter-proximal region of genes (Arnold et al., 2021). Whereas an acute BRD4-specific loss induced an immediate accumulation of transcriptionally engaged Pol II, acute pan-BET degradation led to a reduction (Arnold et al., 2021). This finding implies a potential role of BRD2 and/or BRD3 in Pol II transcription elongation which is distinct from BRD4. Consistent with this view, a study identified a new domain downstream of the ET domain in BRD2 and BRD3, termed the CC domain, that is absent in BRD4 (M. T. Werner et al., 2020). The CC domain binds to elongation factors such as the PAF complex (M. T. Werner et al., 2020).

| BRD2-selective roles
Acute BRD2 degradation led to a genome-wide loss of Pol II at enhancers in human DLD-1 and HCT-116 cells and to an increased Pol II occupancy at a set of genes (Zheng et al., 2021). BRD2 but not BRD4 co-localized and interacted with CTCF genome-wide in murine cells forming a transcriptional boundary that restricted enhancer activity to nearby Pol II-transcribed genes (K. L. Cheung et al., 2017;. These findings suggest that BRD2 plays a general role in transcription control through regulating the higher-order chromatin architecture.

| BRD3-selective roles
BRD3 binds through BD1 to the acetylated form of the hematopoietic transcription factor GATA1, required for erythroid differentiation, to control the transcription of target genes . A recent study found that BRD3 can form liquid-liquid phase-separated condensates in vitro and in cells (Daneshvar et al., 2020). Condensate formation was promoted by the interaction of BRD3 with the long non-coding RNA DIGIT in human embryonic stem cells, which both occupy enhancers during endoderm differentiation (Daneshvar et al., 2020). However, acute BRD3 depletion in mature DLD-1 cells had only a minimal effect on the genomic Pol II density (Zheng et al., 2021). Together, it is unclear whether BRD3 plays a general role in RNA synthesis or whether it serves specific functions in cell differentiation.

| CONCLUSION AND FUTURE PERSPECTIVES
Previous functional studies of BET proteins have largely relied on perturbation methods that simultaneously disrupt functions of all BET proteins, or that require prolonged perturbation times (Figure 5a, RNAi, and BET inhibition). As a consequence, direct functions of individual BET proteins in transcription and transcription-coupled RNA maturation have largely remained unclear. An important future direction is to disentangle the roles of individual BET proteins. Several recent studies have made substantial progress toward this goal by developing and applying methods to rapidly eliminate specific BET proteins in cells as well as methods that capture the direct consequences on nascent transcription with high precision. Notably, findings of the new studies argue against a general implication of BRD4 in the recruitment of P-TEFb to sites of Pol II transcription challenging longstanding views. The data rather supports a model according to which BRD4 is required for the assembly of a functional Pol II elongation complex during a general 5 0elongation checkpoint to enable processive RNA synthesis and transcription-coupled RNA processing genome-wide. We expect that these new functional multiomics approaches along with biochemistry methods will further illuminate the molecular mechanisms of gene regulation by BET proteins including their roles in the dynamics of enhancerpromoter communication, 3D genome organization, and transcriptional phase separation. Future studies are also needed to clarify BET protein functions in the temporal coordination of nascent transcription and RNA processing.
Given the strong implication of BET proteins in a broad range of human diseases, a more complete understanding of the mechanisms that underlie the regulation of transcription and RNA processing by BET proteins may inspire new and more targeted therapeutic interventions in future.

ACKNOWLEDGMENTS
We apologize to our colleagues whose work was not cited due to space limitations. We thank Jana Henck, Serkan Meydaneri, Yelizaveta Mochalova, and Jelena Ulicevic for critical comments on the manuscript. Several figures of this manuscript were created using Biorender.com.

CONFLICT OF INTEREST
The authors have declared no conflicts of interest for this article.

DATA AVAILABILITY STATEMENT
Data sharing is not applicable to this article as no new data were created or analyzed in this study.