Mystery Rays from Outer Space

TRegs in normal skin

Tumors are supposed to be destroyed by our immune system. So how come we still see tumors?

A big part of the answer is probably that our immune system is very good at destroying proto-tumors, but is not so good at handling those that manage to sneak through and grow to the point of detectability. That splits the first question into two questions: Why do some proto-tumors manage to sneak through, not being eliminated by the immune system? And why is it that detectable tumors are not effectively handled?

The first part, I think, may often be related to cell-intrinsic immune escape mutations. That is, pre-cancerous cells are constantly being attacked by the immune system; in turn (if they survive long enough) they constantly mutate, doing things like damaging the antigen-presentation pathway that makes them recognizable by the immune system. Eventually, they find some configuration that reduces the rate at which they’re killed. Once cancer cell replication is even fractionally greater than destruction,¹ a tumor can begin to grow.

So that’s probably the earliest stage of tumor growth. But once tumors reach a certain size, a second factor kicks in. Chronic immune responses are dangerous; after all, the whole point of the immune system is to kill things. The chronic immune response against the growing tumor is now shut down. This has been understood for quite a while — the immune system often becomes “tolerant” of a tumor. More recently, it’s become clear that it’s not merely “tolerance” (which implies that the immune system is simply benignly ignoring the tumor); the presence of a tumor actively forces the immune system to shut itself down, slamming on the brakes rather than just peacefully coasting by.

Brakes are a fundamental part of an active immune response. If you look at diagrams of normal immune responses, they show inverted “U” shaped curves (in here and here, for example), where the response is triggered, rapidly ramps up, hopefully does its thing, and then just as rapidly shuts down to near-background levels once again. There used to be a sort of general feeling that this was a rather passive thing — pathogen stimulates response, response destroys pathogen, no more stimulus, response goes away — but now we understand that the shut-down phase is just as active and dynamic as the upward curve. Just as with the upward phase, there are all kinds of different mechanisms to control the response; one of the most important is the “Regulatory T cell” (TReg).  And it’s pretty clear that TRegs are involved in controlling the immune response to tumors (I talked about that here, and links therein).

TRegs have been known for a while (I gave a brief history, including the I-J fiasco, here). The usual description of a TReg includes a number of markers;² one of the most basic is CD4. CD4 T cells used to be lumped together as “T Helper” cells, but now we have multiple sub-specialties in the CD4 category, and TRegs are one of those specialities.

More recently, TRegs — or at least cells that function the same way as TRegs — have been described in the CD8 population of T cells.³ CD8 T cells are traditionally called “Cytotoxic T lymphocytes” (CTL) (although it’s been increasingly clear that cytotoxicity is just one of many functions a CD8 T cell can offer), but it seems that these variants of CD8s can actively shut down an ongoing immune response, in a specific and targeted way. There seems to be a trend to calling these cells “suppressor cells” rather than “TRegs”. “Suppressor T cells” is an older term that was out of favor for a while, but it’s probably useful to bring it back and distinguish between the natural TRegs and some of the other cells that can do something similar but that have different sources and origins.

At least some of the CD8 suppressor T cells can arise from apparently-conventional CD8 T cells. That is, you can pull CD8 T cells out of a normal mouse’s spleen, and depending on what those cells see and are exposed to, they could progress to being conventional CTL — killing tumor cells, producing interferon and other cytokines, generally being a destructive force — or they could become suppressor CD8 T cells, and actively prevent that destruction from happening.

It turns out that one of the forces that can drive a CD8 T cell into being a suppressor T cell is a tumor. A recent paper from Arthur Hurwitz’s lab⁴ shows this quite clearly. They had shown previously that transferring specific CD8 T cells into a tumor-bearing mouse resulted in what they called “tolerance”.⁵ But now they demonstrate that it’s more than that; the transferred CD8s are converted into suppressor T cells that actively shut down immune responses.

Tumor-infiltrating TcR-I cells suppressed the in vitro proliferation of both melanoma Ag-specific CD8+ (37B7) T cells and OVA-specific CD4+ (OT-II) T cells. … Even at a ratio of one TcR-I cell to four responder T cells, we observed 30% suppression of proliferation. ⁴

This isn’t the only way that tumors escape immune recognition, but (at least for some tumors) it may be an important one. It’s clearly an important consideration for things like tumor vaccines and immune therapy, because it suggests that immunizing with tumor antigens (and thereby generating lots of tumor-specific CD8 T cells) may actually increase the suppressive effect of the tumor.

The conversion of CD8+ effector T cells into suppressor cells may be one mechanism by which tumors restrict the immune response from effectively controlling tumor growth. As subsequent effectors infiltrate the tumor, either following peripheral sensitization or as a result of adoptive transfer therapy, the induced regulatory cells may suppress these new effectors and reduce their ability to confer tumor immunity. This cyclic suppressive process may contribute to the profound loss of antitumor responses following adoptive immunotherapy. ⁴

(My emphasis.) On the other hand, if this is a common mechanism, then overriding it — which should be possible, using cytokines, specific T cell subsets, and/or targeted receptor ligands — may switch the suppressive population abruptly back into an effector group, turning the brainwashed traitors into resistance fighters.

1. Destruction would include far more than immune destruction, of course — it would include cells that become differentiated and no longer replicated, cells that outgrow their oxygen supply, cells that undergo apoptosis, and so on[↩]
2. FoxP3, CD25, and so on[↩]
3. I’m not sure who made the first identification; this looks as if it’s one of those fields where there were incremental advances, hinting more and more strongly at the presence of these cells, but with no single clearcut starting point. Papers in the early 2000s start to point at regulatory CD8s, and by 2004 a handful of relatively high-profile papers fairly solidly identified them. A 2004 review paper is
   Zimring, J., & Kapp, J. (2004). Identification and Characterization of CD8+ Suppressor T Cells Immunologic Research, 29 (1-3), 303-312 DOI: 10.1385/IR:29:1-3:303[↩]
4. Shafer-Weaver, K., Anderson, M., Stagliano, K., Malyguine, A., Greenberg, N., & Hurwitz, A. (2009). Cutting Edge: Tumor-Specific CD8+ T Cells Infiltrating Prostatic Tumors Are Induced to Become Suppressor Cells The Journal of Immunology, 183 (8), 4848-4852 DOI: 10.4049/jimmunol.0900848[↩][↩][↩]
5. Anderson MJ, Shafer-Weaver K, Greenberg NM, & Hurwitz AA (2007). Tolerization of tumor-specific T cells despite efficient initial priming in a primary murine model of prostate cancer. Journal of immunology (Baltimore, Md. : 1950), 178 (3), 1268-76 PMID: 17237372[↩]

The Nervous System (Fritz Kahn (1888-1968))

Is “immunodominance” just what you get when you measure the wrong place?

Usually, when you look at T cell immune responses to a virus, they’re pretty strongly biased. That is, although the T cells are theoretically, and often observably, able to recognize a wide range of target peptides, the immune response is strongly focused on just a handful of these peptides, while the remaining pool of potential targets is either ignored altogether or given a cursory glance by a handful of T cells. This phenomenon is known as “immunodominance“, and it’s seen with  immune responses to all sorts of pathogens. In some cases — such as for HIV — it’s likely that a strongly immunodominant response is harmful, because it makes it easier for the infecting virus to mutate away from immune control. But in the vast majority of cases the immune response, be it never so immunodominant, does a perfectly good job of controlling the virus; which is why we’re able to easily control most of the viruses that we’re exposed to.

Usually when you measure an immunodominant response, you’ll take lymphocytes from the most abundant, easily-accessed place you can find. That would be blood, in humans; in mice you’d probably take a spleen or lymph nodes.  Some viruses like to hang out in these places, and these include some of the more popular research viruses.

But most of the viruses we’re exposed to don’t infect blood or secondary lymphoid organs; they infect the lungs, or the skin, or neurons, or some other tissue. When we measure the blood response, we believe we’re measuring a good approximation of the real response ongoing in the infected tissue, but that’s mostly been an assumption, not a demonstrated fact.

Recently there’s been some work starting to feel out how similar the tissue response is to the blood/lymphoid organ response. For example, I talked here about work establishing the timing of immune responses in the lungs, vs. the blood. In this case, the overall patterns were similar, though the details were somewhat different.

But that was only really looking at a fairly big picture — overall patterns. What about specifics of target recognition? In particular, is the immunodominance we measure in the blood what actually happens on the battlefield?

I’m only aware of a couple of studies that look at this at all, and those were mainly as asides, noticed in passing. Yewdell’s group has shown in a couple of paper that  infecting mice with poxviruses by different routes leads to differences in immunodominance:¹

The latter point is underscored by our observation that the ID hierarchy varies with the route of infection, the first observation of its kind to our knowledge. It will be of great interest to determine the underlying mechanism. ²

I’ve been told of unpublished data that show different immunodominant responses between lung and spleen, as well; also with a poxvirus.

But in those few examples, the epitopes were all known ones.  Known epitopes moved up or down a notch or two in the immunodominance hierarchy. A recent paper from Bob Hendricks’ group³ shows that T cells in the tissues can recognize things that are apparently not seen at all in the blood or spleen.

Electron tomogram of HSV4

Here they used herpes simplex virus (HSV) in C57BL/6 mice, which have long been believed to almost entirely focus their CD8 T cell response on a single peptide. Hendricks’ group has been looking at the immune response to HSV in the brain, where the virus sets up a latent infection  (I’ve talked about some of his findings here and here).  Contrary to more traditional concepts, it’s now becoming clear (from Hendricks’ work, and that of others) that T cells in the brain are important in controlling latent HSV infection.

In this paper, he found that the immune response in the brain is much more diverse, fairly strongly recognizing at least one  peptide other than the known dominant job.  Because the “normal” (that is, non-neuronal) immune response is so focused, this almost certainly means that the active immune response, down at the pointy end where the T cells are actually working, are responding to altogether different peptides.

It’s generally been assumed, as I say, that the easily-accessed blood or secondary lymphoid tissue is a reasonable approximation of what’s going on in the actual sites of action, in the peripheral tissues — in other words, the idea has been that there’s more or less equal flow of cells between the tissues and the blood and lymph. The recent work on timing and kinetics that I mentioned here sort of supported that assumption, but now we have to wonder whether in fact there’s some kind of filter that keeps some sets of T cells from entering, or staying in, the blood.

We also have to wonder if the whole “immunodominance” paradigm is what we think it is. Could immunodominance represent the filter between blood and tissues, rather than the actual formation of responses? I actually don’t think that would explain immunodominance in general (for one thing, we see strong immunodominance for viruses of lymphocytes, where the blood is the site of infection, so there shouldn’t be a filter) but it’s something to factor in.

1. D. C. Tscharke (2006). Poxvirus CD8+ T-Cell Determinants and Cross-Reactivity in BALB/c Mice Journal of Virology, 80 (13), 6318-6323 DOI: 10.1128/JVI.00427-06
   D. C. Tscharke (2005). Identification of poxvirus CD8+ T cell determinants to enable rational design and characterization of smallpox vaccines Journal of Experimental Medicine, 201 (1), 95-104 DOI: 10.1084/jem.20041912[↩]
2. D. C. Tscharke (2005). Identification of poxvirus CD8+ T cell determinants to enable rational design and characterization of smallpox vaccines Journal of Experimental Medicine, 201 (1), 95-104 DOI: 10.1084/jem.20041912[↩]
3. B. S. Sheridan, T. L. Cherpes, J. Urban, P. Kalinski, R. L. Hendricks (2008). Reevaluating the CD8 T cell response to HSV-1: Involvement of CD8 T cells reactive to subdominant epitopes Journal of Virology DOI: 10.1128/JVI.01699-08[↩]
4. Electron tomogram of a HSV nucelocapsid completing envelopment , from Baines, J. D., C. E. Hsieh, E. Wills, C. Mannella, and M. Marko. 2007. Electron tomography of nascent herpes simplex virus virions. J Virol 81: 2726-2735.[↩]

Polarized CD8 T cell responding to a HSV-infected neuron

Just because something is called a “cytotoxic T lymphocyte” doesn’t mean it’s actually, you know, cytotoxic. And just because something is called a “lytic granule” doesn’t mean it’s actually lytic.

I’ve posted earlier on the range of functions that CD8 T cells — the so-called “cytotoxic T lymphocytes”, or CTL — actually have. CD8 T cells can certainly deliver cytotoxic signals to their target cells; but it’s become increasingly obvious that this isn’t all they can do. For example, it’s recently been shown that in HIV infections, CD8 T cells that show more than one function (“multifunctional” or “polyfunctional” T cells) are correlated with better control of the virus, whereas T cells that only show one or two functions don’t seem to control HIV. (The diagram to the right is from a seminar by Mario Roederer, and it shows the average functions — for example, the ability to secrete cytokines such as interferon, IL-2, MIP1b, and TNF — in long-term non-progressors, who control HIV relatively well, vs. those who progress and don’t control the virus.)

In fact, Robert Hendricks’ group just showed that T cell functions are even more complex than that, and they did it in the context of a fascinating problem — control of herpes simplex virus latency and reactivation.¹

The movie below shows sort of the traditional view of CD8 cell functions. ² Here we see a CD8 T cell 9in blue) and a target cell (filled with a green dye). At about 10 minutes (the film is speeded up, don’t worry!) the T cell makes tight contact with the target. Within five minutes, the target loses its dye; this is probably because the T cell is punching holes in the target cell’s membrane, so that internal contents can leak out. This is the caused by the T cell protein “perforin”. But T cells are capable of killing their targets in several ways, and we see a second mode of killing kicking in over the next 45 minutes or so. The target cell starts to bubble up, showing dense internal structures; this is probably the target entering a programmed cell death (apoptosis) pathway, and this is caused by a number of T cell proteins, especially “granzyme B”.

So, perforin and granzyme B are both killer proteins. They’re part of the “lytic granules” found in activated CD8 T cells. What Hendricks’ group has found is that perforin and granzyme B can also protect against herpes simplex infections without actually killing the target cell.

Herpes simplex virus first infects the skin or mucous membranes, then rapidly jumps into the neurons that innervate that patch of tissue and track up the axons to the ganglion. For the familiar and ubiquitous herpes simplex type 1, this is usually the trigeminal ganglion. In the neuron bodies of the ganglion, the virus (supposedly! — this is the traditional view) shuts itself down, reducing its protein levels to an undetectable level, and enters a latent state, where the viral genome hangs out but it’s otherwise pretty much a passive blob. Intermittently, the virus can reactivate from latency (especially after local immunity is reduced, for example due to “stress”), and then it tracks back down the axon to the original site of infection, and sheds into the environment once again. For HSV-1 the reactivations are usually seen the common and fairly harmless cold sores.

That traditional view has been changing. For example, we now know that the virus reactivates far more often than was believed;³ the reactivations come in such short bursts that unless you monitor very closely (swabbing 4 times a day, in the study in question) you’ll miss most of them. And they’re not associated with any lesions, usually.

Another change in the traditional view is that, at least for some, and perhaps most or even all of the infected neurons, the virus doesn’t really shut down protein expression to zero. Levels are drastically reduced, but T cells are incredibly sensitive, and it is now clear that T cells do in fact detect HSV-infected neurons in the ganglia. I posted about this research HERE, noting the evidence that infected neurons are often surrounded by HSV-specific T cells.

So if cytotoxic T lymphocytes are constantly detecting target infected neurons, they should be killing the neurons, right? That’s what “cytotoxic” implies. But clearly that’s not the case. Most of you, sitting reading this now (if anyone has in fact made it this far) have lots of “latent” herpes simplex in your trigeminal ganglia — the vast majority of people are infected. And yet your trigeminal ganglia are not slowly disintegrating under the assault of lytic T cells. The virus will still be there when you’re 70 years old, and your ganglia will still be intact (at least, as far as this is concerned; I make no further promises or guarantees).

So what Hendricks’ group has shown is that, yes, CD8 T cells do recognize HSV-infected neurons (this was already known). And the T cells suppress the virus, preventing it from reactivating; this was already known too. What’s new is that they showed that the T cells need perforin and granzyme B to prevent the reactivation, even though the neurons are not killed! They went on to show that granzyme B (which is a protease, that’s how it causes apoptosis) chops up a critical viral protein, blocking the virus from further protein production. So lytic substances can protect without actually lysing — a new function for CD8 T cells.

It’s not clear to me exactly how this works. For one thing, they also showed that a little later in infection there are other factors that suppress the virus (multifunctional!), and suggested that interferon might play this role. Still, it’s a very cool finding, and reminds us that viruses and immunity are both more complex than we know, and put together are complex cubed.

Note that this does not mean that perforin and granzyme B are not cytotoxic proteins! That’s very clearly their major function.⁴ What Hendricks’ work does show is that cytotoxicity is not their ONLY function.

1. J. E. Knickelbein, K. M. Khanna, M. B. Yee, C. J. Baty, P. R. Kinchington, R. L. Hendricks (2008). Noncytotoxic Lytic Granule-Mediated CD8+ T Cell Inhibition of HSV-1 Reactivation from Neuronal Latency Science, 322 (5899), 268-271 DOI: 10.1126/science.1164164[↩]
2. And I don’t remember where I got this from. Can anyone remind me of the author?[↩]
3. Karen E. Mark, Anna Wald, Amalia S. Magaret, Stacy Selke, Laura Olin, Meei?Li Huang, Lawrence Corey (2008). Rapidly Cleared Episodes of Herpes Simplex Virus Reactivation in Immunocompetent Adults The Journal of Infectious Diseases, 198 (8), 1141-1149 DOI: 10.1086/591913[↩]
4. for example, see S MIGUELES, C OSBORNE, C ROYCE, A COMPTON, R JOSHI, K WEEKS, J ROOD, A BERKLEY, J SACHA, N COGLIANOSHUTTA (2008). Lytic Granule Loading of CD8+ T Cells Is Required for HIV-Infected Cell Elimination Associated with Immune Control Immunity DOI: 10.1016/j.immuni.2008.10.010 [↩]

“Death and the Doctor” Published by William Humphrey, 1777

Last April I commented on a series of experiments  that used intravital microscopy to visualize cytotoxic T lymphocytes (CTL) attacking a tumor. ¹ Immensely cool though the movie is, I noted that I was surprised by their estimate of the rate of cell killing:

Another surprising finding — which is so different from previous work in different systems that I’m hesitant to believe it — is the timing of cell killing. Previous studies (such as the von Andrian paper² that produced this video) have suggested that CTL kill their targets in something under an hour; maybe 30 minutes or even less. Here. Bousso’s group find that the tumor cells take something like 6 hours to be killed. That’s such a large difference — and has such important implications for effectiveness of CTL killing — that, as I say, I’d like to see it confirmed before I take it to the bank.³

A new paper⁴ has run another estimate of the time it takes for a CTL to kill its target, and like most of the previous work, they conclude that it takes about a half-hour, give or take, to kill a target. They do come up with a fairly wide range of killing times, that depend on the target and the timing of the immune response — at the peak of the immune response when there are many cells the targets are killed faster (between 2 and 14 minutes), while at later stages, when there aren’t so many CTL, targets have half-lives of 48 min and 2.8 hr.

CTL killing a target cell (From a video by von Andrian)

This is not quite looking at the same thing as the video showed, though. In this paper, they were looking at the bulk effects, and that’s what almost all the previous studies have also looked at. The video was looking at a one-on-one interaction. What if targets are killed faster when several CTL gang up on them? Here, having different numbers of CTL caused the half-life of the targets to increase between about 10 and 20-fold. But this is probably simply because, with fewer CTL present, it took longer for them to find the target: Once a CTL found the target, the rate of killing was if anything faster than effectors at killing (“we find that LCMV-specific memory CD8 T cells kill more target cells per day than effectors”). ⁵

This is actually a disagreement with a previous paper ⁶ that also looked at killing rates, and offered evidence that different types of CTL can have different killing rates:

We reanalyse data previously used to estimate killing rates of CTL specific for two epitopes of lymphocytic choriomeningitis virus (LCMV) in mice and show that, contrary to previous estimates the “killing rate” of effector CTL is approximately twice that of memory CTL. ⁶

However, whichever of those studies is correct , both suggest that different types of CTL can have different killing efficiencies. This goes back to a point I’ve made several times, as have others (see e.g. Michael Palm’s TAG post here and references therein, including the comments by me and by Otto Yang) — CTL aren’t a uniform batch, and different kinds of CTL may have different types as well as rates of activities.

Returning to the intravital microscopy killing rate of 6 hours:⁷ I wonder if that reflects the nature of the CTL there, perhaps influenced by the tumor environment. Tumors are notoriously resistant to killing (probably because those tumors that are not resistant to killing were, um, killed, before they ever become clinically detectable) and it seems quite likely that an immunosuppressive tumor environment may change CTL types, or activities. I wonder if that would offer some way of intervention. Speeding up the rate of CTL killing from 6 hours to 30 minutes seems like it would be a huge influence of clearance of tumors. On the other hand, of course, it may be that the targets themselves are much more resistant to killing (again because tumor cells have been through selection to be resistant to the immune system) and cranking up CTL won’t make much difference.

1. Breart, B., Lemaître, F., Celli, S., Bousso, P. (2008). Two-photon imaging of intratumoral CD8+ T cell cytotoxic activity during adoptive T cell therapy in mice. Journal of Clinical Investigation, 118(4), 1390-1397. DOI: 10.1172/JCI34388 [↩]
2. Mempel, T. R., Pittet, M. J., Khazaie, K., Weninger, W., Weissleder, R., von Boehmer, H., and von Andrian, U. H. (2006). Regulatory T cells reversibly suppress cytotoxic T cell function independent of effector differentiation. Immunity 25, 129-141.[↩]
3. From this post[↩]
4. V. V. Ganusov, R. J. De Boer (2008). Estimating In Vivo Death Rates of Targets due to CD8 T-Cell-Mediated Killing Journal of Virology, 82 (23), 11749-11757 DOI: 10.1128/JVI.01128-08[↩]
5. There are also other videos of one-to-one killing, at least in vitro, that are more consistent with the 30-minute ballpark; see the image to the right for one example.[↩]
6. Yates A, Graw F, Barber DL, Ahmed R, Regoes RR, et al. (2007) Revisiting Estimates of CTL Killing Rates In Vivo. PLoS ONE 2(12): e1301. doi:10.1371/journal.pone.0001301[↩][↩]
7. Which I have become more relaxed about since my earlier skeptical comment[↩]

The 1918 influenza pandemic that killed between 20 million and 100 million people world-wide was unusual in a lot of ways. One of the most extraordinary things about it was not just the high mortality rate, but the mortality pattern. Normally influenza kills the very old and the very young; but the 1918 flu killed young adults as well — people in the prime of life, who normally are highly resistant to death from influenza. The famous graph below¹ (click for a larger version) shows this; the dashed line is mortality vs. age for “normal” influenza outbreaks (the “U-shaped curve”), and the solid line shows mortality vs. age for the 1918 flu (a “W-shaped curve”).

There are two ways to look at the W-shaped curve. You can either ask, Why did the young adults die? Or you can ask, Why did the older adults NOT die? In other words, was there something that protected the older adults that the younger ones didn’t have?

One explanation (and I can’t find the original paper to give credit) is that somewhere around 1870-ish (that is, 45 or 50 years before 1918) some other influenza strain infected the population, and gave a little bit of cross-protection against the 1918 strain. People alive in 1870 were exposed to this (hypothetical) strain of virus, developed immunity, and fifty years later were protected against the 1918 strain.² Of course, supporting this idea is the recent paper³ that received a fair bit of attention, showing that survivors of the 1918 pandemic still have specific immune responses to that virus, 90 years later — so certainly immunity could last fifty years.

One of the problems with influenza vaccination, of course, is that the virus changes. New strains arise and mutate, and the vaccines have to match the major circulating virus pretty well. Antibodies against one major strain of influenza don’t do a good job against a shifted strain. We see this is the paper I just mentioned — hardly anyone younger than 90 responded to the 1918 flu, even though they had been exposed to influenza viruses (of different types) every year for up to 80 years.

It would be nice if there was a vaccine that gave resistance to many strains of influenza, and we didn’t have to develop new vaccines ad hoc each year, based on imperfect guesses as to which viruses will circulate six months in the future. What’s more, we don’t really have good vaccine strains for novel influenza strains — like, say, whichever strain of avian influenza eventually succeeds in making the jump to humans.

Our present vaccines against flu are designed to raise protective antibodies. Antibodies in general target the outside of the viruses, which are intrinsically variable; hence the need to customize the vaccine strain with the circulating strain. Cytotoxic T lymphocytes, on the other hand, at least have the potential to target internal components of the viruses, which may have different constraints and therefore not be able to change as much. Could these CTL be more cross-reactive than antibodies?

It turns out that, yes, CTL are quite cross-reactive. A paper from Tao Dong’s and Sarah Rowland-Jones’ lab⁴ looked at anti-influenza T cells from normal volunteers (who had, of course, been exposed to the usual influenza A strains that sweep around the world each year).

Memory CD4+ and CD8+ T cells isolated from the majority of participants exhibited human influenza-specific responses and showed cross-recognition of at least one H5N1 internal protein.

The bad news is that these cross-reactive T cells are not very effective even against conventional influenza strains. Most of us are walking around with cross-reactive T cells, yet we still need antibody (such as from the vaccine) to actually be resistant to infection with one of the cross-reactive viruses. The good news is that the T cells aren’t entirely useless. It’s been shown several times that they do offer some protection, and in fact may be one reason that healthy adults are relatively resistant to influenza (the U-shaped curve above) — the very young would not have been previously exposed so wouldn’t have any cross-reactive T cells, while the very old often have defects in T cell responses. As Peter Doherty notes in his accompanying commentary:

The argument is that if seasonal influenza infection does promote cross-reactive T cell responses, then why do so many people get sick every one or two years? The counter-argument is, of course, that the majority of individuals may be protected from more serious disease by their T cell response.⁵

If the T cells that are normally present are partially protective, could they be cranked up to offer more protection? For example, what if vaccines (as well as triggering antibody responses) also tried to boost these cross-reactive memory responses? Potentially, this could lead to people who have long-term resistance to many influenza viruses, avoiding the need for annual re-vaccination, as well as having immunity to new influenza viruses that aren’t otherwise covered by the vaccinations; such as avian influenza.

Most influenza vaccines used today are killed (inactivated) vaccines, which are not very good at inducing CTL responses. However, cold adapted live attenuated influenza vaccines (LAIVs) are also used in some cases, and these should act as boosters for CTL:

… it would be worth evaluating the extent of cross-protection against H5N1 potentially conferred by currently available seasonal human LAIVs. Memory T cell populations boosted by these vaccines may in theory cross-react and provide partial protection against H5N1 by targeting highly conserved internal virus proteins. … The aim of such T cell-based approaches would be to provide broader partial protection against overwhelming infection and help lower morbidity and mortality rather than to provide complete protection against establishment of infection. This would be a highly relevant and perhaps more realistic public health goal in a pandemic situation.⁴

1. taken from Taubenberger JK, Morens DM. 1918 influenza: the mother of all pandemics. Emerg Infect Dis. 2006 Jan. http://www.cdc.gov/ncidod/EID/vol12no01/05-0979.htm[↩]
2. There were influenza pandemics in 1830-33; 1889-1890; and 1900; as well as epidemics in some other years. That was long before influenza virus was identified, and we have no idea what those strains were.[↩]
3. Yu X, Tsibane T, McGraw PA, House FS, Keefer CJ, Hicar MD, Tumpey TM, Pappas C, Perrone LA, Martinez O et al. (2008) Neutralizing antibodies derived from the B cells of 1918 influenza pandemic survivors. Nature[↩]
4. Laurel Yong-Hwa Lee, Do Lien Anh Ha, Cameron Simmons, Menno D. de Jong, Nguyen Van Vinh Chau, Reto Schumacher, Yan Chun Peng, Andrew J. McMichael, Jeremy J. Farrar, Geoffrey L. Smith, Alain R.M. Townsend, Brigitte A. Askonas, Sarah Rowland-Jones, Tao Dong (2008). Memory T cells established by seasonal human influenza A infection cross-react with avian influenza A (H5N1) in healthy individuals Journal of Clinical Investigation DOI: 10.1172/JCI32460[↩][↩]
5. Doherty PC, Kelso A (2008) Toward a broadly protective influenza vaccine. J Clin Invest doi:10.1172/JCI37232. [↩]

It’s become pretty clear that one way HIV persists in spite of an active, powerful immune response is to mutate its immune targets: “immune escape”.  HIV isn’t the only virus that does this, and we can learn from the others.

Cytotoxic T lymphocytes (CTL) identify virus-infected cells by recognizing short peptides, usually about 9 amino acids long. Each CTL is fairly specific; it only recognizes a single sequence. That means that if even one of the nine amino acids in its target mutates, the CTL is blind to the virus. If the CTL response to a virus is limited to a single peptide (which is, to a first approximation, the most common situation) then this single mutation will allow the virus to escape from the immune system, at least until new CTL arise targeting some different peptide.

The downside of this, from the virus’s viewpoint, is that it doesn’t really “want” to mutate itself. There are a limited number of sequences which allow the virus to replicate and spread efficiently, and to the extent that the mutations drive the virus away from these fairly optimal sequences, the virus is less “fit”. That means that the immune response can actually limit the virus’s replication even after (in fact, because) the virus has mutated away from CTL recognition.

What are the conditions under which this sort of continual partial control and constant escape can take place? One fairly obvious point is that the virus must not be eliminated by the immune system. If the first (or even tenth) CTL response to arise gets rid of all the virus from the host, then there’s no more viral immune evasion. I pointed out a parallel instance of this earlier, where poliovirus undergoes mutation in a person, but only when the person has a deficient immune system and can’t completely clear the virus. The virus needs to undergo constant replication over a long period, which makes this less of an issue for things like herpesviruses — they’re long-standing infections, but typically establish a latent infection rather than replicating throughout the infection. It probably would be helpful for the virus to mutate relatively fast, as well — DNA viruses like adenoviruses (which possibly do continue replication in persistent infection) have a relatively low mutation rate compared to RNA viruses like poliovirus or retroviruses like HIV.

Immune escape by hepatitis C virus

There are several lab animal and veterinary examples where something parallel to HIV immune evasion probably takes place (mouse hepatitis virus; perhaps feline infectious peritonitis virus), but the most popular choice for a close parallel to HIV is another human virus, hepatitis C virus (HCV). HCV, like HIV, can establish a chronic infection in immunocompetent people, and continually replicates. It’s been known that HCV mutates over time, and there’s decent evidence that much of this mutation is driven by escape from CTL. The parallels have become stronger with new evidence¹ suggesting that virus fitness changes are an important factor in HCV immune evasion, as well, but there’s a twist that I, as least, haven’t seen in HIV.

In these experiments, the authors experimentally infected a chimpanzee with HCV (as with HIV, there are not many good animal models for HCV) and tracked the immune response, and the dominant viral genome sequences, over some seven years. What’s more, the authors then tested those mutant viruses for their ability to replicate, persist, and evade immunity.

As expected, HCV threw out a bunch of mutations, especially in the early stage of infection, and those mutants were not recognized by the immune system. On the other hand (also as predicted, from the work on HIV) these early mutants were not as good at replicating as was the original (immunogenic) virus: They were less fit.

HCV escape variants can be fit

But — and here’s the twist — there were at least three variants that evaded CTL, arising early (3 months, for one variant), or moderately late (10 months; two variants). The one that arose early was clearly less fit.  It didn’t replicate well even when there was no immune system controlling it (that is, in cultured cells in the incubator), and given the chance, it mutated back to the original sequence.

The later ones, though, were not obviously as damaged; they replicated pretty well in cultured cells, and they did not mutate back to the original parent sequence.

What’s more, one of these relatively-fit variants persisted, more or less unchanged, in the infected chimp for years after it arose. (The other variant, though “fit” in culture, was not in the host, because a new immune response arose that targeted that variant.)

So for HCV, CTL escape mutants can arise, and there is not necessarily a loss of fitness associated with the immune evasion. I don’t remember seeing this established with HIV, but it’s quite likely the same thing happens. (Perhaps it’s even been described in the literature and I’ve missed it.) When immune escape variants are fit and healthy viruses, the immune system hasn’t even imposed a fitness cost on the mutation, and the virus isn’t being significantly controlled by the immune response.

It would be nice to understand better where fitness costs arise, which immune responses drive viruses to these un-fit variants, and how to focus the immune response on a vulnerable target.

1. Luke Uebelhoer, Jin-Hwan Han, Benoit Callendret, Guaniri Mateu, Naglaa H. Shoukry, Holly L. Hanson, Charles M. Rice, Christopher M. Walker, Arash Grakoui, Darius Moradpour (2008). Stable Cytotoxic T Cell Escape Mutation in Hepatitis C Virus Is Linked to Maintenance of Viral Fitness PLoS Pathogens, 4 (9) DOI: 10.1371/journal.ppat.1000143[↩]

Mouse polyomavirus

When we talk about anti-viral T cells, we’re usually talking about cytotoxic T lymphocytes (CTL) that recognize a peptide in association with class I major histocompatibility complexes (MHC class I). MHC class I is extremely polymorphic; there are many hundreds of different MHC class I alleles.

At any rate, that’s true for the classical MHC class I genes. But as well as classical MHC class I, there are also (you’ll never guess) non-classical MHC class I genes. Lots of them. For the most part we don’t really know quite what they do. Typically they’re not very polymorphic, compared to classical class I alleles. Since the major hypotheses explaining which there are so many classical MHC class I alleles involve protection against pathogens, this might hint that non-classical MHC class I don’t behave like classicals — either they don’t protect against pathogens at all, or they do so in a very different way.

Both of these possibilities are true for various non-classicals. For example, some of the non-classical MHC class I genes act as ligands for natural killer (NK) cells, which do recognize pathogens but do so in a very different manner than do CTL. Other non-classicals seem to have little to do with pathogen recognition — there are iron-binding molecules, neuronal molecules, and so on.

But a paper from Aron Lukacher’s lab¹ suggests that at least some non-classical MHC class I genes can act much like the classical genes, which has interesting implications for anti-viral vaccines.

There are mice that have no classical MHC class I genes at all. (Hidde Ploegh’s lab generated them, by crossing mice lacking the H-2K MHC class I gene with those lacking the H-2D MHC class I gene. Since these genes are side by side, the frequency of crossing-over is very low, and Ploegh got the mice by pure brute force, examining thousands of mice to find the single mouse that had crossed over appropriately to generate a double knockout. I have always felt a mixture of admiration and sympathy for the post-doc assigned that project.) These mice actually do reasonably well as far as controlling viruses — the immune system is highly redundant, and there are many antiviral systems in play – so it wasn’t a huge surprise to find that the classical MHC class I-less mice could control infection with polyomavirus quite well.

What was a surprise was that eliminating CD8 T cells also eliminated polyomavirus control. ² If there’s no classical MHC class I, what could the CD8 T cells be recognizing? The answer turned out to be, non-classical MHC class I. ³ WIth a bit more mapping, it seems that the CD8 T cells were recognizing a non-classical MHC class I gene called “Q9″ (catchy, eh?).

Q9 is a member of the Qa-2 family, which I have always believed were involved in natural killer cell recognition,⁴ although in retrospect I see that there’s evidence that CD8 T cells can recognize them.⁵  That’s a picture of Qa-2 off to the right (click for a larger version), and there are more images of it in my previous post comparing the different types of MHC.

What’s more, the CD8 cells recognize a really pretty conventional epitope. Classical MHC class I alleles present peptides that are around 9 amino acids long, while various non-classical proteins present anything from glycolipids to nothing at all. The Q9 complex turned out to present a rather boring 9-amino acid peptide⁶ that, to my eye, could have been cheerfully presented by any of a hundred classical MHC class I complexes.

So basically, this is a very conventional-looking anti-viral response, but directed against an unconventional MHC. There are occasional hints in the literature that this might be more than a one-off,⁷  though it’s not clear how common it is.  (This unconventional response might normally be drowned out by the conventional response, so that it’s hard to see unless you look in the MHC class I knockout mice.)

Why is this interesting (apart from the obvious point that anything remotely to do with MHC is intrinsically interesting, of course)? As I said, classical MHC is highly polymorphic, while non-classical MHC is not. There are only a handful of Q9 alleles known. If you made an antiviral vaccine with a peptide that binds to Q9, it should work in most mice, whereas if you make a similar vaccine directed to classical MHC class I, you’d need to tailor the vaccine to each individual mouse strain. Humans don’t have Q9 (the non-classical MHC are much less conserved between species than are the highly conserved classicals), but they do seem to have analogous proteins that are non-polymorphic and that may be able to work in antiviral contexts.

There’s a couple of other interesting things about this (Q9 binds to a wide range of peptides,⁸ for example, which reminds me of an existential question I had that was prompted by chicken MHC; and Lukacher makes the interesting suggestions that polyomavirus and Q9 might be the product of specific co-evolution), but I have a pair of intrepid little boys who camped out in our back yard for the first time last night, and they are proudly hiking the five feet to the house to tell me all about it now.

1. Swanson, P.A., Pack, C.D., Hadley, A., Wang, C., Stroynowski, I., Jensen, P.E., Lukacher, A.E. (2008). An MHC class Ib-restricted CD8 T cell response confers antiviral immunity. Journal of Experimental Medicine, 205(7), 1647-1657. DOI: 10.1084/jem.20080570[↩]
2. At any rate, the virus titre went up something like 50-fold.[↩]
3. A clue came from their previous finding the mice lacking β₂-microglobulin were highly susceptible to polyomavirus infection. Many, though not all, non-classical MHC class I proteins need β₂-m to form a normal structure.[↩]
4. The nonclassical major histocompatibility complex molecule Qa-2 protects tumor cells from NK cell- and lymphokine-activated killer cell-mediated cytolysis. Chiang EY, Henson M, Stroynowski I. J Immunol. 2002 Mar 1;168(5):2200-11. [↩]
5. Chiang, E.Y., and I. Stroynowski. 2005. Protective immunity against disparate tumors is mediated by a nonpolymorphic MHC class I molecule. J. Immunol. 174:5367-5374.
   and
   Correction of defects responsible for impaired Qa-2 class Ib MHC expression on melanoma cells protects mice from tumor growth. Chiang EY, Henson M, Stroynowski I. J Immunol. 2003 May 1;170(9):4515-23. [↩]
6. HALNVVHDW[↩]
7. For example, Braaten, D.C., J.S. McClellan, I. Messaoudi, S.A. Tibbetts, K.B. McClellan, J. Nikolich-Zugich, and H.W. Virgin. 2006. Effective control of chronic {gamma}-herpesvirus infection by unconventional MHC Class Ia-independent CD8 T cells. PLoS Pathog. 2:e37.[↩]
8. Promiscuous antigen presentation by the nonclassical MHC Ib Qa-2 is enabled by a shallow, hydrophobic groove and self-stabilized peptide conformation.  He X, Tabaczewski P, Ho J, Stroynowski I, Garcia KC.  Structure. 2001 Dec;9(12):1213-24.[↩]

CTL (green) and HSV-infected cells (red) (from Akiko Iwasaki)

Last time I talked about herpesvirus immune evasion of cytotoxic T lymphocytes, I cautiously wondered if there might be a theme emerging: Do these genes mainly help the virus with latent infection?

Immune evasion of CTL seems to be pretty much universal among the millions of different herpesvirus species — at least, as far as I know, in every case where people have looked for it, the virus has some way of blocking antigen presentation. Although other virus families also block antigen presentation (HIV, some poxviruses, and human adenoviruses are probably the best known instances), immune evasion isn’t as universal among those other families.

For example, although human adenoviruses mostly have immune evasion function, adenoviruses of other species do not (as far as we can tell); and for that matter not even all human adenoviruses have the ability to block antigen presentation. What’s more, there is a trend for those non-herpesvirus viruses that do evade CTL, to also establish long-term latent or persistent infection.

A recent paper from Frank Carbone’s lab¹ offers a little more, indirect, support for this theme. They asked what CTL actually do to herpes simplex virus in the initial infection.

We usually blithely call CTL “antiviral lymphocytes”, but what exactly does that mean for specific virus infections? For example, I’ve previously pointed out experiments that show that CTL have more than one way of clearing away virus infections — they can use cytokine secretion as well as, or instead of, cell lysis, as their weapon, which opens up the opportunity for activity over a broad range rather than one cell at a time. In another example, Luis Sigal’s lab has shown that CTL can protect against extromelia (mousepox) infection at a very early stage, by blocking the virus’s spread from the skin to the liver, cutting them off at the bottleneck of the lymph through which the virus intially spreads.

On the other hand, herpes simplex virus often seems to do just fine even when CTL are present. The virus sets up an initial infection in the skin, and rapidly tracks up through neurons to ganglia, where it sets up a latent infection. By the time CTL are up and running, the virus is comfortably snuggled down in the neuron, shutting down all its protein expression to the point where CTL don’t see it very well. Even if you have an active CTL response already, the virus seems to be able to get in to the neurons and set up latency anyway.

So what do CTL do to herpes simplex? Carbone’s lab set up mice with and without specific anti-HSV CTL, and infected them with the virus. As you’d expect (and as has been shown lots of times in the past) the CTL markedly reduced the amount of virus replication and shedding, but did not prevent the virus from setting up a latent infection.

Though the presence of herpes-specific effector CD8+ T cells early during viral challenge attenuated the primary infection and prevented the development of disease, these cells failed to block the skin to nervous system transmission of the virus, and hence substantial latent infections were established in the face of this CTL immunity.

(My emphasis.) How come? Part of the answer seems to be that the CTL didn’t respond quite fast enough. ² Virus infects the skin, replicates, and moves up into neurons in about 24 hours. (If they surgically removed the infected skin prior to 24 hours after infection, neurons weren’t infected; if they removed the infected region more than 24 hours after infection, neurons were infected.) CTL, on the other hand, move into the infected region of skin within about 15 hours of infection. At this point the CTL start to shut down virus replication; but the window of opportunity, as you can see, is very narrow. The CTL need to shut down the virus in the skin very rapidly, and to very low levels, within just a few hours of entering the site of infection.

In fact, if you start off with a relatively small amount of virus, then CTL can shut the replication down enough to make a difference.  It’s mainly when there’s a lot of virus to start with that the CTL can’t get the virus down under some threshold level that allows efficient latent infection:

Thus, virus-specific CTL can reduce the average viral copy number of the residual latent infection, but this is only achievable when a substantial attenuation of the skin infection is observed.

There are a bunch of fascinating points that arise from this work. First, it helps account for the fact that superinfection with herpes simplex is actually quite rare — that is, if you’re infected with HSV already, then you’re unlikely to get re-infected with a second virus. Normal exposure to HSV probably is at a very low level, with only a handful of virions entering the skin; it’s more like the low-range infection in Carbone’s experiments than the high-range, where they put in some 10 million virions, and at the low range, CTL can move in and check the initial infection fast enough to make a difference.

Second, a critical point about these experiments is that they were done in the absence of CTL evasion. That’s because there experiments were done in mice, and the HSV immune evasion molecule ICP47 doesn’t work in mice, as opposed to in humans.

One of the puzzling things about immune evasion genes, to me, has been how ineffectual they seem to be in authentic infections. But these experiments suggest if your interest is in establishing latency, then immune evasion doesn’t need to be hugely effective: It just needs to keep the window open a crack for a few more hours, letting the virus replicate through the first wave of CTL invasion. If ICP47 can hold off the CTL for an extra 8 hours, then it’s probably done its job, allowing HSV to set up a latent infection and thus reactivate and infect new hosts on and off over the next 60 or 70 years.

So, even though this paper really didn’t look at immune evasion per se, I think it does offer some support for the concept that (for herpesviruses, anyway) immune evasion really isn’t for the acute infection at all.  Instead, it’s a mechanism to help the virus establish latent infection.

1. Wakim, L.M., Jones, C.M., Gebhardt, T., Preston, C.M., Carbone, F.R. (2008). CD8+ T-cell attenuation of cutaneous herpes simplex virus infection reduces the average viral copy number of the ensuing latent infection. Immunology and Cell Biology DOI: 10.1038/icb.2008.47[↩]
2. I wonder whether they miigh have seen a faster response if they had started with skin-specific T cells, though.[↩]

Particularly important are questions regarding the effect of the quality, quantity and specificity of the vaccine-induced CD8 T-cell response on post-infection viral load control. These questions could not be addressed without a vaccine approach that actually induced CD8 T-cell responses in most recipients.

The Merck vaccine was the first such candidate, so it is a misleading exaggeration to claim that its failure is a “crisis” for HIV vaccine research. A journal with Nature’s long history is well placed to know how likely first-time successes are in science.

–Richard Jefferys,¹ in Correspondence to Nature:
Vaccine failure is not a ‘crisis’ for HIV research
Nature 453, 719-720 (5 June 2008) | doi:10.1038/453719d

1. See the Michael Palm Basic Science, Vaccines & Prevention Project Weblog[↩]

The other day I was talking about immune recognition of human endogenous retroviruses (HERVs) in tumors. (HERVs are the husks of ancient retroviruses, now trapped in our genomes. Some of them still express various proteins, either under normal conditions or when stimulated, as in tumors.) One of the reasons this is an interesting finding, is that HERVs may offer a relatively constant antigen, even though the tumors themselves may be highly variable.

There are other, rather obvious, scenarios in which we would like to have a constant antigen in the face of an antigenically-variable disease. For example, HERVs have been proposed to be useful vaccine targets in HIV infection.

One of the many obstacles to overcome in developing a vaccine against HIV is the virus’s rapid mutagenesis. Because of its error-prone replication, HIV can readily escape a lot of immune recognition. Especially when cytotoxic T lymphocytes (CTL) recognize a limited number of antigenic targets, all the virus needs to do to escape immune control is mutate a single amino acid. Usually once the virus does this and escapes immune control, new CTL arise and once again shut down the virus, but only to have new escape variants arise and replicate. Over the multi-year course of an HIV infection, there may be dozens or hundreds of major HIV variants, each escaping temporarily from CTL and destroying T cells during their limited period of freedom.

There are several strategies aimed at reducing the effectiveness of immune escape: targeting multiple HIV antigens, so that the virus would have to simultaneously find many mutations at once; targeting regions in the virus that are so essential that they can’t tolerate mutation; and so on. But wouldn’t it be nice if there was an antigen that wouldn’t change?

HIV and HERVs are distant cousins, both retroviruses, so it seems reasonable that HIV infection might turn on sleeping HERVs. In fact, for nearly 10 years there have been intermittent studies suggesting this might be the case; first based on antibody responses in HIV patients¹, and recently with more specific evidence of reactivation of HERVs both in patients² and in the lab, in infected cells.³

Last fall Douglas Nixon’s group took this to the next step.⁴ Although the antibody responses¹ had suggested that HERVs were immunogenic when turned on by HIV, antibodies aren’t believed to be terrible important in control of HIV; rather, CTL are thought to be critical.⁵ Nixon’s group showed that in HIV-infected people, there were often functional CTL responses to HERVs; what’s more, the higher the anti-HERV response, the lower the HIV plasma load, implying that the anti-HERV CTL might actually be controlling HIV. (See the figure to the left; click for a larger version.)

As endogenous retroviral sequences are fixed in the human genome, they provide a stable target, and HERV-specific T cells could recognize a cell infected by any HIV-1 viral variant. HERV-specific immunity is an important new avenue for investigation in HIV-1 pathogenesis and vaccine design.

Let’s go back to a paper⁶ I mentioned last year, where a group looked at genomic variation linked to disease progression in HIV. They found three genomic regions that were linked to viral set-point; one is an RNA polymerase, one is in the MHC region and affects levels of the MHC class I gene HLA-C, and the third … well, the third is a HERV, called HCP5.

The authors pointed out that HCP5 might not be the actual factor involved, because it might be riding along with HLA-B*5701, an MHC class I allele that’s associated with HIV resistance (and I noted that natural killer ligands MICA and MICB are also close by). Still, they clearly like the idea that HCP5 is itself directly involved. They suggested that it might act by an antisense mechanism or something, but I think it might be very interesting to look at CTL responses to HCP5 proteins.

1. Stevens RW, Baltch AL, Smith RP, McCreedy BJ, Michelsen PB, Bopp LH, Urnovitz HB (1999) Antibody to human endogenous retrovirus peptide in urine of human immunodeficiency virus type 1-positive patients. Clin Diagn Lab Immunol 6:783-786.[↩][↩]
2. Contreras-Galindo R, Kaplan MH, Markovitz DM, Lorenzo E, Yamamura Y (2006) Detection of HERV-K(HML-2) viral RNA in plasma of HIV type 1-infected individuals. AIDS Res Hum Retroviruses 22:979-984.[↩]
3. Contreras-Galindo R, Lopez P, Velez R, Yamamura Y (2007) HIV-1 infection increases the expression of human endogenous retroviruses type K (HERV-K) in vitro. AIDS Res Hum Retroviruses 23:116-122.[↩]
4. Garrison, K.E., Jones, R.B., Meiklejohn, D.A., Anwar, N., Ndhlovu, L.C., Chapman, J.M., Erickson, A.L., Agrawal, A., Spotts, G., Hecht, F.M., Rakoff-Nahoum, S., Lenz, J., Ostrowski, M.A., Nixon, D.F. (2007). T Cell Responses to Human Endogenous Retroviruses in HIV-1 Infection. PLoS Pathogens, 3(11), e165. DOI: 10.1371/journal.ppat.0030165[↩]
5. Because of the way CTL recognize their targets, by the way, it doesn’t matter if the HERVs produce defective proteins — even a truncated protein that is unstable and rapidly destroyed might be a good CTL target.[↩]
6. Fellay, J., Shianna, K. V., Ge, D., Colombo, S., Ledergerber, B., Weale, M., et al. (2007).A whole-genome association study of major determinants for host control of HIV-1. Science, 317(5840), 944-947.[↩]