Modulated contact frequencies at gene-rich loci support a statistical helix model for mammalian chromatin organization

Abstract

Background: Despite its critical role for mammalian gene regulation, the basic structural landscape of chromatin in living cells remains largely unknown within chromosomal territories below the megabase scale.

Results: Here, using the 3C-qPCR method, we investigate contact frequencies at high resolution within interphase chromatin at several mouse loci. We find that, at several gene-rich loci, contact frequencies undergo a periodical modulation (every 90 to 100 kb) that affects chromatin dynamics over large genomic distances (a few hundred kilobases). Interestingly, this modulation appears to be conserved in human cells, and bioinformatic analyses of locus-specific, long-range cis-interactions suggest that it may underlie the dynamics of a significant number of gene-rich domains in mammals, thus contributing to genome evolution. Finally, using an original model derived from polymer physics, we show that this modulation can be understood as a fundamental helix shape that chromatin tends to adopt in gene-rich domains when no significant locus-specific interaction takes place.

Conclusions: Altogether, our work unveils a fundamental aspect of chromatin dynamics in mammals and contributes to a better understanding of genome organization within chromosomal territories.

Figure 1

Random collision frequencies at five mouse gene-rich loci. (a) Maps of mouse loci investigated. Genes are indicated by full boxes and promoters by thick black arrows. The scale bar indicates the size of 10 kb of sequence. The names of the loci and chromosomal location are indicated above each map. The HindIII (Usp22, Emb, Lnp, Mtx2 and 11qA5 gene-desert loci) or EcoRI (Dlk1 locus) sites investigated are indicated on the maps. Arrows indicate the positions of the primers used as anchors in 3C-qPCR experiments. (b) Random collision frequencies at five mouse gene-rich loci. Locus names are indicated above each graph. Random collision frequencies were determined by 3C-qPCR in the 30-day-old mouse liver at the indicated anchor sites (for further details see Materials and methods). They were determined in three independent 3C assays each quantified at least in triplicate and the data were normalized as previously described [19]. Error bars are standard error of the mean of three independent 3C assays. Grey circles, triangles or squares are data points obtained from distinct genomic sites as indicated on the graphs. In each graph, red squares represent the floating mean (20-kb windows, shift of 10 kb). P-values (Mann-Whitney U-test) account for the significance of the differences observed between the higher and the lower points of the floating mean. They were calculated from the values of the average random collision frequencies in a window of 30 kb around these points (values indicated in the figure) (One asterisk indicates a P-value < 0.1 and > 0.05; double asterisks a P-value < 0.05 and > 0.01 and triple asterisks a P-value < 0.01).

Figure 2

Random collision frequencies in gene-rich and gene-desert regions. (a) Experimental data obtained for mouse gene-rich regions (shown in separate graphs in Figure 1b) have been plotted into a single graph. A few data points at separation distances above 150 kb, which were omitted in Figure 1b, are included. Statistical analyses were performed on the floating mean (red squares) as explained in Figure 1b. The dashed lines delimit supranucleosomal domains (D.I to D.VI) that encompass separation distances where random collision frequencies are alternatively lower and higher: 0 to 35 kb (domain I), 35 to 70 kb (domain II), 70 to 115 kb (domain III), 115 to 160 kb (domain IV), 160 to 205 kb (domain V) and 205 to 250 kb (domain VI). (b) Random collision frequencies were determined by 3C-qPCR at four sites (R9, F25, F35 and F48; Figure 1) located in an AT-rich/gene-desert region located on mouse chromosome 11. Red squares represent the floating mean (20-kb windows, shift of 10 kb). Error bars are standard error of the mean (the triple asterisks indicate a P-value < 0.01).

Figure 3

Influence of modulated random collision frequencies on long-range interactions and mammalian genome evolution. (a) Separation distances between conserved sequences (cs) and transcription start sites (TSS) of co-expressed mouse genes were determined as explained in the Materials and methods section. Black triangles depict the relative count of separation distances obtained for each supranucleosomal domain. Black squares indicate the mean of relative counts obtained from 30 random samples of genes. Error bars represent the 95% confidence intervals for randomization. Separation distances are significantly over-represented in domains III and V (+7.9% and +6.6%, respectively) while they are significantly under-represented in domain IV (-8.6%) (P-values of t-tests are indicated on the graph). (b) Histogram depicting the relative counts of cis-interactions in human GM06990 or K562 cells (Hi-C experiments from [4]) occurring in Giemsa-negative (gene-rich regions, white bars) or Giemsa-positive (gene-poor regions, gray bars) bands. For each set, the number of interactions was counted in each supranucleosomal domain (as defined in Figure 2a). Counts in each domain were normalized against the total number of sequence-tags counted over all domains (D.I to D.VI). Error bars represent standard error of the mean of two Hi-C experiments. The P-value indicated on the figure was obtained from a t-test (double asterisks indicate a P-value < 0.05 and >0.01, and triple asterisks a P-value < 0.01).

Figure 4

Fitting the statistical helix polymer model to random collision frequencies quantified at mouse gene-rich loci. 3C-qPCR data shown in Figure 2a and Additional file 1 (Usp22PE) were compiled into a single graph (upper panel). Error bars are standard error of the mean. The dashed lines delimit supranucleosomal domains as defined in Figure 2a. The graph shows the best fit analyses obtained with the linear polymer model (Equations 1 and 2a; black curve) or the statistical helix model (Equations 1 and 5; red curve). Correlation coefficients (R²) are indicated in the lower panel, which shows the same graph where collision frequencies are represented in a logarithmic scale. Best fit parameters for the statistical helix model are indicated within the graph (lower panel) and have been used to calculate the expected theoretical means of random collision frequencies for each supranucleosomal domain (numbers in brackets in upper panel), which are in good agreement with the means obtained from the experimental data (values indicated above the expected means). P-values (Mann-Whitney U-test) account for the significance of the differences observed between the experimental means of two adjacent domains. One can note, amongst the experimental points, a few outliers. To minimize the weight of these data points, we chose a non-parametric statistical test (double asterisks indicate a P-value < 0.05 and > 0.01 and triple asterisks a P-value < 0.01).

Figure 5

The statistical helix model. The statistical helix model that we propose in this study (Equations 1 and 5) suggests that, in the absence of strong locus-specific interactions, some gene-rich domains of the mammalian chromatin tend to adopt a helix shape. This helix is averaged over the whole population of cells analyzed (5 million nuclei in each 3C sample) and thus more likely represents a statistical shape arising from the global dynamics of the chromatin than a fixed organization. It is characterized by a mean diameter 〈D〉 and mean step 〈P〉, and it thus likely corresponds with the place where the probability of finding the chromatin at a given t time is the highest (black helical curve).