All too often, cancer patients relapse. Tumours apparently vanquished by surgery and chemotherapy return months, or even years, later. In recent years, an increasing number of scientists have seized on the notion of so-called 'cancer stem cells' as the culprits. This hypothesis proposes that the ability to initiate a new tumour is restricted to just a small population of cells. According to this model, the reason cancer returns is because the drugs that kill the rapidly dividing cells making up the bulk of the tumor do little harm to those few cells that really matter — cancer stem cells that are able to regrow the tumour. Since cancer stem cells were first identified in leukaemia in 1994 (ref. 1), they have been reported in human cancers of the brain, breast, colon, pancreas and other tissues.2

An improved assay for finding these putative cancer stem cells goes a long way to deflating their 'rare' status, however. Previous research indicated that such cells were exceedingly infrequent, perhaps only one in a million cells in the tumour. Now, work by Sean Morrison at the University of Michigan in Ann Arbor reveals that cells capable of generating new tumours may not be uncommon at all, at least in melanoma3.

Human cancer stem cells, or 'tumour-initiating cells' as they are less controversially called, are identified by transplantation assays. Cells from a tumour are isolated, sorted according to surface markers thought to yield the appropriate 'stem cell' population, and these cells are injected into a mouse. If the injection results in a tumour that generates the same mixture of cells as found in the original, then it presumably included cancer stem cells.

Morrison tweaked the standard assay in several ways. He transplanted the human melanoma cells along with some extracellular matrix to help them survive after transplantation, he waited longer for tumours to appear in the mice, and he used a different kind of mouse as recipient. Previous assays used NOD-SCID mice, immunocompromised mice that still have functioning natural killer (NK) cells, immune-system cells that can destroy tumour cells. Morrison used more severely immunocompromised mice that lacked NK cells. When he used the standard assay, he found the frequency of melanoma tumorigenic cells was 1 in 837,000, consistent with previous reports for melanoma. With his new assay system, it was more than 1 in 4.

What's more, the researchers examined dozens of cell-surfaces markers that might have enriched populations for stem cells. Though tumour cells varied greatly in their expression of these markers, says Morrison, none could distinguish tumorigenic from non-tumorigenic cells.

Morrison's is not the only work to question the rarity of tumour-initiating cells. Last year, Andreas Strasser at the Walter and Eliza Hall Institute in Melbourne, Australia, showed that mouse-to-mouse transplants of genetically engineered murine blood cancers4 revealed a much higher frequency of tumour-initiating cells than had been reported in human leukemia. Strasser thinks that much of the cancer stem cell work reflects the inability of human cancer cells to establish themselves in a mouse, rather than the existence of a tiny subset of cells that need to be particularly targeted by cancer drugs. “Is the cell that we have to destroy rare or is it frequent?” he asks. “That's what matters. Whatever keeps the tumours we've been studying going, it's certainly not rare.”

If cancer stem cells aren't rare, then researchers have been studying the right cells all along, says William Kaelin of the Dana-Farber Cancer Institute in Boston, Massachusetts. “What the Morrison result says is that if you change the strain of the mouse and you do some simple tricks, you get a very different answer, at least in terms of the frequency of the cell, raising the question of what you were measuring in the first place.”

Nevertheless, some researchers believe that the cells they've identified as tumour-initiating cells are special. Peter Dirks at the Hospital for Sick Children in Toronto, Ontario, who has identified cancer stem cells in the brain tumour glioblastoma5, says that he's not sure how much of a difference a more severely immunocompromised mouse would make in his experiments, as the immune response in the brain is already constrained. And Morrison's techniques to help the transplanted cells survive also wouldn't be relevant. Unlike cells from other cancers, which are injected in the mouse's skin or kidney regardless of origin, human brain tumour cells can be engrafted into the mouse brain, where they grow well. Michael Clarke of Stanford University in California has identified stem cells in breast and other cancers6. He also thinks that his results wouldn't change much if he transplanted his sorted cells into the severely immunocompromised mice used by Morrison, because he suppressed NK cell activity with immunosuppressive drugs.

Rarity isn't necessarily a defining property of a cancer stem cell. Peter Dirks, Hospital for Sick Children, Toronto, Ontario

More importantly, researchers who have identified cancer stem cells say that rarity is not the point. "Rarity isn't necessarily a defining property of a cancer stem cell," says Dirks, but even enriching for a relatively common cell could be important if those cells serve a different function than others.

The real point is that there is a subpopulation capable of reforming the tumour and that this subpopulation is resistant to standard therapies. In fact, based on cells' appearance by cell sorting, Clarke is not surprised to learn that the frequency of cancer stem cells is more common than transplantation models predict. “We've looked at 40-odd colon cancer tumours, 20 head and neck tumours, 20 breast tumours, and I can tell you the frequency of the stem-cell population by phenotype varies from a low of about 0.6% to 45 or 50%, and the average is probably 5 or 10%.” Furthermore, he and other scientists who have selected tumour-initiating cells using cell-surface markers say that these cells' behaviour is consistent with the stem cell model. Breast tumour-derived cells that lack stem-cell markers can survive for weeks in NOD-SCID mice. But unlike those with stem-cell markers, they eventually stop growing and the cell mass begins to shrink.

Melanoma might be particularly susceptible to immune surveillance — the ability of the immune system to seek out and destroy incipient cancer cells — and so a mouse's immune system could dramatically affect the assay. Furthermore, melanoma is highly metastatic, and this might also indicate a higher percentage of tumour-initiating cells than in other tumours. After all, commercially available cancer cell lines, which have been selected for aggressive growth, also form tumours at a high rate in transplantation assays.

We're suggesting that the cancer stem cell model will be true in some cancers and not in others. Sean Morrison, University of Michigan, Ann Arbor

Morrison says that the point of his research is not that all cancer cells have the potential to regenerate a tumour, but that the experiments used to identify those that do may need to be re-evaluated, particularly in solid tumours. “We're suggesting that the cancer stem cell model will be true in some cancers and not in others,” he says.

A cell by any other name

And this is where the problem of terminology rears its head. Researchers have argued as much over what the term 'cancer stem cell' means as whether the entity exists. The main source of confusion is whether the term refers to any cell capable of perpetuating the tumor or whether it means the very first cell that begins to go wrong. This 'cell of origin' is sometimes called a cancer stem cell because it is generally presumed to begin as a normal tissue stem cell or a progenitor that acquires stem-cell capacities. However, the cells capable of perpetuating a tumour, also called cancer stem cells, may have acquired several additional mutations and epigenetic modifications distinct from the cell of origin. “To name these two cells with the same name is absolutely insane; until we clean that up, there will not be any progress because people don't know what they are talking about,” says Strasser.

“Cancer stem cells are hard to identify because a practical operating definition has never been achieved,” says Scott Kern, an oncologist at John Hopkins who sees no reason for a “binary distinction” between cells that are cancer stem cells and those that are not. “Even if 100% of cells in a tumour could re-initiate a tumour, inefficiencies of the assay would make the number identified less than 100%.”

Operational definitions of cancer stem cells are feasible, however. “If you think about metastasis — a single cell going from a primary tumour to a distant site and re-creating a complex tumour — that by definition is a cancer stem cell,” says Craig Jordan, who studies leukaemia at the University of Rochester School of Medicine and Dentistry in New York.

But although scientists disagree over how distinct and small a subset of cells capable of reforming tumours might be, they tend to agree that the assays for finding such cells can be lousy. “If you really want to understand cancer, you need to take time to improve the assay, rather than taking a one-size-fits all approach,” says Morrison. Where cells are injected, what kinds of mice are used, whether cells are coinjected with other cell types or substances, and how long one waits for cells to grow can all make a big difference. As well as mouse transplantation assays, these include examination of cell phenotypes and looking at the cancers that crop up in genetically tumour-prone mice.

The current assays are also difficult to perform. A sample from a myeloma patient might only contain enough cells to inject two or three mice. While not all researchers wait this long, others say transplantation assays can take as long as six months to read out. And although it might be possible to identify normal stem cells using a standard set of markers, cancer stem cells can't be expected to be so consistent, says Glenn Merlino, who heads the cancer-modeling group at the National Cancer Institute in Bethesda, Maryland. “Everything we've learned about cancer can be summarized as 'unpredictable and wildly varying',” he says. “To go into a tumour and think that if there's a cancer that marker will always be there, I think that's naive.”

That's probably one reason why cancer stem cells are so hard to find, and why different searches for cancer stem cells can pull up different markers. “Whatever we're defining as a cancer stem cell is not very stable, so the frequency and phenotype may be changing,” says Jordan.

Applied science

Even worse, there's no way to determine whether the cells that show themselves to be capable of regenerating tumours in lab assays are the ones that actually cause cancer to recur. “No paper in the entire field of cancer stem cell biology has been able to determine which cells are actually fated to contribute to disease progression in the patient,” warns Morrison.

In the clinical setting, the only importance is 'what are the characteristics of the resistant population that appear to be able to make new cancers'? Jenny Chang, Baylor College of Medicine, Houston, Texas

For Jenny Chang, who studies breast cancer at Baylor College of Medicine in Houston, Texas, the crux of the cancer stem cell model is simply that tumours contain different populations, some of which have increased tumorigenicity. “The purist may not buy that,” she says, “but in the clinical setting, the only importance is 'what are the characteristics of the resistant population that appear to be able to make new cancers?'”

She thinks cancer stem cells may resist chemotherapy for several reasons, such as the expression of efflux pumps that eject the drugs intended to kill them. After all, stem cells would reasonably activate mechanisms that enable them to survive and divide long term. Jeremy Rich of Duke University in Durham, North Carolina, has shown that cancer stem cells in brain cancers seem to have superior mechanisms for repairing DNA damage and thus are resistant to radiation therapies7. And most cancer drugs kill rapidly proliferating cells, which might be addressing the symptoms rather than the cause of the disease. Many tissue stem cells are quiescent and divide only in response to injury; even then these cells may not themselves divide rapidly but instead produce cells that do. “We need,” says Chang, “to find new targets that are important in self-renewal.”

In pursuit of this, Chang compared tumour biopsies before and after standard chemotherapies and found a greater percentage of cells expressing CD44, a breast cancer stem cell marker, in samples collected after 12-weeks of conventional chemotherapy8. Cells from post-treatment samples showed greater propensity for self-renewal as measured by their ability to form mammospheres in culture and tumours in mice. (Interestingly, treatment with a newer 'targeted' drug, lapatanib, did not increase the frequency of tumorigenic cells.)

William Matsui and his colleagues at Johns Hopkins University School of Medicine in Baltimore, Maryland, recently presented evidence that the cells that perpetuate tumours in the patient might have stem-like properties. Using a direct biochemical assay on clinical samples, the researchers examined tumours from 268 pancreatic cancer patients for expression of aldehyde dehydrogenase (ALDH), an enzyme highly expressed in blood-forming and neural stem cells. Matsui found that patients with tumours whose invasive margin contained this marker survived, on average, four months less than those whose tumours did not. Other groups have also reported a correlation with stem cell markers or gene expression and shorter life expectancy.9,10,11.

Matsui's team also compared primary pancreatic cancer lesions and distant metastases and found that whereas ALDH was at very low concentrations in the former, it was abundant in the latter. The cells in the metastases also looked different; they had what Matsui calls a “semi-mesenchymal” phenotype, meaning that they were more mobile and adhered less well to each other than is typical for epithelial cells.

This observation is consistent with the process called the epithelial-to-mesenchymal transition described by Robert Weinberg, a cancer biologist at the Massachusetts Institute of Technology, which enables cells to become invasive and metastatic and also creates cells with stem-cell properties12. Weinberg does not subscribe to what he calls the 'orthodox stem cell hypothesis', but he says that tumours certainly contain stem-like cells. Although it's hard to know precisely what goes wrong first, Weinberg believes that a diseased microenvironment delivers signals that cause certain at-risk cells to go astray.

This might be another reason why studies identifying markers for cancer stem cells have given inconsistent results. Cells within a tumour are not irrevocably in stem-cell and non-stem cell states, Weinberg says. The microenvironment of the tumour could cause non-stem cells to revert to cancer stem cells. How this idea can be exploited for therapy is currently unclear, he says. “The more we study cancer stem cells, the more we find that they are very similar to normal stem cells.”

Hard questions

Of course, what researchers hope to find are differences between cancer cells and normal cells that could be used to cure cancer. A better understanding of a tissue's healthy stem cells is likely to be key to figuring out how to deal with its cancer stem cells, says John Dick at the University of Toronto, who credits his discovery of leukaemia stem cells1 to the fact that the assays for blood-forming stem cells and the knowledge of their biology were already so well worked out.

Already, modulators of stem-cell pathways such as the signal proteins Hedgehog and Notch are in clinical trials at Genentech and Merck, respectively. Matsui hopes that the stem-cell pathways that cancer cells rely on for self-renewal could be used to “re-phylogenize” tumours, although he acknowledges it's almost too early to imagine such a possibility. If a cancer has corrupted a stem-cell pathway that is important only during embryonic development, for instance, shutting that pathway down could stop the cancer — with few side effects for the adult patient.

Then there are questions about the role of the stem-cell niche. What signals sent between the environment and a cancer cell cause the epithelial-to-mesenchymal transition? If the tumour environment is incapable of forcing cells down the lineages for proper differentiation, the cells within it might have a chance of reverting into stem cells. In fact, the ability to reprogram cells to pluripotency in the laboratory has led several researchers to speculate that perhaps the tumour-cell niche reprograms cells to a diseased, stem-like state. Forcing cancer cells to differentiate could therefore be a powerful therapeutic approach.

The cancer stem cell concept — maybe its greatest utility is just a better perspective to look at the cancers. You have stem cell haters and stem cell lovers, and both groups are just entirely too dogmatic. William Matsui, Johns Hopkins University, Baltimore, Maryland

The tools to explore the genomic and proteomic diversity within individual cells of a tumour are evolving and becoming more powerful, which might make it easier to delineate how a cancer changes over time from a benign collection of aberrant cells to a deadly population of aggressive cells. Work with leukaemia indicates that as cancers become more aggressive, the percentage of stem- like cells increases13. But even getting enough samples of late-stage human cancers to study is difficult. Even if one could control for the effects of chemotherapy, monitoring human tumours over time is not feasible.

New techniques for rapid individual genome sequencing and mutation detection could help address the issue of what cells begat what. “You could go into a specimen, laser capture single cells that look more undifferentiated, and could be a stem cell; you could sequence their genomes and you could see if they all have the same oncogenic mutations,” says Strasser. This could be used to figure out how the variety of different cells with different genotypes or cells with different tumorigenic potential found in tumours are generated, which could be a boon, says Dirks. “We haven't yet laid down the genetic heterogeneity of solid tumours with the cancer stem cell model.”

If that can be done, maybe it will help to reconcile some of the differences, and animosity, that the idea of cancer stem cells has generated. "The cancer stem cell concept — maybe its greatest utility is just a better perspective to look at the cancers," says Matsui. “You have stem cell haters and stem cell lovers, and both groups are just entirely too dogmatic.”

Related articles

Stem cells not all to blame

Cancer stem cell sightings and slightings