Oncology researchers have embraced the notion of cancer stem cells (CSCs): certain tumour cells, it seems, are more equal than others. But some think that the flurry of attempts to identify, purify and study CSCs in one tumour after another may be neglecting the role of the surrounding microenvironment in tumour formation.

While CSCs account for a minority — sometimes a very small minority — of a tumour's bulk, they apparently have a lock on the ability to propagate that tumour, as determined by the xenograft transplantation assay pioneered by John Dick of the University of Toronto in Canada. Human tumour cells are suspended, separated into two fractions depending on the presence or absence of a candidate surface marker, and transplanted into severely immune-deficient mice. If only one fraction leads to full-blown tumour formation, the tumour-propagating fraction from the human sample can be subdivided according to a second candidate surface marker and introduced to new mice to further characterize the tumorigenic cells, and so on as long as the sample supply lasts. Through such iterations, tumour-initiating cell populations can be increasingly differentiated from those composing the bulk of the tumour.

Some cells in a tumour are better targets than others Credit: Getty/Jessica Kolman

Dick's isolation of leukemia CSCs in 1997(1) has since been followed by a series of CSC sightings in breast, brain, colon2,3, head-and-neck, and other solid tumors. In some cases, CSCs reportedly number only a few in a million tumor cells. They seem to arise when genetic or epigenetic insults accumulate in normal stem cells, whose innate prowess at division is generally kept in check by internal controls as well as by the stromal niche, a grouping of supporting cells and substances in which stem cells reside. When those restraints are broken, the stem cell (or one of its downstream progeny, a still-frisky progenitor cell) transforms into cells whose self-renewal is no longer tightly regulated. “A CSC is not a completely aberrant, wildly crazy kind of cell, just a normal cell that's gone a little bit bad,” says Dick.

The hunch is that pinpointing the cell type from which a CSC arises would allow close comparisons with CSCs in corresponding tumours. This is spurring efforts to identify and characterize normal stem and progenitor cells in one tissue after another — most recently stem cells of the intestinal tract4, by a group led by Hans Clevers, director of the Hubrecht Institute in Utrecht, the Netherlands. Looking at similar cells that are just different enough for one to be healthy and the other cancerous will, it is hoped, turn up subtle differences — in, say, surface markers or self-renewal pathways — that could serve as targets for intervention by antibodies in the first case and inhibitors in the second.

Whether most or just a few cells are culprits in the tumor's initiation, growth, and metastasis is no angels-on-the-head-of-a-pin debate.

Whether most or just a few cells are culprits in a tumour's initiation, growth and metastasis is no angels-on-the-head-of-a-pin debate. Cancer therapies have largely aimed at trying to shrink tumours. But if 15% of a tumour's cells — or, for that matter, 0.01% — are responsible for its aggressive expansion, declaring a therapy successful because it has reduced a tumour's size by 75% is premature. Shrunken or even apparently eradicated tumours recur with alarming frequency in patients treated with chemotherapy or radiotherapy. For scientists who believe CSCs are to blame, the few cells that got away were precisely the ones that most needed to be reigned in.

“The assumption now is not that a small set of cells, inherently no different from the other cells that did die, just somehow managed to escape treatment, but rather that these cells are inherently different,” says Clevers. Last December, a group led by Jeremy Rich at Duke University, Durham, North Carolina, reported heightened radiation resistance in brain CSCs compared with other tumor cells5. But while CSCs are more resistant to some insults, they might be less resistant to others. This differential sensitivity to drugs or radiation could provide a window for more efficient, less toxic ways of eliminating tumours.

Cell or environment?

But the designation of particular tumour-cell subsets as CSCs has been questioned recently in a brief article in Science6 by a group led by Andreas Strasser at the Walter and Eliza Hall Institute of Medical Research, in Melbourne, Australia. Strasser and colleagues challenge the validity of the standard mouse xenograft assay. They argue that it was not the rarity of tumor-initiating cells, but rather the vagaries of interspecies intermolecular interactions that explained why transplanted cancer cells so rarely launch tumours.

Reasoning that transferring human material into a mouse, no matter how immune-impaired, might yield false-negative results due to faulty signaling between human and mouse tissues, these investigators used a mouse-to-mouse experimental system — transplanting progressively smaller numbers of cells from two mouse lymphomas and one acute myeloid leukemia (AML) into syngeneic mice — and found that as many as 10% of the transferred cells could regain a foothold in the second mouse. The previously reported one-in-a-million frequency of human AML cells producing tumours in xenograft assays, the authors suggest, might reflect the rarity not of cells capable of initiating tumors but of human tumour cells that can grow in a mouse.

“For a human cell to grow in a mouse is not so simple,” says Clevers, who did not participate in the study. He notes that even within the same human body a kidney cell transplanted into the lung will fail to thrive, and the same kind of thing can happen with a tumour cell. “The cell needs to be very malignant to leave its own niche and grow somewhere else,” he says.

“A tissue's phenotype is a collaboration between the cells that are around it and the cells that are in it,” says pathologist Thea Tlsty at the University of California in San Francisco, and so is a tumour's, she adds. Her 1999 study7 showed that fibroblasts from the area of a carcinoma sent signals to non-tumorigenic cells that made them tumorigenic. “We found that if non-tumorigenic cells were placed with these carcinoma-associated fibroblasts, in contrast with normal fibroblasts, you had a growth that was 500 times bigger than with the normal control and looked exactly like a human prostatic carcinoma.” The literature also indicates that normal stromal fibroblasts, placed in proximity with a malignant tumour cell, can suppress the phenotype of malignancy, she says. To Tlsty, this means that a functioning niche might calm a wild stem cell, while an aberrant niche could goad otherwise healthy stem cells into malignancy.

CSCs defended

Proponents of CSCs offer several lines of defence. The Strasser group's high reported frequencies aside, they argue, what counts is that some cells propagate tumours and others do not. “Nobody ever argued that a CSC had to be rare,” says Craig Jordan, a researcher at the University of Rochester, New York. “The key point is that there's heterogeneity in the tumor, where some proportion of its cells – it could be 10%, it could be 20% — has a different capacity,” he says.

Were a CSC to turn out to be not one in a million but more like one in ten, “it still would mean 90% of the cells are less relevant than the key 10%,” says Peter Dirks of Toronto's Hospital for Sick Children, who in 2004 identified cells that propagated human glioblastoma, an infamously aggressive cancer, in the brain of an immunodeficient mouse8. A cell may not need to be rare to be relevant, he says. “It may be that some cancers are so aggressive because they have a lot of stem cells.”

Michael Clarke at Stanford University, whose lab's identification of CSCs in human breast cancer9 was the first such finding in solid tumours, says his group has corroborated xenograft assay results using other routes. After isolating human head-and-neck CSCs in January, Clarke's team showed that pathways associated with self-renewal were activated in those CSCs and not in the non-tumorigenic cells10. Moreover, differentiation pathways were activated in the non-tumorigenic cells but not the CSCs. “That's pretty strong evidence that it wasn't just microenvironment,” says Clarke. “And we were able to show this in the human tissue — not just in the mouse, but in the human tissue that we got the samples from.”

Both Jordan and Clarke have publications pending in which they have identified CSCs, in, respectively, chronic myelogenous leukaemia and breast cancer in mouse models, where microenvironmental hostility to human tumors is a non-issue because the cancers themselves came from mice.

The genetically re-engineered mouse models used in Strasser's study may not truly reflect human cancer, say some researchers. “They used a lymphoma model,” says Jordan. “There is no report of any kind that I'm aware of suggesting that lymphoma arises from a CSC,” he says. “Nobody's claiming that the CSC hypothesis is true in every single cancer.” Besides, says Jordan, neither Strasser's lymphoma nor leukemia model recapitulate human disease very effectively.

But researchers studying CSCs also need to question their models, says Dirks. He suggests that scientists working in nonhuman environments might try humanizing the site of engraftment — although that may be easier said than done — or look for other means to assess the results of human cancer cells in mouse models, for example. Dirks remains confident in the validity of his own finding that glioma, at least, is driven by CSCs, but he grants that Strassers's work bears attention. When researchers expect to find what they are looking for, they often do. “People jumping on this bandwagon need to look at their data more critically and carefully,” he says.

Irving Weissman, director of Stanford's Institute for Stem Cell Biology and Regenerative Medicine, has the last word: “The final proof of the CSC hypothesis is when we take those CSCs, find markers unique to them, and use those markers to treat the cancer in a patient, and the CSCs die, and the cancer goes away.”