The Pill That Could · The Compound

The Compound, part 3

This was the assessment of CGP 57148 inside Ciba-Geigy in early 1993. It was the kind of compound that made a useful chemical biology paper. It was not the kind of compound that any normal pharmaceutical project would advance toward the clinic.

* * *

Brian Druker, in late 1992, was a few months from leaving the Dana-Farber Cancer Institute to take a faculty position at Oregon Health Sciences University. He had been corresponding with Nick Lydon for some time — since the late 1980s, intermittently, on the strength of overlapping interests in tyrosine kinases. Lydon had sent Druker compounds before; some of the early Ciba-Geigy molecules had been useful in Druker's antibody work, as research tools to manipulate kinase activity in cells. The relationship was collegial and informal. Lydon was a chemist working on inhibitors; Druker was a clinician-scientist with the antibodies needed to read out what those inhibitors did to cells. They were natural collaborators.

Druker, by 1993, knew about the 2-phenylaminopyrimidine series. He knew, in general terms, that some of those compounds inhibited ABL. He had not yet had access to the more recent compounds, the ones that were more selective; those were, internally at Ciba-Geigy, being held closely while the chemistry was still being optimized. But the trajectory was visible. He wrote to Lydon, around the time he was starting at Oregon, and asked whether he could test a specific compound — the most ABL-selective one Lydon had — against BCR-ABL in CML cells.

Lydon agreed. Internally at Ciba-Geigy, sending compounds to outside investigators required navigating some bureaucracy — you had to make sure they wouldn't be used in ways the company didn't approve of, that intellectual property was protected, that the compound wasn't being shared with competitors. Lydon worked through it. By 1993 or 1994, Druker had a vial of CGP 57148 in his lab in Oregon.

His test plan was straightforward, and it was straightforward because he had spent his postdoctoral years developing exactly the right tools. He had antibodies that could detect tyrosine phosphorylation in cells. He could expose CML cells to the compound, harvest them, and ask whether BCR-ABL was still active inside the treated cells — whether the kinase, exposed to the compound, was still adding phosphates to its substrates the way it normally would.

It wasn't. The treated CML cells, on the antibody readout, looked like cells with no BCR-ABL activity at all. The compound was reaching its target, fitting into the active site of the kinase, and shutting it down. This was, in itself, encouraging but not yet decisive; many compounds inhibit kinases in cells without doing anything useful for the cells' fate.

The decisive experiment was the colony assay we opened the chapter with. CML cells, plated in soft agar with the compound, didn't form colonies. CML cells without the compound did. Normal bone marrow cells, with or without the compound, formed colonies at about the same rate. The drug was killing CML cells specifically. It was not killing normal cells in any meaningful way.

Druker did the experiment more than once. He did it with cells from multiple CML patients. He did it at multiple drug concentrations. He did it with normal cells from multiple donors. Every time, the same result: the compound selectively killed BCR-ABL-positive cells while sparing the BCR-ABL-negative cells alongside them.

* * *

The reason this is an unusually clean result, the reason Druker spent his evenings in the OHSU laboratory in 1995 counting colonies several times to make sure he had not mis-pipetted somewhere, is that this is what a targeted cancer drug is supposed to look like, and almost no cancer drug had ever looked like this before.

The standard cytotoxic chemotherapies — the drugs that had been the backbone of cancer treatment for decades — work by killing rapidly dividing cells. They damage DNA, or interfere with cell division, or block enzymes that proliferating cells need disproportionately. Cancer cells, dividing more than most cells, are caught preferentially. But normal cells that also divide — the cells of the bone marrow, the lining of the gut, the hair follicles — are also caught. This is why chemotherapy makes patients lose their hair, get nauseous, and get dangerously low blood counts. The drugs aren't selective. They simply happen to hurt cancer slightly more than they hurt the patient.

A drug that kills CML cells while leaving normal bone marrow cells alone is doing something else entirely. It is not exploiting a generic feature of dividing cells. It is exploiting a specific molecular difference — the presence of BCR-ABL — between the cancer cells and the normal cells. Without BCR-ABL, the drug has nothing to do, which is why the normal cells don't notice it. With BCR-ABL, the drug shuts down the one enzyme the cancer cells absolutely depend on, and the cancer cells die.

This is what people had meant, all those years, by the phrase ‘targeted therapy.’ The phrase was older than the achievement; the dream of a magic bullet that picked out diseased cells from healthy ones, leaving the second alone, went back to Paul Ehrlich and the early twentieth century. The dream had survived, mostly intact and mostly unrealized, through ninety years of cancer drug discovery. There were a few partial cases — antibodies against specific cell-surface markers, certain hormones — but not really, in 1995, an oral small molecule that did exactly what targeted therapy was supposed to do. The colonies in Druker's dishes were, in retrospect, the first clear demonstration in human cells that this was possible.

This was not yet obvious. Druker himself was cautious. The cells were one piece of evidence; you would also need to show that the drug worked in animals, that it could be given safely, that it would do the same thing in a body that it did in a dish. There was a great deal of work between this colony assay and a treatment for a real patient.

But the result, in the lab, was unmistakable. He had a compound that killed CML cells and spared normal cells. It came from a chemist who could send him more of it. The molecule existed, in a freezer, in Oregon and in Switzerland.

* * *

There was now a problem that none of the science had prepared anyone to solve.

Druker wanted to take the compound forward. He wanted to test it in animals, then in patients. He wanted Ciba-Geigy, which owned the molecule, to develop it as a drug — to commit the resources required to manufacture it at scale, run the toxicity studies that regulators required, write the regulatory documents, sponsor the clinical trials. None of this was something an academic clinician at OHSU could do alone. The molecule belonged to the company. The decision to develop it was the company's.

And the company, in 1995, was not enthusiastic.

From inside Ciba-Geigy, the case for development looked weak. CML was a disease of perhaps five thousand new patients a year in the United States, similar numbers in Europe. The market, in pharmaceutical terms, was small. The compound's class — selective kinase inhibitors — had no approved precedent and a long history of disappointing expectations. The compound itself had real chemistry issues that would need to be solved before it could be made at scale. There were fears that any clinical benefit might be brief, with cancers evolving resistance the way they did to other targeted approaches that had been tried. There were fears that the toxicology would surprise everyone unpleasantly. There were fears about a more general principle: a class of drugs that took years to develop and might be quickly outpaced by resistance was not the kind of bet a publicly-traded pharmaceutical company was set up to make. There were, internally, more reasons to walk away than to keep going.

From inside the small group that wanted to keep going — Lydon, Matter, and a few allies — the case looked different. The compound did exactly what a targeted therapy was supposed to do. It came out of a careful, principled research program that had identified a specific target and a specific molecule with all the right properties. Druker's data in CML cells were as clean as anything anyone had seen. CML, while uncommon, was uniformly fatal under the existing standard of care; the unmet need was real and obvious. And if the approach worked here, it would set a template for a generation of similar drugs. Letting CGP 57148 sit on a shelf because the market was small was, in this view, a profound failure of strategic imagination.

Both views were defensible. Both views were held by people who were trying, in their respective frames, to do good work. The disagreement was not a disagreement between heroes and villains. It was a disagreement between people who saw an unusual opportunity and people whose job it was to make sure their company didn't pursue every unusual opportunity it saw, because most of them, on examination, turn out not to be the opportunity they looked like.

The next chapter is about how that disagreement got resolved.

It took five years.

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