A Decade of Histone Acetylation: Marking Eukaryotic Chromosomes with Specific Codes

Abstract

Post-translational modification of histones, a major protein component of eukaryotic chromosomes, contributes to the epigenetic regulation of gene expression. Distinct patterns of histone modification are observed at specific chromosomal regions and affect various reactions on chromosomes (transcription, replication, repair, and recombination). Histone modification has long been proposed to have a profound effect on eukaryotic gene expression since its discovery in 1964. Verification of this idea, however, was difficult until the identification of enzymes responsible for histone modifications. Ten years ago (1995), histone acetyltransferases (HATs), which acetylate lysine residues in histone amino-terminal tail regions, were isolated. HATs are involved in the regulation of both promoter-specific transcription and long-range/chromosome-wide transcription. Analyses of HATs and other modification enzymes have revealed mechanisms of epigenetic regulation that are mediated by post-translational modifications of histones. Here we review some major advances in the field, with emphasis on the lysine specificity of the acetylation reaction and on the regulation of gene expression over broad regions.

1. Introduction: Before 1995

The nucleosome, a fundamental unit of eukaryotic chromosomes, is composed of DNA and histone proteins (Fig. 1) (1, 2). Histones have a mass roughly equal to that of the DNA which they are associated with. Each of the core histones (H2A, H2B, H3, and H4) exhibits a similar structural feature called the “histone fold,” which consists of a long central α-helix flanked by shorter helices and loops that interact with DNA (3). All the core histones have 15–30 unstructured residues at their amino termini, which are commonly referred to as “tails”; their carboxy termini consist of similarly unstructured tails.

The tail regions of core histones are subject to various post-translational modifications that are considered to be key reactions in the modulation of chromatin structure and function (Fig. 1A). Post-translational modifications of histones include the acetylation of lysine residues, lysine and arginine methylation, serine and threonine phosphorylation, lysine ubiquitination, lysine sumoylation, and the poly-ADP-ribosylation of glutamic acid (4, 5). Among these, acetylation has been the primary subject of research.

Histone acetylation was first discovered by Allfrey et al. in 1964 and proposed to regulate gene expression (6). This idea was supported by the observation that hyper-acetylation of histones correlates with transcriptional activation (7). Because acetylation of lysine residues neutralizes their positive charges, it was assumed that a decrease in the electrostatic interaction between DNA and histones is the major acetylation-dependent mechanism that regulates gene expression.

Turner et al. observed acetylation of distinct lysine residues in specific chromosomal regions in Drosophila melanogaster polytene nuclei (Fig. 1B) (8). Lysine 5 of histone H4 (H4-K5) or H4-K8 is frequently acetylated in euchromatic regions, where transcription is potentially active. In contrast, acetylation of H4-K12 is increased in heterochromatic regions, where transcription is potentially inactive. Acetylation of H4-K16 is found along the transcriptionally hyperactive male X chromosome (8). These observations suggested that residue-specific acetylation, rather than bulk neutralization of electrostatic charge, is important in regulation of gene expression through histone modification. They further suggested that histone modification mediates not only promoter-specific gene expression but also longer-range (and even chromosome-wide) gene expression.

Although assumed for more than 30 years, the cause-and-effect relationship between histone acetylation and transcriptional activity was not confirmed, mainly because the enzymes responsible for histone acetylation were unknown. Therefore the identification of such enzymes was a major objective in the field of eukaryotic gene expression.

2. Identification of HATs

In 1995, Brownell and Allis developed a histone acetyltransferase (HAT) assay and detected a single polypeptide of 55 kDa (p55) in macronuclear extracts of Tetrahymena thermophilia (9). Cloning of the cDNA encoding p55 revealed striking sequence similarity of the protein with a yeast transcriptional coactivator, Gcn5 (10), and this latter protein was then found to have HAT activity. The identification of a HAT in the transcriptional coactivator was a breakthrough in understanding the causal relationship between histone acetylation and gene expression.

Another gene encoding a HAT was identified in 1995. Kleff et al. screened for mutants defective in HAT activity among a collection of yeast temperature-sensitive mutants with an enzymatic assay that used fractionated cell extracts (11), which led to the discovery of the Hat1 protein. Because the hat1 mutant conferred no obvious phenotypes other than the enzymatic defect, the identification of Hat1 did not directly link histone acetylation and gene expression. However, Hat1 was found to share structural similarity with Gcn5, and both proteins were later assigned to the GNAT (GCN5-related N-acetyltransferases) superfamily (12). Hat1 was the first HAT whose tertiary structure was solved (13), thus contributing to an understanding of the relationship between HAT structure and function.

HAT activities of mammalian proteins were also reported. In 1996, HAT activity of PCAF (p300/CBP-associated factor), a protein that competes with the adenoviral oncoprotein E1A to bind to the coactivator p300/CBP, was reported (14). p300/CBP itself also turned out to be a HAT (15, 16), as did TAF_II250 (CCG1/TAF1), the largest subunit of the general transcription initiation factor TFIID, which is conserved among eukaryotes (17). To date, various types of promoter-associated transcriptional coactivators and proteins in transcription machineries have been tested for HAT activity and some of these were shown to be HATs (Table 1) (18, 19). These findings strengthened the idea that localized acetylation of histones by transcription factors contributes to the activation of promoter-specific gene expression.

Meanwhile, our group identified a novel class of HATs that regulates long-range/chromosome-wide gene expression (20). A member of the MYST (MOZ, YBF2/SAS3, SAS2, Tip60) family of proteins, Tip60, was shown to possess HAT activity with novel substrate specificity. The HAT activities of other members of the MYST family, including Esa1, a yeast counterpart of Tip60, have been confirmed (Table 1). HAT activities for MYST family proteins were anticipated since they contain sequences similar to the acetyl-CoA binding motif (21) in an evolutionarily conserved region, the MYST domain (20). Interestingly, Esa1 has a sequence motif (the “ER motif ”) also found in a histone deacetylase (HDAC), yeast Rpd3 (22). Other MYST family members, yeast Sas2 and Sas3, are involved in long-range gene repression dependent on chromosomal location (23, 24), and Drosophila MOF is involved in hyperactivation of the male X chromosome (25). The identification of MYST proteins as HATs linked histone acetylation and long-range/chromosome-wide gene expression.

3. The residue-specificity of histone acetylation: A two-step classification and the allocation strategy

Many, but not all, lysines in the amino-terminal tail (N-tail) regions of core histones are acetylated in vivo (Fig. 2A) (26, 27). The site specificity of identified HATs is also divergent (18, 28). Analysis of primary sequences in the vicinity of lysines of core histone N-tails revealed that lysines acetylated in vivo can be classified into three classes and six groups (Fig. 2B) (28). This classification distinguishes lysines acetylated in vivo from others, and fits well with the in vitro site specificity of the catalytic domains of HATs. Therefore, it has been hypothesized that the catalytic domains of HATs recognize classes and groups according to this “two-step classification” hypothesis (28, 29).

The hypothesis provides several insights into how HATs select specific lysines. First, the flanking sequences alone are unlikely to determine whether the lysine will be acetylated by a HAT. Some lysines that are not acetylated by HATs in the non-N-tail regions of histones or other proteins meet the classification criteria. HATs also acetylate non-histone substrates (Table 2). Some lysines in these substrates do not fall into any group in the classification. Because the higher order structure of the N-tail regions is considered flexible (3), this structure might be important for the N-tails of histones and non-histone substrates to be acetylated in addition to the flanking sequences of the target lysines. Second, the hypothesis provides a possible explanation for the broad but non-random specificity of Tip60 or p300 (Fig. 2B). For example, there is no apparent and strict consensus sequence for substrate recognition by a class I-specific HAT like Tip60. Because the ‘class I’ lysines consist solely of ‘group A’ and ‘group B’ lysines, the substrate recognition surfaces of the class I-specific HATs may consist of a surface that can recognize both sequences or two distinct surfaces of which one recognizes the consensus sequence of group A and the other that of group B. This speculation, based on the proposed classification, can explain specific lysine selection by HATs without an apparent target consensus sequence.

Most HATs form multisubunit complexes. In 1998, subunit compositions of human PCAF complex and yeast Gcn5 complexes (SAGA and ADA) were reported (30, 31). To date, subunits of various HAT complexes have been identified and revealed to play critical roles in regulating HAT activity, for example by targeting activity to specific chromosomal regions (Table 3) (32). The formation of multisubunit complexes also modulates the substrate specificity of HATs. The catalytic domains of most HATs alone are unable to acetylate the histones in the nucleosomal context. HAT complexes, however, can acetylate nucleosomal histones, and specific subunits of these complexes are required for targeting them to nucleosomes. For example, Gcn5 alone is unable to acetylate nucleosomal histones efficiently, whereas the Gcn5 complexes ADA and SAGA can (33). Ada2 and Ada3, subunits common to these two complexes, are required for their association with and acetylation of nucleosomal histones (34). Similarly, the Epl1 and Yng2 subunits of the Esa1 complex NuA4 are sufficient for Esa1 to acetylate nucleosomal histones (35).

The lysine specificity of HATs is also modulated through multisubunit complex formation. In some cases, the lysine specificity of HAT complexes toward nucleosomal histones is more restricted than that of the catalytic subunit alone toward free histones. This is the case for HATs in the MYST family. Interestingly, there appears to be a systematic pattern underlying the in vivo lysine specificity of MYST family HATs, which we now call “allocation” (Fig. 3A lower). Six lysine residues in histone N-tails (K5 of H2A, K14 of H3, and K5/8/12/16 of H4) are potential targets for acetylation by MYST family HATs (Table 4, Fig. 3A upper) (29, 36). The MYST family has three members (Esa1, Sas2, and Sas3) in Saccharomyces cerevisiae. Esa1 has major effects on H2A-K5 and H4-K5/8/12 acetylation, both in vivo and in the form of multisubunit complexes (37–42), whereas Sas3 affects H3-K14 (43, 44) and Sas2 affects H4-K16 (Table 4 and Fig. 3A, lower) (41, 42, 45, 46). Based on these observations, we propose that there is an “allocation” strategy, such that HATs cover all potential acetylation sites while narrowing down the specificity of each member with little overlap (Fig. 3B, lower). Potential acetylation sites are defined by local structures of catalytic domains of HATs and histone N-tails, which may follow the “two-step classification” hypothesis (Fig. 3B, upper). This strategy does not decrease the number of lysines that are acetylated by all family members, while narrowing the specificity of an individual family member. This may explain why HATs form a family. Since the number of the MYST family members is increased in mammals (Table 1), it is intriguing to examine whether further allocation is observed and how it relates to complexity of gene regulation in mammals.

The formation of multisubunit complexes is also reported to broaden the lysine specificity of HATs. For example, the Gcn5 complexes ADA and SAGA can acetylate histone H2B in vitro, whereas Gcn5 alone cannot (33). Mechanisms underlying this gain of lysine specificity are unknown, although the Ada2 and Ada3 subunits of the complex are necessary and sufficient for broadening the lysine specificity of Gcn5 (34). Tertiary structure analyses of these complexes may provide insight into the underlying mechanisms.

4. The histone code hypothesis

The residue specificity of identified HATs and HAT complexes was a strong evidence against the original proposal that the neutralization of the positive charge of lysine residues and a resultant decrease in electrostatic interactions between DNA and histones is the major consequence of histone acetylation. Instead, the idea that residue-specific modification of histones has unique and specific effects on chromatin function has become widely accepted (8). Post-translational modifications of histones, including acetylation, constitute a code that allows specific interactions or reactions with chromatin-associated components to take place in a chromosomal context. This idea has gained currency as the “histone code hypothesis”, which holds that the code is generated by histone-modifying enzymes of defined specificity and read by nonhistone proteins that bind in a modification-sensitive manner (Table 5 and Fig. 4) (5, 47). The mechanism is likely to be functioning in reactions other than transcription which are regulated by post-translational modification of histones, such as DNA replication, repair and recombination (Fig. 4) (48–50).

The histone codes are read by proteins that interact with histones in modification-dependent manners (Fig. 5A). One group, bromodomain proteins, is considered to bear acetylated histone interaction domains (51). To date, the bromodomains of various proteins have been reported to interact with acetylated histones in a lysine-specific manner in vitro (Table 6). Bromodomain-dependent binding of proteins to acetylated chromatin supports the binding of bromodomains to acetylated histone in vivo (52–56). These bromodomain-containing proteins are components of nucleosome-modulating complexes that also include ATPases and HATs. Acetylation at specific lysines is thought to stabilize these complexes through bromodomain interaction and to stimulate nucleosome remodeling, further acetylation, or the recruitment of TFIID (52, 57). Initial recruitment of a HAT to chromatin may require nucleosome remodeling. In vivo observations indicate that a HAT is recruited to a promoter after a nucleosome remodeling ATPase complex (58, 59). HAT is proposed to be required for the subsequent stable binding of the ATPase complex in these cases (60). HAT may be recruited to chromatin before the ATPase complex and recruit ATPases in some instances (52, 57, 60).

Sir3 (61, 62) and Tup1 (63) are proposed to interact with hypoacetylated histones, and both repress gene expression in S. cerevisiae. Sir3 spreads along chromatin and contributes to gene repression over a range of several kilobases (64). Deacetylation of H4-K16 by an HDAC, Sir2 (65), stimulates binding of Sir3 to chromatin and thus gene silencing, whereas acetylation of this lysine by a MYST-HAT, Sas2, prevents Sir3 from spreading on chromatin and contributes to anti-silencing (41, 42) (see next section). In contrast, Tup1 represses gene expression in a promoter-specific manner. The local recruitment of Tup1 is accomplished by sequence-specific DNA binding proteins such as α2/Mcm1, Mig1 and Sko1 (66). Tup1 acts in concert with a histone H2B/H3-specific HDAC, Hda1, to repress gene expression, possibly by binding to hypoacetylated histones (67).

Other modifications of histones also regulate their interaction with proteins, and these alterations also function as codes (Table 5). For example, methylation of H3-K9 is known to stimulate the binding of chromodomain-containing proteins such as HP1 and Swi6 to chromatin, leading to gene repression (68–70).

Modification of histones also influences other histone modifications, a phenomenon called cross talk (71) (Fig. 5B). In cis-tail crosstalk, a given modification affects modification of neighboring residues by physically stabilizing or inhibiting interaction between enzymes and substrates. For example, phosphorylation of serine 10 of histone H3 (H3-S10) enhances acetylation of H3-K14 by Gcn5 (72). In trans-tail crosstalk, a given modification affects modifications on other histone tails. For example, ubiquitination of H2B-K123 by Rad6/Ubc2 is required for methylation of H3-K4 and H3-K79 (73, 74). The interdependency of histone modifications led Fischle et al. to propose that histone modifications function as binary switches (75).

5. Long-range/chromosome-wide gene expression regulated by histone acetylation

Most HATs identified to date are involved mainly in promoter-specific gene expression. Analyses of these HATs have advanced our understanding of the molecular mechanisms by which histone acetylation controls gene expression at specific promoters. In contrast, the regulation of long-range/chromosome-wide gene expression by histone acetylation is poorly understood (Fig. 6). The MYST-type HAT family is a group of HATs involved mainly in long-range/chromosome-wide gene expression.

Transcriptional silencing/anti-silencing of genes located near telomeres and the cryptic mating loci of yeast S. cerevisiae is a model system of long-range gene regulation in eukaryotes (64, 76–78). Deacetylation of H4-K16 by the Sir2 deacetylase is thought to enhance the binding of Sir3 to chromatin and to cause gene silencing at these regions. However, it was not clear whether deacetylation of this lysine residue was the primary determinant for Sir3 localization and gene silencing in vivo because the HAT acetylating this lysine in vivo was unknown. A MYST protein, Sas2, has been identified as the HAT that acetylates H4-K16 at telomere-distal regions (41, 42). The role of H4-K16 acetylation was evaluated by manipulating the acetylation level in vivo. Mutation of Sas2 and Sir2 caused the boundary between the hyper- and hypo-acetylated regions to advance and recede from its original position. Importantly, the position of this boundary coincided with that of Sir3 localization and of overall transcriptional activity. These results indicate that H4-K16 plays a decisive role in establishing transcriptionally active and inactive regions (Fig. 7A).

More importantly, these results inspired a novel view of chromosomal border establishment (79). A prevailing view of chromosome borders, which we term the “fixed border” model, is that they are ‘walls’ that actively inhibit the function of transcriptional enhancers or silencers between distinct regions on chromosomes (Fig. 7B; reviewed in Ref. 80). Positions of the borders will be defined by specific DNA elements and fixed in chromosomes (Fig. 7C). Analysis of Sas2, however, implied another mechanism that does not require ‘walls’ at chromosomal borders. Instead, it provided evidence for a mechanism (the “negotiable border” model) that defines a border through the balance of opposing enzymatic activities (Fig. 7B) (79). Positions of the borders depend on the strength of each activity and are thus movable (Fig. 7C).

Deacetylation of H4-K16 by Sir2 at the telomere-proximal regions stabilizes binding of Sir3 and Sir4 on the chromosome and inactivates gene expression (76). In contrast, acetylation of H4-K16 by Sas2 at the telomere-distal regions stabilizes binding of a bromodomain containing protein Bdf1 on the chromosome (53). Bdf1 is a component of a chromatin remodeling SWR1 complex, which exchanges conventional histone H2A with a histone variant H2A.Z (Htz1) in nucleosomes (81–83). This recruitment of H2A.Z on chromatin antagonizes telomeric silencing (84). Methylation of H3-K4 by Set1 and H3-K79 by Dot1 at telomere-distal regions may also play a role in anti-silencing by preventing Sir proteins to associate with chromatin of these regions (85–87). Ubiquitination of H2B-K123 by Rad6/Ubc2 should regulate methylation of these lysines (73, 74).

To establish a negotiable border, histone (or DNA) modification enzymes should be recruited to specific regions on chromosomes. A straightforward scenario for enzyme recruitment is to assume specific DNA elements that interact directly or indirectly with these enzymes. We previously predicted that boundary DNA elements (or insulators) originally assumed to define fixed borders may establish negotiable borders instead (79). Consistent with this prediction, a boundary DNA element in vertebrates recruits HATs to prevent the spread of silenced chromatin (88), and DNA elements that recruit histone modification enzymes function as boundary DNA elements in yeast (89). A temporal analysis of silent chromatin assembly in mammalian cells detected the bidirectional spread of histone and DNA modifications nucleated at a specific chromosome region (90). Other explanations of the recruitment of histone/DNA modification enzymes involve repetitive DNA elements and noncoding RNAs (91). Such mechanisms are important in dosage compensation in mammals and Drosophila and in gene silencing mediated by RNA interference (RNAi) (92). Establishment of eukaryotic chromosomal boundaries to regulate long-range/chromosome-wide gene expression involves many mechanisms (93, 94). How these mechanisms act in concert to regulate long-range/chromosome-wide transcriptional activity is one of the major challenges in the field of eukaryotic gene expression.

6. Perspectives

Here, we have reviewed major advances in histone acetylation research with emphases on lysine specificity in the acetylation reaction and on the regulation of long-range/chromosome-wide gene expression. Post-translational modification of histones also plays critical roles in other reactions involving chromatin, such as DNA replication, repair and recombination (48–50). Research in the past ten years has greatly advanced our knowledge not only of histone acetylation but also of acetylation of non-histone proteins, histone modification in general and epigenetic regulation in eukaryotes. Considering its highly packed structure and dynamic modulation throughout the cell cycle, and the tight control over efficient and specific gene expression, the eukaryotic chromosome is still full of mystery. Understanding the post-translational modifications of histones is a key for unlocking its secrets.

We are grateful to H. Fukuda and N. Sano for helpful discussions. We thank Drs. Wendy Zhow (http://www.meet-greatwall.org) and Tim Harrison (http://www.swcivilwar.com) for permission to use the images presented in Figure 7C. During this study, A. K. was a Research Fellow of the Japan Society for the Promotion of Science. This work was supported in part by Grants-in-Aid for Scientific Research from the Ministry of Education, Culture, Sports, Science and Technology of Japan and by Exploratory Research for Advanced Technology (ERATO) of the Japan Science and Technology Agency (JST).

REFERENCES

Author notes

1Laboratory of Developmental Biology, Institute of Molecular and Cellular Biosciences, University of Tokyo, 1-1-1 Yayoi, Bunkyo-ku, Tokyo 113-0032; and 2Horikoshi Gene Selector Project, Exploratory Research for Advanced Technology (ERATO), Japan Science and Technology Agency (JST), 5-9-6 Tokodai, Tsukuba, Ibaraki 300-2635.