Linear Evolution

Summary

Linear evolution is the pattern of clonal-evolution in which each new dominant clone arises from and completely replaces its predecessor, producing a sequential chain of clonal successions. It is the mode Nowell (1976) originally described — “survival of the fittest” sublines, each outcompeting the last — and it produces a phylogenetic-tree with a single trunk and no contemporaneous branches. Linear evolution occurs when the waiting time for the next driver-mutation exceeds the time for the current driver-bearing clone to sweep to fixation (τ_k > sweep_time), a condition that holds in small populations (early tumors, N ~ 10³–10⁵) but breaks down as tumor size grows (Greaves & Maley, 2012; Bozic et al., 2010). In clinically detected solid tumors (N ~ 10⁸–10¹¹), linear evolution is rare — clonal-interference and branching-evolution dominate — but it remains important as the theoretical null model, the pattern expected in early tumorigenesis, and the mode that may re-emerge after severe population-bottlenecks (e.g., therapy, metastasis).

Definition

Linear evolution is clonal succession without branching: clone A → clone B → clone C → …, where each clone derives from and completely replaces its immediate predecessor. Only one lineage survives at each stage; extinct intermediates leave no living descendants. The phylogenetic tree is a single chain — a “line” — with no contemporaneous branches.

This is contrasted with branching-evolution, in which multiple subclonal lineages diverge from a common ancestor and coexist, and with neutral-evolution, in which no clone has a selective advantage and diversity arises from genetic-drift alone. Together with punctuated-evolution (rapid bursts of genomic change), these four modes form Turajlic et al.’s (2019) taxonomy of cancer evolutionary patterns.

Nowell’s Original Model

Nowell (1976) described cancer evolution as a linear succession of increasingly fit clones:

“Within such a mutant subpopulation, an additional mutant may arise with an additional selective advantage with respect to the original tumor cells as well as normal cells, and this mutant becomes the precursor of a new predominant subpopulation.” (p. 24)

Nowell envisioned a stepwise process: mutation → selective advantage → clonal expansion → replacement of predecessor → new equilibrium → next mutation. Each stage is a clonal-sweep — the fitter clone expands to dominate, displacing its predecessor. The process is linear because only one lineage survives each transition; the predecessor goes extinct (or is reduced to undetectable frequency), and the successor becomes the new dominant population.

This model was the canonical description of cancer evolution for three decades and remains the conceptual starting point for all subsequent elaborations. It captures the essential Darwinian logic — variation, selection, expansion — and it correctly predicts that tumors should be clonal (derived from a single founding cell).

When Linear Evolution Occurs

Linear evolution requires clean sequential sweeps, which in turn requires that the waiting time for the next driver exceeds the sweep time of the current one. From the Bozic et al. (2010) branching process model, the waiting time for the (k+1)-th driver in a tumor of size N is:

$τ_{k} \approx \frac{T}{k s} lo g (\frac{2 k s}{u})$

where T is cell division time, s is the selective advantage per driver (~0.4%), and u is the driver mutation rate per cell division. The sweep time scales with N/s: larger populations take longer to traverse.

The condition for linear evolution — τ_k > sweep_time — holds when:

N is small (early tumor, ~10³–10⁵ cells). The sweep distance is short; a single clone can traverse the population before the next driver arises elsewhere. This is Nowell’s regime.
s is large (strong driver, s ≫ 0.4%). A high-fitness clone sweeps quickly, reducing the window for competing drivers. Some oncogenic fusions and hotspot mutations may have s > 1%, enabling linear dynamics even in moderately sized populations.
u is low (few new drivers per division). If the driver mutation rate is low, the waiting time between drivers is long, and sweeps can complete between arrivals. This may describe tumors with low genomic instability.

The condition breaks down — and linear evolution transitions to clonal-interference — when N grows large. A 1 cm³ tumor (~10⁹ cells) has a sweep time measured in years and a driver waiting time measured in months. Two or more drivers arise before any one can complete its sweep; the clones compete; neither achieves full fixation; the evolutionary pattern becomes branched (Greaves & Maley, 2012). See clonal-interference for the quantitative transition.

flowchart TD
    N1["Normal cell"] -->|"1st driver"| C1["Clone A<br>Dominant population"]

    C1 -->|"Time passes<br>mutations accumulate"| C1div["Clone A diversifies<br>(neutral passengers)"]

    C1div -->|"2nd driver arises<br>in one cell of Clone A"| C2emerg["Clone B emerges<br>from Clone A"]

    C2emerg -->|"τ₁ > sweep_time:<br>B sweeps before C arises"| C2dom["Clone B<br>New dominant population<br>Clone A extinct"]

    C2dom -->|"3rd driver arises<br>in one cell of Clone B"| C3emerg["Clone C emerges<br>from Clone B"]

    C3emerg -->|"τ₂ > sweep_time:<br>C sweeps before D arises"| C3dom["Clone C<br>New dominant population<br>Clone B extinct"]

    C1 --> Time1["Time →"]
    C2dom --> Time2["Time →"]
    C3dom --> Time3["Time →"]

    subgraph Phylogeny["Phylogenetic Tree: Linear (single trunk)"]
        N["Normal"] --> A["Clone A"]
        A --> B["Clone B"]
        B --> C["Clone C"]
    end

Figure: Linear evolution as sequential clonal sweeps. Each new driver-bearing clone arises from and completely replaces its predecessor. The phylogenetic tree is a single trunk with no branches — only one lineage survives at each stage. Extinct clones (gray) leave no living descendants. Synthesized from Nowell (1976), Greaves & Maley (2012), and Bozic et al. (2010).

Evidence and Prevalence

Where Linear Evolution Is Observed

Linear evolution is most commonly observed in:

Early tumorigenesis. The transition from normal tissue to benign lesion to invasive carcinoma often follows a linear sequence of driver acquisitions, with each stage characterized by specific mutations (e.g., APC → KRAS → TP53 in colorectal adenoma-carcinoma sequence). At these early stages, N is small, and linear dynamics are expected.
Hematologic malignancies. Some leukemias and myelodysplastic syndromes show linear clonal succession, particularly in the chronic phase before transformation to acute leukemia. The well-mixed nature of blood cancers (no spatial structure) and the strong selective advantage of certain drivers (e.g., BCR-ABL in CML) favor linear dynamics.
Post-bottleneck recovery. After a severe population-bottleneck — chemotherapy that eliminates >99% of cells, or a single-cell metastasis founding a new colony — the surviving population is small (N ~ 1–10³), and linear dynamics resume temporarily. The first driver to arise in the post-bottleneck population can sweep before competitors appear. This is why metastatic lesions often appear clonally simpler than the primary tumor from which they derived: the bottleneck resets the population size, enabling linear evolution in the newly founded colony.

Why Linear Evolution Is Rare in Established Solid Tumors

In clinically detected solid tumors, linear evolution is uncommon for three reasons (Greaves & Maley, 2012):

Large N makes sweeps slow. A clone must traverse 10⁸–10¹¹ cells to reach fixation. Even with s = 0.4%, this takes years — during which new drivers continue to arise elsewhere in the tumor.
Spatial structure prolongs coexistence. In solid tumors, clones are territorially segregated. A driver-bearing clone in one region of the tumor cannot easily outcompete clones in distant regions because cell migration is limited. Spatial structure slows sweeps and promotes branching.
Mutation supply is continuous. Mutational processes — APOBEC, clock-like signatures, therapy-induced mutagenesis — continue to generate new driver candidates throughout the tumor’s lifetime. The supply of variation is ongoing, not confined to early tumorigenesis.

Distinguishing Linear from Branching Patterns

Linear and branching-evolution produce different phylogenetic-tree structures, distinguishable by multi-region sequencing:

Feature	Linear evolution	Branching evolution
Tree shape	Single trunk, no branches	Multiple contemporaneous branches
Clonal mutations	All are truncal (present in all regions)	Truncal + branch-specific private mutations
Regional heterogeneity	Low — all regions share the same dominant clone	High — different regions carry different subclones
crossing-rule	No crossing — ancestor always has higher CCF	Crossing patterns indicate sibling relationships

Gerlinger et al. (2012) demonstrated that in renal cell carcinoma — the first multi-region sequencing study — the pattern was branching, not linear: 128 mutations resolved into a tree with multiple contemporaneous branches, with metastatic clones branching from one region of the primary tumor. Greaves & Maley (2012) subsequently noted that “the evolutionary trajectories are complex and branching, exactly as Nowell proposed and in parallel with Darwin’s iconic evolutionary speciation tree” (p. 309) — acknowledging that Nowell’s linear model was a simplification, not a refutation.