VIRTUALLY EVERY TIME A cell divides, some small error is introduced into its genomic code. That error might mean nothing at all, or it might lead to a modest change in the function of a cell. On a geologic time scale, a pileup of these tiny mutations is the basis for all evolution, how one species becomes another by developing claws, a shell or a complex brain. But at the speed of a single human life, that progression is slow and mostly silent. “The enzymes in your cells have to copy six billion letters with every division,” says Moritz Gerstung, a computational cancer biologist at the European Bioinformatics Institute in Cambridge, England. “That they only make about one mistake per division is remarkably good.”

Still, with cells dividing roughly 10 quadrillion times during a human lifetime, an accumulation of tiny mistakes can sometimes yield deadly effects. One of the most familiar is cancer. This happens when DNA “processing errors,” coupled with damage from external carcinogens and other factors, cause mutations that allow cells to go into reproductive overdrive, growing out of control and eventually overtaking healthy cells, bypassing the body’s ability to police and repair errors and eventually crowding out the body’s healthy tissue.

One of the most successful recent frontiers in cancer research, powered by advances in genomic sequencing, has been to pinpoint which mutations initiate cancer and explore how each one may help tumor cells thrive. Creating a rogue’s gallery of mutations and their functions has led to earlier and more accurate diagnoses, treatments that can narrowly target the mutation’s effects and an overall better prognosis for many cancer patients.

Now a flood of new research is vastly expanding what is known about mutations—how they arise and what transforms them into agents of disease. In one landmark development, an international team of scientists analyzing the largest set of cancer sequencing data ever assembled has produced the most extensive catalog of cancer-causing mutations. The Pan-Cancer Analysis of Whole Genomes (PCAWG) Consortium launched a flurry of publications in February 2020 that detailed a host of insights both profound and useful.

These and other recent discoveries are bringing into focus a more complex and fascinating picture of the role of mutation in cancer. Mutations—even those that are quite dangerous—may be more widespread than researchers had thought, and may lie dormant in the body for much longer than previously believed, relying on a particular cascade of factors to kickstart the disease. Researchers are also seeing how the position of mutations in chromosomes may hold clues to understanding their impact, and how their location in the body may lead to wildly different outcomes for the same genetic error.

One major result of this work is a changing view of mutation itself. It is increasingly seen not as an error of the body but as its natural background activity, one that has a profound effect as humans age and minor transcription errors add up. Although it can’t be stopped, mutation can be better understood, and today’s efforts to trace the course of the mutations may lead to opportunities to intervene early and more effectively, signaling a turning point in the way that cancer is diagnosed and treated.

ONE SURPRISE OF RECENT research is just how widespread potentially deadly mutations may be. That was clear in the results of a study published in November 2018 in Science, in which researchers at the Wellcome Sanger Institute near Cambridge, England, looked at esophageal tissue from deceased donors. New sequencing methods allowed them to examine small populations of cells and see which mutations those cells held in common. None of the donors had esophageal cancer, and the researchers expected to see fewer mutations than in, for example, the skin, which is subjected to the mutagenic effects of the sun. While there were, indeed, fewer than in the skin, the esophageal mutations were still manifold. People in their early twenties carried several hundred mutations per cell, and in samples from older donors, there were more than 2,000 mutations per cell.

More surprising than the sheer number of mutations in normal tissue, though, was how many of those alterations involved genes known to be mutated in cancer. Tissue samples from middle-aged and elderly donors, although they showed no signs of cancerous lesions under a microscope, had “mutant clones”—clusters of cells with cancer-like mutations—colonizing more than half of their surface. These mutant clones behaved like cancer in that they multiplied rapidly in order to gain a competitive advantage over neighboring cells. But they weren’t cancer.

In these samples, the most prevalent cancer-driving mutations affected the TP53 and NOTCH genes. Mutations in TP53 are found in about half of all cancers, but they were also found in up to 37% of these esophageal cells from healthy donors. Even more unexpected was the prevalence of mutations in the NOTCH1 gene, which helps to control cell division. Because it’s mutated in about 10% of esophageal tumors, the NOTCH1 gene has been widely assumed to be a “driver mutation,” crucial for helping cancer cells proliferate. But the Sanger Institute researchers were shocked to find NOTCH1 mutations in up to 80% of the noncancerous esophageal cells taken from older donors. That suggested that mutant versions of this gene, by themselves, might not be sufficient  to push cells into malignancy.

Iñigo Martincorena, who co-led the study, has speculated that in a healthy body, clones with different mutations arise and compete for available space and resources—and that rivalry somehow keeps each of them in check, by not allowing any single population of mutated genes to dominate. And tolerating a certain amount of DNA damage as normal seems to be biologically advantageous, says Serena Nik-Zainal, a clinician scientist at the Medical Research Council Cancer Unit at the University of Cambridge. Because the vast majority of mutations aren’t particularly harmful, responding to every nick and scratch as if it were a three-alarm fire can be too “expensive” from a cellular survival standpoint. In conditions of high DNA damage—such as exposure to ultraviolet light or to carcinogenic chemicals—focusing too much energy on repairing the genome perfectly could exhaust a cell and kill it. Nik-Zainal hypothesizes that the abundance of mutations in normal cells reflects not a compromised ability to repair DNA, but rather a management strategy.

The research underscores the idea that the mere presence of certain mutations isn’t sufficient to initiate the disease. It takes additional outside factors to create an environment in which cancerous cells take over. “A mutated genome may contribute to the potential for malignant transformation, but it does not on its own always determine it,” Nik-Zainal says.


WHY DO SOME CELLS, even when they are riddled with driver mutations—like the cells in that healthy esophageal tissue—not progress to cancer? One reason that mystery has been so difficult to solve is that malignancies don’t develop all at once, or steadily. Rather, they appear to grow in a series of clustered events that may stretch over long periods, even decades, and a typical cancer diagnosis occurs at age 60 or older. By that time, a tumor’s genome already reflects a life’s worth of genetic changes, and it can be almost impossible to reconstruct when and how mutations led to cancer.

It would be ideal to flash back in time and somehow take genetic snapshots of a growing tumor from the time it was a single healthy cell. A team led by Gerstung at the European Bioinformatics Institute has figured out how to do that, virtually, devising a method to reconstruct the evolution of a tumor from a single biopsy. These researchers detailed “life histories” of 38 cancer types for an analysis published in Nature in February 2020.

They took advantage of the fact that tumors contain cells from multiple generations. Sets of mutations in each generation evolve further away from their common genetic ancestor and also tell a story of what mutations happened at each phase. So by sequencing cells from different parts of one tumor, researchers were able to deduce the most recent common ancestor for all of them. From there, they could continue to work backward to infer what happened during previous rounds of mutation and cell division. Gerstung’s team identified other mutations in the cell that occur as a normal part of aging and used them as markers—something like the rings of a tree—to gauge when particular cancer-specific mutations occurred.

The scientists were surprised to find how early some important cancer-causing mutations showed up. In brain cancer, for example, one crucial driver mutation that alters chromosomal structure sometimes develops even before birth. “That is absolutely astonishing,” Gerstung says. “How these people lived so long with such a dramatic alteration before it led to disease is a big unanswered question.”

For most cancers, though, the period between the first cancer-causing mutation and diagnosis is shorter—a matter of several years, not several decades. The team also found that, once the first mutation occurred, a cell typically required an increasingly narrow set of additional changes to become cancerous. It turns out that common driver mutations that are shared by many cancer types—involving TP53 and KRAS genes, for example, and noncoding changes affecting the TERT gene—tend to occur early in cancer evolution. In fact, half of all early-stage mutations in cancerous tumors involve just nine genes. It’s only later that a tumor is likely to differentiate itself with a more specific, diverse set of mutations that involve about 35 genes. Gerstung’s group identified timelines of mutation in colorectal cancer, glioblastoma and pancreatic cancer, among others, about which little had been known.

BUT KNOWING WHEN CANCEROUS changes occur and what genes they’re likely to involve still leaves the question of how and why mutations sometimes cascade into cancer. Important clues can be found by examining underlying patterns of DNA damage known as mutational signatures, and in recent years, there has been a push to catalog the signatures that specifically give rise to driver mutations. The Pan-Cancer (PCAWG) Consortium has put out the most extensive analysis yet of these signatures.

Mutational signatures take researchers a step beyond knowing which genes are mutated, to understanding how they got that way—an interplay of natural failure of DNA repair, for example, and damage caused by internal processes, smoking or too much sunlight. Each type of damage leaves a particular mark, or signature—a characteristic pattern of messing with the cellular DNA. Tobacco, for example, changes the DNA base chemical cytosine to adenine.

The PCAWG researchers identified 97 distinct mutation signatures across 38 types of tumor. The majority of these involved so-called single-base substitutions, in which a single DNA base letter replaces another. Others involved double-base substitutions, affecting two DNA bases, and insertions or deletions of small sections of DNA.

Knowing these signatures can help clinicians spot weaknesses—and even offer clues for treatment. For instance, some cancer cells carry a signature that indicates they have a limited ability to make routine DNA repairs. That characteristic helps the cancer cells mutate and expand, but it can also be used against them. With more mutations, “the cell has more liabilities, more degraded functioning,” says Gad Getz, director of bioinformatics at the Mass General Cancer Center, who co-led the PCAWG Consortium study. “It’s sicker than those without mutations.” Through radiation or chemotherapy, clinicians can prod tumor cells to mutate so much that they become no longer viable. PARP inhibitor drugs, for example, are specifically designed to target tumor cells with defective DNA repair mechanisms.

But this approach to treatment doesn’t always work, and deliberately mutated cells may evolve resistance to the therapies. So Getz and others are now working to characterize the mutational signatures and drivers of treatment-resistant tumors, too, which could improve treatment of recurrent cancers.

By using whole genome sequencing data, another team of researchers in the PCAWG Consortium was also able to identify patterns of larger-scale DNA damage—so-called “structural variants.” Those involve rearrangements of large chunks of DNA across chromosomes, rather than just a few DNA letters getting altered within specific genes. These seismic rearrangements of DNA are a significant factor in many cancers.

“When people think about mutations, they think of changing one DNA base letter into another letter,” says Rameen Beroukhim of the Dana-Farber Cancer Institute in Boston. “But a primary way that cancer becomes cancer is to add copies of genes that it likes”—in other words, those that help promote malignancy—“and delete copies of genes that it doesn’t like.”

In fact, 23 of the 25 most frequent genetic changes in cancer involve structural changes of whole chromosome arms. During those changes, sections of a chromosome can break away, adding or eliminating one or more copies of hundreds or thousands of genes all at once.

For a February 2020 paper published in Nature as part of the PCAWG Consortium, an international team led by Beroukhim analyzed nearly 2,700 whole cancer genomes in the largest study to date of genomic rearrangements. The researchers identified 16 structural variant patterns that play a role in many cancers.

These changes, Beroukhim says, “can screw up the biology of a cell in ways we still can’t understand.” But solving those mysteries could hold enormous potential for new treatment approaches. “If we can understand structural rearrangements, the therapeutic possibilities are very large and pretty much untapped,” he says.


ANOTHER RECENT INSIGHT IS that driver mutations work in remarkably specific ways depending on where in the body they occur. Although a handful operate in similar ways across multiple types of cancers, they are the exception, not the rule, says Stephen Elledge, a geneticist at Brigham and Women’s Hospital in Boston.

In research published in March 2018 in Cell, scientists in Elledge’s lab ran experiments on three types of cells—breast, pancreatic and connective-tissue cells called fibroblasts—and were startled by how differently each tissue responded to the same genes that regulate cell proliferation. Genes that drove proliferation in one kind of tissue had no effect in another and actually suppressed proliferation in still another.

How can the same DNA code be translated so differently in different locations? Elledge’s team had the idea that such tissue-specific responses to cancer mutations might be largely the result of the “epigenetic” landscape—the array of chemical markers that attach to DNA and alter how its code is read, and the distinct chemical environments different types of cells have, which affect how their genes operate. Epigenetic differences can turn cancer genes on or off, for example.

The epigenetic state of a cancer cell is also likely to influence how particular tissue types respond to therapies and how they may evolve resistance. Inhibiting the gene RAF, for example, is effective in slowing down melanomas. But it has little impact on colorectal cancer in which the same mutation plays a role. That’s because tumor cells in the colon express a tissue-specific “growth factor” protein that’s not present in skin cells, and that protein helps tumor cells in the colon survive the treatment. Better understanding of such differences could help researchers find tissue-specific vulnerabilities that could be targeted for true precision treatments.

Understanding in much finer detail how cancer develops—including a more precise knowledge of driver mutations and the events that cause them, and a better grasp of how the same mutation might operate in different parts of the body—could lead to new ways to intercede and stop its growth. “As sequencing costs keep decreasing, one can imagine a day when every patient will have their tumor genome sequenced as a standard step,” says MGH’s Getz. “This information could flow into diagnosis and early detection, and could help identify potential vulnerabilities of newly discovered cancers as well as those already being treated.”