MODELS OF IMMUNOLOGIC TOLERANCE
Day 5: Future Experiments to Compare the Models of Immunologic Tolerance
(Issue 11 ·  posted June 27, 1997  ·  11 messages)

Moderator Kenneth Schaffner - 08:44pm May 14, 1997

Can you think of any experiment that would distinguish between alternative tolerance models - that might clearly support one and not the others?

Rationale: 

The debate has identified some importantly different accounts of how tolerance (and the immune response) is induced and has clarified these differences. The different accounts say that there are different things at work. These different things include a danger signal, a primer (effector) T cell, a loss-of-integrity signal, an antigen-presenting cell (APC)-provoked cytokine cascade, and a temporally evolving network, among others. 

Without my necessarily asking you to commit to any one unitary model, can each of you propose one or more "crucial experiments" that would have different results depending on which model holds and that thus might confirm or disconfirm these different models? Relatedly, although it is difficult to prove a negative, how long is it reasonable to search for experimental support for a hypothetical signal or a process, given the likelihood that not all these things are likely to exist (or work the way the model proposes)? Are there other non-experimental (maybe theoretical?) factors that need to be considered in connection with the model that one accepts, and if so, what might they be? 

Finally, because this debate also has an educational purpose for the nonparticipants, I will ask each of you to summarize the current state of your views concerning immunologic tolerance. In particular, can you indicate whether your position (or your view of others' positions) has evolved during this debate and, if so, how and why? 

Zlatko Dembic - 8:51am Apr 12, 1997 (#1 of 11)

It is hard to disprove or prove some of the proposed hypotheses. As we all know, absence of evidence is not the same as evidence of absence - so, for example, inability to show suppressor cells on the clonal level has no proof value. The factors that might select between the integrity and danger models can be obtained, perhaps, by studying the details like the nature of the dendritic cell activation, the generation of memory, and the mechanism of costimulation (cross-talk, signal[2]). However, the general concept should allow some flexibility in adapting to certain details. This means that the mechanistic part of concepts might grow in the future with new results. 

I am sure that each hypothesis can give either simple or more complicated answers to experimental observations but, perhaps, never a simple answer to all questions. Yet there seems to be a consensus in establishing a scientifically based ground rule for selecting the best working hypothesis. We start with the simplest if not the best model and then go on from there. Clearly, according to this approach, the self-nonself discrimination principle is the simplest, and its generation by the great predecessors was a breakthrough and broke ranks with the instructive hypothesis. This can be compared with changing a viewpoint from a flat to a round Earth. This concept, in its elaborated form by Bretscher and Cohn or Langman and Cohn, still shines with its simplicity. Others, like Coutinho and Bandeira, have introduced a slightly more complicated viewpoint in proposing a (dominant) suppression of activation over self-nonself differences. Then a novel breakthrough appeared, through Janeway and Fuchs and Matzinger, in assigning an alarm signal for the initiation of the immune responses, increasing the level of complexity of the initial self-nonself discrimination. These views, in my opinion, have nothing to do with a self-nonself discrimination principle and can be a part of a novel one - let us call it a dialectic principle - to which these authors paved the way.

The dialectic principle of how the immune system functions is best described by the integrity hypothesis, which developed alongside the infectious nonself and danger models. The immune system operates on a selection force that protects and allows restoration of integrity. The immune system has evolved from the primordial integrity protection-like mechanism. This claims that an immunocyte does not have a choice between reacting to nonself (activation) and not reacting to self (tolerance) but rather competes for the activation to a pool of signals within its microenvironment. Sounding an alarm is just a small part of it. The fittest response to any disintegration is selected, and this is dictated by the local influence. Here, I see some points indicated by Bill and Doug that can be in agreement with this. Because integrity "takes care" of the contextual part of the specific response (including nonspecific immunity), it allows for cross-talk.

Tolerance can be divided into central (clonal deletion by receiving only signal[1]) and peripheral (deletion without activation or premature shutoff due to cross-talk). The immune system adheres to laws of selection in its operation: The response is being selected to make the individual fitter for maintaining its integrity. After an initial disruption of integrity, a state of heightened alertness toward similar disruptions, remembered as the ways to counteract past intruders, helps in this process. However, accessing memory includes different codes (different cross-talk). 

I am especially thankful to Rod and his challenge with the dreaded "thingies" (Day 2, #9). It proved to be an exercise of thought that allowed me to generate something that looks like the present-day MHC complex region with its receptors, starting with a single conceptual premise: integrity protection.

I believe that Antonio's exam (Day 3, #10) gives a hint of how much more needs to be done experimentally to narrow the choice for the best simple theory (not the simplest theory). Each hypothesis should be able to explain his points. I intend to e-mail him my answers and, perhaps, publish elsewhere, because the answers themselves would not bring us closer (anyone who wishes to have them can e-mail or fax me). Only the experiments based on these explanations, to prove certain points, would give us a substance that can be used in assessing the credibility of each hypothesis. With the accumulation of data and the explanations offered by various hypotheses, we might be able to choose to like the one that uses the lowest number of complicated explanations. So, with time, other concepts might be deselected, and the fittest theory would remain. 

One of Antonio's points in his exam message addresses disruption of integrity after a massive heart attack (point 8), which needs to be answered here. The self-reactive antiheart cells get deleted, i.e., tolerized, by constant passing through the heart receiving only signal[1]. The heart is a special kind of muscle, but it shares many properties with other muscles in the body - so many muscle-specific cells get deleted elsewhere, for example, in skeletal muscles. Thus there are few, if any, heart-specific cells left. During the heart attack with massive necrosis, an upregulation of signal[3] by the dendritic cells is seen. These cells then migrate to regional lymph nodes and offer signal[1] and signal[2], but there are almost no cells to respond to them. If, let us suppose, there are some residual antiheart reactivities around, then they might indeed be fatal. This could be the one of the reasons why the heart-attack patients are monitored for 5-6 weeks after the attack in intensive care units. It is not completely clear why this relatively long aftermath is such a critical period for the patient. One of the reasons could be that it might provide just enough time for the weak autoimmune attack to subside. If it does not, the patient might die even 5-6 weeks after the original attack. If the patient survives, he continues to delete his antiheart reactivities. And so, if he also continues to drink, smoke, and like women the way he did before, his heart might stop tolerating him (for physicochemical reasons). 

William O. Weigle - 3:32pm May 15, 1997 (#2 of 11)

I don't think that I can contribute too much in terms of convincing anyone what experiment they need to run to prove or disprove their theories. I have spent considerable time debating theories during my career and have yet to come to an agreement with any of the holders of the various theories as to experiments that we both would accept to disprove or prove the theory. Thus, I would like to present my idea of how tolerance is induced and maintained at the cellular and subcellular level with supporting data. This information may support or suggest some changes in the theories proposed throughout the past few days.

A major mechanism for the induction of tolerance occurs in early life through negative selection in the thymus. However, to maintain tolerance caused by transient leakage of sequestered antigens or changes in antigenic composition of the developing animal, there is obviously a need for peripheral tolerance. It is this latter tolerance that is induced in the periphery in CD4+ T-cell subsets that I wish to dwell on.

I would first like to give my idea of how antigen is presented to the T cell and how the T cell receives the second signal resulting in its activation. It is my belief that either B or professional antigen-presenting cells can deliver either immunogenic or tolerogenic signals. I disagree with Ephraim that dendritic cells are only created to deliver an immunogenic signal, while other APCs with less affinity in their interaction with the T cell can only present antigen in a tolerogenic fashion. It doesn't matter how strong or weak a signal is; if it is delivered effectively to the T cell in the absence of a second signal, tolerance will result. If not delivered effectively, nothing will happen. Thus, I would like to propose that the dendritic cell is like the cap (igniter) on a stick of dynamite, and it is this cell that initiates T-cell commitment in both immunogenic and tolerogenic events. The evidence suggests that the dendritic cells present antigen to the precursor T cell (Finkelman et al., 1996), and it is the B cells, via IL-4, IL-1, and possibly other factors, that are responsible for expansion of committed Th2 CD4+ T cells (Stockinger et al., 1996). Conversely, upon deliverance of the signal to the precursor T cell, macrophages via IL-12 and possibly other cytokines are responsible for the expansion of committed Th1 cells. The expansion of these cells is dependent upon activation of cytokines and thus activation of the lymphokine cascade. Tolerogenic signals most likely occur at the precursor T cell in the absence of subsequent activation of APCs and release of cytokines and subsequent regulation of the costimulatory factors. 

That tolerance occurs at the precursor T-cell level is suggested by (1) the silencing of both the Th1 and Th2 antigen-specific subsets upon successful tolerization (Romball and Weigle, 1993; Chu et al., 1995) and (2) the fact that the limiting dose for induction of tolerance is identical for the silencing of both subsets (Romball, unpublished observations). The support for a role for cytokines in the conversion of an otherwise tolerogenic signal to one that is immunogenic is the fact that generators of cytokines (e.g., endotoxin) and cytokines themselves (IL-1-beta, IL-1-alpha, and TNF-alpha) are all capable of interfering with the induction of tolerance (Weigle et al., 1987; Gahring and Weigle, 1990; Vella et al., 1995). Furthermore, in the human gamma globulin (HGG) mouse model, the exposure in vitro of APCs to aggregated HGG results in rapid degradation of HGG into small peptides and activation of the APCs to release cytokines (Levich et al., 1987). Conversely, the exposure of monomeric HGG to APCs under identical conditions does not result in degradation or APC activation. In the latter case, peptides needed for antigen (tolerogen) presentation result from normal in vivo catabolism. It is also suggested that the only difference in the mechanism of the induction of tolerance in newborn animals is a quantitative one, with different emphasis on negative selection in the thymus and induction in the periphery. 

Rod Langman - 8:00pm May 15, 1997 (#3 of 11)

I think it is good to be suspicious of all models because no doubt one day they will all be replaced. However, models can serve a useful purpose in helping to distinguish principles from details. For example, Bill correctly points to many weaknesses in the danger model (#2) as explained here by Ephraim, but it seems that the main difference between your two positions rests on what I see as detail, not principle. If I were to simply lump all kinds of flavors of ice cream together, it may taste bad, but it is still ice cream. Thus, I see the agents you describe as the sources of second signal (i.e., some signal that is not antigen) as different from those described by Ephraim, but they are different flavors of the thing I tried to call in the most general sense "alarm." I think you would not have the second signal expressed constitutively but dependent on some inductive event, and this is the step that is least well detailed in your views of the decision between tolerance and induction of immunity. 

Rod Langman - 7:00am May 16, 1997 (#4 of 11)

To distinguish between the "alarmists" and the associative antigen recognition (AAR) model, which differs only with respect to the origin of eTh, requires an assay of whether associative recognition is required for the induction of eTh or whether alarm in the near vicinity is sufficient. To my mind, the two-stage breaking of tolerance with a cross-reactive antigen comes awfully close. The delayed appearance of antibodies to epitopes unique to the tolerogen (i.e., not found on the cross-reacting antigen) requires eTh specific for the unique antigens to associatively recognize iB cells also specific for the unique epitopes; these eTh could be generated either by the same process that generated eTh anti-nontolerogen or by specific eTh that recognize epitopes common to the nontolerogen and the tolerogen. To my knowledge, this has not been a subject of intense investigation, and it might bear further analysis to make sure my interpretations are correct. 

To distinguish between the networked dominant-suppression model and the AAR model requires me to better understand how the suppressive and immune networks are linked within themselves and between the two. It would help me a lot to have an explanation of the general observation that the hapten and the carrier have to be on the same molecule for a humoral response and on the same cell (using H-Y and Qa1 as the carrier and hapten) for a cytotoxic response - mixtures of the two components do not have the same result as physically linked pairs. 

Ephraim Fuchs - 12:44pm May 16, 1997 (#5 of 11)

It has been said that although it is possible to disprove a theory, it is never possible to prove it. As the official advocate of the danger model in this discussion, I will therefore propose experiments that would disprove the danger model.

The two key features of the danger model are that (1) only "professional" APCs, such as dendritic cells, can initiate immune responses among naive T cells and (2) the "professional" APC must be activated by an exogenous or endogenous danger signal to become immunogenic for naive T cells.

Thus, one experiment that would prove the danger model wrong would show that either a B cell or a parenchymal (nonhemopoietic) cell could activate a naive T cell. Many may say that this has been done already, but I must stress the word "naive" in the previous sentence. There are indeed experiments that show that a purified population of CD4+ T cells with putative markers of the naive phenotype (such as high expression of the CD45R molecule, low expression of the CD44 molecule, etc.) do respond to B-cell-presented antigens, but we would claim that these markers do not reliably distinguish naive from previously activated T cells. There are other studies showing that purified T cells from unprimed T-cell receptor transgenic mice are activated by B cells, but such T cells may actually express two distinct types of receptors on their surface, one pairing the transgenic alpha and beta chains and one in which transgenic beta chains are paired with endogenous alpha chains. The TcR combining the endogenous alpha with the transgenic beta could be specific for environmental antigens, providing the stimulus for this bispecific T cell to become a memory cell. The bispecific cell could then respond in vitro to the antigen for which the transgenic pair is specific, even when presented by a B cell. 

In response to Bill Weigle (#2), I would say that only dendritic cells can deliver the immunogenic signal, not that dendritic cells only deliver immunogenic signals. Also, I would agree totally that B cells can activate T cells, but not naive T cells. 

In response to Rod's question (#4) about why an immune response to Qa-1 requires that H-Y and Qa-1 have to be on the same cell, Polly Matzinger and I would say that in this case there are no Qa-1 specific CD4+ T cells in the unprimed animal, that help for Qa-1-specific CD8+ T cells is delivered by the CD4+ T cell specific for H-Y, and that this CD4+ T cell delivers "help" by inducing costimulatory signals on the APC. This might be achieved in the case of B-cell APCs by the CD4+ T cell delivering a signal through CD40 that upregulates the expression of B7-2 on the B cell. Obviously, if H-Y and Qa-1 are not on the same cell, the appropriate costimulator for the CD8+ cell will be expressed on the H-Y+ cell and not the Qa-1+ cell. 

Also for Rod: I have never been able to explain the result I obtained (Fuchs and Matzinger, 1992) by the associative recognition model. The result is that an adult female mouse that receives syngeneic male B cells is rendered tolerant to H-Y, but a littermate female that receives syngeneic male dendritic cells is primed to make an anti-H-Y CTL response. Wouldn't there be the same amount of effector T help in both situations? 

Rod Langman - 1:33pm May 16, 1997 (#6 of 11)

In response to Ephraim (#5): It's tough to devise a test that depends on "naive" cell populations without some objective assay of this property. It is not too helpful to argue, for example, that the behavior of cells as predicted by the model is our only assay of naivete. Is there a clear definition we all can use experimentally before going further to test the model? 

I don't have the papers handy, but as I recall, once the H-Y-Qa1 cells have been used to induce anti-Qa1, then H-Y is no longer needed for future boosting in vivo. In any case, can the H-Y-Qa1-primed mice then be used to raise Qa1-Mx response, where Mx is another help-dependent antigen like Qa1, or is Qa1 the only one of its kind? 

Rod Langman - 4:01pm May 17, 1997 (#7 of 11)

Response part 2 to Ephraim (#5): The use of a few dentritic cells or a gizillion B cells (I can't recall the numbers and I don't have my stack of papers at home) is reminiscent of using few erythrocytes to immunize for suppression and many erythrocytes to immunize for antibodies. In your case, can you be certain that the B cells did not induce antibodies and thereby fix the class of response in a non-CTL form, whereas the dentritic cells were unable to switch the class out of a cell-mediated immunity response? I recall a recent paper of Matzinger et al. where numbers of cells were titrated, and there were always many fewer dendritic cells than B cells. I know that the dendritic cells are hard to come by, but using fewer B cells would be reasonable. The shorter answer to your question is that the amount of eTh is only part of the determination of the response, as the serum globulin work of Bill Weigle discussed here in detail shows. So far as I'm aware, precious little is known of the primary process that determines the class of the response. 

As an aside, I would like to think that there is more to immune regulation than the two signals that allow i-state cells to be directed to tolerance or immunity. Host- and parasite-derived factors must play a huge role in modulating the progress of cells after the self-nonself decision has been taken and they head toward becoming effectors. And there are more modulating effects that influence the non-immune effector mechanism that actually perform the ridding function. The self-nonself discrimination is just the first small step in the life of a T or B cell. 

Doug Green - 9:59pm May 17, 1997 (#8 of 11)

Our final question in this debate is probably the toughest, and yet what could be fairer than to ask for the experiments that could possibly distinguish between conflicting theories? Isn't this what we do in science? 

Well, yes and no. Certainly we test hypotheses (e.g., IL-2 is required for T-cell proliferation), but we don't generally do this with theories - not in the same sense. In immunology, for example, no experiments have rigorously tested the clonal selection theory or the network theory. Before this statement raises "alarms," I don't mean that the theories have not generated lots of testable ideas that have been explored to great effect. It's just that they haven't been tested in the same sense that we test a hypothesis. 

This is partly because theories are so slippery. As soon as we propose a test, and that test fails, the proponents of the theory will usually retort that it wasn't a definitive test. Not because it wasn't properly designed, but because experiments rely on concrete entities and theories do rather well with abstract thingies. If we propose to test the theory that costimulation is an integral part of tolerance, we could examine a mouse in which a costimulatory molecule is absent. If this doesn't have the predicted effect, we do not scrap the theory; we simply look for another costimulatory molecule. And we'd be right to do so. 

So what good are these theories if they're not testable? Several of us have stated here - and I think it's worth repeating - that theories have value if they generate interesting and testable hypotheses. When the well of new hypotheses dries up, or when they are no longer interesting, the theory may either be replaced or become dogma. For some of us, the theory itself holds a certain beauty in its own right; for others of a more practical bent, the theory is useful because it leads us to new findings. 

Which is why I've worked to try to convert a theory, rife with constructs and models, into a story with real-life players. It's kind of a middle ground. It's a story of how the immune system might work, and because it's built of real players, all the elements are testable. But just as with any theory, if any testable idea fails, we don't scrap the story - we just swap players. But the value lies in the questions that are asked. So what are the questions? Here are a few. 

I've stressed the role of cellular trauma in initiating immune responses. What are the components released from a necrotic cell that actually recruit an inflammatory response? (There have been a few suggestions, but surprisingly little work has been done to identify this well-known response to cellular injury). How does cellular stress (e.g., of epithelial cells) trigger cytokine release? How important are mast cells in initiating the vascular changes that occur early in inflammation, and how critical are these for initiating an immune response? (We can envision knocking out genes that are critical for mast-cell differentiation or function.) 

And of course I've noted that trauma isn't always completely necessary. The system is clearly set up to respond to molecules that are commonly associated with parasites - one example is the activation of complement via the alternate pathway. How many such signals are hardwired into the system? How critical are such signals (under what circumstances will tissue damage be sufficient to give us responses in the absence of exogenous adjuvants)? 

We haven't talked much about CD8 T cells, but they pose a special problem in the story I've told. If a virus infects a tissue, where is the viral peptide (plus class I MHC) presented to CD8 T cells? (That is, if this occurs in a lymph node, which is what I suggest, how does the ligand get there? Is it that infected cells move to the node, or do we need an APC capable of the exogenous pathway of processing peptide for class I MHC?) 

With regard to the developmental process by which lymphocytes die upon contact with their ligand, I've suggested that it would solve some problems if this process persisted for a while after the cells actually left the primary lymphoid organ. Is this true? This could be tested in cells that mature in fetal thymus organ cultures, for example - how long do the developing cells retain this characteristic? 

There are many more questions, but I think I've made my point. Clearly, none of these questions discriminate between theories. The theory has simply pointed the way to some interesting (I hope!) questions. Other theories will point to completely different questions. And here's the thing - even if the theories are incompatible, the answers to the questions they generate may not necessarily be. That's what makes this so much fun. 

So my suggestion? Don't try to test one theory over another. Pick one you like, study it, and see what new questions it makes you ask. Then test them. If the answers are satisfying, ask some more. The theory will move to accommodate nearly any answer you get. Most of them do, anyway. (I know that my interpretation of what a theory is for will be a bit inflammatory - but I stand by it in this particular case. Theories in biology are a special problem, and technically, I don't think that any of the things we've been discussing are really theories in the formal sense - in the sense of a theory in physics. There are those who are trying to develop true theories in biology. Us? I contend that we tell stories, like the nature of self-nonself discrimination, and the stories have value for the reasons I've said. It's perfectly fine to call them theories.)

Rod Langman - 12:53am May 18, 1997 (#9 of 11)

Response part 3 to Ephraim (#5): From figure 2 of your 1992 Science paper (Fuchs and Matzinger, 1992), the numbers were 10 million B cells or 300,000 dendritic cells intravenously. These two cell types are very different; B cells normally circulate in blood and lymph, whereas the dendritic cells were selected as adherent cells from spleen. It would not be surprising to have these very different forms of antigen treated differently, rather like deaggregated HGG, which circulates freely in serum, and aggregated HGG, which is taken out of the circulation very quickly. 

Rod Langman - 1:09am May 18, 1997 (#10 of 11)

I wonder if Doug really does determine the direction of his experiments with a divining rod in search of water (#8). To be brutal, I doubt that even 10% of what is published makes a difference even two years hence. There is a vast effort to increase the data base, but little ends up in the knowledge base because experiments are arranged to support the flavor of the month, rather than testing well-formulated theories. The waste level is staggering. The complaint that there are no well-formulated theories cannot be due to more than a lack of motivation. However, I suspect that Doug's sentiments are not very different from those of the typical immunologist, and this polite scoffing at theories is not leading to a better way of making real progress. 

Rod Langman - 1:16am May 18, 1997 (#11 of 11)

One final comment to record my pleasure and appreciation to all of the members of this debate team, who made this one of the best meetings of the minds I've experienced. I think Ken did an excellent job managing the discussion; merely controlling the rate of explosion was a major achievement.