MODELS OF IMMUNOLOGIC TOLERANCE
Day 4: Do New Models of Immunologic Tolerance Have Novel Experimental or Clinical Implications?
(Issue 11 ·  posted June 27, 1997 ·  14 messages)

Moderator Kenneth Schaffner - 4:42pm May 13, 1997

Do new models of immunologic tolerance have novel experimental or clinical implications, including but not restricted to vaccine development, autoimmune diseases, cancer, or organ transplantation? Relatedly, does the model(s) you prefer provide new insights into classical experimental findings in immunology?

Rationale:

Several participants have suggested that we look at the experimental bases and implications of the various contrasting models (at the cellular and molecular levels) that have been presented in this debate. Some have urged that we should, in particular, consider the areas of autoimmune diseases, cancer, infectious disease, and organ transplantation (allograft acceptance). 

We might best do this in terms of general empirical findings - for example, addressing whether there are experimentally supported differences in tolerance in neonates vs. adults, and to what extent this varies from one species to another. A closely related area might be how tolerance is broken (in autoimmune diseases) or how it might be artificially induced (in organ transplantation). If the above clinically related areas are perceived by some to be too complex to likely yield clear experimental results, that issue too might be discussed.

Zlatko Dembic - 7:03pm May 13, 1997 (#1 of 14)

There are two basic ways in which the tolerance of specific immunocytes can be achieved:

(1) The cells die if they receive command[1] alone (I use the command instead of the signal because there is no way to avoid it) and (2) after they have received command[1] and signal[2], they downregulate signal[2], for example, due to possible modulatory influence of the microenvironment (including cytokines). (See summary of possible outcomes of the initiation of the immune response in Day 3, message 3.) With this, I will first try to explain the theory and then combine it with potential experimental and clinical potentials. 

Peripheral tolerance: 

The repertoire of naive T cells that exit from the thymus might not be deleted for all self-antigen reactivities that are available in the organism, because, for one reason or another, the self antigen simply was not around in the thymus. These cells would be tolerized by encountering self peptides on a variety of self tissues, thus receiving signal[1] only, and, as in the danger model, die. Also, if by chance command[1] is connected with signal[3], but they are separated in space, a possibility exists that a novel antigenic peptide would be integrated into the macrophages in the novel site for presentation. Maybe even some dendritic cells would pick up this antigen that is away from the influence of the signal[3], thus providing a modulated command[1M] (novel epitope) to the cells. This would cause deletion of peripheral cells. Alternatively, if a cell received command[1M] and signal[2] (a result of disruption of integrity, i.e., signal[3]) and then immediately downregulated the signal[2] (for T cells by switching, for example, CD28 with CTLA4 [CD152]), the result would be death, for example, after 3 weeks (tumor-antigen primed, after subcutaneous injection of the tumor, and "tolerized" CD4 cells die after their transfer into SCID recipients in that time; Bogen, 1996), during which time the cells might appear refractory to antigenic stimuli, and for some, then, anergic (Perez et al., 1997). The reasons for the immediate switching off of signal[2] might be, for example, that a tumor has adapted itself to "mimic" the signals of integrity; translated into the reality, this could look for one particular case as though the tumor, by secreting a kind of cytokine (IL-10 for example), might in the absence of other class-directing cytokines like IL-4 cause switching of CD28 to CD152 and premature shutoff of the activation-initiation part of the immune response to tumor. Thus, with only command[1M] remaining, the cell dies.

Central tolerance: 

Central tolerance of T and B cells is an interesting process to be considered. Although B-cell tolerance can be explained by a negative selection of the A type of tolerance (i.e., clonal deletion to all self antigens if encountered on other cells' surfaces during the development in the bone marrow; thus receiving only signal[1]), negative selection of thymocytes cannot be easily explained by either the A or B type of tolerance. The problem here is the so-called positive-selection phenomenon, which is a selection of a population of double-positive (CD4+CD8+) thymocytes that bear a T-cell receptor (TCR) with neither too high nor too low affinity to counteract the endogenously derived peptide-MHC combination on thymic epithelial cells. If thymocytes would have too low avidity for recognizing self peptides on MHC class I or II molecules, they would die by programmed cell death (neglect). Many believe that recognition of self peptide-MHC complexes rescues double-positive thymocytes from such apoptotic death. That a clonal deletion exists in the thymus is a fact documented and mentioned earlier in the discussion. However, if we apply the A or B type of tolerance, then how is the thymocyte/epithelial cell affinity/avidity being measured? I propose that, due to special characteristics of the thymic microenvironment, which secretes small amount of glucocorticoids required for the development of T cells (King, 1995), all cells that receive command[1] and signal[2], which would normally lead to activation in periphery, die because of these glucocorticoids produced in their vicinity (perhaps a B type of tolerance) in a way similar to depletion of thymocytes caused by injection of steroids.

Now, how are the positively selected thymocytes spared? They are because they receive only signal[2] and a "fake" command[1]. The command is fake because, by recognizing partial agonist-like peptide-MHC complex, TCR does not provide a full signal[1] yet (as Ephraim pointed out earlier (Day 3, #8), this does not include seeing the antigenic peptide-MHC complex), but because coreceptors are engaged, a part of the signal[1] is present. If we increase the level of expression of a coreceptor on thymocytes carrying a to-be-neglected TCR, they would end up being positively selected, thus overcoming the need of TCR specificity for positive selection (my own experiments). Thus, this suffices for the rescue from neglect. And it is also not a complete activation that would kill them. Whether signal[2] gets downregulated or not becomes perhaps irrelevant. Neglect implies a complete misfit between the thymocyte and thymic epithelial cell (only perhaps signal[2] would be available but perhaps also unrecognized). It follows that any antigen or peptide derived from it can cause deletion in the thymus if it can penetrate inside. Because dendritic cells can perhaps pick such in the periphery and move and bring them into the thymus, thymocytes (of single-positive lineages) can be additionally negatively selected (clonally deleted) within the medulla. Alternatively, soluble antigen itself can, perhaps, penetrate through the microcirculation barrier and be presented by the local dendritic and epithelial cells.

Experimental and clinical implications: 

The malignant tumors would appear because they mimic the integrity signals. They would be selected for by the ongoing immune response toward them. The first malignant cell might arise, perhaps, after a benign tumor (or slow-growing cells) "learned" how to mimic signals of integrity (or how to downregulate the signal[3]), thus fooling the immune response into tolerance. Here, immunotherapy or vaccination with tumor-specific antigens or (as P. Matzinger suggested) by necrotically destroyed tumor cells to prime dendritic cells would make it possible for the immune system to fight off the tumor, provided the same individual still has some antitumor cells left (for example, if the peripheral or central tolerance mechanisms did not destroy all available cells to react against the tumor). The rationale for this is that although tumors might rarely express a novel antigen, they are formed, perhaps, by accumulating mutations in about 5 to 10 genes that would make them malignant cells; thus, such mutated forms of antigens (antigenic peptides) might serve as targets. 

Autoimmune disease would be a state that causes an aberrant expression of signal[2] within the affected tissue or organ or an organism. Because there is no signal[3] to modulate the signal[1], activation with destructive response targeted towards individuals' own tissue ensues, provided the same individual still has some reactive cells that were not inactivated (type A tolerance) before the aberrance started. Here, downregulators of costimulation would be a target for therapy. Glucocorticoids, for example, as they kill the activated T cells, work indiscriminately, with a lot of adverse effects. However, antagonists of costimulation would be also indiscriminate and would decrease the potential repertoire but might have less adverse effects (and can be used for achieving transplantation tolerance). Upregulators of CD152 might be also welcome, and these could be tested experimentally already.

For preventing graft rejection and maintenance of transplantation tolerance, the simple rules of achieving type A and/or type B tolerance can be applied. As in the danger model, blockers of costimulation are welcome, but upregulators of CD152 could be tried also. To achieve a state in between the resting and activated types, as mentioned with partial agonists in the thymus - the fake signal[1] - one can try to manipulate signal[1] using promiscuous partial agonists that would bind to known HLA differences between the donor and recipient. Whether the effect (a tolerance to donor tissues) would be permanent is doubtful, and they might generate a lifetime-needed drug therapy, like the already existing ones (cyclosporine, FK506).

Ephraim Fuchs - 12:00am May 14, 1997 (#2 of 14)

As the official advocate of the danger model of immunology, I will try to address its experimental and clinical implications as well as the insights into previous immunologic findings. These are exceedingly rich areas and, although I will provide a brief outline here, I will try to come back to these questions in a subsequent post.

As a clinical oncologist, I will focus on the cancer problem (Fuchs and Matzinger, 1996). 

The main features of the danger model relevant to cancer are:

(1) Only dendritic cells can activate naive T cells; all other cell types presenting antigen to naive T cells turn them off.

(2) Dendritic cells must be activated by nonphysiological tissue distress or death to become immunogenic antigen-presenting cells (APCs) for naive T cells.

The implications for cancer are several: 

(1) There should be no occurrences of tumors of dendritic cells, unless of course they mutate to become noncostimulatory.

(2) Tumor cells derived from parenchymal tissues (lung, colon, ovary, pancreas, prostate, breast, etc.) are tolerogenic APCs for both resting naive and memory T cells.

Thus, if you are to have any hope of generating an immune response to a tumor-specific antigen:

(3) Get the antigen onto an activated dendritic cell. This can be achieved by (a) pulsing dendritic cells with tumor lysates or the antigen itself, if known; (b) transfecting tumor cells with genes for molecules that enhance dendritic cell maturation and/or chemotaxis, such as GM-CSF (Dranoff et al., 1993) or flt-3 ligand; (c) inject a source of danger signals, such as bacteria (Coley, 1894), pus, bacille Calmette-Guerin, etc., into the tumor to alert dendritic cells; or (d) trying to turn the tumor cell into a dendritic cell. People have tried to do this by transfection with costimulatory molecules, such as B7-1 (Chen et al., 1992). The problem might be that activation of naive T cells may take place only in lymph nodes, so that it might not help to have a tumor in a tissue expressing B7. 

(4) Because tumor cells are tolerogenic APCs for even resting memory T cells, it is not sufficient to initiate an immune response against a tumor. One must constantly provide danger signals to ensure that hematopoietic APCs are presenting tumor antigens in an immunogenic fashion. 

Now, what insights does the danger model provide for previous experimental findings? 

(1) Neonatal tolerance (Billingham et al., 1953). We have proposed previously that neonates are easily tolerized because they have very small numbers of T cells, which are all in the naive state. Thus, when one injects allogeneic spleen cells, which contain >90% tolerogenic APCs for naive T cells (our model predicts that anything other than an activated dendritic cell is tolerogenic), it is strongly likely that all naive T cells will encounter antigen on a tolerogenic APC, and tolerance is the resulting phenotype (Fuchs and Matzinger, 1992). Thus, the only difference between the neonatal and adult immune systems are the number of T cells and the proportion of memory T cells. We then predicted that purified dendritic cells should be immunogenic even for so-called weak antigens and, indeed, male dendritic cells injected into neonatal syngeneic females induce priming to the male-specific antigen, H-Y (Ridge, 1996). 

(2) High and low zone tolerance (Mitchison, 1964). At low antigen concentrations, antigen is captured preferentially by antigen-specific B cells, and tolerance results. At medium antigen concentrations, antigen is presented by both antigen-specific B cells and dendritic cells, and immunization results. At high antigen concentrations, the APCs expand to include all B cells, and tolerance again is likely, because B cells greatly outnumber dendritic cells. 

(3) The requirement for adjuvant in inducing immune responses to "innocuous" antigens. Adjuvant is the source or inducer of danger signals.

There are many more issues to be covered, and so little space.

William O. Weigle - 2:58am May 15, 1997 (#3 of 14)

If we accept that there are two models of tolerance, one occurring in the thymus and the other occurring in the periphery, then I believe the difference in tolerance during early life and adulthood is only quantitative. During embryonic life, tolerance to self is entirely carried out in the thymus by native selection (Sprent, 1993). The possible constant changing of antigen during adult development requires a mechanism of peripheral tolerance (discussed by Zlatko in message 1). Thus, the role of peripheral tolerance dramatically increases following neonatal life and, as we all know, the thymus atrophies and becomes less effective in discriminating against nonself. In the aged adult, the absence from any significant input of the thymus and the age-associated limitation of T cells capable of responding to new antigens results in a qualitative and quantitative change in immune reactivity accompanied by autoimmune reactivity. Because of the dramatic shift in T-cell subset usage and the cytokine profile in the aged (Ernst et al., 1993), progressive autoimmune disease does not occur.

The original postulate by Medawar and coworkers, who suggested that tolerance could be more readily induced in neonates with allogeneic cells than in adults, has recently been questioned, based on the ability to immunize neonatal mice with viruses and adjuvants (Forsthuber et al., 1996; Sarzotti et al., 1996). However, it has been known for approximately 40 years that the environment and not the lymphocytes in neonatal animals is responsible for their immune deficiency; through the years, a number of investigators have been able to overcome this deficiency by supplementing neonatal animals with adult APCs. Studies using serum protein antigens demonstrated that tolerance is certainly more permissive in the neonate than in the adult (Dietrich and Weigle, 1963). However, it should also be pointed out that permissiveness of neonates to tolerance induction is limited. Such permissiveness does not apply to most environmental pathogens. This point was discussed thoroughly yesterday. Thus, it can be concluded that despite any deficiency in neonates, appropriate immune responses can be elicited by viruses, bacteria, etc., that are capable of being processed by APCs, resulting in activation of the latter and subsequent release of cytokine. This scenario is the most reasonable one because the induction of tolerance to every antigen in the neonate environment would be suicidal.

It is well established that the immune system is capable of maintaining immune responses to self components and that such responses can be accompanied by disease. Self constituents normally do not stimulate the immune response, but occasionally the immune system turns on its host environment in such an aggressive manner as to cause disease. Significantly, the cellular and subcellular events leading to and regulating this destructive autoimmune reactivity are the same as those involved in beneficial immune responses to foreign antigens. All the elements in the repertoire of immune defense (antibody of various subclasses, antibody-dependent cell cytotoxicity, delayed-type hypersensitivity, and T-cell lympholysis) also participate in autoimmunity. However, before we can understand the cellular parameters involved in autoimmune disease, one must first appreciate the conditions favoring the recognition of self as foreign. The various mechanisms that may be responsible for the loss of tolerance to self antigen can be divided into three categories: 

(1) Abnormalities may occur in regulatory mechanisms that control the immune responses in general. For example, genetic differences in immune regulation may permit self recognition to proceed to an autoimmune response and then to disease. 

(2) A component of self that was once sequestered may become exposed and present in an antigenic form to the immune system. 

(3) A normal tolerated self component may for some reason circumvent the prevailing regulatory mechanism and activate one or more arms of a normal immune system. Such conditions may result from polyclonal activation of B lymphocytes by viral or microbial infections, alterations of self components, or contact with antigens with which self cross-reacts and thus promote bypassing of tolerance at the T-cell level, permitting activation of self competent (nontolerized) B cells. Alterations of self could result from a genetic error in protein synthesis or as a consequence of infection or other trauma.

Therefore, the cause of autoimmune phenomena may range from a single condition to any combination of the above categories, as may be the case in some complex autoimmune diseases. All the above scenarios involve antigen reacting with the specific lymphocyte accompanied by activation by APCs in the immediate microenvironment. As with the response to foreign antigens, there is a release of the cytokines necessary for the second signal.

As an example, I would like to concentrate on the circumventions of tolerance by the antigens sharing epitopes with self. It is well documented that the induction of tolerance in B cells requires hundred- to thousandfold higher concentrations of antigen than the induction of tolerance in T cells. This difference in dose requirements has implications for antibody-mediated autoimmunity in that it suggests that the host may be tolerant in both T and B cells to body constituents in high concentrations but in only the T cells to the body constituents in lower concentrations (Weigle, 1980). For more completely sequestered antigens, neither T nor B cells may be tolerant. When the B cell is competent in the presence of tolerant T cells, the B cells can be rendered responsive by injection of antigens in the presence of lipopolysaccharide or a related (cross-reacting or altered) antigen that contains epitopes that are shared with the tolerogen and epitopes that are different. Such cross-reacting antigens can be natural antigens (even pathogens) or altered self proteins. Thus, CD4+ T cells, activated by a nonrelated epitope, give help to competent B cells for a response to the related determinants. The result is antibody production to the shared epitope for which it previously lacked T-cell help, while the CD4+ T cell remains tolerant to the shared epitopes.

A similar approach has been used to induce autoimmune disease to autoantigens that are present in a concentration sufficient to maintain a tolerant state in T but not B cells. The injection of either altered self (new epitopes) or cross-reacting (nonrelated epitopes) thyroglobulin into rabbits resulted in antibody to rabbit thyroglobulin and antibody-mediated thyroiditis (Weigle, 1980). More recently, this phenomenon was rediscovered with ubiquitin molecularly engineered to yield a new T-cell epitope (Dalum et al., 1996). A T-cell-mediated immune response can also result from this approach if the level of antigen is low enough that it permits a leaky T-cell response. The induction of an immune response in mice to autologous cytochrome c can be accomplished by injecting guinea pig cytochrome c (Lin et al., 1991). However, if B cells become hyperresponsive, they act as super APCs and activate and expand the low frequency of self-reactive T cells to an autologous cytochrome c. A variation of the above model has been termed molecular mimicry, which defines autoimmunity induced with viruses that share amino acid sequences with self proteins (Fujinami and Oldstone, 1985). 

In any event, as with normal immune responses, the autoimmune response would require altered self or cross-reacting antigens that activate antigen-presenting cells, causing the release of cytokines in the immediate microenvironment to supply the second signal to T cells. In the simplest example of this model, circumvention of the unresponsive state would result in a transient autoimmune response but not a progressive disease. Thus, to sustain progressive autoimmune phenomena resulting in progressive disease, the insult must persist in the form of self-perpetuating trauma, chronic viral infections, or other permanent changes, e.g., incorporation of viral DNA in the host genome.

Rod Langman - 7:46am May 15, 1997 (#4 of 14)

Transporting the associative antigen recognition model to the clinic is not very exciting in the short term, but it does suggest new places to look for solutions. 

In the clinic, it would be ideal to manipulate responses in an antigen-specific manner, much the way vaccines manipulate the system in an antigen-specific manner. There are two classes of problem that fall into the domain of inhibitory effects: first, reducing the autoimmune response; and second, establishing tolerance to organ grafts. 

The experimental record is glum in terms of converting immunity to tolerance. Lethal whole-body irradiation would be effective in principle but is obviously out of the question. However, there is good news if one thinks in terms of the different classes of immunity - cell mediated versus humoral - and all the immunoglobulin subclasses. For any particular immunogen/pathogen, there are protective and nonprotective classes of response, and, conversely, in autoimmunity there are responses that effectively produce disease because they tend to rid the host and responses that are apparently harmless. Thus, while adult tolerance can't be induced with any certainty, switching the class of the response should be doable because the immune system does it normally (e.g., initial antiviral cytotoxic T cells followed by long-lived recirculating IgG for future protection). 

The experimental record for inducing unresponsiveness in transplant patients is quite respectable. The answer to whether this is tolerance is probably no. Most transplants last at best several years, but eventually, in 10-20 years, there will be serious rejection episodes in the best of cases. Aside from generalized immunosuppressants, some form of prior "immunization" with donor cells (often blood transfusions from various donors) is beneficial - a situation that is likely tied to inducing a response in an ineffective class. But, if one were going to use immunosuppressants, the one to look for would block the expression of eTh function but not eliminate iTh, which we want to drive to tolerance with antigen. Keeping this up long enough for the eTh to all revert to iTh would in principle do the trick. However, we have no good estimates on the rate of reversion of eTh to iTh and few ideas of how to speed it up if it is rather slow. These are critical numbers because maintaining suppressed eTh function is an invitation for a pathogenic takeover bid. 
In the absence of new data collected with these concepts in mind, it is difficult to look into the crystal ball and make the precise predictions that are needed in the clinical setting.