Lysine acetylation is similar to protein phosphorylation in its prevalence as a post-translational modification and also has a large effect on the physicochemical property of the modified residue. The addition of an acetyl moiety to the side-chain nitrogen of lysine leads to neutralisation of charge, which can significantly influence protein conformation and protein–protein interactions, thus resulting in the modulation of enzyme activities and protein assembly (Ref. Reference Kouzarides1). The central role of ɛ-N-acetylation of lysine residues (K _ac) is reflected by the large number of acetylation sites that have been identified in proteins (Ref. Reference Choudhary2). Acetylation is particularly abundant in large macromolecular complexes that are present in the cell nucleus, suggesting a key role of acetylation in the regulation of chromatin and transcriptional control. In particular, the unstructured tails of histones are hotspots of acetyl lysine modification. Histone acetylation levels have been associated with an open chromatin architecture and transcriptional activation, but specific marks have also been linked to chromatin condensation (e.g. H4K16) (Refs Reference Shogren-Knaak3, Reference Kouzarides4), regulation of metabolism (Ref. Reference Guan and Xiong5) and DNA repair (Ref. Reference Celic6). Acetylation of transcription factors can either stimulate or silence gene transcription, and inappropriate acetylation levels have been associated with aberrant transcription of disease-promoting genes in cancer and inflammation, instigating the development of inhibitors for histone deacetylases (HDACs) (Ref. Reference Bertrand7) and histone acetyltransferases (HATs) (Ref. Reference Bowers8).

Recruitment of proteins to macromolecular complexes by acetylated lysine residues is mediated by bromodomains (BRDs), which are evolutionarily highly conserved protein-interaction modules that recognise ɛ-N-lysine acetylation motifs (Ref. Reference Mujtaba, Zeng and Zhou9). However, BRDs in BRD4 have recently been shown to bind propionylated and butyrylated lysine residues (Ref. Reference Vollmuth and Geyer10). BRDs are named after the Drosophila gene brahma where the BRD sequence motif was first reported (Refs Reference Tamkun11, Reference Haynes12). Since then, BRDs have been identified in a number of nuclear proteins such as HATs (Ref. Reference Nagy and Tora13), ATP-dependent chromatin-remodelling complexes (Ref. Reference Trotter and Archer14), methyltransferases (Refs Reference Malik and Bhaumik15, Reference Gregory16) and transcriptional coactivators (Refs Reference Bres, Yoh and Jones17, Reference Venturini18, Reference Jacobson19) (Table 1).

Table 1. Bromodomain-containing proteins and their functions

BRD, bromodomain; HAT, histone acetyltransferase; MOZ, monocytic leukaemia zinc finger protein; PHD, plant homology domain; SNF, sucrose nonfermenting.

Role of BRD proteins in chromatin biology

BRDs have an important role in the targeting of chromatin-modifying enzymes to specific sites. Often they act with other protein-interaction modules to guarantee a high level of targeting specificity for these essential enzymes. For example, the methyltransferase ASH1L contains a combination of one BRD and one plant homology domain (PHD), as well as a bromo-adjacent homology domain (BAH) (Ref. Reference Nakamura20). ASH1L is a member of the trithorax group of transcriptional activators. In Drosophila, ASH1L activates ultrabithorax expression, and mammalian homologues have been associated with actively transcribed genes. Another example of a multidomain methyltransferase containing a BRD is the mixed lineage leukaemia (MLL) gene product (Ref. Reference Mohan21), which is an essential gene and acts as a key regulator of the expression of many genes. MLL is required for proper segment identity in mammals, it displays haplo-insufficiency and regulates self-renewal of haematopoietic stem cells by controlling HOX (homeobox) gene expression (Refs Reference Yu22, Reference Yagi23, Reference McMahon24).

In addition, the HATs CREBBP and EP300 contain several protein-interaction modules, including one BRD, and zinc finger and KIX domains (Ref. Reference Radhakrishnan25). Both proteins share a high degree of sequence similarity and act as transcriptional coactivators that control a large variety of biological processes, including cell growth, genomic stability, development, neuronal plasticity and memory formation, as well as energy homeostasis (Ref. Reference Kalkhoven26). CREBBP is a coactivator of the cAMP response element-binding CREB transcription factor. The fundamental role of CREBBP is reflected by the severe phenotype of homozygous knockout mice, which die in utero with signs of defective blood vessel formation in the central nervous system, developmental retardation, and delays in both primitive and definitive haematopoiesis (Ref. Reference Tanaka27). Similarly, homozygous deletion of Ep300 results in mice that die between days 9 and 11.5 of gestation as a result of defects in neurulation, cell proliferation and heart development (Ref. Reference Yao28). Two additional HAT-containing BRDs have been reported and these interact with EP300 and CREBBP: PCAF [also known as K(lysine) acetyltransferase 2B (KAT2B)] and the related GCN5. Both proteins acetylate histones and transcription factors, and act as transcriptional coactivators. Gcn5-knockout mice die during embryogenesis because of severe growth retardation, failure in the development of dorsal mesoderm lineages and anterior neural tube closure (Refs Reference Bu29, Reference Xu30). By contrast, homozygous deletion of the closely related Pcaf gene does not show gross abnormalities, but leads to short-term memory deficits and an exaggerated response to acute stress and conditioned fear, associated with increased plasma corticosterone levels (Ref. Reference Maurice31).

Recent data identified evolutionarily conserved AAA ATPase ANCCA (AAA nuclear coregulator cancer-associated protein)/ATAD2 as a protein required for recruitment of transcription factors of the E2F family to their target sites, and as a transcriptional coregulator of Myc, oestrogen and androgen receptors (ARs). ATAD2 associates through its BRD with histone H3 acetylated at Lys 14 during late mitosis, regulating the expression of genes required for cell cycle progression (Refs Reference Revenko32, Reference Ciro33, Reference Zou34).

Dual BRD proteins of the BET (bromodomain and extra-terminal) family also have a pivotal role regulating the transcription of growth-promoting genes and cell cycle regulators. The BET family is represented by four members in humans (BRD2, BRD3, BRD4 and the testis-specific isoform BRDT), with each containing two N-terminal BRDs. BRD4 and BRD2 are key mediators of transcriptional elongation by recruiting the positive transcription elongation factor complex (P-TEFb). The P-TEFb core complex is composed of cyclin-dependent kinase-9 (CDK9) and its activator cyclin T. CDK9 phosphorylates the RNA polymerase II (RNAPII) C-terminal domain, a region that contains 52 heptad repeats. RNAPII undergoes sequential phosphorylation at Ser5 during promoter clearance and at Ser2 by P-TEFb at the start of elongation. It has been shown that BRD4 couples P-TEFb to acetylated chromatin through its BRDs. Interestingly, in contrast to other BRD-containing proteins and transcription factors, BET proteins remain associated with condensed and hypoacetylated mitotic chromosomes (Ref. Reference Dey35), suggesting a role in epigenetic memory (Refs Reference Kanno36, Reference Dey37). Homeostasis of BET expression levels is important for cell cycle control because both inhibition of BRD4 by microinjected specific antibodies and overexpression of BRD4 lead to cell cycle arrest in the G2M and G1S phases, respectively (Refs Reference Dey38, Reference Maruyama39), and genetic knockdown of BRD4 in cultured human cells significantly reduces cell growth (Ref. Reference Wu40). BRD2 associates with the E2F transcription factors and with the SWI/SNF (switch mating type/sucrose nonfermenting) complex to regulate the expression of diverse genes (Ref. Reference Denis41) such as cyclin D1 (CCND1) (Ref. Reference Leroy, Rickards and Flint42). BRD2 can function as a transcriptional coactivator or corepressor in a promoter-specific or tissue-specific manner. Deletion of either BRD2 or BRD4 in mice is lethal, and Brd4 ^+/– mice also show severe developmental defects (Refs Reference Houzelstein43, Reference Shang44, Reference Gyuris45). Mutagenesis of the Brd2 promoter region resulted in mice that expressed reduced levels of BRD2 without causing gross developmental abnormalities. However, these mice are extremely obese without developing glucose intolerance (Ref. Reference Wang46). The testis-specific BET family member BRDT is essential for normal spermatogenesis, and specific deletion of the first BRD in mice results in abnormal spermatids and sterility (Ref. Reference Shang47). In agreement with studies in mice, altered histone modifications have been observed in the BRDT promoter region of subfertile patients (Ref. Reference Steilmann48), and genome-wide association studies linked polymorphism in BRDT to sterility in European men (Ref. Reference Aston49).

Tandem BRDs are also present in TAF1 [RNAPII, TATA box binding protein (TBP)-associated factor, 250 kDa formerly called TAFII250], the largest subunit of the general transcription factor TFIID. TAF1 binds to the core promoter sequence encompassing the transcriptional start site, and also interacts with other transcriptional regulators, thereby modulating the rate of transcription initiation (Ref. Reference Wassarman and Sauer50). It acts as a general transcriptional activator and as such regulates a variety of essential biological processes, including myogenesis, DNA-damage response, the cell cycle and apoptosis (Refs Reference Deato and Tjian51, Reference Buchmann, Skaar and DeCaprio52, Reference Lin53, Reference Kimura54). The C-terminal tandem BRDs have been shown to specifically recognise the diacetylated histone H4 tail at K5/K12 or K8/K16, as well as diacetylated P53 at K373/K382 at the p21 promoter (Refs Reference Jacobson55, Reference Li56). TAF1L is a testis-specific homologue of TAF1. TAF1L is X-linked and might act as a functional substitute for TAF1 during male meiosis, when sex chromosomes are transcriptionally silenced. Similarly to TAF1, TAF1L can bind to the TATA-binding protein (TBP) and can functionally substitute for TAF1 in a temperature-sensitive hamster cell line (Ref. Reference Wang and Page57).

The WD repeat proteins BRWD1 (WDR9) and BRWD3 also contain tandem BRDs. Members of this family are involved in a variety of cellular processes, including cell cycle progression, signal transduction, apoptosis and gene regulation (Refs Reference Kalla58, Reference Ramos59). Mutations in mice revealed a role for BRWD1 in spermiogenesis and the oocyte–embryo transition (Ref. Reference Philipps60). Despite the specific phenotype in germ-cell maturation, BRWD1 is widely expressed, and its expression levels are dynamic during mouse development. It associates with the SWI/SNF complex component and functions as a transcriptional regulator involved in chromatin remodelling (Ref. Reference Huang61). Little is known about the biological function of BRWD3. However, in Drosophila, BRWD3 function has been genetically linked to the JAK–STAT pathway (Ref. Reference Muller62).

Single BRD modules are present in some members of the tripartite motif (TRIM) family of transcriptional regulators (Ref. Reference Borden63). TRIM proteins are characterised by the presence of a RING finger, one or two zinc-binding motifs named B-boxes, and an associated coiled-coil region (Ref. Reference Meroni and Diez-Roux64). TRIM24 (Tif1α), for instance, contains an N-terminal TRIM domain, a nuclear receptor (LxxLL) interaction motif and a C-terminal PHD-BRD (Ref. Reference Peng, Feldman and Rauscher65). TRIM24 associates with chromatin (Ref. Reference Remboutsika66) and mediates ligand-dependent activation of AR and the retinoic acid receptor (RAR), and has been shown to interact with other nuclear receptors such as thyroid, vitamin D₃ and oestrogen receptors (Ref. Reference Le Douarin67). TRIM28 (TIF1β) is a corepressor for Krüppel-associated box-domain-containing zinc finger proteins (Ref. Reference Friedman68), which have a crucial role in early embryogenesis. TRIM28 associates with heterochromatin-associated factors HP1α, HP1β and HP1γ to promote the silencing of euchromatic genes (Ref. Reference Cammas69), and recruitment of TRIM28 to centromeres is required for induction of the parietal and visceral endoderm differentiation pathways (Refs Reference Li, Kirschmann and Wallrath70, Reference Bartova71, Reference Cammas72). Interestingly, the PHD domain of the TRIM28 corepressor functions as an intramolecular E3 ligase, leading to sumoylation of the adjacent BRD. Sumoylation is required for TRIM28-mediated gene silencing, suggesting that the tightly linked PHD-BRD module functions as an intramolecular ubiquitin-like modifier (SUMO) E3 ligase (Refs Reference Ivanov73, Reference Zeng74).

TRIM33 (Tif1γ) is a ubiquitin ligase that targets SMAD4 (Ref. Reference Dupont75). Formation of transcription regulatory complexes of SMAD4 with receptor-phosphorylated SMAD2 and SMAD3 is a key event in canonical TGFβ signalling. Consequently, depletion of TRIM33 in human cell lines inhibits SMAD4-dependent cell proliferation by competing with SMAD4 for selective binding to receptor-phosphorylated SMAD2 and SMAD3 (Ref. Reference He76). Mice deficient in Trim33 die in utero, demonstrating that TRIM33 has an important role in development (Ref. Reference Kim and Kaartinen77). The relatively poorly studied TRIM66 (Tif1δ) is mainly expressed in testis and, similarly to TRIM24/33, associates with heterochromatin-associated factors (HPs) but not with nuclear receptors, and functions as a transcriptional silencer (Ref. Reference Khetchoumian78).

The TRIM family member PML (promyelocytic leukaemia protein TRIM19) has no BRD itself but associates with SP proteins, a family of three proteins in humans (SP100, SP110 and SP140) that all contain a PHD-BRD tandem module N-terminally flanked by a SAND DNA-binding domain. The complex of PML and SP100 is found in nuclear bodies, which are nuclear structures that have been associated with the pathogenesis of acute promyelocytic leukaemia (Ref. Reference Bernardi and Pandolfi79). Nuclear bodies are implicated in the regulation of many cellular functions, including chromatin organisation (Ref. Reference Boisvert80), DNA repair and genome stability (Refs Reference Dellaire81, Reference Zhong82), as well as regulation of transcription (Refs Reference Boisvert, Hendzel and Bazett-Jones83, Reference Wang84, Reference Wu85). In addition, the nuclear body is a target of autoantibodies in patients with primary biliary cirrhosis (Ref. Reference Stinton86) and is involved in viral response (Ref. Reference Ishov and Maul87). However, little is known about the precise mechanisms whereby nuclear body proteins exert their functions.

BRDs have an essential role in the assembly and correct targeting of SWI/SNF complexes, which are particularly rich in BRD interaction modules. SWI/SNF complexes, also called Brahma-associated factors (BAFs), remodel chromatin structure, contributing to either transcriptional activation or repression of target genes, depending on the composition of the various complexes. The components of SWI/SNF complexes were originally identified in screens for mutants that result in defects in mating-type switching in yeast or that were unable to grow on sucrose (Refs Reference Neigeborn and Carlson88, Reference Stern, Jensen and Herskowitz89, Reference Winston and Carlson90). Microarray studies later showed that SWI/SNF functions as a transcriptional regulator that affects about 5% of all genes in yeast (Ref. Reference Sudarsanam91). Mammalian SWI/SNF complexes have a key role in cell differentiation and proliferation, and represent an essential component of the embryonic stem cell core pluripotency transcriptional network (Refs Reference Ho92, Reference Singhal93). All SWI/SNF complexes contain a core subunit, which alters chromatin structure in an ATP-dependent manner, resulting in an open and accessible conformation with increased affinity for transcription factors (Ref. Reference Peterson, Dingwall and Scott94). In humans, two related SWI/SNF ATPase components are expressed. These two proteins are mutually exclusive in SWI/SNF complexes and have been named after the Drosophila homologue Brahma as BRG1 (Brahma-related gene-1, SMARCA4) and the related protein BRM (SMARCA2). BRG1 and BRM contain a C-terminal BRD that has been implicated in the recognition of acetylated lysines within histone H3 and H4 tails (Ref. Reference Shen95). Several SWI/SNF complexes have been shown to mediate critical interactions between a number of hormone and other nuclear receptors (Refs Reference Link96, Reference Aoyagi, Trotter and Archer97, Reference Debril98, Reference Trotter and Archer99). In addition, BRG1 has been shown to associate with Rb proteins, inducing cell cycle arrest and transcriptional repression in an HDAC-dependent manner. BRG1/HDAC-containing complexes have been shown to repress expression of genes involved in cell cycle regulation (Refs Reference Pal100, Reference Zhang101). The chromatin-remodelling activity of BRG1 has also been shown to be important for traversal of the nucleosome by RNAPII (Ref. Reference Subtil-Rodriguez and Reyes102). The SWI/SNF complex PBAF (polybromo-associated BRG1-associated factor) is characterised by the presence of the polybromo protein (PB1) (also called BAF180) (Refs Reference Lemon103, Reference Ryme104). PB1 is required for ligand-dependent transactivation by nuclear hormone receptors and contains six BRDs, two bromo-associated domains (BAH) and a homeobox DNA-binding domain. PBAF complexes, but not BAF, activate vitamin-D-receptor-dependent transcription in response to vitamin D, and mice lacking Pb1 have defects in heart development (Ref. Reference Wang105) because of impaired epithelial-to-mesenchymal transition and arrested maturation of the epicardium as a result of the downregulation of FGF, TGF and VEGF signalling (Ref. Reference Huang106). PB1 also has a role in cell cycle regulation and is a key regulator of senescence (Ref. Reference Burrows, Smogorzewska and Elledge107).

BRDs are present in chromatin-remodelling complexes of the ISWI (imitation SWI) family that assemble into at least seven different complexes containing a central core ATPase of the two SNF2-like mammalian homologues SNF2L and SNF2H of yeast ISWI. ISWI complexes are key regulators of transcription, heterochromatin replication and chromatin structure. The ISWI complex NURF (nucleosome remodelling factor) contains the BRD PHD finger transcription factor BPTF. BPTF contains a C-terminal PHD-BRD and was identified as a highly expressed protein in patients with Alzheimer disease as fetal Alz-50 reactive clone 1, and in fetal brain in patients with neurodegenerative diseases (fetal Alzheimer antigen, FALZ) (Refs Reference Bowser, Giambrone and Davies108, Reference Jones, Hamana and Shimane109). The PHD domain in BPTF associates with trimethylated histone H3 Lys4, an interaction that is required for the recruitment of SNF2L1 to promoters (Ref. Reference Wysocka110). The ISWI complex ACF/WCRF (ATP-utilising chromatin remodelling and assembly factor/Williams syndrome transcription factor) contains BAZ1 (also called WCRF or ACF1), a protein of the BAZ (BRD adjacent zinc finger) family, which is represented by four related genes in humans (BAZ1A, BAZ1B, BAZ2A and BAZ2B), with similar domain organisation, including a PHD-BRD interaction module. BAZ1A was first identified in HeLa cell nuclear extract as a factor associating with SNF2H forming a complex with ATP-dependent chromatin-remodelling activity (Ref. Reference Bochar111). Later, the SNF2H/BAZ1A remodelling activity was shown to be required for the DNA-replication machinery to penetrate condensed chromatin structures. SNF2H/BAZ1A is particularly enriched in replicating pericentromeric heterochromatin, and knockdown of BAZ1A by RNAi impairs replication of condensed chromatin (Refs Reference Collins112, Reference Poot113).

BAZ2A (TIP5, TTF-1-interacting protein 5) is a key subunit of the NoRC (nucleolar remodelling complex), which mediates transcriptional silencing of ribosomal RNA (Ref. Reference Strohner114). Interestingly, mutation of a tyrosine residue in the BAZ2A BRD in yeast impairs interaction with acetylated histones (Ref. Reference Ladurner115) and the mutation Y1775F represses NoRC interaction with chromatin and RNA polymerase I transcription (Ref. Reference Zhou and Grummt116). A table containing all human BRD proteins identified to date and a phylogenetic tree of this protein family is shown in Table 1 and Figure 1a, respectively.

Figure 1. Phylogenetic tree of the human bromodomain family and substrate recognition of bromodomains. (a) Phylogenetic tree based on sequence alignments of predicted BRDs. For targets with multiple BRDs, the domains have been numbered starting from the N-terminus and the number is shown in parentheses. (b) Interaction of mouse BRD4 (Ref. Reference Vollmuth, Blankenfeldt and Geyer117) and mouse BRDT (Ref. Reference Moriniere118) with monoacetylated Lys14 in histone H3 and a diacetylated H4 peptide monoacetylated on both Lys5 and Lys8. (c) Surface representation in similar orientation. See Table 1 for an explanation of protein symbols.

BRD substrates

Given the central role of BRDs in epigenetic gene regulation, it is surprising that only a few substrates have been reported and mapped to specific sites. Reported affinities range from low micromolar to millimolar K _D values, raising questions regarding the physiological relevance of described weak in vitro substrate interactions, as well as which additional factors contribute to binding specificity (Table 2). BRDs are often associated with other protein-interaction modules, a mechanism that is thought to generate high target selectivity and increased binding affinity with substrates owing to avidity that is generated on simultaneous binding of several interaction domains. This property led to the suggestion that epigenetic regulation recognises patterns of post-translational modifications (words) rather than single modifications (letters) (Ref. Reference Wu, Lessard and Crabtree130). In addition, the reading process might require combinations of several modifications for high-affinity interaction with a single BRD. Recently, Moriniere and coworkers showed that the testis-specific BET isoform BRDT requires the presence of several acetylation sites for high-affinity binding to histone tails (Ref. Reference Moriniere118). Interestingly, both acetylated lysines interact with the same binding pocket in BRDT (Fig. 1b). It is also likely that other post-translational modifications, such as phosphorylation and methylation, influence substrate recognition, providing the basis for crosstalk of transcription control and cellular signalling. Similarly, the related BRD protein BRD3 also requires two adjacent acetylation sites for tight interaction with the transcription factor GATA1 (Ref. Reference Gamsjaeger131).

Table 2. Bromodomain substrates with known affinity

^aMethod of affinity experiments are in parentheses. ^bHistone H3 residues 1-25 with single acetylations on K4, K9, K14, K18 or K23.

Abbreviations: FA, stopped-flow fluorescence anisotropy; FP, fluorescent polarization; ITC, isothermal titration calorimetry; NMR, nuclear magnetic resonance; SPR, surface plasmon resonance.

BRDs as therapeutic targets in cancer

Many proteins that use BRDs for their recruitment to specific regulatory complexes have been implicated in the development of cancer. BRD-containing proteins are usually multicomponent, and often the reported disease association has not been directly linked to defects in the BRD module itself. However, a number of dominant oncogenic rearrangements and correlation of overexpression of BRD proteins with patient survival provide a strong case for targeting BRDs in cancer.

Genetic rearrangements of BRD-containing proteins have been linked to the development of a number of extremely aggressive tumours. A very aggressive poorly differentiated carcinoma that originates mainly from midline locations such as the head, neck or mediastinum is NUT (nuclear protein in testis) midline carcinoma (NMC) (Ref. Reference French132). NMCs are genetically characterised by translocations that involve the NUT protein with BRD4, BRD3 or an unknown partner gene. BRD4–NUT rearrangements are most frequent, occurring in two-thirds of cases. Both BRD4–NUT and BRD3–NUT fusion genes encode proteins composed of the N-terminal tandem BRDs and almost the entire NUT gene (Fig. 2). BRD–NUT blocks cellular differentiation, and depletion of this oncogene by RNAi results in squamous differentiation and cell cycle arrest (Refs Reference French133, Reference French134). BRD4–NUT specifically recruits CBP/p300, leading to stimulation of CBP/p300 HAT activity, formation of nuclear foci and inactivation of p53 (Ref. Reference Reynoird135). Selective inhibition of BRD4–NUT by recently developed acetyl lysine competitive inhibitors results in epithelial differentiation, tumour shrinkage and survival in BRD4–NUT xenograft mice (Ref. Reference Filippakopoulos136).

Figure 2. Domain organisation of bromodomain proteins and translocations in cancer. BRD modules are shown in green (labelled BRD). Other domain types are labelled directly in the figure and breakpoints are indicated by arrows. Wild-type domain arrangements are shown in the upper panel. See Table 1 for an explanation of protein symbols.

Chromosomal translocations of CREBBP with the MLL protein and the monocytic leukaemia zinc finger protein (MOZ) have been described in myeloid and lymphoid acute leukaemia and myelodysplasia secondary to therapy with drugs targeting DNA topoisomerase II (Refs Reference Sobulo137, Reference Panagopoulos138) (Fig. 2). CREBBP also contributes to tumourigenesis of NUP98–HoxA9 and MOZ–TIF2 fusion proteins by activating transcription (Refs Reference Kasper139, Reference Deguchi140). In addition, CREBBP mutations have been identified in relapsed acute lymphoblastic leukaemia (Ref. Reference Deguchi140) and are very common in diffuse large B-cell lymphoma and follicular lymphoma, constituting the major pathogenetic mechanism shared by common forms of B-cell non-Hodgkin's lymphoma (Ref. Reference Pasqualucci141). CREBBP and the related HAT EP300 are also highly expressed in advanced prostate cancer, and expression levels have been linked with cancer patient survival (Ref. Reference Bouchal142).

Overexpression of several BRD proteins has been reported in cancer and has been linked to patient survival. For instance, a recent study showed that ATAD2 is overexpressed in more than 70% of breast tumours and that higher protein levels correlate with tumour histological grades, poor overall survival and disease recurrence (Ref. Reference Ciro33). Revenko and coworkers showed that ATAD2 is required for recruitment of specific E2F transcription factors and for chromatin assembly of the host cell factor 1–MLL histone methyltransferase complex. As a result of its association with the MLL methyltransferase, depletion of ATAD2 results in a marked decrease of trimethylation of Lys4 in histone H3, which has been linked to transcriptional activation. BRD mutations disable ATAD2 function as an E2F coactivator and its ability to promote cancer cell proliferation (Ref. Reference Revenko32). The closely related protein ATAD2B has recently been shown to be highly expressed in glioblastoma and oligodendroglioma, as well as in breast carcinoma (Ref. Reference Leachman143).

Aberrant expression has also been reported for TRIM24 in breast cancer, and high expression levels have been shown to negatively correlate with survival of breast cancer patients (Ref. Reference Tsai129). In liver, however, TRIM24 seems to function as a liver-specific tumour suppressor (Ref. Reference Khetchoumian144). TRIM24 also interacts with AR and enhances transcriptional activity of the AR by dihydrotestosterone in prostate cancer cells (Ref. Reference Kikuchi145). These data suggest that TRIM24 function and its role in tumourigenesis might be highly context dependent.

The testis-specific BET family member BRDT is frequently overexpressed in non-small-cell lung cancer (Ref. Reference Grunwald146) and several other cancers (Ref. Reference Scanlan147), but the functional consequences of BRDT overexpression have not been investigated so far. The role of BRD4 in cancer is better understood. BRD4 has been shown to be a key regulator of cell cycle control and transcriptional elongation of growth-promoting genes. In particular, the key role of BRD4 in the recruitment of P-TEFb (CDK9/cyclinT) to transcriptional start sites provides an alternative strategy to targeting CDK9, which emerged as a validated target in chronic lymphocytic leukaemia (Ref. Reference Tong148). In breast cancer, however, BRD4 has been identified as an inherited susceptibility gene for disease progression and its expression levels have been associated with patient survival (Ref. Reference Crawford149). BRD4 and BRD2 also have a key role for the transmission of tumour viruses during mitosis by providing a chromatin anchor to viral episomes. For instance, during latent viral infection of herpes viruses associated with development of Kaposi sarcoma, the transmission of viral genomes to daughter cells during mitosis is mediated by the episome's latency-associated nuclear antigen 1, which is tethered to chromatin through its interaction with BRD4 (Ref. Reference You150). Also, papilloma viruses that have been linked to the development of cervical cancers and Epstein–Barr viruses associate with BRD4 in order to anchor their viral genomes to mitotic chromosomes (Refs Reference Weidner-Glunde, Ottinger and Schulz151, Reference Lin152).

BRDs as therapeutic targets for the treatment of inflammation

Transcriptional control of proinflammatory cytokines is the central mechanism in the aetiology of inflammatory disease, and given the success of HDAC inhibitors in this area, it is likely that selective BRD inhibitors will modulate these processes. A first example has been provided by the recent pan-BET inhibitor iBET, which leads to the disruption of chromatin complexes responsible for the expression of inflammatory genes and conferred protection against lipopolysaccharide-induced endotoxic shock and bacteria-induced sepsis (Ref. Reference Nicodeme153). Three sites of polymorphism in BRD2 have recently been linked to rheumatoid arthritis (Ref. Reference Mahdi154) and Brd2-hypomorphic mice are severely obese and have reduced inflammation in fat tissue (Ref. Reference Wang46).

The BRD-containing HATs EP300/CREBBP have been proposed as therapeutic targets in inflammatory diseases such as lung inflammation and asthma (Ref. Reference Rajendrasozhan, Yao and Rahman155). Activation of proinflammatory genes is intimately linked to activation of nuclear factor κB (NFκB). The activated p65 subunit translocates to the nucleus, where its affinity to its target genes and transcriptional activity is regulated by acetylation by EP300/CREBBP. Compounds that inhibit NFκB acetylation such as the natural product gallic acid have anti-inflammatory properties (Refs Reference Choi156, Reference Jung157). EP300 and PCAF also regulate inflammatory responses through their regulation of cyclooxygenase-2 (COX2) expression. COX2 is a key enzyme of prostaglandin biosynthesis that is well established as a major player in inflammatory response and a clinically successful target for the development of anti-inflammatory drugs. Stimulation by bacterial lipopolysaccharides and other cytokines leads to increased binding of PCAF and EP300 to the COX2 promoter, and its activation. Conversely, inhibitors of EP300 have been shown to reduce COX2 protein levels and promoter activities (Ref. Reference Deng, Zhu and Wu158).

The concerted activation of several proinflammatory genes is regulated by the SWI/SNF class of ATP-dependent remodelling complexes, which make the promoters of inflammatory genes permissive for transcriptional induction. The presence of the catalytic ATPase subunit BRG1 at the promoter of proinflammatory genes such as IL6 has been shown to be necessary for activation of these genes, and termination of transcriptional activation is regulated by proteasomal degradation of BRG1, ensuring a timely and adequate immune response (Ref. Reference Cullen, Ponnappan and Ponnappan159). Although experimental data are still missing, it is intriguing to speculate that removal of BRG1 from promoter regions might have an effect on inflammatory conditions.

BRDs as therapeutic targets for the treatment of neurological diseases

Increasing evidence points to the fact that epigenetic targets have a role in the molecular manifestation of stress and related disorders. Because BRD inhibitors have only just been discovered, no study has addressed the role of BRD inhibition in neurological disorders so far. However, several studies report important functions of BRD-containing proteins in several diseases. TRIM28, for instance, is highly expressed in the mouse hippocampus and cerebellum. Inducible deletion of Trim28 in the forebrain of adult mice resulted in stress-related behaviour and cognitive impairment of these mice similar to effects observed in behavioural disorders such as borderline personality or bipolar disorder. Chromatin immunoprecipitation experiments confirmed changes in histone methylation and acetylation patterns in the promoter regions of TRIM28 target genes such as Mkrn3 and Pcdhb6 (Refs Reference Alter and Hen160, Reference Jakobsson161).

Two other BRD-containing proteins, SMARCA2 (BRM) and BRD1, have been identified in genome-wide association studies as susceptibility genes for schizophrenia and bipolar disorder in several independent studies, but the molecular mechanisms are still unclear (Refs Reference Bjarkam162, Reference Nyegaard163). In addition, low levels of SMARCA2 have been found in the post-mortem prefrontal brains of schizophrenic patients, and the gene expression profiles in the diseased brains match those after downregulation of SMARCA2 in cells and in SMARCA2-knockout mice, which show impaired social interaction and prepulse inhibition. Interestingly, SMARCA2 expression can be increased in the mouse brain on application of antipsychotic drugs, providing further evidence of the potential of this protein as a target for the treatment of schizophrenia (Refs Reference Loe-Mie164, Reference Koga165).

Mutations in CREBBP, and less frequently in EP300, are the genetic background for Rubinstein–Taybi syndrome (RTS), a rare human genetic disorder characterised by mental retardation and physical abnormalities; many patients with RTS have either breakpoints or microdeletions in chromosome 16p13.3 where the CREBBP gene is located, but also heterozygous point mutations can lead to RTS (Refs Reference Rouaux, Loeffler and Boutillier166, Reference Zimmermann167, Reference Viosca168). Several of the pathological features can be mirrored by heterozygous Crebbp-deficient mice strains (Refs Reference Valor169, Reference Oike170, Reference Tanaka171). Although the precise mechanisms underlying the disease are not yet understood, it is thought that the HAT activity of CREBBP and reduced transcriptional activity result in altered synaptic plasticity, which ultimately influences long-term memory, leading to mental retardation (Ref. Reference Saura and Valero172). EP300 also has a role in the aetiology of amyotrophic lateral sclerosis, Alzheimer disease and Huntington disease. Huntington disease is a polyQ disease in which polyglutamine repeats are added to the Huntingtin protein, causing its translocation to the nucleus and formation of aggregates. CREBBP and PCAF interact directly with Huntingtin aggregates, resulting in their depletion (Refs Reference Bartsch173, Reference Petrij174). Indeed, HDAC inhibitors have long been used as mood stabilisers and are studied for the treatment of Huntington and Alzheimer diseases (Ref. Reference Narayan and Dragunow175).

Development of BRD inhibitors

BRDs share a conserved fold that comprises a left-handed bundle of four alpha helices (α _Z, α _A, α _B, α _C), linked by highly variable loop regions (ZA and BC loops), which form the rim of the substrate-binding pocket and determine substrate recognition (Refs Reference Jacobson55, Reference Dhalluin127) (Fig. 3a). Despite the conservation of the overall BRD fold, the surface and loop regions of BRDs are highly diverse, suggesting that inhibitors with high specificity can be designed. Cocrystal structures with peptidic substrates have demonstrated that the acetyl lysine is recognised by a central deep hydrophobic cavity, where it is anchored by a hydrogen bond to an asparagine residue present in most BRDs (Ref. Reference Owen177). Acetylation of lysine residues neutralises the charge of the ɛ-amino group. As a consequence, the central cavity of acetyl lysine binding sites in BRDs is quite hydrophobic and particularly rich in aromatic residues; it also has sufficient size to accommodate potent acetyl lysine competitive ligands. These properties make BRDs attractive targets for the design of pharmacologically active molecules that compete with protein interactions mediated by these modules.

Figure 3. Structural overview of a bromodomain and binding mode of bromodomain inhibitors. (a) Ribbon diagram of the first BRD of BRD4. The main structural elements as well as the acetyl lysine binding site residues are labelled. (b) Superimposition of a diacetylated BET substrate peptide and the inhibitor JQ1. Inhibitor and peptide molecules are shown in stick representation and are coloured according to atom types. (c) Binding of JQ1 to the bromodomain of BRD4. Conserved water molecules in the active site are highlighted and hydrogen bonds are shown as dashed lines. (d) Complex of ischemin with CREBBP (Ref. Reference Sachchidanand176).

Potent and very selective inhibitors have recently been published for BET BRDs (Refs Reference Filippakopoulos136, Reference Nicodeme153, Reference Chung178). All inhibitors that have been published so far are based on a triazolo-diazepine scaffold that successfully mimics interactions observed in BET peptide complexes (Fig. 3b). Interestingly, a number of tightly bound and conserved water molecules remain in cocrystal structures of BET triazolo-diazepine complexes, which interact with the inhibitor through a network of hydrogen bonds (Fig. 3c). Two BET inhibitors have been studied in two different disease models, providing compelling support of BET BRDs as targets in drug discovery. The inhibitor JQ1 has been studied in midline carcinoma where inhibition of BRD4–NUT led to terminal differentiation, cell cycle arrest and apoptosis of carcinoma cells, and significant reduction of tumour growth in patient-cell-line-derived xenograft models (Ref. Reference Filippakopoulos136). The inhibitor iBET led to significant reduction of the expression of proinflammatory genes in activated macrophages, and conferred protection against lipopolysaccharide-induced endotoxic shock and bacteria-induced sepsis, supporting inhibition of BET BRDs as a strategy for the generation of immunomodulatory drugs (Ref. Reference Nicodeme153).

Acetyl lysine mimetic inhibitors have also been reported in the case of CREBBP, competing for its interaction with p53. These inhibitors were identified by NMR screening using a library of compounds that consists of one aromatic ring connected to an −NHCOCH₃ group by different types of linkers (Ref. Reference Borah179). The same laboratory also reported a series of cyclic peptides with improved binding affinities over natural substrates (Ref. Reference Borah179) and azobenzene-based inhibitors such as 4-hydroxyphenylazo-benzenesulfonic acid (MS456) and ischemin. Ischemin binds to the BRDs of CREBBP with a dissociation constant (K _D) of 19 µM and shows at least fivefold selectivity over other human BRDs. The binding mode of ischemin in CREBBP is shown in Figure 3d. In cellular assays ischemin alters post-translational modifications of p53 and histones, inhibits p53 interaction with CBP and transcriptional activity in cells, and prevents apoptosis in ischaemic cardiomyocytes (Ref. Reference Borah179). Early lead compounds such as N1-aryl-propane-1,3-diamine have also been identified for PCAF (Ref. Reference Zeng180). A summary of the chemical structures of the currently most advanced BRD inhibitors is shown in Figure 4.

Figure 4. Chemical structures of bromodomain inhibitors. Specificity and dissociation constants are also indicated.

Research in progress and outstanding research questions

Targeting BRDs for the development of protein-interaction inhibitors has recently emerged as a strategy for the design of pharmacologically active reagents. The relatively weak interaction of BRDs with their substrates, the diversity and physicochemical properties of the acetyl lysine binding site, and the large number of available crystal structures will facilitate the rational design of such inhibitors. However, BRDs usually constitute only one of the interaction domains found in BRD-containing proteins, and whether selective inhibition of the acetyl lysine interaction alone will result in the desired phenotype needs to be investigated in future research projects. The large number of diseases that have been linked to BRD-containing proteins and the success of particular HDAC inhibitors indicate that BRD inhibitors will find a large number of applications in pharmaceutical sciences and basic research.