William O. Weigle - 7:03pm May 13, 1997 (#5 of 14) In response to Ephraim (#2), I know of no dogma that has persisted so tenaciously throughout the years with such meager experimental evidence for its existence as low-high zone tolerance. This entire concept is based on one single antigen in one single laboratory with an experiment that involved a series of multiple weekly injections of antigen over a considerable period of time. To my knowledge, this experiment has not been repeated with any other antigen by any other worker in the field. However, many are quick to quote the term "high zone tolerance" when large amounts of antigen are required for the induction of tolerance. In none of these studies have the authors also shown tolerance with a low dose of antigen, with immune responses occurring with intermediate doses. The ability to induce tolerance in T cells in B-cell knockout mice with high doses of monomeric human gamma globulin (HGG) (Phillips et al., 1996) conflicts markedly with the low-high zone model, especially with the explanation offered by Ephraim and others. According to the theory offered, at extremely low doses of antigen, specific B cells would preferentially pick up antigen by high-affinity binding to surface Ig. Because resting B cells do not express appropriate T-cell costimulatory signal, any interaction with an antigen-specific virgin T cell would be tolerized. At an increased antigen concentration, uptake by other MHC class II positive cells would occur. These cells, the so-called professional APCs, would present antigen to the virgin T cell in an immunogenic fashion. At high antigen concentrations, all B cells would take up antigen by pinocytosis. Because B cells far outnumber other MHC class II positive cells, the result would again be tolerance of the virgin T cell (Eynon and Parker, 1992; Fuchs and Matzinger, 1992). This explanation presumes that, first, only B cells are able to induce T-cell tolerance; second, B cells are tolerogenic even in the presence of their specific antigen; and third, all non-B-cell antigen-bearing cells are constitutively activators of virgin T cells. This assumption provides a framework explaining the midrange immunogenic dose of antigen and implies that a B-cell-deficient animal would not be able to induce peripheral T-cell tolerance to high doses of exogenous antigen. However, the fact that T cells from B-cell knockout mice given monomeric HGG are tolerized rather than activated indicates that low-high zone tolerance cannot be explained by the above hypothesis. Actually, there is no evidence for two-zone tolerance with HGG over a wide range of tolerogen doses (10 mg/mouse to 5 pg/mouse) (Golub and Weigle, 1969). In fact, in a second series of experiments by Mitchison (1968), attempts were made to induce low-high zone tolerance with five different antigens. The only antigen that could successfully induce tolerance was bovine serum albumin (BSA). With such meager evidence for low-high zone tolerance, I personally believe this dogma can be disregarded without further comment. However, if one needs to explain low-high zone tolerance with BSA, it can be readily done on the basis of contamination with endotoxin. BSA and possibly other mammalian serum albumins, because of their charge, have been shown to be associated with endotoxins (Dvorak and Bast, 1970). Thus, the injection of low doses of BSA may not contain sufficient endotoxin to interfere with the induction of tolerance. Intermediate doses contain sufficient endotoxin to mount an antibody response to BSA, whereas with higher doses of BSA the interference of tolerance induction by endotoxin may be overridden. A number of years ago in my laboratory, Jim Clagett, using endotoxin-tolerant mice, was unable to demonstrate a low-high zone-tolerant state to BSA. Mitchison's study with BSA is the only study I am aware of that has been carried out in adult animals. There is one other study that was carried out in Gus Nossal's laboratory by Shellam (1969), in which a low-high zone tolerance in neonatal rats was suggested with polymerized flagellin. This antigen was prepared from Salmonella adelaide, which has the most potent endotoxin we have ever seen in my laboratory. Without additional support for the low-high zone tolerance, I am not impressed with any attempt to use it to explain the theory of self-nonself recognition. I question some of the interpretation that Ephraim alluded to in the neonatal model (#2). Apparently he believes that the permissiveness for the induction of tolerance in neonatal mice is the result of a deficiency in uncommitted T cells. However, I think the majority of evidence for permissiveness is in the deficiency of the environment of these cells rather than in the cells themselves. It is most likely that this deficiency lies in antigen presentation (Lu et al., 1979; Nadler et al., 1980). A number of years ago, we showed that the transfer of neonatal cells to adult animals results in a normal antibody response, whereas transfer of adult cells into neonatal animals fails to mount a T-helper cell response upon appropriate challenge (Dixon and Weigle, 1959). Furthermore, a number of investigators have demonstrated throughout the years that the transfer of antigen-presenting cells to neonatal animals corrects this deficiency (Martin, 1966; Argryris, 1968; Ridge et al., 1996). I also question Ephraim's suggestion that dendritic cells generate only an immunogenic signal while other antigen-presenting cells, such as B cells and macrophages, send only tolerogenic signals to naive T cells. If this is true, macrophages would, out of necessity, be the only tolerogen-presenting cells in B-cell knockout mice tolerized to monomeric HGG. One must also address the experiments of Finkelman (1996), who reported that targeting immunoglobulin epitopes to dendritic cells results in a sound tolerant state. It is my opinion that any cell (including dendritic cells) is capable of delivering a tolerogenic signal in the absence of additional signals to the T cell. It is important to note that one cannot write off the
HGG model as being associated with downregulation of the immune response
through the Fc receptors. Recently, we were able to readily tolerize Fc-receptor
knockout mice obtained from Jeff Ravetch. These mice lack functional Fc-gamma-RI
and Fc-gamma-RIII and have a disrupted Fc-gamma-RII. This experiment eliminated
a possible role for Fc-gamma-Rs in T-cell tolerance. We were also able
to induce a solid unresponsive state in Fc-gamma-RII knockout mice, demonstrating
that neither T- nor B-cell tolerance was associated with Fc-receptor engagement.
Furthermore, we were able to effectively tolerize Fc-gamma-RII knockout
mice at both the T- and B-cell levels despite the fact that these mice
showed a 10-fold enhancement in the antibody response in comparison to
controls (Whitmer, in press).
Rod Langman - 8:09am May 16, 1997 (#6 of 14) I heartily agree with Bill (#5)
that low dose tolerance should be stricken from the record as a nonevent.
As I recall, Parish used a fragment of the same flagellin to induce humoral
unresponsiveness more easily than with the whole monomer, but it turned
out that the "tolerization" scheme resulted in a switch in class when he
found the antibody-tolerant mice were in fact delayed type hypersensitivity
immune to flagellin.
Antonio Bandeira - 2:09pm May 16, 1997 (#7 of 14) Do we accept that there are two models of tolerance, one
occurring in the thymus and the other in the periphery (#3)?
Antonio and I say no. This is an important point that we would like to
clarify. We consider that the thymic (central) tolerance - or, better,
thymic selection - is tightly linked to the establishment of peripheral
tolerance to tissue-specific antigens during normal development. And why
so? Many reasons, but first, just a quick example. Sakaguchi
et al. (1982) have shown, following a number of similar observations
in the past (including in humans), that in several mouse strains neonatal
thymectomy performed 2-4 days after birth leads to the generation of so-called
organ-specific autoimmune diseases (stomach, thyroid, ovaries, testicles,
etc.); these are CD4 T-cell-mediated, because disease can be transferred
by these cells to T-cell-deprived recipients. After day 4, the effects
of thymectomy will no longer promote disease. Again, as a blunt example
of regulation, normal adult CD4 T-cell populations can prevent the onset
of disease if cotransferred with aggressive T cells. Again, a window in
development for tolerization is described. How then could we envisage the
link betweeAF 0 5 _ 0 0 3 b . g i f F05_003BGIF S¤„>'Ž' ¥„>'÷\ What is the repertoire of the first regulatory T cells? We postulate that many recognize "ubiquitous" ligands shared by TE and many tissues in the body. Thus, once in the periphery, they can locally migrate into the tissues together with naive tissue-specific T cells. As some of us agree in this forum, a major discrimination between self and nonself ligands is historical: Almost all (but not all) self antigens are already there, available from the start to the emerging immune system. This brings us immediately to another huge difference between perinatal and adult immune systems: T cells have the opportunity to recognize ligands in the periphery as soon as they exit the thymus - that is, as recent thymic emigrants (RTE) - while in the adult, 95-99% of peripheral T cells are no longer RTE, because most naive T cells have life spans of 1-2 months. Why is this important? Our own experiments have shown that RTE (and by far much less effectively "older" T cells) can "learn"! If at the stage of naive RTE they see antigen in the presence of TE-selected regulatory T cells, they become committed to the same regulatory effector function and are now able to suppress the activity of nontolerant T cells. This phenomenon we named T-cell "education" (a similar phenomenon, "infectiousness," was coined by Waldmann's group using his experimental system). As a consequence of T-cell education, the repertoire of regulatory T cells drifts during development from TE specific to tissue specific. The role of TE-selected regulatory T cells (the first wave of regulatory cells) is to trigger the process of peripheral education. This is necessary because, as shown in many experiments (including Sakaguchi et al.), RTE can be immediately activated into destructive functions, in contrast with the classical notion that, as thymocytes, they are very susceptible to anergy induction. Once the peripheral dynamics is established, newly formed autoreactive T cells produced throughout life can be immediately recruited in the pool of regulatory cells and thus cause no harm to the organism. In this way, the model explains the major problem of Lederberg's concept: new autoreactive lymphocytes are produced all the time, but natural tolerance is only established early in life, as Medawar and colleagues brilliantly showed 40 years ago. At this point, I reiterate the conceptual differences
between the models, but now in relation to therapies. One fundamental difference
separates the Coutinho-Bandeira model from all the others: the "horror
autotoxicus." We say that we need functional regulatory self-reactive T
cells and self-reactive antibodies for the maintenance of tolerance to
our own tissues and tolerance to our many bacterial friends that colonize
our gut. We therefore consider that the problem of autoimmune diseases
is not the presence of potentially aggressive autoreactive cells (we have
them all the time in any case) but failure of the regulatory mechanisms.
Consequently, autoimmune diseases should be treated as immunodeficiencies
(the association of the two in clinics is remarkable), and therefore forms
of immunostimulation are to be considered. In transplantation, we already
know how to stop autoaggressive cells. The problem is how to educate graft-reactive
cells and stably establish the proper regulatory mechanisms. For tumors,
it is the inverse problem. Strategies to locally unbalance ratios of regulatory/aggressive
cells should be found (this could actually be the reason why, with some
tumors, local adjuvant therapies seem to have some effect) while preserving
the systemic regulatory system.
Zlatko Dembic - 7:15pm May 16, 1997 (#8 of 14) In response to Bill (#5): We have observed peripheral and central tolerance in our TCR-transgenic (tg)/SCID mouse model, in which the TCR is specific for the peptide derived from the mutated form of Ig lambda2 light chain together with MHC class II Ed molecule. The mutation lies in the third hypervariable region of the V domain and comprises a three-amino-acid difference in comparison to the germ-line sequence. The antigen, lambda2-315 light chain (L-Mut), is produced by the MOPC315 plasmacytoma in association with IgA heavy chain. Subcutaneous injection of 160,000 MOPC315 cells per mouse in syngeneic combination gave no tumors in almost 90-100% of TCR-tg and SCID/TCR-tg mice when monitored for more than 4 months after the injection, whereas, in control mice, tumors appeared regularly after 14 days. The response was antigen specific, because an unrelated plasmacytoma (J558) had an unimpeded growth (Lauritzsen et al., 1994; Bogen, 1996; Bogen et al., 1995; for a description of the model, see Bogen et al., 1992 and Bogen et al., 1993). So, the anti-L-Mut TCR-tg CD4 cells were able to protect these mice, which had neither B nor CD8 cells (SCIDs). Like other plasmacytomas, MOPC315 has no MHC class II molecules on its surface, but it secretes a large amount of monoclonal antibody. The rejection mechanism remains unknown. However, when 2 million MOPC315 cells were injected per mouse, the tumor broke the resistance or induced tolerance and grew to kill the mice. Big-tumor-bearing mice neither had CD4 cells that were able to mount a proliferative response to antigen in vitro nor expressed IL-2, IL-4, or IFN-gamma. To find out whether they were able to recover, CD4 cells from tumor-bearing mice were purified and parked into a SCID mouse, where they could be observed only for about 3 weeks after the transfer (CD4 cells from control mice were abundant; Bogen et al., 1995). Thus, they were dead or dying. In tumors, many dendritic cells were found, and these might have had something to do with the deletion of transgenic L-Mut-specific CD4 cells. Furthermore, in TCR-tg SCID mice with no tumor, injection of purified protein (a monomeric L-Mut) also caused similar deletion. The loss of peripheral CD4 cells was correlated with the increase of the dose and started with about 50 µg/ml serum. This is a large dose, but tumor-bearing mice had even a five-times-higher concentration of the antigen (Bogen, 1996). The requirement for such a large dose might be that Ig molecules show greater resistance to processing and presentation. When thymuses of mice with big tumors (no metastases were found) were examined, a deletion of the antigen-specific TCR-tg thymocytes was observed (Bogen et al., submitted). An explanation for the central tolerance (deletion) could be either that soluble antigens have crossed into the thymus or that dendritic cells or other APCs have migrated from tumor, or another site where they might have captured the antigen, into the thymus, where they could have caused deletion (Bogen et al., 1992). So, this is an example where we do not have B cells, and the antigen causes peripheral and central tolerance. On the other hand, we have evidence that B cells alone can also cause tolerance with the same antigen. For this, we have transfected a lymphoma line (A20) with a specially modified antigen (L-Mut) that could not be secreted but only presented on the cell surface by the MHC class II molecules (F67) (Weiss and Bogen, 1991). Such F67 B-lymphoma cells kill TCR-tg and TCR/SCIDs when injected in dosages of about 300,000 cells per mouse up to 2.4 million per mouse (these concentrations parallel the ones observed for the nontransgenic control mice) (Lauritzsen et al., 1994; ongoing experiments). My explanation would be that F67 cells can deliver signal[1] but not signal[2] (i.e., they were B7 negative), in which case the naive T cells would die. In conclusion, the F (T-cell) epitope presented on B cells does not cause activation but tolerance in the absence of signal[2]. The soluble SF-like protein can perhaps cause peripheral tolerance by the dendritic-cell mechanism that I described previously (Day 3, #22) and can cause central tolerance, penetrating the blood-thymic barrier either as a molecule or carried via dendritic or other APC. Following this example, one can assume that in the B-cell-depleted
mice an antigen could tolerize T cells via dendritic cells in the periphery
as well as in the thymus as a molecule or via migrating cells. This is
also sufficient to explain the findings of Coutinho and Bandeira mentioned
in point 9 of Day 3, message 10.
Rod Langman - 08:13pm May 16, 1997 (#9 of 14) I may have given the wrong impression (#6)
with the strongly worded banishment of low zone tolerance. The phenomenon
stands, but as suppression or some other induced response, including switches
in the class of response that give the appearance of tolerance when assayed
only in narrow window of responses.
William O. Weigle - 9:24pm May 16, 1997 (#10 of 14) I agree with Antonio (#7) that the thymus may influence the induction of peripheral tolerance. However, a soluble tolerant state can readily be induced to monomeric HGG (Gahring and Weigle, 1990) in the absence of regulatory events. Thymectomy of adult mice 30 days prior to injection of tolerogen has no effect on the induction of peripheral tolerance. Thus, it is safe to state that a tolerant state that mimics self can be induced in the absence of the thymus. Zlatko's point (#8) that B
cells can induce tolerance is well taken. This point was also made by Eynon
and Parker (1992), who induced tolerance to Ig epitopes by targeting
them to B cells. Also, B cells present tolerogen extremely well to antigen-specific
T-cell clones (Gilbert and Weigle, 1994).
Doug Green - 12:03am May 17, 1997 (#11 of 14) Sorry I've been offline for a while - other pressing matters, I'm afraid. First, a couple of points of clarification. In the simple models that I've been outlining, it may seem that I focus predominantly on negative selection and dendritic cell function to explain all immune phenomena. It may also seem that there's little or no difference between what I'm talking about and the "danger" idea (except that I tend to substitute "damage," which is easier to understand but does not include important stimuli such as endotoxin, which I certainly do include in the models). I am also in very good agreement with those who suggest that the initiating signals for specific immune responses have been hardwired into the system by evolution (release of cytokines upon damage, role of mast cells, response to endotoxin, complement activation, etc.) for the reasons you've outlined. Perhaps most importantly, I'm in agreement with the idea that the immune system responds, initially, to changes in the host rather than the presence of something "foreign." I do differ to some respect in my emphasis on local responses, because I think that this is a key feature of the immune system. Local sites of tissue damage, infection, and inflammation produce profound local changes, both at the site and in the draining lymphatic tissue. Specific immune responses, if they are engaged, can function to amplify the response while restricting it to appropriate sites (either the original site or any other inflammatory site that contains the original peptides to which the specific response was directed). Autoimmunity remains a major challenge to our understanding of the immune system. We do have some insights based on animal models and analysis of human diseases, and we can define several different types of autoimmune responses. How well does our current knowledge explain such diseases? (I mean explain in the sense that we might be able to do something to prevent them). Clearly, one important aspect involves an issue I think most of us agree on - whether a particular protein present in our bodies is continuously available for recognition by the system. If it isn't, then its sudden appearance might trigger a response if the initiating signals for an immune response are present. We've been over some of the settings in which a response may or may not occur. There has been an interesting shift, however, in our thinking about autoimmunity, which probably shows the impact of the sorts of ideas we've been discussing. For most of the history of thought on autoimmunity (which has not been very long - since the late fortiess or early fifties), we've thought about something going wrong with the lymphocytes. Because clonal selection removes potentially autoreactive cells, the logic goes, autoimmunity must involve some problem with clonal deletion. Similarly, if we think about one- or two-signal ideas, again, most of the consideration is that something has gone wrong with the lymphocyte (either in terms of an intracellular defect or in terms of the signals the lymphocyte is receiving). The immune system discriminates self from nonself - when it doesn't, there must be something wrong with the immune system. Right? Many studies on receptor repertoires and cytokine profiles among autoreactive lymphocytes (valid studies that have been informative in some cases, less so in others) are based on this idea. Well, lately, we're starting to think that it's not all in the lymphocytes. After all, many autoimmune diseases are tissue specific, and it's difficult to entirely blame defects on general tolerance mechanisms. But because the tissue is beginning to play a more important role in the initiation and maintenance of immunity, we may be starting to think in terms of more active roles of the tissue. In some ways, it's a return to the old "self marking" ideas, but general, nonspecific self markers, not MHC molecules (for example). In addition, studies on the regulation of autodestructive responses are bringing home the point that the response has to be initiated. A useful example of a triggered autodestructive response (along with something like self marking) concerns the phenomenon of immunologic privilege. Some tissues are so sensitive to the destructive effects of an immune response that regulatory mechanisms appear to be in place to prevent them at those sites. The eye is a classic example. When a virus is placed into the anterior chamber of the eye, a massive inflammatory response (involving neutrophils and T cells) ensues, but there is no long-term effect, because all these cells are dead in a few hours. This is because the exposed surfaces of the eye express a molecule, Fas ligand (or CD95L), which triggers apoptosis in the infiltrating cells (Griffith et al., 1995). In an animal in which FasL is defective, introduction of virus produces the same inflammatory response, but the cells don't die. The result after a few days is that the eye is destroyed by the ongoing response. These defective animals have normal eyes; the destruction only occurs if the response is triggered, by either virus or other triggers of inflammation. Once an immune response causes extensive damage to a tissue, there is now the risk that the effect will escalate to true autoimmunity, as normally hidden proteins are released. This relates to the phenomenon of "epitope spreading," where an autoimmune response to a specific peptide grows to include other peptides and molecules released from the damaged tissue. From this point of view, it's hard to understand why we don't always make autoimmune responses (assuming that there are lots of "hidden" proteins in the body, which is not unlikely). One idea that emerges from this is the testable hypothesis that many tissues of the body respond to the extent of an ongoing inflammatory (or immune) response by inducibly expressing FasL, thus defending themselves from extensive damage. Of course, if this were generally true, we would never reject tissues (but I think that this may be a mechanism for eventual acceptance of some kinds of grafts, such as liver). Ideally, the system should respond to some signal produced by responding lymphocytes, as a measure of how extensive a response might be. Again, this is testable. I'm not saying that this is how tissues protect themselves from damage; I'm merely suggesting that it is one of many testable ideas. The main idea is that the tissues, not only the immune system, function in their own protection from the system. If tissues have the capacity to regulate immune responses, then loss of controls on these tissues will open up the possibility of the sort of escalating autoimmune response I mentioned above. This relates to the sort of premature "aging" that Zlatko mentioned in an earlier discussion. Signals that maintain (or induce) expression of regulatory molecules such as FasL may decline, and thus, if there is a strong inflammatory response in that tissue (but not necessarily others), the effect will be not only to sustain damage but also to trigger autoimmunity. If these signals have common sources in different tissues (e.g., neuroendocrine input), then some general defects will affect a variety of tissues, increasing susceptibility to onset of autoimmune disease in those tissues. Which diseases then occur in that individual will depend on which tissues generate initiating signals through infection, damage, etc. As with most of the things I've talked about, there's
nothing very new in what I describe (from the point of view of new theories).
The ideas themselves can only be tested in terms of specific molecules
(e.g., are Fas/FasL interactions required to prevent organ-specific autoimmunity
in any case?) - if the answer is no, then we move on to other molecules
(e.g., B7/CTLA4 interactions) that might fit the bill, until we run out.
Practically speaking, this is how we learn how the system works. If we
can't find an answer, we either drop the idea or wait for some new possibilities.
Rod Langman - 09:45pm May 18, 1997 (#12 of 14) I understand why Doug is tempted to the logical conclusion of self markerism (#11), but this cannot be the critical step in exercising the self-nonself discrimination in the immune system. Consider the following simplified experiment: Two animals, AA and BB, reject skin from each other, but their offspring AB does not reject itself. After grafting AB skin to AA or BB, there are two basic outcomes, depending on whether a dominant self marker or a somatically selected self-nonself discrimination is at work. (1) If the AB skin is accepted by AA and/or BB, then the presence of either A or B is sufficient to identify self, and the self-nonself discrimination is via a germ-line-selected self-marker system. (2) If the AA and the BB parents reject the AB skin, then either A or B is sufficient to cause rejection, and the self-nonself discrimination has the form of a negative somatic selection against antiself. This is the classic immune reaction and is a denial of a germ-line self-marker means of making the self-nonself discrimination. There is no question that being able to get our hands
on the levers of regulation in the immune system would be very helpful
(even profitable). However, my bet is that the self-nonself discrimination
lever is buried deep in embryology, and the better bets would be on manipulating
the class of the response for specific immunity and tweaking the effector
mechanisms, from complement to cytokines to Fas, etc. Deciding in which
bucket each phenomenon should be put is enormously helpful, and having
it in the wrong bucket makes the problem far worse than having no bucket
at all.
Doug Green - 10:09pm May 18, 1997 (#13 of 14) I completely agree with you, Rod (#12), and certainly I do not suggest that there are true "self markers" in the sense of discriminating between individuals. But the sort of marker I've suggested is simply something that can kill off lymphocytes that enter a site where they may inflict too much damage. FasL happens to be the one that works in the eye (and probably other privileged sites such as the testes). At least we can show that it's necessary for the phenomenon of privilege, and the observations are consistent with the role I've suggested. I think it's very likely that there are other molecules that do the same thing in other settings. So, rather than a self marker, I suggest that there may be a common sign that a tissue can express when the extent of the ensuing immune response is likely to cause life-threatening damage. It doesn't allow discrimination between self and nonself or between A and B; it allows discrimination between allowable and unacceptable damage by the immune system (before that damage actually occurs) during the process of normal immunologic defense. And because I like such ideas to be concrete, I'll propose that it's the FasL that does this (with some help from tumor necrosis factor). If it isn't, I'll look around for something else that might do it. But you're quite right - it isn't self marking in the
original sense at all, and it doesn't explain self-nonself discrimination
(or even the illusion of self-nonself).
William O. Weigle - 01:29am May 19, 1997 (#14 of 14) Although I admire Rod for defending (although lukewarmly)
a cherished dogma (#9), I would like to remind
him that he is no longer speaking of tolerance but of immune regulation.
Thus, if he wishes to retain the term "low zone," he should state it as
"high zone immune regulation," because the word "tolerance" no longer applies.
This phenomenon, which has only been seen in adults with one antigen (although
by an excellent investigator), has yet to be repeated. The main reason
that this term has survived is that theorists, by hook or by crook, have
been able to fit it into their theories.
|