In early 2014, Laura Brealey was visiting her daughters in Singapore when she slipped on a marble floor and cracked her hip. She had it replaced, but in the process, the surgeons noticed that her breathing sounded odd, and told her to speak to a respiratory specialist. At her daughtJer’s urging, she did so when she returned home to London, England—and was told that she had lung cancer.

The doctors moved quickly. That same August, they cut out part of her right lung, and gave her both radiotherapy and chemotherapy. The tumor disappeared. Things were looking promising. But the following summer, Brealey started experiencing fresh pain in her ribs. The cancer was back.

When I meet Brealey in April 2016, in a brightly colored room at University College Hospital, she’s about to start her third drug—a new one called nivolumab that her doctor, a young oncologist named Charlie Swanton, is hopeful about. As he preps her for the new treatment, he is also thinking about how to manage the severe pain that the tumor is causing, as it presses against the nerves that run along Brealey’s ribs. “It hasn’t been easy for Laura,” says Swanton. “When I first told her she couldn’t go back to Singapore to visit her family...”

“... I nearly fell off my chair in shock,” she finishes. Brealey, now 75, has bright blue eyes, short white hair, and preternaturally high spirits. “My best friend’s brother calls me the Indestructible Laura. I’m very active. Before this, I was going to the gym, doing aqua-aerobics, pilates, gardening. Now, I just sleep. This has been a nuisance.” She misses the theater, longs to return to Asia, dreams about going on the Trans-Siberian Railway. When she mentions the tumor that is curtailing all of those plans, she talks as if it was an errant child. “It’s so stubborn,” she says.  

Brealey’s story is typical. An initial wave of treatments seemed to knock a tumor out, but despite the scalpels, the radiation, and the drugs—not to mention relentless optimism—the tumor recurs, seemingly more resilient than before. Stubborn, as she says. Even the latest drugs, tailored to strike at the specific genetic faults behind a person’s tumor, fail to permanently halt the backswing of a tumor’s pendulum.

The problem, according to Swanton and a growing group of other like-minded scientists, is that we’ve been neglecting one of the most fundamental aspects of cancer—that it is a disease of evolution. That might sound strange. We’re used to the idea that animals, plants, or microbes can evolve, but cancer isn’t a free-living organism. It’s part of us. And yet, it also evolves, adapting to new challenges as surely as Galapagos finches.

In the classical view of cancer, a cell picks up mutations until it shakes off the checks and balances that restrain its growth, allowing it to divide uncontrollably and turn into a tumor. This linear process is a macabre version of that famous image where a chimp walks to the right and gradually morphs into a human hunter. And both visuals are wrong. In reality, tumors quickly become seething masses of varied cells, all with their own mutations. One area might start growing faster; its neighbor might come to evade the immune system. Over time, the fittest lineages produce more descendants and rise to dominance—the essence of Darwinian natural selection. So forget the linear march. The better visual is that of a tree, with an initial trunk radiating into a web of branches. In 1837, Charles Darwin drew one such tree in one of his notebooks to represent how species evolve from a common ancestor. He could just as easily have been sketching the birth of a tumor.

This realization goes some way to explaining why the war against cancer has been so entrenched and unexpectedly difficult. Clinicians often diagnose these diseases by taking a biopsy from a tumor, but a single sample could miss important mutations with very different prognostic implications just centimeters away. And when we hit tumors with drugs or radiation, we create a potent source of artificial selection, effectively breeding for hardier tumors. That’s why relapses occur. “We deal with patients, day in and day out, whose disease initially benefits from chemo,” says Swanton. “But over time, they develop resistance against multiple drugs. The only way you can really fathom how a tumor can be one step ahead of the clinician at all times is through Darwinian selection.”

The legendary biologist Theodosius Dobzhansky said that “Nothing in biology makes sense except in the light of evolution.” Cancer researchers are now learning that lesson. By understanding how cancer evolves, they hope to start thinking several moves ahead of the disease—halting, redirecting, or even preventing its evolution.

To start, in 2013, Swanton launched an ambitious study called TRACERx, which stands for Tracking Cancer Evolution through Treatment (Rx). It’s a multi-million-dollar initiative to nail down the evolution of lung cancer. Swanton’s team of over 200 collaborators followed 850 patients through their treatments, taking several samples from across their tumors at several points in time—from diagnosis to treatment to either cure or relapse. They then used this information to understand how the cancers change over time, how that affects a patient’s fate, how to detect that evolution, and how to beat it.

Brealey, whose name has been changed for this story, was the 39th patient to sign up for TRACERx—and when I meet her, she is dressed for the part. She holds up the pale stone necklace around her neck, which depicts a bare-branched tree. “Look, I’m wearing the tree of life,” she says.

“How appropriate!” says Swanton, delighted.

“Maybe it’s a good omen,” she adds.

* * *

The idea of cancer as an evolving entity isn’t new. In 1976, cancer researcher Peter Nowell published an article called The Clonal Evolution of Tumor-Cell Populations in the prestigious journal Science. It was a prescient essay, outlining much of what researchers would later confirm. “More research should be directed towards understanding and controlling the evolutionary process in tumors before it reaches the late stage seen in clinical cancer,” he wrote.

His suggestions went largely unheeded because he made them at the wrong time. In the 1970s, molecular biology was the hot new thing, and scientists were swept up by new techniques for studying cells in incredible detail. Many biology departments split into separate factions, walling off those who studied tiny cells and molecules from those who studied entire organisms. Cancer researchers clustered with the former, and evolutionary biologists with the latter.

Mel Greaves was an exception. Now the director of the Center for Evolution at The Institute of Cancer Research, Greaves studied at University College London in the 1960s, in a department that focused heavily on the principles of evolution. His mentors drummed into his head that cells within a body are also subject to natural selection, just like individuals within a species. So when he read Novell’s paper, it blew his mind. “I thought: This is fundamental. This is how cancer works. It just smelled right. It made me think about everything I was doing in that light,” he says.

Greaves carried that realization with him throughout his work on childhood leukemia. But he could only act upon it at the turn of the century—when new technology allowed him to efficiently sequence the genomes of individual cancer cells. In 2010, Greaves and his team used to tools to study 30 leukemia patients. In each one, they identified the mutations within 200 cancerous cells, and compared them to see how the cells were related. “Immediately, we saw trees,” he says. “You could find the simplest cell and work out its descendants and their descendants. Within two years, everyone was doing this.”

For example, in 2012, Swanton and his colleagues found a similar metaphorical tree in a fist-sized tumor that they had cut out of a kidney-cancer patient. Even by eye, the variation was obvious. “It was a large lump, of course, but you could see individual nodules that had different colors,” recalls team member Marco Gerlinger. “Some bled more. Some had this yellowish translucent appearance because they were storing fats. Others were brown and very dense because they lacked blood vessels.”

In analyzing the tumor, Gerlinger and Swanton showed that this cancer was a world unto itself. Cells at one end had a different set of mutations from those at the other, or from secondary tumors that had spread around the patient’s body. Of the 128 mutations that Swanton had detected, only a third of these were found throughout the patient’s tumors. A quarter were unique to just one part. The same was true for three more kidney cancers that the team studied. Others found the same pattern in cancers of the esophagus, bowel, brain, breast, and more. “Every patient has several essentially independent cancers at one time,” says Greaves. “Three to 20 in leukemia. Goodness knows how many in lung cancer.”

Propelled by these discoveries, cancer research started evolving, too. The biggest cancer conferences were devoting keynote talks and entire sessions to the evolutionary side of cancer. When I went to one such session in 2014, a researcher joked to a huge, packed auditorium that “we used to be in the small room.” But this flurry of activity was tempered. “The reaction from oncologists was: Oh my God, this is bloody awful,” says Greaves. “Look at all this complexity! It’s so pessimistic. And all I, Charlie, and others could say was: This is just what cancer is. We haven’t understood the nature of the beast until now.”

“I don’t feel daunted,” he adds, smiling. “Now we know why the therapies are failing, why we’re having such difficulty in eradicating the disease. It’s just like drug resistance in bacteria—a rapidly evolving disease that we’re racing to defeat. It’s an evolutionary arms race.” That insight, he says, is half the battle. The next half is working out what to do about it.

* * *

The first generations of cancer drugs were designed to hit anything that grew quickly. But in the years since the human genome was completely sequenced, researchers have developed more precise weapons. They looked for genes whose mutated forms can fuel the development of cancer, and designed drugs to hit those genes. These “targeted therapies” were meant to usher in a new age of personalized medicine, but “although people have spectacular responses, the drugs are not cures,” says Udai Banerji, a medical oncologist from the Institute of Cancer Research. With rare exceptions, “they work well for a while and then sometime between nine months and two years, people become resistant.”

The diversity that Greaves, Swanton, and others have found explains why. A drug might successfully take out the vast majority of a tumor’s cells, but even if a few survivors contain mutations that allow them to resist these assaults, they will repopulate. The cancer evolves its way around the drug, and the tumor recurs. Worse, tumors are diverse, so targeted therapies only might hit one part of the entire mass, leaving behind a reservoir of genetically different cells to grow anew. Targeted therapies are powerful, but cancer is a target with many bullseyes that shift position whenever struck.

“This realization prompted me to change the way I think,” says Alberto Bardelli, an oncologist at the University of Torino, who studies targeted therapies. Rather than going after specific cancer genes, “we need to target the evolution of tumors—how they change over time. Then we’ll make progress much more quickly.”

For a start, reconstructing the evolution of cancer can help researchers to pick the right targets. Some mutations occur early in a tumor’s life and are found in all of its cells—they’re called trunk mutations. Others occur later are found in small regions—branch mutations. “If you’re going to target mutations, for goodness sake go for the trunk ones,” says Greaves. “Otherwise it’s like pruning a bush—you’ll just stimulate growth.” It’s no coincidence that Gleevec, one of the most successful target therapies around, works by targeting a trunk mutation in leukemia—one that almost always occurs at the very start of the disease. “It hits every cell, which is why it works pretty well,” says Greaves.

Other scientists have started to identify common trunk mutations in other cancers. Attacking these is certainly a better strategy than ineffectually hacking at the branches, but it doesn’t change the fact that tumors could adapt and regrow. A better tactic might be to use combinations of targeted drugs—one to target mutations that already exist, and others to attack those that are likely to occur. Many cancers involve mutations in a gene called EGFR, and a drug called cetuximab was designed to block this gene. But tumors that initially shrink in the face of cetuximab invariably bounce back—many because of mutations in a second gene called MEK. By using cetuximab and a MEK-blocking drug, Bardelli managed to successfully treat bowel cancers in mice, without any relapse. He had cut the tumors off from their usual evolutionary escape routes, intercepting resistance rather than simply reacting to it. As Robert Gatenby from the Lee Moffit Cancer Center and his colleagues once put it, “It is chess, not whack-a-mole.”

But Banerji thinks this approach is unlikely to work in practice because the side effect of chemotherapies become too toxic when drugs are given in combination. You might kill a tumor—but you’ll take the patient along with it. “We’ve looked at multiple ways of doing it and exhausted all the combinations,” he says. “It’s not going to be game-changing.” Their opponent is nothing less than evolution itself—the most dangerous foe of all, with a counter for every move. As the British chemist Leslie Orgel once said, “Evolution is cleverer than you are.” So rather than opposing the inevitable evolution of cancer, it might be better to redirect it.

“Everything comes at a cost in evolution,” says Banerji’s colleague Andrea Sottoriva. “By standing up on two feet, we free up our hands but we can’t outrun a lion.” In the same way, a mutation that allows cancer cells to resist a drug might also, say, slow their growth. This means that resistant cells only do better than their sensitive cousins in the presence of a drug—under normal conditions, they’re less competitive. Doctors could potentially exploit this weakness by giving treatments in waves. Hit a tumor with a drug, and the sensitive cells die off while the resistant ones start to grow. Stop the drug, to let the sensitive cells bounce back. Rinse and repeat. The goal of this “adaptive therapy” isn’t to destroy the tumor, but to keep it in balance. Gatenby tried this approach on breast cancers in mice. By varying doses, he managed to control the growth of the animals’ tumors, despite frequently skipping treatments for weeks at a time.

It might even be possible to lure cancers into traps. When tumors adapt to drugs, it’s usually a small number of cells that develop resistance and re-seeds the cancer. This means that for a time, the tumor passes through a bottleneck, where it turns from a large, genetically diverse mass into a small ball of identical clones. If those cells are all susceptible to a second drug, and if researchers can predict that susceptibility, they could use one drug to set the tumor up for a knock-out blow. Banerji and Sottoriva are testing this strategy, dosing cancer cells with sequential rounds of different drugs, to see if they can turn a diverse tumor into one that’s entirely vulnerable to the same drug. “If we can cure, that would be fantastic,” says Sottoriva. “But if we can make cancer a chronic manageable disease, as happened with HIV, I’ll take it. If we can tell a person who’s 70 that we can delay their disease by 30 years, it’ll be absolutely amazing.”

Other scientists are going after the evolutionary process itself. Swanton has shown that some kidney tumors have trunk mutations in a gene called SETD2 that helps to repair DNA damage. “Genes like this maintain the integrity of the genome,” he explains. Mutating them doesn’t lead to cancer directly, but it does nix a cell’s ability to repair errors in its own DNA and ensures that further mutations accrue more quickly. In other words, these mutations make cells more evolvable. By targeting them, it might be possible to slow the evolution of cancers altogether.

Even simple substances like aspirin might do the trick. This drug is known to reduce the risk of bowel, esophageal, and pancreatic cancers. Partly, this might be because it puts the brakes on a tumor’s early evolution. In a study of people with Barrett’s esophagus—a condition that often leads to esophageal cancer—Carlo Maley from Arizona State University found that mutations arise 13 times faster when people don’t take aspirin than when they swallow the drug.

“The study was small It was just 13 patients, and ideally, I’d want to confirm it in a large cohort of people, half of whom are randomly given aspirin,” Maley admits. But he adds that the broader idea of averting evolution before it has a chance to get going—and prevent cancers before drug resistance is even an issue—is the right one. “We’re not trying to beat evolution but to manage it, which sounds plausible to me,” he says. “The field is still dominated by cancer therapy but we should focus on cancer prevention. We should intervene before it becomes such a hard problem.”

* * *

Laura Brealey passed away last November. “She was incredibly stoic, and was always one of TRACERx’s most avid supporters. We all miss her a great deal,” says Swanton. “It’s still a ghastly disease. Despite all this research, we’re not chipping away at survival as much as I’d like.” As a human being, every loss tears at Swanton, but he remains hopeful that people like Brealey will provide the information that he and his colleagues need to make better progress against their ever-evolving adversary.

The first results from their TRACERx study are finally published today in two companion papers. In the first, the team focused on the first 100 patients—Brealey included—of the 842 they have since recruited. Again, they found that each tumor is a world in itself. In total, one-third of the mutations they found were in the branches of the tumor, restricted only to particular parts of the mass. Some patients had just a handful of these branches. One had 2,310.

These mutations are like single typos in a book, but the team found evidence of larger chromosomal rearrangements—the equivalent of entire paragraphs being duplicated, deleted, or rearranged. On average, these large-scale changes affected half of a tumor’s genome, and patients with high levels were five times more likely to experience recurrence than those with low levels. “We hadn’t anticipated that, in just the first 100 patients, we’d identify something that’s clinically relevant,” says Mariam Jamal-Hanjani, a medical oncologist who has worked on TRACERx for years. “The more chaotic your tumor’s chromosomes are, the more likely it is that your cancer comes back.”

“It’s an incredibly complex disease,” says Swanton. “It has a lot more instability than what we saw in kidney cancer.” Worse still, this instability can rewire a lung cancer’s evolutionary history, altering the trunk mutations that ought to be present and consistent throughout a tumor. In other words, as lung cancer evolves, the trunk can turn into branches, which means that going after the trunk mutations is unlikely to be lung cancer’s downfall.

Still, Swanton sees reason for hope. His team showed that many of these large-scale changes are caused by a family of gene-editing enzymes called APOBEC. These are usually involved in immune responses, but seems to go rogue in lung cancer—and several other types—providing what Swanton calls “mutagenic fuel for cancer evolution.” Even in heavy smokers, APOBEC enzymes can create even more mutations than the potent DNA-damaging chemicals in tobacco smoke. “There are multiple efforts going on around the globe to target the enzyme and arrest evolution in its tracks,” says Swanton. “I’m really excited about this.”

Such efforts may take a long time to bear fruit, but the acceptance of cancer’s evolutionary side is already changing how clinicians approach the disease.

Doctors typically take a single biopsy of a patient’s tumor at the time of diagnosis, and use that sample to make decisions about treatments, even if the tumor recurs. That’s like using last decade’s stock information to guide tomorrow’s investments. “We’ve accepted that profiling those archived tumors could be doing the patients a disservice,” says Samra Turajlic, one of Swanton’s colleagues. “We now try to get the most recent and representative verison of the patient’s disease.

But biopsies won’t cut it—it’s impractical to keep on slicing chunks out of a patient’s tumor over time, and as Swanton’s work has shown, you need several such slices from different regions to really capture a tumor’s diversity. Fortunately, there’s another option.

When cancer cells die, they release fragments of their DNA into their host’s bloodstream. This circulating tumor DNA (ctDNA) acts as a real-time autobiography of a cancer—a “liquid biopsy” that gives a snapshot of a tumor in its entirety, without the need for invasive slicing. In dozens of studies across many types of cancer, researchers have shown that these free-floating phantoms of dead cancer cells can reveal how big tumors are, whether they are threatening to recur, and whether they’ve evolved resistance.

In the second of the two TRACERx papers, Swanton’s team showed that ctDNA provides an edifying look at lung cancer, too. In some cases, patients showed increasing levels of ctDNA immediately after surgery, showing that their tumors were bouncing back despite the chemotherapy that they were taking. In other cases, the free-floating DNA revealed the presence of new branch mutations, suggesting that tumors were relapsing several months before they became visible on CT scans. It could act as a marker that reveals how cancers evolve in real-time, and it could reveal the efficacy of measures designed to curb that evolution.

As ever, there are caveats. Modern techniques for sequencing ctDNA are still too expensive to be used regularly in the clinic, and they aren’t sensitive enough to detect cancers at their earliest stages. It’s also still unclear exactly how representative the floating DNA is of tumors as a whole. But Swanton thinks that these hurdles can be overcome. “Ten years from now, I’d like to see us doing this regularly—using circulating DNA to constantly monitor disease over space and time, and alter our therapies accordingly,” he says.

Five years ago, his big kidney cancer left him pessimistic and unsettled. But like the diseases he studies, he too has rebounded. “I was pretty depressed when we first started on this route, but I’ve come out the other side,” he says. “And I’ve been getting pretty optimistic in the last couple of years. By understanding and embracing the complexity, we can fight it.”