Next-generation cancer drugs will take aim with unprecedented certainty, but making them requires a new discovery and development paradigm.
By Malorye Branca
August 13, 2003 | First came Herceptin, then Gleevec. Now suddenly there's Iressa, and right behind it, Erbitux, Avastin, and Tarceva. Oncologists are finally getting what they and their patients have been waiting for: targeted therapies. Unlike the cytotoxics, or "cell killers," that have long been the mainstay of cancer treatment, these drugs specifically knock out cancer cells without harming healthy ones. It's a boon to patients who have often struggled as mightily with the side effects of chemotherapy as they have with the actual disease.
"Targeted therapies are the hope we are all counting on for the future," says Todd Golub, of the Dana-Farber Cancer Institute and director of cancer genomics at the Whitehead Institute for Biomedical Research.
But the way forward is by no means clear. The first wave of these targeted drugs followed multiple paths and faced numerous surprise hurdles. To achieve the dream of truly individualizing cancer therapy requires new technologies and insights gleaned from genomics. The tools, the trials, the very business structures of companies, all need to be re-evaluated.
"We had a well-established paradigm for the development of cytotoxic drugs. Now we have to re-engineer the process for these targeted therapies," said Lee F. Allen, vice president of clinical research and development in oncology at Wyeth Research, speaking at the Strategic Research Institute's recent Anti-Cancer Drug Discovery and Development Summit.
| Resistance fighter: A "range of drugs that match mutations" could be key, says Paul Herrling of Novartis.
In the future, cancer will increasingly be defined not so much by the body part affected, but by molecular characteristics. "We should be talking about mechanism-one, -two, or -three cancers, irrespective of the organ these occur in," says Paul Herrling, head of corporate research at Novartis International.
Even before the genomics revolution, attractive cancer targets were emerging from molecular biology labs. With the genome sequence in hand and growing interest in systems biology, researchers hope to not only map the many roads to cancer, but also find the best points to choke cancer cells.
Gene-expression studies offer a hint of how this will pan out. Comparisons of RNA levels are revealing biological differences between cancers that look identical under the microscope. It's not just a more accurate way to classify cancers — according to which genes are activated or turned off — but also provides a swelling list of new targets.
|In the wake of the genomic technology boom, many groups are now concentrating on 'getting it right' by refining parts and processes in their platforms.
Other tools such as proteomics and comparative genomics are also adding to that list. Indeed, it's growing so long that most companies are concentrating on how to sift through the interesting candidates and determine which ones will pay off fastest. There is no universally agreed "best way" to do this just yet, although some firms say they've found an edge (see "Technology Tune-Up," right).
"Of course we think we have found the way," Herrling says. "But this is science, so we'll probably have to make some changes." Novartis has tried to make its high-throughput genomic system "flexible" so those changes can be easily achieved.
Pinpointing New Opportunities
One major starting point is to compare tumors to normal tissues, although "there is incredible diversity between tumors," says Joe Gray, director of the life sciences division at Lawrence Berkeley National Lab, and a professor at the University of California at San Francisco.
|The discovery in 1962 of the notorious Philadelphia chromosome caused by the splicing of chromosomes 9 and 22 ultimately led to Novartis' development of Gleevec.
To tease out common paths, Gray's lab pioneered comparative genomic hybridization (CGH) — a method for analyzing the chromosomal breaks and rearrangements that are the hallmark of cancer cells. His group now performs CGH on high-density arrays, which provides much finer resolution than the 10-megabase limit before. "The hard part is managing thousands and thousands of bits of DNA," Gray says. It's also tough to measure very small changes in DNA quantity. "We're interested in the gain or loss of a single copy," he explains. That's in contrast to gene-expression analysis, where the RNA levels measured differ greatly.
CGH should pinpoint some new opportunities. "Once you find a recurrent abnormality in the DNA, that should be a strong guide to a therapeutic target," Gray says.
Indeed, it was the discovery in 1962 of such an abnormality — the notorious Philadelphia chromosome caused by the splicing of chromosomes 9 and 22 — that ultimately led to Novartis' development of Gleevec. The drug formerly known as STI-571 is a small molecule that snugly binds to the active site of the BCR-ABL fusion protein, produced by the chimeric Philadelphia chromosome, that results in chronic myelogenous leukemia (CML).
Gleevec has aroused immense excitement because it has precisely the qualities that oncologists have long sought: It's highly efficacious, with few side effects. Clinical trials were so successful that the FDA approved Gleevec in just 10 weeks in 2001.
Pockets of Resistance
Road to cancer: BCR-ABL fusion proteins are just one route to cancer. Gleevec specifically targets this protein, raising hopes that drugs will be found for other targets.
Additional studies showed the drug also inhibits c-kit and PDGF, which, like BCR-ABL, are kinases. Physicians are taking advantage of this relative lack of specificity, showing that it is effective in patients with gastrointestinal stromal tumor, where the drug targets the c-kit oncoprotein, and in hyper-eosinophilic syndrome, where it hits PDGF.
These happy coincidences have led researchers to rethink some of their ideas about targeted therapies. It's difficult, and apparently unnecessary, to try to find compounds that make clean hits. The emphasis now is on making sure that what the compound does hit doesn't cause side effects.
But Gleevec is not quite the magic bullet that 60 Minutes and Time magazine proclaimed it to be. Several leukemia patients relapsed and died. Just as they outmaneuver traditional chemotherapies, cancer cells acquire mutations that protect them against Gleevec. Novartis and others are urgently seeking "sons and daughters of Gleevec" to use as follow-up treatments.
The resistance typically arises from point mutations in and around the Gleevec binding pocket. With some tinkering, Herrling says, they will be able to make new compounds that overcome the mutation. Work on these compounds has begun. "These drugs will be variations on a theme," Herrling says. He envisions a day when cancer will be treated with a "range of drugs that match mutations in the protein." Alternatively, cancer could be managed the way HIV is — first with a powerful cocktail, and then with tailored therapy.
Now Find the Patient
Beyond the mammoth task of developing such drugs, companies also need to be able to easily match them to the right patients. Tests for this are not required, but the pressure is building to have them. "If a therapy is targeted, then you had better be able to tell which population it's most going to benefit," said Richard Pazdur, director of the FDA's Division of Oncology Drug Products, at this year's American Society of Clinical Oncology (ASCO) meeting.
That's not as easy as it might seem.
"The biggest challenge for pharma will be getting into diagnostics," says Karol Sikora, a consultant with the London Cancer Group and visiting professor of cancer medicine at London's Imperial College. This is an area that most large pharmas have avoided because it's not financially attractive. But that could change.
"We are not a diagnostic company in the classical sense," says Novartis' Herrling. "But with this kind of biological specific therapy, you must have a way to distinguish the mechanisms you are targeting." Even if they don't want to develop the tests themselves, companies will want those markers available somehow.
Finding markers is tough. "It is almost as difficult to develop a validated biomarker as it is a target," says Nicholas Dracopoli, vice president of clinical discovery technologies at Bristol-Myers Squibb.
For one thing, most targets are not simply "on" in some cancers and "off" in others — but are expressed at different relative levels. Determining which patients are positive and which are negative is tricky: Where do you draw the line? It's also challenging to get assays that are simple and reliable. Moreover, adding biomarkers to a trial can double the cost, according to Wyeth's Allen.
|"It doesn't help that the preclinical staff is being rewarded for the number of candidates they get into development. We need these guys to also make biomarkers in animal models." Karol Sikora, Imperial College, London
The emphasis now is on getting the markers early. "When you throw [the compound] over the fence to development, you want to have a good marker already," Allen says. And it's not enough just to have the marker — researchers need to know how to tell whether the drug is working. That means figuring out how much inhibition is needed, and for how long, to obtain a good effect. "If that range isn't determined preclinically," Allen warns, then "you need to do that study in the clinic, and it's going to cost a lot."
Karol Sikora agrees. "It doesn't help that the preclinical staff is being rewarded for the number of candidates they get into development," he told the SRI meeting. "We need these guys to also make biomarkers in animal models." That's going to raise its own challenges, he admits, since "there are some very elegant animal models out there that don't have anything to do with what's going on in the clinic."
Just proving a drug hits the target isn't enough. Iressa, for example, is aimed squarely at the EGFR receptor, but Mary Lynn Carver, director of public affairs at AstraZeneca, notes that "in the clinic, there has not been a correlation between having high EGFR expression and getting a good response."
With lingering uncertainty over Iressa's precise mode of action, it is not yet a bona fide targeted therapy. Still, it is being prescribed broadly for non-small-cell lung cancer patients who have already tried at least two other therapies. AstraZeneca and others are searching for a good biological indicator of response. "From what we've seen so far, it may be easier to find who it won't work for than who it will," Carver says. Selecting patients will be particularly important if it turns out the drug has serious side effects: A rash of unexplained pneumonia cases has occurred in Japan, where the drug was first approved.
Having biomarkers will be even more important for tomorrow's targeted drugs, when the market has become more crowded. Giants like Pfizer, GlaxoSmithKline, and Bristol-Myers Squibb are gathering this information as early as they can. "We are all trying to do it," Dracopoli says. Because of the difficulty in obtaining sufficient samples and doing the studies in time, it may take a while before the results are useful. "Most of what we are doing now will impact the follow-up compounds and finding new indications [for marketed drugs]," he says.
The other hope is that new biomarkers will tell when not to use a particular drug. Because cancer drugs are so hard on patients, and can be a matter of life and death, this is an urgent need.
For this application, many researchers are looking for gene-expression signatures. Two landmark papers published last year in The New England Journal of Medicine reinforced the power of previously established gene signatures. One group at the Netherlands Cancer Institute collaborated with Rosetta Informatics on a breast cancer prognosis signature. In the other study, Louis Staudt and colleagues at the National Cancer Institute defined signatures of lymphoma types. These studies convinced many clinical researchers that gene expression had truly arrived.
In another such study, Jenny Chang and colleagues at Baylor College of Medicine are trying to confirm some intriguing findings from DNA microarrays. They have a 90-gene signature that seems to predict which tumors are resistant to Taxotere, one of the most powerful chemotherapies for breast cancer. The Baylor group looked at tumors that either started out or became resistant to the drug. Interestingly, the two sets of tumors showed similar gene-expression patterns. Now, the researchers are testing the signature against a larger set of samples, and studying the genes that are turned on in Taxotere-resistant tumors — genes that help cancerous cells escape normal cell death.
Ultimately, Chang admits, it will be more useful to have biomarkers that tell which particular therapy will work. "What's the point of a general prognostic signature?" she says. That kind of signature leaves too many questions hanging. "Hopefully, we will end up with more specific signatures," she adds.
This impending explosion of clinical microarray studies may not be an entirely good thing. Larry Norton, head of the division of solid tumor oncology at Memorial Sloan-Kettering Cancer Center, warned ASCO members of the pitfalls associated with "dealing with one thousand times more data points than usual." As he noted, "With this much data, it's easy to find something." Researchers need to thoroughly validate their findings, or a morass of false leads will be generated.
Even if reliable gene-expression signatures are found, it won't be easy to get them approved. The FDA is studying the issue, but few companies are eager to be first to try to get a DNA chip test approved. Bristol-Myers Squibb, for example, is boiling the signatures down so they contain many fewer markers and can be done using traditional immunoassays. The company has already found one such marker for an anticancer drug in Phase I trials, and more are on the way. "We consider pharmacogenomics in every program," Dracopoli says.
|"The development of Gleevec is the shot in the arm that everyone was waiting for. If you know what you are doing, you can design new drugs based on mechanisms and get spectacular results." Todd Golub, Dana-Farber Cancer Institute
CGH arrays and protein expression are also being used to generate biomarkers. "We need to look at this from different angles," Gray says. He's involved in a large-scale study that will compare the strength of gene expression to CGH. DNA is easier to work with than RNA, but Gray thinks that gene expression may provide a better signal. And don't forget proteomics. "It will take looking at all three to understand the best possible descriptor," he says.
With markers in hand, companies should be able to do speedier clinical trials that require fewer patients. That will be critical because there are more than 400 anticancer drugs in the pipeline.
Despite these challenges, researchers are enthused about the prospect for cancer diagnostics and treatment. "The development of Gleevec is the shot in the arm that everyone was waiting for," Golub says. "It showed that if you know what you are doing, you can design new drugs based on mechanisms and get spectacular results."
A giant bolus of targeted therapies is working its way toward the clinic. Even Imclone's Erbitux, which had an uncertain future for a while, seems to be on its way to approval thanks to good clinical data from Europe, where Merck KGaA carried out the trials. "We will have more cases like Gleevec," Herrling says confidently. "And every one of them is absolutely a breakthrough, for the patients and for medicine."
ILLUSTRATION BY TUKIMOTO OFFICE; PHOTO CREDIT: JENNY CHANG BY MARK GREEN