Oncogenes: When the Switch Sticks

In 1911 Peyton Rous passed a ground-up hen tumour through a filter that stops bacteria and injected the clear fluid into healthy hens. New tumours grew where the needle went in. Something smaller than a bacterium had carried the cancer from bird to bird. Sixty-five years later the gene this agent uses to cause cancer turned up somewhere surprising: in the DNA of healthy birds, as one of the birds’ own genes. This lesson gathers what earlier lessons said about cancer into one argument: how a signalling switch gets stuck on, and why one stuck switch is seldom enough to make a tumour.

The story: the cancer gene was the chicken’s own

The hen’s tumour was a sarcoma, a cancer of connective tissue. Before Rous trusted his filter he tested it against a bacterium, and it held the bacterium back. So the agent was smaller than a bacterium, the size of a virus, and that was all the filter could say. By Rous’s own account his result was doubted for about fifteen years.

The agent is now called Rous sarcoma virus. By the early 1970s Peter Vogt and others had isolated mutants of it that still multiplied but could no longer turn a normal cell into a tumour cell, and Peter Duesberg and his colleagues found that their RNA was missing a piece. The missing piece held a gene, named src (say “sark”) for sarcoma. A gene that can turn a cell cancerous is an oncogene.

Where did a virus get such a gene? In the laboratory that Michael Bishop and Harold Varmus shared in San Francisco, a French postdoctoral fellow, Dominique Stehelin, made a radioactive strand of DNA that matched src and no other part of the virus. A single strand of DNA sticks to its matching sequence, so this probe could ask whether any sample of DNA held a relative of src. It stuck to DNA from chickens that had never met the virus.

That alone proved little. A leading idea of the day held that old infections had left viral genes behind in chicken chromosomes. So Stehelin tried duck, quail and emu, a flightless bird at a great evolutionary distance from the chicken. The probe stuck to all of them, and the matches were close enough that Varmus judged the sequence had been conserved for at least 100 million years. A sequence shared across that distance is older than any infection of chickens. It was a gene of the birds themselves, and later work found it in mammals, people included. The paper appeared in Nature in March 1976 under four names: Stehelin, Varmus, Bishop and Vogt.

A normal gene that can become an oncogene when it is altered is a proto-oncogene. The virus had picked up the cell’s gene, c-src, by accident, and what it carries now is an altered copy, v-src.

What does the gene make? In 1978 two groups reported that the Src protein is a kinase: Marc Collett and Ray Erikson, and Art Levinson with colleagues in the Bishop and Varmus laboratory. Two years later Tony Hunter and Bart Sefton showed that it puts its phosphate on tyrosine, a story told in the phosphorylation lesson. And what has v-Src lost? The lesson on interaction modules showed how c-Src holds itself shut: a tyrosine near its tail end (tyrosine 527 in the chicken protein) carries a phosphate, the protein’s own SH2 domain grips it, and the kinase is clamped. In v-Src the tail has been replaced and that tyrosine is gone. Nothing closes the clamp, so the kinase runs all the time, and a few changes elsewhere in the protein raise its activity further.

The idea: five ways to stick a switch

Every pathway in this course has an on-step and an off-step. Jam an on-step or remove an off-step, and a cell that should divide only when told will divide on its own. Most kinases have been associated with disease, and the ways a kinase or any other switch goes wrong fall into a short list.

What changes	Example	Why the switch sticks	Pathway taught in
One amino acid, breaking the off-step	Ras G12V or Q61	Ras cannot hydrolyse its GTP, even with a GAP	GTPase switches, Ras to MAPK
The piece that holds a kinase shut is lost	v-Src	No tail tyrosine for the SH2 domain to grip	Interaction modules
Too many copies of a normal gene	HER2	Crowded receptors pair without growth factor	Receptor tyrosine kinases
Two genes joined	BCR-ABL	The BCR part holds ABL kinases together	Receptor tyrosine kinases, the phosphorylation switch
The off-protein is lost	PTEN, neurofibromin, APC	PIP3, Ras-GTP or β-catenin is not cleared	PI 3-kinase and Akt, GTPase switches, Wnt signalling

Rows one and two you have met. In row three, roughly 15 to 20 percent of breast cancers have the HER2 gene copied many times over. The protein is normal, but there is far too much of it, and crowded receptors pair without growth factor.

Row four began with a small chromosome, seen in 1960 by Peter Nowell and David Hungerford in the blood cells of people with chronic myeloid leukaemia. In 1973 Janet Rowley showed that the missing piece of this chromosome 22 was attached to chromosome 9. Later work found that the join fuses two genes, BCR and ABL. ABL makes a tyrosine kinase related to Src. The BCR part makes the fusion proteins stick to each other, so their kinase domains sit side by side and switch each other on, as paired receptors do.

“Gain of function mutations in RTKs” could head a table of its own: EGFR and its relatives, KIT, MET and RET are each found stuck on in particular human tumours.

Row five is different in kind. PTEN removes the phosphate that PI 3-kinase adds to make PIP3, neurofibromin is a GAP for Ras, and APC is part of the complex that marks β-catenin for destruction. No on-step is jammed here. An off-step is missing. A gene whose loss lets cells grow when they should not is a tumour suppressor.

One copy or two: gain acts alone, loss needs both

You carry two copies of nearly every gene, one from each parent. For the first four rows a change in one copy is enough. The altered protein signals without being told to, and the normal protein from the other copy cannot stop it. This is a gain of function, and it is dominant: it shows even when the other copy is normal.

For row five, one good copy usually makes enough protein to do the job, so both copies must go before the off-step fails. Alfred Knudson reached this idea in 1971 from the ages at which children were diagnosed with retinoblastoma, a tumour of the eye. The inherited form, often in both eyes, appeared on a schedule that fitted one chance event, and the form that is not inherited fitted two. He proposed that the tumour takes two “hits”, and that children with the inherited form are born with the first in every cell. The two hits turned out to be the two copies of one gene.

Gene class On-step, e.g. Ras (oncogene) Off-step, e.g. neurofibromin, the GAP for Ras (tumour suppressor) Copy 1 Copy 2

Exceptions the lesson names. A broken protein can still act through one copy if it gets in the way: the laboratory tool Ras S17N stays in its GDP form and ties up the exchange factor that normal Ras needs, and many p53 missense mutants poison the tetramer they form with normal subunits, so they behave as dominant negatives. Two-hit behaviour holds only when one copy gives enough protein, and that does not always hold: PTEN shows dosage effects in mice. The lamp here follows the usual rule only.

Rule to keep: one altered copy of an on-step is enough, while an off-step usually has to lose both copies. In this sketch the lamp shows one step stuck on, not a tumour, because one stuck switch is seldom enough for a tumour.

There are many ways to break a protein and few ways to jam it on. That is why Ras changes in tumours cluster at positions 12, 13 and 61, while changes in a tumour suppressor are scattered along the gene. A broken protein can still act through one copy if it gets in the way: the laboratory tool Ras S17N stays in its GDP form and ties up the exchange factor that normal Ras needs.

Design problem

Take the chain receptor, Ras, Raf, MEK, ERK, with an off-switch at each step. List every kind of single change you can think of that would leave ERK on with no growth factor present. Sort your list into changes that work when one copy of a gene is altered and changes that need both copies lost. Then say which kind you would expect to find more often as an inherited risk for cancer, and why.

One way to do it

One altered copy is enough (gain). A receptor that pairs without growth factor, because of a point mutation, a missing outer part, too many copies, or fusion to a partner that pairs it. A Ras that cannot hydrolyse GTP. A Raf or a MEK fixed in its active shape (BRAF V600E, common in melanoma, is one).

Both copies must be lost (loss). The GAP for Ras, neurofibromin. The phosphatases that strip phosphate from the receptor, MEK or ERK. The ligase that tags an active receptor for removal.

Inherited risk. Expect the loss kind more often. An inherited change is in every cell from the first division. A strongly stuck on-switch in every cell disturbs the embryo, and the milder inherited activating changes in this pathway cause developmental syndromes called RASopathies. One lost copy of an off-switch is silent while the second copy works, so the person is born healthy. Every cell is then one hit away from having no off-switch, and in a body of many billions of cells a second hit somewhere is close to certain. Neurofibromatosis type 1, one inherited bad copy of the neurofibromin gene, works this way. There are exceptions: inherited activating changes in the receptor RET cause a cancer syndrome.

Why one stuck switch is not enough

A normal cell treats a growth signal that never stops as a fault. In 1983 Robert Weinberg’s laboratory put an always-on ras gene into cells taken fresh from rat embryos. The cells did not become tumour cells. Nor did they with myc alone, a gene for a transcription factor that drives growth. With both genes together they did.

What stops a cell with one stuck switch? Checkpoints do. In normal cells p53, from the stress lesson, answers excess growth signalling much as it answers damaged DNA, and it halts the cell or kills it. A cell that gets past p53 still needs the survival signals of the cell death lesson, or it dies by apoptosis. So a tumour collects changes: a stuck on-switch, a lost off-switch, a broken checkpoint, a block on death. In the colon a common order is known, with loss of APC early, an always-on Ras later and loss of p53 later still.

A tumour that needs its oncogene

How can we exploit this to treat cancer? A cell with several changes sounds like a cell with no single weak point. The surprise is that many tumours stay dependent on one of their stuck switches. In mice built so that an oncogene can be turned off at will, tumours that grew with the oncogene on have shrunk when it was turned off. The other mutations were still there, yet the cells stopped dividing or died.

Why should that be? This dependence is called oncogene addiction, and its cause is still debated. One idea uses the negative feedback of the circuits lesson: a pathway stuck on keeps its own feedback inhibitors high, and those inhibitors may quiet the routes beside it, so that nothing is ready to take over when the stuck signal is removed. Another idea is that survival signals fade faster than death signals once the oncogene is switched off. The ideas are not exclusive, and which one matters may differ from tumour to tumour. Whatever the cause, the dependence is why a drug aimed at one protein can sometimes shrink a tumour that carries many mutations. Imatinib against BCR-ABL, from the biochemistry course, is the best-known case. The next lesson is about those drugs.