Volatility of Mutator Phenotypes at Single Cell Resolution

Abstract

Mutator phenotypes accelerate the evolutionary process of neoplastic transformation. Historically, the measurement of mutation rates has relied on scoring the occurrence of rare mutations in target genes in large populations of cells. Averaging mutation rates over large cell populations assumes that new mutations arise at a constant rate during each cell division. If the mutation rate is not constant, an expanding mutator population may contain subclones with widely divergent rates of evolution. Here, we report mutation rate measurements of individual cell divisions of mutator yeast deficient in DNA polymerase ε proofreading and base-base mismatch repair. Our data are best fit by a model in which cells can assume one of two distinct mutator states, with mutation rates that differ by an order of magnitude. In error-prone cell divisions, mutations occurred on the same chromosome more frequently than expected by chance, often in DNA with similar predicted replication timing, consistent with a spatiotemporal dimension to the hypermutator state. Mapping of mutations onto predicted replicons revealed that mutations were enriched in the first half of the replicon as well as near termination zones. Taken together, our findings show that individual genome replication events exhibit an unexpected volatility that may deepen our understanding of the evolution of mutator-driven malignancies.

Fig 1. Mutation rate of pol2-4 msh6Δ mother yeast cells at single cell resolution.

(A) Polymerase errors (orange, green, and blue boxes) arising in maternal double-stranded DNA (dsDNA) as mismatches become mutations during S-phase DNA replication (see rectangle) and segregate to the mother (M) or daughter (D) cells. Subscript numbers following M or D indicate the division number that produced the cell (e.g. M₁ is the mother cell after one division). Red arrows indicate only one of several segregation scenarios. Single cell mutation rates (M ₁ μ, M ₂ μ, M ₃ μ) are defined as the number of new mutations fixed in the maternal lineage at each cell division divided by the total number of nucleotides sequenced in all members of a lineage. (B) Genomic distribution of the 237 mutations observed in individual cell divisions (blue lines) among the 16 yeast chromosomes (gray lines). Red lines, centromeres. (C) Mutation spectra of pol2-4 msh6Δ cells from whole genome sequencing (blue) compared to published spectra (red).

Fig 2. Distribution of single cell mutation counts in mother cells.

(A) Mutation counts of mutator cells during yeast aging. Each lineage is plotted separately, except Lineage G, whose mother cell became multi-budded, producing two distinct lineages: G1 (blue) and G2 (light blue). The gap in Lineage F is due to sequencing failure of one of the daughter clones. Bottom right, all lineages are plotted together, each represented by a different line color. (B) The observed distribution (blue bars, combined data) of mutation counts is plotted against the predicted Poisson distribution based on the average genome-wide mutation rate (2.6 x 10^-7 mutations/bp/cell division) (orange bars, Poisson Model) and a composite distribution resulting from two overlapping Poisson distributions with mutation rates of 4 x 10^-8 (contributing 35%) and 4 x 10^-7 (contributing 65%) mutations/bp/cell division (green bars, Two-Poisson Model).

Fig 3. Co-occurrences of mutations in the same chromosome and cell division.

The random accumulation of 237 mutations over 87 cell divisions was simulated using 10,000 iterations. green histogram, co-occurrences in the simulation with 3 mutations on the same chromosome; green vertical line, actual co-occurrences with 3 mutations; black histogram, co-occurrences with 2 or more mutations on the same chromosome; gray shading, 95% confidence intervals; black vertical line, actual co-occurrences with 2 or more mutations.

Fig 4. Distribution of mutations within replicons.

Top. Histogram of the fractional distances of mutations to the nearest origin using the replicons defined by Raghuraman et al. [39] (see S3 Dataset). Bottom. Schematic shows how fractional distances are determined.