Less is more

Jeremy Gunawardena pointed me to this terrific paper that results from a collaboration between Bruce Walker’s and Arup Chakraborty’s groups (Košmrlj A, Read EL, Qi Y, Allen TM, Altfeld M, Deeks SG, Pereyra F, Carrington M, Walker BD, Chakraborty AK 2010. Effects of thymic selection of the T-cell repertoire on HLA class I-associated control of HIV infection. Nature 465 350-4. PMID: 20445539).  On reading it, I realized that I had heard Arup present the work at the ICSB 2009 conference; for me, it was one of the highlights of the meeting.

This paper looks at the reasons why a few patients with HIV never progress to AIDS.  Despite what Peter Duesberg might think, and irresponsibly say, the answer is not (repeat, not!) that HIV doesn’t cause AIDS.  I will tell you why, if you are patient.  First, I need to give you a bit of background.

The immune system is classically divided into two “arms”, the innate immune system (which includes the monocytes Ralph Weissleder has been watching) and the adaptive immune system.  The innate immune system is all about fighting off pathogens that always look the same, and/or that always look different from you; for example, bacteria.  The adaptive immune system is meant to catch what the innate immune system can’t, including sneaky viruses like HIV that regularly change the way they look. Of course, this is a gross over-simplification and the two “arms” have lots of cross-talk and regularly work together.  But it highlights the central problem the adaptive system has: it needs to be able to respond to challenges whose nature is unknown.

The MHC, or major histocompatibility complex, is a key part of the adaptive immune system’s solution to the problem of recognizing the unknown.  It comes in two flavors, class I and class II.  Class I is found on all cells, class II only on a few.  What class I does is to go fishing in the interior of the cell for peptides, and display a sampling of what it finds on the surface.  (Class II, in contrast, looks at what is outside the cell, using an elaborate strategy of rescuing partially digested peptides from the endosome.)  A peptide binding to an MHC molecule looks a bit like a hot dog in a bun: the MHC forms the two sides of the bun, with a “cleft” in between; although some parts of the hot dog are covered by bun, other parts aren’t, and the overall shape of the bun + hot dog is unique to the particular MHC-peptide complex. When a cell is infected by a virus, class I MHC molecules will start carrying viral peptides out onto the surface, where the immune system can see them.

But just putting the peptides out there is not enough.  How is it possible to recognize a viral peptide as foreign, especially when the virus can mutate and change?  The answer is not so much that you actively respond to foreignness as that you fail to respond to self.  T cells have a system for producing highly variable receptors, very similar to the system that produces antibodies.  These receptors are expressed on the cell surface instead of primarily secreted; each T cell carries only one receptor, which is structurally predisposed to bind to MHC molecules. Developing T cells are educated early in their development to avoid responding to MHC that carries “self” peptides.  It’s a pretty severe form of education: T cells that respond too strongly to “self” MHC-peptide complexes die.  But afterwards, the T cells left (assuming all went well) are the ones that didn’t respond to what’s normal, and therefore anything they respond to must be strange.

In humans, there are three subtypes of MHC class I genes, named HLA-A, B and C.  Each has hundreds of variants in the human population.  The likelihood that you will get the same variants from both mother and father is small, so the average human cell has 6 different class I proteins on it (two types each of A, B and C).  It turns out that one specific variant of one particular MHC subclass — HLA-B57 — is over-represented in this mysterious class of patients that never develop AIDS.

Why do we need so many subtypes of MHC molecules?  The shape of the peptide-binding cleft in a class I molecule is highly variable; different MHC variants have different shapes of cleft, and therefore bind different subsets of peptides from the very large number of peptides available in the cell.  This is both good for the individual, and good for the population.  An individual generally has six different peptide-binding clefts, and therefore has six different chances to catch a peptide that will trigger an effective immune response.  Unlucky individuals may have a combination of class I MHC molecules that don’t manage to pick up any useful peptides from a particular virus, and those people will have an especially hard time getting rid of that virus.  But even if some people are susceptible to a virus, the diversity of class I MHC in the population usually means that the population as a whole can fight back.

There is one caveat, however: the more different MHCs, the more of your developing T cells will be killed.  Perhaps this is why evolution stopped at 6.  You could think of your T cell repertoire as a petri dish with lots of small filter papers soaked in antibiotic scattered across a lawn of bacteria.  Around each filter paper (each MHC) is a zone of exclusion: these are the areas where T cells are being killed because they can recognize the combination of a self peptide and this particular MHC molecule.  MHCs that bind a larger variety of peptides have a larger exclusion zone.   The area where the bacteria can grow is the area where you can have an immune response.

I’m sorry, this analogy didn’t turn out too well — the immune response would get rid of the bacteria.  But I hope you see what I mean.  The point is that every MHC creates a characteristic “hole” in your potential immune response.

What Košmrlj et al. now show is that the peptide-binding cleft of HLA-B57 is unusually selective.  There are several algorithms to predict the affinities of peptide-MHC binding events that do quite well when tested against a large database of experimental measurements.  With these algorithms you can run through all the peptides in the human proteome, and pick out the ones that will bind to your particular MHC.  It turns out that HLA-B57 binds only 70,000 out of 10 million possible peptides, whereas a “normal” MHC molecule binds 180,000.  So the “exclusion zone” around HLA-B57 is particularly small.

What this means is that most of the T cells that can bind to HLA-B57 are still in your immune repertoire — they haven’t been killed in the process of T cell “education”.  Košmrlj et al. test this quantitatively using an in silico T cell education experiment that simulates the residues on the T cell receptor that contact peptides, and vice versa, as strings.  They use experimental data to define a “statistical potential” that simulates the likely free energy of interaction between the amino acid pairs on the T cell receptor and peptide, and they count up the interactions along the paired strings.  If the summed free energy of interaction exceeds a certain threshold, they consider that T cell receptor to be eliminated.  They did this for 1 million T cell receptor sequences, all the peptides in the human genome, and a variety of MHCs.  [For aficionados: I have ignored MHC-T cell receptor interactions and positive selection here, but the authors didn’t.]

Using this simulation strategy, they find that in the case of selection on HLA-B57 fewer T cells are killed in the education process — but also, and more important, they find that T cell receptors that use a small number of contact points to bind to a peptide have a much better chance of surviving.  These few-contact receptors would be expected to be especially cross-reactive; any peptide that preserves the few important contacts needed for binding will be recognized and attacked. In contrast, MHCs that bind more self peptides have more chances to delete few-contact receptors: the T cell receptor repertoires resulting from education on these MHCs typically require many weak peptide contacts for binding, and the loss of any one of these contacts will result in loss of recognition.

The bottom line is that any HIV peptide that binds to HLA-B57 (Košmrlj et al. estimate there are about 40 of them) is going to have a much harder time finding a mutation that will evade the immune response.  And this would explain why people who are lucky enough to have HLA-B57 among their class I MHC molecules would be better at controlling HIV infection.

I find this beautifully counter-intuitive.  Who would have thought that an MHC subtype that binds fewer peptides would be better at generating an immune response to control HIV?  And yet — wouldn’t you think that having more cross-reactive T cell receptors would lead to more problems with reactions to self peptides?  Indeed, it turns out that HLA-B57 is also associated with autoimmune psoriasis and other hypersensitivity reactions.

What does this do for those of us who aren’t lucky enough to carry HLA-B57? Košmrlj et al. point out that although cross-reactive T cells are rarer in such individuals, they are not non-existent.  What we need to do, they suggest, is to identify cross-reactive T cells that bind HIV peptides in people who don’t carry HLA-B57, and develop a vaccine that’s aimed at activating them.  I am not sure whether they are advocating this strategy for a so-called “therapeutic” vaccine — which is aimed at activating the immune system in people who are already infected — or a true vaccine, intended to protect you from becoming infected in the first place.  One concern might be that activating cross-reactive T cells could, perhaps, also increase the likelihood of autoimmune diseases.  For a therapeutic vaccine you might not worry so much about this — the benefits would be likely to outweigh the potential costs — but I would think you would want to be very sure this wouldn’t happen before giving this kind of vaccine to the general, uninfected population.  The immune system has been known to give unexpected reactions before.

A lovely paper.  Not just for immunologists.  Happy reading.

Kosmrlj A, Read EL, Qi Y, Allen TM, Altfeld M, Deeks SG, Pereyra F, Carrington M, Walker BD, & Chakraborty AK (2010). Effects of thymic selection of the T-cell repertoire on HLA class I-associated control of HIV infection. Nature, 465 (7296), 350-4 PMID: 20445539