Clonal Evolution

Clonal Evolution

13 min read

Clonal Evolution

Definition

Clonal evolution is the process by which a neoplasm arises from a single cell of origin and progresses through the stepwise acquisition of somatic genetic variation, followed by selection of variant subpopulations with greater fitness (Nowell, 1976). It is the application of Darwinian natural selection — mutation, heritable variation, differential reproduction — to somatic cell populations within a multicellular organism.

The Evolutionary Process

flowchart TD
    A["Single Cell<br/>of Origin"] --> B["Acquired Mutation(s)<br/>+ Genetic Instability"]
    B --> C["Heritable Variation<br/>within Expanding Clone"]
    C --> D{"Fitness<br/>Effect?"}
    D -->|"Advantageous<br/>(driver)"| E["Positive Selection<br/>Clonal Expansion"]
    D -->|"Neutral<br/>(passenger)"| F["Genetic Drift<br/>Stochastic Fluctuation"]
    D -->|"Deleterious"| G["Negative Selection<br/>Immune Elimination"]
    E --> H["New Dominant<br/>Subclone"]
    F --> H
    H --> I["Continued Mutagenesis<br/>Ongoing Diversification"]
    I --> J["Subclonal<br/>Branching"]
    J --> C
    J --> K["Branched Tumor<br/>with ITH"]
    G --> L["Extinct Lineage"]

The clonal evolutionary cycle. A single cell acquires mutations that confer heritable variation. Variants face one of three fates: positive selection (driver mutations expand), neutral drift (passenger-harboring lineages fluctuate stochastically), or negative selection (deleterious variants, including immunogenic neo-antigen-bearing cells, are eliminated). Surviving lineages continue to mutate, producing the branched subclonal architecture characteristic of solid tumors. This feedback loop — mutation, variation, selection/drift, diversification — operates throughout the tumor’s lifetime. Synthesized from Nowell (1976), Greaves & Maley (2012), McGranahan & Swanton (2017), Turajlic et al. (2019), and PCAWG Consortium (2020).

Origin

Nowell (1976) laid out the model in its canonical form. Tumor initiation occurs by an induced change in a single normal cell that provides it with a selective growth advantage over adjacent normal cells. As the neoplastic clone expands, its acquired genetic-instability produces mutant subpopulations. Most variants are eliminated (metabolic disadvantage, immune destruction), but occasionally one has an additional selective advantage and becomes the precursor of a new predominant subpopulation. Over time, there is “sequential selection by an evolutionary process of sublines which are increasingly abnormal, both genetically and biologically” (Nowell, 1976, p. 24).

Greaves & Maley (2012) updated the theory for the genomic era, formalizing cancer as “a complex, Darwinian, adaptive system” (p. 306). They emphasized that the dynamics are complex — not always successive clonal-sweeps, but often involving clonal-interference, parallel expansions, and long periods of stasis. This complexity arises because driver mutations confer relatively small fitness advantages (~0.4% on average, estimated from glioblastoma and pancreatic cancer data; Bozic et al., 2010, as cited by Greaves & Maley, 2012), meaning multiple subclones with similar fitness can coexist rather than one rapidly outcompeting the rest.

The Darwinian character of cancer is not metaphorical — it is the same evolutionary logic operating at the somatic cell level. Greaves & Maley (2012) define a Darwinian evolutionary system as having two essential components: (i) “purposeless genetic variation of reproductive individuals who are united by common descent” and (ii) “natural selection of the fittest variants” (p. 306). The PCAWG Consortium (2020) frames the same principle with three preconditions: “characteristics must vary within a population; this variation must be heritable from parent to offspring; and there must be competition for survival within the population” (p. 83). Both formulations converge on the same structure: heritable variation among individuals of common ancestry, subjected to differential reproductive success. Cancer satisfies these conditions: mutations arise blindly through endogenous and exogenous mutagenic processes, all tumor cells descend from a single cell of origin, and clones bearing driver-mutations are selected for expansion relative to less-fit competitors. The recurrence of the same driver mutations across independent tumors is natural selection’s fingerprint — many variants arise blindly, but only the fit ones repeatedly dominate.

The pan-cancer genomic landscape, revealed by PCAWG Consortium (2020) across 2,658 whole-genome-sequenced cancers spanning 38 tumor types, provides the empirical substrate: 91% of tumors harbor at least one identified driver mutation, with an average of 4.6 drivers per tumor. In ~5% of cases, no drivers were identified — possibly reflecting driver events in non-coding regions, epigenetic alterations, or detection limits of current methods.

Core Principles

Common descent. All cells in a neoplasm are “united by common descent” (Greaves & Maley, 2012, p. 306) — they trace back to a single cell of origin. This is what makes clonal evolution a Darwinian system rather than a collection of independently transformed cells. Common descent is why all tumor cells share truncal (clonal) mutations, and it is the foundation on which phylogenetic-tree reconstruction rests. Without common descent, the very concept of a “clonal” mutation — present in every tumor cell — would be incoherent (McGranahan & Swanton, 2017).

Variation. Somatic mutations arise continuously through endogenous processes (replication errors, APOBEC-mutagenesis, defective DNA repair) and exogenous exposures (tobacco carcinogens, UV light, chemotherapy). The mutation rate itself can be subject to selection through mutator-phenotypes (Greaves & Maley, 2012). Critically, mutagenic processes are not confined to early tumorigenesis — APOBEC family cytidine deaminases can remain active late in tumor evolution, generating ongoing subclonal mutational diversity (McGranahan & Swanton, 2017). Germline genetic variation also shapes the somatic mutation landscape: PCAWG Consortium (2020) identified rs12628403 at 22q13.1 associated with APOBEC3B-like mutagenesis, BRCA1/2 protein-truncating variants associated with specific structural variant classes, and MBD4 protein-truncating variants associated with increased CpG>T mutation rates. Variation in cancer is thus the product of an interaction between inherited mutational propensity and acquired mutagenic processes.

Selection. Clones with fitness-conferring driver-mutations expand relative to competitors. positive-selection drives tumor progression; negative-selection eliminates cells with strongly deleterious mutations, including potent neo-antigens that provoke immune responses (Turajlic et al., 2019). The average selective advantage conferred by a driver mutation is modest — approximately 0.4% (Greaves & Maley, 2012, citing Bozic et al., 2010) — which explains why tumors typically require multiple drivers to reach clinical detectability. PCAWG Consortium (2020) found an average of 4.6 drivers per tumor, consistent with the accumulation of individually small-effect mutations over years to decades. Selection operates on pre-existing variation: treatment resistance mutations frequently pre-exist as minor subclones before therapy begins (Turajlic et al., 2019).

Drift. Between selection events, neutral-evolution dominates. Random fluctuations in birth and death rates cause some lineages to expand by chance while others contract. Turajlic et al. (2019) emphasize that “selection is not operative at all times; neutral evolution dominates between selection events.” This neutral diversity provides the standing variation from which adaptive clones can emerge under changing conditions — including the selective pressure of therapy. The balance between drift and selection is not fixed: drift dominates in small populations and during periods of stasis; selection dominates during population expansions and under strong environmental change.

Branching architecture. Evolution is always branched because mutation and cell division continuously produce genotypic divergence. Branching of a phylogenetic-tree does not always imply selection — it is the natural product of mutational processes in proliferating tissues (Turajlic et al., 2019). Selection prunes the tree by favoring some branches over others, but the underlying branching structure persists. Chromosomal instability (CIN) and whole-genome doubling (WGD) events amplify branching complexity: CIN generates large-scale copy-number heterogeneity, and WGD provides a permissive background for further chromosomal aberrations — both are associated with worse clinical outcome (McGranahan & Swanton, 2017). PCAWG Consortium (2020) found chromothripsis in 22.3% of samples, predominantly clonal (early) and enriched for driver events (3.6% of all drivers), illustrating how catastrophic genomic events can instantaneously generate branched architectures from which selection then acts.

Evolutionary Modes

Turajlic et al. (2019) formalized a taxonomy of four evolutionary modes observed across human cancers. These modes are not mutually exclusive within a single tumor — a neoplasm may transition between modes at different stages of its evolutionary trajectory — and all arise from different combinations of the same three fundamental processes: mutation, genetic drift, and natural selection.

flowchart LR
    subgraph Processes
        M["Mutation<br/>(rate, spectrum, timing)"]
        D["Genetic Drift<br/>(population size, turnover)"]
        S["Natural Selection<br/>(fitness differential, environment)"]
    end

    subgraph Modes
        L["Linear Evolution<br/>Sequential selective sweeps<br/>Strong selection, low diversity"]
        B["Branching Evolution<br/>Subclonal diversification<br/>Ongoing selection + drift"]
        N["Neutral Evolution<br/>No strong selection<br/>Drift dominates, high diversity"]
        P["Punctuated Evolution<br/>Rapid genomic catastrophe<br/>Chromothripsis, WGD, early CIN"]
    end

    M --> L
    M --> B
    M --> N
    M --> P
    D --> B
    D --> N
    S --> L
    S --> B
    S --> P

Evolutionary modes arise from different combinations of the three fundamental processes (mutation, drift, selection). Linear evolution: strong positive selection drives sequential clonal sweeps, producing low subclonal diversity. Branching evolution: ongoing mutation and intermediate selection produce coexisting subclones. Neutral evolution: drift dominates between selection events, generating high diversity without strong fitness differentials. Punctuated evolution: catastrophic genomic events (chromothripsis, WGD) rapidly restructure the genome in a single cell division, followed by strong selection on the resulting clones. Synthesized from Turajlic et al. (2019).

Linear evolution. Sequential selective sweeps in which a new driver mutation produces a clone that outcompetes and replaces the previous population. This is the mode Nowell (1976) originally described. It produces relatively low subclonal diversity at any given time because each sweep purges competing lineages. Linear dynamics are most visible when selection coefficients are large relative to mutation rate and population size.

Branching evolution. Multiple subclones coexist, each bearing distinct mutations, with no single clone achieving complete dominance. This is the most commonly observed pattern in solid tumors by multi-region sequencing. Branching arises when driver fitness advantages are modest (~0.4%; Greaves & Maley, 2012) and mutation rates are sufficiently high to continuously generate new variants before any one sweeps to fixation. Clonal interference — competition between multiple adaptive lineages — is a hallmark of branching evolution.

Neutral evolution. Between selection events, subclonal mutations accumulate without detectable fitness consequences. Allele frequencies fluctuate through stochastic birth-death processes. Turajlic et al. (2019) note that the observation of neutral dynamics does not imply that selection never operates in that tumor — only that at the time of sampling, no clone has a fitness advantage large enough to be detected above the background of drift. Bulk sequencing can overestimate the prevalence of neutral evolution because subclones below ~7 doublings of growth fall below detection at standard 100× sequencing depth.

Punctuated evolution. Rapid, large-scale genomic alteration — chromothripsis, whole-genome doubling, or explosive chromosomal instability — occurring early in tumor evolution, followed by relative stasis. This mode produces tumors with high clonal aneuploidy, low subclonal diversity (because the catastrophe occurs as a single early event), and aggressive clinical behavior. Turajlic et al. (2019) report that punctuated vs. gradual evolution predicts clinical phenotype: tumors with early clonal aneuploidy tend to grow fast, metastasize widely, and seed metastases monophyletically, while gradual Darwinian evolution produces slower growth, oligometastases, and intermetastatic heterogeneity. PCAWG Consortium (2020) confirms chromothripsis is predominantly clonal (early), present in 22.3% of samples, and enriched for driver events.

Clinical Significance

The evolutionary capacity of tumors — their ability to generate genetic diversity and select for resistant variants — is the primary reason for therapeutic failure in advanced cancers. As Nowell (1976) first noted, the “capacity for variation and selection which permitted the evolution of a malignant population from the original aberrant cell also provides the opportunity for the tumor to adapt successfully to the inimical environment of therapy” (p. 27). Greaves & Maley (2012) reaffirmed this half a century later: “The inherently Darwinian character of cancer is the primary reason for therapeutic failure” (p. 306). Every subsequent source — McGranahan & Swanton (2017), Turajlic et al. (2019), PCAWG Consortium (2020) — converges on this point.

This evolutionary understanding underpins several modern therapeutic strategies:

Truncal (clonal) targeting. Because truncal mutations — including canonical driver events in EGFR, KRAS, TP53, and KEAP1 — occur early and are shared by all tumor cells, they represent the most homogeneous therapeutic targets. McGranahan & Swanton (2017) argue that clonal neoantigens derived from truncal mutations are expressed more uniformly across the tumor mass than subclonal neoantigens and may therefore represent superior targets for immunotherapy. A therapy directed at a subclonal alteration will, by definition, spare the subclones that lack it — providing them with a selective advantage and setting the stage for relapse.

Adaptive therapy. Rather than attempting maximal tumor cell kill (which applies intense selective pressure favoring resistant variants), adaptive therapy aims to maintain a stable population of chemosensitive cells that suppress resistant subclones through competition. Greaves & Maley (2012) proposed this as an “ecological therapy” approach — cytostatic rather than cytotoxic — drawing on principles from pest management and bacterial antibiotic resistance. The goal shifts from eradication (which evolution often defeats) to control (which works with, rather than against, evolutionary dynamics).

Evolutionary mode as prognostic marker. The evolutionary mode of a tumor — punctuated vs. gradual — predicts its clinical trajectory. Turajlic et al. (2019) report that tumors with early clonal aneuploidy (punctuated evolution) grow fast, metastasize widely, and seed metastases monophyletically, while tumors following gradual Darwinian evolution produce slower growth, oligometastases, and genetically distinct metastases. Knowing the evolutionary mode at diagnosis could inform surveillance intensity, surgical margins, and adjuvant therapy decisions.

Early detection. The earlier a tumor is detected, the fewer cell divisions it has undergone, and the less subclonal genetic diversity it harbors. Greaves & Maley (2012) demonstrated that subclonal genetic diversity in Barrett’s esophagus is a robust biomarker for predicting progression to esophageal adenocarcinoma, and they advocated for “prevention and early detection before genetic diversification becomes extensive.” A tumor detected before it has generated extensive subclonal diversity is less able to evolve around any single therapeutic intervention.

Pre-existing resistance. Treatment resistance mutations frequently pre-exist as minor subclones before therapy begins, rather than arising de novo during treatment. Turajlic et al. (2019) document polyclonal resistance — multiple independent resistant clones emerging simultaneously under therapy — and note that fitness costs of resistance mutations can sometimes be exploited therapeutically (e.g., drug holidays that allow chemosensitive clones to re-expand).

Pan-cancer heterogeneity. The PCAWG Consortium (2020) revealed that the driver landscape varies enormously across tumor types — from near-zero identified drivers in ~5% of cases to highly mutated tumors with numerous drivers. Therapeutic strategies keyed to clonal evolution must therefore be tailored to the specific evolutionary parameters of each cancer type, and ideally to each individual tumor. Nowell’s (1976) prescient conclusion that “each patient’s cancer may require individual specific therapy” is borne out by the genomic data: every tumor has a unique evolutionary history and a unique repertoire of resistance pathways.

Graph View

Backlinks