Following stints as a bass player, a dog trainer, a carpenter, a biology student, and a Playboy bunny, Polly Matzinger had taken a job as a cocktail waitress in a restaurant near the University of California at Davis. One evening she was serving a couple of scientists, one of whom was Robert Swampy Schwab, then chairman of wildlife and fisheries at Davis and now a professor emeritus. Schwab and his colleague were talking about a recent experiment when their waitress pointed out a few errors in their reasoning. We said, ‘Holy cow, who’s that?’
Matzinger and Schwab struck up an acquaintance. She had a penetrating mind, says Schwab. She could spot the problem, the picture, and half the solution. There are people who are suited for research, and you can spot them. Says Matzinger: He started a nine-month campaign. He said, ‘You’re a scientist! You should be a scientist.’ And I said, ‘Go away--it’s going to get boring, like any other job. Everything gets boring after a while.’ And then he’d come back at closing time and we’d go to an all-night Denny’s and talk science.
Those talks, says Matzinger, helped her decide to return to school and become a scientist. A few years later, when Schwab got wind that Matzinger was receiving funding to do immunology research, Schwab’s colleague dismissed her as a charlatan. He said, ‘She’s just a barmaid,’ Schwab remembers. But I said, ‘No, I think it’s true. If more barmaids had minds like hers, you and I would be out of a job.’
Nearly 20 years later, the erstwhile barmaid is now chief of an immunology lab at the National Institutes of Health, the author of dozens of scientific papers, and the creator of an award-winning film about the immune system. But she is still poking holes in the reasoning of scientists. In fact, Matzinger thinks there is a big hole right in the heart of immunology. For almost 50 years, immunologists have thought they understood the fundamental job of the immune system: it was to recognize the difference between self and nonself, and to defend the body against the alien. Matzinger disagrees. She says the immune system doesn’t bother with a sense of self--it just has a clear sense of danger.
The self/nonself model dates from the end of World War II, and some historians have argued that it--and the wide support it has enjoyed-- may have been shaped by cold war xenophobia. But the man whose research laid the experimental foundations for the model was a prophet of immunologic tolerance. One day in 1944, Peter Medawar, then a zoologist at Oxford--and not yet Sir Peter the Nobel Prize winner--was sitting in his garden with his family when a Royal Air Force bomber buzzed over at low altitude and crashed a couple hundred yards down the road. A badly burned pilot was pulled from the wreckage, and the surgeon at the local hospital, a colleague of Medawar’s, asked for Medawar’s help in treating the victim with skin grafts.
By then it was already known that the body, though it will accept grafts of its own skin, will swiftly reject skin grafts from another person. The airman was nevertheless treated with foreign grafts, in hopes of giving his own skin a chance to grow back. Medawar and his colleague thereupon made an interesting observation: second grafts from a particular donor were rejected faster than the first ones had been, and faster than grafts from fresh donors. The implication of that observation, which Medawar went on to confirm with controlled experiments in rabbits, was clear: the rejection was immunologic. The airman’s immune system was rejecting foreign skin as if it were a virus, and having seen it once, was remembering it the next time--just as an immune system would remember and be prepared for, say, chicken pox.
That observation didn’t help the airman any (although he did recover), but it set Medawar on a course of research that not only made him an immunologist but also transformed immunology. If rejection of skin grafts was an immunologic phenomenon, it was natural to ask, as Medawar did, whether it might be possible to make the immune system more tolerant. Experiments in cattle twins had already suggested such tolerance was possible. Because cattle twins share a placenta, their blood-forming cells mingle as they develop in the womb. As a result, even fraternal cattle twins--which are not genetically identical--can accept skin grafts from each other. Somehow the early exposure to the foreign blood cells trains the immune system to accept the foreign tissue as self.
In a seminal series of experiments, Medawar showed that such tolerance could be induced artificially. He took pregnant brown mice and injected their fetuses with cells from a white mouse. When the brown offspring grew up, he grafted patches of white skin onto them. The white skin was not rejected; instead it established healthy islands of white in a sea of brown fur.
If Medawar’s experiments with rabbits had suggested that the immune system distinguishes self from nonself, his experiments with mice now suggested that this sense of self, rather than being fixed in our genetic makeup, is learned while the immune system is still developing in the fetus--which is why it could be modified at that stage by exposure to foreign cells. An Australian virologist named Frank Macfarlane Burnet had predicted as much; it was he who provided the first theoretical formulation of the self/nonself model. In 1960, Medawar and Burnet shared the Nobel Prize in medicine for their pioneering work, which paved the way for transplantation biology. No organ transplants are done today without the careful matching of tissues--that is, without the matching of immunologic selves--and the administering of drugs that suppress the immune response against the foreign organ.
The self/nonself model, however, has gotten considerably more sophisticated since Medawar and Burnet’s day. Immunologists now know, for instance, that the responsibility for rejecting foreign skin grafts falls on T cells--the brokers of the immune response, which must be activated against a specific target if the response is not to grind to a halt. There is even a standard theory of what it is on foreign skin that T cells recognize as foreign: a distinctive molecular badge called MHC, for major histocompatibility complex. MHC is a set of proteins, found on the surface of nearly all body cells, whose function is to provide a window display of a cell’s contents--both normal proteins and, if the cell is infected, viral ones. But because the structure of MHC itself is different from person to person, it is considered a marker for the self. Thus the immune system sees cells that bear viral proteins (displayed by their own body’s MHC) as foreign. It sees skin grafts as foreign because they contain cells displaying a nonself MHC.
How do T cells learn what is self and what is nonself? The long- standing explanation is that immature T cells, which are born in the bone marrow, undergo a clever audition as they migrate from the bone marrow and pass through the thymus, a small gray gland that lies beneath the breastbone. What the T cells encounter there are protein-MHC complexes on the surface of thymus cells, and this surface decoration is pretty similar to the surfaces of most cells in the body. Maturing T cells that respond to any of those proteins, the theory goes--and there will inevitably be some, as T cells are manufactured by the bucketload--are automatically killed off. Only T cells that don’t respond to self proteins will survive and migrate out into the bloodstream and around the body. There they can attack anything nonself--such as a skin graft.
Although to some extent this winnowing of T cells may go on throughout life (just as a person’s psychological sense of self is always being refined), the bulk of it is thought to occur during early development. Moreover, during that formative stage, young T cells cannot be activated. That is supposedly why Medawar’s newborn mice did not reject foreign blood cells. Medawar, of course, knew nothing of T cells and had only an inkling of MHC at the time he did those experiments. But his and Burnet’s fundamental insight--that the immune system’s job is to distinguish self from nonself, and that it learns what is self largely during fetal development--has survived largely unchallenged. Except that Polly Matzinger is now challenging it.
In his autobiography, Memoir of a Thinking Radish, Medawar described the effect of the bomber crash on his development as a scientist. I understood now, he wrote, recalling his carefree days as a young Oxford don, how much of my time had been wasted on unimportant projects, intellectual pastimes, and reveries. Thereafter, he was to devote most of his energy to the problem of self/nonself discrimination that Burnet had described. Matzinger’s career path was also indirect, though it required something less than an airplane crash to get her into immunology. After Robert Schwab pointed her back to school, she ended up a graduate student in biology at the University of California at San Diego. To fulfill her Ph.D. requirements, Matzinger had to design an experiment, and she was looking for an idea to work on. One day in 1976 she was playing chess with a friend, Salk Institute immunologist Rodney Langman, when he suggested she explore how it is that--as had then just been discovered--T cells become so selective in what they respond to. It was in the middle of an argument with Rod, Matzinger remembers, that I had my first creative scientific thought.
But the more Matzinger learned about the theoretical foundation for her chosen field, the more dissatisfied she became. Change is a hallmark of life, she says, and yet models based on self/nonself discrimination didn’t seem to explain why our immune systems don’t kill us off when our body changes. What happens when we go through puberty? When a woman who was never breast-fed as a baby has her first child and starts lactating, why isn’t the lactating breast rejected? Nor were changes in the body the only thing the self/nonself model had trouble accommodating. Why, for instance, don’t mature T cells routinely attack proteins they’re never exposed to in the thymus, such as the ones on cells in the skin, brain, and liver?
Some researchers had already developed the germ of an answer. It takes more to activate a T cell, they hypothesized, than binding to a protein it recognizes on the surface of another cell. Somehow--and researchers differed on the details--it won’t get activated unless it also receives a second signal. The second signal might come from molecules secreted or borne by the same cell, or it might come from a different cell. Regardless of the mechanism, the cornerstone of this model is that only when that second signal is present will the T cell multiply and begin to attack.
Conversely, any time a T cell gets the first signal but not the second, it will be inactivated. This is how potentially self-attacking T cells get winnowed out in the thymus, but the winnowing continues with T cells that make it out of the thymus. As T cells circulate through the bloodstream, they are exposed to proteins they never saw in the thymus, on the surface of skin cells, say. When they bind to such proteins in the absence of the second signal, they die. But if they do get the second signal, they are activated and head for a lymph node--the fire stations of the immune system--and activate other immune cells.
But what is the nature of the second signal? There has never been a good answer. The current consensus, though, is that it comes from a class of cells called antigen-presenting cells, or APCs. An antigen is a molecule that an immune cell can bind to, and antigen-presenting cells include macrophages and dendritic cells. Macrophages are scavenger cells; they migrate through tissue, eating up dead cells--but also bacteria, bits of which they display on their surface, where T cells can detect them. Dendritic cells are the most recently discovered APCs and still the most mysterious. Thought to be our bulwark against viral infection, they are scattered about in the lymph tissue, as well as throughout the other tissues of the body.
The APCs produce some kind of molecule on their surface that can activate T cells: that is the second signal. When a T cell binds to such an APC, the T cell multiplies and launches an attack on cells that display the telltale protein-MHC complex--cells infected by a virus, foreign skin cells, or whatever. For a brief period, activated T cells can kill those cells without additional stimulation; then they will quiet down. But as long as the virus is being made--and is activating APCs--these T cells will continue to be reactivated. Ultimately, after the virus is eliminated, virus-specific T cells will quiet down, and these cells linger as memory cells. They require additional provocation to kill again but will do so much faster the next time they encounter the same pathogen--or as Medawar observed, the same foreign skin.
The two-signal idea sounds neat enough. But as Matzinger points out, it begs an important question: How exactly do APCs avoid launching an autoimmune response? Unlike T cells, they haven’t gone through a winnowing process; they can’t tell which antigens are self and which are nonself. Macrophages, for example, are constantly gobbling up dead tissue cells. If they were thereupon to activate T cells, we’d face a continual immune assault on our own tissues. That’s been a major sticking point for 20 years, says Matzinger. If APCs can’t discriminate between self and nonself, then they couldn’t be the cells that control the immune response. On the other hand, if you’ve concluded that telling self from nonself isn’t what the immune system does, then the indiscriminate nature of APCs is no problem at all.
Matzinger did not come to that conclusion at the outset of her career. For a while, as she gradually became absorbed into the immunologic establishment, she buried her doubts about the self/nonself model. Then in 1989, a young scientist named Ephraim Fuchs came to work as a postdoc in a neighboring National Institutes of Health lab. Fuchs, now an oncologist at Johns Hopkins, had the same initial doubts that had worried Matzinger a decade earlier. By then I had my old-fogy glasses on, Matzinger recalls. But here was an intelligent, highly trained person, and he was having the same problems with the model that I had had. He didn’t think the immune system could discriminate between self and nonself. And he reawakened it all.
Fuchs reminded Matzinger of all the evidence indicating that if the immune system is designed to keep out foreigners, it is a rather lax border guard. Mothers don’t reject fetuses, not to mention their own lactating breasts; none of us attack the food we eat, rife with foreign antigens as it is; nor do we reject the millions of bacterial organisms that colonize our gut. Indeed, some of those bugs make the vitamin K that helps our blood clot. In general, Fuchs argued, infection is not always a bad thing; a virus, for instance, might provide us with helpful genes we would otherwise never acquire. In the long march of evolution, an immune system that took a wait-and-see attitude might be more advantageous than one that summarily evicted all invaders.
Which left Matzinger and Fuchs to ponder this question: If the immune system doesn’t discriminate between self and nonself, what does it do? And that’s when we came up with the idea that an immune system that discriminates dangerous from nondangerous would be much more sensible than a system that discriminates self from nonself, Matzinger recalls. The only thing left to do then was to figure out how it could do that.
The answer came to Matzinger in the bathtub one day. A natural way for the immune system to detect danger, she realized, would be for it to detect when tissue cells somewhere in the body were being subjected to horrible deaths. Cell murder is different from normal cell death. When a cell dies normally, it shrivels up and displays signals that invite a macrophage to eat it. The dead cell’s contents are never released into the extracellular environment. But when cells are infected or otherwise killed, they leak, spilling their insides all over the place. One could hardly ask for a better signal that the body was under attack.
Now all Matzinger needed was some kind of sentry that could detect that alarm. The answer came while Matzinger--who still trains prizewinning border collies--was working with one of her dogs. During a break in a training session, Matzinger sat down and watched her pupil curl up with its nose under its tail and go to sleep. She thought: I need a sheepdog that can wake up when wolves attack and the sheep--the injured cells--start bleating. Dendritic cells, the least-known type of antigen presenters, were tailor-made for this role of immune sentry. All tissues are laced with them; their long, fingerlike extensions--the dendrites-- reach into every cranny of the body, forming a lacy network. Under normal conditions they rest quietly. But when dendritic cells are activated by whatever it is that activates them, they get moving: they literally crawl through the tissue. And their typical destination is a lymph node.
Soon the whole thing seemed clear to Matzinger. Maybe what activates dendritic cells, she thought, is danger--danger as signaled by prematurely dying cells floating around the tissue. Maybe they provide the signal that prompts dendritic cells to sound the alarm. The scenario wasn’t hard to imagine. Somewhere in a piece of tissue, a cell dies a nasty death- -perhaps it is a skin cell burst open by a chicken pox infection. Viral proteins and skin proteins both spill out of the cell. A nearby dendritic cell, sensing the danger, somehow becomes activated and begins drinking up the dying cell’s contents. Then it starts crawling toward the nearest lymph node.
There, like a good APC, the dendritic cell presents antigen--it displays viral proteins and skin proteins on its surface. The T cells that can recognize the telltale proteins bind to them, thus getting the first signal. But along the way to the lymph node, prompted somehow by the evidence of cell death it has absorbed, the dendritic cell has produced another substance on its surface, one of a group of molecules whose identity is still being worked out. That substance gives T cells the second signal: it tells them to get going. Multiplying rapidly, they fan out in search of virus-infected cells to kill. Eventually they eliminate the virus. A few of them remain as memory cells to respond more quickly the next time it invades.
In this scenario, Matzinger realized, the immune response would inevitably cause collateral damage. Some of the T cells--the ones that happened to bind to skin proteins on the surface of the dendritic cell rather than to chicken pox proteins--don’t head for infected tissue; they simply head for skin. They even kill a few perfectly healthy skin cells-- but then they rest down. In this state they can’t kill unless restimulated with both the first and second signals. As long as the infection continues, they are likely to become reactivated. But once the infection is cleared, they rest down. When they bind in this condition to skin protein, they receive signal one without signal two. That kills them.
T cells also die in the self/nonself model when they get only signal one, and that’s how self-attacking ones are eliminated. The difference between that model and Matzinger’s lies in where and when the elimination takes place. In the self model the bulk of it happens in the thymus early in life. In the danger model it happens all the time. Whenever T cells bind to a protein in tissue but don’t get signal two--whenever there is no danger--they die. Because our tissues provide a large and continuous source of signal one, this dying off goes on continuously throughout life, as T cells circulate through the bloodstream and lymph nodes. Says Matzinger: We are constantly acquiring tolerance to our own proteins. The self, in other words, is constantly being defined anew-- which is another way of saying that it doesn’t really exist at all.
Saying that the immune system responds only to real danger provides a natural explanation for why it doesn’t attack fetuses, gut bacteria, or other useful things that it ought to consider nonself: those things aren’t dangerous, they don’t burst cells, and they don’t activate dendritic cells. But why does it sometimes attack things it ought to consider nondangerous? Why does autoimmune disease occur? Matzinger has an answer, albeit a speculative one. The conventional view is that an autoimmune disease arises from a mistake of the immune system--essentially from a poorly developed sense of self. But Matzinger thinks there are several ways in which patients without defects in their immune response may suffer autoimmune disease.
One possibility is that the real culprit is a hard-to-detect, slow-growing infection. Patients with rheumatoid arthritis and multiple sclerosis often suffer intermittent flare-ups; these may be the result of a viral infection in which T cells are activated against normal body proteins. When the infection is cleared, the autoimmune response dies down. Another possibility is that a stretch of a person’s MHC may bear an unfortunate resemblance to a viral protein. Thus, a bout with a virus could activate an ongoing T cell response to a person’s MHC. In either case, the immune system is responding to real danger signals. Yet another possibility is that the tissues themselves are provoking the ongoing immune response-- perhaps because of some genetic defect that causes abnormal cell stress or death. Matzinger notes that a defect in the gene that regulates normal cell death causes lupus in mice.
What about organ transplants? Why are they so often rejected? Some have objected that my model can’t work, says Matzinger, because a transplanted liver isn’t dangerous. But what happens when organs are transplanted? They get cut up and reattached. All that surgery causes injury. The dendritic cells in the donor organ start heading for the lymph nodes. There they bind to recipient T cells that have never seen the donor organ’s brand of MHC. The T cells get activated and start attacking the donor organ.
Indeed, Matzinger’s theory can explain experimental results in transplantation that have been puzzling immunologists for nearly 20 years. In the early 1970s, a couple of researchers had succeeded in grafting tissue from one mouse into an unrelated mouse without using immunosuppressive drugs. Why did it work? Before grafting, the donor tissue had been depleted of passenger leukocytes, the cells that some thought aroused the host immune response. Researchers now know that those cells were actually dendritic cells--which were first isolated in 1973 by Rockefeller University researchers Ralph Steinman and Zanvil Cohn.
Even more impressive is how Matzinger can reinterpret the pioneering experiments by Medawar. The reason Medawar’s bomber pilot and later his rabbits rejected their skin grafts, she explains, was not simply because the skin graft was foreign but because it contained activated dendritic cells. The skin is laced with them. If you take a piece of skin and put it in a culture dish, says Matzinger, just the bruising that is created by the incisions is enough to activate the local dendritic cells, and they will do what they do, which is to leave. In fact, that’s a nice way of purifying them, because they crawl out into your petri dish.
Conversely, says Matzinger, the reason Medawar’s newborn brown mice didn’t reject the white-mouse blood cells he injected into them was not that their sense of self was still malleable: it was that the injection didn’t give them a clear danger signal, because it didn’t contain enough activated dendritic cells to arouse a newborn mouse’s small supply of T cells. What it did contain was stem cells--the cells that give rise to all types of blood cells--which took up residence in the bone marrow of the brown mice, giving them a lifelong supply of white-mouse blood. The newborn brown mice became tolerant of that blood and of the MHC it displayed in the same way they became tolerant of their own MHC--any maturing T cells that could respond to it got signal one without signal two, and promptly died. By the time Medawar grafted white skin onto them--thus providing them with large amounts of signal two, in the form of activated dendritic cells-- there were no T cells around that could respond to it. So the mice tolerated the graft.
But that tolerance, Matzinger stresses, was not something they learned as newborns. Recently she, Fuchs, and their colleague John Ridge showed as much: they repeated Medawar’s experiment. But instead of injecting foreign blood cells into newborn mice, as Medawar had done, they injected activated foreign dendritic cells--far more than Medawar’s original injection had contained. If the self/nonself model were true, the change shouldn’t have made a difference; any foreign cells should have been tolerated, if administered early enough, when the mice were still developing their sense of self. But that’s not what happened. Matzinger’s mice rejected the dendritic cells. The experiment shows that the immune systems of the newborn mice are not inherently tolerant, she says. They can respond to danger.
Does it make any practical difference whether Polly Matzinger is right instead of Peter Medawar? It may. One area that could be affected is transplantation biology, the very field that motivated Medawar. Current transplantation procedures, derived from the self/nonself model, require surgeons to carefully match the donor’s MHC to the recipient’s and to suppress the immune response with powerful drugs. But if the activation of the dendritic cells can be controlled, it may one day be possible to transplant organs without as much fuss about who the donor is, and without subjecting patients to risky drugs.
One way might be to deplete the donor organ of dendritic cells before transplantation--although that would be difficult, because dendritic cells tend to be buried deep within the tissue. Another solution may be to create antibodies that can bind to dendritic cells and prevent them from activating T cells in the recipient. Some studies have already shown that such antibodies can block the immune response--and promote long-term tolerance--in animals receiving transplanted tissue.
The danger model may also have implications for cancer therapy. The immune system is notoriously blind to cancer cells, and the reason, as Matzinger sees it, is that tumor growth doesn’t injure tissue. But cancerous growth might be controlled by repeatedly injuring cancerous tissue, thus continually reawakening an immune response. In fact, researchers are currently developing a cancer vaccine that contains heat shock proteins--substances a cell produces whenever it is stressed. By injecting heat shock proteins into tumors, they hope to stimulate an ongoing immune response against the tumor.
Not surprisingly, though, considering that Matzinger is going up against a theory that has been entrenched for half a century, the reaction to her danger model has been mixed. Transplantation biologists and surgeons, she says, generally give it good notices. A strong supporter, for instance, is Kevin Lafferty, now director of the John Curtin Medical School of Research in Australia. Lafferty was one of the first to propose a two- signal model of the immune response, and he also did some of the graft experiments in mice that are so well explained by the danger theory.
One of the major problems of immunology has been to develop a theoretical explanation for self/nonself discrimination, Lafferty says. But maybe we’ve been going down the wrong path--if things don’t fit, it’s time to change the metaphor. I think we’re at that stage in immunology, and I think that’s where Polly’s been very valuable. It’s going to take a while, though. People aren’t going to swallow this too fast.
And indeed, a lot of immunologists aren’t swallowing it, at least not whole. Some agree that the immune system may be attuned to signals of danger but point out that Matzinger has yet to do the really hard work of pinning down what those signals are. To be sure, some molecules that turn on T cells are already known, but nobody knows what substances in damaged tissue activate the dendritic cell. Others see no reason the immune system can’t do both: why it can’t develop a basic sense of self early in life, through the careful winnowing of T cells in the thymus, and still respond to danger later on. And still others think that the whole argument is semantic--that Matzinger has just found a different way of phrasing what the immune system does, one that doesn’t change anything important. Among this last group is Rod Langman, the Salk Institute immunologist who got Matzinger into the field.
Matzinger finds the last group of critics the hardest to understand. At least nobody is throwing rotten tomatoes, she says. But the people who are saying, ‘Well, Polly isn’t saying anything new. What she’s done is a very nice synthesis of everything we knew into a model that works’--they haven’t really jumped off the cliff with me. They haven’t said, ‘What this says is that the self/nonself model is wrong.’