That's an interesting idea, but there is little or no evidence that an altered post-translational modification is a common target for naturally-arising autoantibodies. Autoantibodies are typically polyclonal, reacting with multiple epitopes on a macromolecule or a molecular complex. In fact, they are generally not even species specific, tending to react with epitopes conserved between quite disparate species, and these epitopes are unlikely to be only post-translationally modified domains. Autoantibodies reflect a loss in tolerance to true self proteins, nucleic acids or their complexes. There are a few examples of autoantibodies apparently arising by cross-reaction with infectious agents, but these are rare.

This is an intriguing question indeed. I think it is correct to differentiate between naturally occuring and disese-specific autoantibodies in this respect. To miy mind, disease-specific autoantibodies mainly interact with conformational epitopes. Post-translational modifications seem to play a critical role in the formation of such autoantigenic epitopes as has been shown for antitransgluatminase antibodies in celiac disease. Usually, we may expect that the central tolerance established in the thymus and bone marrow by deleting self-reactive T and B cells, respectively, is not able to cover all post-translational modifications.

One of the best examples of altered post-translational modifications lies in rheumatoid factor (RF), which can not be ignored as it accounts for around 80% of occurrences in rheumatoid arthritis patients and other autoimmune diseases too. The aberrant glycosylation induces the change in the structure of the immunoglobulin IgG, so RF is indeed the antibody against the Fc portion of IgG.

The reason why antibody gets formed against the self protein is because of its structural alteration induced by improper glycosylation, phosphorylation, nitrosylation, methylation, etc.

I would agree that post-translational modifications could create neo-epitopes for antibody selection at the B cell level, and there are some good examples of this. But robust autoantibody production requires T cell help, and most of these modifications are invisible to T cells because they are unlikely to be presented on MHC class II. If an aberrant AA side chain modification did get presented, it does run the risk of activating a T cell because central or peripheral tolerance to that modification might not have developed, as Dr. Roggenbuck pointed out. One possible example of this is protein citrullination, resulting in anti-citrullinated protein antibodies which are common in RA. Anti-transglutaminase antibodies, which Dr. R. mentioned, may be another example of this. But this is not the only way or even the probable way autoantibodies spontaneously develop because they are typically polyclonal, reacting with multiple epitopes on native proteins, nucleic acids, or their macromolecular complexes.

Dr Rubin is right indeed in stating that a sustained autoantibody production requires efficient T cell help. However, to the best of my knowledge, we need to accept the fact that we do know little about the respective T cell epitopes responsible for this help. Even more challenging, non-proteinaceous autoantigenic targets such as glycolipids (gangliosides) do not contain T cell epitopes despite their pathogenitic role in autoimmune peripheral neuropathies and high specificity. The more interesting, anti-ganglioside antibodies recognize just one epitope determined by the number and position of sialic acid attached to no more than four sugar residues. The loss of tolerance in such a case is still poorly understood and a great enigma to me.

As I’m sure you understand, Dr. Roggenbuck, just because autoantibodies target non-proteinaceous epitopes does not preclude that there were proteins physically associated in vivo with some form of the autoantigen and which produced the T cell peptidic epitope underlying this autoimmune response. But as you pointed out there is usually little if anything known about the T cell specificity underlying autoantibody elicitation. One recent study was particularly remarkable to me in this regard because it reported on the presence of CD4+ T cells autoreactive with peptides derived from hypocretin/orexin in narcolepsy patients (http://stm.sciencemag.org/content/5/216/216ra176.full.html), yet there is no associated autoantibody response in these patients. So, it may be possible to have pathogenic autoreactive T cells without autoantibodies, but probably not autoantibodies without T cells, as elusive as they might be.

Inıtiation of autoantibody production (from what i understand from the discussion) is largely random and non-specific, however you end-up finding monoclonal/oligoclonal aggregates of plasma cells in RA synovium (i actually had an RA patient diagnosed as having a plasmacytoma in the elbow joint, which responded very well by the way to a local corticosteroid injection :)). And we also know from aCCP studies that as you get more IgG and affinity maturation you're more likely to develop RA.

On the other hand it would really be interesting to see whether and how post-translational modifications change throughout the course of life, at the earlier stages of which you're more likely to develop tolerance.

Many thanks indeed, Dr. Rubin, for the interesting hint. I am trying to get hold of the paper. What do you think about naturally occuring autoantibodies? Do B cells generating these autoantibodies also require T cell help?

An interesting discussion. An irony that has struck me is that antibodies to post translational modifications may not in fact strictly qualify for being 'autoantibodies' since they may react with a whole range of proteins, both self and non-self, carrying something like a citrulline residue in one of a range of suitable peptide sequences, as in RA. The MHCII shared epitope association with ACPA indicates that this is almost certainly a T cell dependent response and it looks likely that citrulline may bind preferentially to the SE. But this begins to play merry hell with the interpretation of the likely respective T and B cell specificities, since if the citrulline is binding to DR then the T cell receptor binding to the combination is probably not.

The more I look at autoimmunity the more it seems to me that we should assume that the B and T cell specificities are likely to obey totally different rules and may never actually be to 'the same antigen'. I agree with Robert Rubin that the T cell responses to hypocretin reported are very remarkable and might actually indicate a form of 'classical T cell model' autoimmunity that does not actually apply to all the other autoimmune diseases we know, in which the T cell specificities are chiefly unknown (coeliac yes, but again the B cell antigen is different). Yet the absence of autoantibody maybe suggests that there might be an autoantibody to something else that we do not know how to look for, that somehow encourages the antihypocretin T cells.

One thing that might turn out to be true is that abnormal antibody responses to PTM epitopes do not in fact turn up often and that citrulline is the exception because citrulline is probably some sort of innate immune 'danger' or other signal. That might allow it to induce promiscuous help. An interesting twist to the story is that Dr G Cambridge has found ACPA of different isotype profile in a proportion of men with coronary artery disease. What that might mean will take a while to work out I think.

Concerning Dr. Roggenbuck’s question about naturally occurring autoantibodies, I believe these generally derive from B cells, mainly B-1 cells, with little somatic mutation of the H- and L-chains, consistent with the view that they are driven in a T-independent fashion. So, much of their specificity is probably germ line encoded, and they may be stimulated by multivalent microbial products with cross-reaction to self-epitopes. However, some naturally-occurring autoantibodies could be T-dependent with the T-cell recognizing an epitope also derived from the microbe and which somehow was unaffected by T cell tolerance mechanisms; this would be the classical “molecular mimicry” concept.

Dr. Edwards brings up all sorts of fascinating dilemmas in the world of autoimmunity. Here’s my take: 1) I don’t think the capacity of an antibody to bind a post-translational modification of both foreign- and self-derived material disqualifies it from being called an autoantibody, although it does raise questions about the origin of the inducing agent. This makes the need to identify the responsible T cell epitope particularly important. 2) The presumed mutual exclusivity of the peptide side-chains that interact with the MHC pockets and those that are complementary to the residues in TCR would indeed make for some difficulties in imagining the nature of the T cell epitope. One idea could be that PTMs increase the affinity and therefore density of self-antigens presented on MHC, but the T cell response is truly autoreactive to the unmodified self. Dr. Edward’s suggestion that citrulline might be particularly immunostimulatory could enhance this scenario. However, just because the MHC on the B cell might be interacting with the citrulline moiety, doesn’t mean its BCR isn’t also, consistent with the development of T cell dependent anti-citrullinated protein antibodies in RA. 3) I’m not willing to discard the “classical rules” of T-B interaction when applied to autoimmunity. I think the immune system has had a very hard time distinguishing self from foreign so has evolved many tolerance mechanisms to try to keep us alive. 4) Years ago we published a study on narcolepsy patients in which we failed to find any of the classical autoantibodies to non-organ-specific antigens or to antibodies to neuronal extracts, but we could easily have missed the target(s). Alternatively, maybe narcolepsy is a B-independent, T cell autoimmune disease driven by professional APCs in the CNS.

Yes, I agree with pretty much all of that Robert.

I was not challenging the classical rules of T-B interaction. I think those have a firm basis in the work on the mechanism of antigen presentation in the 1980s and subsequent work on costimulatory factors etc etc. What I do challenge is the 'classical T cell model of autoimmunity' by which I mean the presumption I spent ten years proving wrong for things like RA; that the process is always driven by autoreactive T cells, with B cell responses tagging along or just being epiphenomenal. In most of the autoimmune rheumatic diseases I have worked with we have no real evidence for autoreactive T cells. We have abundant evidence for T cell help but in most cases there are mechanisms available that would allow T cells with specificities only for foreign antigens being seduced into helping autoreactive B cells. Roosnek and Lanzavecchia is just the tip of the iceberg. I amy be proved wrong in the end but having worked out an alternative theoretical model we did try it out with rituximab for RA and the showed that B cells drive inflammation through antibody and that on their own T cells don't seem to do much at all! There are layers of further complexity but the idea that RA was a 'T cell disease' proved to be a dead duck.

Dr. Edwards: I am curious about your interpretation of the Melet, et al, 2013 study in A&R on Rituximab effects in RA in which they showed substantial CD4+ T cell depletion (in addition to B cells) in most patients, but patients resistant to T cell depletion had poor clinical response. This seems to support the role of T cells in the pathogenesis of RA with B cells serving as their personal APCs.

I have not looked at the Melet paper in detail, so I am going on the abstract. I would not be surprised if there were shifts in CD4 T cell numbers actually during the treatment period -which is what they measured. We saw quite big swings early on in all subsets which we assumed were due to the steroid premedication (we saw similar changes in patients given steroid plus placebo). But by two months all the T cell subsets are back to normal. Maria Leandro had masses of data on this from all the early work and over the years we have done lymphocyte studies on several thousand treatments. It is worth noting that the therapeutic benefit does not tend to kick in (beyond a brief steroid holiday) until three month or longer. Any immediate shift in T cells is clearly not relevant to that in any direct way. Patients have been known to remain well for up to five or more years after a single course of rituximab, during which time their lymphocyte profiles are normal. (Even the B cells are often back to normal for years.) What is different is their autoantibody levels.

It may be worth pointing out that our suggestion that RA was driven by B cells through antibody after all was very unpopular. It was particularly unpopular with a politically astute group of immunologists who thought they were going to make their reputations by showing that B cells were relevant to autoimmunity after all because they -surprise surprise - presented antigen to T cells - maybe some of them were too young to know we had known that for twenty years. The pharmacodynamics with rituximab show that this makes no sense. However, the drug company involved was keen to be in the in crowd so the ghost writer for the NEJM paper was told to include a statement about how the effect might be via an effect on T cells. The fact that this did not seem to make any sense in the context of the data was not picked up by many rheumatologists or immunologists as far as I know - the only person who complained about it to me was a rather sharp Wall Street analyst who like to feel he understood the science he was betting on. Ever since (as well as before) the popular story has been that B cells are important because they present antigen to T cells. Nobody stops to think that we did not see any pathology that looked like direct T cell attack so presumably the effect of talking to CD4 T cells would be maturation of B cells (which were now gone) into antibody secreting plasma cells anyway. One of the reasons I retired from immunology was the extraordinary slowness of the community to take note of what data actually show rather than what has been in fashion for the last thirty years! (Like the idea of molecular mimicry, which was a vague speculation in the 1960s that makes little sense and as far as I know has never been shown in human autoimmunity.) People are still trying to show it is an effect on T cells, in whatever way they can - and I guess the Melet paper is in this tradition. One day the penny will drop.

Best wishes

Jo E

Many thanks indeed to Drs Rubin and Edwards for their most enlightening discussion. I still remember the days when RA was considered to be a T cell mediated disease. Furthermore, it was very intriguing to note how difficult it was for the community to accept a new concept taking into account B cells as a driving factor in RA. After the discovery of the role of PTM such as citrullination and carbamylation in antigen presentation, I suppose that there is now a better understanding of the inflammatory process in RA. However, we still do not know the triggers of inflammation let alone the cause thereof. Seems to be an uphill task which I hope will not deter the younger generation of immunologists to take on.

Dear Dirk,

The beauty of the B cell model that led us to use rituximab is that means that there are no triggers. So the causality is entirely covered by threshold factors (shared epitope, two X chromosomes, smoking, PTPN22 (probably through its role in receptor editing)) and the inherent metastability of an antibody production system that makes use of a positive feedback loop (low affinity antibody drives more, higher affinity antibody). Any plausible design of such a regulatory system has inherent 'system bugs' that can lead to inappropriate chain reactions. Whether these actually occur is dependent on the threshold factors.

The mistake in the received wisdom is that these diseases are caused by a combination of internal genetic factors and environmental triggers. This is not good epidemiology. External factors can be triggers or threshold modifiers. Internal factors can be genetic or stochastic (and that's making it simple). The epidemiology of most autoimmune disease is non epidemic so we are not expecting environmental trigger factors, only threshold factors. The epidemiology is predominantly stochastic - even with the genetic bias you have no idea when the disease will set up. That makes complete sense for an error in regulation in a system that operates entirely from a stochastic base. Antibody species are generated at random. So we can expect to be looking for genetic and maybe environmental thresholds, which we find, and chance. Put differently, all you need to do is generate an antibody that sets up an inappropriate positive feedback because of some cross talk with generic (innate) signalling ligands and the whole thing snowballs. We are not looking for a virus or a bacterium that starts things off like rheumatic fever. The sad truth is that rheumatologists have still not woken up to the fact that they are not looking after children with rheumatic fever like their mentors were 60 years ago!

The work to be done in RA, to my mind, is the hard slog of defining the detailed feedback mechanisms and their thresholds with respect to potential subversion mechanisms. We understand a lot about the rules of feedback (CD21, FcgammaRIIb etc etc) and we know some thresholds but we need to know details that are hard to come by. One thing that looks interesting is that usage of the VH4-34 gene by a subset of B cells that rarely makes much antibody in the normal situation but often does in autoimmunity.

The concept of autonomous B cell mutation and RF autoantibody-mediated pathogenesis in RA pioneered by Dr. Edwards deserves commendation and serious attention by the clinical and basic research community. However, aversion to attributing autoreactive T cells as a necessary component of this process strikes me as a bit obstinate. The K/BxN mouse model is informative in this regard. While the capacity of anti-glucose-6-phosphate isomerase (GPI) autoantibodies, which spontaneously develop in these mice, have by themselves joint disease pathogenic capacity in passive transfer studies, their elicitation is entirely dependent on autoreactive T cells to a peptide derived from GPI. The same kind of thing could be happening in RA – that is RF+ B cells presenting mu- or gamma-derived peptides to autoreactive T cells. The origin of anti-CCP would be along similar lines with the added assist of PTM we previously discussed. So, while there may be “…mechanisms available that would allow T cells with specificities only for foreign antigens being seduced into helping autoreactive B cells.”, that would be a contrived and unnecessary scenario in my opinion. Furthermore, joint inflammation seems to be at high risk as Dr. Roggenbuck pointed out, and synovial FcR involvement is a good candidate for IgG RF and citrullinated proteins for anti-CCP in mediating damage. But autoreactive T cells recognizing the same epitope presented by B cells for autoantibody production could also be contributing to joint disease by making osteoclasts even more active in cytokine production. Local autoantibody production could also contribute as Dr. Edwards has proposed. I don’t think the Melet, et al study tried to resurrect an “old” idea on the role of T cells in RA – it was just an observation that with RA patients treated with Rituximab “…The decreases in CD4 and CD3 cell counts were significantly greater for moderate responders (on average, 43% and 37%, respectively) and good responders (on average, 47% and 38%, respectively) than for nonresponders (on average, 7% and 7%, respectively)”. This is consistent with an important and probably necessary contribution of T cells to the disease process, and there is no reason to believe these would not also be autoreactive T cells.

Dear Robert,

I am happy if you wish to stick to the old view, but that seems to me more where the obstinacy and unnecessary contrivance lie! What could be more unnecessary and contrived in studying naturally occurring autoimmunity than inserting a T cell transgene in a mouse. The K/BxN mouse can tell us nothing about the way autoimmunity develops in man because it was engineered to create autoimmunity in a way that cannot be the basis of the human disease! Think about it.

A small point before recounting a little of the T cell story. The autonomous B cell mutation in our theory of RA pathogenesis is of course nothing other than the way we know B cells work – all antibodies come from autonomous mutation. Sometimes people have thought we were talking about something unusual (even Diane it seems).

The T cell story for RA arose in the wake of Alero Thomas and I demonstrating the presence of what T cells and what we now call DR in inflamed synovium in 1979. It seemed to me to be unsurprising and what people seemed to miss was that the first sample we used was from a meniscal injury, not RA! We had known for a decade that most of the lymphocytes in synovium were T but it later became clear that this is true of all inflammation – B cells have no normal local function in inflammation and are usually hard to find. We had also known since the 1960s that macrophage activation (then called intimal cell hypertrophy) and endothelial swelling precede T cell infiltration (Maarten Kraan showed this nicely). The T cells come in because of the TNF from the macrophages. Of note Fc receptor ligation produces about 100 times more TNF than was seen with T cell stimulation in the synovial co-cultures. So there was never any histological reason to think there was a T cell mediated reaction.

T cells were sexy in London in 1980, that’s all, and some people wanted to get some grants in to make monoclonals. There never was a story of autoreactive T cells. People have looked for them for nearly 40 years and found nothing. How much more negative evidence does one need? And it wouldn’t make any sense for RA to be due to inappropriate T cell responses. T cell repertoires are formed in early life. Encounters with environmental antigens occur mostly in early life. So a disease triggered by T cell responses should be a childhood disease and RA is the opposite! The T cell emperor in autoimmunity has never had a stitch on I am afraid. I have no aversion to implicating T cells – if they turn out to be relevant in narcolepsy that would be absolutely fascinating, and I am sure they are relevant in some rare endocrine syndromes but for routine autoimmune disease the **evidence** condemns the idea to the trash can.

The political pressure in Europe to persist with the T cell explanation has been huge. Vast grant programmes depend on it. It makes no difference to me since I left immunology four years ago. I will never write another grant. There have also been powerful commercial reasons for hoping that rituximab did not work the way we said it did. If you had seen the letters flying about in patent cases you would understand!

Nobody is suggesting that T cells are not involved. That is the non-sequitur. Of course the disease is T cell dependent. The Melet data are, I believe, still irrelevant because the pharmacodynamics make no sense, as I said. But of course T cells contribute to the disease. It is just that we have not the slightest hint, despite thirty plus years of looking, of an autoreactive T cell in RA as far as I know. Why should we expect it? Just because in the 1980s T cells were sexier and people like Av Mitchison wrote on some tablets that they ‘drove’ the B cell response. We have learnt it is more complex since then. The Roosnek and Lanzavecchia mechanism is much more plausible because it explains why normal people have transient RF after infection. Autoreactive T cells are redundant for RF responses. The situation for ACPA is more difficult to work out, but it is easy enough to see why rules should be broken for a PTM.

Yes, if you induce autoreactive T cells either with vast quantities of adjuvant or transgenes in mice you will get disease, but spontaneous human disease by definition does not work that way! It is a stochastic disease of advancing years - a time when autonomous mutation in B cells goes on being stochastic.

All best wishes

Jo E

Well, Jo, I am more that a little surprised that you did not take better advantage of the entrée into the KBxN mouse model of RA that I opened with. Here’s an example where passive transfer of autoantibody – no T cells – can cause inflammatory joint disease, exactly as you are proposing for RF. And just as a historical point of information: as I remember, the discovery by Valerie Kouskoff with Diane Mathis/Christoph Benoist of arthritis in these mice was completely serendipitous; the KBxN was not engineered to create autoimmune disease. In addition to anti-GPI T cells, the I-Ag7 MHC class II is also required to develop spontaneous arthritis, reflecting, I believe the propensity of this MHC molecule to cause tolerance abnormalities in T cells selected on it. A similar thing might happen as we humans age because not the entire “T cell repertoires are formed in early life.”; even people as old as me continue to produce T cells de novo from the thymus, but I would propose that these are increasingly likely to become autoreactive because of failing tolerance mechanisms. That could be why most autoimmune diseases are late onset. And I totally agree with you that whether or not autoimmune disease develops is largely due to internal stochastic events, the basis of which lies in the origin of the TCR and BCR.

Concerning your point about the dearth of autoreactive T cell reports and “How much more negative evidence does one need?”, I might remind you that up to the early ‘80s, many people were beginning to question if antigen-specific T cells really existed because nobody could successfully clone T cells despite many attempts. As Efraim Racker wrote, “A good hypothesis [such as pathogenic autoreactive T cells – my words] is worth a few ugly little facts and a few hundred negative experiments.”

In the end, while it is true that arthritic-type joint disease can be mediated largely by antibodies, immune complexes and cells with FcR, this mechanism comes across as unnecessarily restrictive within the context of a highly interdependent immune system that has a strong probabilistic component. If “Of course the disease is T cell dependent”, then I assume you mean that at least somatic mutation of the BCR is not autonomous. So, if autoreactive T cells “help” in autoantibody production (which you seem to reject for reasons that are obscure to me), how could such effector T cells not go on the live another day, home to tissue where their target antigen is abundant and contribute to inflammatory synovitis? That’s what T cells do in an anti-microbe response – the fundamental problem with autoreactive B and T cells is that the “pathogen” can never be cleared. While “Autoreactive T cells are redundant for RF responses”, redundancy is practically the definition of the immune system.

Fair enough Robert,

I enjoyed hearing the work on the KBxN pathways, but we had known that immune complexes cause arthritis since anti-tetanus horse serum and the role of small complexes we had worked out in the human context over the period 1993-6 with our work with CD16 and DAF in the synovial environment. KBxN was nice confirmation but I don't think Diane has ever forgiven me for standing up at Keystone and congratulating her on finding the real mouse disease for which human RA was such a good model!

I am still worried by the word 'autonomous'. I have never suggested that RF B cells are autonomous in the sense of being T cell dependent and somatic mutation is just somatic mutation - neither autonomous or otherwise. It is always random. This is where I have been puzzled by people thinking we were talking of some 'special' sort of mutation that broke rules. The mutation that generates RF would be just like any other Ig gene translocation/mutation. What would be different would be the ability of the resulting B cell to acquire help in the context of that particular BCR species - which for RF would partly be the usual physiological mechanism but partly something more subtle, very likely involving cross-signalling between several clones with different fine specificities. We went into this a bit in the 1999 Immunology paper but really one can only speculate which of the many possibilities is likely to occur.

I do not reject help from autoreactive T cells, I am just saying that so far we have little or no evidence and we do not need to postulate it. It is against Ockham's razor to invoke things we do not need - truly redundant, rather than 'redundant' in the biological sense of there often being multiple pathways available.

I am not sure why we should think that in RA T cell antigens should be any more abundant in synovium than elsewhere? Nobody has ever found a convincing joint related antigen in RA. RA is not a synovium specific disease any more than serum sickness. The reason why synovium gets inflamed looks to have nothing to do with where antigens are - it looks to be due to the unique combination of the presence of CD16+ tissue macropahges (which normally occur only in synovial intima, serosae, sclera, bone marrow, secondary lymphoid tissue, alveoli, salivary gland and subcutaneous tissue specifically at sites of stress - the exact places where RA produces disease) and the high level of expression of VCAM-1 and DAF on synovial intimal fibroblasts - which is only shared with the bone marrow nurse cells that are needed for B cell survival (probably of the same embryological origin). This should give a 'double whammy' for B cell/small complex mediated disease. I did not entirely agree with the original interpretation of the susceptibility of joints in KBxN (G6PI is obviously not a joint antigen) because anti-complement activity in synovial fluid was shown to be high by Medof way back, but I don't think we fully understand the relation between DAF and CD16 ligation.

So I see no reason to invoke autoreactive T cells homing to joints in RA and anyway, after good B cell depletion the CRP normalises and the joints no longer show signs of inflammation, despite normal CD4 counts.

Some good questions María,

When I say small complexes I mean really small complexes, and in particular complexes too small to fix C1q. CD16 seems to have a particular ‘early warning’ function in generating responses to complexes with maybe two or three IgG Fc sites only. These complexes will have molecular radii barely larger than IgG and smaller than a single IgM. Being so small they can cross endothelium but would not form any deposits in glomerular BM regions and since they do not bind complement they would not activate or damage endothelium. My understanding is that glomerular disease occurs when larger complexes are not cleared by complement, for whatever reason. Paradoxically, once deposited these may engage complement locally and cause endothelial damage.

T cell responses are more restricted than B cell in general (an antibody can bind to even inorganic macromolecules) but both are restricted in complicated ways.

If CRP normalizes following B cell depletion that will persist for as long as there is depletion and for anything up to five years longer despite B cell return. Basically it persists for as long as the autoantibody level drop is maintained.

You are right to say that the joint involvement in RA is different from serum sickness. This is where we think the VCAM-1 story comes in. In serum sickness you just have circulating complexes causing inflammation either through Fc ligation or maybe complement. In RA the activation of intimal macrophages calls in inflammatory cells including plasma cells and all the cells of secondary lymphoid tissue. Because of the propensity of synovial stroma to express VCAM-1, DAF and in deep stroma CD21, shared with bone marrow nurse cells and follicular dendritic cells, this provides a way for all later stages of the B lineage to colonise and survive actually in the joint. That ramps up the level of local antibody and complex production to the extent that TNF and other cytokine levels rise to something similar to septic arthritis. At this level the density of immigrant cells is such that glucose levels fall to zero, pH changes and, probably also due to other changes, the chondrocytes die. Once chondrocytes are dead the matrix has no defense against proteinases. Cytokines also drive osteoclast activity both in periosteum and subchondral marrow. In comparison to synovium lesions at other sites like pleura and sclera are much less often affected by matrix destruction, presumably because they do not support local ectopic lymphoid aggregate formation.

After B cell depletion synovial samples show a return to more normal architecture. Bone and cartilage damage is inhibited, although not as completely as with direct TNF blockade.

Both T and B cells can drive TNF production from macrophages. I do not think one can make a simple comparison but when we looked at release of TNF from macrophages in response to CD16 ligation by cross linking with antibody the levels seemed to be about 100 times higher than those seen after co-culture of T cells with macrophages. Traditionally, directly T cell mediated responses are rather slow and low key as in the tuberculous ‘cold abscess’ whereas complex mediated inflammation can be violent and even rapidly fatal.

Concerning Maria Zavala-Cerna’s comment/question about T cells vs. B cells, the B cell repertoire is highly promiscuous, with capacity to bind just about anything we throw at it. Even individual B cells (via monoclonal antibodies) are often multireactive, capable of binding dissimilar epitopes. Autoreactive B cells are abundant in the peripheral immune repertoire because of the limitations of central B cell tolerance. That is why nature does not entrust B cells with autonomy in antibody response and generally requires their dependence on “help” from CD4 T cells, which will recognize peptidic antigen only in the context of MHC class II and which are subjected to numerous tolerance mechanisms to inhibit autoreactivity. However, the antigen that mediates a B cell response needs only to have substantial affinity for the BCR and have capacity to be processed into a T cell epitope, and the B- and T-cell epitopes need only to be physically linked as the B cell endocytosis this material.

This brings me to rheumatoid factor, anti-IgG. This is a potentially very dangerous antibody because the T cell antigen providing help for this B cell can be derived from essentially any and perhaps all immune complexes; other than the IgG component, the T cell epitope is not necessarily self-derived so is not restricted by tolerance mechanisms.

However, while IgM RF is easy to detect in normal individuals, it is low affinity and/or low concentration, suggesting that these B cells do not generally receive T cell help. It is unclear to me why class switching and affinity maturation generally does not occur, but this is a very good thing. IgM RF is probably incapable of binding soluble IgG, and even if it bound IgG-containing immune complexes, there are no mu-chain specific Fc receptors in humans, so inflammatory sequelae would be limited. Nevertheless, IgM RF is a common test in clinical labs to support the diagnosis of RA, although specificity and sensitivity are only ~70-75%.

Dr. Edward’s work implicates IgG RF in pathogenesis of RA. I think this has been an uphill battle in part because IgG RF is a technically difficult assay and is not routinely performed, at least in the U.S. But I think, Jo, you make a compelling case for implicating IgG RF immune complexes in joint disease. And I acknowledge that there seems no reason why such antibodies would not readily develop as long as there are IgM RF precursor B cells and T cell help to some component of potentially any immune complex that could bind such B cells via the IgG Fc. Fortunately, this does not seem to happen very often, possibly because of the transient nature and/or inaccessibility of circulating immune complexes and perhaps the need for additional phenomena such as “…cross-signaling between several [B cell] clones with different fine specificities”. So, for what it is worth, I have now become a Jonathan Edwards disciple, and I wish you well in expanding your followers. And thanks to Debasis for initiating this conversation and to Maria for forcing me to think a little harder about all this.

Dear Robert,

Thanks for the comments. As you say, there are several paradoxes in the RF story. We tried to produce a complete mechanistic theory in our 1999 Immunology paper and earlier in a chapter I have just had to scan in for somebody because there are no copies of the book in the USA! The diagram in that chapter has 25 steps. My memory is that my most complicated diagram had 55 steps, with four positive feedback steps and more negative ones than I could count. Balancing effects from FcgammaRIIb and CR1 and various modes of linking of more than one receptor figured, together with the need for several clones to feed off each others' signals. So your point about IgM RF being perhaps ordinarily T cell independent is a good one, but my guess is that in RA persistent IgM RF production is bound in to the IgG RF and IgA RF production that IS T cell dependent and which generate, at least for IgG RF, complexes that encourage IgM RF B cells.

We failed to get a watertight story, because the number of possibilities are just too great. But maybe that is the lesson - RA may come about in lots of slightly different ways. And of course since that time we have had to reconsider the story in the light of ACPA. Do they feed off RF, or the other way around or are they runners in the same race or what? There is now an embarasse de richesse available for mechanisms. Yet Dr Cambridge and I hang on to the idea that there is a general principle here of 'cheating on the signals'. The 9G4 epitope on VH4-34 increasingly looks to us to be of much wider relevance to this cheating than was realised. It seems to mark a completely different low threshold system that is supposed to 'burn out' maybe by 'overstimulation induced death' but for some reason sometimes does not.

I have had to admit defeat on a tight story, and become a philosopher, but Jo Cambridge has not and the truth will one day emerge I think!

Dear all,

thanks for such a lively and informative discussion. I myself will take much time to find out the conclusion. Hope, I may come up with some interesting inference.