MADRID (03/22/2004) - "Cancer Jab Soon" was a British newspaper headline that greeted the discovery of the first human oncogene. Media hype has changed little in 25 years, nor has the annual number of casualties of cancer. But speakers at the Spanish National Cancer Centre (CNIO) 2004 Symposium* presented copious signs of a detailed new taxonomic understanding of tumor cell biology, thanks largely to advances in DNA profiling of gene expression, producing new clues about the hallmarks of cancer and new strategies in treating the disease.
Census of Cancer
Genome census has been conducted by Michael Stratton (Wellcome Trust Sanger Institute) and colleagues of all cancer-causing mutations in the two decades since the first such mutation was described in the RAS oncogene by Mariano Barbacid (CNIO). Mutations have been logged in 291 genes, or 1 percent of the genome, although most are restricted to sporadic, rather than inherited, cancers. Twenty-seven of these are genes for protein kinases (enzymes that attach phosphate groups to proteins), whereas only six would be expected by chance, which helps explain why the kinase family is such a popular therapeutic target. Stratton's Cancer Genome Project ultimately aims to screen the entire genome for cancer mutations.
Cancer is often associated with changes in gene expression as well as sequence, spawning interest in the field of epigenetics. Manel Esteller (CNIO) spoke of "the five bases of DNA," referring to 5-methylcytosine, which makes up 4 percent to 8 percent of the genome. The use of demethylating agents to re-activate tumor suppressor genes reduces lymphoma in mice, and has "big potential for cancer treatment," Esteller said.
Despite the intense focus on genetic factors, environmental triggers remain shrouded in mystery. Mary-Claire King (University of Washington), who pinpointed the breast and ovarian cancer gene BRCA1 in 1990, reported that the risk of breast cancer in relatives of BRCA1/2 carriers is significantly higher in women born after 1940, than in those born before 1940. The key environmental determinant turns out to be teenage physical activity. Girls who were active tend to have a later age-of-onset of cancer, presumably because of the effect of physical activity on regulating estrogen levels (a known cancer risk factor). King also described a new technology called FLASH (fosmid library allele separation and haplotyping), which she is using to identify new mutations and susceptibility genes.
A cursory analysis of Medline citation trends by Olli Kallioniemi (VTT Technical Research Centre of Finland) reveals that interest in microarrays has outstripped all other postgenomic buzzwords (bioinformatics, systems biology, genomics, and so on). Kallioniemi's team is studying numerous types of biochips, including cell carpet assays -- overlaying cells on slides printed with biomolecules such as Qiagen's "druggable genome" siRNA library. Tissue microarrays, with 1,000 tissue specimens per slide, and lysate microarrays, are also paving the move "from pathology to pathomics."
However, most presenters focused on the diagnostic and therapeutic applications of DNA chips. In 2002, Laura van't Veer (Netherlands Cancer Institute) described a breast cancer gene profile that correlated with poor prognosis. New profiling studies of primary breast tumors reveal close similarities to distant metastases, suggesting that metastatic potential is an inherent property of breast tumors (Weigelt, B. et al. PNAS 100, 15901-5; 2003).
Louis Staudt (National Cancer Institute (NCI)) has extended profiling studies in lymphoma to follicular lymphoma, seeking the "supervised discovery" of key genes linked to clinical outcome and survival. Staudt found that the synergism of three groups of gene signatures -- two in immune response, one in B-cell differentiation -- was the best predictor of survival outcome. Miguel-Angel Piris (CNIO) also described novel gene signatures in non-Hodgkin's lymphoma. And Chris Boshoff (University College London) presented striking findings on the gene-expression changes associated with HIV-induced Kaposi's sarcoma.
Todd Golub (Dana-Farber Cancer Institute/Broad Institute) expressed surprise that his group's molecular taxonomy studies should yield a therapeutic target -- FLT3 kinase in leukemia -- and highlight the broad role of kinases in cancer. Lately, Golub's team has shown that some primary lung tumors exhibit metastatic gene signatures, using either Affymetrix or Rosetta inkjet arrays. But Golub cautioned that the current studies "are not robust enough to hallucinate about clinical significance."
In another vignette, Golub asked: "Can gene-expression signatures be used directly as a starting point for a small-molecule screen?" The answer is yes. Taking a small-molecule library of 2,000 compounds produced by Brent Stockwell, Golub has revealed 10 compounds that reproducibly induced a neutrophil signature in leukemia cells. Further studies have used a technique called gene set enrichment analysis, which Golub's colleague David Altshuler has used successfully to highlight gene signatures in diabetes, to classify gene signatures in lung adenocarcinoma.
Not every speaker enthused about DNA microarrays, however. "DNA is an information archive," said the FDA's Emmanuel Petricoin, "but proteins do all the work!"
Petricoin stridently advocated serum proteomics for the early detection of ovarian cancer, but rather than focus on a single biomarker (such as the current CA-125 test), Petricoin and Lance Liotta (NCI) have developed a diagnostic test called OvaChek -- a serum proteomic pattern revealed by mass spectrometry (MS) in conjunction with Correlogic Systems -- without identifying the protein fragment peaks. Early diagnosis of ovarian cancer would save lives and prevent unnecessary prophylactic oophorectomies. The first bioinformatic analysis relied on MS patterns produced by the Ciphergen SELDI system (Petricoin, E.F. et al. Lancet 359, 572; 2002), but now relies on the new high-resolution Applied Biosystems QSTAR time-of-flight system. "These are real features that are being predicted," Petricoin said, noting the excellent profile reproduction after 90 days, and hopes the test could be in the clinic within five years.
John Quackenbush (The Institute for Genomic Research) is "building resources to do science, to ask fundamental biological questions." Some of his open-source bioinformatics tools (see www.tigr.org/software) include MADAM (microarray data manager), MIDAS (Java data analysis), and MeV (data mining). The former physicist stressed the need to link genes to proteins because of the generally poor correlation between RNA and protein levels. "We're at a point where we can develop a theoretical biology, moving from biology being deterministic to being stochastic," he said.
Supporting that assertion, Iya Khalil, co-founder of Gene Network Sciences Inc., delivered a spellbinding overview of the company's colon cancer cell model using its diagrammatic cell language (DCL). The model incorporates more than 500 genes to model processes including apoptosis, cell cycle, and proliferation in five cellular compartments. For one unspecified pharma customer, GNS could predict the top 100 and lower 100 targets from a group of 775 targets. Khalil hopes to publish a portion of the in silico model later this year.
Cracking the 'Kinome'
The success of Gleevec has focused attention on the kinase family as enticing drug targets. Some 50 kinase inhibitors are currently in clinical trials, although only two have been approved so far. But building on that success will not come easy -- or cheaply.
"Drug discovery takes about five years," said GlaxoSmithKline PLC's Peter Goodfellow. One-and-a-half years are required for high-throughput screening, conducted in state-of-the-art facilities such as Tres Cantos, just outside Madrid; the remainder is devoted to "chemical tinkering to produce the optimal structure." Clinical testing could take seven years, but the prevalence of related kinases means many drugs could exhibit unexpected toxicity. Nevertheless, with the full kinase family now known, GSK can test drug candidates against the "kinome."
One drug in the clinic, GW572016, targets both EGFR and ErbB2 (the targets of Herceptin and Iressa/Tarceva). Sixty percent of tumors overexpress either of these proteins, raising the intriguing idea of targeting both simultaneously. Positive responses have been recorded in more than half the patients.
Ultimately, success will hinge on the cost-benefit ratio -- if the molecule is safe, it could be taken by a large number of people. "Aspirin is the world's most dangerous drug," said Goodfellow, yet still very safe, considering the number of people who take it.
Carlos Garcia-Echeverria (Novartis Institutes for Biomedical Research) also focused on kinase drugs, while reiterating the numerous "druggability hurdles" facing his medicinal chemistry team, including potency, cell permeability, toxicity, stability, and ease of manufacture. Novartis' pipeline includes PKC412, a drug that targets FLT3 in acute myelogenous leukemia, having yielded results in a mouse leukemia model comparable to Gleevec. Another co-targeting drug, AEE-788, looks promising as an inhibitor of angiogenesis (new blood vessel growth) in solid tumors.
Nick Dracopoli (Bristol-Myers Squibb) lobbied for pharmacogenomics in designing clinical trials, particularly in population screening, to combat the low efficacy of most oncogenic drugs. Finding good responders may not reduce the time of enrollment in clinical trials, but the success of "Herceptin shows that pharmacogenomics is economically practical in the industry," he said.
Sir John Maddox closed the meeting with the observation that the "public needs a sense of cultivated stoicism ... the bad luck of cancer is something we have to live with. Eventually it will be tractable, treatable, manageable ... But not yet."
Curtailing the cytological anarchy that claims one life in the United States every 60 seconds will take decades more work. But deep within the new molecular taxonomy of cancer almost certainly lies the answer.