Lessons from genome-wide studies: an integrated definition of the coactivator function of histone acetyl transferases

Abstract

Histone acetylation is one of the key regulatory mechanisms controlling transcriptional activity in eukaryotic cells. In higher eukaryotes, a number of nuclear histone acetyltransferase (HAT) enzymes have been identified, most of which are part of a large multisubunit complex. This diversity, combined with the large number of potentially acetylable lysines on histones, suggested the existence of a specific regulatory mechanism based on the substrate specificity of HATs. Over the past decade, intensive characterisations of the HAT complexes have been carried out. However, the precise mode of action of HATs, and particularly the functional differences amongst these complexes, remains elusive. Here we review current insights into the functional role of HATs, focusing on the specificity of their action. Studies based on biochemical as well as genetic approaches suggested that HATs exert a high degree of specificity in their acetylation spectra and in the cellular processes they regulate. However, a different view emerged recently from genomic approaches that provided genome-wide maps of HAT recruitments. The careful analysis of genomic data suggests that all HAT complexes would be simultaneously recruited to a similar set of loci in the genome, arguing for a low specificity in their function. In this review, we discuss the significance of these apparent contradictions and suggest a new model that integrates biochemical, genetic and genome-wide data to better describe the functional specificity of HAT complexes.

Figure 1

Substrate specificity of histone acetyltransferases (HATs) in vitro. Review of the in vitro substrate specificity described in mammalian systems for the studied five HAT enzymes, either as a free protein or within their respective macromolecular complexes. Dark grey boxes represent lysine residues highly acetylated, light grey boxes represent residues where weak acetylation activity has been observed, and white boxes represent residues where no acetylation is detected. ND represents residues that have not been tested for acetylation. References [15,18,27-34] cited in figure.

Figure 2

Phenotypes observed in gene disruption studies of HATs. Phenotypes associated with HAT genetic knockout (KO) or RNAi targeting in mouse development and embryonic stem cell (ESC) pluripotency, respectively. For mouse KO, the day of embryonic death (ED) for homozygous animals is presented as a phenotype readout parameter. For the effect on mouse embryonic stem cell (mESC) pluripotency, any observation (e.g., flattening of the cells, differentiation) was simplified as a positive phenotype (+). ND represents HATs for which the parameter was not determined. References [19-22,35,36] cited in figure.

Figure 3

Similar genome-wide binding patterns of 15 HAT marks at binding sites of HATs. Raw chromatin immunoprecipitation sequencing (ChIP-seq) data were extracted from [12,13]. (a) Co-occurrence of 15 acetylation marks on CBP binding sites: average binding densities of 15 HAT marks (marked in black), control (immunoglobulin G (IgG)) (marked in dashed black) and CBP (marked in red) surrounding ± 5-kb region of a collection of 10,360 CBP binding sites. From the raw data sets, enrichment clusters representing CBP binding sites were determined. Around each CBP binding site, four hundred 25-bp bins were created, and densities were collected for each bin for the 15 acetylation tracks. The mean was calculated for each bin and used to represent average acetylation densities around the CBP binding sites. (b) Co-occurrence of 18 acetylation marks on all the genome-wide CBP binding sites: binding densities of regions (± 5) surrounding the 10,360 binding sites of CBP. Densities are shown for control (IgG), CBP and 18 HAT marks (as indicated). In the heat map, each line represents a genomic location of a binding site with its surrounding ± 5-kb region. CBP binding sites were used as references to collect ChIP-seq tag densities over a 10-kb (± 5 kb) window. This matrix was subjected to k-means clustering. The heat map representing the clustered density matrix is displayed. In (a) and (b), similar results were obtained with the other four other HATs (data not shown).

Figure 4

HATs are corecruited at high frequency on their binding loci. Raw ChIP-seq data were extracted from [12,13]. (a) Heat map showing colocalization frequency of all the five HATs, namely, MOF, p300, CBP, TIP60 and PCAF. Colors in the heat map reflect the colocalization frequency (Pearson correlation coefficient) of each pair of HAT (red means more frequently colocalized) over a 5-kb region surrounding the complete set of HAT binding sites. HATs were clustered along both axes based on the similarity in their colocalization with other HATs. (b) Co-occurrence of five HATs on CBP binding sites: Binding densities of PCAF, p300, MOF and TIP60 were clustered according to 10,360 CBP binding sites. In the clustering, each line represents a genomic location of a binding site with its surrounding ± 5-kb region. HAT binding sites were used as a reference to collect ChIP-seq tag densities over a 10-kb (± 5 kb) window in each HAT density map. This matrix was subjected to k-means clustering. The heat map representing the clustered density matrix is displayed.

Figure 5

Schematics describing the possible modes of action of HATs. (a) Gene specificity model. This model can be proposed on the basis of the conclusions from biochemical and genetic studies. Each HAT complex is recruited by a particular DNA binding transcriptional activator to a defined set of genes, allowing their activation. Thus HATs seem to exert a high functional specificity. (b) Systematic co-occurrence model. This model arises from conclusions of genome-wide mapping studies. HATs would be corecruited to all the transcriptionally active loci, creating a hyperacetylated environment that would favor the activation of the corresponding genes. (c) Dynamic model. Model proposed to reconcile the observations of the genome-wide mapping of HATs with previous biochemical and genetic observations. HATs play a dual role in the gene activation process. In the first phase, a specific HAT recruited by a specific activator allows the initiation of the activation process. Later other HATs can bind the activated loci in a less-specific manner, thus maintaining a nonspecific hyperacetylated environment.