Mystery Rays from Outer Space

One of the interesting things about the retrovirus family is that it contains some of the most lethal viruses we know of, and also some of the least harmful. For obvious reasons, the lethal family members (HIV) get lots of recognition, while their harmless cousins get little more than an occasional mention. There are more — far more — articles on HIV in the past two weeks, than there have been on the spumaviruses in the past 50 years. In fact, it’s only very recently that people have demonstrated why the spumaviruses are actually harmless. ¹

Spumaviruses, or foamy viruses, infect many species — though not, apparently, humans,² even though they’re common infections of old world monkeys and apes. Normally I’d assume that the lack of human spumaviruses was simply because nobody has looked hard enough, but I think in this case that simple answer isn’t true; lots of groups have been looking and haven’t found any evidence. Perhaps they’re only found in a limited number of tissues, or under certain circumstances, but I believe the work has been pretty thorough. Indeed the work has been thorough enough to turn up people infected with spumaviruses, but only because of cross-species infections from apes and so on³ — even though the detection methods are apparently adequate, there’s no evidence for human to human transmission.⁴

Anyway, spumaviruses infect many species, and in all of them appear to be completely harmless. ⁵ Yet in tissue culture, these viruses are deadly, rapidly killing infected cells.⁶ I’ve always assumed that the reason spumaviruses are harmless in an authentic infection is that the host immune system controls the virus, preventing it from replicating so that its host cells don’t get killed by the virus. This turns out not to be the case.

You may be wondering, by the way, why anyone should care why they’re harmless. Why not accept that they are harmless, and concentrate on dangerous pathogens? Apart from being interesting in their own right, I think part of the answer is probably obvious — is there some underlying mechanism that can be applied to a different pathogen, to make it apathogenic?

But another part of the question is viral vectors. There’s intense interest in developing viruses as gene therapy vectors, and one of the obstacles to many of the present candidate vectors — adenoviruses, adeno-associated viruses, lentiviruses, and so on — is that the vector is intrinsically harmful. To use these vectors to treat disease you need to render the virus harmless, first and foremost; and almost ineviatably, this reduces the virus’s ability to infect, which is why you want it in the first place. If you can start with a virus that’s already harmless, then you’re halfway to your goal from the start. There’s increasing interest in developing spumaviruses as gene therapy vectors,⁷ and so we really would like to know why they are harmless.

It turns out, surprisingly (to me, anyway) that spumaviruses are probably just as cytopathic in their natural host, as they are in cultured cells. However, they strictly limit their replication to expendable cells, that are going to die anyway: epithelial cells that are about to be sloughed off and shed anyway.

… FV replication is limited to cells at a very late stage of epithelial cell differentiation, when the cells are about to be shed from the tissue. … The most superficial terminally differentiated cells in the epithelial cell layer turn over in about 3 hours. Thus, the limitation of viral replication to a relatively expendable, superficial cell type could account for the innocuous character of FV infection in vivo. … The contrasting in vitro situation, in which FV infection spreads rapidly and is cytopathic to the entire culture, may be an artifact of culture systems that do not represent the differentiated cell types or cell turnover rates within mucosal epithelial tissues.

My first reaction is that this isn’t very encouraging for spumaviruses as gene therapy vectors, but that’s not necessarily true; the viruses can infect many other tissues, though probably latently, and that may well be perfectly adequate for gene therapy.

I’ve pointed out in the past that pathogens don’t necessarily evolve toward reduced virulence; rather, pathogens evolve toward efficient transmission, which may or may not involve a healthier host. This is a lovely example of efficient transmission coupled with reduced virulence:

This mechanism of viral replication promotes efficient virus transmission via shed cells while limiting viral replication to a superficial site and thus minimizing host tissue damage.

The authors make a comparison to some other virus families that do something quite similar — herpesviruses and papillomaviruses, which as a group often infect one cell type latently but undergo lytic, active replication in differentiated skin cells. They say

… herpesviruses and papillomaviruses, with similar replication patterns, are also ancient viruses. Differentiation-specific viral activation within differentiated cells of the oral mucosa or the skin arose independently in these diverse virus families. … Indeed, in the majority of infections, none of these viruses leads to pathogenic sequelae.

What they mean by “ancient” here, I think, is that they’ve co-evolved with their hosts for a long time, and are well adapted. Herpes and papillomaviruses are more often pathogenic than are spumaviruses, I think, but it’s an interesting suggestion.

1. Murray, S.M., Picker, L.J., Axthelm, M.K., Hudkins, K., Alpers, C.E., Linial, M.L. (2008). Replication in a Superficial Epithelial Cell Niche Explains the Lack of Pathogenicity of Primate Foamy Virus Infections. Journal of Virology, 82(12), 5981-5985. DOI: 10.1128/JVI.00367-08[↩]
2. The so-called human foamy virus that was isolated in the 1970s is actually a chimp virus that was probably acquired through interspecies transmission.[↩]
3. Calattini S, Betsem EB, Froment A, Mauclere P, Tortevoye P, Schmitt C, Njouom R, Saib A, Gessain A (2007) Simian foamy virus transmission from apes to humans, rural Cameroon. Emerg Infect Dis 13:1314-1320.[↩]
4. Boneva RS, Switzer WM, Spira TJ, Bhullar VB, Shanmugam V, Cong ME, Lam L, Heneine W, Folks TM, Chapman LE (2007) Clinical and virological characterization of persistent human infection with simian foamy viruses. AIDS Res Hum Retroviruses 23:1330-1337.[↩]
5. It’s possible that they are associated with very rare disease and because humans aren’t involved there isn’t the intense search for cause that human disease often leads to; but the virus is really very common, infecting at least a third of cats, cattle, and primates in the wild; at that frequency of infection, even rare diseases would probably be picked up in cats or cattle.[↩]
6. Saib A (2003) Non-primate foamy viruses. Curr Top Microbiol Immunol 277:197-211.[↩]
7. For example, Bastone P, Romen F, Liu W, Wirtz R, Koch U, Josephson N, Langbein S, Lochelt M (2007) Construction and characterization of efficient, stable and safe replication-deficient foamy virus vectors. Gene Ther 14:613-620.
   Also, reviewed in Rethwilm A (2007) Foamy virus vectors: an awaited alternative to gammaretro- and lentiviral vectors. Curr Gene Ther 7:261-271.[↩]

A while ago I talked about evolution of the herpesviruses, and I said:

We know of 200-odd herpesviruses so far, and more are being identified practically daily. It’s likely that virtually every animal species has its own set of unique herpesviruses. This is probably because herpesviruses are very host-restricted (rarely infecting more than a single species) and set up latent, life-long infection. When an animal population speciates, its complement of herpesviruses speciates along with it.

Word of my blog post spread like wildfire (as is inevitable for a blog that is read by upward of five people, including my mother) and Duncan McGeoch hastened to correct me with a fascinating paper now in pre-print form at the Journal of Virology:
Ehlers, B., Dural, G., Yasmum, N., Lembo, T., de Thoisy, B., Ryser-Degiorgis, M., Ulrich, R.G., McGeoch, D.J. (2008). Novel mammalian herpesviruses and lineages within the Gammaherpesvirinae: Cospeciation and interspecies transfer. Journal of Virology DOI: 10.1128/JVI.02646-07

McGeoch’s group went herpesvirus hunting (”Be vewwy, vewwy quiet!”) and — supporting at least the first part of my quote here — found no less than 14 brand-new gammaherpesviruses. (They would have found 16 new viruses, but their elephant gammaherpesvirus was described earlier this year when Wellehan et al earlier this year found 6 new gammaherpesviruses of elephants, rock hyraxes, and manatees,¹ and their gorilla virus was described a few years ago by Gessain’s group² )

The new gammaherpesviruses come from shrews, tapirs, rhinocerous, zebras, babirussas, pygmy hippos, and so forth (see the table to the right, and see the paper itself for accession numbers, I’m not going to copy them all down). They used a fiddly, if not conceptually difficult, technique to get a reasonable sequence length (3.4 kb) for phylogenetic analysis, rather than the couple hundred base pairs that this sort of virus fishing expedition usually yields, and then lined up the newly expanded complement of gammaherpesviruses, which now includes no less than 45 viruses.

I won’t reproduce their phylogenetic tree here (the actual figure from the preprint has the “ACCEPTED” watermark stamped all over it, and I don’t have time to get 45 virus sequences and re-build the alignment according to their description). The interesting thing about this tree is that it tells a somewhat different story than an earlier and more limited analysis.

In earlier analyses of Herpesviridae phylogeny, it was possible to discern within each of the subfamilies a substantial degree of congruence in tree branching pattern with the corresponding pattern for lineages of the mammalian hosts, and this was taken as indicating extensive cospeciation of herpesviruses and hosts.

Even in these previous analyses, the gammaherpesviruses were the worst match for the co-speciation hypothesis — that is, closely-related gammaherpesviruses were found in mammalian hosts that were not very closely related, suggesting that the viruses’ ancestors might have jumped hosts at some point. That observation was even more true in this larger set, especially in the more ancient divisions:

… applicability of cospeciational interpretation declines further with the extensive detail now available for the GHV tree. … the deeper branching details of the tree prove rather unproductive for constructing any unified coevolutionary correspondence across host lineages. In particular, the two deepest distinct lineages, i.e. LCV and EmaxGHV-1, are not simultaneously compatible with a single cospeciational scheme, and in the MF2 clade the unresolved nodes for major lineages do not enable any compelling interpretation. On the other hand, clear dispersed examples of cospeciation can be seen in the terminal branchings within major lineages. … In summary, there are substantial indications in the GHV tree of evolution both by cospeciation with host lineages and by transfer between widely distinct hosts.

(My emphasis.)

There are at least two really interesting implications from this work. First, McGeoch’s group supports the previously-proposed notion that there may be more human herpesviruses yet to be found, because there are major divisions of the gammaherpesvirus families that don’t yet have human representatives. Second, although I’m not aware of any instances of non-human herpesviruses successfully³ infecting humans, this work suggests that at least in the past, and therefore likely today, gammaherpesviruses could make interspecies leaps.

Just something else to watch for.

1. Wellehan JFX, Johnson AJ, Childress AL, Harr KE, Isaza R (August 19, 2007) Six novel gammaherpesviruses of Afrotheria provide insight into the early divergence of the Gammaherpesvirinae. Vet Microbiol. doi:10.1016/j.vetmic.2007.08.024 [↩]
2. Either Lacoste V, Mauclre P, Dubreuil G, Lewis J, Georges-Courbot M, et al. (2000) Virology: KSHV-like herpesviruses in chimps and gorillas . Nature 407.:151-152
   or
   Lacoste V, Mauclère P, Dubreuil G, Lewis J, Georges-Courbot MC, et al. (September 2001) A novel gamma 2-herpesvirus of the Rhadinovirus 2 lineage in chimpanzees. Genome Res 11.:1511-9.
   I didn’t check the sequences to see which reference is McGeoch’s gorilla virus.[↩]
3. Where “successfully” means not merely infecting and killing occasional individuals, but actually spreading from person to person — I know about herpesvirus B and so on[↩]

TcR interacting with artificial membrane1

Earlier this week I talked about the phenomenon of viruses that downregulate MHC class II. The “purpose” of this blockade is kind of unclear to me, because the immunity driven by MHC class II is not focused on the cell it’s attached to, but rather spills out broadly over a wide area; it seems that a virus would have to very rapidly infect a very large number of MHC class II-expresing cells to have much effect on the anti-viral immune system.

One possible explanation is that the downregulation may be only peripherally related to MHC class II-based immunity. Instead, the virus could be simply going about its cell-biological business and either targeting MHC class II as an accidental side-effect, or because of some function of MHC class II that isn’t directly related to immunity. Here’s a parallel case that may help think about the problem.

The paper is
Sullivan, B.M., Coscoy, L. (2007). Downregulation of the T-Cell Receptor Complex and Impairment of T-Cell Activation by Human Herpesvirus 6 U24 Protein. Journal of Virology, 82(2), 602-608. DOI: 10.1128/JVI.01571-07

The T-cell receptor (TcR) is what recognizes MHC class I or II. Human herpesvirus 6 (HHV6) infects T helper (CD4) cells and (depending on viral strain) also does a number of things to modulate the immune system.² Sullivan and Coscoy show here that the virus also down-regulates the TcR on infected T cells. (It’s altogether a more solid paper than the last one I mentioned, with nice experiments that directly show what’s happening to the TcR: “HHV-6 U24 protein inhibits CD3 recycling to the cell surface and, as a consequence, downregulates CD3 cell surface expression and prevents T-cell activation.“)

U24 blocks CD3ε access to Rab11 recycling endosomes.3

At first glance this raises exactly the same question as does Vpu’s effect on MHC class II. How does reducing TcR on infected cells benefit the virus? The infected T cell will be less able to recognize its target, but what are the odds that its target is HHV6? Pretty minimal; there are (at least) tens of billions of different TcRs and only a handful of them recognize any particular antigen. The virus might be causing generalized immune suppression if it infects a large fraction of the T cells, but that’s not a particular benefit for the virus. If HHV6 specifically infected cells with TcRs that are specific for the virus then this would be a targeted immune evasion technique, but as far as I know there’s no evidence for this, nor is there an obvious mechanism by which HHV6 could target antigen-specific CD4 T cells.

There is, however, a nice explanation for TcR downregulation that doesn’t involve direct effects on T cell recognition. HHV6 (like all herpesviruses) has two choices when it infects a cell. It can either enter lytic replication — replicating the genome, producing more viruses, and eventually destroying the infected cell — or enter latency — a long-term, perhaps life-long, infection with minimal protein expression and little if any effect on the infected cell. As with any virus, the more prepared a cell is to replicate, the easier it is for a virus to replicate it’s own genome. T cells that receive a signal through their TcR become activated⁴ and divide very rapidly. In this environment, it’s very easy for HHV6 to replicate — that is, to enter lytic replication and kill the cell, releasing more viruses into the system.

Human herpesvirus 6

Probably HHV6 downregulates the TcR “because” it prevents its host from becoming activated by whatever its random antigen is. That prevents the virus from entering lytic replication and allows it to enter a persistent state, where it can hang about and await the best opportunity to infect a new person.

I know of one other viral protein that downregulates the TcR — the herpesvirus saimiri Tip protein ⁵ — and there seems to be controversy⁶ over whether this protein activates T cell signalling (potentially driving the virus into lytic replication) or blocks it (preventing lytic replication and facilitating persistence). ⁷ The original paper describing the TcR downregulation found that Tip blocked signaling, and proposed the same explanation as Sullivan and Coscoy:

… these associations ultimately block lymphocyte receptor signal transduction. … these interactions likely play an important role in the establishment and maintenance of HVS persistent infection by protecting infected cells from surveillance by the immune system. In fact, animals infected with recombinant HVSΔTip have been shown to have higher levels of cell-associated infectious virus titer compared to other recombinant HVS.

So in this case the downregulation of the TcR (a quintessentially immune molecule) apparently isn’t directly related to immune evasion, but is a way of switching between the lytic and the persistent lifestyles. (It’s also a reminder of the fairly obvious point that we shouldn’t think of viruses as blind replicators, desiring nothing more than maximal replication. At least some viruses have a range of lifestyle options, and can switch between them quite comfortably.)

I still don’t see a direct analogy to the MHC class II downregulation imposed by the HIV Vpu protein. but it’s an example of why we shouldn’t get too focused on single causes and single functions. Life is more complicated than that.

1. By Raghuveer Parthasarath, then in the Groves lab[↩]
2. Lusso, P., 2006. HHV-6 and the immune system: mechanisms of immunomodulation and viral escape. Journal of Clinical Virology, 37(Supplement 1), p.S4-S10. doi:10.1016/S1386-6532(06)70004-X [↩]
3. Sullivan & Coscoy, Fig 4[↩]
4. I am simplifying immensely![↩]
5. Park, J. et al., 2002. Herpesviral Protein Targets a Cellular WD Repeat Endosomal Protein to Downregulate T Lymphocyte Receptor Expression. Immunity, 17(2), p.221-233. doi:10.1016/S1074-7613(02)00368-0 [↩]
6. Brinkmann, M.M. & Schulz, T.F., 2006. Regulation of intracellular signalling by the terminal membrane proteins of members of the Gammaherpesvirinae. J Gen Virol, 87(5), p.1047-1074. DOI 10.1099/vir.0.81598-0[↩]
7. I’m more convinced by the argument for blocking signaling, but only because of the very bad reason that I know the people involved. I haven’t looked at the papers pro and con very carefully.[↩]

Oncolytic VSV (gold) infecting lung tumors1

Oncolytic viruses are a concept I’d like to be more excited by than I am.² It’s an idea that seemed really exciting when I first came across it, but the more I thought about it the more dubious I was. But a recent paper helps me feel better about at least two of my worries.

The concept is a straightforward one. Viruses are good at killing cells.³ Why not have them infect cells that we want to die? That would be, for example, cancer cells. So all you need to do is find or make a virus that only grows in cancer cells, and you’re cured. Simple! Tomorrow we’ll fix global warming!

There’s the obvious problem with this: How can you find (or make) a cancer-specific virus? In principle the answer is the same as with chemotherapy; you use the ways cancer cells are different from normal as targets. This isn’t as hard as you might think. Lots of the things that make cancer cells cancerous are similar to the things viruses like. Viruses often drive infected cells into a cancer-like state that is more hospitable to the virus — friendly to nucleic acid replication, replication, unresponsive to death signals, independent of the signals that normally regulate growth. So lots of viruses are already kind of pre-adapted to replicate well in cells with a cancerous phenotype, and it doesn’t take all that much tweaking to make them adapted to only replicate well in cancer cells.

(After writing this post, it occurred to me that this is actually topical! I don’t usually do the topical blog post thing, but the background in “I Am Legend” has an anti-cancer virus, isn’t it? I haven’t seen it myself, with the grant-writing and the teaching and the two little kids,⁴ but have I actually tied a current entertainment topic into “Mystery Rays”? Fame and fortune is certain to come my way!)

Oncolysis through the ages

The first runs at this technique that I knew of⁵ used mutant herpesviruses,⁶ but I think that much of the buzz came from work with defective adenoviruxes, especially the ONYX-015 virus.⁷ The approach here was based on the observation that adenoviruses (like many other viruses) normally inactivate p53 during infection. p53 is a multifunctional growth regulator that is very often also inactivated in cancers, for the same reason as viruses like to inactivate it:it oten triggers death in cells with unchecked growth. Adenoviruses lacking the gene that inactivates p53 (their E1B gene) can only efficiently infect cells lacking p53 — which would usually be, of course, cancer cells.

As well as herpesviruses and adenoviruses, though, all sorts of other viruses have been used.⁸ One interesting approach is vesicular stomatitis virus. This is a very, very innocuous virus in normal people, partly because VSV is extremely sensitive to interferon. (VSV is used in bioassays for interferon release, because even tiny amounts of interferon completely block the virus’s replication.) So which kind of cells often aren’t responsive to interferon? Right; cancer cells, as part of their own immune evasion pathway, frequently disable their interferon responses. VSV doesn’t infect normal cells, but does infect, and kill, tumor cells.⁹

Questions and (maybe) answers

Anyway, the first question, of specificity, is more or less under control.¹⁰ Three questions that had made me rather dubious about the concept, though, still remained:

1. Getting the virus to the tumor …
2. Especially in the face of an immune response.
3. Killing all of the cancer cells, not a mere 99% of them (from which the cancer will rapidly recover).

Malignant melanoma cells in lymph node

A paper in Nature Medicine¹¹ offers encouragement on all of those.

They used VSV as their cancer killer, and their twist here was to deliver it by loading it onto T cells. T cells naturally traffic to lymph nodes, and quite a few tumors metastasize through lymph nodes; the T cell therefore acts as a ferry to deliver its deadly viral cargo to the metastasizing tumor. (The goal here was not to clear the primary tumor, but to prevent metastases, which are often the major problem.  However, they did see some effect on the primary tumor, too, in some cases.) When it reaches the lymphoid tissue, it delivers the passenger virus to the cancer cells, the only ones that the VSV can productively infect (since the cancer cells are the only ones that have mutated their interferon pathway). This is an interesting idea, though limited in this form — I wonder about using antigen-specific T cells instead, to target the virus to a specific site — and it seemed to work quite well.

The two more interesting points to me were kind of peripheral to their main point. First, they find that once the virus killed some cancer cells, there was anti-tumor protection even after the virus was all cleared, and this was probably because of the immune response,¹² which was triggered by the cell death initially caused by the virus:

In vivo tumor cell purging resulted both from direct viral oncolysis by virus released from the T cell carriers and from the priming of protective antitumor immunity, which prevents repopulation by further waves of cells metastasizing from the primary tumor.

– just as described in the paper by Apetoh et al¹³ that I talked about here. The authors suggest that because the cancer metastases are being killed in the lymph nodes, rather than in the bulk of the tumor (which is generally a highly immunosuppressive environment) the immune response was more efficient. That starts to get past my concern #3 above, because it offers multiple attacks on the tumor, not just the virus.

The other point is that the virus could reach the cancer reasonably well even in the face of an anti-viral immune response; the trick was to use just enough virus to kill the tumor cells, without getting enough on the T cells to trigger an immune response:

In virus-immune mice, T cells loaded with large amounts of VSV (MOI 1 or 10) could not keep DLNs or spleens free of tumors. However, T cells loaded with fewer viruses (MOI 0.1) still protected even virus-immune animals from tumor colonization of the DLN and spleen

The data are still very preliminary and inconclusive, but certainly it’s a step in the right direction, and I feel better about this whole approach than I did before reading the paper.

1. Carrier Cell-based Delivery of an Oncolytic Virus Circumvents Antiviral Immunity. Anthony T Power, Jiahu Wang, Theresa J Falls, Jennifer M Paterson, Kelley A Parato, Brian D Lichty, David F Stojdl, Peter A J Forsyth, Harry Atkins and John C Bell. Molecular Therapy (2007) 15, 123-130. [↩]
2. That sentence needs a road map, but you got here eventually, didn’t you.[↩]
3. At least, lytic viruses are.[↩]
4. And the running and the screaming and the monkeys in the hair[↩]
5. I realize now, though, that the concept arose long before that, apparently in the 1950s. For example: Love R, Sharpless GR. Studies on a transplantable chicken tumor, RPL-12 lymphoma. II. Mechanism of regression following infection with an oncolytic virus. Cancer Res. 1954 Oct;14(9):640-7.http://dx.doi.org/10.1126/science.1851332 though I don’t know much about those studies other than the titles [↩]
6. Experimental therapy of human glioma by means of a genetically engineered virus mutant. Martuza RL, Malick A, Markert JM, Ruffner KL, Coen DM. Science. 1991 May 10;252(5007):854-6. [↩]
7. ONYX-015, an E1B gene-attenuated adenovirus, causes tumor-specific cytolysis and antitumoral efficacy that can be augmented by standard chemotherapeutic agents. Heise C, Sampson-Johannes A, Williams A, McCormick F, Von Hoff DD, Kirn DH. Nat Med. 1997 Jun;3(6):639-45.
   and
   An adenovirus mutant that replicates selectively in p53-deficient human tumor cells. Bischoff JR, Kirn DH, Williams A, Heise C, Horn S, Muna M, Ng L, Nye JA, Sampson-Johannes A, Fattaey A, McCormick F. Science. 1996 Oct 18;274(5286):373-6. [↩]
8. And I have no idea which, if any, is the most promising.[↩]
9. Exploiting tumor-specific defects in the interferon pathway with a previously unknown oncolytic virus. David F. Stojdl, Brian Lichty, Shane Knowles, Ricardo Marius, Harold Atkins, Nahum Sonenberg & John C. Bell. Nature Medicine 6, 821 - 825 (2000) http://dx.doi.org/10.1038/77558 doi:10.1038/77558[↩]
10. One other point is that you can probably get away with a virus that isn’t completely restricted to tumor cells, because these are usually viruses that cause very mild disease anyway, so even if they can spread to normal cells it’s no more worry than exposure to a standard subway car. Maybe more concern for immunosuppressed cancer patients, of course, but likely not an insurmountable worry.[↩]
11. Qiao, J., Kottke, T., Willmon, C., Galivo, F., Wongthida, P., Diaz, R.M., Thompson, J., Ryno, P., Barber, G.N., Chester, J., Selby, P., Harrington, K., Melcher, A., Vile, R.G. (2007). Purging metastases in lymphoid organs using a combination of antigen-nonspecific adoptive T cell therapy, oncolytic virotherapy and immunotherapy. Nature Medicine DOI: 10.1038/nm1681[↩]
12. The short-term clearance worked in immune-deficient mice, but the long-term did not[↩]
13. Toll-like receptor 4-dependent contribution of the immune system to anticancer chemotherapy and radiotherapy. Apetoh, L., Ghiringhelli, F., Tesniere, A., Obeid, M., Ortiz, C., Criollo, A., Mignot, G., Maiuri, M. C., Ullrich, E., Saulnier, P., Yang, H., Amigorena, S., Ryffel, B., Barrat, F. J., Saftig, P., Levi, F., Lidereau, R., Nogues, C., Mira, J. P., Chompret, A., Joulin, V., Clavel-Chapelon, F., Bourhis, J., Andre, F., Delaloge, S., Tursz, T., Kroemer, G., and Zitvogel, L. (2007). Nat Med 13, 1050 - 1059. [↩]