Figure 1.
Heterochromatic and euchromatic histone modifications form non-overlapping domains on a coarse scale.
We used the measurements obtained in a genome-wide experiment on human CD4+ T cells from [59] and [57] and analyzed them using the CCAT (version 3.0) tool [73]. We adapt the analysis using slightly less stringent parameters than the default ones allowing for noisy measurements with lower significance in order to obtain a most complete modification landscape. The plot exhibits the significance scores of the histone modifications H3K9me2 and H3K9me3 related to heterochromatin (red) and H3K4me2, H3K4me3, H3K18ac and H3K23ac related to euchromatin (black). We visualized the distribution over entire chromosome 1. Heterochromatin marks were plotted upside down for better visualization. Both sets of modifications form very similar patterns and form regions of higher and lower abundance. We marked some of the regions with high (low) euchromatic and low (high) heterochromatic content green (blue).
Figure 2.
Euchromatin and heterochromatin marks become anti-correlated on a coarse scale.
Pearson's correlation between modifications within bins of 100 kbp for different chromosomes and two cell lines. Instead of taking into account the individual scores in a bin, we simplify the content by the sum of all scores. It can be clearly observed that euchromatic (H3K9/14ac, H3K18ac, H3K23ac, H3K4me2 and H3K4me3) and heterochromatic histone marks (H3K9me2, and H3K9me3) oppose each other for all considered chromosomes as well as for two different cell lines. We therefore show that these modifications form long domains that are still detectable on a scale of about 1000 histones.
Figure 3.
Illustration of the processes nucleation, propagation, competition and deletion in the computational model.
Only nucleosomes with nucleation sites can directly be modified with the respective modification with rate (probability per histone and time step). Empty nucleosomes with neighboring modified nucleosomes obtain a modification of the same type with rate
. This parameter can vary for different modification types (euchromatic or heterochromatic). Multiply modified nucleosomes are not allowed in the model with only competing marks and therefore a new mark will not be set if the histone is already modified. Finally, every modified nucleosome looses its modification with rate
.
Figure 4.
Pearson's correlation between simulations and experiments on chromosome 1 of CD4+ cells for different values of the parameters ,
and
.
A higher value corresponds to a higher correlation between the simulated chromatin distribution averaged over the last 100,000-seq analysis. The heterochromatin marks H3K9me2 and H3K9me3 were compared to the simulated heterochromatin distribution and the marks H3K4me2, H3K4me3, H3K18ac and H3K23ac to the simulated euchromatin distribution. Hence, a good match between simulations and experiment is obtained for all 6 fields being green. The heterochromatin marks are rather sparsely distributed and therefore only low correlation values could be reached, especially for the H3K9me3 mark.
Table 1.
Parameters of the computational model giving positive correlation with the experimental data on all simulated chromosomes.
Figure 5.
Simulation results (blue) and experimental data (red/black) of CD4+ T cells exhibit similar distributions for euchromatin and heterochromatin on chromosome 1.
The red dots show both H3K9me2 and H3K9me3 marks together, i.e. plotting their scores. The black dots exhibit all scores for the marks H3K4me2, H3K4me3, H3K18ac and H3K23ac. For each histone we depict its occupation frequency averaged over the last 100,000(green dots) merely function as initiators of the process whereas propagation acts as the main competitor in the system. The blue dots show the histone mark distribution. The black and red dots correspond to the same experimental data from Figure 1, this time normalized for better visualization. Model parameters were ,
,
.
Figure 6.
Example for rearrangements of heterochromatin and euchromatin due to changes of the propagation rate on chromosome 1.
We zoomed on a region of 0.5-seq scores of euchromatin marks and black dots to the ones of heterochromatin marks. Green dots are respective nucleation sites and blue denotes simulations. On this scale, ChIP-seq data becomes sparse even when taken from multiple marks. However, the heterochromatin marks mostly disappear at the euchromatin domains where the euchromatin marks reach higher scores. Although this zoom was taken on an arbitrary region of the first chromosome, we see 3 different reactions of the system to alterations of the propagation rate, : (a) the euchromatin domain disappears for higher propagation rates; (b) a large euchromatin domain develops at low propagation rates; (c) the euchromatin domain remains unchanged. (a) and (b) exhibit characteristic regions for potential gene regulations. Refseq gene locations where adopted from http://genome.ucsc.edu ([74]). Other parameter values were
,
,
.
Figure 7.
Switch-like behavior for competing histone marks.
We take the temporal average of the number of with mark modified histones after the simulation reached a stationary-like state,
, presenting now the average frequency of a histone mark. A clear transition between two saturated states is observed, where the number of modifications fluctuates maximally for
. Inner panels: evolution of the number of modifications for different parameter sets. The other parameters were
,
,
.
Figure 8.
Complete transition of the chromatin landscape for different propagation rates.
The figure shows the spatial distribution of modifications averaged over the last iterations, , for different values of
. Despite the purely random distribution of nucleation sites, chromatin domains form around accumulations of nucleation sites in the upper two panels. The other parameters were
,
,
.
Figure 9.
New nucleation sites can lead to different effects.
Upper panel: Black and red lines denote the spatial distribution of histone mark 1 and 2, respectively. The 90 (100) nucleation sites for mark 1 (2) are shown as black (red) points. Lower panels: Zooms on the spatial distribution of mark 1 for 90 nucleation sites (black line) and 100 nucleation sites (green line). The 100 nucleation sites of mark 2 remain unchanged. The arrows tag new inserted nucleation sites of mark 1. Three different nucleation effects can be observed: (a) no change ; (b) narrow spike around nucleation site ; (c) activation of large region. The simulation parameters were ,
and
.