Bibliographic Reference
Sottoriva, A., Kang, H., Ma, Z., Graham, T. A., Salomon, M. P., Zhao, J., Marjoram, P., Siegmund, K., Press, M. F., Shibata, D., & Curtis, C. (2015). A Big Bang model of human colorectal tumor growth. Nature Genetics, 47(3), 209–216. https://doi.org/10.1038/ng.3214
Core Argument
After initial transformation, colorectal tumors grow predominantly as a single clonal expansion in which numerous intermixed subclones co-exist without stringent selection. This “Big Bang” model predicts that both public (clonal) and most detectable private (subclonal) alterations arise early during growth, with the timing of an alteration — not selection for it — being the primary determinant of its prevalence in the final tumor. Selective sweeps are rare because the population expands too rapidly for selection to operate effectively before spatial constraints freeze the clonal architecture in place.
Methods
Genomic profiling of 349 individual tumor glands sampled from opposite sides of 15 colorectal tumors (11 carcinomas, 4 advanced adenomas), averaging ~23 glands per tumor. Not all glands received all assays: 127 glands (7–10 per tumor) were profiled by whole-genome SNP arrays for copy number alterations; 102 glands were subjected to targeted deep sequencing of patient-specific mutations; 65 glands were evaluated by FISH for single-cell copy number heterogeneity at the HER2 locus; and 55 glands underwent ultra-deep single-molecule methylation tag sequencing. Alterations were classified into six spatial categories (public, side-specific, side-variegated, variegated, regional, unique) based on presence across individually sampled glands. A single Approximate Bayesian Computation (ABC) inference framework was used, built on a three-dimensional cellular automaton simulating gland-level tumor growth on a 400×400×400 lattice (~8 million glands, ~80 billion cells, ~5.3 cm diameter). The framework infers patient-specific mutation rates and subclone fitness changes by varying a fitness parameter sigma (sigma in {0, 0.2, 0.6}, corresponding to neutral, moderate, and large fitness effects respectively).
Key Findings
- Big Bang growth validated. All 15 tumors exhibited Big Bang dynamics: public alterations were uniformly present, and most private alterations were pervasive with variegated spatial patterns (detectable in glands from opposite tumor sides), rather than regional or unique as would be expected from late-arising clones under sequential selection. The paper consistently reports an absence of selective sweeps across all tumors.
- Selection is not required to explain ITH. The ABC modeling showed that the observed spatial distribution of alterations is explained by the timing of mutations during expansion rather than by fitness differences. Early-arising private mutations become pervasive simply because they had more time to expand before spatial constraints froze the architecture. Late-arising mutations, even if advantageous, remain localized.
- “Born to be bad” signature — variegation restricted to carcinomas. Variegated alterations (identical private mutations found in glands from opposite tumor sides, several centimeters apart) were detected in the majority of carcinomas but in none of the adenomas. This is a cross-sectional observation: the adenomas and carcinomas are separate tumors from different patients. The authors hypothesize that early subclone intermixing within the primordial tumor — made possible by loss of normal cell adhesion and disorganized growth — reflects the early emergence of an invasive phenotype. They propose that the degree of subclone mixing could serve as a biomarker for predicting which adenomas may progress to invasion.
- Primordial tumor profile recoverable. Because late selective sweeps are rare, the genomic profile of the early (primordial) tumor — when it contained fewer than ~10,000–100,000 cells — can be reconstructed from the final tumor by identifying pervasive private alterations found across both tumor sides at subclonal frequencies.
Concepts Introduced or Used
- Big Bang model: tumor growth as a single expansion with intermixed subclones, where alteration timing rather than selection determines prevalence
- clonal-evolution — the Big Bang model is a specific mode of clonal evolution characterized by early burst followed by spatially constrained growth
- clonal-expansion — modeled as a single dominant expansion rather than successive waves
- neutral-evolution — selective sweeps are argued to be extremely rare; ITH arises from mutation timing during expansion, not selection
- subclonal-architecture — characterized by pervasive early private alterations and variegated spatial patterns
- intratumor-heterogeneity — uniformly high ITH observed, with subclone mixing in distant tumor regions in carcinomas
- clonal-sweep — explicitly tested for and found absent in all tumors
- driver-mutation — early driver alterations (APC, KRAS, TP53) are public/clonal; later drivers would be localized if they exist
- passenger-mutation — private alterations are mostly passengers arising during expansion
- branching-evolution — the Big Bang model produces star-shaped/branched phylogenies compatible with long-lived cancer stem cell lineages
Entities Referenced
- Cancer type: Colorectal carcinoma (11 tumors) and advanced adenomas (4 tumors)
- Genes: APC, KRAS, TP53 (clonal driver events), PIK3CA, CTNND1, SOX7, CDH1, HER2 (ERBB2)
- Methods: Approximate Bayesian Computation (ABC), 3D cellular automaton, whole-genome SNP arrays (Illumina OmniExpress), whole-exome sequencing (HiSeq), targeted deep sequencing (Ion Torrent PGM), FISH (HER2/CEP17 probes), neutral methylation tag sequencing, MEDICC (phylogenetic reconstruction from copy number)
- Sampling: 349 individually isolated tumor glands total, averaging ~23 per tumor, from 15 patients; 22 bulk samples (left and right sides)
Limitations (as stated by authors)
- Not every tumor may exhibit Big Bang dynamics; “selective bottlenecks” may be common for markedly different environments, such as metastatic seeding to foreign sites or during treatment.
- The microenvironment was modeled as a static entity; the model does not account for dynamic interactions between tumor cells and their microenvironment, nor for clonal cooperation or interference.
- Complex and poorly characterized processes — cellular migration, apoptosis within a gland, contribution of surrounding normal tissue, angiogenic factors — are not modeled.
- Model developed and validated in colorectal tumors only; generalizability to other cancer types was not established in this study.
Relevance to Clonal Evolution
This is a landmark paper in cancer evolution. By demonstrating that the timing of mutations during a single early expansion — not ongoing selection — produces the ITH patterns observed in established colorectal tumors, it challenges the assumption that Darwinian selection is continuously operative throughout tumor growth. The Big Bang model bridges neutral-evolution theory and spatial growth dynamics, and its finding that growth kinetics (expansion rate, spatial constraints) override selection in determining clonal architecture has shaped subsequent work on neutral tumor evolution and the interpretation of cancer-cell-fraction distributions.