Fig 1.
A CFSE-based method to sort mammalian cells by proliferation rate.
(A) Cells were stained with CFSE and a subpopulation of cells with identical CFSE levels was collected by FACS. Growth for several generations resulted in a heterogeneous cell population with a broad CFSE distribution, and cells with high, medium, and low CFSE signal (slow, medium and fast proliferation, respectively) were sorted by FACS for RNA-sequencing. (B) The change in the CFSE distribution over time, for fibroblasts and ESCs. (C, D) The population-level doubling time can be calculated by fitting a line to the median of the log2(CFSE) signal. We discard data from time 0, cells immediately after the sort, because the CFSE signal decreases in the initial hours, even in the absence of cell division, likely due to efflux pumps. (E, F) Bromodeoxyuridine (BrdU) was used to measure the % of cells in S-phase for FACS-sorted fast and slow proliferating subpopulations. Fibroblasts: 4 replicates, p = 0.0002441. ESCs: 3 replicates for ESCs, p = 0.001953. p-values are for binomial tests across all biological replicates that the two populations have the same percentage of cells in S-phase. (G) Examples of genes whose expression positively or negatively correlated with proliferation rate. Each line is one biological replicate, and the error bars are 95% confidence intervals for each expression value.
Fig 2.
Slow-proliferating ESCs display a more naive pluripotent stemness character than fast-proliferating ESCs.
(A) Comparison of lineage commitment-related gene expression between fast and slow proliferating subpopulations. (B) Comparison of pluripotency-associated gene expression between fast and slow proliferating subpopulations. (C) Comparison of 2C-like state markers expression between fast proliferating subpopulation and slow proliferating sub-population. Dashed lines separate genes expressed preferentially in slow- (left of dashed line) or in fast-proliferating (right of dashed line) ESCs. P-values are from binomial tests, testing if genes are more often highly expressed in slow cells than would be expected by chance (53.5% of all genes are more highly expressed in slow cells).
Fig 3.
Functional pathways for which cell-to-cell heterogeneity in expression correlates with proliferation rate across cell types and species.
(A) In Gene Set Enrichment Analysis, genes are sorted by their fast/slow expression value (left panel, bottom), and each gene is represented by a single black line (left panel middle). The enrichment score is calculated as follows: for each gene not in the GO preribosome gene set, the value of the green line decreases, and for each gene in the gene set, the value of the green line increases. The ES score will be near zero if the genes in a gene set are randomly distributed across the sorted list of genes, positive if most genes are to the left, and negative if most genes are to the right. (B) The heatmap (right panel) shows the expression (z-scored read counts) of preribosome genes in fibroblasts across four biological replicates of the CFSE sorting experiment. (C) Gene sets enriched (FDR<0.1) in both fibroblasts and ESCs were mapped as a network of gene sets (nodes) related by mutual overlap (edges), where the color (red or blue) indicates if the gene set is more highly expressed in fast (red) or slow (blue) proliferating cells. Node size is proportional to the total number of genes in each set and edge thickness represents the number of overlapping genes between sets. (D) GSEA results (FDR<0.1) of S. cerevisiae [8] that sorted by cell-to-cell heterogeneity in proliferation rate. (E) Pearson correlations of mean expression (average of log2(TPM+1)) of ribosome biogenesis genes vs proteasome genes across organ developmental time courses in seven species (see also S2 Fig).
Fig 4.
A proliferation signature model can predict relative growth rates from gene expression for species and cell-types on which it was not trained.
(A) Genes and proliferation signature spearman correlation with proliferation rate (sorted by CFSE). Compare with KI67 or PCNA, proliferation signature has a better correlation with proliferation rate. (B) Using the proliferation signature to predict growth rate in budding yeast, we apply ssGSEA to calculate the enrichment score of proliferation signature for each sample. The Pearson correlation of proliferation signature score with growth rate is 0.82 (p = 8.9×10−7), the grey shading is a 95% confidence band. (C) Using the proliferation signature to predict growth rate in cancer cell lines, the Pearson correlation is 0.65 (p = 1.9×10−8), the grey shading is a 95% confidence band. (D) Comparison of proliferation signature score between 2C-like ESC and non-2C-like ESC (paired t-test, p = 0.04669).
Fig 5.
Proliferation signature of cell development.
(A) A cartoon showing four terminal cells, and a partial linage showing the final four generations of preterminal cells. Comparison of single-cell proliferation signatures between preterminal cell lineage and terminal cell types (t-test, p = 4.9×10−41). (B) UMAP projection of 89,701 cells. Cells in the left panel are colored by estimated embryo times; in the right panel by proliferation signature score. (C) To calculate the proliferation signature score (y-axis) at each time point (x-axis) cells are binned by embryo time, and the mean proliferation signature score for all cells in the same bin is calculated. The spearman correlations are -0.65 (p = 9.3 ×10−19) for binned data and -0.42 (p < 2.2e-16) for unbinned data. (D) Boxplots (line shows median, boxes interquartile range) of proliferation signature score for all cells with embryo time > 650min. (E) Temporal dynamics of proliferation scores of select cell lineages, showing the average proliferation score for all single cells in that lineage, at each time point. (F-G) Boxplot of C. elegans (F) and human (G) proliferation signatures as a function of developmental time, from scRNAseq data.
Table 1.
Gene sets whose expression exhibits opposite correlations with growth between fibroblasts and ESCs.
Fig 6.
Expression of proliferation-related gene sets in cells sorted by intra-population heterogeneity in mitochondria membrane potential.
(A) Cells were stained with Hoechst and CFSE and a homogenous population of equally sized cells in G1 with equal CFSE was obtained by FACS. These cells were stained with TMRE sorted by TMRE, and then used for RNA-seq, or allowed to proliferate to measure the doubling time of each TMRE sub-population. (B, C) Enrichment maps of fibroblasts and ESCs sorted by TMRE. (D) Doublings times, as estimated by the measured by the decrease in CFSE signal over time, for high, medium and low TMRE sorted cells. P-values are from ANOVA, testing if TMRE is predictive of doubling time (see Materials and Methods).
Fig 7.
Growth-rate and cell-type specific effects of mitochondria inhibitors on proliferation rate, cell viability and cell state.
(A) Schematic of the experimental setup for measuring the effects of mitochondria inhibitors on slow and fast proliferating cells. (B) Fast proliferating Fibroblast and ESCs sorted by CFSE signal maintained a higher fraction of cells in S phase over two days of growth in medium+DMSO. (C) Effect of antimycin treatment on fast and slow proliferating fibroblasts and ESCs (Drug effect: log2 of the fraction of cells in S-phase when treated with DMSO divided by the fraction of cells in S-phase when treated with drug). (D) Fast fibroblasts changed morphology after the treatment with oligomycin. Scale bars = 80 μm. (E) Immunostaining of fibroblasts for N-Cadherin and DAPI after drug treatment and corresponding quantifications. Fast fibroblasts lose N-Cadherin staining specifically after oligomycin treatment. Scale bars = 15 μm. The fluorescence intensity of N-Cadherin has been quantified on the right. Medians and the 95% confidence intervals are shown as error bars. Kruskal–Wallis test was used for statistical comparison (ns, not significant, **** p < 0.0001). (F) Effect of oligomycin treatment on fibroblast viability. The % viable cells is measured as % trypan blue-negative cells.