We show that yeast isochores share characteristics with those found in higher eukaryotes in addition to those described before. Our results indicate that GC-rich and AT-rich domains are both structurally and functionally distinct. First, interaction frequencies within GC-rich chromatin tend to be lower than those in AT-rich chromatin, which is in agreement with a more extended chromatin conformation, as observed in higher eukaryotes [12
]. Second, similar to mammalian isochores, genes located in the most GC-rich regions of the yeast genome are, on average, more highly expressed (for example, [4
]). Importantly, we found that GC-rich genes display higher levels of H3 and H4 acetylation compared to more AT-rich genes. Finally, we identify Rpd3p as a molecular component involved in base composition-dependent control of chromatin structure and function. This role of Rpd3p may be conserved in higher eukaryotes as it is also associated with less condensed interbands in Drosophila
]. This activity appears to be specific for Rpd3p as we did not detect a base composition-dependent activity of another globally acting histone deacetylase, Hda1p.
Rpd3p has been shown to have two distinct modes of action. First, Rpd3p is recruited to specific target genes to modulate their expression. Second, Rpd3p acts in a global and non-targeted fashion to deacetylate bulk chromatin. We propose that the base composition-dependent effects of Rpd3p are related to its global activities. First, the magnitudes of these GC-content dependent effects are subtle, similar to the previously described effects of Rpd3p on global histone acetylation [40
]. Second, deletion of Ume6p, a protein involved in recruitment of Rpd3p to many of its specific target genes [38
], does not result in up-regulation of GC-rich genes, indicating that Rpd3p interacts with GC-rich genes in a Ume6p-independent manner. Third, the GC content-dependent effects are not correlated with the steady-state expression of genes, and thus seem unrelated to local promoter controls.
We favor the model that the global and untargeted activity of Rpd3p acts predominantly and/or has the largest effect on GC-rich chromatin. First, Rpd3p binds GC-rich genes more prominently than AT-rich genes. Second, deletion of RPD3 results in increased H4 acetylation, particularly of GC-rich genes. Finally, treatment of wild-type cells with the histone deacetylase inhibitor TSA activates GC-rich genes more strongly than AT-rich genes. However, we did observe that TSA induced activation of GC-rich genes requires more time than induction of many direct target Rpd3p genes. This relatively slow effect could be interpreted to mean that the base composition-dependent effects of deletion of RPD3 are indirect and are due to altered expression of a specific Rpd3p target gene that, in turn, encodes a protein that acts in a GC content-dependent fashion. Alternatively, and consistent with the Rpd3p localization and acetylation data, Rpd3p does directly affect expression of GC-rich genes, but this more global and non-targeted process occurs at a longer time scale or requires passage through a specific phase of the cell cycle.
An alternative or additional molecular explanation of the observed phenomena is related to potential base composition-dependent differences in wrapping of DNA around histones. AT-rich DNA may be more flexibly and more easily wrapped around nucleosomes than GC-rich DNA [44
]. This physical model implies intrinsic differences in nucleosome organization dependent on base composition and does not require that histone modifying enzymes act in a base composition-dependent fashion per se
. In this model, histone modifying enzymes recognize differences in intrinsic conformation of GC- and AT-rich chromatin. Rpd3p may preferentially act on the nucleosome organization of GC-rich chromatin. Similarly, acetyl transferases may preferentially modify GC-rich domains in wild type, resulting in higher levels of histone H3 and H4 acetylation, as we observed here. Based on these considerations, we predict the presence of histone acetyl transferases that act most prefentially on GC-rich chromatin.
In light of these observations, we can interpret our 3C analysis more precisely. The 3C results show that deletion of RPD3 differentially affected the conformation of GC- and AT-rich isochore domains along chromosome III, but did not allow determination of which of the two types of domains, or both, displayed an altered conformation. When Rpd3p activity affects GC-rich genes most prominently, the most parsimonious explanation of our 3C data is that deletion of RPD3 most strongly affects the conformation of the GC-rich domain, resulting in a more extended and transcriptionally active chromatin conformation, consistent with predicted relationships between transcription, histone acetylation and chromatin conformation.
GC-rich chromatin displays lower interaction frequencies, as detected by 3C, than AT-rich chromatin. Analysis of cross-linking efficiency suggests that both types of domains are cross-linked with similar frequencies (Additional data file 2) and, therefore, have similar protein densities. Histones are the most abundant chromatin proteins, and thus our results suggest that GC-rich and AT-rich regions have similar levels of histone binding. Consistent with this hypothesis, Nagy et al
] found no correlation between base composition and regions depleted in nucleosomes. Previously, we found a decrease in interaction frequencies upon activation of the FMR1
gene in human cells [19
], similar to the observed changes in rpd3Δ
cells described here, suggesting that reduced 3C interaction frequencies may be a general characteristic of active chromatin.
The base composition-dependent effect of Rpd3p activity affects expression of genes independent of their steady state level of expression. Genes with the same steady state expression level in wild type are more strongly repressed by Rpd3p when they are GC-rich than when they are AT-rich. This implies that GC-rich genes are intrinsically more active, consistent with higher steady state levels of H3 and H4 acetylation, as we observed here, and that Rpd3p acts as an attenuator of these genes. Based on these considerations, we propose that chromatin status is regulated through a homeostatic and highly dynamic mechanism involving counteracting activating and repressing activities. A similar model of dynamic global acetylation and deacetylation has been proposed by Katan-Khaykovich and Struhl [46
] and by Clayton et al