The role of parvovirus B19 in the pathogenesis of autoimmunity and autoimmune disease (J Kerr)

[Jonathan Edwards]

No, I meant it´s not possible to search a database for them, so it´s hard to get an accurate idea of their relative frequency in nature.

How do we know that the clonal deletion system gets rid of low-affinity antibodies (which according to the article I posted a link for above, can still sometimes actívate complement)? Do we have any evidence for how it does this? I.e. is there any evidence that antibodies that bind to self antigens are deleted because they actívate complement? When I looked it up, there seem to be several ways in which tolerance to self is produced - why would there need to be many different ways if the clonal deletion worked so well?

The system must get rid of complement fixing autoantibodies, otherwise you would have complement fixation and disease - called lupus in fact. We know a lot about how this is done but it is complex. There are a t least four selection checkpoints. Review articles tend to cover one or two - a full review would be a chapter of at least thirty pages and you would need to know quite a lot of background to follow that.

Complement seems to be crucial to selection at at least two checkpoints. I think it was Carroll who showed around 1990 that complement is involved in deleting autoreactive B cells in bone marrow. It is also involved in amplifying foreign -reacting B cells in follicle centres so has a paradoxical role. In lupus complement is defective - either genetically or through over-consumption. Lupus is the one disease where you get lots of different auto-antibodies formed and it was suggested that this reflects failure of complement in marrow.

So tolerance depends on weeding out self-reactive cels at several stages - bone marrow, follicle centre, T cell zone etc. You need several checkpoints because antibodies mutate once B cells have got into lymph nodes. The system provides itself with a moving target and presumably has developed a quadruple checking system to cope with this. The system also has the Achilles heel that it depends on positive feedback loops for antibody amplification, so any mistake is potentially amplified once it has got past checkpoints. That is probably one reason why there are two recognition systems to start with - T and B. It is all hideously complicated I am afraid and we do not know half of it.

But I come back to the issue that I think people are thinking that cross reactivity is a property of the antigens. It is a property of unique antibody business end conformations.

 

[msf]

Very interesting, thanks. No I understand it´s about conformations. In a way though, it´s a property of both, but I get your point - it´s the antibody conformation that decides what is cross reactive and what isn´t.

 

[msf]

With regards to the way self-tolerance works, doesn´t the fact that some autoantibodies are quite common in the healthy controls suggest that it doesn´t actually work that well? Or are these autoantibodies not fixing complement?

 

[Jonathan Edwards]

Very interesting, thanks. No I understand it´s about conformations. In a way though, it´s a property of both, but I get your point - it´s the antibody conformation that decides what is cross reactive and what isn´t.

Bingo!

With regards to the way self-tolerance works, doesn´t the fact that some autoantibodies are quite common in the healthy controls suggest that it doesn´t actually work that well? Or are these autoantibodies not fixing complement?

This is the grounding for the approach Jo Cambridge and I took in 1999 that led us to discover the B cell depletion is effective in a wide range of autoimmune disorders. Certain self-antigens appear to sneak through tolerance and generate antibody responses - thyroid proteins, IgGFc (rheumatoid factor antigen), and we may now find citrullinated proteins come under that too (but they are not really auto because the citrullination does not distinguish self from non-self). And several more self antigens sneak through quite often and are associated with disease. But 95% of self proteins never seem to produce an autoimmune response. The reason for this, we and others have suggested, is that the proteins for which tolerance is dodgy are proteins that are somehow involved themselves in immune signalling - IgGFc is the most obvious one. That means that they can confuse the immune system by acting both as antigen and as regulatory signal. The analogy is when a computer confuses a data string with a command string - like when you think you did capital X but the computer does 'delete'.

The theory is that for any very complex information processing system like the immune system there is always a trade off between fine grained function and introducing 'programme bugs'. The more exhaustive you want discrimination the more you are likely to introduce mechanisms that bring new ways of screwing up. I think you will agree that that is how Microsoft programs go - the fancier they are the more they crash. So we can expect the immune system to have crash prone 'blind spots'. That is what I think we call autoimmunity.

 

[msf]

When you say 95% of self proteins never produce an autoimmune response, do you mean in healthy people or in anyone? And do the autoantibodies found in the blood of healthy controls actívate complement?

It´s definitely a fascinating area - I can see why you went into it.

 

[Jonathan Edwards]

When you say 95% of self proteins never produce an autoinmune response, do you mean in healthy people or in anyone? And do the autoantibodies found in the blood of healthy controls actívate complement?

It´s definitely a fascinating area - I can see why you went into it.

As far as I know proteins like albumin and haemoglobin never seem to have autoantibodies - or vanishingly rarely.

The issue of activating complement is tricky because you have to consider it in the physiological microenvironment, which is very complex too. You can fiddle around with conditions in a dish until almost anything happens.

 

[M Paine]

I think there is a continuing misunderstanding in this conversation that cross reactivity is a property of a microbial protein and a host protein. It is not. It is a property of an antibody species that happens to bind to both - the two proteins bound to do not actually need to be similar in shape, they just need to have a complementary charge pattern to the antibody

I wasn't suggesting anything contradictory to what you have mentioned here, or implying that similar shape determines binding affinity. That said, shape is of coarse very important nonetheless, as without knowing the shape of a protein, it would be impossible to infer potential epitope sites. Knowing which atoms are grouped and burried, and which exposed atoms constitute possible epitope locations etc.

In this way, before trying to identify possible epitopes in silico, the structure needs to be solved, (not so in the wet bench environment), and that could be done using a protein structure algorithm. As I pointed out, this is a functional limitation involved in this due to things like quaternary structure, regulation of viral gene expression by codon usage, post translational protease activity or protein folding involving difficult to predict marshaling events, and other events which happen in vivo that are not easily predicted just by looking at a sequence.

The technique wherein you used 3D printed models of antibodies to try and identify binding sites, twisting and turning your proteins (protein containing potential epitope and antibody binding domain protein) to see if the lock fits the antibody binding domain's key in some way with electrostatic binding affinity, is the same approach that can be used in silico using epitope prediction algorithms in a CAD environment.

Similar techniques are used in the search for potential drug candidates, or in the search for peptide-based vaccines, where billions of small peptides derived from target proteins (eptiopes) are screened to find those likely to provoke an immune response to a particular cell type.

I very much duoubt even now that anyone could predict dissociation constant.

Thankfully traditional techniques, such as X-ray crystallography and NMR spectroscopy provide a clear picture of how proteins interact taking a reverse engineering approach. Salt bridges, hydrogen bonds, water interactions etc have been analyzed in atomic resolution, and the information is available to use when designing algorithms.

Clearly it's an extremely complex undertaking, but at least in this way algorithms can be tested against known protein-protein complexes to test accuracy. I'm sure that there's a long way yet to go to get to a 100% accurate prediction algorithm for dissociation constant, but I for one am impressed with what is currently possible in this area, this article documents things pretty well -On the binding affinity of macromolecular interactions: daring to ask why proteins interact

 

[daisybell]

http://www.medscape.com/viewarticle/869526_1
This looks to be a related article by J Kerr.
I find this interesting as I was well before I got Parvovirus - now I have limited scleroderma, Graves' disease and ME......

 

[Kati]

http://www.medscape.com/viewarticle/869526_1
This looks to be a related article by J Kerr.
I find this interesting as I was well before I got Parvovirus - now I have limited scleroderma, Graves' disease and ME......

Quite interesting regarding your history @daisybell.