Andrew Oliveira Silva

Senescence is a cellular state in which the cell loses its proliferative capacity, often irreversibly. Physiologically, it occurs due to a limited capacity of cell division associated with telomere shortening, the so-called replicative senescence. It can also be induced early due to DNA damage, oncogenic activation, oxidative stress, or damage to other cellular components (collectively named induced senescence). Tumor cells acquire the ability to bypass replicative senescence, thus ensuring the replicative immortality, a hallmark of cancer. Many anti-cancer therapies, however, can lead tumor cells to induced senescence. Initially, this response leads to a slowdown in tumor growth. However, the longstanding accumulation of senescent cells (SnCs) in tumors can promote neoplastic progression due to the enrichment of numerous molecules and extracellular vesicles that constitutes the senescence-associated secretory phenotype (SASP). Among other effects, SASP can potentiate or unlock the tumor plasticity and phenotypic transitions, another hallmark of cancer. This review discusses how SnCs can fuel mechanisms that underlie cancer plasticity, like cell differentiation, stemness, reprogramming, and epithelial-mesenchymal transition. We also discuss the main molecular mechanisms that make SnCs resistant to cell death, and potential strategies to target SnCs. At the end, we raise open questions and clinically relevant perspectives in the field.

Several phenotypes that impact the capacity of cancer cells to survive and proliferate are dynamic. Here we used the number of cells in colonies as an assessment of fitness and devised a novel method called Dynamic Fitness Analysis (DynaFit) to measure the dynamics in fitness over the course of colony formation. DynaFit is based on the variance in growth rate of a population of founder cells compared with the variance in growth rate of colonies with different sizes. DynaFit revealed that cell fitness in cancer cell lines, primary cancer cells, and fibroblasts under unhindered growth conditions is dynamic. Key cellular mechanisms such as ERK signaling and cell-cycle synchronization differed significantly among cells in colonies after 2 to 4 generations and became indistinguishable from randomly sampled cells regarding these features. In the presence of cytotoxic agents, colonies reduced their variance in growth rate when compared with their founder cell, indicating a dynamic nature in the capacity to survive and proliferate in the presence of a drug. This finding was supported by measurable differences in DNA damage and induction of senescence among cells of colonies. The presence of epigenetic modulators during the formation of colonies stabilized their fitness for at least four generations. Collectively, these results support the understanding that cancer cell fitness is dynamic and its modulation is a fundamental aspect to be considered in comprehending cancer cell biology and its response to therapeutic interventions.Significance:Cancer cell fitness is dynamic over the course of the formation of colonies. This dynamic behavior is mediated by asymmetric mitosis, ERK activity, cell-cycle duration, and DNA repair capacity in the absence or presence of a drug.

Supplementary Fig. S1 DynaFit principles. a, numerical representation of the rationale presented in Fig. 1a. Cells with static growth rates will produce colonies with the same variance in growth rate, regardless of colony size (upper scenario). Colonies containing cells with dynamic growth rates will present a lower variance of growth rates as they grow in number (lower scenario). b, schematic representation of the colony fitness experiment. Single cells are seeded, and colony fitness is quantified between two distinct time points. c, Example of experimental data obtained from b. For DynaFit, the colonies are grouped by their initial colony size. The growth rate variance is measured for each colony size.

Supplementary Fig. S6  Growth characteristics of individual colonies. a, GR (mean {plus minus} variance) of U251 glioma colonies according to the colony size at the beginning of the analyzed GR during the presence of TMZ and b. after the withdrawal of TMZ. In orange, the mean {plus minus} variance GR of all colonies combined c. GR d. variance of untreated or TMZ-treated colonies, either including or excluding all colonies that had at least one event of cell death, as evaluated by the single-cell tracking of colonies in the presence of TMZ. e. GR of U251 cells treated with the indicated concentration of cisplatin, f. CVP of U251 cell line treated with 1 microM of cisplatin or g. 4.5 microM cisplatin.

Supplementary Fig. S7  Comparison of Dynafit results with growth rate. a, GR and DynaFit Cumulative Hypothesis plot results b. GR2 versus CS1-GR2 c. GR3 versus CS1-GR2 d. GR3 versus CS1-GR3 of untreated (green), treated only with cytotoxic drugs (red) and the combination of epigenetic modulators and cytotoxic drugs (blue).

File describing in more detail the methods.

Supplementary Fig. S3Analysis of the desynchronization of colonies of cells expressing FastFUCCI. a, FUCCI simulation for a single, synchronized colony with 4 cells. At a given time point, all cells in the colony will all either be green or not. b, simulating multiple colonies gives rise to multimodal distributions of green cell percentage in colonies. c, adding a phase shift parameter allows cells in each colony to be out of sync, producing percentages of green cells in each colony other than 0% or 100% for a given observation point. d, by modifying the simulation parameters, various frequency distributions of the percentage of green cells can be obtained for several colonies. Note that a, b and c represent three specific time steps of a full simulation run, which must walk all cells around the cell cycle at least once, and produced a final distribution considering all evaluated time steps.

Supplementary Fig. S4 Tagging of the endogenous locus of Ki67 with EYFP. Sequence coding mVenus-P2A-Neomycin was PCR amplified from the eFlut-mVenus-Neomycin plasmid generously donated by Galit Lahav (Harvard University) and used to insert flanking regions of about 500 nt, amplified with the indicated primers. This was used for homologous recombination in the CRISPR-Cas9 targeted C-terminal region of Ki67 gene.

Supplementary Fig. S2 DynaFit implementations. a, data collected from Supplementary Fig. S1c is given as input to the two DynaFit apps. The Python bootstrap app and the R predictive modeling app are based on different analytical strategies. b, Colony Variance Plot formed by the Python app and c, its hypothesis plot. d, Colony Variance Plot formed by the R app and e, its hypothesis plot.

Supplementary Fig. S5  Growth characteristics of individual treated colonies. a, GR of all colonies used in Fig. 1e. Mean {plus minus} SD in red. b. Pairwise correlations among the three GRs. c. Mean {plus minus} variance of GR2 of U251 glioma colonies according to the number of cells in CS1 or all colonies. Growth rate of colonies with 1 and 2 cells were different from all growth rates (ANOVA, Tukey's multiple comparison post-hoc test). d, Distribution of the number of colonies in each bin for Fig. 1e. e, GR of the indicated cell lines at high and low density. f, CVP and g, hypothesis plot of the glioma cell line A172wt (WT), GFP-tagged alone (GFP) or in the presence of 100 fold excess of untagged cells (GFP 1:100).