Cancer cures in mice are like Mark Twain's observation that "quitting smoking is easy - I've done it a hundred times." The frequent successes that are reported in the scientific literature are a particularly frustrating form of easy come, easy go; the chance that a given cancer-fighting compound will progress from Phase I to FDA approval is less than 5 percent, not to mention the attrition from preclinical.

Part of the problem is that the mouse models simply are not very good. Cheryl Marks, program director of the National Cancer Institute's Mouse Models for Human Cancers Consortium, told BioWorld Today that the most common cancer model - the xenograft - is "not really a mouse cancer model. It is a human tumor model in an immunoincompetent animal."

And what's wrong with that? Well, any number of things.

The implanted cells are often derived from a single tumor; they have been grown on plastic for generations to decades before being implanted into mouse flanks, which is not the organ in which the original tumor developed; and the animals have a severely altered immune system. All of that makes the conditions under which the xenografts grow even more unrealistic, since, as Marks pointed out, the immune system plays "an enormous role" in how cancer originates, develops and spreads.

But researchers are developing new forms of mouse and, to some degree, rat models that will improve their predictive powers for human disease. The models, which mostly consist of inducible knockouts or transgenes in the cell types and pathways that go awry in human cancers, "are rodents - they are not us," Marks said. "But the gene similarities are remarkable."

In general, Marks said, there's bad news and good news about new transgenic mouse models. The bad news is that "we do a very bad job of curing cancer in these animals." The good news is that this is because the transgenics "are really good mimics of the response in the clinic - unlike the xenografts, which often show gangbuster responses to things that don't pan out in the clinic."

The push to develop such models started about eight years ago, when National Institutes of Health in Bethesda, Md., took stock of cancer research efforts and decided that "we were not learning, nor were the pharmaceutical and biotech companies learning, what we needed to know about human disease" from the available mouse models. And the newest example of the effort appears in the April 24, 2006, issue of the Proceedings of the National Academy of Sciences.

The paper described a mouse model for the pediatric brain cancer medulloblastoma. The authors of the paper, from Harvard Medical School, compared two strains of mice that had different genes knocked out specifically in neural progenitor cells.

One group lacked a protein that is part of the nonhomologous end-joining pathway, a major pathway by which cells repair double-stranded DNA breaks, and another that had an additional deficiency in the tumor suppressor gene p53.

The mice appeared to develop normally early on, but the double knockouts invariably developed medulloblastomas by the time they were 12 to 14 weeks old.

Analyzing the tumors for genetic abnormalities, the scientists found that several genes that are activated or inactivated in human medulloblastomas also were frequently altered in association with recurrent chromosomal translocations in their mouse model, which showed frequent translocations involving chromosomes 6 and 13.

Marks said that in the PNAS study, the translocation findings in particular were "representative of what you would find in humans," and called the paper "very elegant from a scientific standpoint."

But she added that because the animal model was designed to answer certain basic science questions, other models might be best for drug development.

Rather than considering one particular mouse - or rat, or zebrafish, or cell line - as the model to end all models, the question for both researchers and drug developers should be "where and when do we use which models?"

Frederick Alt, professor of both pediatrics and genetics at Harvard Medical School, and the paper's senior author, agreed that the model was developed to answer basic science questions. But he added that the model does possess features that could make it of interest to drug developers, as well. Specifically, the mice develop tumors fairly rapidly and have "alterations of genes that come up in human tumors."

So maybe, with some more discoveries in the same vein, the biotech industry will do better than Mark Twain, who smoked on and off until his death in 1910.