A mutation found in cancer can also turn up in healthy cells

THE LINING of the food pipe is replaced throughout life. Cells at its surface are shed, and cells in the deeper layers divide to take their place. Each of those cells carries DNA, the material that contains the instructions a cell follows to build its proteins and do its work. A stretch of DNA with instructions for a product such as a protein is a gene. When a cell divides it must copy that DNA, and the copying is not perfect. DNA can also change through chemical reactions within cells. Most of the changes this leaves behind do nothing noticeable. Once a change is established in a cell's DNA, it can pass to descendants when that cell divides. A change acquired in this way during a person's life, rather than inherited from a mother or father, is called a somatic mutation. An inherited change is present in nearly every cell of the body from the start. A somatic one is present only in the cell where it arose and in the cells that descend from it.

Over the years, the descendants of a single mutated cell can become a large family of related cells sitting side by side in the tissue, all carrying the same change. Such a group is called a clone. A clone can spread across more and more of a lining while the lining still looks entirely ordinary to a doctor or under a microscope. Most changes give the cells that carry them no particular advantage, and their clones stay small or disappear as cells are shed. A few changes help their carriers divide more often, or be lost less often, than the cells around them. Over time the descendants of such a cell come to outnumber their neighbors. Biologists call this positive selection. The advantage belongs to the cells, not to the person, and a cell that out-competes its neighbors has not thereby become a tumor.

This matters because many of the genes in which these advantageous changes occur are the same genes that are found mutated in cancers. So a familiar question arises for anyone who reads a genetic test report. If a change known from cancer is found in a piece of tissue, what does that say about the tissue, and about the person it came from?

Reading normal tissue

A study published in Science in 2018 set out to measure how common such changes are in a normal food pipe. Iñigo Martincorena and Joanna Fowler, the paper's joint first authors, and Philip Jones, a corresponding author, all of the Wellcome Sanger Institute in Hinxton, in Britain, worked with tissue from nine deceased organ donors aged between 20 and 75. None of the donors had a known history of disease of the esophagus, the medical name for the food pipe, or of other chronic disease. The researchers peeled the lining of the upper and middle esophagus away from the tissue beneath it and cut it into a continuous grid of small pieces, each two square millimeters in area. That gave them 844 samples. Because the pieces lay next to each other, a clone that spread across several could be traced between them.

Each piece was examined under a dissecting microscope, and no abnormal patches were seen. Neighboring tissue was also examined in thin slices and with a form of three-dimensional microscope imaging, and it too looked normal. The team then read the order of the chemical letters in selected stretches of DNA, a process called sequencing. Rather than reading every letter of every cell, they read selected stretches from 74 genes known to be involved in cancer, and read each position many times over. After duplicate readings were removed, the middle value of the coverage across targeted positions, the median, was about 870 readings.

Reading so many times is necessary because a change carried by a small clone appears in only a small fraction of the readings from a sample. Sequencing machines also make occasional errors, and a rare error can look much like a rare genuine change. The researchers handled this with a statistical method that used the error rates observed at each position across a large collection of normal samples to build a model of the errors expected at that position. That model helped distinguish likely mutations from errors. Using this approach, they identified 8,919 somatic mutations in the protein-coding parts of the targeted genes across the 844 samples. When the same mutation turned up in neighboring pieces, they treated it as one clone, which left 6,935 independent mutation events. The number of mutations found, and the size of the clones, tended to be greater in older donors.

A label reconsidered

The most surprising result concerned one gene in particular. Cells are constantly exchanging signals with the cells beside them, and some of those signals are received by proteins that sit on a cell's surface and bind to partner proteins on neighboring cells. One gene contains the instructions for making such a receiving protein, and its surface portion contains the regions that bind its partners, called Jagged and Delta. The gene is called NOTCH1. Across the nine donors, the researchers found 2,055 coding mutations in NOTCH1. The patterns of those mutations, many of which would be expected to disable the protein or alter the regions that bind its partners, showed strong positive selection. In other words, cells carrying a damaged copy of the gene tended to leave more descendants than their neighbors. The team's estimates, based on how often the mutations appeared among the readings rather than on a count of individual cells, suggested that NOTCH1-mutant clones made up a large share of the lining in most of the middle-aged and older donors and a much smaller share in the three donors under 40.

NOTCH1 had been regarded as a gene that drives a cancer of the flat cells in the lining, called squamous cell carcinoma, because it is mutated in a minority of those tumors. Martincorena and his colleagues found that the gene appeared to be mutated more often in the normal lining of middle-aged people than in the cancers that arise from that lining. Those two figures do not share a denominator, since one is an estimate of the fraction of normal tissue carrying a mutation and the other is the fraction of tumor samples carrying one, so the comparison says nothing directly about any person's chance of cancer. It does, however, change the meaning of the label "cancer gene". A gene can appear repeatedly in catalogs of tumor mutations simply because mutations in it are already common in the normal tissue from which tumors arise. The authors wrote that drivers of noncancerous clonal growth "could still appear as recurrently mutated genes in cancer genomes owing to their high mutation frequency in the normal cells from which tumors evolve". They suggested that cancers of this type may be more likely to arise from cells without NOTCH1 mutations.

A different gene, TP53, gave the opposite pattern. It is mutated in nearly all of these esophageal cancers but was found in a minority of the cells in the normal lining. The authors suggested that cancer risk may therefore vary across the aging lining, depending on which mutations the clones in it carry. They also listed ways in which the normal clones differed from cancer cells. The normal cells carried roughly a tenth as many mutations as many of these cancers, showed no sign of a particular pattern of DNA damage common in such tumors, and showed no apparent sign of the unstable chromosomes seen in these cancers. The cells in these clones had expanded within normal tissue, which is a separate step from becoming malignant.

The study has clear bounds. It looked at nine people at one moment, after death, so it cannot say how common such clones are across the population, and it did not follow anyone to see whether cancer developed later. It does not show that NOTCH1 mutations protect people against cancer, and it tested no treatment. Tissue that looks normal is also not proof that a clone in it can never cause harm.

Following people over time

That leaves the question a patient would most want answered, which is whether carrying such a clone changes what happens later. The esophagus study could not address it, but an earlier study of blood could. Many blood cells are short-lived and are continually replaced by a small population of parent cells in the bone marrow. When one of those parent cells acquires a mutation that gives it an advantage, a growing share of the circulating blood cells can come to descend from it and carry the same change. This is known as clonal hematopoiesis, from the medical term for blood formation.

In 2014 Siddhartha Jaiswal and Benjamin Ebert of the Division of Hematology at Brigham and Women's Hospital and Harvard Medical School, working with colleagues at the Broad Institute and in several population studies, reported in the New England Journal of Medicine on blood DNA from 17,182 people. These people had been selected without regard to any blood disorder. Most, 15,801, were patients and comparison participants from studies of type 2 diabetes. The researchers looked for previously described changes in 160 genes that are repeatedly mutated in blood cancers. Their ability to detect a change depended on how many times each position had been read, so smaller clones could have been missed. Such mutations were very rare in people younger than 40 and became more common with each decade of age after that. Among people aged 70 to 79, 219 of 2,300 carried a detectable mutation, or 9.5%. The authors noted that the true share depends on how such mutations are defined and on the sensitivity of the method, and may be substantially higher.

Two of the cohorts, comprising 3,342 people, had recorded which participants were later diagnosed with cancer. Of those 3,342 people, 134 carried a detectable mutation. Over a median follow-up of 95 months, 16 blood cancers were reported among the 3,342, and five of them occurred among the 134 carriers.

Comparing risks

The way such numbers are reported can make them seem larger or smaller than they are. One way is to compare the rate at which illness occurred in one group with the rate in another, which tells how much more often something happened to carriers than to similar people without the mutation. Another is to count how many people in a group actually became ill. Comparing how quickly cases occur during follow-up gives a hazard ratio. Here the analysis also accounted for deaths that prevented a later cancer diagnosis. After adjusting for age, sex and diabetes, the researchers estimated a relative rate of blood cancer of 11.1 among carriers compared with non-carriers. The uncertainty range, a 95% confidence interval, ran from 3.9 to 32.6. That interval is wide because the estimate rests on few cancers. A hazard ratio of 11.1 is a comparison of rates. It is not a probability, and it does not mean that 11 of any particular number of carriers will become ill.

Counted directly, five of the 134 carriers developed a blood cancer over a median of 95 months, and the other 129 did not during that time. The authors described this as an absolute risk of about 4% over the study period, or about 0.5% a year, and called it small. They were unable to examine the later cancers to establish whether each had arisen from the clone detected earlier. They also wrote that most carriers did not have a blood cancer when their blood was sampled, and that it would be premature to screen healthy people for such clones because a positive result predicts cancer poorly. Those statements come from 2014 and are not current clinical guidance.

The two studies involved different tissues, different people and different research groups. Together they show that cells carrying mutations known from cancer are common in normal human tissue, that they become more common with age, and that in blood, at least, carrying a detectable clone was linked to a higher rate of later cancer although most carriers did not develop a blood cancer during follow-up. The blood results cannot be applied directly to the esophagus, and neither study shows that every such mutation causes cancer. Finding a cancer-linked mutation in a sample of tissue therefore does not by itself diagnose cancer. Judging what it means requires knowing which gene is affected, in which tissue, how large the clone is and what happened to people who carried similar ones.