Rethinking the Natural Killer

NK cells have long been considered blunt instruments of the innate immune system. But a growing body of evidence suggests that some NK cells can specifically recognize—and later remember—novel antigens, behaving very much like members of the adaptive immune response.

By Andreas von Bubnoff

Every day, our body makes millions of T or B cells specific to unique molecular signatures, or antigens, most of which will never enter the body. The specificity of each of these cells is determined by a surface receptor—known as the T- or B-cell receptor—that is generated by a constrained but random rearrangement of genes. Though distinct molecules, the receptors borne by healthy B and T cells share the ability to detect antigens that are foreign to the body and so represent a potential threat of infection or disease. When it detects its antigen, the T or B cell becomes activated and proliferates, mounting an immune response against the bearer of that molecule before its progeny die away. A few of its descendents, however, become memory cells, which stay alive for a long time and are rapidly activated if the antigen turns up again.

This “antigen-specific memory” is the hallmark of what immunologists call the “adaptive” immune response. The innate immune system, by contrast, doesn’t have the ability to rearrange antigen receptor genes and therefore cannot respond to or remember specific antigens; its constituent cells—such as granulocytes, patrolling macrophages, and dendritic cells—are less specific in their recognition of quarry, detecting molecular patterns that are characteristic of certain types of pathogens (viruses, bacteria and so on), rather than specific pathogens.

Natural killer (NK) cells, which destroy stressed or infected cells, have long been thought to belong to this compartment of the immune system. A series of recent discoveries has, however, challenged this long-standing view of their role and capabilities.

Just like B and T cells, it seems, some NK cells can recognize—and later remember—antigens they have never seen before. This has led to a frantic search for some kind of NK receptor analogous to the receptors borne by T and B cells that can somehow change randomly to enable NK cells to specifically recognize random antigens. To further complicate the emerging picture, researchers now have evidence that certain kinds of NK cells can dampen the adaptive T and B cell immune response.

What the new findings could mean for vaccine development is far from clear. But some researchers daringly suggest that NK cells could represent a “third arm” of the adaptive immune system. “People get really excited, [because] this could be huge,” says Stephen Waggoner of Cincinnati Children’s Hospital.

Serendipitous beginnings

It all started with a study of mouse immune responses to haptens in the early 2000s in the lab of Ulrich von Andrian at Harvard Medical School. Haptens are chemicals that can elicit an immune response when they touch the skin. At the time, von Andrian, like everybody else, believed that only T and B cells could possibly mediate hapten-specific memory responses. And, as expected, when he and his colleagues painted the shaved backs of mice with a hapten named dinitrofluorobenzene and, a few weeks later, injected the same hapten into their bladders, the massive inflammatory response involved many T cells.

But as he was preparing to submit the paper, von Andrian asked his postdoc to do just one more experiment: a negative control to show that there is no memory response when B and T cells are absent. To their surprise, the researchers found that the memory response was just as large in mice that lacked those cells, suggesting that perhaps NK cells—one of the remaining immune cell types in these mice—might have been a major driver of the hapten-specific memory response. “So here, suddenly, there was a learned response in an animal that had no T and B cells that led to inflammation,” von Andrian says.

How could this be? At the time, no one had ever reported on NK cells the kind of variable receptors required to elicit an antigen-specific immune and memory response. So von Andrian’s first thought was that the result was a mistake—a mix-up of mice. But a quick check revealed that the mice used in the experiment really had no B and T cells, and when the researchers redid the experiment using similar mice from another source, they got the same results.

Still, von Andrian was skeptical: Perhaps, he thought, this hapten-specific NK memory response was specific to the bladder. After all, it was known that one way to treat bladder cancer is to induce an NK cell mediated immune response to cancer cells by injecting a vaccine. But when the researchers painted one ear of a sensitized B and T cell–free mouse with the hapten (instead of injecting the hapten into the bladder), they got the same result: Only the ear they painted with the hapten was swollen as a result of inflammation. What’s more, this only happened with the same hapten they had used before to paint the shaved back of the mouse a month earlier. “Now you have the hallmark features of adaptive immunity,” says von Andrian. “It’s a learned response, it’s remembered long-term, and it is antigen specific.”

Around that time, von Andrian heard from other researchers who had made similar observations but dismissed them. “I heard from [a] colleague who had a grad student who showed him these data way before we saw it, and he basically dismissed [it],” he says.

But von Andrian and his team kept digging. They soon found that NK cells were required and sufficient for the hapten-specific memory response, and that only NK cells from the liver of the mice, but not from the spleen, could remember the hapten when they transferred them to another mouse that had no NK cells and had never been exposed to the hapten. This suggests that most memory NK cells reside in the liver, von Andrian says.

To add a final layer of confidence in these results, von Andrian let his postdoc repeat the key experiments in a blinded fashion. “It always came out right,” he remembers. “So I was very confident at that time that what was observed was a real reproducible effect that was not tainted by observer bias.”

Still, he had sleepless nights because he felt that this was “just too heretical to be true.” And he wasn’t the only one with doubts: The manuscript spent eight months at one of the major scientific journals, went through three revisions, and an editor told him at one point that a reviewer had called the study “the cold fusion of immunology.” Around the same time, the third postdoc who had been working on the project quit. “[She] actually decided that [she] did not want to stay in science,” he says. “At the time, this project was so frustrating that no one in my lab would touch it with a ten-foot pole.”

That’s when von Andrian had enough. On a Friday night, he sent the study to Nature Immunology, which accepted it the following Monday without further review (von Andrian had informed them about the many previous revisions and changes to the manuscript; Nat. Immunol. 7, 507, 2006).

The paper made quite a splash. “That was a big discovery,” says Marcus Altfeld, an NK cell expert who still has a lab at the Ragon Institute in Boston but is currently moving to the Heinrich Pette Institute in Hamburg. Still, researchers were scratching their heads: How could NK cells respond to antigens they had never seen before without the kinds of variable receptors that T and B cells are known to have? The only types of NK cell receptors known at the time were hardwired (encoded in the germline, so that all NK cells had the same receptors), and they did not appear to undergo routine genetic rearrangement. For example, some NK cells in certain strains of mice that are naturally resistant to cytomegalovirus (CMV) infection carry a CMV-specific receptor called Ly49H on their surface, probably because CMV has been infecting mice for millions of years, which is long enough for NK cells to evolve the receptor.

But when von Andrian reported in 2010 that mice can also show NK memory responses to antigens of HIV (Nat. Immunol. 11, 1127, 2010), a hardwired receptor seemed like an unlikely explanation because HIV is a relatively new virus and doesn’t infect mice. “I think the data really challenged that idea of a germline-encoded receptor that has emerged through evolution,” Altfeld says. They suggest, he adds, that NK cells might indeed have a receptor that can randomly rearrange to bind novel antigens.

Meanwhile, several other groups have also detected NK cell-based hapten responses, von Andrian says, and Norman Letvin at Harvard Medical School and his colleagues reported that an NK cell-based memory response can protect mice from vaccinia virus infection (PLoS Pathog. 7, e1002141, 2011). Even NK cells that carry hardwired receptors like the CMV receptor Ly49H have been shown to form long-lived memory NK cells: When UCSF’s Lewis Lanier, who initially found the Ly49H receptor, transferred CMV-specific NK cells from one mouse to another mouse that lacked these CMV-specific NK cells, he and his colleagues found that the transferred cells rendered the mice protected from CMV infection, persisted for at least six months, and enabled the mice to respond to a secondary CMV challenge with a stronger immune response (Nature 457, 557, 2009).

Is NK cell memory hardwired?

While there’s strong evidence that the kind of antigen-specific NK memory von Andrian found  is real, nobody knows how it is formed. “Could it be,” von Andrian asks, “that there is perhaps a [genetic] locus in NK cells where some kind of gene becomes edited or rearranged, where you generate [receptors with] highly diverse, more or less randomly formed binding specificities, similar to the T- and B-cell receptor locus?”

Such a scenario isn’t completely improbable. After all, von Andrian argues, adaptive immune memory cells that undergo genetic rearrangements of receptor genes that differ from B- and T-cell receptors have been found in primitive vertebrates called agnathans (jawless fish), which include hagfish and sea lampreys. Their immune memory is formed by so-called “variable lymphocyte receptors.”

Another possibility is that NK cell memory could be encoded not by rearrangements of the genes, but by so-called “epigenetic changes,” such as chemical modifications of the genetic material itself to silence various genes in response to environmental stimuli.

von Andrian reasoned that if the memory can be transferred between different cells together with the genetic material, it’s likely based on genetic rearrangements, since any epigenetic changes would be erased after transferring genetic material. And indeed, he has done preliminary experiments in mice that suggest that the NK cell memory doesn’t get erased and has even started to identify the part of the genome that might carry the NK cell memory. “The most likely, simplest explanation,” von Andrian says, “is that there is actually a gene, a locus where there is some modification at the DNA level that’s heritable that determines NK cell specificity for a large diversity of different antigens.”

Despite these results, not everyone is convinced that an epigenetically encoded NK cell memory can be excluded. For one, DNA sequence itself has recently been found to encode instructions as to how to change DNA epigenetically, says Waggoner. In addition, Lanier says that when he and others compared the expression of all of the about 20,000 genes of NK cells before and after immunization of mice, they didn’t find any obvious candidate receptor genes that were expressed differently. That’s why Lanier is currently checking whether changed methylation patterns of genes could explain NK cell memory.

But, in general, the field seems to be slowly accepting that antigen-specific NK cell memory exists, and is eagerly waiting for the identification of the elusive NK cell receptors. “I think there is [now] much greater acceptance,” says von Andrian. “As of two months ago, for the first time after having worked on this for ten years, I was able to get NIH grant support.”

Altfeld agrees. “I think [most] people are now—even in the NK cell field—accepting it,” he says. “That was different two years ago. But the big question now is what kind of receptor is mediating that. The Holy Grail right now is trying to identify that receptor that mediates this antigen specificity.” 

If von Andrian is right, he might have identified a third arm of the adaptive immune system—which was actually the title of one of his rejected grants. There are fewer NK cells than B and T cells, he admits, but on a per-cell basis, NK cells seem more powerful, because fewer virus-specific NK cells than T or B cells need to be transferred to naive mice to render them protected from influenza infection.

While most of the evidence for NK cell memory comes from mice, there are already hints that it might also exist in nonhuman primates and perhaps even humans. Keith Reeves of Beth Israel Deaconess Medical Center, in collaboration with von Andrian and Dan Barouch of Beth Israel Deaconess Medical Center and the Ragon Institute, studied NK cells from rhesus macaques that had been vaccinated years ago, and found that liver and spleen NK cells from the animals could kill dendritic cells from the same animal in vitro, but only if these dendritic cells had been exposed to the same antigen that was in the vaccine the animal had seen five years ago. This shows that in rhesus macaques, the NK cell-based antigen-specific memory can last as long as five years.

One hint that even humans might have such responses is von Andrian’s observation of NK cell memory responses to viruses or haptens in humanized mice, which carry human NK cells.

But if the macaque experiments are any indication, studying human NK cell responses could be a challenge. That’s because the NK memory cells in rhesus macaques reside mostly in the spleen and liver, from where they probably migrate to the sites of infection. If this is also the case in humans, Altfeld says, it will be difficult to get samples of memory NK cells, unless researchers can identify an NK memory cell marker that can be used to identify such cells in blood, where they are much rarer. The ideal marker, of course, would be the elusive antigen-specific memory NK cell receptor itself, Altfeld says. “The critical step really is to identify that receptor that mediates antigen specificity,” he says. “Once that receptor is identified, then I think we can really characterize these NK cells in more detail.”

NK cells as wet blankets

If the recent findings on NK memory cells aren’t complicating things enough, researchers now also have evidence that NK cells can have a dampening effect on the B- and T-cell response.

While it’s been known for decades that NK cells can regulate adaptive immunity, the first evidence that this regulation is of vital importance for control of virus infection came from a 2011 study by Waggoner and his colleagues. Waggoner says that one motivation for the study was when he heard of a somewhat paradoxical finding by Mary Carrington and her colleagues a few years ago: An NK cell receptor that inhibits NK cell activity was associated with better control of HIV and hepatitis C infection, which would imply a stronger immune response.

So how could lower NK cell activity be associated with better immune responses? Most people, Waggoner says, tried to reconcile the result with their assumption that NK cells primarily kill virus-infected cells, and thought that perhaps only NK cells with a functioning inhibitory receptor (or “brakes”) get the license to kill: “They are only putting gas in the cars that have a brake pedal,” he says.

But, Waggoner wondered, what if NK cells dampen the adaptive T-cell immune response? This would suggest a different explanation of how an NK receptor that inhibits NK cell activity could lead to better control of HIV and hepatitis C infection: A better adaptive immune response. To see if this could be the case, he infected mice with a strain of lymphocytic choriomeningitis virus (LCMV) called clone 13, since clone 13-infected cells are less likely to be killed by NK cells. That’s important because he didn’t want to have any NK cell effects on the adaptive immune response obscured by the fact that NK cells kill infected cells.

It was known at the time that mice can not only clear low doses of LCMV, but also survive high doses, because the virus overwhelms the immune system so much that it stops fighting the virus, which then simply persists because the animal’s immune response is weakened. In addition to low and high doses, Waggoner and colleagues also tried a medium dose in their study. They found that the medium dose could kill the mice because there was enough virus to keep the immune system fighting, but not enough to overwhelm it sufficiently to give up the fight. The continuing immune response eventually killed the mice (Nature 481, 394, 2011).

But when the researchers depleted the NK cells in the mice, they saw the opposite. The mice survived the medium dose, but died at the high dose. Waggoner’s explanation was that at both doses, NK cells kill activated CD4+ T cells, dampening the adaptive immune response, and that removing NK cells strengthens the CD4+ and CD8+ T-cell response. At the medium dose, this helps the mice clear the virus. At the high dose, though, the stronger CD4+ response so boosts the response that it continues fighting the virus, which eventually kills the mice.

The NK cell-mediated killing of CD4+ T cells also has a dampening effect on the antibody response, Waggoner and colleagues found. Without NK cells, mice that are acutely infected with LCMV not only have stronger CD4+ and CD8+ memory T-cell responses, but also develop LCMV-specific neutralizing antibodies sooner.

Not much is known about similar effects of NK cells in humans, but researchers have recently reported that in HIV elite controllers, there is an inverse relationship between the strength of NK cell responses and virus-specific CD8+ T-cell responses, Waggoner says (AIDS 26, 1869, 2012). To find out more, Altfeld is checking the effects of NK cell activity on adaptive immune responses in people acutely infected with HIV, and is studying whether depleting NK cells in humanized mice can improve the adaptive immune response to vaccination.

Implications for vaccine design

Waggoner’s results would suggest that a vaccine that can inhibit NK cells that kill activated CD4+ T cells should do a better job at inducing broadly neutralizing antibodies. On the other hand, vaccines that can induce memory NK cells that can kill infected cells could also be quite advantageous, because such NK cells seem to respond more quickly than T cells when it comes to migrating to the places where they are needed, such as mucosal sites, says Altfeld. “NK cells have this very quick response time before they kill, much faster than T cells,” he says. What’s more, von Andrian says, NK cells can’t be infected by HIV, whereas vaccines that induce T cells might make HIV infection worse by creating additional HIV target cells.

Still, the opposing effects NK cells can have on the immune response—improving the immune response by killing infected cells, but dampening the adaptive immune response by killing activated CD4+ T cells—represent quite a challenge for vaccine developers. For example, a vaccine or drug that inhibits NK cells that kill activated CD4+ T cells would have to do so without also inhibiting NK cells that can kill infected cells. Alternatively, a vaccine that induces NK cells that kill infected cells would have to do so without also inducing NK cells that inhibit adaptive immune responses. Selectively modulating just one of these opposing NK cell functions will require that researchers completely understand both, Waggoner says. Importantly, von Andrian notes, these opposing effects of NK cells on the immune response could very well reflect the effects of different types of NK cells.

While it’s unclear how to develop a vaccine that induces effective NK cell responses, researchers are starting to study the effects of existing vaccines on NK cells. Altfeld and colleagues reported a few years ago that flu vaccination results in changes in the types of NK cells that persist for at least 150 days after vaccination. Altfeld and Reeves plan to study NK cell responses in volunteers of vaccine trials, and Lanier is part of the human immunology project consortium where researchers use microarray or similar analyses to check responses to vaccines such as yellow fever, flu, and, in Lanier’s lab, to varicella zoster (chickenpox) vaccine in a more comprehensive way than ever before.

As is often the case in scientific discovery, the recent findings have raised more questions than they have answered. But it is equally clear that scientists, like all people, are often reluctant to shed their prejudices; were that not the case, the NK cell’s versatility might have been discovered earlier. “Challenging [old] ideas is just tough,” Altfeld says. “We are probably more likely to discard [a] result rather than thinking that it might be an initial indication that there is a completely different kind of arm of the immune system.”