Mystery Rays from Outer Space

There are maybe 10¹¹ naive T cells in a human body. How many of those T cells can recognize any particular antigen?

About twenty.

OKTHXBYE!

… Well, if you want a little more expansion of that (and all the weasel words that go along with it) …

T cells have to recognize the entire universe of all possible pathogens, and they generally manage to do so; it’s not often that people are infected with a pathogen that simply doesn’t elicit a T cell response. On the other hand, for all the possible antigens present in Generic Joe Pathogen, T cells typically only recognize a handful of them; you don’t see a massive upwelling of T cells that recognize every possible epitope in the pathogen.

Floating around your body, there are somewhere between, let’s say, 10⁸ (if you’re a mouse) and 10¹¹ (if you’re a human) naive T cells. ¹ It’s that population that must be prepared to take on our hypothetical universe of pathogens. In other words, the largest number of antigens you could possibly recognize is 10⁸ - 10¹¹, if each naive T cell recognized a distinct antigen.

Is there redundancy among T cell specificities? If so, how many T cells typically recognize an individual antigen? And therefore, how may distinct antigens can your body detect?

When a T cell is allowed to exit the thymus, where it matures, it has a T cell receptor (TcR). That TcR is what interacts with, say, a viral antigen, and what allows the T cell to respond in its specific and (hopefully) appropriate way. TcRs are formed by genomic rearrangement, shuffling a moderate handful of possible segments to form, by combinatorial multiplication, a very large number of possible sequences. (If you want a mechanism, see any introductory immunology text, or Wikipedia. ) How large is a “very large number of possible sequences”? In theory, it could be as many as 10¹⁵ different TcRs², but in practice it’s probably more like 10⁸.³ (And that’s 10⁸ possible clones — precise TcR sequences. There’s more than one way to skin a virus: TcRs with different sequences can recognize the same epitope.)

At any rate, it seems TcR diversity is, very roughly, on the same order of magnitude as naive T cell abundance; or perhaps a little less. We would expect maybe up to a thousand, maybe a few more, maybe a lot fewer, T cells per epitope. That doesn’t help us all that much with the question; we’re left with having to measure directly.

Directly measuring the frequency of naive T cells is, as you can imagine, very difficult. You’re looking for an event with a frequency of at most 1/100,000, with the positives spread out among an entire mouse (or an entire human). Several groups have tried, and have published their results to widespread raised eyebrows. Just recently, Marc Jenkins’ group has taken another run at the problem,⁴ and this time there’s more of the thoughtful nodding and less of the skeptical frowns.

The paper is almost entirely technical, so I won’t go into any details. Suffice it to say that they show fairly convincingly that they are counting what they say they are, and that they’re not missing many of them. (Mark Davis has a commentary⁵ on the paper, in which he points out some caveats and cautions — though I agree with his points, I don’t think they’re likely to throw the estimates way out of whack. For now, let’s accept the numbers but mentally add some grey fuzz to the upper side.)

Here’s what they found. They looked at three T cell epitopes. One yields a large, one a medium, and the other yields a smallish T cell response when you infect with the appropriate conditions. For the “large” epitope, they estimated their mice contained 190 naive T cells specific for it; the “medium”, about 20; the “small”, about 16.

Sixteen T cells, swimming about among the vast pool of irrelevant T cells and distributed randomly through the body’s lymphoid tissue, are capable of generating an immune response that, in less than 6 days, will expel invading pathogens.

The next cool thing was the link to the ultimate T cell response. Over the first 6 days of an immune response, the “large” epitope response went from around 190 T cells to around 80,000; the medium, from 20 to 5000; the small, from 16 to 3000 cells. (The figure at right shows the cell counts for each T cell group, over time. Note that it’s a log Y axis.) The expansion is quite similar for all three epitopes: 400-fold, 250-fold, and 200-fold. Here I’m going to quibble with the Moon et al interpretation. They call these all “about 300″ (fair enough, I suppose) and argue that each ultimate response was proportional to the number of naive cells. While I can see that for the biggest response, I’m skeptical that 20 is actually different from 16 — though the error bars aren’t spelled out, they clearly overlap a lot — and I’m also skeptical that 400-fold is the same as 200-fold. Also, of course, this is just three epitopes. I think it’s equally likely that while the size of the naive precursor pool is one factor, you can also get different T cell responses out of the same number of precursors, for any of a variety of reasons.

(Of course, this is a part of the immunodominance equation that I’ve touched on before.)

Still, it’s an interesting suggestion, and their data certainly are suggestive. I’m sure there will be more epitopes examined by this technique over the next little while, so we’ll see how well it holds up.

Incidentally, it’s been stated (I don’t know the data well enough to judge how accurately) that after naive T cell clones⁶ leave the thymus, they divide a little bit — just ticking over, compared to the vast expansion after they meet their antigen, but enough to expand each clone up to maybe 10-fold or so. If so, the two smaller naive populations here may have originated with just a handful of T cell clones. Jenkins’ group actually looked at TcR sequences, and their findings are roughly consistent with this idea. Certainly these small pools had a very limited number of TcR clones within them, and the larger pool had a lot more T cell clones, but there wasn’t enough material to tweeze it down much finer than that.

1. ”Naive” means they haven’t encountered their cognate antigen yet. After they encounter antigen, they’ll typically divide and multiply immensely.[↩]
2. T-cell antigen receptor genes and T-cell recognition. Davis, M. M., P. J. Bjorkman. 1988. Nature 334:395. [↩]
3. T. P. Arstila et al., Science 286, 958 (1999).[↩]
4. Naive CD4(+) T Cell Frequency Varies for Different Epitopes and Predicts Repertoire Diversity and Response Magnitude. Moon JJ, Chu HH, Pepper M, McSorley SJ, Jameson SC, Kedl RM, Jenkins MK. Immunity. 2007 Aug;27(2):203-13. [↩]
5. The αβ T Cell Repertoire Comes into Focus. Davis MM. Immunity. 2007 Aug;27(2):179-80. [↩]
6. A clone being a T cell with a specific TcR sequence[↩]

“Typically, viruses that rapidly kill their host have a very short history, as they rapidly run out of places to reproduce.”

I’m quoting John Timmer from Ars Technica’s Nobel Intent, from a couple of weeks ago. I feel kind of bad about this because I’m only quoting to disagree with him, and I always like Nobel Intent and find it interesting — but this is my most recent sighting of what I think is a very widespread misunderstanding. I commented on it briefly in the thread there (“This is one of those widely-believed rules that’s not nearly as universal as people think. … “), but here’s a chance to expand a bit. ¹

The concept is intuitively satisfying: A pathogen that rapidly and inevitably kills its hosts runs out of new hosts; if the host remains alive longer, then there’s a continuing supply of new hosts; therefore rapidly-lethal pathogens hastily evolve toward reduced virulence. It seems to make all kinds of sense, and there are some famous examples that fit this theory beautifully. Myxomavirus is the type specimen. It also fits with the observation that some of the most virulent virus infections we see are recent introductions into humans — HIV, obviously; also SARS, Ebola, and so on — that may not yet have had time to evolve toward avirulence.Unfortunately, there are also lots of counterexamples to the theory, starting with rabies (a beautifully-adapted virus that is invariably lethal), and aside from myxomavirus there aren’t all that many good examples pro. The reality is probably that evolution toward reduced virulence is a special case rather than a general rule. What’s more, the lay understanding of the theory — viral evolution toward avirulence — has little if any support, and may not occur at all.Everyone² knows about myxomavirus in Australia. Myxomavirus was introduced there in the early 1950s as a biological control agent for the rabbit plague. At first, the virus killed virtually every rabbit it infected (99.8% lethality), reducing the rabbit population by 85%, to a mere 100,000,000; but after some years of adaptation, most rabbits survived infection, and the rabbit population rebounded. While on the one hand the rabbits were obviously selected for resistance to myxomavirus,³ the virus also did in fact evolve to reduced virulence. This was shown in a classic study by Fenner and Marshall in 1957.⁴ In this unusual situation, they still had samples both of the original virus, and of a non-evolved rabbit population in Europe, so they could do direct comparisons. The new strains of the virus circulating in Australia were less lethal. What’s more, those rabbits that did die, took much longer to do so, surviving for several weeks instead of 5 days or so as with the original highly lethal strain.But Fenner’s work also showed why this isn’t a general law, and showed one of the problems with extrapolating this to the extreme of avirulence — because in fact the virus did not evolve to avirulence, it evolved to moderate virulence and then stayed there, killing about 50% of the rabbits it infected. Fenner & Marshall said: “The overall trend towards moderate virulence (grade III) … can be explained by the selective advantage for mosquito transmission of strains which cause extensive and long-persisting infectious skin lesions in rabbits.”In other words:⁵

The less-virulent virus took 3 to 4 weeks to kill a rabbit instead of 6 to 10 days, so that sick rabbits could be bitten by mosquitoes and fleas for 3 to 5 times as long as a rabbit suffering from the highly virulent strain. The milder strain was therefore more successful in infecting rabbits, and it spread rapidly. Through this selection the virus evolved to a less-virulent form.

(The map at right is from a later paper by Fenner,⁶ showing a similar phenomenon in British myxomavirus. Note that most of the strains isolated here, a decade or so into enzootic myxomavirus, are Grade III, “moderate”, killing “just” 70-90% of infected rabbits, rather than the relatively avirulent grade V or the brutally lethal grade I that was the original infection.)This highlights a key for this sort of evolution to work. There needs to be a direct link between increased transmission of the disease, and reduced virulence. The issue of a new supply of hosts, which is what most people seem to think is the critical factor, seems to be relatively minor. Conversely, if there’s a link between increased transmission and increased virulence, then the balance will not favour the pathogen becoming benign. If, for example, you are a virus that spreads by causing your host’s blood to explode out of its body, or if you destroy your host’s brain and force it to run about furiously biting anything in sight, or if you are spread through insect vectors that find your host an easier target when it’s moribund — then becoming less lethal is unlikely to help you. This has been proposed in detail, and to some extent experimentally tested, most prominently by Paul Ewald.⁷ I don’t know enough about the evolutionary and epidemiological sides to comment intelligently,⁸ so I’ll stop here, but with this quote from Ewald that explains why this sort of theoretical work can be important:⁹

Insights into the evolution of virulence may aid efforts to control or even prevent emerging diseases. Specifically, dangerous pathogens can be distinguished from those that pose relatively little threat by identifying characteristics that favor intense exploitation of hosts by pathogens, hence causing high virulence. Studies to date have implicated several such characteristics, including transmission by vectors, attendants, water, and durable propagules.

1. Also, as is usually the source for whatever I’m blathering about here, it’s something I ran across in my reading anyway, and this is one way I help solidify things in my mind.[↩]
2. Everyone who is anyone, at least[↩]
3. A major complication in interpreting this sort of phenomenon — if you don’t have an original population of the hosts, how can you tell if it was the pathogen or the host that evolved?[↩]
4. A comparison of the virulence for European rabbits (Oryctolagus cuniculus) of strains of myxoma virus recovered in the field in Australia, Europe and America. Fenner F, Marshall ID. J Hyg (Lond). 1957 Jun;55(2):149-91. [↩]
5. From a commentary on rabbit calicivirus at the Australian Academy of Sciences’ Nova page[↩]
6. Evolutionary Changes In Myxoma Virus In Britain. An Examination Of 222 Naturally Occurring Strains Obtained From 80 Counties During The Period October-November 1962. Fenner F, Chapple PJ. J Hyg (Lond). 1965 Jun;63:175-85. [↩]
7. For example, Pathogen survival in the external environment and the evolution of virulence. Walther BA, Ewald PW.. Biol Rev Camb Philos Soc. 2004 Nov;79(4):849-69.) [↩]
8. Not that that usually stops me[↩]
9. The evolution of virulence and emerging diseases. Ewald PW. J Urban Health. 1998 Sep;75(3):480-91.[↩]

When I was talking about Microsoft’s epitope prediction software, and when I discussed Kotturi’s update on LCMV epitopes, I made the point that predicting MHC class I epitopes is hard. How come it’s so hard?

First let’s define the question. MHC class I, the target ligand for cytotoxic T lymphocyte recognition, binds peptides of about 9 amino acids. These peptides are generated during proteolysis within the cytosol of the target cells¹. CTL recognize those peptides that are derived from abnormal proteins (viral or tumour, for example), while ignoring those that come from normal cellular proteins (”self”). An average virus might encode, let’s say, 10,000 amino acids, ² so there’s 10,000 or so overlapping 9mers. Out of that potential ocean of peptides, there might be 10 or 20 that CTL see at all, and of those couple dozen only two or three of those are going to be good (”immunodominant”) epitopes.

So the question is: Given the sequence of amino acids encoded by a virus, can we point to the particular 9mers that CTL will react to?

To get an accurate answer, you’d need to do exhaustive scanning of all possible viral epitopes. This hasn’t been done much, but Kotturi et al³ did it and compared their findings to epitope prediction. Twenty-five of 160 predicted epitopes were real (16%) and their predictions missed three of 28 altogether (11%). ⁴

The two granddaddies of epitope prediction are BIMAS and SYFPEITHI. Kotturi used, I am pretty sure, either ARB MATRIX⁵ or something very close to it. (The figure at the top here is from Peters et al., Figure 2A: ARB Predictions for HLA-A*0201.) A more recent paper⁶ claims that pooling together multiple predictive methods gives higher accuracy than individual methods alone, but this isn’t available online:

The authors have elected not to make the HBM available online, for two reasons: first, frequent server outages and other problems with individual web-based tools often prevent acquisition of all the requisite scores. Automatic operation is therefore not possible. Second, the querying of all the web-based tools can take a long time, making the tool inconvenient for real-time web-based access. Interested researchers may, however, contact the authors regarding obtaining the scripts implementing the HBM.

There’s also the Microsoft tool I mentioned previously, as well as a bunch of other tools — the Trost and Peters papers both compare many of them.

I haven’t tested these myself, even to the extent of comparing predictions to database results (a crude measure). ⁷ So as far as I know (with the caveat that I haven’t followed this with rabid attention) the 16% positive/11% negative that Kotturi et al got is just about as good as anyone has done (and the ranking of tools in Trost et al shows ARB MATRIX as used by Kotturi et al. is only slightly worse than the pooled prediction tool they describe, so I wouldn’t expect much better results than that from other technologies). But still, some 15 years after MHC class I motifs were described — with the pathway at least reasonably well understood — 16% and 11% isn’t all that great. Why can’t we just point to the epitopes?

Here’s the components of the pathway that need to be taken into account to successfully predict a CTL epitope:

1. Protein expression. Is there enough of the precursor protein available to yield enough epitope?
2. Proteasome cleavage. The proteasome has to cut precisely at the carboxy terminus of the epitope, though there’s a little room for error at the amino terminus. Also, the proteasome must not cleave in the middle of the potential epitope.
3. Peptidase destruction. The epitope has to survive destruction by a bunch of very active peptidases in the cytosol.
4. Transport into the ER. The TAP peptide transporter that carries peptides across the ER membrane has clear sequence preferences.
5. Trimming and destruction by ER peptidases. If the TAP-transported peptide is too long, can it be converted into the right form? If the mature epitope is there, will it be destroyed?
6. Transport out of the ER. There’s a system that pumps peptides out out of the ER, but little if anything is known about it. Perhaps it’s just diffusion out of the Sec61 channel, or maybe it’s ERAD-related, or who knows what else..
7. Binding to the MHC complex in the ER.
8. Stimulating CTL. There’s a whole complicated set of interactions in that, too, but I’ll summarize it as a single step.
9. Mystery factors that we don’t understand.

Of those 9 steps, I’d say that only one (TAP transport) is reasonably well defined as far as sequence requirements. Peptide binding to MHC class I is the next-best understood, though it’s not as simple as some people think. Protein expression level should be relatively easy, but it’s still not clear whether we need to look at total expression or levels of defective ribosomal products, or what. Predicting cleavage by the proteasome has been the subject of a lot of work, but it’s turned out to be a really difficult task; even the best algorithms are not, I think, very accurate. And I think there’s very little clue about most of the other factors.

I’ll talk more about each of the steps in other posts.

1. I’ve said and typed that phrase so often that I’m pretty much on autopilot with it[↩]
2. Divide by ten, multiple by ten, doesn’t much change the conclusion.[↩]
3. Kotturi, M. F., Peters, B., Buendia-Laysa, F. J., Sidney, J., Oseroff, C., Botten, J., et al. (2007). The CD8+ T-cell response to lymphocytic choriomeningitis virus involves the L antigen: uncovering new tricks for an old virus. J Virol, 81(10), 4928-4940. [↩]
4. The quality of the predictions were not good, either, in that many of the strongly predicted epitopes only stimulated a very few CTL. As well, I’m being a little generous is granting them just 160 predictions; that’s the number they came up with post hoc as what they would have needed — in fact they tested 400 predicted epitopes. [↩]
5. Peters, B., Bui, H. H., Frankild, S., Nielson, M., Lundegaard, C., Kostem, E., et al. (2006). A community resource benchmarking predictions of peptide binding to MHC-I molecules. PLoS Comput Biol, 2(6), e65.[↩]
6. Trost, B., Bickis, M., & Kusalik, A. (2007). Strength in numbers: achieving greater accuracy in MHC-I binding prediction by combining the results from multiple prediction tools. Immunome Res, 3, 5.[↩]
7. I’ll run some trials when I have time.[↩]

I swear this was coincidence. Last Thursday, after talking about CTL-based selection on HIV sequences, I blogged the musical question, “What other selection pressures on HIV, within a single individual, have been shown?”. One day later,¹ a paper in Science offers a partial answer.

The paper is:
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.

There’s a huge amount of individual variation in the outcome of HIV infection. Some people progress rapidly to AIDS, others more slowly, and a very few are “long-term non-progressors” (LTNP) who seem to control the virus for long periods, even without treatment. Over the years there’s been a lot of attention focused on LTNP, and one of the factors that’s been linked to slow progression is a particular MHC class I allele, HLA-B*5701.Different HLA alleles are able to bind to different viral peptide epitopes and act as targets for CTL. The presumption about the role of HLA-B*5701 has been that the particular peptides it binds to, in the virus, include very critical regions of the virus. That means (the reasoning has gone) that as HIV mutates to avoid the CTL that are targeting those particular peptides, the virus gives up a lot of fitness — it’s had to trade, say, replication ability, for the ability to avoid CTL. There were a couple puzzles with that explanation — what’s so special about the regions these particular CTLs targeted? And how come the virus seems to be somewhat inhibited quite early on in infection, even before CTL were thought to be important? — but it was a decent explanation. And it may even be right, but Fellay et al throw another possible explanation into the mix.

What these guys are did is look at more than the extremes of the resistance spectrum (which let them use far more patients than in LTNP studies — some 30,000 patients), and compare variations in HIV susceptibility to genomic markers (looking at 535,101 variations in the genome):

One striking and largely unexplained difference is the level of circulating virus in the plasma during the nonsymptomatic phase preceding the progression to AIDS. This is known as the viral set point and can vary among individuals by as much as 4 to 5 logs. We aimed to identify human genetic differences that influence this variation.

They found three links to viral set point.² “Together, the three polymorphisms explain 14.1% of the variation in HIV-1 set point.”

One of the links was to an RNA polymerase, and I’m not going to talk about that one because I don’t know enough about the potential biological relevance.³

The other two links are both MHC class I-associated. (The complicated figure at the top is from Fellay at al, showing part of the MHC class I region, the genomic polymorphisms they were analyzing, and the links. To the right is a much simpler diagram of the same region,⁴ with the two genetically-linked genes highlighted in red and HLA-B in green.One link, a relatively simple story, was with HLA-C. That’s an MHC class I molecule, and it’s recognized by CTL. The genetic polymorphism they found turns out⁵ to influence the amount of HLA-C expressed:

The protective allele leads to a lower VL⁶ and is associated with higher expression of the HLA-C gene. This strong and independent association with HLA-C expression levels suggests that genetic control of expression levels of a classical HLA gene influences viral control.

Simple and straightforward, but there’s a potential and very interesting complication. The HLA-C genes are indeed recognized by CTL, but they also interact with natural killer (NK) cells. Could this be a sign that NK cells are important in controlling HIV? Certainly that wouldn’t be surprising, and there are some relatively recent studies that link NK cells to HIV resistance,⁷ but as far as I know this HLA-C thing hasn’t been previously connected. It may have nothing to do with NK cells, but hey, it’s worth looking at.

The other link in the MHC class I region is more complicated. The actual genetic polymorphism they found was in a gene called “HCP5″, but because HCP5 variants are linked to HLA-B variants, it’s possible that the actual resistance comes from HLA-B*5701, and the HCP5 variation piggybacks along with that:

Given the strong functional data supporting a role for HLA-B*5701 in restricting HIV-1, our first hypothesis is that the association observed here is due to the effect of HLA-B*5701, reflected in its tagging a SNP within HCP5. ⁸

However, the biological plausibility of HCP5 having a role in HIV protection is high, because HCP5 is a HERV!OK, when I read that part I was all, like, “Whoa! No way, man!” but maybe not everyone reacts that way. “HERVs” are human endogenous retroviruses, fossilized retroviral sequences that, long ago, were viruses that inserted into the genome and then stayed there. There are huge numbers of these HERVs in the human genome, and it’s generally accepted that they have little if any function.⁹ In mice, though, (which are also riddled with endogenous retroviruses) resistance to various retroviruses is linked with some of their endogenous retroviral sequences.¹⁰ Could HCP5 do something similar in humans? The authors clearly like this idea, pointing not only to aspects of HCP5 that make the idea plausible (it has polymorphisms in the right place, contains proteins, and is expressed in the right cells), but also to the weaknesses of the HLA-B*5701 connection:

In fact, as a human endogenous retroviral element (HERV) with sequence homology to retroviral pol genes and confirmed expression in lymphocytes, HCP5 is itself a good candidate to interact with HIV-1, possibly through an antisense mechanism. Moreover, HCP5 is predicted to encode two proteins, and the associated polymorphism results in an amino acid substitution in one of these proteins.A model in which HCP5 and HLA-B*5701 have a combined haplotypic effect on the HIV-1 set point is consistent with the observation that suppression of viremia can be maintained in B*5701 patients with undetectable VL, even after HIV-1 undergoes mutations that allow escape from cytotoxic T lymphocyte (CTL)-mediated restriction.

So the bottom line is that, while this large-scale study hasn’t offered direct answers to what makes HIV progress slowly or rapidly, it has raised a couple of really intriguing questions. I’m looking forward to seeing more studies on HIV resistance and HCP5 and, perhaps, NK cells, in the next few years.

1. That would be Friday, if you’re playing along at home[↩]
2. They also found a link or two to another aspect of HIV resistance — slow progression — but I won’t talk about those now.[↩]
3. Though it seems pretty biologically plausible that an RNA virus might have its life cycle affected by an RNA polymerase.[↩]
4. Drawn with XPlasMap v.0.96 from the GenBank sequence[↩]
5. They actually tested this; it’s great to see these hypotheses actually being tested at the same time they’re generated by these big-science type studies[↩]
6. VL: “Viral load”, like viral set point[↩]
7. Reviews include Barbour, J. D., Sriram, U., Caillier, S. J., Levy, J. A., Hecht, F. M., & Oksenberg, J. R. (2007). Synergy or independence? Deciphering the interaction of HLA Class I and NK cell KIR alleles in early HIV-1 disease progression. PLoS Pathog, 3(4), e43. and Fauci, A. S., Mavilio, D., & Kottilil, S. (2005). NK cells in HIV infection: paradigm for protection or targets for ambush. Nat Rev Immunol, 5(11), 835-843. [↩]
8. Looking at the map, I wonder if there might (also? Or instead?) be a link with the MICA and/or MICB genes that flank HCP5. MICA and MICB are NK receptors, so the same argument as with the HLA-C link applies. The authors don’t mention this possibility, and I don’t know if MICA and MICB are as tightly linked to HCP5 variation as is HLA-B. [↩]
9. There are lots of papers that have looked for functions, but few if any are convincing. For a review, see Griffiths, D. J. (2001). Endogenous retroviruses in the human genome sequence. Genome Biol, 2(6), REVIEWS1017. [↩]
10. For example, Best, S., Le Tissier, P., Towers, G., & Stoye, J. P. (1996). Positional cloning of the mouse retrovirus restriction gene Fv1. Nature, 382(6594), 826-829. and Ikeda, H. & Sugimura, H. (1989). Fv-4 resistance gene: a truncated endogenous murine leukemia virus with ecotropic interference properties. J Virol, 63(12), 5405-5412.[↩]

They’ve done it again! The online table of contents for today’s issue of Science has an article on a ubiquitin ligase ¹ and the editors have added this helpful blurb:

In developing worms, the pruning of excess synapses requires proteosome-mediated protein degradation and is selectively prevented by a neural adhesion molecule.

No, no, no, no, no! There’s no such word as “proteosome“! It’s “proteasome“. It’s a horribly common mistake (PubMed has 7280 cites for the misspelled version, and 11813 for the correct spelling — a ratio that’s actually nearly ten times worse than the generic web’s)  ² but it’s a mistake nonetheless.

This is far from the first time Science has done this³ and I wrote to them the last time they did it, which was in June for the teaser “Selective Proteosomes”. I wrote to them again today, but I’m not optimistic; the misspelled version from June is still there.

Some editor at Science needs a sharp smack upside the head.

1. Spatial Regulation of an E3 Ubiquitin Ligase Directs Selective Synapse Elimination. Mei Ding, Dan Chao, George Wang, and Kang Shen. Science 17 August 2007: 947-951. [↩]
2. Google claims “about 240,000″ hits for -o- and 2,990,000 for -a-, a 12:1 ratio compared to the presumably more technical literature’s 1.6:1 ratio.[↩]
3. 82 hits in Pubmed for “Proteosome AND science[Journal]“![↩]

HIV is a genetically unstable virus, and exists as a “quasispecies”,¹ a cloud of variations surrounding a platonic ideal virus. Over time, selection pushes the cloud in various directions. What’s the main push behind that movement?

Because I’m interested in T cell immunity I tend to think of HIV mutation as being driven by, well, T cell immunity. This is the CTL escape² I’ve mentioned before, and the paper that most dramatically reinforced that viewpoint for me was:
Constraints on HIV-1 evolution and immunodominance revealed in monozygotic adult twins infected with the same virus.
Draenert R, Allen TM, Liu Y, Wrin T, Chappey C, Verrill CL, Sirera G, Eldridge RL, Lahaie MP, Ruiz L, Clotet B, Petropoulos CJ, Walker BD, Martinez-Picado J.
J Exp Med. 2006 Mar 20;203(3):529-39.

This study found a pair of identical twins, infected with the same HIV strain at the same time, and tracked the appearance of new variants of HIV that popped up over time, correlating with immune responses. Remarkably, the mutations of HIV that appeared to be CTL escape variants were almost identical:

Of four responses that declined in both twins, three demonstrated mutations at the same residue. In addition, the evolving antibody responses cross-neutralized the other twin’s virus, with similar changes in the pattern of evolution in the envelope gene. These results reveal considerable concordance of adaptive cellular and humoral immune responses and HIV evolution in the same genetic environment, suggesting constraints on mutational pathways to HIV immune escape.

The conclusion I drew from this paper is that, in a particular genetic environment, the immune system shepherds HIV along a particular trail.

Now, this paper only tracked the twins for 3 years. Another paper³ had tracked a similar pair of twins over a much longer time, 17 years, and their findings were rather different:

Seventeen years after infection, their CTL targeting of HIV-1 was remarkably similar. In contrast, their overall TCR profiles were highly dissimilar, and a dominant epitope was recognized by distinctly different TCR in each twin. Furthermore, their viral epitopes had diverged, and there was ongoing viral phylogenetic divergence between the twins between 12 and 17 years after infection. These results indicate that while CTL targeting is predominately genetically determined, stochastic influences render the interaction of HIV-1 and host immunity, and therefore viral escape and CTL efficacy, unpredictable.

(The figure at right is the concluding figure from Yang et al., showing quite dramatically how each twin’s HIV had moved in different directions: “Phylogenetic relationships between pol (A), env (B), and nef (C) sequences from 1995 and 2000 are shown. Open and closed circles represent twin 1-05 sequences from 1995 and 2000, respectively; open and closed triangles represent twin 1-06 sequences from 1995 and 2000, respectively.“)

Still, even the Yang et al. paper don’t change my overall impression that HIV mutation is mainly CTL-driven; even if the viruses trotted down different paths, they were both (probably) being chivvied along those paths by CTL pressure.

What made me think about this today is an interesting variant on CTL escape variants, described in the latest issue of Journal of Virology:
A Rapid Progressor-Specific Variant Clone of Simian Immunodeficiency Virus Replicates Efficiently In Vivo Only in the Absence of Immune Reponses.
Takeo Kuwata, Russell Byrum, Sonya Whitted, Robert Goeken, Alicia Buckler-White, Ronald Plishka, Ranjini Iyengar, and Vanessa M. Hirsch.
Journal of Virology, Sept. 2007, p. 8891-8904 Vol. 81, No. 17

These guys were looking at monkeys infected with SIV, a subset of which were “rapid progressors” (RP). These monkeys show an early immune response, but lose their anti-HIV immunity very quickly — within 4 weeks. Worse than that, they also lose their ability to mount any new immune response, even to related antigens like tetanus toxoid. It turns out that these RP monkeys also contain a unique variant of SIV, with specific mutations in the env gene. Is this mutant virus responsible for the rapid progression of disease?

In fact, when they tried infecting monkeys with the new variant SIV, the recipients did not progress rapidly. If anything, the new variant virus was actually worse than wild-type virus at causing disease. And what’s more, when they took sampled the virus circulating in the newly-infected virus, what they found was that as immune responses to the virus developed, the RP variant disappeared and was replaced by … wild-type SIV, the parent of the RP variant. The only way the SIV could survive in their new hosts, in the face of an immune response, was to mutate back to the original wild-type sequence. It looks as if the causality was backwards; the RP variant didn’t cause the rapid progression, rather the rapid progression permitted these new viruses (that are very sensitive to immune responses) to be able to replicate.

These studies suggest that the SIV variants commonly selected in RP macaques are not the direct cause of rapid disease de novo in naive macaques. The evolution of RP-specific variants appears to be the result of replication in a severely immunocompromised host.

So perhaps this is an exception that proves the rule, and the major force behind HIV mutation and selection really is immune pressure: The virus doesn’t develop other variants until the immune system is completely screwed.

There’s at least one obvious exception to this: The change in receptor usage that HIV shows after infection. According to my primitive perception of this, the HIV types that are most readily spread between individuals (those that use the CCR5 receptor), are not the same type as most efficiently spread within an individual (those which can also use CXCR4); so the former are more likely to infect, but then mutants with the latter arise after infection. This is selection that’s not CTL-based. What other selection pressures on HIV, within a single individual, have been shown? I don’t know, I’m asking.

1. The figure at the top of this post is “Quasispecies complexity of kidney and PBMC-derived from 2 patients with HIVAN“ from: Replication and compartmentalization of HIV-1 in kidney epithelium of patients with HIV-associated nephropathy. Daniele Marras, Leslie A. Bruggeman, Feng Gao, Nozomu Tanji, Mahesh M. Mansukhani, Andrea Cara, Michael D. Ross, G Luca Gusella, Gary Benson, Vivette D. D’Agati, Beatrice H. Hahn, Mary E. Klotman & Paul E. Klotman. Nature Medicine 8, 522 - 526 (2002) [↩]
2. That is, as the host’s T cells target specific regions of the virus, any new versions of the virus that mutate the targets, are more likely to thrive than the wild-type sequence. Antibodies also probably select HIV mutants, through the same mechanism — i.e. escape from neutralizing antibodies. I don’t know the relative importance of CTL and antibody selection, but I suspect that CTL are more important because antibodies mainly target a limited region of a limited number of proteins, whereas CTL attack the entire HIV genome.[↩]
3. Genetic and Stochastic Influences on the Interaction of Human Immunodeficiency Virus Type 1 and Cytotoxic T Lymphocytes in Identical Twins. Otto O. Yang, Joseph Church, Christina M. R. Kitchen, Ryan Kilpatrick, Ayub Ali, Yongzhi Geng, M. Scott Killian, Rachel Lubong Sabado, Hwee Ng, Jeffrey Suen, Yvonne Bryson, Beth D. Jamieson, and Paul Krogstad. Journal of Virology, December 2005, p. 15368-15375, Vol. 79, No. 24 [↩]

Having beaten the theme of viral immune evasion into the ground, I’ll just change my aim a tiny bit and keep right on clubbing immune evasion. Cancers, like viruses, are (probably) controlled by the immune system,¹ and, also like viruses, cancers evade cytotoxic T lymphocytes (CTL) in various ways.

Cancers are not viruses, however,² and they don’t evade CTL in the same ways. One difference is that cancers evolve from square one each time. When you’re infected by a virus, that virus is the end-product of an unbroken chain of evolution that goes back millions or billions of years;³ it’s had time to evolve its own specialized molecules, which may or may not have been based on the host genome at some time but is now generally a standalone, distinct gene. Cancers don’t have that history. Each individual cancer arose independently within you,⁴ and it only has your lifespan⁵ in which to experiment with immune evasion.

Accordingly, what cancers tend to do is evade immunity by destruction rather than creation. Viruses are creative in their evasion techniques, crafting solid, workmanlike wooden shoes to cast into the gears of the immune system, while cancers blindly bludgeon their genomes until they break something the immune system needs. Most cancers (especially solid tumours), for example, contain genomic deletions that eliminate MHC class I alleles,⁶ and MHC class I alterations are associated with poor clinical outcomes.⁷

A part of this selective destruction can lead to cancer immune escape. I’ve talked about the way chronic viruses (like hepatitis C and HIV) alter their protein sequences in such a way as to mutate MHC class I epitopes, so that immunodominant CTL no longer recognize infected cells. Cancers are much like chronic infections and they too mutate immunodominant epitopes, so that CTL can no longer recognize them.⁸ This is, obviously, a real concern in cancer immunotherapy trials.

However, there are clearly lots of other ways that tumours avoid immune clearance that don’t involve mutation of their MHC class I system. In particular, CTL in the presence of tumours are often fairly ineffective; their targets may be present and the cells may be nominally susceptible, but the cells are simply not capable of dealing with the tumour. Again there are many reasons for this phenomenon. A new one was recently published in Nature Medicine:
Altered recognition of antigen is a mechanism of CD8(+) T cell tolerance in cancer.
Nagaraj, S., Gupta, K., Pisarev, V., Kinarsky, L., Sherman, S., Kang, L., et al. (2007).
Nat Med, 13(7), 828-835.

This shows that tumours can also avoid CTL recognition by targeting the other side of the equation, the T cell receptor. (The TcR on a CTL recognizes MHC class I, and yes, editors always get exercised about using too much jargon and contractions.) The authors started with the previous finding that one cause of T cell tolerance of tumours is a suppressor cell, that infiltrates into the tumours and turns off — tolerizes — CTL that could otherwise attack the tumour. They show that, via release of peroxynitrite, these suppressor cells can physically alter the T cell receptor of the CTL, nitrating some of the tyrosines on the TcR (and also on the co-receptor, CD8); as a result, the TcR couldn’t bind to its target MHC class I. (The figure to the right shows the predicted location of the nitrated tyrosines on the TcR. At the top of this post: a stain for nitrotyrosines lights up T cells in a lymph node from a cancer patient — lymph nodes from normal individuals showed few if any nitrotyrosine-containing T cells. Both figures are from the supplementary info for the paper.)

To test their hypothesis, Nagaraj et al. treated mice with uric acid (which, they say, specifically neutralizes peroxynitrite) at the same time as they transferred specific CTL into tumor-bearing mice. Transferring the CTL alone didn’t do much: they were tolerized, and the tumours weren’t affected. Treating with uric acid alone also did nothing.⁹ The combination of uric acid — to reduce peroxynitrite-triggered nitrotyrosine formation — with adoptive transfer of specific T cells, though, led to a significant reduction of tumor growth (but not to complete of the tumour). (Figure on the left shows tumour size after the various treatments — the “Vaccine + UA [Uric acid]” trace is the open circles. The tumours in those mice are less than 1/3 the size of the mice that just got CTL.)

Uric acid treatment is unlikely to be very practical for cancer patients, but if this work holds up, it may offer a way to enhance tumour immunogenicity, perhaps with some more specific targeting of the suppressor cells or some more specific way of reversing the tyrosine nitration. In any case, it suggests new ways that tumours (and perhaps pathogens) could block T cell recognition.

1. The whole question of whether cancers really are controlled by the immune system is still somewhat controversial, but the weight of the evidence is that they are. The back-and-forth on this issue is probably worth a blogpost some time, if only because it will help me get the story straight in my own mind.[↩]
2. Even the cancers that are caused by viruses aren’t viruses[↩]
3. Depending on which origin of viruses you’re considering[↩]
4. Aside from a few weird and whacky things like transmissible canine venereal tumours and the Tasmanian Devil transmissible tumours.[↩]
5. What’s left of it … [↩]
6. A lot of references cite relatively low numbers for this phenomenon, like “15% - 60% of solid tumours”. Until quite recently, screening of tumors for loss of MHC class I was relatively perfunctory and coarse, and probably missed many of the smaller and more focused deletions, so I think the higher end of the ranges is probably much more accurate, or may even still be an underestimate. Work by Soldano Ferrone, where he has looked very carefully at a relatively small set of tumours, turned up MHC class I/antigen presentation defects in virtually every one, and many of those would have been missed by standard approaches. A review is: Chang, C. C. & Ferrone, S. (2007). Immune selective pressure and HLA class I antigen defects in malignant lesions. Cancer Immunol Immunother, 56(2), 227-236.[↩]
7. Bangia, N. & Ferrone, S. (2006). Antigen presentation machinery (APM) modulation and soluble HLA molecules in the tumor microenvironment: do they provide tumor cells with escape mechanisms from recognition by cytotoxic T lymphocytes? Immunol Invest, 35(3-4), 485-503. [↩]
8. Singh, R. & Paterson, Y. (2007). Immunoediting sculpts tumor epitopes during immunotherapy. Cancer Res, 67(5), 1887-1892.[↩]
9. Uric acid is a potent immune stimulant, but only in the crystal form. The authors treated the mice with soluble uric acid and argue this avoids this potential confounder. However, as I recall, uric acid in the body is very close to the crystallization point normally, and adding in more, even soluble, uric acid might kick it over the edge into crystallizing; so this is still a potential artifact in this study. But that might be less of a concern with intraperitoneal injections.[↩]

My last couple mentions of viral immune evasion of T cells may have left the impression that immune evasion in general has a minor contribution to viral pathogenicity. Far from it. There are lots of examples — some very dramatic (I know of a couple really spectacular experiments that aren’t yet published) — of viral immune evasion genes that are important, or essential, for virulence. One example is influenza virus’s evasion of interferon.

Everyone’s heard of influenza. It’s a relatively small RNA virus with many different strains, which differ in their virulence, host range, antigenicity, and so forth. As with many viruses, the interferon response (part of the innate immune system) can be a very powerful inhibitor of virus infection, and so it’s not surprising that influenza has evolved a way around interferon. It does this by means of a non-structural protein, NS1, which prevents interferon induction in infected cells. How important is this protein to viral infectivity, in vivo?

The experiment has been done in a number of species. (One nice thing about influenza, from the researcher’s viewpoint, is that it does infect many different species, including some — mice, chickens, pigs — which are relatively easy to study.) For example, Donelan et al¹ made a mutant influenza virus that lacks the NS1 protein, and looked at its virulence in mice. Wild-type influenza infection made the mice very sick: they lost weight (Figure to the right here; the diamond traces are the wild-type virus) and, around day 5, died of the infection, but mice infected with the mutant virus didn’t lose any weight (squares). As well, the virus replicated thousands of times worse in the mice. Although in this case mice aren’t the natural host of the virus, there are similar findings in, for example, chickens using avian influenza viruses² so it’s pretty clear that this is an authentic virulence factor.

Speaking of avian influenza (nice segue, eh?), we all know about the concerns about an avian influenza pandemic. One of the reasons for the fear of this virus is previous experience with avian influenza in humans. In 1997, H5N1 influenza viruses jumped from chickens to humans and proved to be highly virulent — this was the Hong Kong outbreak of avian influenza, in which something like 6 of 18 known-infected people died. ³ It turns out that the NS1 in this H5N1 virus somehow is even more potent than normal, wild-type, virus in its ability to block the effects of interferon.⁴ Taking the H5N1 NS1 gene, and plugging it into a generic influenza virus (PR8, the most common lab strain) turned the normally fairly-innocuous PR8 virus into a vicious bastard, causing nearly a 50% weight loss in infected pigs (the circles in the figure to the left; the squares and triangles show the weights of pigs infected either with wild-type PR8, or with PR8 containing a mutated version of the H5N1 NS1). (The adorable pig pictures at the top of this post are from a later paper by the same authors.⁵ )

So immune evasion genes can be really potent virulence factors, which really contrasts to the apparently-minor effects of the CTL immune evasion genes.

1. A Recombinant Influenza A Virus Expressing an RNA-Binding-Defective NS1 Protein Induces High Levels of Beta Interferon and Is Attenuated in Mice. Nicola R. Donelan, Christopher F. Basler, and Adolfo Garcia-Sastre.