Tumour cells must hide from the immune system to form a cancer. Most research so far shows how malignant cells avoid the patrols of the B cells, T cells and natural killer cells of the adaptive immune system. But new studies suggest that cancer cells may also have to sneak past the innate immune system and avoid being eaten by macrophages. To do this, they borrow the same defence mechanism used by normal stem cells.

When normal haematopoietic stem cells (HSCs) leave the bone marrow to colonize new niches elsewhere, they move through blood vessels full of macrophages ready to eliminate errant cells. A team led by Irving Weissman of Stanford University in California suspected that HSCs may have to protect themselves from macrophages as they move in and out of the bloodstream and set out to establish how they do so.

Using two different experimental approaches, the researchers observed a marked yet transient increase in the levels of a protein called CD47 on the surface of HSCs mobilized from the bone marrow1. When they transplanted HSCs into mice whose own blood stem cells had been depleted, the amount of CD47 in the injected HSCs correlated with their ability to maintain engraftment. However, if the host mice were treated with drugs to eliminate macrophages, even HSCs with little CD47 could engraft. Perhaps, the researchers thought, transient upregulation of CD47 in HSCs helps them evade macrophages. After all, CD47 is already known to interact with macrophages and inhibit phagocytosis.

Weissman's team reasoned that leukaemic cells may employ the same strategy to evade macrophages. “We knew that nearly every mouse model of leukaemia and, more recently, that most human leukaemia samples had elevated CD47,” says Siddhartha Jaiswal, the lead author of the one of two studies published in Cell. His colleague Ash Alizadeh then found that higher CD47 levels correlated with a poorer clinical prognosis among leukaemia patients2. And a human leukaemia cell line with lower CD47 expression, normally unable to engraft in mice, would do so if forced to express high levels of CD47 or if macrophages had been depleted from the animals.

Weissman was shocked by the fact that “something as fundamental as the dominant role of macrophages in tumour surveillance had not been revealed by most experimentalists”. He was also surprised by the “beautiful 'on-off' defence against macrophage killing” used by normal HSCs “and how every tumour that we looked at carefully have acquired the on and disabled the off”.

Next, Mark Chao and Ravindra Majeti in the Weissman lab engrafted mice with leukaemia stem cells and later injected them with anti-CD47 or control antibodies. The leukaemic cells accumulated in the control group but rapidly disappeared from the blood and bone marrow of mice treated with the CD47-blocking antibodies. Additional experiments indicated that the leukaemic cells had been engulfed by macrophages2.

Kornelia Polyak, a breast cancer specialist at the Dana-Farber Cancer Institute in Boston, believes that CD47 antibodies should be carefully examined for both beneficial and unwanted effects. CD47 is involved in multiple other interactions, she explains, so researchers hoping to bring CD47 to the clinic must hunt vigilantly for potential side effects. On the other hand, she says, at least one study has shown decreased bone metastases in the tumours of mice that lack CD47 expression3. Incredibly, Weissman found that targeting CD47 did not seem to affect the healthy blood stem cells in the animal. “We need to find out if it's generally true that antibodies against CD47 are not toxic in mice and if this is true for primates as well,” he says. He speculates that leukaemia cells may show stress signals that attract macrophages, making them particularly vulnerable to blocking their CD47 'don't eat me' signal. Jaiswal adds that “it remains to be seen if leukaemic cells co-opt the same regulatory mechanisms or novel ones” to upregulate CD47.

The Weissman group also recapitulated their observations in vitro. CD47-blocking antibodies allowed macrophages in culture to engulf not only leukaemic cells, but also cells from a human bladder cancer that display high levels of CD47 and are otherwise resistant to phagocytosis4. Weissman suspects that overexpression of CD47 may therefore be a widely used defence among neoplastic stem cells, though it's unclear whether cancer cells would need to protect themselves within tissues or only while in circulation. For instance, Keith Chan in the Weissman lab is now looking into whether CD47 is more highly expressed in normal bladder epithelial stem cells.

Like Weissman, cancer researcher Jerry Adams is most intrigued that targeting CD47 would affect leukaemia stem cells but not their normal counterparts. However, Adams, who works at the Walter and Eliza Hall Institute of Medical Research in Melbourne, Australia, thinks that researchers are often too quick to see cancer stem cells. Groups studying putative leukaemic stem cells could instead be homing in on human cells that are particularly good at engrafting in mice, he explains. Xenograft experiments “may have little to do with the cells' capacity to maintain the growth of leukaemia in a patient”. He points out that CD47 was present on all of the leukaemia cells and, in fact, one of the reasons leukaemia stem cells seem like a rare subpopulation could be that human CD47 proteins are not particularly good at conveying the “don't eat me” signal to mouse macrophages.

Whatever CD47 implies for stem cells, says Adams, it merits further study as a potential target against cancer. “Putting the stem cell issue aside,” he concludes, “these are important and interesting results.”