|Home | About | Journals | Submit | Contact Us | Français|
Acetylation of the ε-amino group of lysine residues, or Nε-lysine acetylation, is an important post-translational modification known to occur in histones, transcription factors and other proteins. Since 1995, dozens of proteins have been discovered to possess intrinsic lysine acetyltransferase activity. Although most of these enzymes were first identified as histone acetyltransferases and then tested for activities towards other proteins, acetyltransferases only modifying non-histone proteins have also been identified. Lysine acetyltransferases form different groups, three of which are Gcn5/PCAF, p300/CBP and MYST proteins. While members of the former two groups mainly function as transcriptional co-activators, emerging evidence suggests that MYST proteins, such as Esa1, Sas2, MOF, TIP60, MOZ and MORF, have diverse roles in various nuclear processes. Aberrant lysine acetylation has been implicated in oncogenesis. The genes for p300, CBP, MOZ and MORF are rearranged in recurrent leukemia-associated chromosomal abnormalities. Consistent with their roles in leukemogenesis, these acetyltransferases interact with Runx1 (or AML1), one of the most frequent targets of chromosomal translocations in leukemia. Therefore, the diverse superfamily of lysine acetyltransferases executes an acetylation program that is important for different cellular processes and perturbation of such a program may cause the development of cancer and other diseases.
The question of how protein functions are regulated in vivo has been and remains a central issue in studies of various biological processes. Acetylation of the ε-amino group of lysine residues, or Nε-lysine acetylation, has recently emerged as an important covalent post-translational modification for regulating protein functions (1–7). Lysine acetylation has been mainly found in eukaryotic cells, but more recently also in archaea and eubacteria (8–11). This type of modification ought to be distinguished from acetylation of the α-amino groups of N-terminal residues, or Nα-terminal acetylation, which occurs in many eukaryotic proteins (12). Nα-terminal acetylation mainly occurs in a co-translational manner and is generally irreversible (12), whereas lysine acetylation is a reversible post-translational process. The dynamic equilibrium of lysine acetylation in vivo is governed by the opposing actions of acetyltransferases and deacetylases. The very first direct links of histone acetyltransferase (HAT) and histone deacetylase to transcriptional co-regulators, made in 1996 (13,14), initiated a gold rush to identify proteins with such enzymatic activities. It is now clear that there are three major classes of histone deacetylases (15–19). In comparison, proteins with HAT activity are more diverse (4,6,20–22). Some known HATs also actetylate other proteins (20,23). Using different protein substrates, several proteins have recently been found to possess intrinsic lysine acetyltransferase activity (24–27). Interestingly, at least two of them are unable to acetylate histones. Since acetyltransferases modifying non-histone proteins have begun to emerge, the generic term ‘lysine acetyltransferases (LATs)’ is used hereafter to refer to enzymes that are able to acetylate specific lysine residues within histones and/or other proteins. In what follows, I will first list various types of proteins modified by lysine acetylation and present an overview of different groups of LATs, with a special focus on the MYST family of acetyltransferases to illustrate how similar catalytic domains are used for different functional purposes in various eukaryotic organisms. I will then discuss how activities of LATs are regulated, what determines their substrate specificity and how lysine acetylation affects protein function. Evidence for the involvement of aberrant HATs in human leukemia will be described at the end, to conclude that LATs are potential molecular targets for therapeutic intervention of leukemia and other diseases caused by abnormal lysine acetylation.
The occurrence of acetyllysine in histones was first discovered in the 1960s (28–30). Core histone proteins were found to be acetylated at the ε-amino nitrogen of specific lysine residues located in the N-terminal tails (reviewed in 31). Importantly, histone acetylation appeared to be associated with active chromatin (28). In 1979, lysine acetylation of high mobility group (HMG) proteins was observed (32). In 1987, Lys40 of α-tubulin from Chlamydomonas was identified as the acetylation site (33). A decade later, in 1997, the tumor suppressor protein p53 was discovered to be acetylated at specific lysine residues in its C-terminal regulatory domain (34). Together with the finding that HATs acetylate the general transcription factors TFIIE and TFIIF in vitro (35), the unexpected discovery of p53 acetylation revealed that HATs are also able to acetylate non-histone proteins, thereby unleashing numerous studies aimed at assessing lysine acetylation in various proteins. It is well recognized now that this type of modification occurs much more widely than anticipated several years ago. As listed in Table Table1,1, over 30 DNA-binding transcription factors have been found to be Nε-acetylated (20,23,36). At least five transcriptional co-regulators are known to be similarly modified (37–41). Reminiscent of autophosphorylation found with protein kinases, several HATs are autoacetylated (42–45). Autoacetylation of PCAF (p300/CBP-associated factor) is important for nuclear localization (46) and acetylation of the general transcription factor TFIIB by itself promotes TFIIF association and transcriptional activation (26). In addition, the chromatin remodeler Brm is acetylated at a lysine residue (47). Although most known Nε-acetylated proteins are histones and transcriptional regulators (Table (Table1),1), this modification also occurs in other cellular proteins, including MCM3 (minichromosome maintenance 3) (24), DNA metabolic enzymes (48–50), the signaling regulator Smad7 (51) and α-tubulin (33). Just as in cellular proteins, Nε-acetylation has been found in viral proteins, including the HIV TAR RNA-binding protein Tat, adenoviral oncoprotein E1A and polyomavirus large T antigen (52–55). Moreover, lysine acetylation is not just unique to eukaryotic and viral proteins. The archaeal architectural protein Alba is Nε-acetylated and this modification blocks oligomerization and DNA binding (8,9). Acetylation of Lys609, a catalytic residue of acetyl-CoA synthetase from the enterobacterium Salmonella enterica, inhibits the enzymatic activity in vivo (10,11). This residue is invariant among a family of AMP-forming enzymes from prokaryotes and eukaryotes, so acetylation may modulate the function of these enzymes (10,11). More amazingly, acetylation of Lys92 dramatically up-regulates the function of Escherichia coli chemotaxis regulator CheY (56,57). Therefore, lysine acetylation has emerged as a general post-translational modification that regulates functions of various cellular and viral proteins.
In 1995, yeast HAT1 (histone acetyltransferase 1) (Table (Table2)2) was identified as the first HAT (58). Although HAT1 was considered to be mainly localized in the cytoplasm to acetylate nascent histones for deposition, recent studies indicate that this acetyltransferase also exists in the nucleus to regulate gene silencing (59). The importance of HATs in gene regulation began to be widely considered in 1996 when HAT activity was shown to be intrinsic to several known transcriptional co-activators such as Gcn5 (general control non-derepressible 5) (13), PCAF (60), p300 (E1A-associated 300 kDa protein) (43), CBP (CREB-binding protein) (43,44) and TAFII250 (TBP-associated factor of 250 kDa) (61). Subsequently, additional proteins were shown to possess HAT activity (Table (Table2).2). As listed in Table Table2,2, these proteins include the nuclear receptor co-activators SRC-1 (steroid receptor coactivator 1) and ACTR (activator of retinoid receptor, also known as AIB1 for amplified in breast cancer 1) (62,63), the transcriptional co-activator CIITA (major histocompatibility class II activator) (64), the DNA-binding transcription factor ATF2 (65), the transcriptional mediator Nut1 (66), the transcription initiation factor TFIIIC (67), the transcription elongation factor Elp3 (68), the yeast protein Hpa2 (69), CDY (chromodomain Y) and its homolog CDYL (CDY-like) (70) and the MYST family of proteins (20,22,71). p300, CBP and PCAF also acetylate transcription factors and other proteins (20,23). Some HATs display autoacetylating activity (42–45). TFIIB acetylates itself, but does not possess any detectable HAT activity (26,69). MCM3AP (MCM3 acetylating protein), Eco1 (establishment of cohesion 1) and ARD1 (a known Nα-acetyltransferase) are three other LATs that were not identified first as HATs (24,25,27). Quite interestingly, acetyl-CoA synthetase is responsible for the acetylation of CheY in E.coli (56,57).
According to sequence similarity, known LATs can be organized into different groups (Table (Table2).2). As one major group of nuclear HATs, the Gcn5/PCAF family is composed of Gcn5, PCAF and related proteins. As illustrated in Figure Figure1A,1A, yeast Gcn5 possesses a HAT domain and a bromodomain and is highly homologous to the C-terminal halves of human PCAF and GCN5L (mammalian GCN5 long form) (60,72–75). Like Drosophila GCN5, mammalian PCAF and GCN5L possess PCAF-specific N-terminal domains (Fig. (Fig.1A)1A) (74). A recent study indicates that GCN5 from the worm Schistosoma mansoni has a similar domain organization (76). Numerous studies indicate that these HATs function as histone- acetylating transcriptional co-activators (4,6,21). Besides histones, PCAF also acetylates non-histone proteins (20,23).
The p300/CBP family is another major group of nuclear HATs that has been extensively characterized (Fig. (Fig.1B)1B) (77–79). Like PCAF, both p300 and CBP are transcriptional co-activators able to acetylate histones and non-histone proteins. Reminiscent of PCAF and GCN5L, p300 and CBP form a pair of homologous HATs in mammals (Fig. (Fig.1).1). Drosophila CBP is larger, but possesses a similar domain organization (Fig. (Fig.1B).1B). As in Drosophila, there is also only one p300/CBP ortholog in Caenorhabditis elegans (78). In Arabidopsis thaliana there are five proteins displaying sequence similarity to the HAT domains of p300 and CBP (Fig. (Fig.1B)1B) (80,81); functions of these novel proteins remain to be determined.
The MYST family of proteins constitutes a third major group of nuclear HATs (Table (Table22 and Fig. Fig.2).2). Compared with the Gcn5/PCAF and p300/CBP groups, the MYST family is larger, more diverse and not so well characterized. Despite their similar HAT domains, MYST proteins play different roles in various cellular processes. In light of their unique structure and function, these acetyltransferases will be discussed in detail in the next section.
In addition to these three major groups of HATs, more than a dozen other proteins have been shown to possess acetyltransferase activity (Table (Table2).2). Like members of the Gcn5/PCAF family, HAT1, Nut1, Elp3, Hpa2/Hpa3, MCM3AP, Eco1 and ARD1 share three or four similar motifs with various N-acetyltransferases and thus belong to the Gcn5-related N-acetyltransferase (GNAT) superfamily (4,6,24,25, 27,82). One of the shared motifs is the classical acetyl-CoA-binding site, which is also present in the MYST family members and ATF2 (6,65,83,84). In contrast, such a motif is absent in other known LATs, so there may be different acetyl-CoA binding modes. Indeed, CDY and CDYL do not have the classical acetyl CoA-binding motif, but display some sequence similarity to several CoA-utilizing enzymes (70).
Most HATs exist as stoichiometric multisubunit complexes in vivo (Table (Table2).2). The complexes are typically more active than their respective catalytic subunits and display distinct substrate specificities (85–88), suggesting that associated subunits regulate the activities of the respective catalytic subunits. In addition, non-catalytic subunits are also involved in recruiting substrates for targeted action to ensure the specificity. Amazingly, one HAT can be the catalytic subunit of multiple complexes. As listed in Table Table2,2, GCN5L forms at least two distinct multisubunit complexes (89–91), and yeast Gcn5 is the catalytic subunit of four complexes (85,92–96; reviewed in 6). Notably, recent studies indicate that Ubp8, a deubiquitinating enzyme present in two Gcn5 complexes, controls the deubiquitination of histone H2B and methylation of histone H3 (97–99). Therefore, the diversity of multisubunit complexes adds another level of complexity to the already diverse superfamily of LATs.
Some known LATs display weak activity towards substrates tested, so an interesting question is whether low levels of activity observed in vitro have any biological significance. One possibility is that the weak activity is not intrinsic, but rather due to an associated HAT. For example, the HAT activity observed with BRCA2 appears to be from associated PCAF (100). It is noteworthy that a weak or null activity observed with a potential LAT in vitro could also be a ‘false negative’. A given substrate used may not be the real one since many proteins are acetylated. Moreover, HATs form multisubunit complexes with different activities and substrate specificities (85–87), so activity data obtained with recombinant catalytic subunits could be potentially misleading. Activities of LATs are dynamically regulated by different mechanisms (see below), so the situation could be even more complicated in vivo. For example, neither Sas2 nor its complex acetylates nucleosomal histone H4 in vitro (101), but Sas2 appears to do so in vivo (102,103). Therefore, different experimental approaches are needed to address whether a potential LAT with weak activity in vitro is an authentic one in vivo.
The acronym MYST is from its four founding members: human MOZ (monocytic leukemia zinc finger protein) (83), yeast Ybf2 (renamed Sas3, for something about silencing 3) (84,104), yeast Sas2 (84) and mammalian TIP60 (HIV Tat-interacting 60 kDa protein) (105–107). As illustrated in Figure Figure2A,2A, a third MYST protein in Saccharomyces cerevisiae is Esa1 (essential Sas2-related acetyltransferase 1) (108,109). In Drosophila (Fig. (Fig.2B),2B), there are five members, including Mof (male-absent on the first) (110,111), Enok (Enoki mushroom) (112), Chameau (camel in French) (113) and two uncharacterized MYST proteins (CG6121 and CG1894). In humans (Fig. (Fig.2C),2C), besides MOZ and TIP60, there are hMOF (ortholog of Drosophila Mof), HBO1 (HAT bound to ORC1, a Chameau ortholog) (114) and MORF (MOZ-related factor) (45). Among these, TIP60 is most similar to Drosophila CG6121 and yeast Esa1 (Fig. (Fig.2A–C).2A–C). There are two uncharacterized MYST proteins in A.thaliana (Fig. (Fig.2D)2D) (81). Similar proteins also exist in Schizosaccharomyces pombe and parasitic protozoa (22). Therefore, this family appears to have members in all eukaryotes.
Members of this family possess highly homologous ~370 residue MYST domains (identity, 36–77%; similarity, 54–84%) (Fig. (Fig.2).2). Structural analysis has revealed that the MYST domain of Esa1 uses an acetyl-cysteine intermediate in the acetylation reaction, so the catalytic mechanism involved is different from that shared by members of the GNAT superfamily of acetyltransferases (115). C2HC fingers are present in all MYST domains except for that of Esa1 (Fig. (Fig.2).2). Despite the absence of zinc and three of the four chelating residues, the corresponding region of Esa1 forms a classical TFIIIA-type zinc finger fold (116). The C2HC fingers of Sas3, Mof and MOZ are known to be essential for HAT activity (104,117,118). A mutation in the C2HC finger of Drosophila Enok affects brain development (112). The C2HC finger of Mof binds to nucleosomes in vitro (117), whereas the C2HC finger of TIP60 is essential for interaction with the translocation E26 transforming-specific (ETS) leukemia protein TEL (119). Therefore, C2HC zinc fingers are important in vitro and in vivo.
Besides MYST domains, this family of proteins contains other structural modules (Fig. (Fig.2).2). One such module is the chromodomain (120), conserved among Esa1, Mof and TIP60 (Fig. (Fig.2A–C).2A–C). The chromodomain of Mof binds to roX (RNA on the X) RNAs and targets the complex to the male fly X chromosome for gene dosage compensation (121). In comparison, the chromodomain of HP1 (heterochromatin protein 1) recognizes Lys9-methylated histone H3 (122,123). Therefore, it will be interesting to determine what roles the chromodomains of Esa1 and TIP60 may have. The PHD (plant homeodomain-linked) zinc finger is another recognizable module (Fig. (Fig.2).2). Two PHD fingers are found in the N-terminal parts of Enok, MOZ and MORF (45,83,112,124). The PHD domains of MOZ and MORF are much more similar to each other than to those of Enok and display high sequence homology to Neuro-D4 and Requiem, two potential transcription factors that do not have MYST domains (83). PHD fingers are also known as LAP (leukemia-associated protein) domains and have been found in many chromatin regulators (125,126). Mutations in PHD fingers of several chromatin regulators contribute to a variety of human diseases (127,128). Atypical PHD zinc fingers with similarity to RING fingers have been shown to function as E3 ubiquitin ligases (129), but it is unclear whether this is a common functional feature of all PHD fingers (130). Interestingly, PHD fingers from several chromatin regulators bind to phosphoinositides and are thus implicated in nuclear lipid signaling (131). The PHD zinc fingers of MOZ display phosphoinositide-binding activity, raising the interesting possibility that phosphoinositides act through PHD zinc fingers and regulate the functions of Enok, MOZ and MORF.
Besides PHD fingers, MOZ and MORF share their extreme N-terminal regions with Enok (Fig. (Fig.2B2B and C), referred to as NEMM (N-terminal region in Enok, MOZ or MORF) domains. The NEMM domains of MOZ and MORF, but not that of Enok, display some sequence similarity to the globular domains of linker histones H1 and H5. The functional significance of this similarity remains to be determined. These H1- and H5-like regions, known as H15 domains, may mediate self-association and interaction with core histones and nucleosomes since the globular domains of histones H1 and H5 are known to have similar activities (132,133). In addition, Enok possesses an uncharacterized neurofilament protein-like domain that is missing in other MYST proteins (Fig. (Fig.2B)2B) (112). At the C-terminal ends of MOZ and MORF (Fig. (Fig.2C)2C) are serine- and methionine-rich regions, which are known as SM domains and possess potent transcriptional activation potential (45,134). Therefore, MYST proteins have diverse domain organizations.
Consistent with their diverse domain organizations, MYST proteins exist as distinct multiprotein complexes in vivo. Yeast Esa1, Sas2 and Sas3 are the catalytic subunits of different multiprotein complexes (Table (Table2).2). The Esa1 complex NuA4 contains 12 subunits, including Tra1 (TRRAP homolog 1), actin, Arp4 (actin-related protein 4), Epl1 (Enhancer of Polycomb-like protein 1), Yng2 (homolog of mammalian ING1, for inhibitor of growth 1) and Yaf9 (homolog of the leukemogenic human protein AF9) (94,135,136). Among these subunits, Esa1, Epl1 and Yng2 form a highly active but much smaller core complex, termed Piccolo NuA4 (88). Sas2 and Sas3 are the catalytic subunits of two trimeric complexes: Sas2 associates with Sas4 and Sas5 (AF9 homolog), whereas Sas3 interacts with Taf30 (AF9 homolog) and Yng1 (ING1 homolog) (137–139). Since Esa1 and Sas3 associate with homologs of mammalian ING1, phosphoinositides may bind to the complexes and regulate their functions. In Drosophila, one MYST complex has been characterized (140). Mof in male flies is part of a dosage compensation complex that contains Msl1 (male-specific lethal 1), Msl2, Msl3, Mle (maleless, homolog of mammalian RNA helicase A) and two non-coding RNA molecules, roX1 and roX2. Notably, this is the only HAT complex known to contain RNA. The chromodomains of Mof and Msl3 are able to mediate RNA binding (121). In mammals, the TIP60 complex has been purified and characterized (141,142). It shares similar subunits with the Esa1 complex, including TRRAP (transformation/transcription domain-associated protein), actin, BAF53 (actin-related protein) and EPC (Enhancer of Polycomb-like protein). However, the TIP60 complex also possesses unique subunits such as p400 (SWI2/SNF2-related ATPase), two RuvB-like proteins and DMAP1 (DNMT1-associated protein 1) (141,142). Originally identified as an essential cofactor for the oncogenic transcription factors c-Myc and E2F (143), TRRAP is an ATM/PI-3 kinase-like protein shared by the PCAF (144), GCN5L (89–91), TIP60 (141) and p400 complexes (145) (Table (Table2).2). As the yeast homolog of TRRAP, Tra1 is a common subunit of the Gcn5 and Esa1 complexes (92–96). In contrast, the Sas2, Sas3 and Mof complexes do not possess proteins similar to TRRAP. It is expected that human MOF does not associate with TRRAP, but it remains unclear whether other MYST proteins, such as HBO1, MOZ and MORF, interact with TRRAP.
As suggested by their unique domain organizations and different complex compositions, MYST proteins are involved in regulating various biological processes. In agreement with the fact that the NuA4 subunit Yng2 is homologous to the candidate tumor suppressor ING1 (146,147), Esa1 is important for DNA repair and cell cycle progression (108,109,148). The Esa1 complex has also been linked to epigenetic control, gene regulation and cellular response to spindle stress (135,136). In support of its role in yeast gene silencing (84), Sas2 opposes the action of the deacetylase Sir2 to establish the boundary between euchromatin and heterochromatin (102,103). Sas3 was originally identified as a regulator of gene silencing (84), whereas its complex was recently shown to be involved in regulating transcriptional elongation (137). Drosophila Chameau is important for heterochromatin-mediated gene silencing and Mof is a major player in gene dosage compensation in male flies (110,112). As for mammalian MYST proteins, TIP60 plays an important role in apoptosis and DNA repair (141).
Different lines of evidence suggest that MYST proteins are also targeted to specific promoters to regulate transcription. Through interacting with DNA-binding transcription factors, yeast Esa1 is recruited to regulate the expression of ribosomal proteins (149). The tumor suppressor p53 interacts with ING proteins and its ability to activate transcription in yeast is modulated by Yng1 and Yng2, two ING homologs that are subunits of the Sas3 and Esa1 complexes, respectively (147,150). TIP60 not only functions as a transcriptional co-repressor for TEL (119) and STAT3 (151), but also as a coactivator for androgen receptor (152,153), NF-κB (154) and c-Myc (155). Through a unique NPxY motif within its MYST domain, TIP60 interacts with the WW domain of Fe65 and potentiates transcription mediated by the APP (amyloid-β precursor protein) cytoplasmic domain (154,156–158). Therefore, TIP60 regulates transcription in a context- dependent manner. Originally identified as a protein interacting with ORC1, an ORC (replication origin recognition complex) subunit (114), HBO1 has recently been found to function as a transcriptional co-repressor for androgen receptor (159). Human MOF is part of a complex implicated in activating the B-myb promoter (160). MOZ and MORF possess transcriptional repression and activation domains (45,134,161), suggesting that these two HATs are potential transcriptional co-regulators. Indeed, MORF is present in a transcriptional co-activator complex associated with the nuclear receptor PPARα (162). Both MOZ and MORF physically and functionally interact with Runx1 and Runx2 (161,163,164), two Runt domain transcription factors important for cell growth and differentiation in different tissues (165–168). In agreement with this, down-regulated expression of mouse MORF, known as Querkopf (squarehead in German), leads to defects in osteogenesis and neurogenesis (169). Therefore, compared with members of the Gcn5/PCAF and p300/CBP families, MYST proteins are much more diverse in domain organization, multiprotein complex formation and biological function.
With more information available on the structure and function of different LATs, the regulation of their enzymatic activities has become an important issue in the past few years. Emerging data suggest that multiple mechanisms are involved. First, as an essential cofactor for different acetyltransferases, acetyl-CoA also stabilizes GCN5 and PCAF (42). Second, as described above, formation of stoichiometric multisubunit complexes modulates the specific activities and substrate specificities of different LATs. Third, the enzymatic activities of PCAF, p300 and CBP are regulated by interaction with transcription factors such as p/CIP (170), Twist (171), vIRF (172), Zta (173), HOX proteins (174), PU.1 (175) and the early B-cell factor (176); the ubiquitin ligase MDM2 (177–179); the protein kinase RSK2 (180); and viral proteins like E1A (170,171,181,182), E1B (183), T antigen (184), E7 (185) and Tat (186). Fourth, LATs are subject to covalent modifications such as phosphorylation (65,187), acetylation (37,46), ubiquitination (188–190) and sumoylation (191). Fifth, LATs are degradated by caspases, calpains and ubiquitin-dependent proteasomes (188–190,192). Sixth, subcellular compartmentalization is an important regulatory mechanism for HATs. For example, HAT1 binds to 14-3-3 proteins (193) and TIP60 is sequestered to the cytoplasm in a signal-dependent manner (154,156,158,194). Finally, while p300, CBP, MOZ and MORF possess PHD fingers (Figs (Figs11 and and2),2), yeast Esa1 and Sas3 associate with PHD finger-containing subunits (6,22). PHD fingers are implicated in phosphoinositide binding and may thus provide structural modules for integrating nuclear lipid signals (131), so activities of these acetyltransferases may be regulated by nuclear signaling events.
In addition to the regulatory mechanisms just described, substrate recognition is another point for controlling acetylation. A relevant question is how the specificity is achieved. The classical ‘hit-and-run’ model can be used to depict how HATs recognize free histone substrates (Fig. (Fig.3A).3A). For example, HAT1 acetylates newly synthesized histones in such a mode (195,196). HATs also adopt ‘attract-and-hit’ mechanisms. To bring substrates into their physical proximity, HATs and other LATs may either bind directly to their substrates (Fig. (Fig.3B)3B) or interact with their substrates through adaptor proteins (Fig. (Fig.3C).3C). Since physical proximity increases the local substrate concentration, ‘attract-and-hit’ modes are more efficient than ‘hit-and-run’ ones. For example, when the WD40-repeat histone-binding protein HAT2 is present, HAT1 is 10 times more active (197). With polymer substrates such as chromatin, region-specific acetylation is sometimes necessary (Fig. (Fig.3D).3D). To specifically modify chromatin, HATs are often recruited to carry out targeted acetylation, which is reminiscent of the indirect ‘attract-and-hit’ mechanism (Fig. (Fig.3C).3C). In this scenario, subunits associated with a HAT may modulate the targeting specificity. As stated above, incorporation of one HAT into different complexes adds another level of diversity for substrate targeting. Like chromatin, microtubules are polymers, so a tubulin acetyltransferase may also need to be targeted. While the DNA sequence marks the position of a given nucleosome within a chromatin array and thus dictates region-specific recruitment of HATs for acetylation, microtubules do not have such a marking system. Region-specific tubulin acetylation on a microtubule may be achieved during polymerization. As illustrated in Figure Figure1,1, most members of the Gcn5/PCAF and p300/CBP families contain bromodomains, so these acetyltransferases are able to participate in ‘relay’ reactions (Fig. (Fig.3E),3E), with one enzyme executing the initial acetylation in a polymer substrate to create a binding site for the bromodomain of another enzyme, which then initiates another round of acetylation to trigger a cascade of reactions. This may be one means to execute chromatin domain-, chromosome- and genome-wide acetylation (198). A slightly different scenario is that with acetylating activities coupled to acetyllysine-binding bromodomains, two LATs cooperate with each other to carry out sequential acetylation of different substrates. Such a mode of action has been elegantly demonstrated for p300 and PCAF in their sequential acetylation of Tat and histones (Fig. (Fig.3F)3F) (199,200). Therefore, LATs utilize different means to cope with various types of substrates.
Once a lysine residue is acetylated, a mechanistic question is how such a modification may affect protein function. It appears that both ‘loss-of-function’ and ‘gain-of-function’ mechanisms are involved. Regarding the former, acetylation of the ε-amino group of a lysine residue neutralizes the positive charge, so the modification may affect interaction of the lysine residue with DNA, RNA and proteins. Such a mechanism may operate with chromatin since the DNA backbone is negatively charged. Indeed, histone acetylation has been shown to affect the nucleosomal structure and the stability of nucleosomal arrays (201–204). Alternatively, acetylation may render the ε-amino group unable to form hydrogen bonds. For example, Lys11 of Alba is involved in forming a hydrogen bond important for oligomerization, so acetylation of this residue inhibits oligomerization (8,9). Acetyl-CoA synthetase from S.enterica provides another unique example of a ‘loss-of-function’ effect (10). Lys609 of this enzyme is part of the catalytic center, so acetylation of this residue inhibits the enzymatic activity.
In eukaryotic cells, the ε-amino group of a lysine residue is also subject to methylation and modification by ubiquitin and ubiquitin-like proteins such as NEDD8 and SUMO (Fig. (Fig.4).4). Different modifications are mutually exclusive, thus leading to their potential competition. It has recently been shown for SREBP and Smad7 that acetylation directly competes with ubiquitination for the same lysine residues to increase protein stability (51,205). Acetylation and methylation of histone H3 at Lys9 have completely opposite functional consequences, with the former associated with active chromatin and the latter linked to heterochromatin or inactive chromatin (2,5). Since acetylation blocks methylation, histone deacetylases are needed to remove the acetyl group from Lys9 and clear the way for subsequent methylation and thus the establishment of inactive chromatin (206,207).
As for ‘gain-of-function’ mechanisms, the addition of an acetyl group to a lysine residue creates a new surface for protein association. Reminiscent of domains that recognize phosphoproteins (208,209), bromodomains function as structural modules specific for acetyllysine-containing motifs (Fig. (Fig.4).4). It has been demonstrated that several chromatin regulators use bromodomains to recognize acetyllysine (210–216). Bromodomains are found in many proteins and sequence variations may dictate their binding specificity (217). Some proteins contain multiple bromodomains, which may cooperate with each other to increase the affinity for binding partners with multiple acetylated lysine residues (211). Acetylation is also known to stimulate the association with proteins that do not contain bromodomains (27,218). Different types of structural modules have been identified for phosphotyrosine (209,219), so an interesting question to be addressed is whether there are additional acetyllysine-recognizing modules (Fig. (Fig.44).
Consistent with the essential roles of LATs in different biological processes, molecular and genetic studies have revealed that these enzymes are also important players in human pathology. Among others, the following lines of evidence strongly suggest that several HATs are directly linked to oncogenesis: (i) the viral oncoproteins such as E1A and large T antigen target p300 and CBP (77,78,220); (ii) E1A also interacts with PCAF and TRRAP, a subunit of multiple HAT complexes (6,143,220,221); (iii) the proto-oncoprotein SYT targets p300 (222); (iv) p300, CBP and PCAF associate with and modify various transcription factors, such as p53, Rb, E2F, H1Fα and E2A, that play key roles in controlling different cellular programs (21,77–79); (v) oncogenic transcription factors c-Myc and E2F bind to TRRAP (143); (vi) the tumor suppressor p53 also binds to ING proteins, homologs of which have been found in HAT complexes (147); (vii) MOZ and MORF interact with Runx1 (161,163,164), the most frequent target of leukemia-associated chromosomal translocations (166,168); (viii) TIP60 associates with the androgen receptor and has been implicated in the development of prostate cancer (152,153,223); (ix) TIP60 is involved in regulating apoptosis and its yeast homolog Esa1 is essential for cell cycle progression (108,109,141).
Among known HATs (Table (Table2),2), p300 and CBP have been considered as tumor suppressors (77–79). Consistent with this notion, monoallelic mutation of the CBP locus is the genetic basis for Rubinstein–Taybi syndrome and patients with this syndrome exhibit an increased risk of developing malignant tumors (224). Biallelic mutations of the p300 locus have been identified in human cancers of epithelial origin (225,226) and exogenous expression of p300 is able to suppress the growth of human carcinoma cells in vitro (227). Moreover, p300 and CBP play distinct but essential roles in hematopoiesis, and mice with inactivated alleles of the p300 and CBP loci develop hematological malignancies (228–231).
Studies of chromosomal abnormalities in leukemia patients have provided additional support for the direct involvement of HATs in human cancer. The p300 and CBP genes are located on chromosomes 16p13 and 22q13, respectively, and have been shown to be rearranged in chromosomal translocations associated with leukemia or treatment-related myelodysplastic syndrome (Fig. (Fig.5A).5A). Known fusion partners are MOZ, MORF and MLL (mixed lineage leukemia) (Fig. (Fig.5).5). The MLL gene, located at 11q23, is fused to the p300 and CBP genes in the translocations t(11;22)q23;q13) and t(11;16)(q23;p13), respectively (Fig. (Fig.5A5A and C) (232–238). These translocations lead to the production of different MLL-p300 and MLL-CBP fusion proteins, in which the bromodomain, HAT domain and Q region of p300 or CBP are linked to the N-terminal part of MLL. The MOZ gene, located at 8p11, is fused to that of CBP in two t(8;16)(p11;p13) translocations (Fig. (Fig.5B),5B), which gives rise to MOZ-CBP proteins containing the N- and C-terminal parts of MOZ and CBP, respectively (83,239). In two t(8;22)(p11;q13) translocations, the MOZ gene is fused to the p300 gene, generating MOZ-p300 fusion proteins (240,241). In the t(8;16) and t(8;22) translocations, the reverse transcripts expressing CBP-MOZ or p300-MOZ are not always produced, suggesting that the MOZ-CBP and MOZ-p300 fusion proteins are responsible for the leukemogenesis. These fusion proteins possess the NEMM, PHD and MYST domains of MOZ, as well as the CH1, KIX, bromodomain, HAT domain and Q region of CBP or p300 (Fig. (Fig.5A5A and B). Notably, the SM domain of MOZ is missing from these fusion proteins.
Besides p300 and CBP, other partners are also involved in the fusion with the MOZ gene. In two slightly different inv(8)(p11q13) chromosomal inversions (Fig. (Fig.5C),5C), the MOZ gene is fused to the TIF2 gene located at 8q13, creating a protein with the N-terminal part of MOZ fused to the C-terminal part of TIF2 (242–244). TIF2 is a member of the p160 family of nuclear receptor co-activators known to interact with p300 and CBP. A recent study indicates that the MOZ gene is also rearranged in a t(2;8)(p23;p11) translocation associated with therapy-related myelodysplastic syndrome, but the translocation partner remains to be identified (245).
The sequence similarity between MOZ and MORF suggests that the MORF gene is rearranged in a manner similar to the MOZ gene (45). Indeed, the MORF gene was recently found to be rearranged and fused to the CBP gene in a recurrent t(10;16)(q22;p13) translocation associated with acute myeloid leukemia (Fig. (Fig.5A5A and C) (246,247). A slightly different t(10;16)(q22;p13) translocation, associated with therapy-related myelodysplastic syndrome, also leads to fusion of the MORF gene to the CBP gene (247). The resulting MORF-CBP fusion proteins are structurally similar to the MOZ-CBP and MOZ-p300 fusion proteins described above.
All of the above chromosomal abnormalities suggest that aberrant acetylation by mistargeted HATs plays a causative role in leukemogenesis. Indeed, one MLL-CBP fusion protein has been analyzed in mice and found to generate a myelodysplastic syndrome that evolves into myeloid leukemia (248). The bromodomain and HAT domain of CBP are the only modules needed for the leukemogenic activity, suggesting that the N-terminal part of MLL directs the two CBP domains for aberrant acetylation and subsequently leads to leukemogenesis. MLL is processed by Taspase 1 cleavage and the resulting N- and C-terminal fragments remain associated within the same histone methyltransferase complex (Fig. (Fig.5C)5C) (249). It is unclear whether the MLL-CBP fusion proteins are still able to bind the processed C-terminal fragment of MLL. Potential targets of MLL-CBP are members of the HOX gene family (250,251), so MLL-CBP may lead to abnormal expression of these genes and cause the subsequent development of leukemia (Fig. (Fig.66A).
A MOZ-CBP fusion protein from t(11;16)(q23;p13) has been characterized (161). This fusion protein inhibits Runx1-dependent transcription and blocks the differentiation of murine myeloid M1 cells to macrophages. Moreover, the HAT domain of CBP appears to be important for the repressive activities. Two slightly different MOZ-TIF2 fusion proteins from inv(8)(p11q13) chromosomal inversions have recently been shown to display oncogenic potential in both in vitro transformation and in vivo transplant assays (118). The C2HC zinc finger of MOZ (Fig. (Fig.2)2) is essential, whereas its acetyl-CoA-binding motif only modulates the penetrance and phenotype of the resulting diseases. For TIF2, only its CBP-interacting domain is essential, so the recruitment of CBP to MOZ is responsible for leukemogenesis. This finding suggests that the underlying molecular mechanism is very similar to that used by MOZ-CBP. Therefore, both MOZ-CBP and MOZ-TIF2 may repress Runx1-dependent gene expression and cause the development of leukemia (Fig. (Fig.6B).6B). Although both MLL-CBP and MOZ-CBP are leukemogenic, their mechanisms of action are quite different, with the former being an activator and the latter functioning as a repressor (Fig. (Fig.66).
As a common post-translational modification, lysine acetylation is known to occur not only at the N-terminal tails of core histones but also within other eukaryotic proteins, including about 40 transcription factors, a chromatin remodeler, one DNA replication factor, three DNA metabolic enzymes, a signaling regulator and a cytoskeletal protein (Table (Table1).1). In addition, viral proteins are also Nε-acetylated. Most of these proteins are nuclear, so it will be interesting to examine whether lysine acetylation plays a wider role in the cytoplasm. Intriguingly, this modification has been found in at least three bacterial proteins (Table (Table1),1), so it is not just unique to eukaryotic and viral proteins. Consistent with the wide spectrum of substrates, a highly diverse superfamily of LATs has been identified (Table (Table2).2). However, enzymes responsible for the acetylation of α-tubulin, Alba and acetyl-CoA synthetase remain elusive. Most known acetyltransferases are catalytic subunits of multiprotein complexes. With distinct sets of subunits, one acetyltransferase can be the catalytic subunit of different complexes, thereby adding another level of diversity. Biochemical and molecular approaches have been and will continue to be fruitful in analyzing the function and regulation of acetyltransferase complexes (6). To understand how lysine acetylation regulates different cellular processes in vivo, genetic analysis has been invaluable to distinguish between functional differences of homologous HATs. Gene targeting in mice has revealed that GCN5L but not PCAF is essential in early embryonic development and that p300 and CBP play distinct roles in hematopoeisis (228,230,252,253). Therefore, such an approach shall continue to yield novel insights into the biological function as well as the spatial and temporal regulation of these and other LATs.
As for molecular mechanisms by which lysine acetylation exerts its effects, one major advance made in the past few years is the discovery that bromodomains are able to recognize acetylated lysine residues (210–216). Many proteins possess such modules (254,255), so one interesting question is whether all bromodomains have such an ability. If so, one major challenge would be to determine the binding specificity of different bromodomains. Degenerate peptide libraries have been successfully used to determine the binding specificity of structural modules involved in cytoplasmic signaling (256). Similar strategies should be valuable for establishing the binding specificity of different chromodomains and understanding their functions in recognizing specific acetylation signals.
The genes for CBP, p300, MOZ and MORF are rearranged by leukemia-associated chromosomal abnormalities (Fig. (Fig.5).5). Some of the resulting fusion proteins have been shown to be leukemogenic (118,161,248). TIP60 has been implicated in the development of prostate cancer (223) and it also interacts with and modulates the function of the cytoplasmic domain of APP, an important regulator of Alzheimer’s disease (154,156–158). Expanded polyglutamine tracks, a root cause of Huntington and other polyglutamine diseases, target the acetyltransferase activity of CBP (257–260). In addition, lysine acetylation affects the important roles that various proteins play in other human diseases (Tables (Tables11 and and2).2). For example, acetylation of HIV Tat is an essential regulatory step for virus production. Therefore, LATs play important roles in the pathogenesis of leukemia (Figs (Figs55 and and6)6) and other diseases.
Small molecule inhibitors and activators of histone deacetylases have been extensively explored for the treatment of cancer and other human diseases (17,261,262). The direct involvement of LATs in leukemia and other diseases suggests that small molecules able to modulate their enzymatic activities should be of therapeutic potential. In contrast to histone deacetylase inhibitors, only a few small molecules have been shown to modulate the activity of HATs (263–265). Model organisms such as S.cerevisiae possess various HAT orthologs (Figs (Figs11 and and2)2) and may be used to screen for cell-permeable compounds using similar assays to those described for histone deacetylases (266,267). The resulting compounds will yield lead structures for further optimization with iterative rounds of structure-based molecular design and chemical synthesis. Therefore, studies of the fundamental process of lysine acetylation will not only yield important further insights into how post-translational modifications regulate various cellular processes, but also shed new light on the development of novel therapeutic means for the treatment of leukemia and other human diseases.
This work was supported by funds from the Canadian Cancer Society through the National Cancer Institute of Canada.