Structure of histone acetyltransferases¹

Abstract

Histone acetyltranferase (HAT) enzymes are the catalytic subunits of multisubunit protein complexes that acetylate specific lysine residues on the N-terminal regions of the histone components of chromatin to promote gene activation. These enzymes, which now include more than 20 members, fall into distinct families that generally have high sequence similarity and related substrate specificity within families, but have divergent sequence and substrate specificity between families. Significant insights into the mode of catalysis and histone substrate binding have been provided by the structure determination of the divergent HAT enzymes Hat1, Gcn5/PCAF and Esa1. A comparison of these structures reveals a structurally conserved central core domain that mediates extensive interactions with the acetyl-coenzyme A cofactor, and structurally divergent N and C-terminal domains. A correlation of these structures with other studies reveals that the core domain plays a particularly important role in histone substrate catalysis and that the N and C-terminal domains play important roles in histone substrate binding. These correlations imply a related mode of catalysis and histone substrate binding by a diverse group of HAT enzymes.

Introduction

The eukaryotic genome is packaged into the compact state of chromatin that forms the scaffold from which the fundamental nuclear processes of transcription, replication and DNA repair occur. Chromatin is composed of nucleosomes that are comprised of 145–147 base-pairs of DNA wrapped around an octameric core containing two molecules each of histones H2A, H2B, H3 and H41. The H1 linker histone stabilizes the assembly of the octameric core into higher-order structures characteristic of chromatin. Each core histone contains a carboxyl-terminal, highly helical globular domain that comprises about 75 % of the amino acid content and forms the interior core of the nucleosome particle. The remaining amino-terminal portion of the core histone proteins contain flexible and highly basic tail regions that are highly conserved across various species and that provide the sites for several different types of post-translational modification, including methylation, ADP-ribosylation, phosphorylation, ubiquitylation and acetylation2, 3. Such post-translational histone modifications have long been correlated with various nuclear activities, including replication, chromatin assembly and transcription4, 5, 6.

The post-translational histone modification that has been most extensively studied is histone acetylation. It has been nearly 40 years since Allfrey and co-workers proposed that the acetylation state of histones within chromatin is correlated with gene regulation2, 7, whereby genes containing hypoacetylated histones were transcriptionally repressed, while genes containing hyperacetylated histones were transcriptionally active. Several subsequent studies have reinforced this proposal8, 9, 10, 11, 12. Despite the time that has elapsed since the initial studies by Allfrey and co-workers, the proteins that mediate histone acetylation have been identified only recently, and structural information has become available only over the last two years. A major breakthrough in understanding the mechanism of histone acetylation came with the cloning of a histone acetyltransferase enzyme from Tetrahymena13 that was a homologue of the previously identified Gcn5 transcriptional coactivator/adaptor from yeast14, 15. Functional characterization of yeast Gcn5 mutants revealed a direct correlation between the ability of the protein to acetylate histones and to activate transcription16, 17. Subsequently, a flurry of studies led to the discovery of a large number of histone acetyltransferase (HAT) enzymes, many of which, such as CBP/p30018, 19, TAFII25020, and SRC-121, were previously identified transcriptional coactivators. Almost in parallel with the discovery of HAT enzymes, came the identification of several histone deacletylase (HDAC) enzymes22, whose activities have been correlated with transcriptional repression. These findings underscore the key role in gene regulation that is played by the enzymes that modulate the acetylation state of histones.

Here, I will review what is known about the sequence and functional properties of HAT enzymes, and what we have learned by correlating this information with the structure of three divergent HAT proteins Hat1, Gcn5/PCAF, and Esa1. I will discuss how a correlation of the structural and functional studies leads to a framework for understanding how HAT enzymes catalyze acetylation and how they bind histone, and possibly also non-histone, substrates. Finally, I will address how these studies may lead ultimately to a better understanding of how histone acetylation leads to transcriptional activation.

A large number of transcription factors are now known to have HAT activity (Table 1). Sequence analysis of these proteins reveals that they fall into distinct families that show high sequence similarity within families but poor to no sequence similarity between families23. Moreover, between families the HAT domains appear to mediate different biological functions, come in different sizes and appear in the context of a different set of other conserved protein domains (Figure 1). For example, the Gcn5/PCAF family of HAT proteins function as coactivators for a subset of transcriptional activators, contain a roughly 160 residue HAT domain and directly C-terminal to the HAT domain contain a conserved bromodomain, which has been shown recently to be an acetyl-lysine targeting motif24, 25, 26. In contrast, the CBP/p300 family of HAT proteins are more global regulators of transcription18, contain a considerably larger HAT domain of about 500 residues, and contain several other protein domains, including a bromodomain and three cysteine-histidine rich domains (TAZ, PHD and ZZ) that are believed to mediate protein-protein interaction. The MYST family of HAT proteins participate in divergent biological functions including positive regulation of transcriptional silencing (Sas2 and Sas3)27, the formation of leukemic translocation products (MOZ and TIF2)28, 29, and dosage compensation in Drosophila (MOF)30. The HAT domain within the MYST family is about 250 residues long and many members contain a cysteine-rich, zinc-binding domain within the HAT regions as well as an N-terminal chromodomain. The ATF-2 protein is the only sequence-specific DNA-binding transcriptional activator that has been shown to have HAT activity31. At the N terminus of ATF-2 is a potent transcriptional activation domain and near the C terminus is a basic-zipper (bZip) DNA-binding domain32.

Each of the HAT proteins that have been characterized in vivo is found to be associated with large multiprotein complexes. For example, yeast Gcn5 (yGcn5) is part of at least two multiprotein complexes, Ada and SAGA33, and PCAF is part of a SAGA-like complex in vivo34. A general feature of HAT proteins is that, although the recombinant proteins will acetylate free histones, nucleosomal acetylation occurs only in the context of the in vivo HAT complexes35. Detailed biochemical characterization of the acetylation activity of HAT proteins reveals that each HAT family generally has unique substrate specificity 23(Table 1). In addition, these specificities are further modulated within the context of the multisubunit in vivo HAT complexes. For example, recombinant yGcn5 will preferentially acetylate Lys14 on histone H3, but will also acetylate to a lesser extent Lys8 and Lys16 of histone H416, 17, 36. However, in the context of the Ada and SAGA complexes, the sites that are acetylated on nucleosomal H3 are expanded 37 and histone H2B is also a substrate16, 17.

In addition to catalyzing histone acetylation, a number of HAT proteins, including CBP/p300 and PCAF, have been shown to possess intrinsic transcription factor acetyltransferase (FAT) activity (Table 1)38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50. In several of these cases, specific acetylation sites have been identified and in many cases acetylation has been shown to enhance the DNA-binding affinity of the affected protein. Interestingly, acetylation by CBP/p300 of non-transcription factor proteins involved in nuclear import, importin-α and importin-α7, has also been reported51.

In addition to the HAT proteins that function in RNA Polymease (Pol) II transcription, several HAT proteins that affect other DNA-regulatory processes have been identified. Cytoplasmic, or B-type HAT proteins, acetylate newly synthesized histones H3 and H4 prior to their deposition in replicating chromatin52. The best studied B-type HAT, Hat1 from yeast (yHat1), is the catalytic subunit of the Hat1/Hat2 complex that acetylates histone H4 for replication-dependent chromatin assembly53, 54. Intriguingly, Hat1 from yeast and human has been localized to the nucleus55, 56, suggesting a role in transcription or replication-related histone acetylation. Very recently, Hat1 has been shown to participate in the nuclear function of telomeric silencing57. There is also evidence that HAT proteins are used for Pol III transcription, since subunits of the TF_IIIC complex (TF_IIIC90, TF_IIIC 110 and TF_IIIC 220) that are required for initiation of Pol III transcription also have HAT activity58, 59. Taken together, these findings suggests that HAT proteins may have very broad function in many DNA-regulatory processes.

Neuwald & Landsman noted that Gcn5/PCAF and Hat1 belong to a functionally diverse superfamily of N-acetyltransferases (called GNAT’s for gcn5 related N-acetyltransferases), containing limited sequence homology within four 15–33 amino acid segments called motifs A, B, C and D60. Gcn5, P/CAF and Hat1, however, do not harbor statistically significant sequence homology within motif C. The crystal structure of two members of the GNAT family, yHat1 bound to acetyl-coenzyme A61, and the Serratia marcescens aminoglycoside 3-N-acetyltransferase (SmAAT) bound to coenzyme A (CoA)62 reveal that motifs A, B and D form a structurally conserved domain (Figure 2(c) and (f)). Within this conserved domain, contiguous motifs A-D form three antiparallel β-strands followed by an α-helix on the underside of the β-sheet and roughly parallel with its direction. Motif B adopts a different structure in the two proteins, although in both cases the CoA cofactor is bound between the A-D and B submotifs63, and CoA interaction with the A-D motif is very similar in the two proteins.

The structure determination of the HAT GNAT members Gcn5/PCAF 64, 65 and yEsa1 (which shows sequence homology only in motif A)66, also reveal a homologous structure corresponding to the A-D motif with analogous CoA interactions (Figure 2(a)-(c)). Interestingly, the structure of N-myristoyl transferase (NMT)67, 68, which is not a member of the GNAT proteins, contains structural homology in the region corresponding to motifs A-D of the GNAT proteins63. These observations imply that other HAT proteins that show sequence divergence from the GNAT proteins, will, nonetheless, contain a structurally homologous A-D motif for CoA interaction.

A detailed comparison of the HAT domains of various members of the Gcn5/PCAF family in several different liganded forms64, 65, 69, with the MYST family member, yEsa1 bound to CoA66, and yHat1 bound to acetyl(Ac)-CoA reveals structural homology that extends beyond the A-D motif of the GNAT proteins (Figure 2). Specifically, the HAT proteins also show a structurally conserved loop- β-strand region immediately C-terminal to the helix of the A-D motif of the GNAT proteins (Figure 2). In each of the HAT proteins, the structurally homologous region (including the A-D motif and the loop-β-strand region) occurs roughly at the center of the molecule and forms a scaffold for the folding of the corresponding N and C-terminal domains and I therefore refer to this region of the HAT proteins as the core domain.

In contrast to the structurally similar core domains of the HAT proteins, the protein segments N and C-terminal to the corresponding core domains show structural divergence (Figure 2(a)-(c)). Despite the overall structural differences within the N and C-terminal segments of the three HAT proteins, subregions within the N and C-terminal segments show good alignment upon superposition of the core domains (purple domains in Figure 2(a)-(c)). Specifically, an α-helix-loop region just N-terminal to the core domain and a loop-α-helix region just C-terminal to the core domain superimpose well. The discussion below will illustrate how the central core domain of the HAT proteins play an important role in acetyl-coenzyme-A binding and catalysis, and how N and C-terminal segments play an important role in histone substrate binding.

The central core domain of the HAT proteins yHat1, Gcn5/PCAF and yEsa1 each play a conserved role in binding the CoA cofactor (Figure 2(a)-(c), and Figure 2(g)-(i)). The CoA is bound in a cavity formed on the surface of the core domain and buries over one-half of the CoA-accessible surface area and approximately 500 Ų of protein surface area. Most of the CoA contacts are mediated by a β-strand-loop-α-helix segment of the central core domain, corresponding to motif A of the GNAT proteins. In each case, the coenzyme A is bound in a bent conformation, which helps facilitate an extensive set of protein interactions that are mediated, predominantly by the pantetheine arm and the pyrophosphate group of the CoA. The pantetheine chain and pyrophosphate groups of Ac-CoA mediates most of the protein contacts, while the adenine base of CoA does not mediate protein contacts in either of the structures. As a result of this, the pantetheine arm-pyrophosphate chain superimposes well in the three structures, while the adenine base adopts a different orientation within each structure (Figure 2(a)-(c)).

The first clue that the central core domain of HAT proteins may play a role in catalysis came from structural and functional studies of yGcn5. Several lines of evidence suggested that a glutamate residue located in the β4 strand of the core domain (Glu173 in yGcn5 and Glu122 in tGCN5) played a role as a general base for catalysis (Figure 2(d)). The structure of yGcn5 revealed that this glutamate residue was centered within an electronegative patch that would be an ideal docking site for a positively charged lysine substrate, and was partially buried within a hydrophobic patch that would serve to raise it’s p K_a value to facilitate proton extraction from the ε-amino group of the target lysine. The functional importance of this glutamate residue was suggested by the fact that it was strictly conserved within the Gcn5/PCAF family of HAT proteins, and mutagenesis studies revealed that a triple alanine mutation including the glutamate position was one of the most debilitating mutations made within the HAT domain16.In vivo and in vitro analysis of an E173Q mutant of yGcn5 further supported the role of Glu173 in catalysis69.In vivo the E173Q mutation (yGcn5-E173Q) resulted in a debilitated growth phenotype and low levels of transcription, similar to that seen for the yGcn5 deletion strain. Detailed kinetic analysis of yGcn5-E173Q 70 revealed that the protein had K_m values for CoA and histone H3 that are indistinguishable from that of the wild-type protein, but a K_cat value that was down 360-fold for the mutant relative to that of the wild-type protein, demonstrating that this glutamate residue played a specific role in catalysis.

The ternary Tetrahymena Gcn5 (tGcn5)/CoA/histone H3 complex allows one to derive a detailed mechanism of catalysis that is mediated by the glutamate residue within the core domain of Gcn564(Figure 3). Glu122 of tGcn5 (Glu173 in yGcn5) is surrounded by several hydrophobic residues that likely shield it from solvent and help raise its p K_a to facilitate proton extraction. Glu122 is also bound to a water molecule located between its side-chain and the reactive Lys14 residue of the histone H3 peptide, and this water molecule is held in place by hydrogen bonds from the backbone carbonyl group of Val123 and the backbone NH of Tyr160. This water molecule is ideally located to shuttle a proton from the reactive Lys14 of the histone to Glu122 of the protein. Once the lysine proton is extracted, the acetyl group of the acetyl-CoA, which is hydrogen bonded to the backbone NH of Leu126 (in the tGCN5/acetyl-CoA structure), is transferred to the reactive Lys14 side-chain of the histone. The backbone NH of Leu126 probably functions to polarize the carbonyl group of the thioester prior to nucleophilic attack of the amino group and stabilize the negative charge that develops on the oxygen atom in the tetrahedral transition state. One issue that is still unclear is how the acetylated lysine product leaves once the reaction is complete.

Recent studies on the HAT domain of yEsa1 indicate that its core domain also plays an important role in catalysis66. A superposition of the yEsa1 core domain with the core domains of Gcn5/PCAF reveals that the residue corresponding to Glu173 in yGcn5 (122 in tGcn5) superimposes most closely in three-dimensional space to Glu338 in yEsa1 (Figure 2(d) and (e)). This superposition occurs despite the fact that the two glutamate residues arise from different secondary structure elements of their corresponding proteins. Glu338 of yEsa1 is conserved within the entire MYST family and is in an ideal chemical environment to function as a catalytic residue. Functional characterization of a E338Q mutant of yEsa1 (yEsa1-E338Q) further supports its catalytic role. An in vitro HAT assay reveals that wild-type yEsa1 has robust activity, while yEsa1-E338Q has an activity that is only marginally higher than background levels. Moreover, in vivo studies demonstrate that while wild-type yEsa1 effectively complements an Esa1 deletion strain, yEsa1-E338Q does not. Transformation of a wild-type gene into a strain bearing yEsa1-E338Q also does not complement the deletion, further suggesting that the mutation acts as a dominant negative.

Interestingly, a superposition of the core domain of yHat1 with the core domains of Gcn5/PCAF and yEsa1 also reveals that Glu255 of Hat1 superimposes well with the putative general base residues of the other HAT proteins, suggesting that Glu255 may play an important catalytic role in yHat1 (Figure 2(f)). In support of this, within the Hat1 protein family, Glu255 is strictly conserved within a stretch of sequence that shows poor conservation. Taken together, these observations suggest that the structurally conserved core domains of Gcn5/PCAF, yEsa1 and yHat1 mediate functionally conserved roles in CoA binding and catalysis. An implication from this is that other more divergent HAT proteins, such as CBP/p300 and TAF_II250, will also have a related catalytic mechanism.

The mode of histone binding by Gcn5/PCAF is visualized in the structure of the ternary complex of tGcn5 bound to CoA and an 11 residue histone H3 peptide64. The histone H3 peptide adapts a random coil structure and is bound in a pronounced protein cleft of the tGCN5 protein above the core domain and flanked on opposite sites by the N and C-terminal protein segments (Figure 2(a)). Most of the protein-peptide interactions are mediated by the loop-α2 N-terminal segment and the loop-α4 C-terminal segment (Figure 4). Remarkably, most of these interactions involve the backbone of the histone H3 peptide, and about 75 % of these interactions involve the Lys14 target and the five residues C-terminal to it. In addition to Lys14 of the histone H3 substrate, Gly13 and Pro16 play important roles in histone H3 binding specificity. The requirement for Lys14 and Pro16 stem from the extensive protein interactions and the requirement for Gly13 appears to derive from its unusually constrained packing environment within the protein. Significantly, of the mapped acetylation sites of recombinant yGcn5 for K14 in histone H3, and for K8 and K16 in histone H4, only positions 13 and 16 (and the reactive lysine residue at position 14) show homology among the three sequences. Position 13 is a glycine residue in all three cases and position 16 is either a proline, leucine or histidine residue, each of which can provide hydrophobic surfaces to mimic the interactions mediated by Pro16 of the histone H3 peptide in the complex. Thus, the histone H3 binding determinants of tGCN5 appear to be restricted to a small G-K-X-P recognition sequence.

The comparison of the of Gcn5 structure in various liganded forms reveals that the binding of CoA and the histone peptide do not change the protein structure substantially64, 65, 69. However, CoA does appear to induce localized changes that appear to have ramifications for histone substrate binding. Specifically, it appears that CoA binding widens the histone H3 peptide binding groove by moving the C-terminal protein segment outward. This groove is widened further upon binding of the histone H3 peptide. A comparison of CoA-protein contacts between the binary and ternary complexes shows that CoA interactions to the core domain are left largely unchanged, while CoA interactions with the N and C-terminal protein segments, that also interact with the histone H3 peptide, make different CoA interactions. Taken together, it appears that CoA-mediated protein contacts facilitate histone H3 contacts, and it is likely that acetyl-CoA binding must precede histone binding in vivo.

Neither the yHat1/Ac-CoA 61 nor yEsa1/CoA 66 structures have bound peptide, so one cannot directly visualize HAT protein-histone interactions. However, a mapping of conserved and surface-exposed residues in the respective HAT families highlights the α-helix-loop region in the N-terminal segment (α2-loop in yEsa1 and α7-loop in yHat1) and a loop-α-helix region in the C-terminal segment (loop-α4 in yEsa1 and loop-α9 in yHat1) (Figure 2(g)-(i)). These segments flank opposite sides of the core domain and, as in Gcn5, form a cleft that is proximal to the CoA and the putative general base for catalysis (Figure 2(a)-(c)). A superposition of the core domains of Gcn5/PCAF, yEsa1 and yHAT1 shows that these same regions have similar conformation; namely, a helix-loop region N-terminal to the core domain and a loop-helix domain C-terminal to the core domain (purple regions in Figure 2(a)-(c)). Taken together, comparison of the three known HAT structures, Hat1, Gcn5/PCAF and Esa1, suggests a conserved structural framework for histone substrate binding whereby sequence divergence within this framework may modulate the binding to specific histone targets. This proposal is consistent with the observation that the analogous structural regions of GNAT proteins that bind non-histone targets have significantly more divergent N and C-terminal structures66.

Despite the fact that yEsa1 shows weak sequence homology to PCAF/Gcn5 and yHat1, the entire core domain of the three HAT proteins superimpose remarkably well, and mediate very similar CoA interactions. The studies described above suggest that the core domains of Gcn5/PCAF, yEsa1 and possibly also yHAT1, contain a structurally superimposable general base for catalysis and thereby share a related catalytic mechanism. Taken together, it is reasonable to propose that protein members from other HAT families, such as CBP/p300 and TAF_II250, also contain a structurally conserved core domain that mediates functionally conserved roles in CoA binding and catalysis.

Regions N and C-terminal to the core domain of the HAT proteins Gcn5/PCAF, yEsa1 and yHAT1 show overall structural divergence. Nonetheless, a superposition of the core domains of these proteins reveals striking superposition between an N-terminal α-helix-loop and a C-terminal loop-α-helix. These superimposed regions colocalize to the precise regions of tGcn5 within the ternary tGcn5/CoA/histone H3 peptide complex that play particularly critical roles in histone substrate binding. Moreover, in each case, these regions contain surface-exposed residues that are highly conserved within their respective HAT families, and directly flank the core domain that harbors the putative catalytic base. Taken together, the three HAT proteins analyzed here appear to have a conserved structural framework for histone substrate binding that likely extends to other more divergent HAT proteins. Moreover, it is likely that sequence divergence within this framework modulates the binding to specific histone targets.

Despite the availability of the ternary tGcn5/CoA/hisotone H3 peptide complex, the mode of substrate binding specificity by HAT proteins is far from clear. Indeed, the fact that the majority of HAT domain-histone interactions are mediated by the backbone of the histone peptide substrate argues that there is more to binding specificity then these interactions. Functional data demonstrate that HAT proteins and HAT complexes have similar but distinct substrate specificities, supporting the notion that substrate specificity is modulated, at least in part, by other proteins and/or other protein domains, within the in vivo HAT complexes. In support of this, the Gcn5/PCAF64, yEsa166 and Hat161 structures each reveal conserved patches of residues that do not appear to be involved in substrate binding directly but may highlight contact sites for other proteins or protein domains within the in vivo HAT complexes (Figure 2(g)-(i)). Of special note in the MYST family of HAT proteins is a cysteine-rich region in the N terminus of the HAT domain. This region contains the consensus CxxC x₁₂HxxxC, that has been proposed to ligand zinc and has been shown to be essential for HAT activity for several of the MYST family members71, 72, 73. In the case of MOF, this domain has been shown to interact with the globular region of the nucleosome core72, and in the case of HBO1 this region has been shown to interact with the minichromosome maintenance (MCM) protein for DNA replication73.

Yeast Esa1 is one member of the MYST family of HAT proteins that does not harbor the cysteine-rich consensus sequence. Surprisingly, however, the structure of yEsa1 reveals that this region of the protein forms a classical TFIIIA zinc finger-type fold (Figure 5).74 Zinc fingers of this type have been proposed to mediate protein/DNA interaction or protein-protein interaction. In support of the proposal that this region may indeed be involved in protein interactions within the in vivo HAT complexes, is the observation that this is one of the most highly conserved regions of the MYST family and many of the conserved residues are solvent-accessible and therefore available for potential interaction. The exact mode of this interaction, and how such interaction may modulate histone substrate binding/specificity, must undoubtedly await structural analysis of a relevant protein complex. Nonetheless, the available data suggest that the cysteine-rich region within the MYST family of HAT proteins plays a very important role in their HAT function.

Further clues into the mode of histone substrate specificity by HAT proteins comes from the study of histone phosphorylation. Two studies 75, 76 show a direct functional link between phosphorylation of Ser10 and acetylation of Lys14 on histone H3. Specifically, in vitro acetylation at Lys14 by yGcn5 is enhanced five- to tenfold when the histone tail is pre-phosphorylated at Ser10. Moreover, this enhancement is seen for the Gcn5/PCAF and CBP/p300 families of HATs, but not for the Esa1 member of the MYST HAT family, suggesting that the enhancement of acetylation due to phosphorylation is specific to distinct HAT families. Berger and co-workers 76 also show that a specific arginine to alanine mutation in the histone binding site of yGcn5, and proximal to Ser10, disrupts the enhanced effect of acetylation due to phosphorylation. Importantly, this mutation in yGcn5, as well as a histone H3 serine to alanine mutation as position 10, has reduced activity at the same subset of yGcn5-dependent promoters. These studies suggest that transcriptional activation of certain Gcn5-dependent genes requires histone H3 phosphorylation at Ser10 prior to histone H3 acetylation on Lys14. Phosphorylation at Ser10 of histone H3 may therefore increase the specificity for acetylation at Lys14 of yGcn5.

The recent finding that bromodomains, a domain found in many, but not all HAT proteins, have binding specificity for acetylated lysine residues within histones24, 25, 77, suggests that bromodomain function may be coordinated with HAT activity in vivo. Such coordination may involve targeting HAT proteins to acetylated histones, or possibly contribute to the off-rate of the acetyl-lysine product of HAT proteins. Taken together, although the mode of histone binding specificity by HAT proteins is still unclear and will require further biochemical and structural analysis, the data described above suggest that it may involve a combination of other histone modifications and the cooperation of other conserved transcriptional regulatory domains.

Structural analysis of the catalytic HAT subunit of several of the in vivo HAT complexes has provided important insights into the mechanism of histone substrate binding and catalysis. However, several key questions underlying HAT activity remain unanswered. Among the most important of these questions is, what are the factors that influence substrate binding specificity by HAT proteins for histones, as well as non-histone targets, and how does this specificity influence gene activation? A related question is, what are the specific roles of the other subunits of the in vivo HAT complexes in modulating HAT function? Resolution of this question will require the structure of relevant multiprotein HAT complexes and ultimately the structure of intact in vivo HAT complexes. Detailed biochemical analysis of the intact in vivo HAT complexes as well as complexes missing specific subunits or specific protein domains will be required to complement the structural work.

Aberrant HAT activity has been associated with diseases such as cancer. For example, in acute myeloid leukemia (AML), the CBP HAT gene is translocated with the HAT gene encoding MOZ28, or another homeotic regulator called MLL78. In both cases, the HAT domain of CBP remains intact in the fusion protein. The p300 HAT gene is also found rearranged in a separate subset of AMLs79. In addition, point mutations are found in the p300 gene in a significant proportion of colorectal and gastric carcinomas80, where a loss of heterozygosity coincident with the p300 levels is observed in 80 % of glioblastomas81. Finally, analysis of primary breast cancers indicates gene amplification and protein overexpression of the AIB1 HAT82.

In light of the involvement of HAT proteins in human cancer, it would appear that these enzymes would be attractive targets for the design of specific inhibitors. A relatively recent study reports on the design of a group of bisubstrate peptide inhibitors that exhibit specificity for different HAT families83. Since these studies did not use any prior structural information in inhibitor design, it is reasonable to think that incorporation of structure-based optimization may lead to small molecule compounds that may be effective as in vivo HAT inhibitors with potential therapeutic applications. Aside from the disease connection, the availability of HAT-specific inhibitors would be a useful research tool, since it would help to elucidate which HAT is acetylating which substrate and when. One thing that is clear is that HATs come in many sizes and many flavors, and determining which HAT is doing what and when and for what purpose will occupy the literature for many years to come.