By Christopher Vaughan
Illustration by Jeffrey Fisher
“It’s no fun to manage a patient’s death,” says Stanford physician and researcher Ravi Majeti, MD, PhD, with grim and deliberate understatement. As a physician specializing in treating leukemia and lymphoma, Majeti is often required to perform that task with as much compassion and empathy as possible. In the case of acute blood cancers, even in young patients who undergo aggressive treatment, the odds of surviving more than five years is less than 50-50. For those over age 65, Majeti says, five-year survival rates are about 10 percent.
What’s particularly frustrating is that these atrocious odds haven’t changed in decades, says Majeti, an assistant professor of hematology and a member of the Stanford Institute for Stem Cell Biology and Regenerative Medicine and the Stanford Cancer Institute. “The primary drugs and the combinations we use have not changed in 30 years. They have not,” he says. If a patient fails a first round of chemotherapy, it is difficult to halt the malignant blood stem cells from multiplying uncontrollably, crowding out nearly all the normal blood cells that protect us from infection, carry oxygen and clot to keep us from hemorrhaging. Lung infections are usually the ultimate cause of death, he says.
In large part because of his frustration with these terrible and unchanging statistics, Majeti spends much of his time in the laboratory, looking for new ways to shift the odds in the patients’ favor. Some of his work involves taking samples of leukemia cells from patients in the hospital and putting them into mice to test new therapies.
So it was with some small hope, on a day in November 2007, that Majeti tested a new antibody discovered in the laboratory of institute director Irving Weissman, MD. Majeti and MD/PhD student Mark Chao first injected a mouse with aggressive human acute myelogenous leukemia cells. This particular leukemia had, in fact, eventually killed the patient who donated a blood sample for research. Injected into a mouse, leukemic blood cells normally do the same thing they do in humans — multiply out of control until they kill the host. What happened next was one of the most astonishing moments in Majeti’s scientific career.
“For the first test, we were just guessing a dosage and hoping we could observe some small effect,” Majeti says. But there was no small effect — it was huge. A day after injecting the antibody, Majeti and Chao couldn’t observe any cancer cells in the mouse at all. “One dose, one day, and the cancer was gone,” Majeti says. At first Chao thought he had done something wrong, but there was no mistake. The same lethal cancer cells that could not be stopped in the human patient, the same cancer that was so irritatingly resistant to everything the doctors could throw at it in the clinic, had simply disappeared in the mouse after being exposed to a single dose of experimental antibody.
As dramatic as that experiment was, further research kept producing new amazements, suggesting that the applications of this antibody to cancer therapy are far broader and more powerful than anyone dared hope. The experimental antibody that Weissman and his collaborators discovered blocks a cell protein called CD47 — a cellular cloaking device that offers cancer safe passage from immune cells that eat damaged cells or foreign matter.
Investigation into the role of CD47 began slowly 14 years ago in Weissman’s laboratory, but like a snowball kicked off a hilltop, it has picked up speed and mass as it has rolled along. It now seems on a course to blast through the traditional cancer treatment community. As the researchers prepare to test the treatment in humans, they have dared to hope that they’re on the trail of something many have dreamed about but most had begun to think impossible: a single therapy that uses our own immune system to effectively attack all cancers with almost no side effects.
It is the nature of life that things will go wrong eventually. Our cellular software, our DNA, can get damaged in many ways. Eventually “bugs” in that software accumulate, and cells stop following instructions written and revised over billions of years to make sure they do their proper jobs. One result of this process is cancer — cells that are supposed to behave within the rules of the body’s decorum begin breaking those rules and multiplying out of control.
In 1998, Weissman and his postdoc David Traver, PhD, were crossbreeding mice with various genes that block programmed suicide in cells that have been damaged, genes that are known to be associated with cancer. They created a breed that was particularly prone to developing leukemia, then analyzed all the genes being manufactured (or “expressed”) by the blood-forming cells in these mice. “The first gene that we saw that was overexpressed in the mice that got leukemia was CD47,” Weissman says. In fact, it turned out that high levels of CD47 were common in every kind of leukemia, in mice and humans both.
But no one knew what CD47 did. Then, two years later, a group in Sweden discovered that one role of the CD47 protein was to act as an age marker on red blood cells. They discovered that red blood cells start out with a lot of CD47 on their cell surface and slowly lose the protein as they age. At a certain level, the dearth of CD47 allows macrophages to eat the aging red blood cells, thus making way for younger red blood cells and a refreshed blood supply. CD47 thereafter became known as a “don’t eat me” signal to the macrophages.
For Weissman, the Swedish work and the work in his lab suggested an explanation for leukemia cells’ invincibility. Leukemia, which is a disease of excess production of specific sorts of blood cells, always features high levels of CD47. And CD47 naturally protects red blood cells from being cleared away by the immune system. Could blood cancer cells be boosting levels of CD47 to protect themselves from being consumed by macrophages?
Oddly enough, no one in Weissman’s lab was interested in looking for an answer to that question. Principal investigators with large labs generally don’t conduct experiments themselves, relying instead on an army of postdocs and students to do research under their direction. Weissman, unlike many principle investigators, doesn’t order people in his lab to carry out specific research, preferring instead to let them pick projects that interest them. “I’ve found over my career that the people I had to direct more closely never developed as scientists,” Weissman says. “Whereas those people who take on problems and figure out how to approach them go on to become accomplished scientists.”
So for three years, Weissman says, he found himself saying, “Come on you guys, how could it be more apparent? Every mouse leukemia has CD47, and CD47 is a ‘don’t eat me’ signal to macrophages. It has got to be important.”
Finally, in 2003, an MD/PhD student named Siddhartha Jaiswal took an interest. Jaiswal showed that human leukemias, like those in mice, also have elevated levels of CD47 on the leukemic cells. He also found another intriguing link with cancer. When blood-forming stem cells leave the bone marrow to move to another site, they dial up their production of CD47 to protect themselves against macrophages. The way these cells do this looks a lot like the way metastatic cancer cells move around the body and invade tissues. These and other experiments performed by Jaiswal showed that CD47 could indeed play an important role in blood cancers. “But showing that it’s possible is different than showing that’s what happens in real life,” says Jaiswal, who just completed his second year of residency in pathology in Boston.
‘we started out small, but in the end we were giving mice really large doses of anti-cd47 antibody, and the mice were just fine.’
If leukemia cells can use CD47 to protect themselves against macrophages, then the obvious next question is whether one can reverse that process. Could doctors get macrophages to eat leukemia cells by blocking the CD47 “don’t eat me” signal? Majeti, who at that time was still a postdoctoral researcher in the Weissman lab, decided he wanted to take on this project. Majeti identified an anti-CD47 antibody that would block the “don’t eat me” signal to the macrophages. He and Chao then mixed labeled leukemia cells with macrophages and the antibody. Under the microscope, they could see the marked leukemia cells inside the immune cells. In some cases, he could see macrophage cells in the act of eating the cancer cells. Blocking CD47 worked, at least in glassware. After these early tests, the two of them performed the dramatic experiments that showed the antibody worked in mice, too.
When Majeti and his colleagues conducted a full series of experiments with human acute myelogenous leukemia in mice, they were able to totally eliminate the cancers in a majority of the mice. Ash Alizadeh, MD, PhD, another postdoc in the Weissman lab, found that CD47 was also present on non-Hodgkin’s lymphoma and performed a similar experiment. He, Majeti and Chao showed that the anti-CD47 antibody, combined with an FDA-approved antibody called rituximab, would eliminate aggressive human non-Hodgkins lymphoma in mice. Rituximab by itself does not.
Killing cancer cells is not that hard. A little household bleach will annihilate the worst cancer doctors have ever encountered. But of course you can’t treat cancer with bleach. The trick to all cancer treatments is to harass, inhibit, contain, cut out, beat down and, you hope, kill cancer cells while simultaneously doing as little harm as possible to normal cells in the body. This is especially hard because cancer cells and normal cells are so closely related.
CD47 is not found only on cancer cells. The protein is also on many normal cells, and the obvious danger is that an anti-CD47 antibody would strip away the protective protein cloak from normal cells. Stephen Willingham, PhD, a postdoctoral scholar in the Weissman lab, took on the task of finding out if the use of the antibody would cause macrophage cells to attack normal tissue. If it did, that would rule out its use as a therapy, no matter how well it eliminated cancer cells.
“The experiments in mice are impressive, but they are rigged in a way because we are using a human form of the antibody against a human cancer, but we are doing it in a mouse,” Willingham says. “The humanized antibodies don’t attach to the mouse’s cells, so the mouse’s immune system won’t attack its own healthy tissue.”
A more fitting test of the safety of the therapy would be to use the mouse version of the anti-CD47 antibody in mice without cancer, which is exactly what Willingham did next. Luckily, these tests showed no major effects on normal tissues in the mouse.
“We started out small, but in the end we were giving mice really large doses of anti-CD47 antibody, and the mice were just fine,” says Willingham. The only change was a temporary anemia as the mice’s red blood cell count fell (because declining CD47 is a sign of age in red blood cells, blocking CD47 makes the young red blood cells look old to the immune system, which eliminates them). But within days, the level of red blood cells in the mice’s bloodstream was back to normal.
“It was actually amazing to me how little effect there was on normal tissue,” Willingham says.
How could it be that blocking CD47 is so devastating against cancer but affects normal cells so little? If healthy cells also have CD47, why doesn’t blocking CD47 also lead to their destruction? The Stanford researchers hypothesized CD47 can’t be the whole story. Cancer cells must also have an “eat me” signal that normal cells do not carry. “It wouldn’t be likely that killing cells was the default action of the immune system,” Majeti says.
The idea that cancer cells would carry the seeds of their own destruction is not really surprising. Cells have many ways of signaling that not all is well inside them. For instance, specialized proteins inside cells carry bits and pieces of what they find to the cell surface and show it to circulating immune cells. If something is wrong inside the cell, the immune cells can then spot it. It’s a bit like a scenario in which parents sit outside their house chatting while the kids play inside. When the kids occasionally come to the window to show them a toy or game they are playing with, the parents know all is well. If a child comes to the window holding a severed human arm, the parents will know something is terribly amiss.
The genetic changes involved in making a cell cancerous disrupt its normal function, making it more likely that the cell will present signs of abnormality, the “eat me” signals that mark it for destruction. Warning signs like these actually make our bodies fairly adept at fighting errant cells. It’s likely that every one of us has had cells that are precancerous or cancerous, and that these cells have been effectively dealt with by our body’s defenses.
Majeti, Chao and Rachel Weissman-Tsukamoto — a high school student who is also Weissman’s daughter — took on the search for an “eat me” signal. They began to focus on a molecule called calreticulin as a possible “eat me” signal because other researchers had shown that it worked together with CD47 to regulate cell suicide. Indeed, the Stanford scientists found calreticulin on a variety of cancers, including several leukemias, non-Hodgkin’s lymphoma and bladder, brain and ovarian cancers.
“Our research demonstrates that the reason blocking the CD47 ‘don’t eat me’ signal works to kill cancer is that leukemias, lymphomas and many solid tumors also display an ‘eat me’ signal,” says Weissman. “The research also shows that most normal cell populations don’t display calreticulin and are therefore not depleted when we expose them to a blocking anti-CD47 antibody.”
Understanding how calreticulin and CD47 balance out each other’s influence in controlling how the immune system reacts to cancer is important because it can affect how anti-CD47 antibodies are used as a therapy. “If calreticulin is displayed in response to cell damage, you might not want to use anti-CD47 immediately after chemotherapy or radiation,” says Majeti. “These treatments can cause damage to normal cells, which might make them vulnerable to macrophage attack when CD47 is blocked.”
One of the many frustrations of cancer is that each type can be so different. New drugs that seem to be effective against one type of cancer turn out to be ineffective against other types. Drugs like Gleevec are miraculous therapies for chronic myelogenous leukemia, but ineffective against acute lymphocytic leukemia. Researchers had high hopes for drugs that block the growth of blood vessels that tumors need, but one such drug, Avastin, while effective against colon cancer, turns out to be ineffective against breast cancer.
As Majeti and his colleagues investigated CD47’s role on leukemias, another team in the Weissman lab began to look at solid tumors. Willingham and Jens-Peter Volkmer, MD, a urologist who came from Germany to study stem cells in the lab, began looking at cancer tumor samples collected from patients at the hospital. They were astounded to find CD47 everywhere they looked.
“When Stephen and I first started, we thought we might get lucky and find CD47 on a few tumors, but nobody expected that every kind of cancer we looked at would have high expression of CD47,” Volkmer says. The list of cancers with high CD47 levels ultimately reached 20 and could still be growing. The natural next step was to find out if the methods that were so successful in treating human leukemias could be replicated with solid tumors.
Using collaborative relationships Weissman had built with oncologists at Stanford Hospital over a decade, Volkmer and Willingham obtained biopsies of human cancers, embedded those cancers in laboratory mice and then treated half the mice with the anti-CD47 antibody. As expected, in the untreated control group the samples of deadly, aggressive cancers of the breast, ovaries, colon, bladder, brain, liver and prostate grew rapidly. Special imaging shows how the cancers whipped like wildfire in the mice, starting as one dot of color but growing inexorably over weeks until they had spread throughout their bodies. But in the mice treated with the anti-CD47 antibody, clusters of cancer cells shrank or even disappeared.
“If the tumors were large, then we could shrink them, but if the tumors were small, the antibody could eliminate them altogether,” Volkmer says.
Even more dramatic, Willingham and Volkmer, along with pathology professor Matt van de Rijn, MD, PhD, showed that the anti-CD47 antibody could stop cancer from metastasizing, or spreading from the original tumor. This finding is highly important because tumors can most often be cut out or controlled by focused radiation therapy when they restrict themselves to one site in the body. Many cancers become really deadly only once they start spreading the seeds of cancer throughout the body.
“I think the most amazing moment in our research was when we saw that anti-CD47 antibody could stop the metastatic spread of cancers, and even treat cancers that had already metastasized,” says Willingham.
As amazing as it seems to the researchers, the CD47-blocking therapy looks as though it could be useful in combating almost every type of cancer (although not every individual cancer — a very small number of cancer samples seem to use other, unknown methods to escape macrophages). It may turn out to be most effective when combined with antibodies that reinforce the positive “eat me” signals from cell surface proteins like calreticulin. “The nice thing is that anti-CD47 antibodies should work with, and boost the effectiveness of, other treatments that use the immune system to fight cancer,” says Volkmer.
With the help of a $20 million grant from the voter-funded California Institute for Regenerative Medicine, the Stanford researchers are pushing ahead with plans to begin human clinical trials of the therapy in late 2013 or early 2014. This will be none too soon for many current cancer patients and even the researchers, many of whom spend time treating patients when they are not conducting lab research.
“We are tremendously excited about the potential of CD47 antibodies for improving the lives of our patients,” says Beverly Mitchell, MD, PhD, director of the Stanford Cancer Institute.
Majeti remembers many of the patients who graciously donated their cancer cells for research, especially those patients whom he could not cure and who succumbed to their illness. He is sensitive to the cruel irony that in mice, he is able to defeat some of the very same leukemic cells that he could not vanquish in his patients.
“Working in the clinic is a very motivating thing,” Majeti says. “When I find myself in a situation that can’t be addressed clinically, I say, ‘We have got to get back to the lab and get on this.’”
E-mail Christopher Vaughan
The following cancers have been shown to make higher than normal levels of CD47.
Cancers listed in red type are those for which anti-CD47 antibodies have been shown to be effective in animal models.
Head and neck cancer
B-Acute lymphoblastic leukemia
T-Acute lymphoblastic leukemia
Acute myeloid leukemia
Chronic myeloid leukemia