After modifying a common experiment used to test the cancer stem cell hypothesis, researchers are reporting that melanoma, a deadly form of skin cancer, may not follow the model predicted by this theory.

The cancer stem cell hypothesis states that some cancers are driven by a small subset of self-renewing cells. One way to explore this idea has been to transplant human cancer cells into mice and observe the results. For a number of cancers, including melanoma, colon, and pancreatic, this experiment has shown that only some cells form tumors.

Citing the importance of this work to the field, Dr. Sean Morrison of the University of Michigan and his colleagues decided to scrutinize the model. Using melanoma as a test case, the team conducted the experiment twice, first with the original study design and then with some modifications.

The changes, such as using mice with weaker immune systems and extending the observation period, led to dramatic differences in the results. With the original design, tumor-initiating cells were rare (about one in a million), but they were much more common in the new model (about one in four), according to findings in the December 4 Nature.

The researchers propose that melanoma may not follow the cancer stem cell model but caution against applying the results to other cancers.

"We expect that some cancers really will follow the cancer stem cell model," said Dr. Morrison. "But there probably will be plenty of other cancers like melanoma in which the ability to form tumors is a common attribute of the cancer cells."

His team transplanted cells from 12 patients with melanoma into mice, including cells from both primary and metastatic tumors. The tumor-initiating cells had diverse features, and no single feature was associated with the ability to form tumors.

The cancer stem cell question has potentially broad implications, because new therapies designed to eradicate these deadly cells may be needed for cancers that follow the model.

"This is an important and rigorous study," said Dr. Jeremy Rich, who studies cancer stem cells at Duke University Medical Center and had no role in the research. "It shows that you can, through the manipulation of your mouse model, get a dramatic difference in the frequency of human cancer cells that are able to grow a tumor."

Whether tumors form in the mice depends highly on how the transplantation is done and on the nature of the recipient, added Dr. Stewart Sell of the Ordway Research Institute, who studies adult stem cells and cancer and also was not involved in the study.

Dr. Morrison's laboratory specializes in purifying cells, and for the first time the researchers describe transplanting single human cancer cells into mice. Of the individual transplanted cells, 27 percent formed tumors.

A big unknown is whether the cells that initiate tumors in mice also play a role in human cancers. The researchers note that an even greater - or a much smaller - fraction of melanoma cells may actually contribute to the disease in patients. They also stress that the results, in their view, do not refute the cancer stem cell hypothesis.

Nonetheless, said coauthor Dr. Timothy Johnson, "the study provides a new tool so we can now go back and repeat previous experiments to prove which cancers are following the cancer stem cell model and which are not."

The findings may apply to a subset of tumors and only under certain conditions, but it is equally possible that the observations are more commonly applicable, wrote Dr. Connie Eaves of the British Columbia Cancer Agency in an accompanying commentary.

Asked which of the two models might be more relevant for investigating cancer stem cells, Dr. Eaves replied, "We don't know. Both are only surrogates for the real test, which is: What causes tumors to grow and cause relapses in humans?"

Tumors do not grow in a vacuum, she continued, and few grow in the complete absence of host immune components. "So it is not yet clear that going to extraordinary lengths to promote human tumor growth in a mouse will ultimately be the best test of what propagates a tumor in a person," she said.

Despite progress, none of the models can yet predict what might happen in humans. This study, noted Dr. Rich, may "push people to do a better job monitoring how well cancer cells grow in different environments. And that is important."

—Edward R. Winstead