3D replicon distributions arise from stochastic initiation and domino-like DNA replication progression

Abstract

DNA replication dynamics in cells from higher eukaryotes follows very complex but highly efficient mechanisms. However, the principles behind initiation of potential replication origins and emergence of typical patterns of nuclear replication sites remain unclear. Here, we propose a comprehensive model of DNA replication in human cells that is based on stochastic, proximity-induced replication initiation. Critical model features are: spontaneous stochastic firing of individual origins in euchromatin and facultative heterochromatin, inhibition of firing at distances below the size of chromatin loops and a domino-like effect by which replication forks induce firing of nearby origins. The model reproduces the empirical temporal and chromatin-related properties of DNA replication in human cells. We advance the one-dimensional DNA replication model to a spatial model by taking into account chromatin folding in the nucleus, and we are able to reproduce the spatial and temporal characteristics of the replication foci distribution throughout S-phase.

Figure 1. Induced firing probability.

The firing probability of origins that are close to forks follows a Gaussian probability density, indicated as shaded areas next to the forks. Firing at positions closer than d_i=55 kbp to a fork is inhibited and the probability density is cutoff at values below 0.1. The relative probabilities of individual origins are indicated by dark grey bars. All four forks to the left of the chromosome boundary belong to a single 1D fork cluster (assuming that neighbouring forks are <1 Mbp apart). The chromosome boundary near the right edge of the image isolates chromatin belonging to different chromosomes and thus cuts off the induced firing range of the rightmost fork.

Figure 2. Several simulated replication characteristics compared with experimental data.

(a) Confocal RFi measurements were used to model the initial increase of the limiting factor with a mono-exponential fit L(t)=L_max(1−e^−t/τ) with timescale τ=15 min. (b) Distribution of distances between adjacent fired origins from DNA combing data for HeLa Kyoto cells. The distribution has a peak below 200 kbp and a heavy tail up to 600 kbp. The corresponding distribution, averaged over 100 simulations, displays similar features. (c) Fraction of replicated chromatin as a function of time. Colours are used to distinguish between the chromatin type specific and total replication. Dotted lines show the simulation results, when only induced firing events are allowed. Dashed lines display the other extreme case, where solely spontaneous firing was used. The combined model includes both firing events and the results are shown with solid lines. (d) Time-dependent number of forks in each chromatin type. (e) Comparison of our model with replication timing data for chromosome 6 from the ENCODE project (cell type GM12878). Sampling positions are identical to the positions in the experimental data. For individual simulations, the euchromatic peaks start at time zero, but because of the specific sampling positions and averaging over 100 simulations, the displayed peaks are less extreme. The Pearson's correlation coefficient between the theoretical and experimental data shown here is 0.60. The Background indicates the Giemsa staining, where white regions are interpreted as euchromatin and shaded regions as facultative or constitutive heterochromatin. The centromere is indicated as a striped pattern. Analogous figures for other human chromosomes can be found in the Supplementary Figs 4–6.

Figure 3. Replication subphase analysis and simulated fork clusters.

(a) DNA content frequency throughout the cell cycle. Cells are binned by DNA content (DAPI signal), with the abscissa showing the DNA content of the bins in arbitrary units. The distribution remains at an approximately constant value throughout S-phase, that is, between the G1 and G2 peaks, meaning that the overall rate of replication is constant. (b) Frequency of specific DNA content intervals in an ensemble of 840 HeLa Kyoto cells from five separate slide areas dependent on their cell cycle position. Through inspection of the PCNA signal, the cells were sorted into early, middle and late S-phase. It is notable that the number of early S-phase cells drops off steeply at 15% of the DNA replicated. (c): The observed subphase durations, where the error bars indicate the s.d. Subphase durations were obtained using live cell microscopy as described in the accompanying manuscript Chagin et al. modified from Reinhart et al. (d) HeLa Kyoto cells stably expressing mCherry-PCNA were labelled with modified nucleotides (20 μm EdU) for 15 min before fixation. Wide-field images of cells going through different S-phase stages show that, while the overall level of PCNA is rather constant, the total amount of incorporated nucleotides is clearly lower in cells going through early S-phase, indicating a lower synthesis rate. From left to right: single channel images, overlay of EdU (red) and PCNA (green) signals, representation of the ratio of EdU to PCNA signal intensity. LUT as indicated. Scale bar, 10 μm. (e) Line profile from the overlaid EdU (red) and PCNA (green) image over six cells going through different S-phase sub-stages as indicated. (f) Number and size of replication clusters over time. EdU, 5-Ethynyl-2′-deoxyuridine; LUT, Lookup Tables; MCM, Minichromosome maintenance protein complex; UCSC, University of California, Santa Cruz.

Figure 4. Comparison between the microscopy pattern during replication in experiment and model.

(a) Experimental maximum intensity z-projections and middle section images of green fluorescent protein (GFP)-tagged PCNA in HeLa cells during replication (as described by Chagin et al. scale bar, 5 μm). (b) The corresponding patterns of the replication model results from a 3D DNA conformation calculated using the random loop model. The fork positions in the simulations were accumulated over 15 min similar to the experimental staining time. A Gaussian blur was applied to imitate the limited experimental voxel sizes of 40 × 40 × 125 nm. In the last row the simulated fork positions are marked depending on the chromatin type (blue, euchromatin; green, facultative heterochromatin; red, constitutive heterochromatin). Images for different parameters and chromatin distributions can be created online at http://sim.bio.tu-darmstadt.de. See also Supplementary Movies 1, 2, 3 for a visualization of the fork movement within the nucleus.