The Pill That Could

Chapter 9: Resistance

Charles Sawyers had been worried about resistance from the beginning.

He had run his half of the phase 1 trial at UCLA, alongside Druker in Oregon and Talpaz in Houston, and he had seen what the rest of the team had seen — the unprecedented response rates, the cytogenetic clearances, the patients who came back for follow-up looking healthier than they had in years. He shared the team's optimism about what the drug was doing. He also kept asking, in his own work and in conversations with his collaborators, what would happen when it stopped.

He asked because he was a cancer biologist as well as a clinician, and cancer biologists have an instinct, born of watching what cancer does, that any treatment that works will eventually fail. The reason is not pessimism. The reason is evolution. A tumor in a patient's body contains, at the time of diagnosis, somewhere between a hundred million and a hundred billion cancer cells, depending on the cancer and the stage. Each of those cells is a slightly imperfect copy of its parent. Each of them carries, in its genome, a small number of new mutations that the parent didn't have. Most of those mutations are silent or harmful or irrelevant to anything. But across a population of a hundred million cells, every single position in the genome has, statistically, been mutated in at least a few cells. The diversity is staggering. It is also, from the perspective of any drug aimed at a specific molecular target, an opportunity. If even one cell in the tumor has, by accident, a mutation that prevents the drug from binding its target, that one cell will survive treatment, while every cell without the mutation dies. The surviving cell will then divide, and its descendants will be a tumor again, this time made entirely of cells the drug can't touch.

This is, in the broadest sense, how cancers become resistant to treatment. It is also how bacteria become resistant to antibiotics, how viruses become resistant to antivirals, how insects become resistant to insecticides. The same Darwinian logic operates anywhere a population of replicating things meets a selection pressure that kills most of them. The few survivors are the ones that, for whatever reason, were already different from the rest. Their descendants inherit the difference. The population that grows back is, on average, less susceptible to the killing pressure than the population that was there before.

Sawyers, in 2000, was one of the people thinking hardest about whether this would happen with imatinib.

* * *

There was a clinical signal, even in the early phase 1 data, that pointed at where to look first.

Patients in chronic-phase CML — the early, slow form of the disease — responded to imatinib durably. They went into hematologic and cytogenetic remission, and most of them stayed there for as long as they kept taking the drug. Some of them, we will see in the next chapter, are still in remission in 2026, on the same daily dose they started in 1999 or 2001.

Patients in blast crisis — the late, transformed form of the disease — responded to imatinib too. The response rates were lower than in chronic phase, but they were not zero; many blast-crisis patients had clear and meaningful responses, with disappearance of leukemic blasts from the bone marrow and improvement in their clinical condition. The trouble was that the responses didn't last. A blast-crisis patient who responded beautifully in the first few months would, very often, relapse within a year. The leukemia that grew back would be, by every measure the trial could apply, less responsive to the drug than the leukemia that had been there to start with. Sometimes substantially less responsive. Sometimes essentially completely refractory.

This pattern — dramatic initial response in advanced disease, followed by relapse — is a classic signature of the kind of evolutionary process this chapter opened with. A treatment that works on most cells but not all of them produces a brief remission, after which the cells that didn't respond grow into a tumor that the treatment can't control. The pattern was familiar in oncology from earlier targeted approaches — from the way some leukemia cells survive cytotoxic chemotherapy, from the way HIV adapts to antiretroviral drugs given as monotherapy. It was the pattern Sawyers had been worried about.

What he wanted to know, by 2000, was what was happening at the molecular level in the relapsed cells. If you took blast cells from a patient who had relapsed on imatinib, and asked what was different about them compared to the cells that had originally responded, what would you find?

* * *

The simplest possibility, in principle, is that the cells had stopped depending on BCR-ABL. Cancer can do this. A cell driven by one signaling pathway can sometimes, given time and selection pressure, learn to be driven by a different pathway instead. The kinase that was running the show is bypassed, and the cell continues to grow on the strength of some other signal. Drugs that targeted the original kinase would, in this case, not work, because the kinase no longer mattered. The cancer had moved on.

There was a second possibility, more specific and in some ways more interesting. The cells could still be depending on BCR-ABL but could have changed BCR-ABL itself in a way that prevented the drug from binding. Imatinib's mechanism, as chapter four laid out, is to fit into the ATP pocket of the kinase and block it. The fit depends on the precise three-dimensional geometry of the pocket. A change in even one amino acid, in the right place, could disrupt the fit — not enough to disable the kinase, just enough to make it stop accepting the drug.

All 13 chapters of The Pill That Could · Library · All courses