Nature Biotechnology
22, 15 - 18 (2004)
doi:10.1038/nbt0104-15
Signal transduction inhibitors—a work in progressJoseph BeckerBoston Is the race to take protein kinase inhibitors to market compromising their success in the clinic?Since Novartis' (Basel, Switzerland) imatinib mesylate (Gleevec) was hailed as a breakthrough in cancer treatment, other compounds that specifically target protein kinases have found success more elusive. AstraZeneca's (London) gefitinib (Iressa), which was approved in Japan one and a half years ago, has been associated with 200 deaths from interstitial pneumonia in people with lung cancer. And late in November, results from a US phase 3 trial in people with colorectal cancer for Genentech's (S. San Francisco, CA, USA) bevacizumab (Avastin), a monoclonal antibody targeted against a tyrosine kinase inhibitor, failed to reach statistical significance, despite a 29% increase in participant survival. Even the long-term effectiveness of Gleevec has been limited because of mutations arising in p210BCR-ABL, its tyrosine kinase target. According to speakers at a recent meeting in Boston*, these setbacks reflect a variety of factors, including a rush to get compounds into the clinic, a lack of validated biomarkers and insufficient characterization of patient populations appropriate for treatment.
Magic bullets?
The rationale behind Gleevec and similar drugs is that by targeting a rogue kinase in a signaling cascade, cancer can be prevented without damage to normal cells and adverse toxicity. The spectacular efficacy and minimal side effects of Gleevec in clinical trials of people with chronic myelogenous leukemia (CML) resulted in its rapid approval by the US Food and Drug Administration (FDA; Rockville, MD, USA) in 2001. This, together with similar successes for Genentech's (S. San Francisco, CA, USA) rituximab (Rituxan; which was codeveloped with Biogen Idec, Cambridge, MA, USA) and trastuzumab (Herceptin)—monoclonal antibodies (mAbs) that target components of cancer signal transduction pathways—convinced the drug industry that molecular-targeted cancer therapy was a realistic goal.
An apparent overnight success, Gleevec actually had several decades of academic and pharmaceutical research behind it. A key factor in this success was the unraveling of the biology underlying CML. Since the late seventies, it was appreciated that in nearly all cases of the disease, a chromosomal rearrangement occurs in which c-abl oncogene (ABL) becomes fused to the breakpoint cluster region (BCR) protein to form p210BCR-ABL. Thus, in CML, Gleevec's target essentially defines the disease it was designed to treat. Similarly, the overexpression of the epidermal growth factor receptor (EGFR) family member ERBB2 (also known as HER-2/neu) defines people with breast cancer suitable for treatment with Genentech's Herceptin.
In contrast, many of the new protein kinase inhibitors that are coming through the pipeline have been targeted at conditions where the biology underlying molecular pathogenesis of disease is less clear. AstraZeneca's Iressa, which is targeted at people with non-small-cell lung cancer (NSCLC), is a case in point.
Iressa on trial
Janet Dancey, of the US National Cancer Institute (NCI, Bethesda, MD, USA), presented a case history for Iressa at the Boston meeting. In its initial phase 1 trial in people with NSCLC, 10% of the participants went into complete response (CR) after treatment with the compound, an "observation that changed development of the drug," according to Dancey. But instead of moving to a series of small trials testing the drug in various combinations with chemotherapy, AstraZeneca then quickly moved into big phase 2 single-agent trials, which confirmed the phase 1 data and laid the groundwork for approval. Indeed, the Iressa early-stage trials data were so promising, they prompted a competitor, Genentech, to design and run large randomized trials for erlotinib (Tarceva)—a small molecule that like Iressa, targets epidermal growth factor receptor (EGFR)—in conjunction with chemotherapy. This was despite the lack of a complete response in people participating in Genentech's own phase 1 trial on Tarceva in people with lung cancer.
In sharp contrast to the early Iressa data, however, subsequent phase 3 studies by both sponsors showed that their drugs provided "no benefit in first-line patients over chemotherapy alone in this patient population," Dancey summarized. Moreover, multivariate analysis did not pull out a significant subset of people, other than women from a single-arm study who seemed to respond better than men (as did nonsmokers, and those with adenocarcinomas).
The mediocre results left AstraZeneca and Genentech with some hard lessons to ponder about trial design. In addition, after considerable time and expense, clinicians were left with few guideposts for how to use EGFR inhibitors in people with lung cancer. Was the lack of efficacy due to a failure to identify a smaller subset of people with NSCLC who would be likely responders (and whose response could be shown with sufficient power in a smallish trial)? With efficacy, but no molecular markers, was there a better way to proceed; for example, should phase 2 combination trials be considered as a matter of course for compounds that haven't already been validated preclinically, despite the fact that small combination phase 2 trials aren't likely to be enough to accurately predict response rates? Was it that three drugs work no better than two? Did the clinical research community really need data from four trials of very similar agents, each enrolling over 1,000 people?
Clinicians knew the EGFR was commonly expressed in people with NSCLC, Dancey explained, and that forced overexpression of the receptor transformed cells in model systems. "EGFR may play a critical role in growth, repair and survival, angiogenesis and metastasis," she noted. "And the drugs appeared safe."
On the other hand, at the time the trials were designed, the epidemiology of the target wasn't understood, including the role of EGFR in the pathophysiology of lung cancer tumors, the effect of target inhibition in humans and whether the molecular phenotype was predictive of clinical response—as p210BCR-ABL was, for example, for Gleevec, or even certain levels of overexpression of ERBB2 were for Genentech's Herceptin. On this evidence, Dancey wondered whether late-stage randomized trials should have been delayed until the molecular phenotype had been worked out.
Stampede to the clinic?
But an industry imperative exists to develop drugs fast—a fact acknowledged by Dancey herself, who noted that in discussing this with industry colleagues, she repeatedly heard that the drug industry "doesn't know slow." As much as anything else, the highly competitive race to bring a first-in-class EGFR drug to market deterred the sponsors from running more empirical studies. Nor does industry tend to spend adequate time extracting information from failures. "There's no time to ask 'What if?' when a compound doesn't act as expected versus a target," commented Paul Workman of the Institute of Cancer Research (Sutton, UK).
When this was raised at the meeting, the industry response was to point a finger at the FDA. When Rick Pazdur, the FDA's representative, responded with the question why were four trials done in this case when "you usually have to hit companies over the head to do one?" Susan Arbuck of Aventis (Bridgewater, NJ, USA) answered that "companies do trials because the FDA calls for randomized trials—plural." Pazdur didn't dispute the comment, but nonetheless suggested that the real reason was that two companies were competing, and they needed data from combination trials for marketing purposes. "It wasn't so much regulatory," he said, adding that to the extent that commercial sponsors are tending toward multiple late-stage trials, it is for risk management purposes, to find the best indications for a drug.
The importance of biomarkers
The case of Iressa illustrates the problem of not systematically evaluating the pharmacodynamic effects within tumor specimens (so it is not clear whether optimal target inhibition is occurring) and perhaps more importantly the lack of information on molecular markers (biomarkers) that are predictive of antitumor activity. Biomarkers are essential to assess pharmacodynamics (to set dosage) and to select the people most likely to respond. Yet such markers—either those that measure discrete changes (e.g., phosphorylation in individual downstream proteins) or patterns of signals from several proteins activated in the pathway—have yet to be identified and validated for use in clinical trials of EGFR drugs. And Iressa and EGFR-targeted drugs are not alone; the next classes of signal transduction inhibitors now in the clinic, led by several against the vascular endothelial growth factor receptor (VEGFR; see Table 1, Fig. 1), also lack validated biomarkers. And the competitive demands of getting to the market rapidly mean that the drug industry is unlikely to wait for science to catch up to commerce.
 | | |
At the Boston meeting, Alex Matter, head of the Gleevec development team at Novartis, emphasized the point: "We need to test the concept [of targeted cancer drugs] with biomarkers in early clinical trials, as pharmacodynamic endpoints." And with the ability to measure multiple parameters in cellular screens now in hand using flow cytometry "the phenotypic gates have opened," he claims.
Indeed, many panelists at the meeting downplayed the importance of mechanistic work around targets as a starting point for drug development, instead favoring a systems biology approach where compounds are first screened in cell-based assays, with mechanistic understanding of the target coming only after validation of its impact on the biology. The disadvantage of that approach is that activity must be deconvoluted after the fact. But that's better than starting with "a closed universe driven by prior knowledge of the target," commented NCI's Edward Sausville. "The drugs will show the importance of a target in disease," added Paul Workman, as Gleevec did with p210BCR-ABL in CML and c-kit in gastrointestinal stromal cell tumors.
Resistance
Another problem lurking for cancer drugs is the emergence of resistance as disease progresses. According to Charles Sawyers of the University of California, Los Angeles, who presented on mechanisms of resistance in CML at the meeting, the pattern of patients' resistance to Gleevec will be a model for kinase inhibitors generally.
CML is defined by the molecular translocation called the Philadelphia chromosome, which produces p210BCR-ABL. For decades, researchers have known that overexpression of the platelet derived growth factor receptor (PDGFR), which is linked to p210BCR-ABL, is necessary to preserve CML, which progresses in three phases: chronic (lasting four to six years); accelerated (one year); and blast crisis (only three to six month survival).
Since its approval in 2001, clinicians have discovered that despite a 95% response rate and remissions lasting for years in chronic patients, as the p210BCR-ABL fusion undergoes additional genetic hits, causing it to mutate. Thus, although the drug is active in people in blast crisis and 75% of such people respond, these remissions last only weeks or, sometimes, months. "The data suggest that late-stage disease can't be cured with a single agent," Sawyers suggested. Indeed, point mutations in p210BCR-ABL account for 80% of relapses, he noted. Some mutations disappear after discontinuing Gleevec—in a similar manner to hypomorphic alleles that cause resistance to azidothymidine (AZT) in AIDS when that drug is used as a single agent, Sawyers suggested. Other mutations appear before the use of the drug and persist (hypermorphic alleles that wipe out nonmutated p210BCR-ABL and so select for and encourage mutations). "You need another drug, one that targets the open conformation of ABL, instead of the closed formation," he said, as therapy in late-stage disease.
Sawyers proposed three take-home lessons to be learned from the Gleevec experience: mutant kinase targets are a smoking gun for kinase dependency; resistance reveals tumor heterogeneity; and the conformation of the kinase—and whether it's active or inactive—may be important when choosing drug leads to take into the clinic.
Gleevec, which hits both p210BCR-ABL and c-kit, turned out to be one of the first examples of a multi-targeted kinase inhibitor—"therapeutic shrapnel," in the words of Judith Sebolt-Leopold of Pfizer Global Research & Development (Ann Arbor, MI, USA). In such molecules, different portions bind to different sites on kinases in what's often referred to as a 'barbell' approach. The benefit is that, unlike using different drugs for inhibiting different kinases, drug-drug interactions and overlapping toxicities are minimized. On the other hand, given the heterogeneity of tumors among people with cancer and even in the same person over time, multiple drugs give clinicians an opportunity to vary dosing in proportion to the specific person's tumor expression profile and the pathways activated in that individual. (The downside has been potency, or rather too much potency: as Alex Matter points out, it's impossible to formulate an oral drug at very low doses.) In fact, most of the new small-molecule inhibitors of VEGF also hit other targets—including PDGFR, FLT3, c-RAF and b-RAF—to inhibit tumors via multiple mechanisms. For example, Bayer's (West Haven, CT, USA) and Onyx's (Richmond, CA, USA) codeveloped drug BAY 43-9006, which is now in phase 3 in malignant melanoma, inhibits both c-RAF (anti-angiogenesis) and b-RAF (proliferation) kinases. Other examples of these next-generation compounds are Pfizer's (New York) SU11248, which targets PDGF, VEGF, c-kit and FLT3 and is in three phase 3 single-agent studies for gastrointestinal cancer, and Novartis's AEE 788, an inhibitor of kdr/VEGFR2, EGFR and ERBB2.
Conclusions
The fundamental role of kinases in cancer biology and the success of pioneering therapeutics, such as Gleevec and Herceptin, have prompted intensive efforts to develop kinase inhibitors, over 30 of which are now in the clinic. However, many of these drugs cry out for validated clinical biomarkers to help set dosage and select people likely to respond. The situation is compounded by the poor interplay between academics, clinicians, regulators and companies, resulting in a dismal record of development of molecularly targeted drugs, despite good science.
Twenty years after beginning to work on genetically well-established targets, only a handful of approved compounds has emerged. Despite the fact that developing a drug from scratch usually takes more than a decade, "It's not a good record," says Alex Matter, citing the industry's lack of creativity and laying blame squarely on inefficiencies in the system. Matter also points out that the collective experience of the past 15 years has shown that, in terms of efficacy models, "xenografts have been misleading."
But companies have little choice than to continue to rush headlong toward approvals with little clinical validation along the way to fill their shallow pipelines. For this reason, they are likely to continue to support rapid development even of those drugs validated only by the clinical experience of related compounds against the same targets (e.g., Tarceva's development on the basis of Iressa's phase 1 results). Witness the deal in September 2003 in which Aventis (Strasbourg, France) received joint development and marketing rights to Regeneron Pharmaceuticals' (Tarrytown, NY, USA) phase 1 VEGF trap, a fusion protein that acts as a soluble receptor to bind VEGF. Aventis paid the unheard-of sum of $80 million up-front for rights to the drug, based largely on data from phase 3 studies in colorectal cancer of a competitor, Avastin, from Genentech. The Avastin data, released early in 2003 at the annual meeting of the American Society of Clinical Oncology, were so compelling that they triggered a bidding war for the VEGF trap. Yet in late November, the second Avastin Phase III trial failed to reach statistical significance. That might be significant for Regeneron's VEGF trap and it might not. But $80 million is a lot to stake partly on the basis of data from another drug.
*AACR-NCI-EORTC International Conference on Molecular Targets and Cancer Therapeutics, Boston, November 17−21, 2003.
|
