by Mary Roach
"In the early eighties, I looked into how Merck and Pfizer went about drug discovery," Chen recalls. "How many compounds are they using? Are they doing the best they can? And I come up with an incredible number. It turns out that mankind had, at this point, made tens of millions of compounds. But Pfizer was screening only six hundred thousand compounds, and Merck even fewer, about five hundred thousand. How could they screen for drugs and use only five hundred thousand, when mankind has already made so many more?"
An early financial backer of Chen's was Michael Milken, the junk-bond king of the 1980s, who, after being treated for prostate cancer, became a major cancer philanthropist. "I told Milken my story," Chen said, "and very quickly he said, 'I'm going to give you four million dollars. Do whatever you want.' Right away Milken thought of Russia. Someone had told him that the Russians had had, for a long time, thousands of chemists in one city making compounds, and none of those compounds had been disclosed." Chen's first purchase was a batch of 22,000 chemicals, gathered from all over Russia and Ukraine. They cost about ten dollars each and came in tiny glass vials. With his money from Milken, Chen then bought a $600,000 state-of-the-art drug-screening machine. It was a big, automated Rube Goldberg contraption that could test ninety-six compounds at a time and do a hundred batches a day. A robotic arm would deposit a few drops of each chemical onto a plate, followed by a clump of cancer cells and a touch of blue dye. The mixture was left to sit for a week and then reexamined. If the cells were still alive, they would show as blue. If the chemical killed the cancer cells, the fluid would be clear.
Chen's laboratory began by testing his compounds against prostate-cancer cells, since that was the disease Milken had. Later he screened dozens of other cancer cells as well. In the first round, his batch of chemicals killed everything in sight. But plenty of compounds, including pesticides and other sorts of industrial poisons, will kill cancer cells. The trouble is that they'll kill healthy cells as well. Chen was looking for something that was selective—that was more likely to kill malignant cells than normal cells. He was also interested in sensitivity—in a chemical's ability to kill at low concentrations. Chen reduced the amount of each chemical on the plate a thousandfold and tried again. Now just one chemical worked. He tried the same chemical on healthy cells. It left them alone. Chen lowered the dose another thousandfold. It still worked. The compound came from the National Taras Shevchenko University of Kiev. It was an odd little chemical, the laboratory equivalent of a jazz musician's riff. "It was pure chemist's joy," Chen said. "Homemade, random, and clearly made for no particular purpose. It was the only one that worked on everything we tried."
Mass screening wasn't as elegant or as efficient as rational drug design. But it provided a chance of stumbling across something by accident—something so novel and unexpected that no scientist would have dreamed it up. It provided for serendipity, and the history of drug discovery is full of stories of serendipity. Alexander Fleming was looking for something to fight bacteria but didn't think the answer would be provided by the mold that grew on a petri dish he accidentally left out on his bench. That's where penicillin came from. Pfizer was looking for a new heart treatment and realized that a drug candidate's unexpected side effect was more useful than its main effect. That's where Viagra came from. "The end of surprise would be the end of science," the historian Robert Friedel wrote in the 2001 essay "Serendipity Is No Accident." "To this extent, the scientist must constantly seek and hope for surprises." When Chen gathered chemical compounds from the farthest corners of the earth and tested them against one cancer-cell line after another, he was engineering surprise.
What he found was exactly what he'd hoped for when he started his hunt: something he could never have imagined on his own. When cancer cells came into contact with the chemical, they seemed to go into crisis mode: they acted as if they had been attacked with a blowtorch. The Ukrainian chemical, elesclomol, worked by gathering up copper from the bloodstream and bringing it into cells' mitochondria, sparking an electrochemical reaction. His focus was on the toxic, oxygen-based compounds in the cell called ROS, reactive oxygen species. Normal cells keep ROS in check. Many kinds of cancer cells, though, generate so much ROS that the cell's ability to keep functioning is stretched to the breaking point, and elesclomol cranked ROS up even further, to the point that the cancer cells went up in flames. Researchers had long known that heating up a cancer cell was a good way of killing it, and there had been plenty of interest over the years in achieving that effect with ROS. But the idea of using copper to set off an electrochemical reaction was so weird—and so unlike the way cancer drugs normally worked—that it's not an approach anyone would have tried by design. That was the serendipity. It took a bit of "chemist's joy," constructed for no particular reason by some bench scientists in Kiev, to show the way. Elesclomol was wondrously novel. "I fell in love," Chen said. "I can't explain it. I just did."
When Freireich went to Zubrod with his idea for VAMP, Zubrod could easily have said no. Drug protocols are typically tested in advance for safety in animal models. This one wasn't. Freireich freely admits that the whole idea of putting together poisonous drugs in such dosages was "insane," and, of course, the first patient in the trial had nearly been killed by the toxic regimen. If she had died from it, the whole trial could have been derailed.
The ALL success story provided a hopeful road map for a generation of cancer fighters. But it also came with a warning: those who pursued the unexpected had to live with unexpected consequences. This was not the elegance of rational drug design, where scientists perfect their strategy in the laboratory before moving into the clinic. Working from the treatment to the disease was an exercise in uncertainty and trial and error.
If you're trying to put together a combination of three or four drugs out of an available pool of dozens, how do you choose which to start with? The number of permutations is vast. And, once you've settled on a combination, how do you administer it? A child gets sick. You treat her. She goes into remission, and then she relapses. VAMP established that the best way to induce remission was to treat the child aggressively when she first showed up with leukemia. But do you treat during the remission as well, or only when the child relapses? And, if you treat during remission, do you treat as aggressively as you did during remission induction, or at a lower level? Do you use the same drugs in induction as you do in remission and as you do in relapse? How do you give the drugs, sequentially or in combination? At what dose? And how frequently—every day, or do you want to give the child's body a few days to recover between bouts of chemo?
Oncologists compared daily 6-MP plus daily methotrexate with daily 6-MP plus methotrexate every four days. They compared methotrexate followed by 6-MP, 6-MP followed by methotrexate, and both together. They compared prednisone followed by full doses of 6-MP, methotrexate, and a new drug, cyclophosphamide (CTX), with prednisone followed by half doses of 6-MP, methotrexate, and CTX. It was endless: vincristine plus prednisone and then methotrexate every four days or vincristine plus prednisone and then methotrexate daily? They tried new drugs and different combinations. They tweaked and refined and gradually pushed the cure rate from 40 percent to 85 percent. At St. Jude Children's Research Hospital in Memphis—which became a major center of ALL research—no fewer than sixteen clinical trials, enrolling 3,011 children, have been conducted in the past forty-eight years.
And this was just childhood leukemia. Beginning in the 1970s, Lawrence Einhorn, at Indiana University, pushed cure rates for testicular cancer above 80 percent with a regimen called BEP: three to four rounds of bleomycin, etoposide, and cisplatin. In the 1970s Vincent T. DeVita, at the NCI, came up with MOPP for advanced Hodgkin's disease: mustargen, oncovin, procarbazine, and prednisone. DeVita went on to develop a combination therapy for breast cancer called CMF—cyclophosphamide, methotrexate, and 5-fluorouracil. Each combination was a variation on the combination that came before it, tailored to its target through a series of iterations. The often-asked question "W
hen will we find a cure for cancer?" implies that there is some kind of master code behind the disease waiting to be cracked. But perhaps there isn't a master code. Perhaps there is only what can be uncovered, one step at a time, through trial and error.
When elesclomol emerged from the laboratory, then, all that was known about it was that it did something novel to cancer cells in the laboratory. Nobody had any idea what its best target was. So Synta gave elesclomol to an oncologist at Beth Israel in Boston, who began randomly testing it out on his patients in combination with paclitaxel, a standard chemotherapy drug. The addition of elesclomol seemed to shrink the tumor of someone with melanoma. A patient whose advanced ovarian cancer had failed multiple rounds of previous treatment had some response. There was dramatic activity against Kaposi's sarcoma. They could have gone on with Phase 1s indefinitely, of course. Chen wanted to combine elesclomol with radiation therapy, and another group at Synta would later lobby hard to study elesclomol's effects on acute myeloid leukemia (AML), the commonest form of adult leukemia. But they had to draw the line somewhere. Phase 2 would be lung cancer, soft-tissue sarcomas, and melanoma.
Now Synta had its targets. But with this round of testing came an even more difficult question. What's the best way to conduct a test of a drug you barely understand? To complicate matters further, melanoma, the disease that seemed to be the best of the three options, is among the most complicated of all cancers. Sometimes it confines itself to the surface of the skin. Sometimes it invades every organ in the body. Some kinds of melanoma have a mutation involving a gene called BRAF; others don't. Some late-stage melanoma tumors pump out high levels of an enzyme called LDH. Sometimes they pump out only low levels of LDH, and patients with low-LDH tumors lived so much longer that it was as if they had a different disease. Two patients could appear to have identical diagnoses, and then one would be dead in six months and the other would be fine. Tumors sometimes mysteriously disappeared. How did you conduct a drug trial with a disease like this?
It was entirely possible that elesclomol would work in low-LDH patients and not in high-LDH patients, or in high-LDH patients and not in low-LDH ones. It might work well against the melanoma that confined itself to the skin and not against the kind that invaded the liver and other secondary organs; it might work in the early stages of metastasis and not in the later stages. Then there was the prior-treatment question. Because tumors quickly become resistant to drugs, new treatments sometimes work better on "naive" patients—those who haven't been treated with other forms of chemotherapy. So elesclomol might work on chemo-naive patients and not on prior-chemo patients. And in any of these situations, elesclomol might work better or worse depending on which other drug or drugs it was combined with. There was no end to the possible combinations of patient populations and drugs that Synta could have explored.
At the same time, Synta had to make sure that whatever trial it ran was as big as possible. With a disease as variable as melanoma, there was always the risk in a small study that what you thought was a positive result was really a matter of spontaneous remissions, and that a negative result was just the bad luck of having patients with an unusually recalcitrant form of the disease. John Kirkwood, a melanoma specialist at the University of Pittsburgh, had done the math: in order to guard against some lucky or unlucky artifact, the treatment arm of a Phase 2 trial should have at least seventy patients.
Synta was faced with a dilemma. Given melanoma's variability, the company would ideally have done half a dozen or more versions of its Phase 2 trial: low-LDH, high-LDH, early-stage, late-stage, prior-chemo, chemo-naive, multidrug, single-drug. There was no way, though, that they could afford to do that many trials with seventy patients in each treatment arm. The American biotech industry is made up of lots of companies like Synta, because small start-ups are believed to be more innovative and adventurous than big pharmaceutical houses. But not even big firms can do multiple Phase 2 trials on a single disease—not when trials cost more than $100,000 per patient and not when, in pursuit of serendipity, they are simultaneously testing that same experimental drug on two or three other kinds of cancer. So Synta compromised. The company settled on one melanoma trial: fifty-three patients were given elesclomol plus paclitaxel, and twenty-eight, in the control group, were given paclitaxel alone, representing every sort of LDH level, stage of disease, and prior-treatment status. That's a long way from half a dozen trials of seventy each.
Synta then went to Phase 3: 651 chemo-naive patients, drawn from 150 hospitals in fifteen countries. The trial was dubbed SYMMETRY. It was funded by the pharmaceutical giant Glaxo Smith Kline. Glaxo agreed to underwrite the cost of the next round of clinical trials and—should the drug be approved by the Food and Drug Administration—to split the revenues with Synta.
But was this the perfect trial? Not really. In the Phase 2 trial, elesclomol had been mixed with an organic solvent called Cremophore and then spun around in a sonicator, which is like a mini washing machine. Elesclomol, which is rock-hard in its crystalline form, needed to be completely dissolved if it was going to work as a drug. For SYMMETRY, though, sonicators couldn't be used. "Many countries said that it would be difficult, and some hospitals even said, 'We don't allow sonication in the preparation room,'" Chen explained. "We got all kinds of unbelievable feedback. In the end we came up with something that, after mixing, you use your hand to shake it." Would hand shaking be a problem? No one knew.
Then a Synta chemist, Mitsunori Ono, figured out how to make a water-soluble version of elesclomol. When the head of Synta's chemistry team presented the results, he "sang a Japanese drinking song," Chen said, permitting himself a small smile at the eccentricities of the Japanese. "He was very happy." It was a great accomplishment. The water-soluble version could be given in higher doses. Should they stop SYMMETRY and start again with elesclomol 2.0? They couldn't. A new trial would cost many millions of dollars more and set the whole effort back two or three years. So they went ahead with a drug that didn't dissolve easily against a difficult target, with an assortment of patients who may or may not have been ideal—and crossed their fingers.
SYMMETRY began in late 2007. It was a double-blind, randomized trial. No one had any idea who was getting elesclomol and who wasn't, and no one would have any idea how well the patients on elesclomol were doing until the trial data were unblinded. Day-to-day management of the study was shared with a third-party contractor. The trial itself was supervised by an outside group, known as a data-monitoring committee. "We send them all the data in some database format, and they plug that into their software package, and then they type in the code and press 'Enter,'" Bahcall said. "And then this line"—he pointed at the Kaplan-Meier in front of him—"will, hopefully, separate into two lines. They will find out in thirty seconds. It's, literally, those guys press a button and for the next five years, ten years, the life of the drug, that's really the only bit of evidence that matters." It was January 2009, and the last of the 651 patients were scheduled to be enrolled in the trial in the next few weeks. According to protocol, when the results began to come in, the data-monitoring committee would call Jacobson, and Jacobson would call Bahcall. "ASCO starts May 29," Bahcall said. "If we get our data by early May, we could present at ASCO this year."
In the course of the SYMMETRY trial, Bahcall's dining-room-table talks grew more reflective. He drew Kaplan-Meiers on the back of napkins. He talked about the twists and turns that other biotech companies had encountered on the road to the marketplace. He told wry stories about Lan Bo Chen, the Jewish mother and Jewish father rolled into one—and, over and over, he brought up the name of Judah Folkman. Folkman died in 2008, and he was a legend. He was the father of angiogenesis—a wholly new way of attacking cancer tumors. Avastin, the drug that everyone cheered at ASCO seven years ago, was the result of Folkman's work.
Folkman's great breakthrough had come while he was working with mouse melanoma cells at the National Naval Medical Center: when the tumors couldn't set up a network of blood vess
els to feed themselves, they would stop growing. Folkman realized that the body must have its own system for promoting and halting blood-vessel formation, and that if he could find a substance that prevented vessels from being formed he would have a potentially powerful cancer drug. One of the researchers in Folkman's laboratory, Michael O'Reilly, found what seemed to be a potent inhibitor: angiostatin. O'Reilly then assembled a group of mice with an aggressive lung cancer and treated half with a saline solution and half with angiostatin. In the book Dr. Folkman's War (2001), Robert Cooke describes the climactic moment when the results of the experiment came in:
With a horde of excited researchers jampacked into a small laboratory room, Folkman euthanized all fifteen mice, then began handing them one by one to O'Reilly to dissect. O'Reilly took the first mouse, made an incision in its chest, and removed the lung. The organ was overwhelmed by cancer. Folkman checked a notebook to see which group the mouse had been in. It was one of those that had gotten only saline. O'Reilly cut into the next mouse and removed its lung. It was perfect. What treatment had it gotten? The notebook revealed it was angiostatin.