Clonal Evolution

Clonal Sweep

Properties1
tagsselection

9 min read

Clonal Sweep

Summary

A clonal (selective) sweep occurs when a subclone harboring a fitness-conferring driver mutation expands to dominate the tumor population, reducing genetic diversity by displacing competing lineages (Nowell 1976; Greaves & Maley 2012). The sweep requires that the time to the next driver mutation exceeds the time for the current clone to reach fixation — a condition formalized in the branching process model (Bozic et al. 2010). Sweeps may be complete (all cells carry the adaptive mutation), incomplete (the dominant clone coexists with residual competitors), or soft (multiple clones carrying the same driver sweep simultaneously). The sweep-bottleneck cycle — sweep → diversity collapse → mutation accumulation → diversity recovery → next sweep — explains the punctuated pattern observed in longitudinal tumor sequencing (TRACERx; Al Bakir et al. 2023).

A clonal sweep (also called a selective sweep) is the process by which a subclone harboring a fitness-conferring driver-mutation expands to dominate the tumor cell population, reducing genetic diversity by driving competing lineages to low frequency or extinction. The term is borrowed directly from population genetics, where a selective sweep describes the reduction of genetic variation near a positively selected allele.

Mechanism

In Nowell’s (1976) model, when a variant subpopulation acquires an additional selective advantage, “this mutant becomes the precursor of a new predominant subpopulation” (p. 24). As this new dominant clone expands, the previously dominant clone is displaced. The sweep may be complete (all cells in the population carry the adaptive mutation) or incomplete (the adaptive clone dominates but does not fully eliminate competitors).

When Sweeps Can Occur

A clonal sweep requires a simple mathematical condition to be met (Greaves & Maley, 2012):

Time to next driver mutation > Time required for the current clone to sweep through the population.

flowchart TD
    N[N = tumor size] --> TauK[τ_k ≈ T/ks × log 2ks/u<br/>waiting time for next driver]
    N --> SweepT[Sweep time ∝ N / ks<br/>time for clone to reach fixation]
    TauK --> Cond{"τ_k > sweep time?"}
    SweepT --> Cond
    Cond -->|"YES: early tumors N ~ 10³−10⁵"| Sweep[Clean sequential sweeps]
    Cond -->|"NO: late tumors N ~ 10⁸−10¹¹"| Interference[Clonal interference]
    Sweep --> Result1["Nowell's sequential sublines<br/>one clone dominates at a time"]
    Interference --> Result2["Branching evolution<br/>multiple clones coexist and compete"]
    Therapy[Therapy] -.->|"kills sensitive cells<br/>artificially shortens sweep time"| Sweep

If the next driver mutation appears before the previous clone has reached fixation, the two clones compete. The first never completes its sweep — the result is clonal-interference, not a clean succession.

This condition explains why sweeps dominate early tumor evolution but become rare later:

Early tumors are small (N ~ 10³–10⁵). The sweep time — the number of generations for a clone to expand from one cell to dominance — is short. Meanwhile, the waiting time for the next driver is long because the target population (the N cells that could mutate) is small. The condition holds: τ_k > sweep time. Clean sequential sweeps are the expected pattern. This is the regime Nowell (1976) described: “sequential selection by an evolutionary process of sublines.”

Late tumors are large (N ~ 10⁸–10¹¹). The sweep time is long — a clone must traverse an enormous population. The waiting time for the next driver is short because the target population is vast. The condition fails: a new driver mutation appears before the previous clone can sweep. Multiple clones expand simultaneously, competing for resources and space. The result is branching evolution with clonal interference, not successive sweeps. Greaves & Maley (2012) emphasized this shift: clonal evolution is “not always successive selective sweeps.”

Mutation rate matters. Higher mutation rates (e.g., from mutator-phenotypes or chromosomal-instability) shorten τ_k, making sweeps less likely at any given tumor size. Lower rates extend τ_k, preserving the sweep regime further into tumor growth. The Bozic et al. (2010) waiting-time formula quantifies this: τ_k ≈ (T/ks) × log(2ks/u). As u increases, τ_k decreases — the next driver arrives sooner.

Fitness advantage matters. A larger s shortens both sweep time (the clone grows faster) and τ_k (more divisions = more mutation opportunities), but the effects are not symmetric. The sweep time scales as 1/(ks); τ_k scales as 1/k × log(1/u). For fixed u, variation in s affects sweep time more strongly, meaning strongly advantageous drivers are more likely to complete sweeps.

Therapy removes competition. Cytotoxic treatment eliminates sensitive cells, clearing the field for any resistant clone. This artificially shortens sweep time while τ_k remains unchanged — the condition for a sweep is imposed by treatment. This is why therapy-induced sweeps are the central mechanism of treatment failure (see therapy-resistance).

Complete vs. Incomplete Sweeps

When a clonal sweep is complete, the within-tumor evolution reverts to neutral-evolution with respect to that selective event — the population is now homogeneous for the adaptive mutation, and further selection acts on new variants arising within the new dominant clone (Turajlic et al., 2019).

When a sweep is incomplete, multiple lineages coexist, and evolution continues to be shaped by both selection and clonal-interference.

Detection

Detecting past clonal sweeps from genomic data is challenging. Once a selected clone has taken over and reached fixation, the population is homogeneous with respect to the selective alteration, and “dense longitudinal sampling is necessary to accurately detect selection” (Turajlic et al., 2019, p. 415). Without longitudinal data, sweeps that occurred before sampling are invisible — evolution appears neutral even though selection was the driving force.

Metastatic Timing

Al Bakir et al. (2023) used the last clonal sweep in the primary tumour as the reference point for timing metastatic divergence in NSCLC. Primary clonal mutations that arose during or before the last sweep are present in all primary tumour cells. If the metastasis shares these mutations, divergence occurred after the last sweep (late divergence, ~75% of cases). If the metastasis lacks them, divergence occurred before the last sweep (early divergence, ~25%), and the primary subsequently underwent a complete clonal sweep that the metastatic clone missed. This framework makes clonal sweeps the fundamental clock for timing metastatic events in tumour evolution.

Punctuated Sweeps

In punctuated-evolution, multiple driver events are acquired in a short burst, producing a rapid and comprehensive clonal sweep that results in a functionally homogeneous tumor mass. Such tumors are characterized by low driver intratumour heterogeneity and high levels of clonal aneuploidy that became fixed early in evolution (Turajlic et al., 2019). In TRACERx Renal, tumors with this pattern grew rapidly, metastasized widely, and had worse outcomes than those with ongoing subclonal diversification.

Sweeps as Compression Breakthroughs

A clonal sweep can be understood as a compression breakthrough — the moment when the evolutionary system discovers a genomic program that encodes the microenvironment more efficiently than any competitor. This framing is adapted from Schmidhuber’s (2009) compression progress principle, which defines a “discovery” as an unusually large compression improvement (see compression-progress-evolution).

In Schmidhuber’s framework, a curious agent generates intrinsic reward when it improves its compression of sensory history. The analogue in clonal evolution: a driver mutation that confers a selective advantage is a new compression of the tumor microenvironment — a shorter algorithmic description of “how to survive and proliferate in this niche.” The clonal sweep IS the moment of compression progress: the period of rapid clonal expansion before the new genotype becomes the dominant compressed representation. Once the sweep is complete, the tumor genome plateaus — well-compressed (high fitness) but “boring” in Schmidhuber’s sense (no further compression progress until the next driver event).

This reframing has practical implications:

  • Sweep likelihood depends on compression gradient steepness. The probability that a mutation triggers a sweep is not merely a function of its absolute fitness effect (s) but of the compression gradient — how much better the new genomic program encodes the microenvironment compared to competitors. A mutation may have a large s (fitness advantage) because it compresses a previously incompressible aspect of the microenvironment, or because it finds a more compact encoding of an already-compressed regularity. The former is a discovery; the latter is an optimization.
  • Sweeps as punctuated learning. Each clonal sweep is a discrete learning event for the tumor — a step improvement in its compression of the microenvironment. Cancer progression is thus a sequence of compression breakthroughs punctuated by periods of stasis, exactly the pattern described by punctuated-evolution.
  • Therapy as forced decompression. Cytotoxic therapy can be understood as forcibly decompressing the tumor’s genomic program — disrupting the compressed representation and forcing the tumor to either find a new compression (resistance) or die. Adaptive therapy, by maintaining competitive dynamics, prevents any single clone from completing its compression of the therapeutic environment.

Clinical Significance

Clonal sweeps under therapy are a primary mechanism of treatment failure. Chemotherapy and targeted therapy exert strong selective pressure that can trigger sweeps of pre-existing resistant subclones. As noted by Nowell (1976), “the same 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).

Therapy-induced bottlenecks as incomplete sweeps. When therapy achieves a deep response, it can create a clonal bottleneck — a severe population reduction that eliminates most subclones, leaving only resistant or dormant survivors. This bottleneck functions as a forced sweep: therapy, not natural selection, clears the field. However, unlike a natural clonal sweep where one fit clone expands to dominance, the therapy bottleneck can produce branching re-diversification at relapse. In the Myeloma XI trial, patients achieving CR/vgPR underwent a bottleneck leading to branched clonal architecture upon relapse, while patients with partial responses maintained linear or stable clonal patterns (Miething, 2019). The bottleneck paradox — that deeper responses produce more diverse relapses — is consistent with compression-progress-evolution: the decompression event is so severe that surviving clones must re-explore the fitness landscape from scratch, generating new diversity rather than converging on a single compression. See population-bottleneck for cross-domain synthesis.

Graph View

Backlinks