Bibliographic Reference
Gerstung, M., Jolly, C., Leshchiner, I., Dentro, S. C., Gonzalez, S., Rosebrock, D., … & Van Loo, P. (2020). The evolutionary history of 2,658 cancers. Nature, 578(7793), 122–128. https://doi.org/10.1038/s41586-019-1907-7
Core Argument
By whole-genome sequencing analysis of 2,658 cancers across 38 cancer types as part of the Pan-Cancer Analysis of Whole Genomes (PCAWG) Consortium, the life history and evolution of mutational processes and driver mutation sequences can be reconstructed. Early oncogenesis involves a constrained set of driver genes; the mutational spectrum changes significantly throughout tumor evolution in 40% of samples; and timing analyses suggest driver mutations often precede diagnosis by many years or decades — highlighting opportunities for early cancer detection.
Methods
PCAWG cohort: 2,658 whole-genome-sequenced cancers across 38 types. Molecular time reconstruction using the relative timing of clonal vs. subclonal mutations (mutations on 1 copy before vs. after copy number gain). Chromosomal gain timing via mutation copy number states. Driver gene identification from recurrently mutated genes with evidence of positive selection. Mutational signature decomposition per clone (early clonal vs. late subclonal). Phylogenetic timing models using the ratio of clonal to subclonal mutations on gained chromosomal segments.
Key Findings
- Early oncogenesis is constrained: A limited set of driver genes (e.g., TP53, PIK3CA, KRAS, PTEN, APC) are mutated early and clonally across cancer types. Specific copy number gains occur early in specific tumor types (trisomy 7 in glioblastoma, isochromosome 17q in medulloblastoma).
- Mutational spectrum shifts: In ~40% of samples, the mutational spectrum (relative contribution of different signatures) changes significantly between early (clonal) and late (subclonal) evolution. Some mutational processes are more active early, others late — implying changing exposures or repair deficiencies over the tumor’s lifetime.
- Near-fourfold diversification of driver genes occurs in later stages — subclonal driver mutations involve a much broader set of genes than clonal ones, indicating that while the initiating events are constrained, the subsequent evolutionary paths are diverse.
- Copy number changes and mitotic crises: Copy number alterations often co-occur in bursts (suggesting mitotic crises), and whole-genome doubling is a common event that precedes extensive subclonal diversification. Gains of chromosomal segments often involve synchronous acquisition of multiple segments.
- Long latency of drivers: Timing analyses suggest that many driver mutations precede clinical diagnosis by many years, often decades. Chromosomal gains in particular occur predominantly early in tumor evolution. This has implications for early detection — the genomic changes that drive cancer are present long before symptoms.
- Chromosomal instability increases late: Increased genomic instability (measured by the rate of copy number changes) is a feature of later stages of tumor evolution.
Concepts Introduced or Used
clonal-evolution, driver-mutation, passenger-mutation, mutational-signature, chromosomal-instability, whole-genome-duplication, mitotic-crisis, subclonal-architecture, molecular-time, cancer-early-detection, tumor-progression, pan-cancer-analysis, positive-selection, copy-number-alteration
Entities Referenced
- PCAWG (Pan-Cancer Analysis of Whole Genomes), ICGC, TCGA
- TP53, PIK3CA, KRAS, PTEN, APC, EGFR, BRAF, CDKN2A, RB1
- Whole-genome sequencing, molecular clock, mutational signature analysis
- 38 cancer types including glioblastoma, medulloblastoma, breast, lung, colorectal, prostate, etc.
Limitations
- Bulk sequencing data — subclonal architecture below the detection limit (~0.1 CCF) is not captured.
- Single time point per patient — longitudinal dynamics are inferred from cross-sectional data and population-level patterns.
- The evolutionary timing model assumes that mutations occurring before a chromosomal gain are on both copies (clonal), while those after are on one copy (subclonal) — this assumption may be violated in aneuploid genomes.
- Driver gene identification uses recurrence-based methods that may miss rare but potent drivers.
Relevance to Clonal Evolution
This is the largest single study of cancer evolutionary history to date, demonstrating that the evolutionary principles established by Nowell (1976) and Greaves & Maley (2012) operate universally across 38 cancer types. The finding that driver events often precede diagnosis by decades provides an evolutionary rationale for early detection. The demonstration that mutational processes shift over tumor lifetime has methodological implications for any study that treats a tumor’s mutational spectrum as static.