Kronos scRT: a uniform framework for single-cell replication timing analysis

Abstract

Mammalian genomes are replicated in a cell type-specific order and in coordination with transcription and chromatin organization. Currently, single-cell replication studies require individual processing of sorted cells, yielding a limited number (<100) of cells. Here, we develop Kronos scRT, a software for single-cell Replication Timing (scRT) analysis. Kronos scRT does not require a specific platform or cell sorting, which allows investigating large datasets obtained from asynchronous cells. By applying our tool to published data as well as droplet-based single-cell whole-genome sequencing data generated in this study, we exploit scRT from thousands of cells for different mouse and human cell lines. Our results demonstrate that although genomic regions are frequently replicated around their population average RT, replication can occur stochastically throughout S phase. Altogether, Kronos scRT allows fast and comprehensive investigations of the RT programme at the single-cell resolution for both homogeneous and heterogeneous cell populations.

Fig. 1. An efficient uniform framework for scRT extraction.

a The pipeline of Kronos scRT with its different modules. The input files, the main modules of Kronos RT, and the optional modules are shown in green, dark blue-grey, and light blue-grey, respectively. b Scatter plot reporting the mean cell ploidy on the x-axis and the bin-to-bin intracellular variability on the y-axis. Each point is a single cell and the colour is assigned based on a cut-off automatically calculated, or manually imposed, relative to the population variability (“Methods”). High variability is associated with S-phase cells (green), while low variability is associated with G1/G2-phase cells (blue). An unknown region can be manually set (yellow). The vertical black line represents the median ploidy of the population. The data are from asynchronous S-phase enriched MCF7 cells. The green and purple arrows show the S-phase progression of the first and second S-phase cell group, respectively (“Methods”). c Data presented in (b) after S-phase progression correction. The colour of the second S-phase group has been changed to purple. As in (b), the green and purple arrows indicate the S-phase progression. The black curve reports the S-phase density distribution that is used to calculate the parameters to adjust the S-phase progression (“Methods”). d Read down-sampling for the MCF7 cell dataset obtained using the 10x Genomics System. G1/G2- and S-phase cells are plotted in orange and green, respectively. Boxplots show the mean ploidy variation (in percentage) after down-sampling compared with the original value. The red dashed line indicates a change of 5%. Cells with more than 450 reads per Mb were used for the down-sampling. The numbers of cells (n) used for each down-sampling are reported below each boxplot, cells are analyzed over one experiment. In the boxplots, bounds of box: 25th and 75th percentiles; centre line: median; lower (upper) whisker: lower (upper) bound - (+) 1.5 x Interquartile range (IQR).

Fig. 2. Extracting scRT from sorted mid S-phase mouse cells.

a scRT calculated from FACS-sorted mid S-phase mouse embryonic stem cells (mESC, left) and mouse neuroectoderm cells at day 7 of differentiation (mNE-7d, right). Bottom panels: binary scRT profiles of single cells sorted from top to bottom in function of the increasing replication percentage. Replicated and non-replicated regions are shown in red and blue, respectively. Upper panels: the pseudo-bulk RT calculated from the scRT profiles (see Methods) are compared with the bulk RT of the corresponding cell type. b Comparison between single-cell and bulk RT data. 2D density plot reporting pairwise comparisons between samples (colour code on the right) and Spearman correlations (circles) between RT data. Diagonal panels: RT distribution of each sample. c T_width values of mESCs (upper panels) and mNE-7d (lower panels) cells for Early (RT > 0.5) and Late (RT ≤ 0.5) replicating regions. Higher T_width indicates higher RT variability among cells. P values were calculated with the Kronos scRT Compare TW module (see “Methods”); **** < 10⁻⁴.

Fig. 3. Extracting scRT from asynchronous human MCF7 cells.

a Scatter plots showing the intracellular bin-to-bin variability in function of single-cell ploidy before (left) and after (right) S-phase progression correction. The colour code is the same as in Fig. 1b, c. Kronos scRT can automatically identify the S-phase cells from the asynchronous scWGS data (see Methods). b scRT calculated from unsorted cycling MCF7 breast cancer cells. Same presentation as in Fig. 2a. c Comparison between single-cell and bulk RT data, as in Fig. 2b. d T_width values calculated for genomic regions classified in five RT categories based on the pseudo-bulk RT values (Very early >0.8, 0.8≥ Early >0.6, 0.6≥ Mid >0.4, 0.4≥ Late >0.2, and Very Late ≤0.2). P values were calculated using the Kronos scRT Compare TW module (“Methods”); * < 0.05, ** < 10⁻², *** < 10⁻³, **** < 10⁻⁴.

Fig. 4. Identification of two sub-populations in MCF7 cell cultures.

a UMAP analysis of the scCNV data of MCF7 cells coloured in function of S-phase progression (only autosomal chromosomes). Both G1/G2 (yellow) and S-phase (colour intensity in function of genome replication as indicated) cells are separated into two major sub-populations. G1/G2- and S-phase cells are gated with colour-coded gates: aqua for sub-population 1 and blue for sub-population 2. b Copy number analysis of autosomes in G1/G2 MCF7 cells separated into two sub-populations based on the clustering results shown in (a). The binning for visualization is 1 Mb (median CN profile in Supplementary Fig. 3c). c, d Examples of FISH experiments for cells with 4 (c) and 5 (d) copies of chromosome 3 that correspond to sub-population 1 and sub-population 2, respectively. Chromosomes are stained with DAPI (grey), chromosome 3 is labelled in green, and its centromere in red. Scale bar: 10 µm. Two examples out of 344 cells obtained over two independent experiments. e Chromosome 3 count based on the FISH results (n=344). The two major groups of cells with 4 (n = 162) and 5 (n = 170) copies of chromosome 3 correspond to the two groups shown in (a): sub-population 1 and sub-population 2, respectively.

Fig. 5. scRT analysis in S-phase enriched human cells.

a Pairwise comparison of the MCF7 cell pseudo-bulk RT and bulk RT profiles. Same as in Fig. 2b. b The scRT profiles of S-phase enriched MCF7 cells along a representative region. Upper: pseudo-bulk RT profiles of the two MCF7 sub-populations and bulk RT profile. Bottom: scRT profiles ordered from top to bottom in function of the genome replication percentage of each cell. c, d Dimensionality reduction analysis of scRT data for the indicated human cell lines (generated in the current study). Cells are colour-coded based on the cell type in (c) and based on the genome replication percentage in (d). e T_widths calculated for the indicated five RT categories based on the pseudo-bulk RT values in the two MCF7 sub-populations. Categories were selected as in Fig. 3d. P-values were calculated using the Kronos scRT Compare TW module (see “Methods”); * < 0.05, ** < 10⁻², *** < 10⁻³, **** < 10⁻⁴.

Fig. 6. scRT data support a stochastic model of replication.

a, b Boxplots showing the replication probability (y-axis) relative to the pseudo-bulk RT (x-axis, 1 is early, and 0 is late) for S-phase-enriched MCF7 cells (sub-population 1) (a) and Jeff cells (b) at the indicated S-phase stages calculated in 200 kb bins. Middle panels: whole distribution; top and bottom panels: zoom on the areas showing the distribution upper and lower extremities, respectively (indicated with the dashed boxes in the middle panels). The numbers of genomic bins (n) are reported in the bottom panel below each boxplot. c Comparison between the pseudo-bulk RT (solid line) and simulated RT (dashed line) for the MCF7 cell sub-population 1 and Jeff cells. The simulation is based on Replicon and uses the probability of being replicated in early S-phase cells (that completed up to 30% of their genome replication) for each 200 kb bin as input. Similar results were obtained with the scRT data of the S-phase-enriched MCF7 cell sub-population 2 and HeLa cells (Supplementary Fig. 5a, b, g). Spearman correlations are shown in Supplementary Fig. 5h.

Fig. 7. Kronos scRT analysis of scRT using scHi-C data.

a Top: pseudo-bulk RT profile of mESC grown in 2i (purple) or in serum (green) compared with bulk RT (black). Bottom: scRT profiles ordered, from top to bottom, in function of the genome replication percentage of each cell. b Pairwise comparisons of the pseudo-bulk RT and bulk RT profiles in mESC, mESC-2i, and mESC Serum. Same as in Fig. 2b. c T_width calculated for the indicated five RT categories based on the corresponding pseudo-bulk RT values, as in Fig. 3d. P values were calculated with the Compare TW module as before; * < 0.05, ** < 10⁻², *** < 10⁻³, **** < 10⁻⁴.