Cycling cancer persister cells arise from lineages with distinct programs.

Non-genetic mechanisms have recently emerged as important drivers of cancer therapy failure1, where some cancer cells can enter a reversible drug-tolerant persister state in response to treatment2. Although most cancer persisters remain arrested in the presence of the drug, a rare subset can re-enter the cell cycle under constitutive drug treatment. Little is known about the non-genetic mechanisms that enable cancer persisters to maintain proliferative capacity in the presence of drugs. To study this rare, transiently resistant, proliferative persister population, we developed Watermelon, a high-complexity expressed barcode lentiviral library for simultaneous tracing of each cell's clonal origin and proliferative and transcriptional states. Here we show that cycling and non-cycling persisters arise from different cell lineages with distinct transcriptional and metabolic programs. Upregulation of antioxidant gene programs and a metabolic shift to fatty acid oxidation are associated with persister proliferative capacity across multiple cancer types. Impeding oxidative stress or metabolic reprogramming alters the fraction of cycling persisters. In human tumours, programs associated with cycling persisters are induced in minimal residual disease in response to multiple targeted therapies. The Watermelon system enabled the identification of rare persister lineages that are preferentially poised to proliferate under drug pressure, thus exposing new vulnerabilities that can be targeted to delay or even prevent disease recurrence.

Connecting Timescales in Biology: Can Early Dynamical Measurements Predict Long-Term Outcomes?

Prediction of long-term outcomes from short-term measurements remains a fundamental challenge. Quantitative assessment of signaling dynamics, and the resulting transcriptomic and proteomic responses, has yielded fundamental insights into cellular outcomes. However, the utility of these measurements is limited by their short timescale (hours to days), while the consequences of these events frequently unfold over longer timescales. Here, we discuss the predictive power of static and dynamic measurements, drawing examples from fields that have harnessed the predictive capabilities of such measurements. We then explore potential approaches to close this timescale gap using complementary measurements and computational approaches, focusing on the example of dynamic measurements of signaling factors and their impacts on cellular outcomes.

A Switch in p53 Dynamics Marks Cells That Escape from DSB-Induced Cell Cycle Arrest.

Cellular responses to stimuli can evolve over time, resulting in distinct early and late phases in response to a single signal. DNA damage induces a complex response that is largely orchestrated by the transcription factor p53, whose dynamics influence whether a damaged cell will arrest and repair the damage or will initiate cell death. How p53 responses and cellular outcomes evolve in the presence of continuous DNA damage remains unknown. Here, we have found that a subset of cells switches from oscillating to sustained p53 dynamics several days after undergoing damage. The switch results from cell cycle progression in the presence of damaged DNA, which activates the caspase-2-PIDDosome, a complex that stabilizes p53 by inactivating its negative regulator MDM2. This work defines a molecular pathway that is activated if the canonical checkpoints fail to halt mitosis in the presence of damaged DNA.

Quantifying the Central Dogma in the p53 Pathway in Live Single Cells.

Transcription factors (TFs) integrate signals to regulate target gene expression, but we generally lack a quantitative understanding of how changes in TF levels regulate mRNA and protein production. Here, we established a system to simultaneously monitor the levels of p53, a TF that shows oscillations following DNA damage, and the transcription and protein levels of its target p21 in individual cells. p21 transcription tracked p53 dynamics, while p21 protein steadily accumulated. p21 transcriptional activation showed bursts of mRNA production, with p53 levels regulating the probability but not magnitude of activation. Variations in p53 levels between cells contributed to heterogeneous p21 transcription while independent p21 alleles exhibited highly correlated behaviors. Pharmacologically elevating p53 increased the probability of p21 transcription with minor effects on its magnitude, leading to a strong increase in p21 protein levels. Our results reveal quantitative mechanisms by which TF dynamics can regulate protein levels of its targets. A record of this paper's transparent peer review process is included in the Supplemental Information.

Identification of universal and cell-type specific p53 DNA binding.

BACKGROUND: The tumor suppressor p53 is a major regulator of the DNA damage response and has been suggested to selectively bind and activate cell-type specific gene expression programs. However recent studies and meta-analyses of genomic data propose largely uniform, and condition independent p53 binding and thus question the selective and cell-type dependent function of p53. RESULTS: To systematically assess the cell-type specificity of p53, we measured its association with DNA in 12 p53 wild-type cancer cell lines, from a range of epithelial linages, in response to ionizing radiation. We found that the majority of bound sites were occupied across all cell lines, however we also identified a subset of binding sites that were specific to one or a few cell lines. Unlike the shared p53-bound genome, which was not dependent on chromatin accessibility, the association of p53 with these atypical binding sites was well explained by chromatin accessibility and could be modulated by forcing cell state changes such as the epithelial-to-mesenchymal transition. CONCLUSIONS: Our study reconciles previous conflicting views in the p53 field, by demonstrating that although the majority of p53 DNA binding is conserved across cell types, there is a small set of cell line specific binding sites that depend on cell state.

Live cell monitoring of double strand breaks in S. cerevisiae.

We have used two different live-cell fluorescent protein markers to monitor the formation and localization of double-strand breaks (DSBs) in budding yeast. Using GFP derivatives of the Rad51 recombination protein or the Ddc2 checkpoint protein, we find that cells with three site-specific DSBs, on different chromosomes, usually display 2 or 3 foci that may coalesce and dissociate. This motion is independent of Rad52 and microtubules. Rad51-GFP, by itself, is unable to repair DSBs by homologous recombination in mitotic cells, but is able to form foci and allow repair when heterozygous with a wild type Rad51 protein. The kinetics of formation and disappearance of a Rad51-GFP focus parallels the completion of site-specific DSB repair. However, Rad51-GFP is proficient during meiosis when homozygous, similar to rad51 "site II" mutants that can bind single-stranded DNA but not complete strand exchange. Rad52-RFP and Rad51-GFP co-localize to the same DSB, but a significant minority of foci have Rad51-GFP without visible Rad52-RFP. We conclude that co-localization of foci in cells with 3 DSBs does not represent formation of a homologous recombination "repair center," as the same distribution of Ddc2-GFP foci was found in the absence of the Rad52 protein.

The Single-Cell Yin and Yang of Live Imaging and Transcriptomics.

A new approach to monitoring both signaling over time and a global gene expression profile from the same cell establishes a functional role for NF-κB dynamics in transcription.

Re-establishment of nucleosome occupancy during double-strand break repair in budding yeast.

Homologous recombination (HR) is an evolutionarily conserved pathway in eukaryotes that repairs a double-strand break (DSB) by copying homologous sequences from a sister chromatid, a homologous chromosome or an ectopic location. Recombination is challenged by the packaging of DNA into nucleosomes, which may impair the process at many steps, from resection of the DSB ends to the re-establishement of nucleosomes after repair. However, nucleosome dynamics during DSB repair have not been well described, primarily because of a lack of well-ordered nucleosomes around a DSB. We designed a system in budding yeast Saccharomyces cerevisiae to monitor nucleosome dynamics during repair of an HO endonuclease-induced DSB. Nucleosome occupancy around the break is lost following DSB formation, by 5'-3' resection of the DSB end. Soon after repair is complete, nucleosome occupancy is partially restored in a repair-dependent but cell cycle-independent manner. Full re-establishment of nucleosome protection back to the level prior to DSB induction is achieved when the cell cycle resumes following repair. These findings may have implications to the mechanisms by which cells sense the completion of repair.

Asf1 facilitates dephosphorylation of Rad53 after DNA double-strand break repair.

To allow for sufficient time to repair DNA double-stranded breaks (DSBs), eukaryotic cells activate the DNA damage checkpoint. In budding yeast, Rad53 (mammalian Chk2) phosphorylation parallels the persistence of the unrepaired DSB and is extinguished when repair is complete in a process termed recovery or when the cells adapt to the DNA damage checkpoint. A strain containing a slowly repaired DSB does not require the histone chaperone Asf1 to resume cell cycle progression after DSB repair. When a second, rapidly repairable DSB is added to this strain, Asf1 becomes required for recovery. Recovery from two repairable DSBs also depends on the histone acetyltransferase Rtt109 and the cullin subunit Rtt101, both of which modify histone H3 that is associated with Asf1. We show that dissociation of histone H3 from Asf1 is required for efficient recovery and that Asf1 is required for complete dephosphorylation of Rad53 when the upstream DNA damage checkpoint signaling is turned off. Our data suggest that the requirements for recovery from the DNA damage checkpoint become more stringent with increased levels of damage and that Asf1 plays a histone chaperone-independent role in facilitating complete Rad53 dephosphorylation following repair.

A Cohesin-Based Partitioning Mechanism Revealed upon Transcriptional Inactivation of Centromere.

Transcriptional inactivation of the budding yeast centromere has been a widely used tool in studies of chromosome segregation and aneuploidy. In haploid cells when an essential chromosome contains a single conditionally inactivated centromere (GAL-CEN), cell growth rate is slowed and segregation fidelity is reduced; but colony formation is nearly 100%. Pedigree analysis revealed that only 30% of the time both mother and daughter cell inherit the GAL-CEN chromosome. The reduced segregation capacity of the GAL-CEN chromosome is further compromised upon reduction of pericentric cohesin (mcm21∆), as reflected in a further diminishment of the Mif2 kinetochore protein at GAL-CEN. By redistributing cohesin from the nucleolus to the pericentromere (by deleting SIR2), there is increased presence of the kinetochore protein Mif2 at GAL-CEN and restoration of cell viability. These studies identify the ability of cohesin to promote chromosome segregation via kinetochore assembly, in a situation where the centromere has been severely compromised.

Caffeine impairs resection during DNA break repair by reducing the levels of nucleases Sae2 and Dna2.

In response to chromosomal double-strand breaks (DSBs), eukaryotic cells activate the DNA damage checkpoint, which is orchestrated by the PI3 kinase-like protein kinases ATR and ATM (Mec1 and Tel1 in budding yeast). Following DSB formation, Mec1 and Tel1 phosphorylate histone H2A on serine 129 (known as γ-H2AX). We used caffeine to inhibit the checkpoint kinases after DSB induction. We show that prolonged phosphorylation of H2A-S129 does not require continuous Mec1 and Tel1 activity. Unexpectedly, caffeine treatment impaired homologous recombination by inhibiting 5' to 3' end resection, independent of Mec1 and Tel1 inhibition. Caffeine treatment led to the rapid loss, by proteasomal degradation, of both Sae2, a nuclease that plays a role in early steps of resection, and Dna2, a nuclease that facilitates one of two extensive resection pathways. Sae2's instability is evident in the absence of DNA damage. A similar loss is seen when protein synthesis is inhibited by cycloheximide. Caffeine treatment had similar effects on irradiated HeLa cells, blocking the formation of RPA and Rad51 foci that depend on 5' to 3' resection of broken chromosome ends. Our findings provide insight toward the use of caffeine as a DNA damage-sensitizing agent in cancer cells.

Caffeine inhibits gene conversion by displacing Rad51 from ssDNA.

Efficient repair of chromosomal double-strand breaks (DSBs) by homologous recombination relies on the formation of a Rad51 recombinase filament that forms on single-stranded DNA (ssDNA) created at DSB ends. This filament facilitates the search for a homologous donor sequence and promotes strand invasion. Recently caffeine treatment has been shown to prevent gene targeting in mammalian cells by increasing non-productive Rad51 interactions between the DSB and random regions of the genome. Here we show that caffeine treatment prevents gene conversion in yeast, independently of its inhibition of the Mec1(ATR)/Tel1(ATM)-dependent DNA damage response or caffeine's inhibition of 5' to 3' resection of DSB ends. Caffeine treatment results in a dosage-dependent eviction of Rad51 from ssDNA. Gene conversion is impaired even at low concentrations of caffeine, where there is no discernible dismantling of the Rad51 filament. Loss of the Rad51 filament integrity is independent of Srs2's Rad51 filament dismantling activity or Rad51's ATPase activity and does not depend on non-specific Rad51 binding to undamaged double-stranded DNA. Caffeine treatment had similar effects on irradiated HeLa cells, promoting loss of previously assembled Rad51 foci. We conclude that caffeine treatment can disrupt gene conversion by disrupting Rad51 filaments.

Functional interplay between the 53BP1-ortholog Rad9 and the Mre11 complex regulates resection, end-tethering and repair of a double-strand break.

The Mre11-Rad50-Xrs2 nuclease complex, together with Sae2, initiates the 5'-to-3' resection of Double-Strand DNA Breaks (DSBs). Extended 3' single stranded DNA filaments can be exposed from a DSB through the redundant activities of the Exo1 nuclease and the Dna2 nuclease with the Sgs1 helicase. In the absence of Sae2, Mre11 binding to a DSB is prolonged, the two DNA ends cannot be kept tethered, and the DSB is not efficiently repaired. Here we show that deletion of the yeast 53BP1-ortholog RAD9 reduces Mre11 binding to a DSB, leading to Rad52 recruitment and efficient DSB end-tethering, through an Sgs1-dependent mechanism. As a consequence, deletion of RAD9 restores DSB repair either in absence of Sae2 or in presence of a nuclease defective MRX complex. We propose that, in cells lacking Sae2, Rad9/53BP1 contributes to keep Mre11 bound to a persistent DSB, protecting it from extensive DNA end resection, which may lead to potentially deleterious DNA deletions and genome rearrangements.

Chromatin modifications and chromatin remodeling during DNA repair in budding yeast.

Double-strand breaks (DSBs) pose a serious threat to genome integrity. Eukaryotes from yeast to humans respond to DSB damage by activating a complex DNA damage response that includes imposing a block to cell cycle progression and the repair of the DSB by one of several pathways. Many of these processes are accompanied by alterations in chromosome and chromatin structure. In this review we focus on the checkpoint responses and DNA repair in the well-studied model organism, the budding yeast Saccharomyces cerevisiae.

The Saccharomyces cerevisiae chromatin remodeler Fun30 regulates DNA end resection and checkpoint deactivation.

Fun30 is a Swi2/Snf2 homolog in budding yeast that has been shown to remodel chromatin both in vitro and in vivo. We report that Fun30 plays a key role in homologous recombination, by facilitating 5'-to-3' resection of double-strand break (DSB) ends, apparently by facilitating exonuclease digestion of nucleosome-bound DNA adjacent to the DSB. Fun30 is recruited to an HO endonuclease-induced DSB and acts in both the Exo1-dependent and Sgs1-dependent resection pathways. Deletion of FUN30 slows the rate of 5'-to-3' resection from 4 kb/h to about 1.2 kb/h. We also found that the resection rate is reduced by DNA damage-induced phosphorylation of histone H2A-S129 (γ-H2AX) and that Fun30 interacts preferentially with nucleosomes in which H2A-S129 is not phosphorylated. Fun30 is not required for later steps in homologous recombination. Like its homolog Rdh54/Tid1, Fun30 is required to allow the adaptation of DNA damage checkpoint-arrested cells with an unrepaired DSB to resume cell cycle progression.