Cancer as a Genetic Disease

Every cell in a child who inherits one broken copy of the gene RB1 carries that broken copy. Yet most such children develop only a few tumors, mostly in the retina of the eye, and often in both eyes.

If the inherited allele were enough to make a cell cancerous, the whole retina would be tumor. The way out is that the inherited mutation is the first of two steps, and the second is a rare event in only a few cells. That idea, worked out from clinic records in 1971, organizes much of what is known about how mutations make cancer.

Cancer is a genetic disease because tumor cells carry altered DNA that they pass to their daughter cells. That does not make it inherited. Most changes in a tumor are somatic: they arise in body cells during life and are not passed to children.

Two kinds of cancer gene: accelerators and brakes

Think of a car that will not stop: either the accelerator is stuck or the brakes have failed.

A proto-oncogene is a normal gene whose product drives growth or division. An oncogene is an altered, overactive form: a gain of function. Because the altered protein acts on its own, one mutant allele (one copy of the gene) is usually enough, so the mutation is dominant within the cell. RAS is a switch that is on while it holds GTP and off once it cuts GTP to GDP; the oncogenes lesson and the RAS-to-MAPK lesson cover the signaling. In 1982 several laboratories, among them those of Mariano Barbacid, Robert Weinberg and Michael Wigler, reported in Nature that the transforming gene from a human bladder carcinoma line differed from normal HRAS by one base change, which put valine in place of glycine at codon 12. A purified Val12 protein hydrolyzed GTP several times more slowly than normal HRAS protein made in E. coli in a 1984 assay.

MYC is usually activated another way: by extra copies (amplification) or by a translocation that moves it next to a highly active gene. In Burkitt lymphoma the commonest change is t(8;14), which places MYC beside an immunoglobulin heavy-chain gene. Riccardo Dalla-Favera and colleagues mapped c-myc to chromosome 8 in 1982 (see chromosome changes).

A tumor-suppressor gene is a brake: its product restrains division or stops a damaged cell, and cancer follows its loss of function. One working copy usually gives enough brake, so a cell must normally lose both alleles. RB1 encodes RB, which in its underphosphorylated form binds E2F transcription factors that switch on genes for DNA replication. TP53 encodes p53, which accumulates after DNA damage and can halt the cell cycle or trigger cell death (see stress signals and death receptors and BCL-2). Returning a cloned RB gene to RB-deficient human tumor cells suppressed their tumor-like growth in a 1988 test.

Now the paradox. An inherited tumor-suppressor allele shows autosomal dominant inheritance in a family: about half a carrier’s children inherit the risk, and affected people appear generation after generation (see reading pedigrees). Yet the allele is recessive in the cell. Both hold because a pedigree tracks risk, and a carrier has a vast number of cells that each already have one hit, so a second hit in at least one is likely. Penetrance, the fraction of carriers who show the trait, is therefore high but below 100%.

p53 warns against tidy rules. It acts as a tetramer, and many TP53 missense mutant proteins can disable wild-type subunits they join, a dominant-negative effect. In mice with one p53-null allele, many tumors kept a functional wild-type allele, so reduced dose may suffice in that model.

Retinoblastoma and the two-hit arithmetic

Retinoblastoma is a tumor of the developing retina. Cases with a family history tend to be bilateral (both eyes) and diagnosed earlier; cases without one usually involve a single tumor in one eye. Alfred Knudson analysed 48 cases with published series and proposed in 1971 that two mutational events are needed. In the hereditary form the first is inherited through the germline and present in every cell; in the nonhereditary form both must happen somatically, in the same cell. He estimated that carriers develop about three tumors on average. Knudson did not find the gene: a cDNA from the RB1 locus, on chromosome 13, was isolated by Stephen Friend and colleagues in 1986.

Worked problem. Assume that in a carrier of one mutant RB1 allele each tumor arises from an independent second hit, so the number of tumors per eye follows a Poisson distribution with mean m = 1.5 (3 per person, Knudson’s approximate figure), and the eyes are independent. What fractions of carriers have no tumor, a tumor in one eye only, and tumors in both eyes?

Step 1. The Poisson probability of zero events is e^(−m).

P(no tumor in one eye) = e^(−1.5) = 0.223

Step 2. So P(at least one tumor in an eye) = 1 − 0.223 = 0.777.

Step 3. No tumor in either eye: 0.223 × 0.223 = 0.050, which equals e^(−3). About 5% of carriers stay tumor free, so penetrance is about 95%.

Step 4. Tumors in both eyes: 0.777 × 0.777 = 0.604.

Step 5. A tumor in one eye only, where either eye can be the affected one: 2 × 0.223 × 0.777 = 0.347.

Check: 0.050 + 0.604 + 0.347 = 1.001, which is 1 within rounding. Most affected carriers have bilateral disease, and about one carrier in twenty has no tumor yet can pass the allele on.

Now a sporadic case. Both rare hits must fall in one cell, so the expected number of tumors per eye is far below 1. If it were 0.0001, a tumor in each eye would have a chance of about 1 in 100 million. Sporadic tumors are therefore single and one-sided, and because two events must accumulate in one cell lineage, they tend to be diagnosed later.

How the second hit happens: loss of heterozygosity

A carrier is heterozygous at RB1: one mutant allele, one wild-type. The second hit can be a new point mutation in the wild-type allele, or silencing of its promoter by DNA methylation (see eukaryotic regulation). Often it is something larger. Loss of heterozygosity (LOH) means that a tumor cell has lost the wild-type allele and kept only the mutant one. It can happen by deletion, by loss of the whole chromosome through nondisjunction, by loss then duplication of the remaining chromosome, or by mitotic recombination, a crossover between homologs in a dividing body cell that can leave a daughter cell with two copies of the mutant region.

In 1983 Webster Cavenee, Thaddeus Dryja and colleagues tested this with DNA markers that differ between the two copies of chromosome 13. In several patients whose normal tissue was heterozygous at those markers, the tumor showed only one allele. The patterns fitted mitotic recombination, or nondisjunction with loss of the wild-type homolog. The second hit was often chromosomal, and LOH in tumor DNA became a way to find suppressor genes in other cancers.

Instability and clonal evolution

A tumor is a clone, the descendants of one cell. In 1976 Peter Nowell proposed that tumors evolve: acquired genetic variability lets more aggressive sublines be selected, and treatment can fail when a resistant subline emerges. As in bacteria under an antibiotic, mutation supplies variants and selection keeps those that divide more or die less.

Genome instability raises the supply. Mutations in DNA-repair genes tend to cause further mutations in other genes. Cells lacking mismatch repair, as in Lynch syndrome tumors, accumulate copying errors in short repeats. Loss of BRCA1 or BRCA2 impairs repair of double-strand breaks by homologous recombination (see DNA repair and recombination and mutation). Faulty chromosome segregation produces whole-chromosome gains and losses, and chromothripsis can produce many clustered rearrangements in one catastrophic event.

Colorectal tumors show the accumulation. In a 1988 series of 172 specimens, Bert Vogelstein and colleagues found RAS mutations in 9% of adenomas smaller than 1 cm, 58% of larger adenomas and 47% of carcinomas, and the changes accumulated as tumors progressed. Eric Fearon and Vogelstein’s 1990 model stressed the accumulation of changes over their exact order; it is a colorectal model, not a fixed sequence for every tumor.

A driver gives its cell a selective advantage; a passenger happened to be present when a clone expanded. Across 2,658 whole cancer genomes from 38 tumor types, each with matched normal tissue, the Pan-Cancer Analysis of Whole Genomes consortium identified an average of 4.6 drivers per tumor with current knowledge, a figure that depends on the rules used to call drivers. Selection also works before cancer, in clones of normal human esophagus.

Reading a tumor genome

To call a variant somatic, you sequence the tumor and a matched normal sample, often blood, from the same person (the sequencing lesson covers the method). A variant in both is inherited; one only in the tumor is somatic. The variant allele fraction (VAF) is the fraction of reads that carry the variant. A tumor sample also contains normal cells, so its purity, the fraction of tumor cells, must enter the reading.

expected VAF (heterozygous, all tumor cells, two copies) = purity × 0.5

Genotype now shapes diagnosis: in chronic myeloid leukemia the 9;22 translocation makes the BCR-ABL kinase that imatinib inhibits (see drugs against a part). In a 2021 study of acute myeloid leukemia and myelodysplastic syndromes, whole-genome sequencing in 117 consecutive patients took a median of five days and changed the genetic risk category for 19 of them.

Design problem

A colorectal tumor sample is estimated at 60% tumor cells. Tumor and matched-normal sequencing gives: KRAS G12D at VAF 0.30 in the tumor; TP53 R175H at VAF 0.60, with the other TP53 allele lost; PIK3CA at VAF 0.08; and a BRCA2 frameshift at VAF 0.50 in normal tissue and 0.80 in the tumor. Assume two copies of each region unless the data say otherwise. Which variants are inherited, which are in every tumor cell, and which are in a subclone? Why is the matched normal needed?