35.03.03 · health-medicine / chronic-disease

Cancer biology: hallmarks of cancer (Hanahan-Weinberg), oncogenes, tumor suppressors

stub3 tiersLean: nonepending prereqs

Anchor (Master): Hanahan, D. and Weinberg, R. A. — Hallmarks of Cancer: The Next Generation (2011)

Intuition Beginner

Cancer is not one disease. It is more than 200 diseases, all united by a single feature: cells that grow uncontrollably. Normal cells divide only when the body tells them to, self-destruct when their DNA is damaged, and respect the boundaries of their neighbors. Cancer cells ignore every one of these rules. They divide without permission, survive damage that should kill them, and invade tissues where they do not belong.

Douglas Hanahan and Robert Weinberg distilled cancer into a set of hallmarks — the capabilities every cancer acquires. A cancer must make its own growth signals, ignore the signals that command it to stop, resist the programmed death that clears damaged cells, divide without limit, recruit its own blood vessels, and invade other organs. Later, they added more: rewiring energy metabolism, hiding from immune attack, hijacking inflammation, and accelerating the mutation rate.

Cancer arises from DNA mutations. Some are inherited — the BRCA1 and BRCA2 changes carried in families prone to breast and ovarian cancer. Most are acquired over a lifetime, caused by carcinogens such as tobacco smoke, ultraviolet light, and radiation, or by random copying errors during cell division. Oncogenes are mutated growth-promoting genes: RAS and MYC behave like a gas pedal jammed to the floor. Tumor suppressor genes are the brakes — when p53 and RB are mutated, the brakes fail.

Most cancers need several mutations in different genes, which is why age is the dominant risk factor — every year of cell division is another roll of the dice. Tumors are not uniform. They contain diverse cell populations competing for resources, and treatment kills the sensitive cells while selecting for the resistant ones. This is Darwinian natural selection, playing out inside a single body. Immunotherapy — releasing the immune system to attack the tumor — has transformed treatment for melanoma, lung cancer, and several other types.

Visual Beginner

Hallmark	Normal function	What cancer does
Sustaining proliferative signaling	Cells divide on external signal	Produce own growth signals
Evading growth suppressors	Stop signals halt division	Ignore the stop signals
Resisting cell death	Damaged cells self-destruct	Survive lethal damage
Enabling replicative immortality	Limited number of divisions	Divide without limit (telomerase)
Inducing angiogenesis	Vessels grow on demand	Recruit a private blood supply
Activating invasion and metastasis	Cells stay in their tissue	Spread to distant organs
Deregulating cellular energetics	Efficient aerobic respiration	Switch to glycolysis (Warburg effect)
Avoiding immune destruction	Immune cells kill abnormal cells	Hide from immune patrols

Gene	Type	Role	Cancer association
RAS (KRAS, NRAS)	Oncogene	Growth signal relay	Pancreatic, colorectal, melanoma
MYC	Oncogene	Transcription factor	Burkitt lymphoma, many others
HER2	Oncogene	Receptor tyrosine kinase	Breast cancer
BCR-ABL	Oncogene (fusion)	Constitutive kinase	Chronic myeloid leukemia
TP53 (p53)	Tumor suppressor	DNA damage response	~50% of all cancers
RB1	Tumor suppressor	Cell-cycle checkpoint	Retinoblastoma, many cancers
APC	Tumor suppressor	Wnt pathway regulation	Colon cancer
BRCA1 / BRCA2	Tumor suppressor	DNA double-strand repair	Breast, ovarian cancer

Worked example Beginner

Why does cancer risk rise so steeply with age?

In 2015, Cristian Tomasetti and Bert Vogelstein compared lifetime cancer risk across 31 tissue types against the number of stem cell divisions each tissue undergoes. Their reasoning was direct: every cell division is an opportunity for a copying error, and every copying error is a potential cancer-driving mutation.

The correlation was striking. Tissues with many stem cell divisions — the colon, with roughly a trillion over a lifetime — have far higher cancer rates than tissues with few, like bone. The correlation coefficient was 0.8, meaning about two-thirds of the variation in cancer risk between tissues tracks the division rate.

This does not mean two-thirds of cancers are unpreventable. Inherited mutations raise the starting point, and carcinogens like tobacco add to the baseline. But the engine of cancer is cell division itself. A seventy-year-old colon has produced quadrillions of cells from a single starting layer; each division was a chance for error.

The practical consequence: a cancer that needs five or six mutations in sequence typically takes decades to assemble them. This is why colon cancer is rare before age forty and common after sixty. Each driver mutation expands a clone, and the next mutation arises within that enlarged population — a staircase climbed one step at a time.

Check your understanding Beginner

Exercise 3 (easy, multiple choice).

According to the Hanahan-Weinberg hallmarks, which capability allows a cancer to spread to other organs?

A. Sustaining proliferative signaling

B. Activating invasion and metastasis

C. Resisting cell death

D. Enabling replicative immortality

Hint

Only one hallmark is specifically about breaking through tissue boundaries and colonizing distant sites. The others govern growth, survival, and division within the primary tumor.

Answer

Option B. Activating invasion and metastasis is the hallmark that enables cancer cells to breach the basement membrane, enter the bloodstream, and establish colonies in distant organs. Metastasis is responsible for the majority of cancer deaths. The other hallmarks (proliferation, death resistance, immortality) govern the primary tumor's growth.

Formal definition Intermediate+

Cancer (synonymous in clinical usage with malignant neoplasm) denotes a family of diseases in which cells acquire heritable genetic and epigenetic alterations that confer uncontrolled proliferation, the capacity to invade surrounding tissue, and the ability to metastasize to distant sites. A neoplasm (tumor) is an abnormal mass of tissue whose growth exceeds and is uncoordinated with that of the normal surrounding tissues, and which persists in the same excessive manner after the evoking stimulus is withdrawn. Neoplasms are classified as benign (localized, well-differentiated, non-invasive) or malignant (invasive, capable of metastasis, often poorly differentiated). The defining criterion of malignancy is metastasis — the establishment of secondary growths at anatomical sites discontinuous with the primary tumor. The treatment below follows Weinberg, The Biology of Cancer (3rd ed., 2019), and Kumar et al., Robbins Basic Pathology (10th ed., 2021), Ch. 5 ^{[Ch. 5 Neoplasia]}.

The hallmarks of cancer

Douglas Hanahan and Robert Weinberg proposed in Cell (2000) that the bewildering diversity of cancer could be organized under six hallmarks — acquired capabilities shared by virtually all cancers. In 2011 they added two emerging hallmarks (deregulating cellular energetics; avoiding immune destruction) and two enabling characteristics (genome instability and mutation; tumor-promoting inflammation) ^{[Cell 144 (2011) 646-674]}. The framework is a organizing schema, not a literal checklist: individual cancers acquire the hallmarks in different orders and through different molecular routes (see 19.08.*, macroevolution, cancer as an evolutionary process).

Sustaining proliferative signaling. Normal cells require external growth-factor signals (EGF, PDGF) to divide. Cancer cells produce their own growth factors (autocrine signaling) or carry mutations that constitutively activate downstream signaling (RAS, PI3K), rendering the cell independent of external instruction (see 17.07.*, signaling).

Evading growth suppressors. Growth-inhibitory signals — TGF-β, contact inhibition, the CDKN2A / p16 checkpoint — normally halt division. Cancer cells disable these pathways, most commonly through loss of RB1 (which releases E2F transcription factors and permits S-phase entry) or loss of TP53 (which eliminates the G1/S damage checkpoint).

Resisting cell death. Programmed cell death (apoptosis) eliminates damaged or superfluous cells. Cancers evade apoptosis by upregulating anti-apoptotic proteins (BCL-2, as in follicular lymphoma) or by losing TP53, which would otherwise trigger apoptosis after DNA damage (see 17.08.04, apoptosis).

Enabling replicative immortality. Normal somatic cells undergo a finite number of divisions — the Hayflick limit — because telomeres shorten with each division. Cancer cells reactivate telomerase (TERT), rebuilding telomeres and conferring unlimited replicative potential.

Inducing angiogenesis. A tumor cannot grow beyond the oxygen diffusion limit (roughly 1–2 mm) without its own blood supply. Cancers secrete vascular endothelial growth factor (VEGF) to recruit vessels, a process targeted by the antibody bevacizumab (Avastin).

Activating invasion and metastasis. Epithelial cancers activate developmental programs — the epithelial-mesenchymal transition (EMT) — that dissolve cell-cell junctions (loss of E-cadherin) and confer motility, enabling invasion through the basement membrane and entry into the circulation (see 17.02.*, membrane proteins, cell adhesion).

Deregulating cellular energetics (Warburg effect). Cancer cells preferentially metabolize glucose to lactate even in the presence of oxygen — Otto Warburg's 1924 observation. The glycolytic shift supports rapid biosynthesis of nucleotides, lipids, and amino acids needed for proliferation (see 17.04.*, metabolism).

Avoiding immune destruction. Tumors that succeed in the immunocompetent host have evolved mechanisms to evade T-cell attack — most prominently by expressing PD-L1, which engages the PD-1 receptor on T cells and delivers an inhibitory signal that disables the T cell (see 17.10.*, immunology, tumor immunology).

Oncogenes

An oncogene is a mutated, amplified, or translocated version of a normal cellular gene — a proto-oncogene — whose product promotes cell growth, survival, or proliferation. Oncogene activation is a gain-of-function event: a single altered allele suffices to drive the phenotype, so oncogenes act dominantly at the cellular level.

RAS (KRAS, NRAS, HRAS) encodes a small GTPase that transmits growth signals from receptor tyrosine kinases to the MAPK cascade (RAS → RAF → MEK → ERK). RAS is the most commonly mutated oncogene family in human cancer — roughly 30 percent of all tumors carry a RAS mutation, with KRAS prevailing in pancreatic (over 90 percent) and colorectal cancer, and NRAS in melanoma (see 17.07.*, signaling, MAPK). The oncogenic mutations (G12, G13, Q61) cripple GTPase activity, locking RAS in its GTP-bound, active conformation so the growth signal never switches off.

MYC is a transcription factor that activates the expression of hundreds of proliferation genes. In Burkitt lymphoma, a chromosomal translocation — t(8;14) — places MYC under the immunoglobulin heavy-chain enhancer, driving constitutive overexpression (see 17.06.*, molecular genetics, translocations). MYC is overexpressed or amplified in a wide range of carcinomas and hematologic malignancies.

HER2 (ERBB2) is a receptor tyrosine kinase amplified in approximately 20 percent of breast cancers. Trastuzumab (Herceptin), a monoclonal antibody against the HER2 extracellular domain, was one of the first molecularly targeted cancer therapies and substantially improves survival in HER2-positive disease.

BCR-ABL arises from the Philadelphia chromosome — the reciprocal translocation t(9;22) — in chronic myeloid leukemia (CML). The fusion gene encodes a constitutively active tyrosine kinase. Imatinib (Gleevec), a small-molecule ABL kinase inhibitor, transformed CML from a near-fatal disease into a manageable chronic condition and became the paradigm of rational targeted therapy (see 35.07.*, pharmacology).

Tumor suppressor genes

A tumor suppressor gene encodes a product that restrains cell division, promotes DNA repair, or triggers apoptosis. Tumor suppressor inactivation is a loss-of-function event: both alleles must be disrupted, so tumor suppressors act recessively at the cellular level. Alfred Knudson's two-hit hypothesis (1971), formulated from retinoblastoma epidemiology, states that both copies must be lost. In familial cancer syndromes the first hit is inherited (germline) and the second is somatic; in sporadic cancers both hits are somatic.

TP53 (encoding p53, "the guardian of the genome") is the most frequently mutated gene in human cancer — altered in approximately 50 percent of all tumors. In response to DNA damage, oncogene activation, or telomere shortening, p53 activates cell-cycle arrest (via the CDKN1A / p21 checkpoint), DNA repair genes, and — if repair fails — apoptosis (see 17.06.02, DNA repair; 17.08.04, apoptosis). Most TP53 mutations are missense substitutions in the DNA-binding domain that abolish sequence-specific transactivation.

RB1 (the retinoblastoma susceptibility protein) governs the G1/S transition. In its hypophosphorylated state, RB1 sequesters E2F transcription factors, preventing entry into S phase. Cyclin-CDK complexes phosphorylate RB1, releasing E2F and permitting DNA replication. Loss of RB1 removes this brake. Germline RB1 mutation causes hereditary retinoblastoma, the cancer from which Knudson derived the two-hit model.

APC (adenomatous polyposis coli) is a negative regulator of the Wnt signaling pathway: it forms a destruction complex that targets β-catenin for proteasomal degradation. Loss of APC causes β-catenin accumulation, nuclear translocation, and constitutive activation of Wnt target genes (including MYC and CCND1). APC loss is the initiating event in over 80 percent of sporadic colorectal cancers. Germline APC mutation causes familial adenomatous polyposis (FAP), in which hundreds to thousands of colonic polyps develop, each a potential carcinoma (see 17.07.*, signaling, Wnt).

BRCA1 and BRCA2 are required for homologous recombination repair of double-strand DNA breaks. Women carrying germline BRCA1 mutations have a 55–65 percent lifetime breast cancer risk and a substantially elevated ovarian cancer risk. PARP inhibitors exploit synthetic lethality: BRCA-deficient tumor cells cannot repair double-strand breaks by homologous recombination and depend on PARP-mediated single-strand break repair; blocking PARP selectively kills the tumor cells while sparing normal cells that retain one functional BRCA allele (see 17.06.02, DNA repair).

PTEN opposes the PI3K-Akt-mTOR pathway by dephosphorylating PIP3. Loss of PTEN activates Akt signaling, driving cell survival, growth, and proliferation. PTEN is deleted or silenced in a wide range of cancers, including glioblastoma, prostate, and endometrial cancer (see 17.07.*, signaling, PI3K-Akt-mTOR).

Multistep carcinogenesis

Cancer develops through the sequential accumulation of mutations over years to decades — a process termed multistep carcinogenesis. Peter Nowell proposed in 1976 that tumors evolve by clonal evolution: an initial mutation gives one cell a growth advantage; its descendants expand into a clone; within that clone a second mutation confers a further advantage; and so on, with natural selection favoring the most aggressive variants (see 19.03.*, selection; 35.02.04, epidemiology, the multistage model of Doll and Hill).

The canonical molecular dissection is Bert Vogelstein's analysis of colorectal cancer. The adenoma-carcinoma sequence proceeds through defined genetic steps: loss of APC (initiating adenoma formation) → mutation of KRAS (adenoma growth) → loss of TP53 and additional hits (carcinoma) → further mutations (metastasis). Each step confers a selective growth advantage and is followed by clonal expansion of the mutant lineage ^{[Science 339 (2013) 1546-1558]}.

The accumulation of multiple mutations over decades explains why cancer incidence rises steeply with age. The quantitative form of this relationship — the Armitage-Doll multistage model — is the subject of the key derivation below.

Tumor microenvironment

A tumor is not a homogeneous mass of cancer cells but an ecosystem of malignant cells embedded in stromal tissue. Cancer-associated fibroblasts (CAFs) remodel the extracellular matrix and secrete growth factors (HGF, TGF-β) that promote invasion and drug resistance. Tumor-associated macrophages (TAMs) can be reprogrammed by tumor-derived signals into a phenotype that promotes angiogenesis, suppresses cytotoxic T cells, and facilitates metastasis (see 17.10.*, immunology). The microenvironment is not merely a scaffold — it is an active participant in every hallmark.

Metastasis

Metastasis — the spread of cancer to anatomically discontinuous sites — accounts for the majority of cancer deaths. The metastatic cascade requires that cancer cells (1) invade through the basement membrane and into surrounding stroma, (2) intravasate into blood or lymphatic vessels, (3) survive circulatory shear stress and anoikis and evade immune attack, (4) arrest and extravasate at a distant site, and (5) colonize — proliferate within the foreign microenvironment. Each step is inefficient: most disseminated tumor cells die or remain dormant; only a small fraction establish clinically significant metastases.

Stephen Paget's 1889 "seed and soil" hypothesis proposed that metastasis is not random. The "seed" (the cancer cell) grows only in hospitable "soil" (the target organ), explaining the non-random organ tropism of metastasis — breast cancer to bone, liver, lung, and brain; colorectal cancer to liver; lung cancer to brain and bone. The molecular basis includes organ-specific adhesion molecules, growth factors, and the pre-metastatic niche (see 18.11.*, organogenesis).

Key derivation: the Armitage-Doll multistage model Intermediate+

One of the foundational quantitative results in cancer epidemiology relates the age-specific incidence of cancer to the number of mutational events required for malignant transformation. Peter Armitage and Richard Doll derived this relationship in 1954, and its prediction — that incidence rises as a power of age — remains one of the most robust empirical regularities in oncology.

The model

Assume that a single cell must accumulate $k$ sequential mutations (in specific driver genes, in any order) to become malignant, and that each mutation occurs independently at rate $μ$ per cell per unit time. The waiting time for any single mutation is exponentially distributed with rate $μ$ . The total waiting time to the $k$ -th mutation is the convolution of $k$ independent exponential( $μ$ ) random variables — an Erlang distribution with shape $k$ and rate $μ$ .

Derivation of the incidence power law

The probability that a cell has accumulated exactly $k$ mutations by time $t$ follows a Poisson form:

P (k mutations by t) = \frac{( μ t ) ^{k}}{k !} e^{- μ t} .

For the biologically relevant regime — $μ t ≪ 1$ , since the per-gene, per-division mutation rate is on the order of $1 0^{- 7}$ — the exponential term approaches 1, and the probability that a cell has become malignant by time $t$ is approximately:

P (cancer by t) \approx \frac{( μ t ) ^{k}}{k !} .

The age-specific incidence (hazard rate) is the derivative with respect to $t$ :

h (t) = \frac{d P}{d t} \approx \frac{k μ ^{k} t ^{k - 1}}{k !} = \frac{μ ^{k}}{( k - 1 )!} t^{k - 1} .

Collecting all constants independent of age into a single term $C$ , the central result is:

h (t) \propto t^{k - 1} .

Taking logarithms of both sides yields:

lo g h (t) = (k - 1) lo g t + lo g C .

A log-log plot of incidence against age should therefore give a straight line with slope $k - 1$ .

Comparison with epidemiological data

For many adult solid tumors — colorectal, gastric, pancreatic, and lung cancer — the age-specific incidence rate between ages 30 and 80 rises approximately as $t^{5}$ or $t^{6}$ . The Armitage-Doll model interprets this slope as evidence that five or six rate-limiting mutational stages separate a normal epithelial cell from a clinical cancer. This prediction is consistent with cancer-genome sequencing: the median number of non-synonymous mutations in a typical adult solid tumor ranges from 30 to 200, but the number of driver mutations — those that confer a selective growth advantage — is typically 3 to 8 (Vogelstein et al., 2013) ^{[Science 339 (2013) 1546-1558]}.

The model has known limitations. It assumes a constant mutation rate (ignoring genomic instability, which accelerates the rate after loss of TP53 or mismatch-repair genes), a single cell lineage (ignoring clonal competition and branched evolution), and sequential or partially ordered mutations. Richard Peto's 1977 refinement showed that the exponent depends on whether intermediate clones expand — but the qualitative prediction holds: multistep accumulation of driver mutations produces a steep power-law rise in incidence with age.

The nontrivial content of the result is that the steep age-incidence curve is not merely a correlate of senescence — it is the quantitative signature of a multistep process. If cancer required only one mutation, incidence would be roughly constant with age; the $t^{k - 1}$ scaling directly measures the number of required stages.

Exercises Intermediate+

Exercise 1 (easy, multiple choice).

Oncogenes are activated by which type of mutation, and what is their pattern of dominance at the cellular level?

A. Loss-of-function; recessive (both alleles required)

B. Gain-of-function; dominant (one allele sufficient)

C. Loss-of-function; dominant (one allele sufficient)

D. Gain-of-function; recessive (both alleles required)

Hint

Consider the gas-pedal analogy: a growth-promoting gene that becomes hyperactive needs only one mutated copy to send a constitutive signal, because the normal copy cannot counterbalance the excess.

Answer

Option B. Oncogenes are activated by gain-of-function mutations (or amplification, or activating translocation) and act dominantly at the cellular level: one mutated allele produces enough hyperactive protein to drive proliferation. Tumor suppressors, by contrast, require loss-of-function in both alleles (recessive), because the remaining normal allele is sufficient to restrain growth.

Exercise 2 (easy, multiple choice).

Knudson's two-hit hypothesis was formulated from epidemiological study of which cancer?

A. Lung cancer

B. Breast cancer

C. Retinoblastoma

D. Colorectal cancer

Hint

Knudson compared familial (hereditary) cases — early onset, bilateral tumors — with sporadic cases (later onset, unilateral) and inferred the number of mutational events each required.

Answer

Option C. Alfred Knudson studied the epidemiology of retinoblastoma (a childhood eye cancer). Hereditary cases presented early and bilaterally (one inherited germline hit, plus one somatic hit in each eye), while sporadic cases presented later and unilaterally (two somatic hits in the same cell). The pattern required two mutational events — the two-hit hypothesis — later confirmed by the identification of the RB1 gene on chromosome 13.

Exercise 3 (medium, short answer).

KRAS mutations are found in over 90 percent of pancreatic ductal adenocarcinomas. Explain why the RAS protein was considered "undruggable" for three decades, and describe the mechanism by which the recent covalent KRAS-G12C inhibitors (sotorasib) overcome this obstacle.

Hint

Consider two obstacles: (1) the protein's surface topology (does it have a pocket for a drug to bind?), and (2) the biochemistry of substrate binding (how tightly does RAS bind GTP, and what is the cellular GTP concentration?). The G12C mutation creates a chemical handle that was previously absent.

Answer

KRAS was considered undruggable for two reasons. First, the RAS protein surface is unusually smooth, lacking the deep hydrophobic pockets that small-molecule inhibitors typically occupy. Second, RAS binds GTP with picomolar affinity, and the cellular concentration of GTP is in the millimolar range, so competitive inhibition at the nucleotide-binding site is biochemically infeasible — the drug cannot outcompete the natural ligand.

The G12C mutation substitutes a cysteine for glycine at position 12, creating a nucleophilic sulfhydryl group that is absent in wild-type KRAS. Sotorasib (and adagrasib) are designed with an electrophilic warhead that forms a covalent bond with this cysteine, irreversibly locking the mutant protein in its inactive (GDP-bound) conformation. Because the cysteine is present only in the mutant (G12C) protein, the drug spares wild-type KRAS. This covalent, allele-specific strategy was the first successful direct pharmacological inhibition of RAS after decades of failure.

Exercise 5 (hard, short answer).

Describe the Vogelstein adenoma-carcinoma sequence in colorectal cancer. Why does loss of APC typically precede the KRAS mutation, and what does this order reveal about the selective advantage of each step?

Hint

Consider what each mutation does in isolation versus what it does in a cell that already carries the previous mutation. The selective value of a mutation depends on the cellular context established by prior mutations.

Answer

The adenoma-carcinoma sequence proceeds: normal epithelium → APC loss (early adenoma) → KRAS mutation (intermediate / late adenoma) → TP53 loss and additional hits (carcinoma) → further mutations (metastasis).

APC loss initiates the sequence because it activates Wnt / β-catenin signaling, which drives crypt stem-cell self-renewal and disrupts differentiation. The cell gains a self-renewal advantage and forms a dysplastic crypt focus, then a small adenoma — but this alone is a slow-growing, non-invasive lesion. Within the expanding APC-null clone, a KRAS mutation (constitutive MAPK signaling) provides an additional proliferative drive, converting a small adenoma into a larger, faster-growing one. Loss of TP53 then removes the apoptosis and checkpoint response to the accumulating genomic damage, permitting invasive carcinoma.

The order is not arbitrary. Each mutation's selective advantage depends on the cellular context established by the preceding mutations. A KRAS mutation in normal, APC-intact epithelium confers little clonal advantage because Wnt-driven stem-cell maintenance is still regulated. APC loss creates the expanded population within which a KRAS mutation is then strongly selected. This context-dependence is the molecular fingerprint of stepwise clonal evolution — the same Darwinian logic Nowell described in 1976.

Exercise 6 (hard, numeric).

A woman carries a germline BRCA1 mutation, which impairs homologous recombination repair and thereby raises the effective somatic mutation rate by a factor of 10 relative to a BRCA-proficient cell. Using the Armitage-Doll incidence formula $h (t) \propto μ^{k} t^{k - 1}$ with $k = 5$ rate-limiting stages, compute the factor by which the elevated mutation rate changes the cancer incidence at a fixed age. Explain why the result is so large, and identify one simplification in the model that may overstate the effect.

Hint

At fixed $t$ , incidence scales as $μ^{k}$ . Form the ratio $h_{new} / h_{old} = (μ_{new} / μ_{old})^{k}$ with $μ_{new} / μ_{old} = 10$ and $k = 5$ .

Answer

At fixed age $t$ , the incidence scales as $μ^{k}$ . With a 10-fold increase in mutation rate and $k = 5$ rate-limiting stages:

\frac{h _{new}}{h _{old}} = (\frac{μ _{new}}{μ _{old}})^{k} = 1 0^{5} = 100, 000.

The amplification is enormous because the mutation rate enters the incidence formula raised to the $k$ -th power: each of the five required mutations occurs ten times more often, and the effects compound multiplicatively. A modest per-division increase in mutation rate translates into a vast reduction in the expected waiting time for all five driver events to accumulate.

One simplification: the model assumes the mutation rate remains constant throughout carcinogenesis, but in reality the elevated mutation rate also increases the passenger-mutation burden, and cells with lethal mutations are removed by apoptosis. The power-law regime ( $μ t ≪ 1$ ) breaks down at very high mutation rates, at which point the exponential term $e^{- μ t}$ in the exact Poisson formula can no longer be neglected. The qualitative conclusion — that DNA repair deficiency accelerates carcinogenesis dramatically — is nonetheless correct, and the BRCA1 lifetime risk (55–65 percent for breast cancer) reflects this amplification.

Exercise 7 (hard, short answer).

Explain the concept of synthetic lethality as exploited by PARP inhibitors in BRCA-deficient cancers. Why are normal (BRCA-proficient) cells spared, and what does this principle reveal about the design of cancer therapies that target tumor-specific vulnerabilities?

Hint

Consider the two DNA repair pathways — base excision repair (BER, PARP-dependent) and homologous recombination (HR, BRCA-dependent). Each alone is dispensable; what happens when both are lost? Why does the tumor cell lose both while the normal cell retains HR?

Answer

Synthetic lethality occurs when two genetic or pharmacological defects are individually tolerable but jointly lethal. PARP inhibitors exploit this in BRCA-deficient cancers as follows. PARP enzymes repair single-strand DNA breaks via base-excision repair. Inhibiting PARP causes single-strand breaks to persist; at replication forks these convert to double-strand breaks. Normal cells repair double-strand breaks by homologous recombination (HR), which requires functional BRCA1 and BRCA2. BRCA-deficient tumor cells have lost HR capacity, so the double-strand breaks produced by PARP inhibition cannot be repaired: they accumulate, trigger chromosomal catastrophe, and kill the cell.

Normal cells are spared because they retain at least one functional BRCA allele and therefore retain HR capacity. The PARP inhibitor creates double-strand breaks in all dividing cells, but only the HR-deficient tumor cells are unable to survive them.

The principle extends beyond PARP: any tumor with a specific DNA repair deficiency can, in principle, be killed by inhibiting a complementary pathway. The broader lesson is that effective targeted therapies need not target the driver mutation itself — they can target a dependency created by the driver. The cancer-specific vulnerability is the loss of redundancy, and the therapy exploits the tumor's narrowed survival margin.

Advanced results Master

Cancer genomics and driver mutations

The Cancer Genome Atlas (TCGA), a multi-institutional project that molecularly profiled 33 cancer types, and the COSMIC (Catalogue of Somatic Mutations in Cancer) database have collectively mapped the genomic landscape of human cancer (see 33.08.*, big science). The central finding, articulated by Vogelstein and colleagues in 2013, is that despite the large total number of mutations per tumor — a median of 30 to 200 non-synonymous mutations in adult solid tumors — only a small subset are driver mutations: alterations that confer a selective growth advantage. Most mutations are passengers: selectively neutral consequences of the increased mutation rate and the large number of cell divisions. Across all cancer types, roughly 140 driver genes have been identified, and a typical tumor carries 2 to 8 driver mutations. These 140 genes cluster into a dozen or so core signaling pathways — RAS-MAPK, PI3K-Akt-mTOR, Wnt, p53, cell-cycle control, apoptosis, DNA repair, and chromatin regulation ^{[Science 339 (2013) 1546-1558]} (see 20.08.03, causation, the driver as causal).

Mutational signatures, catalogued by Ludmil Alexandrov and colleagues, identify the mutational processes active in a tumor from the pattern of base substitutions in context (e.g., C>A transversions at specific trinucleotides). Tobacco smoking produces a characteristic signature; ultraviolet light produces another (C>T transitions at dipyrimidine sites, the signature of UV-induced pyrimidine dimers); defective DNA mismatch repair produces yet another (the signature of microsatellite instability). These signatures connect a tumor's mutation catalog to the exposures and repair defects that generated it (see 19.01., genetics, mutation; 17.06., molecular genetics).

Intratumor heterogeneity — the coexistence of genetically distinct subclones within a single tumor — is revealed by multi-region sequencing and single-cell genomics. Tumors do not evolve linearly; they branch, with different regions of the same tumor carrying different combinations of mutations. This branched evolution has direct clinical consequences: a biopsy samples only one region, and a drug targeted to the dominant clone may miss minor resistant subclones that then expand. Phylogenetic reconstruction of tumor evolution uses the same methods applied to organismal phylogenetics, treating each biopsy region as a "species" (see 19.07.*, phylogenetics, cancer phylogeny).

Immunotherapy: checkpoint blockade and cellular therapy

The 2018 Nobel Prize in Physiology or Medicine, awarded to James Allison and Tasuku Honjo, recognized a mechanistic insight that transformed oncology: the immune system is fully capable of recognizing and killing cancer cells, but tumors erect molecular checkpoints that shut T cells down. Blocking those checkpoints releases the brakes.

CTLA-4 (cytotoxic T-lymphocyte-associated protein 4) is an inhibitory receptor upregulated on T cells after activation; it competes with the co-stimulatory receptor CD28 for B7 ligands on antigen-presenting cells, thereby attenuating the T-cell response. Ipilimumab, an anti-CTLA-4 antibody developed by Allison, was the first checkpoint inhibitor to extend survival in metastatic melanoma. PD-1 (programmed cell death protein 1) is a second inhibitory receptor, expressed on exhausted T cells in the tumor microenvironment; its ligand PD-L1 is expressed by many tumors. Anti-PD-1 antibodies (nivolumab, pembrolizumab) and anti-PD-L1 antibodies (atezolizumab) have produced durable responses in melanoma, lung cancer, renal cell carcinoma, bladder cancer, Hodgkin lymphoma, and a growing list of malignancies defined not by tissue of origin but by molecular features (microsatellite instability, tumor mutational burden) (see 17.10.*, immunology, T-cell activation and checkpoints; 29.10.03, biological treatments).

CAR-T cell therapy (chimeric antigen receptor T cells) engineers the patient's own T cells ex vivo: a viral vector inserts a synthetic receptor — an antibody-derived single-chain variable fragment fused to T-cell signaling domains — that redirects the T cell against a tumor antigen. Anti-CD19 CAR-T cells (tisagenlecleucel, axicabtagene ciloleucel) produce high complete-response rates in relapsed B-cell acute lymphoblastic leukemia and diffuse large B-cell lymphoma. The principal toxicities — cytokine release syndrome and immune effector cell-associated neurotoxicity — are direct consequences of the explosive T-cell activation (see 33.06.*, the double helix, genetic engineering).

Cancer vaccines target oncogenic infections: the HPV vaccine (Gardasil) prevents the infections responsible for most cervical cancers, and the hepatitis B vaccine prevents the chronic infection that drives most hepatocellular carcinoma in endemic regions (see 35.06.03, vaccine science). Oncolytic viruses (talimogene laherparepvec, T-VEC, a modified herpes simplex virus) are engineered to replicate selectively in cancer cells and lyse them, releasing tumor antigens that stimulate anti-tumor immunity (see 35.02.03, viral pathogenesis).

Cancer as an evolutionary process

Peter Nowell's 1976 proposition that tumors evolve by clonal evolution framed cancer as Darwinian selection within the body: genetic variation arises through mutation, and the microenvironment imposes selection, favoring clones with the greatest proliferative, survival, and invasive fitness. This framing unifies cancer biology with evolutionary theory (see 19.03., selection; 19.08., macroevolution).

Drug resistance exemplifies this process. A tumor bearing a sensitizing mutation (e.g., EGFR exon 19 deletion in lung cancer) is treated with an EGFR inhibitor (e.g., osimertinib). The drug kills the sensitive population, but pre-existing resistant subclones — carrying the T790M or C797S mutations — survive and expand. This is the same evolutionary dynamic that drives antibiotic resistance in bacteria, operating on somatic rather than germline timescales (see 35.02.02, bacterial pathogenesis, resistance evolution).

Mathematical oncology applies evolutionary and ecological modeling to treatment design. Robert Gatenby's adaptive therapy paradigm proposes a counterintuitive strategy: rather than administering the maximum tolerated dose to kill as many tumor cells as possible — which strongly selects for resistance — maintain a stable population of drug-sensitive cells that competitively suppress resistant clones. By treating only enough to control tumor burden, the sensitive majority holds the resistant minority in check through resource competition. Early clinical trials in metastatic prostate cancer have shown that adaptive docetaxel scheduling can prolong time to progression, lending empirical support to the ecological model (see 20.05.*, philosophy of biology, evolution and medicine).

Branched evolution and the tumor's internal phylogenetic structure mean that a single biopsy is a census of one branch, not the whole tree. Multi-region sampling and circulating tumor DNA (ctDNA) analysis — the "liquid biopsy" — reveal subclonal diversity that a tissue biopsy misses, and longitudinal ctDNA monitoring can detect resistant subclones weeks or months before radiographic progression.

Cancer stem cells and developmental signaling

The cancer stem cell hypothesis proposes that within a heterogeneous tumor, a subpopulation of cells with stem-like properties — self-renewal and the capacity to differentiate into the non-stem cells that compose the tumor bulk — drives growth, metastasis, and recurrence after therapy. Whether this hierarchy is rigid (a fixed stem-cell compartment) or stochastic (any cell can enter a stem-like state under the right signals) remains debated; the two models are not mutually exclusive. Epithelial-mesenchymal transition and its reversal generate dynamic stem-like states through transcription factors (SNAIL, TWIST, ZEB) that also drive invasion and metastasis (see 18.09.*, fertilization and early development).

Developmental signaling pathways — Wnt (APC / β-catenin), Notch, and Hedgehog — are frequently reactivated in cancer. These are the same pathways that pattern the embryo; their inappropriate reactivation in adult tissue drives self-renewal and blocks differentiation. Hedgehog pathway inhibitors (vismodegib) are used in basal cell carcinoma, where ligand-independent Hedgehog signaling — usually from PTCH1 loss — is the near-universal driver (see 18.11.*, organogenesis).

Cancer prevention and screening

Tobacco remains the single largest preventable cause of cancer, responsible for roughly 20 percent of all cancer deaths worldwide and the overwhelming majority of lung cancer. The Doll and Hill studies of British doctors (1954) established the causal link between smoking and lung cancer, and subsequent tobacco-control policies — taxation, advertising bans, smoke-free legislation — have driven the sharpest reductions in cancer mortality achieved by any public-health intervention (see 32.20.*, the twentieth century, the smoking epidemic; 35.02.04, epidemiology, Doll and Hill).

Infection-attributable cancers account for roughly 15 percent of cancer worldwide. Human papillomavirus (HPV) causes cervical and oropharyngeal cancer; hepatitis B and C viruses cause hepatocellular carcinoma; Helicobacter pylori causes gastric adenocarcinoma and MALT lymphoma; Epstein-Barr virus causes Burkitt and Hodgkin lymphoma and nasopharyngeal carcinoma. Vaccination against HPV and hepatitis B is, in effect, primary cancer prevention (see 35.02.02, bacterial pathogenesis, H. pylori; 35.02.03, viral pathogenesis).

Screening — mammography for breast cancer, colonoscopy for colorectal cancer, the Pap smear and HPV testing for cervical cancer — reduces cancer mortality by detecting premalignant or early-stage lesions. But screening introduces statistical pitfalls that must be interpreted carefully. Lead-time bias inflates apparent survival: detecting a cancer earlier makes patients live longer with the diagnosis without changing the time of death. Length bias over-represents slow-growing, more favorable tumors, because screen-detectable lesions are, by construction, the ones that spent the longest time in the detectable preclinical phase. Overdiagnosis detects indolent lesions that would never have become clinically significant — a real harm, because every detection triggers intervention (see 29.01.03, statistical reasoning; 20.08.*, philosophy of science, values in screening).

The future: precision oncology

Precision oncology matches treatment to the molecular features of the individual tumor rather than its tissue of origin. Tumor genomic profiling identifies actionable mutations — EGFR inhibitors for EGFR-mutant lung cancer, BRAF inhibitors for BRAF V600E melanoma, HER2 antibodies for HER2-amplified breast and gastric cancer, TRK inhibitors for NTRK-fusion cancers irrespective of histology. The basket trial — enrolling patients by mutation rather than by cancer type — is the corresponding experimental design (see 35.08.03, precision medicine and AI).

Liquid biopsies analyze circulating tumor DNA (ctDNA) in a blood sample, detecting tumor-specific mutations, quantifying minimal residual disease after surgery, and monitoring response and resistance noninvasively. ctDNA can rise months before imaging shows progression, offering an early window for changing therapy (see 33.06.*, sequencing technology; 35.08.02, genomic medicine).

Artificial intelligence is being applied across oncology: deep-learning models read mammograms and detect breast cancer with accuracy comparable to specialist radiologists; computational pathology models grade prostate and breast cancer from whole-slide images; AlphaFold-derived protein structures guide drug design; and large language models are being tested for clinical decision support. Algorithmic bias is a central concern: models trained on datasets that under-represent certain populations may perform worse on those populations, and unchecked deployment can deepen rather than narrow disparities (see 33.07.*, computing, deep learning; 20.02.06, AI ethics, algorithmic bias in healthcare).

CRISPR-based gene editing enables precise manipulation of cancer-relevant genes in the laboratory and is entering clinical trials for cancer cell therapy — editing T cells to remove the PD-1 checkpoint or insert a CAR. Personalized neoantigen vaccines, designed from a tumor's unique mutational catalogue and delivered as mRNA, aim to prime T cells against patient-specific tumor antigens, extending the vaccine concept from prevention to therapy (see 35.08.02, genomic medicine, gene therapy; 33.06.*, mRNA).

Connections Master

Metabolism, the Warburg effect, and cancer energetics

The Warburg effect — aerobic glycolysis in cancer cells — connects cancer biology to the regulation of cellular metabolism. The same metabolic reprogramming appears in rapidly proliferating normal cells (activated lymphocytes, embryonic tissue), suggesting that the glycolytic shift supports biosynthesis rather than energy production per se. The PI3K-Akt-mTOR pathway, frequently activated in cancer, directly regulates glucose uptake and glycolysis, linking oncogenic signaling to metabolic rewiring (17.04., metabolism; 17.07., signaling, PI3K-Akt-mTOR).

Signaling pathways: the molecular circuitry of the hallmarks

Every hallmark is implemented by a specific signaling pathway, and the same pathways recur across cancer types. RAS-MAPK drives proliferative signaling; PI3K-Akt-mTOR drives survival and growth; Wnt / β-catenin drives self-renewal; JAK-STAT drives cytokine responses; NF-κB drives inflammation and survival. The hallmarks framework and the signaling-pathway map are two descriptions of the same underlying circuitry, one phenotypic and one molecular (17.07., signaling; 17.06., molecular genetics).

Molecular genetics: mutation, repair, and chromosomal instability

The cancer genome is the product of mutagenesis (exposures, replication errors) and DNA repair (mismatch repair, double-strand break repair, nucleotide excision repair). Defects in repair — BRCA1/2 (homologous recombination), MLH1/MSH2 (mismatch repair, causing microsatellite instability), xeroderma pigmentosum genes (nucleotide excision repair, sensitizing to UV) — accelerate the accumulation of driver mutations and define clinically actionable tumor subclasses (17.06.02, DNA repair; 17.06.*, molecular genetics).

Immunology and tumor immune surveillance

The cancer-immunity cycle — release of tumor antigens, presentation by dendritic cells, priming and activation of T cells, trafficking to the tumor, infiltration, recognition, and killing of cancer cells — can fail at any step. Checkpoint blockade releases the activation step; CAR-T cells bypass the antigen-presentation step; oncolytic viruses stimulate antigen release and infiltration. The same immune biology that governs response to infection (see 17.10., immunology) governs response to immunotherapy, and the toxicities of checkpoint blockade (colitis, pneumonitis, thyroiditis) are autoimmune phenomena produced by the same T-cell activation that attacks the tumor (17.10., immunology, tumor immunology; 35.02.01, infectious disease and immunity).

Evolution, phylogenetics, and clonal dynamics

Cancer is a somatic evolutionary process. The methods of phylogenetics — reconstructing trees from mutation patterns, inferring common ancestors, estimating branch lengths — apply directly to multi-region tumor sequencing data. Drug resistance is a special case of selection under environmental pressure. Mathematical oncology borrows from population genetics and ecology to model and predict tumor dynamics, and adaptive therapy is a clinical translation of ecological competition theory (19.03., selection; 19.07., phylogenetics, cancer phylogeny; 19.08., macroevolution; 20.05., philosophy of biology, evolution and medicine).

Developmental biology: reactivation of embryonic programs

The pathways that pattern the embryo — Wnt, Notch, Hedgehog, BMP, EMT — are the same pathways reactivated in cancer. Teratocarcinomas and embryonal carcinomas blur the line between developmental and oncogenic proliferation. The cancer stem cell concept and the stochastic state-transition model both draw on developmental biology's understanding of cell-fate determination (18.09., fertilization and early development; 18.11., organogenesis; 20.05.03, natural selection and teleology, cancer as de-differentiation).

Pharmacology: targeted therapy and the logic of drug design

Each targeted cancer drug — imatinib for BCR-ABL, trastuzumab for HER2, BRAF inhibitors for V600E, KRAS-G12C inhibitors, PARP inhibitors for BRCA deficiency — targets a specific molecular dependency created by a specific driver mutation. The pharmacology of these agents — pharmacokinetics, resistance mechanisms, combination strategies — is a direct application of the molecular biology of the hallmarks (35.07.*, pharmacology; 29.10.03, biological treatments).

Epidemiology: the causes and distribution of cancer

The methods of chronic-disease epidemiology — prospective cohorts, case-control studies, population attributable risk, Mendelian randomization — established the major causes of cancer (tobacco, infections, obesity, radiation) and quantify how much of the cancer burden is preventable. The same methods that revealed cardiovascular risk in Framingham revealed cancer risk in the Doll-Hill studies and the Nurses' Health Study (35.02.04, epidemiology).

Ethics, society, and the illness experience

BRCA testing raises questions about genetic privacy, insurance discrimination, prophylactic surgery, and reproductive choice (see 20.02., ethics, bioethics). Susan Sontag's Illness as Metaphor (1978) argued that the metaphorical baggage attached to cancer — secrecy, moral taint, retribution — amplifies the suffering of the disease; her critique extends to the language used for HIV / AIDS (see 20.04., aesthetics, illness as metaphor). Occupational and environmental carcinogenesis — asbestos and mesothelioma, vinyl chloride and angiosarcoma, the Bhopal disaster — places cancer within the political economy of industrial harm (30.06., deviance, corporate harm). Breast cancer activism, modeled on AIDS activism, transformed the politics of research funding and clinical-trial design (30.07., social movements, patient advocacy). Medical anthropology frames cancer as "biographical disruption" — an event that reorganizes a person's life narrative and relationships (31.06.02, medical anthropology, illness narratives).

Downstream link to metabolic disease

Cancer and metabolic syndrome share risk factors (obesity, insulin resistance, chronic inflammation) and molecular pathways (PI3K-Akt-mTOR, insulin-like growth factor signaling). Obesity is an established cause of at least thirteen cancer types. Adipose-tissue inflammation, hyperinsulinemia, and altered adipokine signaling provide mechanistic links, motivating the proposed connection to the metabolic-disease unit (35.03.04).

Historical and philosophical context Master

Peyton Rous and the first oncogenic virus (1911)

In 1911, Peyton Rous, a pathologist at the Rockefeller Institute, reported that a cell-free filtrate from a chicken sarcoma could transmit the tumor to healthy chickens. The agent, later identified as the Rous sarcoma virus (RSV), was the first oncogenic virus discovered. The finding was met with skepticism: cancer was not thought to be transmissible, and Rous's result seemed specific to chickens and irrelevant to human disease. Rous was awarded the Nobel Prize in 1966, fifty-five years later, after the viral oncogene src had been characterized and the broader relevance of oncogenes established. Rous's work demonstrated that a single agent could cause cancer — the conceptual foundation on which the oncogene hypothesis was built.

Bishop, Varmus, and the cellular origin of oncogenes (Nobel 1989)

The Rous sarcoma virus carries an oncogene, v-src, that is essential for transformation. In 1976, J. Michael Bishop and Harold Varmus, working at the University of California, San Francisco, demonstrated that a gene closely homologous to v-src exists in the normal cellular DNA of all vertebrates — including chickens, humans, and even fruit flies. The viral oncogene was not a viral invention but a captured and corrupted version of a normal cellular gene, c-src. This finding generalized: every viral oncogene turned out to derive from a normal cellular proto-oncogene. The implication was profound: cancer-causing genes are not foreign agents but corrupted versions of the cell's own growth-regulatory machinery, built into every cell's genome. Bishop and Varmus shared the 1989 Nobel Prize for this discovery.

Knudson's two-hit hypothesis (1971)

Alfred Knudson, working at the Fox Chase Cancer Center, studied the epidemiology of retinoblastoma — a childhood eye cancer that occurs in both hereditary and sporadic forms. Hereditary retinoblastoma presented bilaterally and early; sporadic cases presented unilaterally and later. Knudson modeled the two forms statistically and concluded that hereditary cases required one additional somatic event (because one mutation was inherited), while sporadic cases required two somatic events in the same cell. The two-hit hypothesis, published in 1971, predicted the existence of a gene whose loss — in two steps — caused the tumor. The prediction was confirmed a decade and a half later with the cloning of RB1 on chromosome 13q14. Knudson's insight established the recessive genetics of tumor suppression: a tumor suppressor gene is "invisible" until both copies are lost, which is why inherited loss of one copy creates a predisposition (every cell is one somatic hit away from tumor initiation) while sporadic loss requires two hits in the same cell (hence the later, unilateral presentation).

Otto Warburg and aerobic glycolysis (1924)

Otto Warburg, working in Berlin, observed in 1924 that cancer cells ferment glucose to lactate even when oxygen is abundant — the Warburg effect, or aerobic glycolysis. Warburg believed this metabolic reprogramming was the fundamental cause of cancer (he termed it "the prime cause of cancer"), a hypothesis that proved too narrow: we now know that the metabolic shift is a consequence of oncogenic signaling, not the root cause. But Warburg was right that metabolism is fundamentally altered in cancer. The clinical application — FDG-PET imaging, which visualizes tumors by their avid glucose uptake — relies directly on the Warburg effect. Warburg received the Nobel Prize in 1931 for his work on cellular respiration (though not specifically for the cancer observation).

Hanahan and Weinberg: codifying the hallmarks (2000, 2011)

By the end of the twentieth century, cancer biology had accumulated a vast but disorganized catalog of molecular mechanisms. Douglas Hanahan (UCSF) and Robert Weinberg (MIT, co-author of the first textbook of cancer biology) proposed in Cell in 2000 that the field's accumulated knowledge could be organized under six organizing principles — the hallmarks. The framework was heuristic: it did not claim that every cancer uses every mechanism, only that these capabilities are, in some form, common to all cancers. The 2011 update, "Hallmarks of Cancer: The Next Generation," added reprogramming energy metabolism and evading immune destruction, and emphasized genome instability and inflammation as enabling characteristics ^{[Cell 144 (2011) 646-674]}. The hallmarks framework became the dominant pedagogical and conceptual scaffold of cancer biology, structuring textbooks, grant proposals, and the self-understanding of the field. Its philosophical interest lies in its status as a unifying framework for a phenomenon (cancer) that is genetically and histologically diverse — an organizational schema that makes a complex, heterogeneous disease tractable without denying its heterogeneity (see 20.08.*, philosophy of science, the logic of classification).

The War on Cancer and the imatinib paradigm (1971, 2001)

The U.S. National Cancer Act of 1971, signed by President Nixon, launched the "War on Cancer" with a surge of federal funding and the expectation that a concentrated effort could conquer cancer as polio had been conquered. Half a century of mixed results followed: age-adjusted cancer mortality in the United States has fallen roughly 30 percent since 1991, driven mostly by declines in smoking and improvements in early detection and surgery, with targeted and immune therapies contributing at the margin. Cancer proved not to be a single disease but hundreds, each with its own biology.

The development of imatinib (Gleevec) for chronic myeloid leukemia, approved by the FDA in 2001, became the paradigmatic success of the molecular-targeting strategy that the War on Cancer had sought. Brian Druker, Charles Sawyers, and others designed a small molecule to inhibit the BCR-ABL fusion kinase — the product of the Philadelphia chromosome, the first specific chromosomal abnormality linked to a human cancer (1960). The result was dramatic: CML, previously fatal within five years, became a manageable chronic condition for most patients. Imatinib validated the principle that a cancer defined by a single dominant driver mutation could be controlled by a drug targeting that mutation — but it also revealed the limitation: most cancers are not driven by a single kinase, and resistance mutations (T315I) emerge, requiring second- and third-generation inhibitors.

Allison, Honjo, and checkpoint immunotherapy (Nobel 2018)

James Allison, at UC Berkeley, spent years studying the CTLA-4 receptor, which he recognized as a brake on T-cell activation. Rather than pursuing the conventional approach of stimulating the immune system directly, Allison proposed releasing the brake — blocking CTLA-4 to unleash T cells against tumors. The resulting antibody, ipilimumab (anti-CTLA-4), became the first treatment to extend survival in metastatic melanoma, a disease previously considered essentially untreatable. Tasuku Honjo, at Kyoto University, discovered PD-1 in 1992 and characterized it as a second inhibitory checkpoint. Anti-PD-1 antibodies (nivolumab, pembrolizumab) produced durable responses across a broader range of cancers than anti-CTLA-4 and with less toxicity.

The 2018 Nobel Prize honored the mechanistic insight underlying checkpoint blockade: the immune system can recognize cancer, but tumors suppress the response through inhibitory receptors that evolved (in their physiological role) to prevent autoimmunity and limit inflammation. The therapeutic strategy is not to strengthen the immune attack but to remove the tumor's defense against it. The autoimmune toxicities of checkpoint blockade — colitis, pneumonitis, hepatitis, endocrinopathies — are the direct cost of releasing these brakes, a pharmacological confirmation that the checkpoints exist precisely to prevent T cells from attacking self-tissues.

The Armitage-Doll model and the logic of mathematical epidemiology (1954)

Peter Armitage and Richard Doll, working at the Statistical Research Unit of the Medical Research Council in London, published in 1954 a mathematical model of carcinogenesis that assumed cancer arises from a sequence of $k$ somatic mutations. Their derivation predicted that age-specific incidence should scale as $t^{k - 1}$ , and they showed that this power law fit the incidence data for cancers of the stomach, colon, rectum, and pancreas with slopes of 5 to 7. The model was published before the molecular biology of cancer was understood — before oncogenes were discovered, before the double-helix structure of DNA had even been established (1953). It is a case study in the power of mathematical reasoning to extract structural information (the number of rate-limiting steps) from population-level data (age-incidence curves) decades before the molecular substrate became accessible. The Armitage-Doll model also illustrates the complementarity of mathematical and molecular approaches: the model predicted that cancer requires several mutational events; the molecular biology, forty years later, identified the genes in which those events occur.

Bibliography Master

Hanahan, D. & Weinberg, R. A. (2000). "The Hallmarks of Cancer." Cell, 100(1), 57–70.
Hanahan, D. & Weinberg, R. A. (2011). "Hallmarks of Cancer: The Next Generation." Cell, 144(5), 646–674.
Vogelstein, B., Papadopoulos, N., Velculescu, V. E., Zhou, S., Diaz, L. A. & Kinzler, K. W. (2013). "Cancer Genome Landscapes." Science, 339(6127), 1546–1558.
Armitage, P. & Doll, R. (1954). "The Age Distribution of Cancer and a Multi-stage Theory of Carcinogenesis." British Journal of Cancer, 8(1), 1–12.
Knudson, A. G. (1971). "Mutation and Cancer: Statistical Study of Retinoblastoma." Proceedings of the National Academy of Sciences, 68(4), 820–823.
Nowell, P. C. (1976). "The Clonal Evolution of Tumor Cell Populations." Science, 194(4260), 23–28.
Stehelin, D., Varmus, H. E., Bishop, J. M. & Vogt, P. K. (1976). "DNA Related to the Transforming Gene(s) of Avian Sarcoma Viruses Is Present in Normal Avian DNA." Nature, 260(5547), 170–173.
Tomasetti, C. & Vogelstein, B. (2015). "Variation in Cancer Risk Among Tissues Can Be Explained by the Number of Stem Cell Divisions." Science, 347(6217), 78–81.
Weinberg, R. A. (2019). The Biology of Cancer (3rd ed.). Garland Science.
Kumar, V., Abbas, A. K., Fausto, N. & Aster, J. C. (2021). Robbins Basic Pathology (10th ed.). Elsevier.
Pardoll, D. M. (2012). "The Blockade of Immune Checkpoints in Cancer Immunotherapy." Nature Reviews Cancer, 12(4), 252–264.
Druker, B. J., Talpaz, M., Resta, D. J. et al. (2001). "Efficacy and Safety of a Specific Inhibitor of the BCR-ABL Tyrosine Kinase in Chronic Myeloid Leukemia." New England Journal of Medicine, 344(14), 1031–1037.
Paget, S. (1889). "The Distribution of Secondary Growths in Cancer of the Breast." The Lancet, 133(3421), 571–573.
Warburg, O. (1956). "On the Origin of Cancer Cells." Science, 123(3191), 309–314.