Oncogenes and tumour suppressors

The genes whose mutation drives cancer fall into two complementary categories. Oncogenes are activated forms of normal cellular genes (proto-oncogenes) and behave as dominant gain-of-function alleles at the cellular level: a single activating event suffices. Tumour suppressors restrain proliferation, maintain genome integrity, or promote apoptosis; they are typically recessive, requiring biallelic inactivation per Knudson’s two-hit model. This page is the mechanism-and-examples reference.

| 11 min read

Short version. Oncogenes are gain-of-function and dominant: one allele is enough. Tumour suppressors are loss-of-function and (with the haploinsufficient exceptions noted below) recessive at the cellular level; biallelic inactivation, per Knudson 1971, is the rule. The mechanism of activation or inactivation maps to the gene class and to the kind of evidence that supports it: missense hotspots and amplification for oncogenes, frameshift / nonsense / splice variants plus LOH for tumour suppressors. Vogelstein et al. 2013 consolidated these into the cancer-genome landscape; Kinzler & Vogelstein 1997 distinguished gatekeeper from caretaker functions.

Oncogenes: gain-of-function, dominant

An oncogene is the activated form of a proto-oncogene — a normal cellular gene encoding a component of a growth-signalling pathway, a transcription factor, a survival regulator, or a chromatin remodelling factor. The discovery that the transforming sequences in acutely transforming retroviruses (v-src, v-myc, v-ras, v-erbB) derive from cellular homologues, recounted in Bishop 1991 (Cell 64:235), established the proto-oncogene as the unit of analysis: cancer arises when the cellular gene is dysregulated, not when a foreign sequence is introduced.

Four mechanistic routes account for the bulk of oncogene activation:

  • Activating point mutation. A missense substitution at a recurrent residue locks the protein in an active conformation, blocks an inactivating modification, or destabilises the off-state. RAS-family GTPases are the archetype: KRAS codon 12 (G12C, G12D, G12V), codon 13, and codon 61 substitutions impair intrinsic and GAP-stimulated GTP hydrolysis; NRAS and HRAS show the same hotspot pattern in different tumour spectra. BRAF V600E in cutaneous melanoma (and in a fraction of colorectal, papillary thyroid, and hairy-cell leukaemia cases) constitutively activates the kinase; EGFR L858R and exon-19 deletions activate the receptor in lung adenocarcinoma; PIK3CA H1047R, E545K, and E542K are common in breast, endometrial, and colorectal cancer.
  • Gene amplification. An increase in copy number scales protein expression beyond the threshold for malignant signalling. MYC amplification in neuroblastoma (MYCN) and small-cell lung cancer is the textbook example; ERBB2 / HER2 amplification in 15–20% of breast cancers (and a fraction of gastric, oesophageal, and salivary cancers) defines a distinct molecular subtype with therapeutic implications via trastuzumab; EGFR amplification in glioblastoma; CCND1 in mantle cell lymphoma. Amplification is often visualised as homogeneously staining regions or double minutes on a karyogram — relevant to the karyogram viewer page on this site.
  • Chromosomal translocation. A reciprocal translocation creates a fusion gene whose product is constitutively active or whose normal regulation is bypassed by an alien promoter. The BCR-ABL1 fusion arising from t(9;22)(q34;q11) — the Philadelphia chromosome — was first identified by Nowell & Hungerford 1960 and is the diagnostic hallmark of chronic myeloid leukaemia (CML); EML4-ALK and EML4-ROS1 fusions in lung adenocarcinoma; PML-RARA in acute promyelocytic leukaemia; EWSR1-FLI1 in Ewing sarcoma; MYC juxtaposition to immunoglobulin enhancers in Burkitt lymphoma t(8;14).
  • Insertional mutagenesis. A retrovirus or transposon integrates near a proto-oncogene and brings its enhancer into range. Historically important in the slow-transforming retroviruses (the avian leukosis virus / c-myc story is the canonical case) and a recurring concern in retroviral and lentiviral gene-therapy vectors.

An oncogene is dominant at the cellular level: an activating mutation in one allele is sufficient to confer the phenotype, because the abnormal protein imposes its activity in the presence of the normal protein from the other allele. This is what makes hotspot point mutations such a strong evidentiary signal in tumour-sequencing data — the same residue is hit again and again, in tumour after tumour, because no other class of variant satisfies the gain-of-function constraint.

Tumour suppressors: loss-of-function, recessive (mostly)

A tumour suppressor encodes a protein whose normal function restrains the malignant phenotype. The mechanisms of restraint vary — cell-cycle gatekeeping, DNA-damage response, apoptosis induction, signalling pathway termination, chromosome segregation fidelity — but the genetic logic is the same: malignancy requires inactivation of both alleles, because as long as one functional copy remains, the protein continues to do its job.

This is the genetic content of Knudson’s 1971 two-hit hypothesis. In the heritable form of retinoblastoma, the proband inherits one inactivating event in RB1 on chromosome 13q14, and the second event is acquired somatically in a single retinal cell — commonly by mitotic recombination, deletion, or point mutation, all visible as loss of heterozygosity (LOH) at flanking polymorphisms. In the non-heritable form, both events occur somatically in the same cell, a much less likely scenario, with correspondingly later age of onset and unilateral disease. The bimodal age distribution Knudson observed is the population-level evidence for the two-hit model; LOH analysis of paired germline-tumour DNA is the molecular evidence at the locus.

The canonical tumour suppressors and their associated tumour spectra:

  • RB1 (13q14) — retinoblastoma; osteosarcoma in retinoblastoma survivors; small-cell lung cancer (somatic).
  • TP53 (17p13) — the most-mutated gene in human cancer; germline inactivation underlies Li-Fraumeni syndrome; somatic mutation in roughly half of all sequenced solid tumours.
  • APC (5q22) — familial adenomatous polyposis (germline); colorectal adenoma – carcinoma sequence (somatic), as a near-obligate first hit.
  • BRCA1 (17q21) and BRCA2 (13q12) — hereditary breast and ovarian cancer (BRCAPRO); pancreatic cancer; male breast cancer; prostate cancer. BRCA-deficient tumours are homologous-recombination deficient and show mutational signature 3 (see somatic genomics).
  • PTEN (10q23) — Cowden / PTEN hamartoma tumour syndrome (germline); somatic loss in glioblastoma, endometrial, prostate, breast.
  • CDKN2A (9p21, encoding p16INK4A and p14ARF) — familial melanoma; familial pancreatic cancer (FAMMM); somatic loss in many cancers.
  • NF1 (17q11) — neurofibromatosis type 1; encodes neurofibromin, a RAS GAP.
  • NF2 (22q12) — neurofibromatosis type 2; encodes merlin.
  • VHL (3p25) — von Hippel-Lindau syndrome; somatic loss in clear-cell renal carcinoma.
  • STK11 (19p13) — Peutz-Jeghers syndrome; somatic loss in lung adenocarcinoma.
  • MLH1, MSH2, MSH6, PMS2 — the mismatch-repair genes underlying Lynch syndrome (MMRpro); biallelic inactivation gives a mismatch-repair-deficient (dMMR) tumour with microsatellite instability (MSI-H).
  • SMAD4, BMPR1A — juvenile polyposis.
  • CDH1 (E-cadherin) — hereditary diffuse gastric cancer; lobular breast cancer.

Haploinsufficiency: the exception to the recessive rule

A subset of tumour suppressors are dosage-sensitive: 50% of the normal protein dose is insufficient to maintain the suppressor function, so loss of a single allele is already phenotypically consequential. The cleanest example is PTEN, where mouse heterozygotes show tumour-prone phenotypes without somatic loss of the second allele, and human PTEN-haploinsufficient cells show measurable PI3K-pathway activation. TP53 shows partial haploinsufficiency in some contexts, and a number of mismatch-repair genes show subtle haploinsufficient phenotypes ahead of the second hit. Where haploinsufficiency operates, the tumour-suppressor / oncogene dichotomy is genuinely a continuum rather than a clean binary.

Gatekeeper, caretaker, landscaper

Kinzler & Vogelstein 1997 (Nature 386:761) proposed a functional taxonomy that survives essentially unchanged in modern teaching:

  • Gatekeepers directly restrain growth or promote apoptosis. Loss of a gatekeeper is rate-limiting in the tissue where it operates — APC in the colon, RB1 in the retina, VHL in clear-cell kidney, NF1 in the peripheral nervous system. Inactivation initiates a tumour in that tissue.
  • Caretakers maintain genome integrity. Loss of a caretaker does not initiate a tumour directly, but the resulting genomic instability accelerates the acquisition of mutations elsewhere in the genome, including in gatekeepers. The mismatch-repair genes (MLH1, MSH2, MSH6, PMS2) and the homologous-recombination genes (BRCA1, BRCA2, PALB2, ATM) are the canonical caretakers; this is why caretaker-deficient tumours show characteristic mutational signatures (see somatic genomics).
  • Landscapers shape the tissue microenvironment in ways that alter the niche for transformation. The juvenile-polyposis genes SMAD4 and BMPR1A are the original examples: loss in stromal cells changes the signalling landscape that epithelial cells experience.

The gatekeeper / caretaker distinction is operationally useful: gatekeeper loss demands a second hit at the same locus to release the constraint, while caretaker loss elevates the rate of mutation across the genome. The two contribute to malignancy through different mechanisms and respond to different therapeutic strategies — PARP inhibitors exploit caretaker loss in BRCA-deficient cells; immune-checkpoint inhibitors exploit the high mutation burden of caretaker-deficient tumours such as MSI-H colorectal cancer.

Driver and passenger mutations

Modern tumour-sequencing studies routinely yield hundreds to thousands of somatic point mutations per tumour, the great majority of which are biologically irrelevant passengers — mutations acquired in the lineage leading to the tumour but contributing nothing to the malignant phenotype. The minority are drivers — mutations whose presence has been positively selected because they confer a fitness advantage on the cell. Distinguishing drivers from passengers is one of the central problems of cancer genomics. Vogelstein et al. 2013 argued that the catalogue of true cancer driver genes is small — on the order of 140 across all cancer types — and identified the recurrent-mutation criterion as the strongest evidence: a residue or position hit far more often than the background mutation rate would predict is almost certainly under selection.

Statistical frameworks such as MutSigCV, dN/dS-based methods (e.g. dNdScv), and large-cohort recurrence-based approaches (OncoKB, IntOGen) operationalise driver-versus-passenger discrimination across the cross-cancer landscape. The output of each approach is a per-gene driver list and, ideally, a per-residue hotspot annotation that can be checked against a candidate variant of unknown significance.

The mutational landscape across tumour types

Cross-tumour comparisons from Vogelstein et al. 2013 and the later PCAWG analyses show a stable pattern: most tumour types are dominated by recurrent driver events in 5–10 genes, with a long tail of less-frequent drivers and a much longer tail of passengers. The pattern is highly tumour-type specific. Pancreatic adenocarcinoma is dominated by KRAS, TP53, CDKN2A, SMAD4. Cutaneous melanoma is dominated by BRAF, NRAS, NF1, TP53, with a large UV-derived passenger burden. Colorectal cancer separates into a chromosomally-unstable (CIN) subset dominated by APC / KRAS / TP53 and a microsatellite-unstable (MSI-H) subset enriched for MMR-pathway loss. Glioblastoma is dominated by EGFR, PTEN, TP53, CDKN2A, NF1. The mutation count per tumour also varies by orders of magnitude, from low-burden paediatric tumours to UV-driven melanomas and tobacco-driven lung cancers at the high end — relevant to the somatic genomics page’s coverage of tumour mutational burden as a biomarker.

Why this matters for family-history modelling

Inherited cancer-predisposition syndromes are dominated by germline tumour-suppressor variants, because the two-hit logic naturally accommodates one inherited and one somatic event. Inherited oncogene activation does occur (germline RET activating variants in MEN2; germline MET in hereditary papillary renal carcinoma) but is rare, because constitutive oncogene activation in every cell of an embryo is generally lethal or massively pleiotropic, not a recipe for a viable cancer-prone individual. The genes underlying the syndromes covered on the inherited cancer predisposition page are therefore overwhelmingly tumour suppressors: BRCA1, BRCA2, the MMR genes, APC, MUTYH, TP53, PTEN, STK11, BMPR1A, SMAD4, CDH1, VHL, NF1, NF2, TSC1, TSC2, RB1. The published risk-model algorithms Evagene implements (BRCAPRO, MMRpro, PancPRO) operate on this logic: family-history input plus carrier-frequency priors plus age-specific penetrance functions yield a posterior carrier probability for the relevant tumour-suppressor locus.

Selected sources

  • Knudson AG. Mutation and cancer: statistical study of retinoblastoma. PNAS 1971; 68:820. PMID 5279523.
  • Nowell PC, Hungerford DA. A minute chromosome in human chronic granulocytic leukemia. Science 1960; 132:1497. Reference.
  • Bishop JM. Molecular themes in oncogenesis. Cell 1991; 64:235. PMID 1988146.
  • Kinzler KW, Vogelstein B. Gatekeepers and caretakers. Nature 1997; 386:761. PMID 9126724.
  • Hanahan D, Weinberg RA. The hallmarks of cancer. Cell 2000; 100:57. PMID 10647931.
  • Hanahan D, Weinberg RA. Hallmarks of cancer: the next generation. Cell 2011; 144:646. PMID 21376230.
  • Vogelstein B, Papadopoulos N, Velculescu VE, Zhou S, Diaz LA, Kinzler KW. Cancer genome landscapes. Science 2013; 339:1546. PMID 23539594.
  • HUGO Gene Nomenclature Committee — genenames.org.
  • NCBI Gene (KRAS, BRAF, MYC, TP53, RB1, APC, BRCA1, BRCA2, PTEN, CDKN2A, MLH1) — ncbi.nlm.nih.gov/gene.

Related Evagene pages

Try Evagene’s pedigree platform

In-browser pedigree drawing with NSGC notation, gesture drawing, GEDCOM and CanRisk export, 20 published risk-model algorithms, and a 1,900-entry help catalogue. Free during alpha for clinicians, researchers, educators, and students. For research, education, and family-history documentation; not a medical device.

Join the Alpha Waiting List