Molecular Cell, Vol. 13, 33–43, January 16, 2004, Copyright 2004 by Cell Press Selective Recognition of Acetylated Histones by Bromodomain Proteins Visualized in Living Cells Tomohiko Kanno,1,3 Yuka Kanno,1,4 divisions in order to take part in stable transcriptional Richard M. Siegel,2,5 Moon Kyoo Jang,1 memory such as in developmental processes and cellu- Michael J. Lenardo,2 and Keiko Ozato1,* lar differentiation. Evidence suggests that the covalent 1Laboratory of Molecular Growth Regulation modifications of histone tails serve as physical interac- National Institute of Child Health tion surfaces on nucleosome assemblies to recruit chro- and Human Development matin remodeling proteins and transcription regulatory 2 Laboratory of Immunology proteins (Strahl and Allis, 2000; Turner, 2002). National Institute of Allergy Acetylation of histones correlates with gene activation and Infectious Diseases and can be recognized by chromatin-associated pro- National Institutes of Health teins containing bromodomains (Dhalluin et al., 1999; Bethesda, Maryland 20892 Jacobson et al., 2000; Owen et al., 2000). The bromodo- 3 Department of Microbiology and Immunology main is a module of about 110 amino acids that is con- Tohoku University School of Medicine served in many chromatin-associated proteins including Sendai 980-8575 histone acetyltransferases (HATs) such as PCAF and Japan TAFII250, the BET family of nuclear proteins such as Brd2, Brd4, and Bdf1, and ATP-dependent chromatin remodeling factors such as BRG-1 (Jeanmougin et al., Summary 1997). In yeast, bromodomains are involved in gene acti- vation, and in antisilencing of genes at heterochromatin Acetylation and other modifications on histones com- boundaries through chromatin association (Hassan et prise histone codes that govern transcriptional regula- al., 2002; Ladurner et al., 2003; Matangkasombut and tory processes in chromatin. Yet little is known how Buratowski, 2003). Brd4, a mammalian BET protein different histone codes are translated and put into structurally similar to Brd2, has been shown to associate action. Using fluorescence resonance energy transfer, with chromatin (Dey et al., 2000, 2003). Also, an acetyla- we show that bromodomain-containing proteins rec- tion-dependent recruitment of bromodomain-protein ognize different patterns of acetylated histones in in- complexes to chromatin-associated DNA targets has tact nuclei of living cells. The bromodomain protein been demonstrated when nucleosomes were reconsti- Brd2 selectively interacted with acetylated lysine 12 tuted in vitro (Agalioti et al., 2002; Hassan et al., 2002). A variety of HATs exhibit unique substrate specificity, on histone H4, whereas TAFII250 and PCAF recognized H3 and other acetylated histones, indicating fine spec- discriminating between individual lysine residues of his- ificity of histone recognition by different bromodo- tones when assayed in vitro (Roth et al., 2001). Thus the mains. This hierarchy of interactions was also seen in potential exists for creating diverse patterns of acetyla- direct peptide binding assays. Interaction with acet- tion, thereby establishing fine specificity within the his- ylated histone was essential for Brd2 to amplify tran- tone code. This possibility is further strengthened by the scription. Moreover association of Brd2, but not other implication of essential roles of H3 and H4 acetylations in bromodomain proteins, with acetylated chromatin transcription (Howe, et al., 2001; Agalioti et al., 2002; An et al., 2002; Smith et al., 2002). Since bromodomains persisted on chromosomes during mitosis. Thus the contained in a variety of transcriptional regulators and recognition of histone acetylation code by bromodo- chromatin remodeling proteins recognize acetylated mains is selective, is involved in transcription, and histones, they are predicted to possess the ability to potentially conveys transcriptional memory across recognize a diverse array of acetylated histone codes. cell divisions. Although several studies have recently addressed this question (Agalioti et al., 2002; Hassan et al., 2002; La- Introduction durner et al., 2003; Matangkasombut and Buratowski, 2003), the levels of specificity and the mechanisms by Covalent modifications of histone proteins, such as which the bromodomains embedded in the whole pro- acetylation, phosphorylation, methylation, and ubiquiti- tein recognize chromatin have yet to be fully resolved. nation of the N-terminal tails of histones, have been The universal histone code created by acetylation as hypothesized to constitute a histone code that controls well as other modifications is likely to be complex. How- patterns of gene expression (Strahl and Allis, 2000; ever, even a simple rule by which bromodomain proteins Turner, 2002). Some histone modifications are short- recognize specific acetylated histones in living cells has lived and implicated in inducible gene activation or re- not been well established. pression. Others may be inherited down through cell We wished to develop an approach that would allow the delineation of general patterns of the interactions *Correspondence: [email protected] between bromodomains and acetylated histones and 4 Present address: Molecular Immunology and Inflammation Branch, also their persistence during cell division in vivo. To this National Institute of Arthritis, Musculoskeletal and Skin Diseases, end, we employed a flow cytometric adaptation of the National Institutes of Health, Bethesda, MD 20892. 5Present address: Autoimmunity Branch, National Institute of Arthri- fluorescence resonance energy transfer technique (FC- tis, Musculoskeletal and Skin Diseases, National Institutes of Health, FRET) recently developed to study cell surface recep- Bethesda, MD 20892. tors (Siegel et al., 2000a, 2000b; Chan et al., 2001). FRET Molecular Cell 34 is based on the ability of a high-energy donor fluoro- phore to transfer energy directly to a lower energy ac- ceptor fluorophore. In FC-FRET, spectral variants of the green fluorescent protein (GFP) that carry out energy transfer are quantified by flow cytometry in living cells. A FRET signal between two GFP variants requires prox- imity at the Angstrom level and can identify specific protein-protein interactions. This technique could allow the detection of bimolecular interactions between bro- modomain proteins in macromolecular complexes and target histones within nucleosomal histone octamers in the intact nuclei of living cells. Results FRET Reveals Specific Histone-Nuclear Protein Interactions in Living Cells To perform FC-FRET analysis, bromodomain proteins were fused to cyan fluorescent protein (CFP) as a donor, and histones were fused to yellow fluorescent protein (YFP) as an acceptor. We first expressed YFP-histone H1 (YFP-H1) or YFP-histone H4 (YFP-H4) in HeLa cells and found that these histones localized to the nucleus during interphase and to chromosomes during mitosis (Figure 1A). YFP-H4 was in the nucleosomal fraction and acetylated as efficiently as the endogenous H4 (Figure 1B), as in the previous reports (Lever et al., 2000; Kimura and Cook, 2001; Ahmad and Henikoff, 2002). Impor- tantly, lysines K5, K8, K12, and K16 were all acetylated in YFP-H4, as reported for endogenous H4 (Turner et al., 1989). All other YFP-histones, including mutants used in the present study, were also incorporated into chroma- tin (data not shown). Thus, ectopically expressed YFP- histones behaved identically to the endogenous his- tones. We first tested a bromodomain protein, Brd2 (formerly Ring3) (Denis and Green, 1996), as an effector molecule that may recognize acetylated histones. Brd2 belongs to the BET family of transcriptional regulators conserved from yeast to humans (Dey et al., 2000; Ladurner et al., 2003; Matangkasombut and Buratowski, 2003). Trans- fected CFP-Brd2 was expressed in reasonable propor- tion to endogenous Brd2 (Figure 1C), localized to the nucleus (Figure 1A), and was associated with an essen- tial component of Mediator, TRAP220 (data not shown) in a similar way to endogenous Brd2 (Jiang et al., 1998). Figure 1. Brd2 Interacts with Histone H4 in Living Cells We then performed FC-FRET between YFP-H4 and CFP- (A) Microscopic images of YFP-H4, YFP-H1, or CFP-Brd2 in HeLa Brd2 in living HeLa cells. Three fluorescent signals were cells. (B) Nucleosomal preparations from parental HeLa cells and those separately measured in each cell: CFP, YFP, and FRET, expressing YFP-H4 were analyzed by immunoblot with antibody where FRET is fluorescence in the YFP emission channel specific for H4, tetra-acetyl H4, or monoacetyl H4. triggered by CFP excitation that was optimally compen- (C) Cells transfected with CFP-Brd2 or untransfected HeLa cells sated as described (Siegel et al., 2000a, 2000b; Chan were blotted with Brd2 antibody. With the 40% transfection effi- et al., 2001). In Figure 1D, a substantial FRET signal ciency observed, the amount of CFP-Brd2 was calculated to be was observed between CFP-Brd2 and YFP-H4, but not approximately 52% of the endogenous Brd2 in transfected cells. (D and E) HeLa cells expressing the indicated fluorescent proteins between CFP-Brd2 and YFP-H1. To compare FC-FRET were analyzed by FC-FRET. In (D), dot plots show FRET versus results among different combinations of proteins, we levels of CFP-Brd2 for all cells. In (E), for each sample, the left panels selected cell populations expressing equivalent levels represent dot plots for levels of CFP- and YFP-fusion proteins. Pink of CFP- and YFP-proteins (Figure 1E). When CFP-Brd2