RVX-208, an inhibitor of BET transcriptional regulatorswith selectivity for the second bromodomainSarah Picauda, Christopher Wellsa, Ildiko Felletara, Deborah Brothertona, Sarah Martina, Pavel Savitskya,Beatriz Diez-Dacalb, Martin Philpotta, Chas Bountraa, Hannah Lingarda, Oleg Fedorova, Susanne Müllera,Paul E. Brennana, Stefan Knappa,c,1, and Panagis Filippakopoulosa,b,1

aStructural Genomics Consortium, Nuffield Department of Clinical Medicine, University of Oxford, Oxford OX3 7DQ, United Kingdom; bLudwig Institute forCancer Research, Nuffield Department of Clinical Medicine, University of Oxford, Oxford OX3 7DQ, United Kingdom; and cTarget Discovery Institute, NuffieldDepartment of Clinical Medicine, University of Oxford, Oxford OX3 7BN United Kingdom

Edited by Angela M. Gronenborn, University of Pittsburgh School of Medicine, Pittsburgh, PA, and approved October 23, 2013 (received for reviewJune 5, 2013)

Bromodomains have emerged as attractive candidates for thedevelopment of inhibitors targeting gene transcription. Inhibitorsof the bromo and extraterminal (BET) family recently showedpromising activity in diverse disease models. However, the pleio-tropic nature of BET proteins regulating tissue-specific transcriptionhas raised safety concerns and suggested that attempts should bemade for domain-specific targeting. Here, we report that RVX-208,a compound currently in phase II clinical trials, is a BET bromodo-main inhibitor specific for second bromodomains (BD2s). Cocrystalstructures revealed binding modes of RVX-208 and its synthetic pre-cursor, and fluorescent recovery after photobleaching demon-strated that RVX-208 displaces BET proteins from chromatin.However, gene-expression data showed that BD2 inhibition onlymodestly affects BET-dependent gene transcription. Our datademonstrate the feasibility of specific targeting within the BETfamily resulting in different transcriptional outcomes and high-light the importance of BD1 in transcriptional regulation.

small molecule inhibitor | epigenetics | microarray | ApoA1

Bromodomains (BRDs) are protein-interaction modules thatare selectively recruited to e-N-acetylated lysine-containing

sequences. BRDs are present in 46 diverse, mostly nuclear pro-teins functioning as effector domains of transcriptional regu-lators, chromatin modulators, and chromatin-modifying enzymes(1). BRD-containing proteins have been implicated in the de-velopment of many diverse diseases, and the architecture of theiracetyl-lysine binding pocket makes them attractive targets for thedevelopment of potent and specific inhibitors (2, 3). All BRDmodules share a conserved fold comprising a left-handed helicalbundle creating a deep, largely hydrophobic and aromatic bind-ing pocket for the specific recognition of peptide sequencescontaining one or more e-N-acetylated lysine residues (1, 4–6).In particular the bromo and extraterminal (BET) proteins,

which comprise four members in human (BRD2, BRD3, BRD4,and the testis-specific BRDT), recently received a lot of atten-tion after highly potent and cell-active pan-BET inhibitors weredeveloped (7–10). BETs are transcriptional regulators thatcontrol expression of genes that play key regulatory roles incellular proliferation, cell cycle progression, and apoptosis (11,12). Dysfunction of BET proteins has been associated with thedevelopment of aggressive tumors, such as NUT midline carci-noma (NMC). In NMC, the N-terminal bromodomains of BRD3or BRD4 are fused in frame with the testis-specific protein NUT(nuclear protein in testis), giving rise to an incurable fatal sub-type of squamous carcinoma and in some cases tumors of othertissue origin (13). Importantly, BETs play a critical role in tu-morigenesis also outside NMCs by driving the expression ofgenes that are essential for tumor growth and survival, such asc-Myc (14) and Aurora B (15). The potent pan-BET inhibitors(+)-JQ1 and GSK1210151A (I-BET151) have exhibited significantantitumor activity in murine models of NUT midline carcinoma

(7), multiple myeloma (16), acute myeloid and mixed lineageleukemia (17), lung cancer (18), and glioblastoma (19).BET family members play an essential role in diverse cellular

processes, including general transcriptional elongation (20, 21),replication (22), hematopoiesis (23), adipogenesis (24, 25), andspermatogenesis (26), suggesting that drug discovery effortsshould explore isoform, or even domain-specific targeting, toavoid adverse effects of prolonged pan-BET inhibition duringtreatment in different tissues.The quinazolone RVX-208 (Fig. 1A) has been developed by

Resverlogix Corporation for the treatment of cardiovasculardiseases associated with atherosclerosis (27, 28) and has morerecently entered clinical studies on Alzheimer’s disease (29).RVX-208 is a derivative of the plant polyphenol resveratrol(3,4’,5-trihydroxy-transstilbene) that leads to an increase ofplasma levels of the high-density lipid protein ApoA1. IncreasingApoA1 levels has emerged as a promising approach for thetreatment of atherosclerosis (30), and recent phase IIb clinicaltrial data using RVX-208 as an ApoA1 modulator have beenencouraging (28). ApoA1 expression is regulated by BET pro-teins, and chemical inhibition of BET bromodomains has beenassociated with ApoA1 up-regulation on transcriptional andprotein levels (9, 31, 32). As a consequence, a similar mode ofaction has also been suggested for RVX-208, but no data char-acterizing the RVX-208/BET interaction have been published sofar. The promising clinical outcome of RVX-208 trials and thepresumed function of RVX-208 as a BRD inhibitor prompted us

Significance

Bromo and extraterminal (BET) proteins have diverse roles inregulating tissue-specific transcriptional programs, raising safetyconcerns for their inhibition and suggesting that targeting ofspecific isoforms or even specific domains within this subfamilyis important. We report the discovery and characterization ofRVX-208 as a domain-selective inhibitor of BETs and providea potential mechanism of action of a clinical compound that wasidentified based on phenotypic screens.

Author contributions: C.B., H.L., O.F., S. Müller, P.E.B., S.K., and P.F. designed research; S.P.,C.W., I.F., D.B., S. Martin, P.S., B.D.-D., M.P., and P.F. performed research; S.P., C.W., B.D.-D.,O.F., S.K., and P.F. analyzed data; and S.K. and P.F. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

Freely available online through the PNAS open access option.

Data deposition: The crystal structures reported in this paper have been deposited in theProtein Data Bank, www.pdb.org (4MR3–4MR6). Microarray data have been deposited inthe Gene Expression Omnibus (GEO) database, www.ncbi.nlm.nih.gov/geo (accession no.GSE51143).1To whom correspondence may be addressed. E-mail: panagis.filippakopoulos@sgc.ox.ac.uk or stefan.knapp@sgc.ox.ac.uk.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1310658110/-/DCSupplemental.

19754–19759 | PNAS | December 3, 2013 | vol. 110 | no. 49 www.pnas.org/cgi/doi/10.1073/pnas.1310658110

to study its role regulating BET-dependent transcription. In-terestingly, we found that RVX-208 is specific for BET bromodo-mains and showed preferred binding to their second bromodomains(BD2s). Highest selectivity was observed for BD2 of BRD2 (23-fold) and BRD3 (21-fold). RVX-208 displaces BRD3 fromchromatin at higher concentrations, and inhibition of BET bro-modomains in the liver hepatocellular carcinoma cell line HepG2resulted in weak regulation of only a subset of BET target genes.The study suggests that first and second BDs of BET familymembers can be selectively targeted despite high levels of sequencehomology resulting in distinct transcriptional outcomes.

ResultsHuman BET proteins have a modular architecture comprisingtwo N-terminal BRD modules, an extraterminal (ET) domain,and a C-terminal motif (BRD4 and BRDT only) (Fig. S1A).Their two highly conserved BRD modules share a high degree ofsequence homology. Sequence homology is most pronouncedcomparing all four BD1s or BD2s, respectively, resulting inclustering of these interaction modules in sequence- and struc-ture-based phylogenetic trees (1). Interestingly, sequence com-parisons revealed three residue positions in close proximity tothe acetyl-lysine peptide binding site that differ between BD1 andBD2 domains: with the exception of BRDT, the residue positioncorresponding to BRD4/BD1 Q85 is a lysine residue in BD2s; theposition corresponding to the BRD4/BD1 residue D144 is a histi-dine residue in all BET BD2s; the position corresponding to theBRD4/BD1 residue I146 is a valine residue in BD2s. The locationof these residues suggested that these sequence variations may beexplored for the development of inhibitors that specifically rec-ognize one of the two BET BRDs (Fig. S1B).The established role of RVX-208 up-regulating the BET-

regulated HDL protein ApoA1 without associated antiproliferativeeffects prompted us to synthesize this compound following thesynthetic route disclosed by Resverlogix Corporation [previouslyestablished by Hansen—examples no. 4 and no 7 (33)] with

minor changes (Fig. S1C) and to study its interaction with humanBRD proteins.

RVX-208 Is a Potent Inhibitor of Second BET Bromodomains. To es-tablish a selectivity profile for RVX-208, we used temperature-shiftassays (ΔTm) carried out on 44 of the 61 human bromodomains andfound that its effect on BRD temperature stabilization is limited tothe BET subfamily. Interestingly, we observed significant stabiliza-tion of only second bromodomains (BD2s) of BET proteins (Fig. 1Band Table S1), suggesting that RVX-208 selectively targets BD2s. Incontrast, the synthetic precursor RVX-OH, which lacks thehydroxylether substitution (Fig. 1A), also specifically interacted withBET family members, although the differences in ΔTm betweenBD1s and BD2s were diminished. In addition, RVX-OH showeda number of weak ΔTm shifts on BRD9, CECR2, and some otherBRDs (Table S1). AlphaScreen assays carried out using BRD3, forwhich a large difference in ΔTm values was observed, confirmed theinteractions and demonstrated competitive displacement of histoneH4 acetyl-lysine containing peptides by RVX-208. The IC50 valuesderived from the AlphaScreen data were 87 ± 10 μM and 0.510 ±0.041 μM for BD1 and BD2, respectively, thus about 170-fold se-lectivity (Fig. 1C). In agreement with ΔTm data, RVX-OH exhibi-ted a smaller window of selectivity between the two BRD3bromodomains (IC50, 11.4 ± 5.9 μM and 0.379 ± 0.101 μM for BD1and BD2, respectively) (Fig. 1C). Isothermal titration calorimetry(ITC) led to determination of accurate binding constants in solutionfor both RVX-OH and RVX-208 for all eight BET bromodomains(Fig. 1D and E and Tables S2 and S3). The interaction of RVX-208with BET bromodomains is driven by large negative binding en-thalpy changes resulting in a KD of 4.06 ± 0.16 μM for BD1 andmore than 20-fold stronger binding to BD2 (0.194 ± 0.013 μM) inthe case of BRD3. ITC data measured on other BET bromodo-mains confirmed BD2 selectivity of RVX-208 but resulted in less-pronounced differences in binding affinities. In contrast, RVX-OHshowed KD values between 0.15 and 1.4 μM, with no significantdifferences in ligand affinity for the two BRDs of BRD3. The more

0 1 2 3 4-12-10

-8-6-4-20

kca

l/mol

e of

inje

ctan

t

Molar Ratio0 10 20 30 40 50 60 70-0.3

-0.2

-0.1

0.0

Time (min)

ΔP

(μca

l/sec

)

BD1

BRD4

BD2

RVX-208RVX-OH

RVX-208RVX-OH

0 10 20 30 40 50 60 70 80

-0.15

-0.10

-0.05

0.00

Time (min)

ΔP

(μca

l/sec

)

0 1 2 3 4

-6

-5

-4

-3

-2

-1

0

kca

l/mol

e of

inje

ctan

t

Molar Ratio

BD1BD2

BRD3

-7 -6 -5 -4

0

25

50

75

100

Inhi

bitio

n (%

)

[compound] (log(M))

BD1RVX-208

BD2

BD1RVX-OH

BD2

BRD3/BD2BRD2/BD2

BRD3/BD1

BRD2/BD1

BAZ1A

BRD8B/BD2

BRD8B/BD1

EP300CREBBP

WDR9/BD2PHIP/BD2

BRWD3/BD2BAZ1B

BRD9BRD7

BRPF3BRPF1B

BRPF1ABRD1

KIAA1240 ATAD2

TRIM66

TRIM33BTRIM33A TIF1α

SP110CSP110ASP100

SP140 LOC93349

BAZ2BBAZ2ATRIM28

MLL

TAF1L/BD2 TAF1/BD2TAF1L/BD1

TAF1/BD1PRKCBP1

WDR9/BD1PHIP/BD1BRWD3/BD1

PB1/BD6

PB1/BD5PB1/BD4

PB1/BD2PB1/BD3PB1/BD1

ASH1L

CECR2FALZ GCN5L2

PCAF

SMARCA4SMARCA2B

SMARCA2A

ZMYND11

BRD4/BD1BRDT/BD1

BRD4/BD2

BRDT/BD2

> 43 - 42 - 31 - 20 - 1

< 0

ΔT ( C)om

OO

O N

R

NH

R = -O(CH ) OHR = -OHR = -H

2 2RVX-208:RVX-OH:RVX-H:

BA

C

ED

Fig. 1. In vitro selectivity profile of RVX-208. (A)Structure of the inhibitor RVX-208 and its precursorRVX-OH. (B) Selectivity of RVX-208 within the humanbromodomain family determined using a thermalshift assay. Temperature shifts (ΔTobsm in °C, at 10 μMcompound concentration) are shown as spheres asindicated in the Inset. Screened proteins are shownin bold. (C) Competitive displacement of a tetra-acetylated histone H4 peptide (H41–20K5ac/K8ac/K12ac/K16ac) from BD1 and BD2 of BRD3 using RVX-208 or RVX-OH (as indicated in the Inset) in a bead-based proximity assay (ALPHA assay). (D) Isothermaltitration calorimetry (ITC) binding study. Data col-lected against the bromodomains of BRD3, showingraw injection heats for titrations of protein (BD1 orBD2) into compound. The Inset shows the normal-ized binding enthalpies corrected for the heat ofprotein dilution as a function of binding site satu-ration (symbols as indicated in the figure). Solid linesrepresent a nonlinear least squares fit using a single-site binding model. (E) Isothermal titration calorimetry(ITC) evaluation of RVX-208 and RVX-OH against thebromodomains of BRD4. Data have been correctedand displayed as described in D. All ITC titrations werecarried out in 50 mM Hepes, pH 7.5 (at 25 °C), 150 mMNaCl and 15 °C while stirring at 1,000 rpm.

Picaud et al. PNAS | December 3, 2013 | vol. 110 | no. 49 | 19755

BIOCH

EMISTR

YCH

EMISTR

Y

pronounced differences observed in AlphaScreen assays might bedue to differences in peptide affinity for BD1 and BD2 domains.

RVX-208 Presents a Template for BD2 Rearrangement upon Binding.The observed selectivity profile for RVX-208 and the lack ofselectivity of the closely related compound RVX-OH promptedus to determine the high-resolution cocrystal structures of RVX-208 and RVX-OH with representative BET bromodomains (Fig. 2and Fig. S2). As predicted from our AlphaScreen data, RVX-208bound to the acetyl-lysine binding pocket in a peptide-competitivemanner. In the cocrystal structure of the first bromodomain ofBRD4, the carbonyl oxygen and one of the nitrogen atoms of thequinazolinone ring system act as an acetyl-lysine mimetic moiety,forming a hydrogen bond with the conserved asparagine residue(N140), as well as a water-mediated hydrogen bond with Y97 (Fig.2A and Fig. S2A). The hydroxy-ethylether moiety points out of theacetyl-lysine binding pocket and makes only a few contacts withthe bromodomain surface. The WPF shelf, which contributessignificantly to the affinity of phenyl-isoxazole and methyl-triazoloinhibitors (9, 34), is not occupied by the inhibitor but is occupiedby an ethylene glycol solvent molecule. RVX-208 makes no directinteractions with residues unique to BD1 except to a water-mediated hydrogen bond with Q85. The binding mode of RVX-208 is largely conserved in BD2 domains, mimicking that of ahistone substrate with the ligand occupying the entire channel usedby a single acetyl-lysine. However, the BD2 unique residue H433in BRD2 flips into the acetyl-lysine binding site packing against thephenyl ring of the inhibitor, providing a possible explanation forthe tighter affinity for BD2 domains (Fig. 2B and Fig. S2B). In-terestingly, the RVX-208 binding mode is not conserved in RVX-

OH complexe with the BD1 of BRD4 (Fig. S2C). In this cocrystalstructure, the ligand’s free hydroxyl group from the phenyl ringsystem acts as an acetyl-lysine mimetic moiety, forming a hydrogenbond with N140. Surprisingly, RVX-OH inverts its binding modein the cocrystal complex with the BD2 of BRD2 assuming a similarinteraction as observed in RVX-208 where the quinaxolinonefunction acts as the acetyl-lysine mimetic moiety (Fig. S2D and Fig.S3A). Both RVX-OH and RVX-208 are very well resolved in thehigh-resolution crystal structures (Fig. S3 B and C). The shift inbinding mode of this inhibitor is also evident by the thermody-namic data of this interaction. The RVX-OH interaction withBD1s is characterized by a large negative binding enthalpy changefor all BD1 interactions (∼ −7 to −10 kcal/mol) that is opposed bya negative entropy term (TΔS ∼ −2 kcal/mol). In contrast, in-teraction with BD2s gives rise to a modest negative enthalpychange (∼ −3.5 kcal/mol) associated with a favorable positive en-tropy change (TΔS ∼ +4 to +5 kcal/mol) (Table S3). However, thedifferent binding modes of these two closely related inhibitorscomplicate the interpretation of the structural reasons for theobserved selectivity of RVX-208. To further test the template’sbinding mode, we synthesized RVX-H, an analog lacking the Kac-mimetic free hydroxyl group and tested its ability to stabilize BETBRDs in thermal melt experiments. As expected, this scaffoldsignificantly lost its affinity for BET BD1s (Table S1).We concludethat differences in binding affinity are due to small structuralrearrangements and differences in binding mode that may includecontributions of BD2 unique residues such as H433 in the case ofthe BD2 of BRD2, as well as possible differences in dynamicproperties of BET bromodomains.

Selective Inhibition of BD2 Displaces BETs from Chromatin.Given theweak interaction of RVX-208 with BET BD1s, we were in-terested in establishing whether this inhibitor can dissociate full-length BET proteins from acetylated chromatin. To this end,we established a FRAP (fluorescence recovery after photo-bleaching) assay using full-length human GFP-BRD3 transientlytransfected into U2OS osteosarcoma cells. Exposure of thesecells to the potent pan-BET inhibitor PFI-1 (10, 35) led to sig-nificant reduction of recovery times of the photobleached nu-clear region, suggesting efficient dissociation of BRD3 fromchromatin (Fig. 3), even at concentrations as low as 100 nM,given that the in vitro dissociation constant for this inhibitor is 80nM for the BD1 of BRD3 and 76 nM for BD2 (35). At 250 nM,recovery times reached a plateau that did not decrease further athigher concentrations of the inhibitor. RVX-208 exhibitedslightly weaker activity, displacing BRD3 at concentration of 500nM and higher, thus demonstrating that both bromodomains areneeded for efficient interactions with nucleosomes.

BET Transcriptional Regulation Is Mainly Mediated by First Bromo-domains in Liver. We next investigated the transcriptional regula-tion effect on gene expression in human liver carcinoma HepG2cells by BET bromodomains, by either inhibiting both domains atthe same time using the pan-BET inhibitor (+)-JQ1 (or JQ1 forsimplicity) or RVX-208 seeking to inhibit mainly the secondbromodomain (BD2). A microarray study of cells treated witheither inhibitor for 4 h revealed large differences in gene ex-pression, with the pan-BET inhibitor JQ1 strongly affectingtranscription of genes with almost a 10-fold difference comparedwith the BD2-specific inhibitor RVX-208. Although inhibition ofboth BD1 and BD2 affected the gene expression of 754 geneswithin a 1.5-fold window, only 46 genes were affected by theinhibition of only BD2 using RVX-208 (Fig. 4A and Fig. S4A).Indeed, the top genes that seem to be affected by each inhibitorhave small overlap, and the fold change in their expression issystematically higher when both domains are inhibited, usingJQ1 (Fig. 4 B–D). This trend was also observed when we lookedat the effect on the expression of the top 1,000 statistically

K5K5ac

K8

K12K12acac

R3R3

W370P371

K374

V376

L381

L383

Y386

Y428

H433H433

D434

M438

V435

αZαC

ZA-loop

αA

H4K5ac/K12ac

RVX-208BRD2/BD2

H433

H433H433

R3R3

K5K5acac

K8K8acac

ZA-loop

BC-loop

H4K5ac/K8ac

RVX-208BRD4/BD1

αA

αZ αC

P82

Q85V87

L92

L94

Y97Y97

Y139

ZA-loop

N140 D144D145

M149

I146

W81

K5K5acac

K8K8acac

R3R3Y97Y97

o45o45

B

A

Fig. 2. Cocrystal structures of RVX-208 and RVX-OH with the first and sec-ond bromdomains of human BET proteins. (A) Overview of RVX-208 bindingonto BD1 of BRD4 (Left). A detail of the boxed area is shown on the Right(rotated 45° counterclockwise), highlighting the acetyl-lysine mimetic bind-ing of the inhibitor, compared with a histone H4 di-acetyl peptide (H4K5ac/K8ac, PDB ID no. 3UVW, shown in stick representation, colored in blue),engaging the protein by directly interacting with the conserved asparagine(N140 in BRD4/BD1) in addition to a water-mediated hydrogen bond to Y97.Residues that differ between BD1 and BD2 are highlighted in red. (B)Comparison of the binding mode of RVX-208 and a histone H4 peptide ontoBD2 of BRD2 (Left). The inhibitor engages the protein in the same orienta-tion to that observed in the structure of a di-acetylated H4 peptide (H4K5ac/K12ac, PDB ID code 2E3K, colored as in A). A blow-up of the boxed areahighlights the stabilization of the inhibitor complex, achieved by an inwardsmotion of H433 from the BC-loop region toward the front of the acetyl-lysinebinding pocket, a unique feature of second site-BET bromodomains (Right).Residues that differ from the N-terminal BET BRDs are highlighted in red.

19756 | www.pnas.org/cgi/doi/10.1073/pnas.1310658110 Picaud et al.

significant genes (with an adjusted P value less than 0.05) for eachinhibitor; genes that are strongly up-/down-regulated when bothdomains are inhibited by JQ1 are only weakly regulated when thesecond domain is inhibited by RVX-208 (Fig. S4B) whereas genesthat are strongly up-/down-regulated when the second domain isinhibited by RVX-208 exhibit an even higher degree of regulationwhen both domains are inhibited by JQ1 (Fig. S4C). To furtherprobe and verify the differences in gene expression initiated byBD1 or BD2, we selected a subset of four genes for qPCR studiesthat were strongly down-regulated (AREG, EREG, INHBE, andNR1H4) and four genes that were up-regulated (BRD2,MYBL1, MYLIP, and ZNF117). The effects of pan-BET versusBD2-selective inhibition were monitored in a time-dependent(Fig. 4E) and dose-dependent (Fig. 4F) manner. In the firstinstance, we observed a striking difference in gene expressionwhen both BDs were inhibited by JQ1, with little or no effect whenonly BD2 was inhibited with RVX-208. This effect was verified bythe dose experiment as well, despite the fact that we chose a 10-fold window between the two inhibitors in an effort to account fortheir differences in affinity for the two domains. Despite the lowerconcentrations used for JQ1 (0.3–1.0 μM), as opposed to higherdoses of RVX-208 (0.78–25.0 μM), inhibition of the second do-main was not sufficient to drive a strong transcriptional response.To establish that these effects are not inhibitor-dependent, weconducted a similar experiment and monitored gene expression byquantitative real-time PCR on a set of genes using either the pan-BET inhibitors (+)-JQ1 and PFI-1 in addition to the BD2 in-hibitor RVX-208. We included also the precursor RVX-OH andused (−)-JQ1 as an inactive control compound. As expected, weobserved strong up- or down-regulation when both BD1 and BD2where inhibited by (+)-JQ1, PFI-1, or RVX-OH and a smallereffect when only BD2 was inhibited by RVX-208 (Fig. S4D)whereas the inactive variant, (−)-JQ1, had no significant effect onthe expression of the selected genes. In agreement with the well-

established role of JQ1 down-regulating c-MYC we found thatpan-BET inhibition by JQ1, but not by RVX-208, led to tran-scriptional down-regulation of the c-MYC oncogene in liver cells(Fig. S4D). Interestingly ApoA1, a reported downstream target ofBET proteins (9), was not affected by RVX-208 at the concen-trations tested (27).We found that ApoA1 RNA levels were not affected in a time-

or dose-dependent manner whereas ApoA1 protein levels wereonly slightly affected by (+)-JQ1 treatment (Fig. S5 A–D). Usinga luciferase reporter, we were also unable to observe a significanteffect on ApoA1 transcription using a range of inhibitor con-centrations (Fig. S5 E and F). Taken together, our data suggestthat the transcriptional effect of BET proteins on their targetgenes is mainly dependent on both BDs, which seem to driveinteractions with histones in the context of chromatin, suggestingthat they are not functionally redundant. Indeed, deletion of onlyBD1 in BRDT is sufficient to impair spermatogenesis in mice (26),supporting the dominant role of BD1 in transcription control.

DiscussionBET proteins act as multidomain docking platforms that havea general role in transcription elongation by recruiting the pos-itive transcription elongation factor b (P-TEFb) to acetylatedchromatin (36, 37), as well as tissue-specific functions. The celltype-specific roles have been highlighted by recent reports onBRDT-dependent transcription programs that regulate sper-matogenesis (37–39) and BRD3-dependent recruitment ofGATA1 in hematopoietic cells regulating maturation of ery-throid, megakaryocyte, and mast cell lineages (23, 40), as well asBRD2-dependent roles regulating differentiation of adipose tis-sue (25) and neurons (41). Application of BET inhibitors outsidethe area of oncology, given the pleiotropic nature of BET tran-scriptional regulation, strongly suggests a requirement for moreselective targeting with respect to isoform and/or domains, to avoidadverse effects. Currently developed inhibitors are highly BET-specific but show no or little isoform or domain specificity (7–10).In this work, we have established a first step toward isoform

selective inhibition by characterizing interactions of RVX-208with BET bromodomains, a template currently in phase I/IIclinical trials for the treatment of cardiovascular diseases. Im-portantly, RVX-208 preferentially binds to the second bromo-domain found on BET proteins, exhibiting selectivity over BD1of up to 23-fold. With a KD of 195 nM against BD2 and 4 μMagainst BD1 of BRD3, we were able to show in vitro competitivedisplacement of a histone H4 tetra-acetylated peptide. In-terestingly, cocrystal structures showed that RVX-208 induceda conformational switch upon binding to BD2, resulting in aro-matic stacking accompanied by large entropic changes, as de-termined by isothermal titration calorimetry, which may explainin part its preference for BD2 over BD1 and suggests that dy-namic properties of bromodomains need to be considered in thedesign of domain-specific inhibitors. BET BD1s and BD2s showa high degree of sequence similarity but differ in their recogni-tion of acetylated target sequences (1). It is therefore likely thateach BET bromodomain has a distinct function although thecombined effect seems to be conferring a regulatory output ingene transcription. We observed in this study that targeting theBD2 had only a modest effect on transcription, an observationsupported by previous studies of BRD3 that showed that its re-cruitment to acetylated sites on GATA1 is mediated by BD1(40). Notably, deletion of the first bromodomain in BRDT inmice is sufficient to confer sterility by blocking BRDT-dependentsperm maturation (39). Unfortunately, no genetic knockout stud-ies have been published that target only the BD2 in any of theBET family members, but biochemical studies and data pre-sented here suggest a modulating role of BET function byBD2. In a recent study, Wu et al. showed that phosphorylationof flanking regions of BD2 in BRD4 triggers an intramolecular

* *

*** ***

0 5 10 15 20 25 30 350.0

0.2

0.4

0.6

0.8

1.0

Nor

mal

ized

Inte

nsity

Time (sec)Compound Concentration (nM)

Vehicle100 nM250 nM500 nM

5000 nM

PFI-1

100

nM50

00 n

MPFI

-1

0123456

Rec

over

y t

(s

ec)

1/2

0 100 250 500 5000

RVX-208PFI-1

****** ***

***

0 5 10 15 20 25 30 350.0

0.2

0.4

0.6

0.8

1.0

Nor

mal

ized

Inte

nsity

Time (sec)

Vehicle100 nM250 nM500 nM

5000 nM

RVX-208

Vehi

cle

100

nM50

00 n

MR

VX

-208

Pre 1.4 sec 5.3 sec 9.2 sec 38.9 secDC

A B

Fig. 3. Fluorescence recovery after photobleaching (FRAP) evaluation ofhuman BRD3 dissociation from chromatin. (A) Nuclei of DMSO treated (Top),RVX-208 treated (Middle), or PFI-1 treated (Bottom) cells. Target regions ofphotobleaching are indicated by a white circle. (Scale bars: 10 μm region.) (Band C) RVX-208 and PFI-1 accelerate fluorescence recovery in FRAP experi-ments performed in U2OS cells transfected with full-length GFP-BRD3 ina concentration-dependent manner. (B) RVX-208 can displace the proteinfrom chromatin at concentrations above 500 nM whereas PFI-1 displaces theprotein when used at 100 nM. (D) Quantitative comparison of time to half-maximal fluorescence recovery for FRAP studies using RVX-208 (blue bars)and PFI-1 (red bars) as a function of ligand concentration. Data represent themean ± SEM (n = 15) and are annotated with P values as obtained froma two-tailed t test (*P < 0.05; ***P < 0.001).

Picaud et al. PNAS | December 3, 2013 | vol. 110 | no. 49 | 19757

BIOCH

EMISTR

YCH

EMISTR

Y

rearrangement that unmasks BD2 and directs it toward acetylatedchromatin (42). This observation suggests that BET transcrip-tional regulators are controlled posttranslationally and that spe-cific targeting, such as the one highlighted here toward BD2, maybe able to perturb and modulate BET function in a context-dependent manner. However, further development of domain andisoform specific inhibitors will be necessary to unravel the exactrole of BET bromodomains in gene transcription.

Materials and MethodsCloning, Protein Expression, and Purification. The BRD regions of human BRD3and BRD4 were cloned/amplified as previously described (7). The full-lengthmouse BRD3 ortholog was used to reconstruct the human clone, by firstmutating the C-terminal region and then cloning into pDONR223-hBRD3according to Tillett and Neilan (43) and confirmed by sequencing. Proteinexpression and purification were carried out as previously described (7).

RVX-H, RVX-OH, and RVX-208 Synthesis. The synthesis and characterization ofthese compounds were carried out following the synthetic route previouslydescribed (44) with minor changes, using the synthetic scheme highlighted inFig. S1C.

Protein Stability Shift Assay (Tm Assay). Thermal melting experiments werecarried out using anMx3005p Real-Time PCR machine as previously described(7). Temperature shifts (ΔTobsm ) for three independent measurements perprotein/compound are summarized in Table S1.

Competitive Histone Displacement Assay (AlphaScreen Assay). Experiments wererun on a PHERAstar FS plate reader using an AlphaScreen 680 excitation/570emission filter set. IC50 values were calculated in Prism 5 after normalization

against corresponding DMSO controls. Assays were performed as previouslydescribed (7, 45), with minor modifications from the manufacturer’s protocol.

Isothermal Titration Calorimetry. Experiments were carried out on an ITC200microcalorimeter from MicroCal at 15 °C in 50 mM Hepes, pH 7.5 (at 25 °C),150 mM NaCl by titrating protein into ligand solutions (reverse titrations),and data were corrected for protein heats of dilution and deconvolutedusing the MicroCal Origin software as previously described (1, 7). Dissocia-tion constants and thermodynamic parameters are listed in Tables S2 and S3.

Fluorescent Recovery After Photobleaching. Fluorescent recovery after pho-tobleaching (FRAP) studies were performed in U2OS cells transfected withmammalian overexpression constructs encoding GFP chimeras of BRD3, usinga Zeiss LSM 710 scanhead coupled to an inverted Zeiss Axio Observer.Z1microscope equipped with a high-numerical-aperture (N.A. 1.3) 40× oil im-mersion objective equipped with a heated chamber set to 37 °C, usinga protocol modified from previous studies (7).

Crystallization, Data Collection, and Structure Refinement. Cocrystallizationand structure determination for complexes of BET BRDs with RVX-OH andRVX-208 were carried out following published procedures (7). Data collectionand refinement statistics can be found in Table S4.

Cell Culture and RNA Extraction. HepG2 cells were treated so that a finalconcentration of 0.1% DMSO was achieved. Cells were harvested, washed,and lysed in situ using standard protocols. Total RNA was extracted andprepared using RNeasy columns, and RNA was quantified and quality con-trolled using a Nanodrop spectrophotometer.

DNA Microarray Analysis. An Affymetrix GeneChip WT Terminal Labeling andControls Kit was used according to themanufacturer’s instructions and processed

AR

EG

ER

EG

INH

BE

NR

1H4

BR

D2

MY

BL1

MY

LIP

ZNF1

17

0

1

2

3

Rel

ativ

e Q

uant

ity

0

1

2

3

0.00

0.78

1.56

3.12

6.25

12.5

25.0

[RVX-208]/(µM)

0.00

[(+)-JQ1]/(µM)

0.03

0.06

0.13

0.25

0.50

1.00

(+)-JQ1

RVX-208

AR

EG

ER

EG

INH

BE

NR

1H4

BR

D2

MY

BL1

MY

LIP

ZNF1

17

0

1

2

0

1

2

Rel

ativ

e Q

uant

ity

0

1

2

Time (h) 2 4 8 24 30 48

Not-treated

(+)-JQ1

RVX-208

RVX-208

-3 -2 -1 0 1 2 3Log Fold Change

02

46

810

-log

(p)

10

BRD2

ZNF117

MYLIPNR1H4

MYBL1

AREG

INHBEEREG

ApoA102

46

810

-log

(p)

10

(+)-JQ1

-3 -2 -1 0 1 2 3Log Fold Change

BRD2

MYLIP

MYBL1

INHBE ZNF117

AREG

EREG

NR1H4

ApoA1

qRTPCR

Down-regulatedUp-regulated

HIST2H4ARABL2ACYP1A1KRCC1RFC1DHRS2SRP9RPL23AP82ZNF737MYBL1SLFN5HIST1H2BKZNF43KLRAP1STOX1HIST1H2AMST3GAL3HIST1H2BDC6orf203NPIPDNAJC28PNPLA8LINC00467ZNF280CSNORD59BHIST1H2BCTSTZNF117RPS3ATNS4LIFCACNB4LOC151009EREGFLJ11710DGKKFAM209BSH3TC2PCDHA13TBX4CYP4Z1MAB21L3ZNF208VPS52LINC00652

HIST2H4ASLFN5ARRDC4KRCC1ZSWIM6CNNM4ADARB1TXNIPWDR47HEXIM1MYBL1C1orf63DHRS2SERPINB9BTG1VAMP1STXBP1GTF2BKLHL28HSPA1AHNF4AADAT2ARSIKLHL14ARID5BZBTB38DHRS3DUSP16CCR6NAV2DKK1IRS1ZNF114GCNT3KANK1SLC16A6C3orf33GPAMTNS4HNF1ALOC151009PALMDMAB21L3ART4AREG

-3 0 3

(+)-JQ1 RVX-208

DH

RS

2

CY

P1A

1

KR

CC

1

LIN

C00

467

HIS

T2H

4A

MY

BL1

MA

B21

L3

ZNF2

80C

ST3

GA

L3

TNS

4

-2-1012

Log

Fold

Cha

nge

CN

NM

4

ZSW

IM6

HN

F1A

DH

RS

2

HE

XIM

1

TXN

IP

GC

NT3

SLC

16A

6

TNS

4

AR

EG

-2-1012

(+)-JQ1RVX-208

Log

Fold

Cha

nge

342

314

(+)-JQ1(+)-JQ1

RVX-208

370

128

(+)-JQ1(+)-JQ1RVX-208

712424

(+)-JQ1(+)-JQ1

RVX-208

Total

Down

UpDifferentially ExpressedGenes (> 1.5-fold)

A

C D

FE

B

Fig. 4. Microarray analysis of the transcriptionaleffect following 4-h treatment of HepG2 cells with5 μM RVX-208 or 0.5 μM (+)-JQ1. (A) Venn diagramof statistically significant (P < 0.05) differentialexpression of genes that are up- or down-regu-lated with a 1.5-fold change (or greater) when cellsare treated with compound. The transcriptionaleffect of (+)-JQ1 is 10x higher than that of RVX-208. (B) Log-fold change for the top 10 statisticallysignificant (+)-JQ1 (Upper) or RVX-208 (Lower)regulated genes, highlighting the weaker effect ofRVX-208 on transcription. (C) Top 45 up-/down-regulated genes in the case of (+)-JQ1 (Left) andRVX-208 (Right) treated cells. The color scale in theInset represents log-fold change of expressioncompared with the untreated control. (D) Volcanoplot of the top 1,000 genes that are up-/down-regulated in the case of (+)-JQ1 (Left) and RVX-208(Right) after 4 h of treatment with 0.5 μM (+)-JQ1or 5 μM RVX-208. Top genes are sorted by theirfold-change and are highlighted and colored whenthey are up-regulated (red) or down-regulated(blue). Genes verified by qPCR are highlighted. (E)Quantitative real-time PCR on eight representativegenes identified in A using gene-specific primers(Table S5). Cells were treated with 0.5 μM (+)-JQ1or 5 μM RVX-208. Gene expression was monitoredfor 48 h, as indicated in the Inset. (F) Dose-dependentgene expression on the same set of eight genes testedin E, measured at 4 h. Even at the highest concen-tration, RVX-208 had only a modest transcriptionaleffect compared with (+)-JQ1. Error bars in E and Frepresent SD from triplicate experiments.

19758 | www.pnas.org/cgi/doi/10.1073/pnas.1310658110 Picaud et al.

on an Affymetrix GeneChip Fluidics Station 450 and Scanner 3000. Data wereprocessed in R using Bioconductor (46). Background correction and normaliza-tion were carried out using the Robust Multichip Array (RMA) (47). A linearmodel was applied (limma) followed by empirical Bayesian analysis, and geneswere considered differentially expressed if the adjusted P value, calculated usingthe Benjamini–Hochberg method (48) to minimize false discovery rate, was lessthan 0.05 and the mean level of expression was greater than 1.5-fold.

ACKNOWLEDGMENTS. We are grateful for support received by the Struc-tural Genomics Consortium, a registered charity (number 1097737) thatreceives funds from the Canadian Institutes for Health Research, the CanadaFoundation for Innovation, Genome Canada, GlaxoSmithKline, Pfizer, EliLilly, Takeda, AbbVie, the Novartis Research Foundation, the OntarioMinistry of Research and Innovation, and the Wellcome Trust (092809/Z/10/Z). P.F. and S.P. are supported by Wellcome Trust Career DevelopmentFellowship (095751/Z/11/Z).

1. Filippakopoulos P, et al. (2012) Histone recognition and large-scale structural analysisof the human bromodomain family. Cell 149(1):214–231.

2. Vidler LR, Brown N, Knapp S, Hoelder S (2012) Druggability analysis and structuralclassification of bromodomain acetyl-lysine binding sites. J Med Chem 55(17):7346–7359.

3. Muller S, Filippakopoulos P, Knapp S (2011) Bromodomains as therapeutic targets.Expert Rev Mol Med 13:e29.

4. Dhalluin C, et al. (1999) Structure and ligand of a histone acetyltransferase bromo-domain. Nature 399(6735):491–496.

5. Owen DJ, et al. (2000) The structural basis for the recognition of acetylated histone H4by the bromodomain of histone acetyltransferase gcn5p. EMBO J 19(22):6141–6149.

6. Morinière J, et al. (2009) Cooperative binding of two acetylation marks on a histonetail by a single bromodomain. Nature 461(7264):664–668.

7. Filippakopoulos P, et al. (2010) Selective inhibition of BET bromodomains. Nature468(7327):1067–1073.

8. Dawson MA, et al. (2011) Inhibition of BET recruitment to chromatin as an effectivetreatment for MLL-fusion leukaemia. Nature 478(7370):529–533.

9. Nicodeme E, et al. (2010) Suppression of inflammation by a synthetic histone mimic.Nature 468(7327):1119–1123.

10. Fish PV, et al. (2012) Identification of a chemical probe for bromo and extra C-ter-minal bromodomain inhibition through optimization of a fragment-derived hit.J Med Chem 55(22):9831–9837.

11. Maruyama T, et al. (2002) A mammalian bromodomain protein, Brd4, interacts withreplication factor C and inhibits progression to S phase. Mol Cell Biol 22(18):6509–6520.

12. Dey A, Nishiyama A, Karpova T, McNally J, Ozato K (2009) Brd4 marks select genes onmitotic chromatin and directs postmitotic transcription. Mol Biol Cell 20(23):4899–4909.

13. French CA (2012) Pathogenesis of NUT midline carcinoma. Annu Rev Pathol 7:247–265.

14. Rahl PB, et al. (2010) c-Myc regulates transcriptional pause release. Cell 141(3):432–445.

15. You J, et al. (2009) Regulation of aurora B expression by the bromodomain proteinBrd4. Mol Cell Biol 29(18):5094–5103.

16. Delmore JE, et al. (2011) BET bromodomain inhibition as a therapeutic strategy totarget c-Myc. Cell 146(6):904–917.

17. Zuber J, et al. (2011) RNAi screen identifies Brd4 as a therapeutic target in acutemyeloid leukaemia. Nature 478(7370):524–528.

18. Lockwood WW, Zejnullahu K, Bradner JE, Varmus H (2012) Sensitivity of human lungadenocarcinoma cell lines to targeted inhibition of BET epigenetic signaling proteins.Proc Natl Acad Sci USA 109(47):19408–19413.

19. Cheng Z, et al. (2013) Inhibition of BET bromodomain targets genetically diverseglioblastoma. Clin Cancer Res 19(7):1748–1759.

20. Bartholomeeusen K, Xiang Y, Fujinaga K, Peterlin BM (2012) Bromodomain and extra-terminal (BET) bromodomain inhibition activate transcription via transient release ofpositive transcription elongation factor b (P-TEFb) from 7SK small nuclear ribonu-cleoprotein. J Biol Chem 287(43):36609–36616.

21. Zhang W, et al. (2012) Bromodomain-containing protein 4 (BRD4) regulates RNApolymerase II serine 2 phosphorylation in human CD4+ T cells. J Biol Chem 287(51):43137–43155.

22. Chen R, Liu M, Zhang K, Zhou Q (2011) Isolation and functional characterization ofP-TEFb-associated factors that control general and HIV-1 transcriptional elongation.Methods 53(1):85–90.

23. Gamsjaeger R, et al. (2011) Structural basis and specificity of acetylated transcriptionfactor GATA1 recognition by BET family bromodomain protein Brd3. Mol Cell Biol31(13):2632–2640.

24. Denis GV (2010) Bromodomain coactivators in cancer, obesity, type 2 diabetes, andinflammation. Discov Med 10(55):489–499.

25. Wang F, et al. (2010) Brd2 disruption in mice causes severe obesity without Type 2diabetes. Biochem J 425(1):71–83.

26. Shang E, Nickerson HD, Wen D, Wang X, Wolgemuth DJ (2007) The first bromodo-main of Brdt, a testis-specific member of the BET sub-family of double-bromodomain-containing proteins, is essential for male germ cell differentiation. Development134(19):3507–3515.

27. Bailey D, et al. (2010) RVX-208: A small molecule that increases apolipoprotein A-I andhigh-density lipoprotein cholesterol in vitro and in vivo. J Am Coll Cardiol 55(23):2580–2589.

28. Nicholls SJ, et al. (2012) ApoA-I induction as a potential cardioprotective strategy:Rationale for the SUSTAIN and ASSURE studies. Cardiovasc Drugs Ther 26(2):181–187.

29. McNeill E (2010) RVX-208, a stimulator of apolipoprotein AI gene expression for thetreatment of cardiovascular diseases. Curr Opin Investig Drugs 11(3):357–364.

30. Joy TR (2012) Novel HDL-based therapeutic agents. Pharmacol Ther 135(1):18–30.31. Chung CW, et al. (2011) Discovery and characterization of small molecule inhibitors of

the BET family bromodomains. J Med Chem 54(11):3827–3838.32. Mirguet O, et al. (2012) From ApoA1 upregulation to BET family bromodomain in-

hibition: Discovery of I-BET151. Bioorg Med Chem Lett 22(8):2963–2967.33. Hansen H (2008) Compounds for the prevention and treatment of cardiovascular

diseases. International Patent WO 2008/092231 A1.34. Bamborough P, et al. (2012) Fragment-based discovery of bromodomain inhibitors

part 2: optimization of phenylisoxazole sulfonamides. J Med Chem 55(2):587–596.35. Picaud S, et al. (2013) PFI-1, a highly selective protein interaction inhibitor, targeting

BET Bromodomains. Cancer Res 73(11):3336–3346.36. Yang Z, et al. (2005) Recruitment of P-TEFb for stimulation of transcriptional elon-

gation by the bromodomain protein Brd4. Mol Cell 19(4):535–545.37. Gaucher J, et al. (2012) Bromodomain-dependent stage-specific male genome pro-

gramming by Brdt. EMBO J 31(19):3809–3820.38. Matzuk MM, et al. (2012) Small-molecule inhibition of BRDT for male contraception.

Cell 150(4):673–684.39. Berkovits BD, Wolgemuth DJ (2011) The first bromodomain of the testis-specific

double bromodomain protein Brdt is required for chromocenter organization that ismodulated by genetic background. Dev Biol 360(2):358–368.

40. Lamonica JM, et al. (2011) Bromodomain protein Brd3 associates with acetylatedGATA1 to promote its chromatin occupancy at erythroid target genes. Proc Natl AcadSci USA 108(22):E159–E168.

41. Tsume M, et al. (2012) Brd2 is required for cell cycle exit and neuronal differentiationthrough the E2F1 pathway in mouse neuroepithelial cells. Biochem Biophys ResCommun 425(4):762–768.

42. Wu SY, Lee AY, Lai HT, Zhang H, Chiang CM (2013) Phospho switch triggers Brd4chromatin binding and activator recruitment for gene-specific targeting. Mol Cell49(5):843–857.

43. Tillett D, Neilan BA (1999) Enzyme-free cloning: A rapid method to clone PCR prod-ucts independent of vector restriction enzyme sites. Nucleic Acids Res 27(19):e26.

44. Hansen HC, Wagner GS, Attwell SC, McLure KG, Kulikowski EB (2010) Novel antiin-flammatory agents. International Patent WO 2010/123975 A1.

45. Wigle TJ, et al. (2010) Screening for inhibitors of low-affinity epigenetic peptide-protein interactions: An AlphaScreen-based assay for antagonists of methyl-lysinebinding proteins. J Biomol Screen 15(1):62–71.

46. Gentleman RC, et al. (2004) Bioconductor: Open software development for compu-tational biology and bioinformatics. Genome Biol 5(10):R80.

47. Irizarry RA, et al. (2003) Exploration, normalization, and summaries of high densityoligonucleotide array probe level data. Biostatistics 4(2):249–264.

48. Benjamini Y, Hochberg Y (1995) Controlling the false discovery rate: A practical andpowerful approach to multiple testing. J R Stat Soc, B 57(1):289–300.

Picaud et al. PNAS | December 3, 2013 | vol. 110 | no. 49 | 19759

BIOCH

EMISTR

YCH

EMISTR

Y