What roles do regulatory T cells play in the control of the adaptive immune response?

Abstract

The immune system, like many systems responsive to specific stimuli, requires feedback regulation. The key regulatory element determining antigen-specific responsiveness is the effector T helper. As the response tends to overshoot, a feedback control of the magnitude of the response is critical to avoid immunopathology. This is the proposed role of the effector T suppressor (T_s). The reasons for this interpretation of the data are discussed as are the reasons that the competing postulate is ruled out, namely that T_s function in determining the self-non-self-discrimination. The regulatory T cell family consists of two lineages, T helpers and T_s. Differentiated derivatives of the T helper lineage drive the expression and amplification of specific classes of defensive effector cells. T_s feedback to limit the magnitude of the process so that debilitating immunopathology is acceptably infrequent.

Introduction

This essay is an attempt to put ‘Tregs’ into a biological context. The overriding enthusiasm for this cell as ‘the Holy Grail of transplantation’, ‘the jack of all trades’, ‘the master regulator of immunity’, ‘the guardians for life’, etc., is a tribute to the marketing strategies so popular with high-profile journal editors but is not very useful for understanding. While the major effort of immunologists is directed towards application and not unreasonably so, there is still something to be said for a minor effort being directed toward understanding as a virtue unto itself. ‘Understanding’ means that we can define the function upon which evolutionary selection operates that accounts for the contribution of that function to the physiology, behavior and structure of the normal present day system.

Any discussion of Tregs runs into a problem of meaning and in this case it is particularly important. The term Tregs designated to mean T suppressors (T_s) is inappropriate as regulatory T cells must include the T helper (T_h). Both categories are required to manage the specific response to an antigen. Clearly, if T_s were the only Treg, the immune system would be unable to function. I will use the terms T_h and T_s because they historically and more directly describe their roles.

Background

The evolutionary selection pressure is operating on the biodestructive and ridding output of the immune response. As this response is just as harmful to the host as it is to the invader, it must be somehow confined to the destruction of the pathogen. The selection then affects several levels which determine the characteristics of the response and allows us to divide it into three modules conveniently amenable to analysis: (i) the nature and structure of the repertoire, (ii) the sorting of the repertoire (Decision 1) and (iii) the coupling of the repertoire to effector class (Decision 2).

The immune system is subject to two different levels of selection, germline and somatic. The germline-selected system, often referred to as ‘innate’, is characterized by a repertoire that recognizes epitopes shared by many pathogens. This repertoire recognizes the epitopes on pathogens that are not present in the self-of-the-species. This is why one can transplant tissues without rejection between individuals (vertebrate or non-vertebrate) that express only the germline-selected system. The recognitive sites (paratopes) of the innate system are born coupled to the appropriate effector mechanism; the signal triggered by the binding of ligand is all that is needed to initiate the effector response.

As the interactive selection pressures between non-vertebrates and their pathogenic universe progressed, there came a point in time when germline evolution of the recognitive system could no longer keep pace with the ability of the pathogenic universe to escape recognition. At about the time when jawed vertebrates appeared, evolution invented a remarkable solution, namely to somatically generate a random repertoire that divided the antigenic universe into combinatorials of epitopes. An antigen is a collection of linked epitopes. This invention, often referred to as the ‘adaptive’ system, is the subject of this essay and is a prime example of somatic evolution. It is important to appreciate that the innate system is blind to a large portion of the antigenic universe that the adaptive system recognizes. There must exist, therefore, antigen-specific regulatory pathways unique to the adaptive immune system. Implied is that, if the recognitive elements of the innate immune system were ablated, the adaptive system would remain functional and vice versa. Interestingly enough, the innate and adaptive immune systems share most, if not all, effector mechanisms.

While this somatically generated repertoire solved the problem of escape from recognition, it created two new problems that had to be solved in tandem: (i) The somatically generated repertoire is random with respect to the recognition of self (S) and non-self (NS). Therefore, it had to be sorted into those specificities [anti-Self (S)] which if expressed would debilitate the host and those specificities [anti-Non-self (NS)] which, if not expressed, would result in the death of the host by infection. The sorting of the repertoire (Decision 1) is what is meant by the colloquialism, a self-non-self-discrimination. (ii) Once sorted, the residue anti-NS had to be coupled to the biodestructive and ridding effector mechanisms in a coherent and independent manner for each pathogen. As there are effective and ineffective effector mechanisms for the ridding of each pathogen and the expression of ineffective classes can inhibit the efficacy of the effective class, a decision must be made between them. Further, the magnitude of the effector response must be carefully controlled as overshoot has serious immunopathological consequences. The coupling of the sorted repertoire to the appropriate effector mechanism and the control of its magnitude will be referred to as ‘regulation of effector class’ or Decision 2.

The three modules

How do regulatory T cells, T helpers (T_h) and T_s fit into the framework of the three modules?

The repertoire (Module 1)

There are three non-overlapping families of ligands for the epitope-specific receptors of the cells of the somatically selected (adaptive) system. The TCR recognizes peptide (P) presented by an MHC-encoded restricting element (R). The restricting element (R) is recognized in an allele-specific manner, the specificity of which is determined by germline selection. Our concern here is not with the recognition of R alleles but with the recognition of peptide (P) by the TCR anti-P site which must have a specificity sufficient to make a self-non-self-discrimination (1–3).

The T cells fall into two classes dependent on the category of R that they recognize. Cytotoxic T cells (T_c) are restricted to Class I MHC symbolized here as RI Class I MHC that restricts T_c whereas helper (T_h) and T_s are restricted to Class II MHC (4) symbolized here as RII Class II MHC that restricts T_h. The ligand for T_c is P-RI and for T_h/T_s it is P-RII. The virgin repertoires of T_h and T_s are ‘per force’ the same as they express TCRs encoded by the same gene segments and are positively selected by the same RII elements in thymus.

B-cells recognize via their B-cell antigen receptors (BCRs) a shape-patch (referred to here as a ‘topotope’) on the intact antigen, in contrast to the TCR of T cells which recognizes peptide (P) derived by processing of the antigen. The antigenic universe for each class of cell is distinct, T_c recognize P-RI, T_h/T_s recognize P-RII and B-cells recognize a shape-patch (topotope). There are, therefore, three non-overlapping families of self-epitopes in each individual: self-P(Ps)-RI, Ps-RII and self-topotopes.

Each of the three antigenic universes is subdivided into combinatorials of ligands (epitopes) by the paratopic repertoires of each of the cell types: T_c, T_h/T_s and B. Paratopes recognize epitopes, not antigens. An antigen is a combinatorial of linked epitopes.

The sorting of the repertoire (Module 2)

As pointed out earlier, a random paratopic repertoire cannot be left unsorted if it is to function as a protective mechanism that does not debilitate the host.

An aside on nomenclature: self versus non-self.

No two words used by immunologists engender more ‘confusionism’ than self and non-self, yet after >60 years of usage, it is difficult to communicate without using them. We tried to substitute the not-to-be-ridded—to-be-ridded discrimination without success (5). So now lets try to clarify these terms by limiting their meaning.

Self is defined by the individual's adaptive immune system during a somatic learning process. It is not defined by the New Oxford English dictionary or by the immunologist. Not all autogenously generated or germline-encoded components of the host are self to the immune system. An immune attack on a self-component must have a debilitating consequence in order that a mechanism for the sorting of the repertoire be evolutionarily selectable. The self-components of an individual as discussed above fall into three classes, Ps-RI seen by T_c, Ps-RII seen by T_h/T_s and the self-topotope seen by B-cells, and this has consequences.

The existence of a feedback mechanism to regulate responsiveness is an a priori necessity; there never was a question as to whether feedback regulation (suppression) of the induced response exists. What was in question is whether the particular cell or system that was described fulfilled that function (6–8). The role of feedback regulation was understood in the 1960s by endocrinologists and by molecular biologists studying induced enzyme synthesis in microorganisms. In the very first formulation of the two-signal model of the self-non-self-discrimination in 1970 (9), we opted for deletion and against suppression as the mechanism for ridding anti-self cells. As the model became more sophisticated emerging as the Associative Recognition of Antigen (ARA) Model (10–13) so the arguments against suppression operating to sort the repertoire became stronger (7, 8). Given that suppression regulating responsiveness is an a priori necessity, it could only play a role at the level of Decision 2, the regulation of effector class. Bretscher (6) appreciated this in his 1972 classic paper ‘The control of humoral and associative antibody synthesis’ and this was developed further both by Langman (7) and by myself (8) over the years.

The immunologists who discovered what appeared to be suppression in the 1970s placed its function in the determination of the self-non-self-discrimination (i.e. the sorting of the repertoire) applying the descriptive term ‘infectious tolerance’. This view is still dominant today (14–18), albeit highly questionable as will be developed here.

The immune system has no way to know what is encoded in the germline or the origin of the antigen. In addition, there is no physical or chemical property of antigens that permits the immune system to segregate them into self or non-self as classes. Further, what is self for one individual of a species is non-self for another. The random paratopic repertoire is sorted in an individual-specific manner. Given that the immune system cannot determine the specificity of a paratope that has not interacted with its ligand, the sorting of the repertoire requires the prior sorting of the antigenic universe into self and non-self. How might that be accomplished?

At present, there is only one viable hypothesis. The antigenic universe is sorted into self and non-self as a function of developmental time. The principles involved in sorting the antigenic universe have been discussed many times (10–12, 19–21) so that here only an outline will be presented.

The sorting of the repertoire as a function of developmental time.

In all vertebrates with adaptive immune systems, the fetus is protected by maternal immunity. There must exist a developmental time window when all self and no non-self are presented to the newly arising antigen-responsive cells of the immune system. These will be referred to as initial state or ‘i-cells’. While the window is open, the system is inactivatable only. Interaction with an epitope results in the inactivation of the i-cell. As the only epitopes present are self-epitopes, then by definition, the interaction of i-cells with them inactivates anti-self i-cells leaving as a residue anti-non-self i-cells that accumulate to protect the individual. When the window closes and the system becomes responsive, the persistence of self maintains the state of tolerance to self. Basically then the sorting of the repertoire rests on the mechanism for the appearance of responsiveness. What determines the transition from an unresponsive (inactivatable only) state while the window is open to a responsive state when the window closes?

The crucial role of the regulatory effector T helper.

Under the ARA model (12), the interaction of the antigen receptor with ligand results in an inactivating Signal[1]. If a cell on the pathway of Signal[1] inactivation receives Signal[2], then it is activated and passed on to Decision 2, the regulation of class. The source of Signal[2] is the effector T helper (eT_h). While the developmental time window is open, the system is unresponsive because of an insufficiency of eT_h. When the window closes and the system becomes responsive, it is because a sufficiency of eT_h anti-non-self has appeared to prime induction. As iT_h undergo the same sorting process as iT_c and iB, the iT_h anti-self are inactivated by interaction with Ps-RII while the window is open and continue to be inactivated after the window closes because of the persistence of self (Ps-RII), thereby preventing the appearance of eT_h anti-self but allowing eT_h anti-non-self to be induced. The precise details of this model have been published (22, 23).

The repertoires of iT_h, iT_c and iB are purged of anti-self leaving the residue as anti-non-self. Before discussing the behavior of iT_s, we must specify how the eT_h interacts to deliver the activating Signal[2] to the T_c- and B-cells receiving Signal[1].

Signal[1] is delivered epitope-by-epitope. The antigen receptors, BCR and TCR, see epitopes not antigens which are combinatorials of linked epitopes. Inactivation, then, is mediated epitope-by-epitope. Activation, which is determined by delivery of Signals ([1]+[2]), is mediated antigen-by-antigen. This means that the eT_h can only deliver Signal[2] to an i-cell receiving Signal[1], provided that the two cells are interacting with epitopes that were linked on an antigen. This is referred to as ‘ARA’. The coherence and independence of the effector response depend on ARA as well as does the maintenance of the normal tolerant state to self.

An aside on nomenclature: ‘tolerance’ vs ‘unresponsiveness’.

The manipulation of an animal so that it fails to respond in an antigen-specific way to an immunogen to which it normally responds, is referred to as ‘unresponsiveness’. This term should be applied to the experimental observation.

The extrapolation of the observation of unresponsiveness to a theory of how the self-non-self-discrimination is normally established and maintained, is ‘tolerance’. Tolerance is a conceptualization used to interpret the observation.

There are several ways to establish ‘antigen-specific’ unresponsiveness. There is only one way to establish ‘natural’ tolerance, a point to which we will return.

Is the T_s repertoire sorted to be anti-Ps or anti-nonself P (Pns)?

The functioning repertoires of T_c, T_h and B are sorted to be anti-NS. While in the framework of the ARA model the same would be true of T_s, the immunological community is schizophrenic on this subject (14, 24–29).

If the T_s repertoire is sorted to be anti-S, then it cannot regulate the response to NS and, therefore, cannot regulate the effector class. If it is sorted to be anti-NS, it cannot regulate the response to S and, consequently, cannot play a role in determining the self-non-self-discrimination. If the T_s repertoire is unsorted, it would render the immune system ineffective and be, therefore, unselectable. No matter where, when or how it is accomplished, the repertoire of T_s must be sorted.

So, which is it? Clearly, the repertoire of T_s is sorted to be anti-NS like all other i-cells, iT_c, iT_h, iB. The iT_s follows the same rules of inactivation and activation as do all i-cells, namely it is inactivated by Signal[1] alone and activated by Signals([1]+[2]) where Signal[2] is delivered by an eT_h in ARA (8).

If eT_s anti-S were to sort the repertoire by purging the other classes of i-cells anti-S, then they would have to function in ARA to be self-antigen specific. This would result in the acceptance of grafts between individuals, which, of course, is contrary to fact, not to mention that an inability to respond to non-self-antigens that share epitopes with self would be lethal. Suppression as the mechanism of ‘natural’ tolerance is ruled out (7, 8). Lastly, a functional population of T_s that were sorted to be anti-self cannot regulate the response to non-self-antigens; this too is contrary to fact (26, 29).

An aside on nomenclature: ‘autoimmunity’ versus ‘immunopathology’

Autoimmunity is the consequence of a debilitating attack on ‘self’. It is distinct from both ‘autoreactivity’ and ‘immunopathology’. Autoreactivity is an immune response to autogenously expressed components that is normally salutary, not debilitating. They include senescing and necrosing cells, denatured and effete proteins as well as other macromolecular waste. This housekeeping function of the adaptive immune system (30) is not autoimmunity. Housekeeping (autoreactivity) is a debridement that maintains the integrity of healthy tissues and is involved in wound healing among other functions. The autogenous targets of housekeeping are non-self to the immune system.

Immunopathology is due to an attack by the expressed effector function on innocent bystander tissues. The normal induction of an effector response to a pathogen concomitantly does harm to normal tissues usually to an acceptable extent. The [antigen-antibody]_n complexes activate inflammatory processes which, while protective against invaders, are also damaging to healthy tissues and are clinically manifested when they overshoot. The biodestructive and ridding effector mechanisms must be confined to an acceptable radius, ideally limited to the inducing target.

Autoimmunity results from an effector response to self. This, by definition, requires the breaking of tolerance which is a process that converts a self-antigen into one that behaves as a non-self-antigen. Immunopathology results from an effector response to non-self, the biodestructive activity of which spills over to attack healthy tissues. This distinction, while crucial, is not easily made in any given situation. The reason is that ‘breaking tolerance’ is largely conceptual and difficultly assayable. Consequently, when presented with a complex disease process involving immune effector functions, it is difficult (not impossible) to determine whether the origin is autoimmunity or immunopathology; yet the understanding of the process depends on this distinction. Further, the two processes are not distinguishable by histopathology because they use the same effector mechanisms.

Autoimmunity requires the breaking of tolerance at the level of the T helper. All autoimmune processes are eT_h dependent. Chronic autoimmunity requires that the induction of eT_h anti-‘the self-component’ become a self-sustaining process driven by the self-component as an immunogen. If the target is self-renewing (e.g. fibroblasts, epithelial cells) the likely autoimmune attack is chronic; if it is non-renewing (e.g. endocrine cells), the autoimmune attack only need be transitory. Once tolerance is broken at the level of the T helper, the defensive categories of cell, T_c and B can be engaged. The autoimmune process is in general directed to a single-defined self-component.

Immunopathology, by contrast, resulting from the overshoot of an anti-NS response, attacks a variety of tissues to an extent dependent on their sensitivity to the particular effector mechanism.

A closer look at self.

The R used by the αβTCR have different distributions of expression on cells. RI is expressed by all cells. Therefore, the T_c repertoire is expected to be purged of recognition of all intracellular proteins presented as peptide by RI at a detectable level. RII is only expressed on specialized cells, such as B-cells, and various other categories of antigen-presenting cell (APC). The T_h repertoire, therefore, is purged of recognition of only a subset of the host's germline-encoded proteins, in essence those expressed in B-cells and APC. The vast majority of autogenous proteins, if they were to be expressed as P-RII would be non-self to T_h. This has important consequences as T_h cells are the key regulators of responsiveness. The B-cells see topotopes expressed on the surfaces of cells and on soluble proteins. They do not encounter intracellular components. When the developmental time window is open, the fetal cells die by apoptosis and the resultant granules are phagocytosed without the internal contents of the cells being released. The B-cell repertoire treats those proteins as non-self. When the window closes, the cells die by necrosis due to their senescence or the cytopathic effects of pathogens. The internal contents of the necrosing cells are released to be processed to anti-non-self P (Pns)-RII, the ligand for eT_h, both by APCs and B-cells. This results in the induction of a humoral housekeeping response. These autogenous products are non-self to the immune system.

The relationship of these three classes of ligand to autoimmunity becomes the next question.

Why don't all of us succumb to autoimmunity?

Consider Table 1 summarizing the relationships between self-ligand and cell type that were discussed above.

Table 1.

The sorting of cell types by expressed self-ligands

Cell type	Ps-RII		Ps-RI		Self-topotope
	Expressed	Not expressed	Expressed	Not expressed	Expressed	Not expressed
Th	−	+	…	…	…
Ts	−	+	…	…	…
Tc	…	…	−	+	…
B	…	…	…	…	−	+

− = cell anti-S is purged; + = cell anti-NS is residue and … = not a ligand for the given cell type.

Under the ARA model, eT_h is required for the activation of all i-cells expressing the somatically generated repertoire. The self-ligands that are encountered by iT_h and expressed when the developmental time window is open are largely on B-cells and APC (including thymic medullary cells). When the developmental time window closes and these cells process proteins originating from sources that do not express RII, these autogenous components are treated as non-self and recognition by eT_h initiates a housekeeping response. In the absence of an eT_h-delivered Signal[2], the interaction of the iT_c, iT_s or iB with ligand (Signal[1]) is inactivating.

Thus we have three situations regarding a response to self-ligands (autoimmunity).

First, in the absence of recognition by eT_h of the self-antigen via ARA, all other i-cells that recognize it would be inactivated. The i-cells that are in the residue after sorting have no self-ligand with which to interact and, therefore, pose no problem. They are defined by the immune system as anti-NS.

Second, in the presence of recognition in ARA of the ligand by eT_h there are two situations:

• (i) The persisting presence of the self-ligand purges the defensive iT_c- and iB-cells to a level that makes ineffective induction by self (see later, the autoimmune boundary). However, in the presence of a non-self-antigen that shares epitopes with self (i.e. is cross-reactive), tolerance may be broken because it is driven by eT_h anti-non-self. This is the limitation imposed by the requirement for ARA to maintain the specificity of responsiveness.

As roughly 10% of non-self-antigens share determinants with self, why is autoimmunity rare? There are two reasons.

First, the presence of the self-antigen competes with the crossreactive non-self-antigen to prevent the breaking of tolerance.

Second, the cross-reactive non-self-antigen is ridded before the induction of eT_h anti-S reaches a level high enough to sustain an autoimmune process driven only by the self-antigen (i.e. the breaking of tolerance).

• (ii) A non-self-antigen may share epitopes with an autogenous component that is not encountered as a self-ligand. The autoreactive response to this autogenous component would not trigger autoimmunity because it is not expressed as a self-target. This illustrates why the semantics of self and non-self is important.

Tolerance is broken either by non-self-antigens that share determinants with self or by unspecific agents like LPS that deliver ‘Signal[2]’ non-associatively (not in ARA). In this latter case, the immune system responds with immunopathology and generalized autoimmunity (31). The response due to breaking of tolerance by crossreactive non-self-antigens has a defined target that results in autoimmunity to a specific tissue. A non-associatively delivered Signal[2] would drive predominantly immunopathology and to a much lesser extent generalized autoimmunity which originates from the induction to effectors of the anti-self cells comprising the autoimmune boundary.

What is the source of anti-self cells when tolerance is broken: the autoimmune boundary?

In humans and mice, throughout life there is a steady state production of i-cells that are on the pathway of inactivation upon receiving Signal[1]. If inactivation were instantaneous, then no i-cell could be activated because the ligand–receptor interaction that initiates Signal[1] is fast compared with the cell–cell interaction involving eT_h, the source of Signal[2]. The time that it takes before Signal[1] becomes irreversible (i.e. the half-life of reversibility by Signal[2]) determines the steady state level of the population of anti-self i-cells on the pathway to inactivation. Evolutionary selection had to compromise. If the half-life were too short, the efficiency of activation would be too low to protect the individual from infection. If the half-life were too long the frequency of autoimmunity would be too high due to breaking of tolerance. The autoimmune boundary is a compromise between these extremes, sufficiently high to allow protection from infection and yet low enough to reduce the frequency of autoimmunity to an acceptable level. The steady state level of anti-self i-cells on the pathway to inactivation can be revealed experimentally by immunization with non-self-antigens crossreactive with self that break tolerance, for example, ‘Myasthenia gravis’ resulting from immunization of rabbits with the acetylcholine receptor of the electric eel.

Why wasn't suppression evolutionarily selected as the mechanism of natural tolerance?

In order for suppression to be antigen specific, tolerance must be mediated via ARA. This means that no response would be possible to non-self-antigens that share epitopes with self, a lethal situation. When deletion is the mechanism of ‘natural’ tolerance, responsiveness is mediated by ARA. This means that autoimmunity (breaking of tolerance) becomes a threat. As discussed above, given a deletion mechanism, autoimmunity can be controlled. However, given a suppressive mechanism, responsiveness to a crossreactive non-self-antigen cannot be controlled and yet maintain the state of tolerance to self.

A questionable concept: central versus peripheral tolerance.

It is widely assumed that (16, 18) there is a fundamental difference between central and peripheral tolerance. Central tolerance is deletional whereas peripheral tolerance is mediated by a collection of failsafe negative (recessive) mechanisms like anergy, activation-induced cell death, receptor editing, immune deviation and ignorance, that is sequestered from interaction with i-cells). To this is to be added, the positive (dominant) mechanism of T-suppression (32). The idea that evolution selected for peripheral fail-safe mechanisms, particularly the panoply of those cited above, to correct the central deficiencies in sorting the repertoire, is untenable. While the developmental window is open and the system is inactivatable only due to the insufficiency of eT_h, it makes no difference whether the i-cell interacts with its self-ligand in the thymus, the bone marrow or elsewhere in the animal (i.e. Signal[1] is inactivating) (12). The establishing of tolerance occurs while the developmental window is open and operates throughout the animal because eT_hs are not at a functional level.

Tolerance is maintained after the window closes and the system is responsive. During this period of maintenance, one might invoke the concept of central (thymic) tolerance and peripheral tolerance. Responsiveness is uniquely a property of the periphery where eT_h anti-NS reach a priming level. They are functionally absent from the thymus which remains an inactivatable-only enclave where iT cells are generated, sorted and exported to the periphery.

If the self-component is expressed in the thymus as Ps-RII or Ps-RI, then the iT cells recognizing that component are Signal[1] inactivated. If it is not expressed in thymus or if the iT cell fails to interact with it, then the iT cells must be inactivated upon encountering that self-component (S) in the periphery with the result that eT_h anti-S remains functionally absent.

The mechanism of Signal[1] inactivation is a second-order question. In thymus, Signal[1] is deletional (apoptosis), which if it remained unchanged in the periphery would be adequate. The additional recessive mechanisms of inactivation listed above are not directly selectable. They are experimental ways of revealing either steps in the pathway of Signal[1] inactivation (e.g. anergy) or byproducts of the mechanism of haplotype exclusion (e.g. receptor editing). In order for a mechanism to be fail-safe, it must be superior to the mechanism it is correcting and, therefore, would, over evolutionary time, replace it. In order to sort the repertoire, the inactivation mechanism selected by evolution must be irreversible (ruling out anergy) and must not create more problems than it solves (ruling out receptor editing). In any case, what is key is not the mechanism of Signal[1] inactivation but, rather, what is the factor that distinguishes peripheral self from peripheral non-self (i.e. what maintains tolerance)?

For suppression to be the mechanism of peripheral sorting, the T_s must be somatically selected to be anti-S (not anti-NS) and must function in ARA to be antigen specific. Given this, the i-cells anti-NS would be expected to be activated (not inactivated) by Signal[1] (i.e. the eT_h are not required for activation and would need a proposal as to their role). As pointed out earlier, the prediction of such a model would be that tissues would be transplantable between individuals of a species (contrary to fact) and the individual would be unresponsive to non-self-antigens that share epitopes with self (sure to be lethal).

As one illustration of a proposed activating Signal[1], we are told that ‘one obstacle to the spurious activation of naive peripheral T cells upon TCR recognition of self-ligands is a requirement for an additional signal. This signal emanates from costimulatory activating receptors … that are up-regulated in APCs upon exposure to microbial products that are recognized by a set of evolutionarily conserved pattern recognition receptors’ (33).

The model is that the activating signal is a composite of Signal[1] plus a failsafe auxiliary (costimulatory) signal unrelated to the epitope being recognized by the TCR. The auxiliary signal prevents ‘spurious activation’ of anti-self T cells which have accumulated because they were not inactivated by interaction with ligand. Consider two T cells interacting with an APC, one anti-self, the other anti-non-self, both receiving the auxiliary signal non-associatively; both would be activated. How could such a construct contribute to the decision process that sorts the repertoire? Signal[2] must be non-self antigen specific and in some strict way linked to the epitope on the antigen-triggering Signal[1]. A costimulatory signal, as described above, would result in immunopathology and generalized autoimmunity. It could not protect against ‘spurious activation’ of anti-self cells; it would in fact enhance it. Lastly, how would this regulate the multitude of antigens not recognized by the innate system (34).

Why must Signal[2] be antigen specific?

Signal[1] resulting from the interaction of the TCR/BCR with ligand (epitope) cannot inform the i-cell if it is interacting with a self- or non-self-epitope. Therefore, the additional signal that makes that determination must be antigen specific and derived from a source that has made the discrimination between a self- and a non-self-antigen. A self-antigen is one that expresses self-epitopes only. A non-self antigen can express non-self-epitopes only or any combination of self- and non-self-epitopes. This is why ARA is postulated to be obligatory in the delivery of Signal[2]. The cell-initiating Signal[2] must recognize via its antigen receptor an epitope from the same antigen with which the target i-cell is interacting (receiving Signal[1]). If the source of Signal[2] were an APC, then a mechanism to establish the equivalent of ARA would be required (35). The so-called co-stimulatory signals, as they are described, do not meet this requirement. An unspecific Signal[2] substitute, like LPS, or the delivery of Signal[2] non-associatively, like graft versus host reaction, triggers predominately immunopathology and to a lesser extent, autoimmunity by inducing indiscriminately i-cells anti-self and anti-non-self. ‘Co-stimulation’ functioning non-associatively is not the source of Signal[2].

Why must Signal[1] be inactivating?

Under the view that the repertoire is sorted by suppression, Signal[1] is activating and Signals([1]+[3]) are inactivating. In order to be antigen specific, Signal[3] must be delivered by eT_s anti-self in ARA. Activation then would be mediated epitope-by-epitope and inactivation would be mediated antigen-by-antigen. This inversion of the ARA model is ruled out as it would result in acceptance of transplants between individuals of a species. This untenable view is raised once again to illustrate a contradiction in this suppressor model revealed by asking, how would one explain the breaking of tolerance mediated by a non-self-antigen that shares epitopes with self?

In the absence of a deletional mechanism, the putative anti-self-repertoire of T_s-cells cannot be purged of the given specificity responsible for the recognition of the self-P processed from the NS-antigen. Further, effector T suppressors (eT_s) cannot suppress suppressors and be a functional system. Therefore, tolerance can be broken only by inducing a level of eT_h anti-'the given self-P’ (Signal[2]) that overrides the suppressive effect of eT_s anti-'the given self-P’ (Signal[3]). This requires that Signal[2] be activating in a system where it must be assumed that Signal[1] is activating, or formally speaking, does not exist (i.e. no signal to the i-cell on binding ligand).

In sum, no valid model has appeared or is evident for the antigen-specific breaking of tolerance in a suppressor framework for sorting the repertoire.

Why are T_s class II (RII), not class I (RI) MHC restricted?

As class I MHC (RI) is expressed by all cells and given the assumption that T_s control anti-self reactivity, would it not have been of far greater efficacy for T_s to express an anti-(Ps-RI) repertoire that would include recognition of most germline-encoded host proteins than to express an anti-(Ps-RII) repertoire that lacks recognition of most of them?

One simple answer would be that the evolutionarily selected role of T_s is not to purge anti-self reactivity but to regulate anti-non-self-reactivity. This implies that T_s anti-NS must regulate T_h activity, a point that will be developed under Module 3.

Confronting the question of specificity versus degeneracy.

The argument that degeneracy poses a major problem for the self-non-self-discrimination has become very popular (36, 37). One of the earliest voices for this view is Seddon and Mason (38). To this they added that suppression is a peripheral mechanism to deal with T cells anti-self that escape central deletion in the thymus. Their view was based on the assumption that T_s are sorted to be anti-self and that the self-non-self-discrimination faces a ‘severe problem’ due to the extreme degeneracy (their term, ‘crossreactivity’) of TCR anti-peptide recognition. Their arguments are questionable because they failed to distinguish (i) autoimmunity from autoreactivity and immunopathology and (ii) degeneracy from specificity. This applies also to the present day derivative school (36, 37).

The importance of the distinction between autoimmunity and immunopathology has been analyzed above. They cannot be distinguished by histopathology because the two disorders result from the same effector biodestructive and ridding mechanisms.

The importance of the distinction between degeneracy and specificity has been discussed (39, 40) but deserves a brief comment here. The degree of specificity of the TCR/BCR is driven solely by evolutionary selection to make a self-non-self-discrimination (41–43). Consequently, for our present purpose, the degree of specificity can be loosely defined as the probability that a random TCR/BCR from an unsorted repertoire, will be anti-self. We refer to this as the Specificity Index (43–45).

When a single paratope functionally recognizes a family of epitopes that are chemically distinguishable, we refer to this as degeneracy of recognition. If any one of this family of epitopes is defined by the immune system as a self-epitope, then every epitope in that family is a self-epitope. It makes no difference what the immunologist calls them. The self-epitope in the family played its role in the sorting of the repertoire. The so-called non-self-epitope in that family may have originated from the planet Uranus and be referred to by the immunologist as non-self. However, to that individual's immune system, it is nonetheless a self-epitope.

Specificity is random with respect to the defining of self and non-self. Degeneracy is non-random with respect to the defining of self and non-self, a fundamental distinction. Seddon and Mason claim that T cells anti-self can be used to protect the individual against non-self pathogens, illustrating the failure to distinguish degeneracy from specificity. For the immune system, paratopes that functionally recognize both self and non-self epitopes do not exist. They exist only for the immunologist.

While Seddon and Mason present a viable argument for T_s playing a role in immunopathology, it is weak for autoimmunity. In any case their view, generally accepted even today, leaves us with this puzzle. If the TCR/BCR cannot adequately distinguish self from non-self, how can a T_s which faces the same problem help them make the distinction? Clearly, this idea needs clarification/rationalization.

Lastly, it might be pointed out that there are sharp thresholds for signaling via the TCR/BCR dependent on the degree of occupancy per time interval.

The regulation of effector class (Module 3)

Thus far no role for suppression in the sorting of the repertoire (Module 2) has emerged. In essence, it has been argued that such a role is ruled out. Consequently, T-suppression must play a role in the regulation of class.

In spite of the fact that the problem of regulation of class is rich in information, it lacks a unifying conceptual foundation. What might be the framework within which we could analyze this process?

We might begin with a question with which we have wrestled over the years. Is the determination of the effector class a somatically learned process like the sorting of the repertoire or is it mediated by a germline-selected mechanism or both? Given that there is an enormous variety of pathogens, it has seemed unavoidable that a learning mechanism existed to assay which class was effective in ridding a pathogen. The effective class then would be favored by suppressing the ineffective classes. After exploring countless models, none survived because either the proposed assay used by the immune system for effectiveness of ridding was implausible or the mechanism for suppression required an isotype-restricted, antigen-specific T_s for which no conceptual or experimental basis could be found. As a result, we have assumed that a role for suppression in determining the isotype is unlikely (46). Although failure to come up with a plausible model for a somatically learned determination of effector class is not a very good argument against its existence, we decided to follow Medawar's principle, the art of the soluble, and explore a germline-selected mechanism.

The innate system has an entirely germline-selected expression of effector function because the recognitive site is pre-coupled to the given effector mechanism. Evolutionary selection on the germline has determined that the coupling relationship is adequate to protect. For the adaptive system, a set of decisions must operate to make a response effective (e.g. cell-mediated or humoral and, in each category, which subclass).

Bretscher and associates (47–50) have introduced two central concepts, ‘coherence’ and ‘independence’ and supported them by carefully constructed experiments. As there are effective and ineffective classes in ridding a pathogen, the response to it must be ‘coherent’ (i.e. solely in an effective class) because the ineffective classes inhibit the efficacy of the effective classes. As different pathogens are best ridded by different effector mechanisms, the response to each must be ‘independent’. The regulatory cells determining coherence and independence must then function in ARA.

There must be a relationship between what is to be eliminated and what is induced. If we refer to the unit of elimination as an ‘eliminon’ (e.g. a virus, a bacterium, a fungus, a protozoan, a cell, etc.), then this unit must be viewed by the immune system in a manner that ties the response to an effective effector function. The first step in this pathway is the activated defensive cell, one that has responded to Signals([1]+[2]). This activated defensive cell anti-‘the eliminon’ must now be told which pathway of differentiation to follow that relates effector function to the biodestruction and ridding of the eliminon.

The pathway to activation requires that the eliminon be processed to Pns-RII so that eT_h can be induced. If the eliminon is a virus for which a cytotoxic response is effective then it must also be cross-presented as Pns-RI on the APC to permit an eT_h–APC–iTc interaction. For a humoral response, the iB-cell captures and processes the eliminon to Pns-RII, the ligand for the eT_h-delivering activating Signal[2]. Once activated, the cell must receive additional signals to proliferate, to differentiate into lineages of isotypes and to become effectors.

There are two levels of decision. First, is the response to be cell mediated (defensive eT cells) or humoral (defensive eB-cells)? This decision appears to be initiated by signals dependent on whether the pathogen is intracellular or extracellular. Second, if the response is cell-mediated, which defensive T cell, eT_c, eT_h1, eT_h17 etc., is to be favored? If it is humoral, which defensive B-cell, one secreting which isotype, IgM, G, A, E, is to be favored?

There are many eliminons and only a handful of effector mechanisms to rid them. Therefore, some grouping of eliminons relative to effector mechanisms is required. The information for this grouping cannot come from the paratope–epitope interaction requiring, therefore, that the relationship between eliminon and effector mechanism be derived from some consequence of the infection that might be described under the umbrella, ‘tissue injury’.

For some years, now a school of immunologists (51–61) have been arguing that the self-non-self-discrimination can be determined at the level of effector output. They have called attention to the role of tissue injury in directing the choice of effector response. While the sorting of the repertoire at the level of the effector response is untenable (12), the role of tissue injury as a source of specific signals for the determination of the effector class is a reasonable derivative from their ideas. Tissue injury is a blanket term which must be subdivided to provide the source of specific signals that relate the eliminon to the choice of an effective biodestructive effector mechanism.

In the end, all non-self, no matter how it is destroyed, must be ridded by phagocytosis. Therefore, a humoral response in a class that arms this process is required to rid the detritus from any eliminon. The response switches from cell mediated to humoral, when intracellular pathogens are involved.

There are important proposals for these regulatory switches that begin to open up the field. Matzinger (62) suggests that the injury of different tissues are sources of distinct signals to the immune system determining class. As pathogens can be grouped by the trauma to the specific tissues they invade, this is a heuristic probing idea meriting intensive investigation. Bretscher and associates (6, 63–66) stress the effective level of T-help as the determining factor in switching from a cell mediated to humoral response and from an IgM to Ig other. They have carried out elegant experimental studies to support the concept. The effective level of T-help is a composite of several factors such as concentration of secreted interleukins and the numbers of cells signaling per unit time. Further, it is determined by the opposing rates of ‘autocatalytic’ induction of eT_h versus the level of inhibition of the induction of eT_h by eT_s (Fig. 1).

Fig. 1.

The proposed pathway of regulatory controls (see list of abbreviations).

The signals as to which class should be expressed upon encountering a pathogen is postulated to come from the traumatized or stressed tissue (62). These signals are read by a precursor T_h which differentiates into the T_h lineages that direct defensive cells iT_c and iB to express appropriate effector classes (67). As pointed out earlier, T_s is unlikely to play a role in this decision process. If we rule out T_s as determining the effector class, the only regulatory role that it can play is in determining the magnitude of the response. This is a crucial role if immunopathology is to be limited to below the debilitating threshold. T_s anti-NS can have no direct effect on autoimmunity and yet leave the system responsive to non-self.

One might envisage the following scenario. The eT_h are required for the activation of all i-cells. This results in eT_h driving an increase in effector activity and eT_s limiting this activity at an evolutionarily acceptable level (i.e. a frequency of immunopathology that is not limiting to the procreation of the species). An antigen-specific relationship for communication between T_h and T_s is implied (Fig. 1). Such a relationship requires a signaling communication between cells that is dependent on ARA. The role of the APC platform will be analyzed later. In this schema, the level of eT_h is dampened by the induced eT_s which feeds back to shut off the induction of eT_h. The feedback is critical as the induction of eT_h, being autocatalytic, tends to run out of control (22, 23).

This elemental circuitry (Fig. 1) can be modified to include the regulation of differentiation of iT_h0 into the various lineages that then control the expression of B-cell isotypes.

Comment on a specific interpretation of T_s function.

Several groups (16, 68, 69) make the argument that T_s function via a generalized inhibition of bystander responses, antigen unspecific. If the immune response is described as inflammation and this effector output is localized, the T_s could dampen its magnitude whether or not that effector output is driven by the same eliminon as the one triggering the eT_s response. The proposed evolutionarily selected role of T_s then would be to regulate immunopathology that is driven by non-self. This removes T_s from playing any role in maintaining peripheral tolerance or in sorting the repertoire. To describe the role of T_s as regulating ‘two core Treg cell-mediated phenomena, bystander suppression and infectious tolerance’ (16) is a contradiction because ‘bystander suppression’ is by definition antigen unspecific involving non-self whereas ‘infectious tolerance’ is by definition antigen specific involving self. There is no way that such a system could functionally regulate the response to non-self by innocent bystander suppression; that would be an unregulated fortuity because effective regulation must be different (independent) for each antigen. Lastly, what would be the evolutionarily selected role of the antigen-specific TCR used by T_s if its output is umbrella suppression of inflammation. All that such a regulatory system would need to recognize is the level of inflammation, not the eliminon.

What is surprising is that no evidence for antigen specificity (ARA) in the functioning of T_s has been examined. While T_s specific for a single H-Y peptide inhibits responsiveness by T_h and T_c to the H-Y antigen [hinting at (not proving) functioning in ARA], no effort was expended to determine whether the inhibition was limited to H-Y (70). If the output is as described (16, 68, 69, 71), there will exist systemic suppression to an extent dependent on the magnitude and localization of the effector response.

If the output of the eT_s is the secretion of anti-inflammatory interleukins/cytokines, then the regulation would require only recognition of the level of inflammation. This could be a role of the innate system; no role for induction to eT_s via TCR signaling appears necessary. If, however, the regulation is not at the level of the effector activity itself but at the level of the induction of the effector cells responsible for inflammation, then specificity of recognition of antigen by the TCR would be expected. In essence, we would be dealing with a family of interacting regulatory T cells the output of which determines the defensive effector class.

Detailing the role of T-suppression.

The induction of iT_h anti-NS to eT_h anti-NS would be an expected major target of eT_s anti-NS. While inhibition of the effector activity of eT_h is also a possible target, it would be far less efficient. As T helper activity is the central regulator of responsiveness, eT_s inhibition of induction of iT_h to eT_h could be envisaged to control indirectly all aspects of the immune response. A shutting off of eT_h activity, specific for one epitope, would result in the Signal[1] deletion of all defensive cells that recognized epitopes linked on the eliminon. Mechanisms for interaction between eT_s and T helpers have been envisaged (71) but that discussion has left too many loose ends. A T-T interaction must take place on an APC or a B-cell because the eliminon must be processed and presented as Pns-RII. A model of such a process must include (i) The states of differentiation of the interacting T cells. Presumably an eT_s is signaling (Signal[3]) an aT_h-receiving Signal[1] by making it directly refractory to Signal[2] or by shortening its half-life of reversibility by Signal[2]. (ii) The antigen specificity of the signaling (ARA). This has been largely ignored in spite of the fact that regulation of class is dependent on the accuracy of the signaling pattern (i.e. ARA). Bretscher (13) has tackled this question by proposing that the eT_h–aT_h interaction occurs uniquely on B-cells. As it is known that B-cells can function as APC (72, 73), this is a thoughtful solution because B-cells process only one eliminon at any given time and therefore guarantee ARA for eT_h–aT_h signaling. A minor extension of his proposal would be that the eT_s–aT_h signaling interaction also occurs on B-cells. There are problems with the mandatory use of B-cells as APC (13) but the more critical question of antigen specificity has been squarely faced.

The use of the B-cell as an APC does not obviate a role for the dendritic cell in this process. In fact, most immunologists treat the signaling interactions between T cells as entirely dendritic cell mediated ignoring any role for ARA (16, 71). If the APC can present simultaneously several antigens (self and non-self) as Ps/ns-RII for interaction with T cells then ARA cannot be enforced. Every non-self-antigen processed by the APC would be converted into a non-self-antigen that shares epitopes with self. Not a very salutary situation!

We have proposed (12, 35, 46) two possible solutions to this conundrum.

First, the eliminon might be processed as a unit that displays all of the derived RII-bound peptides together in a signaling patch. The T-T signaling can only occur between cells bound to the P-RIIs expressed in one patch.

Second, the eliminon might be taken up only as an antigen–antibody complex, thus making the presentation one of non-self-eliminons only (12, 46). If signaling occurs in a limited defined enclave, on average, only one non-self-eliminon will be present or, on average, the simultaneously presented non-self antigens might be effectively dealt with by the same ridding mechanism, or ARA might be superimposed at a later decision stage. Whether or not these suggestions are viable, the problem of specificity of signaling must be faced. Using the dendritic cell as a non-specific source of Signal[2] or [3] (i.e. costimulation or cosuppression) is not a viable solution (35).

Autoimmunity versus immunopathology.

In the absence of T_s, the effector response overshoots because the eliminon is ridded as a viable entity long before its derivative immunogenic components. Therefore, most of the disease consequences in a T_s knockout are expected to be due to immunopathology, not autoimmunity. Any apparent specificity for a given tissue-target assayed by its pathology is likely due to the greater sensitivity of the tissue to non-specific bystander destruction. For example, it would be expected that rapidly self-renewing tissues like fibroblasts and epithelial cells would display greater resistance to immune attack than endocrine and neuronal cells which are poorly renewing.

By way of illustration, under the ARA model any self-component that appeared in the periphery after the developmental time window closed would be indistinguishable from non-self and be a target for autoimmunity. In order to deal with this unavoidable problem, evolutionary selection ectopically expressed these components in thymus while the window was open under the control of the transactivating transcription factor, Aire (discussed in 23). This purged the T_h specific for those components leaving all other immune cells that recognized them inactivatable only. Under the ARA model, an Aire knockout would result in autoimmunity to these delayed expression self-components (see detailed discussion in 12). This is to be contrasted to a situation in which T_s are deleted. If, as postulated, T_s regulates the magnitude of the response to non-self, then the result in a T_s knockout would be immunopathology, seen initially in the most sensitive tissues.

These two situations may have similar symptoms and histopathology, but they are conceptually miles apart.

The difficulty with this conceptualization lies in its translation into experimental test. Autoimmunity would require the demonstration of an antibody or other effector activity specific for the self-target that is absent in normal animals. Failure to find it would be a weak argument but would favor immunopathology. The tissue histology cannot distinguish the two processes; only the specificity of an immune effector response can be informative.

All this having been said, the use of T_s as a clinical tool has a greater chance of success than attempts to manipulate the sorting of the repertoire (e.g. attempts to ablate specifically T helpers anti-‘a self-component’ responsible for the autoimmunity). If the target of the autoimmune disease is known, then T_s specific for linked epitopes can be used to reduce the magnitude of the response to a point where no autoimmune clinical manifestation is evident. The use of random populations of T_s may accomplish the same result but it would be at the expense of a shutting off of the responsiveness of a large part of the non-self repertoire. This side-effect would be unavoidable if the disease were due to immunopathology and poses the danger of leaving the individual unprotected from infection.

Identifying the trees in the forest of T-regulatory biology

• (i) The presence or absence of eT_hs is the determinant of ‘natural’ tolerance (i.e. the sorting of the repertoires). T-suppression cannot regulate natural tolerance.

• (ii) The repertoire of T_s is sorted to be anti-non-self peptide [anti-(Pns-RII)] as is that of T helpers. Their naive repertoires are identical.

• (iii) T_s down-regulate the magnitude of the biodestructive and ridding effector response; T helpers up-regulate it.

• (iv) eT_hs (source of Signal[2]) are required for the activation and induction of effector T_s. The primary target of effector T_s function is the antigen-specific inhibition of induction of eT_hs.

• (v) The signaling pathways between effector T_s and their T helper targets require associative recognition of the antigen (ARA).

• (vi) Most of the disease patterns consequent to ablation of T_s are due to immunopathology, not autoimmunity. Autoimmunity is a T helper-dependent aberration.

• (vii) In the framework of a general extrapolation, we can expect a family of regulatory cells to be delineated that interact with each other in a Yin (turn-on)–Yang (turn-off) fashion (e.g. T_h1 versus T_h2), the evaluated output of which will be the signals that determine the proliferation and differentiation of activated defensive cells to effectors appropriate both with respect to class and magnitude. The activation of all regulatory T cells will require a Signal([1]+[2]) from an eT_h and the induction of expression of that eT_h will be regulated by an eT_s. The input signals to this regulatory family will be derived from the cytopathic consequences of the infection. The cross talk between regulatory and defensive cells will be antigen specific mediated via ARA to result in a biodestructive and ridding response that is coherent and independent for each pathogen. The feedback mechanism controlling the magnitude of the response can be expected to operate at two levels:

  the antigen-specific inhibition by effector T_s of the induction of T helpers to effectors and the end products of the defensive effectors (e.g. inhibitory antigen-antibody complexes or interleukins that act at the level of effector output).

The mechanisms regulating class (Module 3) are, in this framework, entirely germline selected unlike those of the self-non-self discrimination (Module 2) which are somatically learned, historical processes.

Addendum

Are T_s (Tregs) engaged in the maintenance of self-tolerance?

This probing question raised by the reviewer must be considered at several levels.

• (i) Are T_s (Tregs) sorted to be anti-self or anti-non-self?

• (ii) Can tolerance be maintained at the level of effectors mediating biodestructive and ridding activities?

• (iii) What failure in the principle mechanism of sorting of the repertoire becomes a selection pressure for a self-specific mechanism of T-suppression?

• (iv) Is there an alternative explanation of the data claiming to support a role for T-suppression in maintaining tolerance?

A summary of the above discussed answers in the ARA framework would be (i) T_s (Tregs) are sorted to be anti-non-self. (ii) Tolerance which is epitope specific cannot be regulated at the level of effector output which is antigen specific. (iii) There is no failure in the “normal” evolutionarily selected mechanism of the self-non-self-discrimination but it has a fundamental limitation. Per force there is a steady state level of anti-self cells on the pathway to inactivation (the autoimmune boundary), from which pool diversion to activation can lead of autoimmunity. (iv) This question deserves a detailed review publication by publication, but in brief.

The depletion of Tregs (T_s) or a mutation that results in what clinically is a dominant syndrome like type 1 diabetes, or lupus, or colitis does not, in itself, demonstrate breaking of tolerance, the conceptualization behind autoimmunity. An effector attack on a virus harbored by the pancreatic cells producing digestive enzymes could generate a localized inflammation that in mice of a given genotype would destroy the β-cells producing insulin because they were uniquely sensitive. A demonstration that a specific immune response to a germline encoded target expressed in the β-cell would be required as a minimum. The adoptive transfer of T cells from an animal that has succumbed to B-cell destruction and that reproduces the diabetes in another animal of the same genotype does not demonstrate that a β-cell specific response was elicited by Treg depletion. It is not clear why it would be expected.

If the T_s (Tregs) are sorted to be anti-non-self, then their depletion can only result in immunopathology as the dominant pathology. If they are sorted to be anti-self, then a mechanism permitting a response to pathogens that share epitopes with self is required. If the Tregs are unsorted, how would they deal with these crossreactive non-self-antigens, not to mention the general question, what were they selected to regulate, the response to infectious agents or the response to the host? They regulate responsiveness without regard to antigen specificity, what role does an antigen-specific MHC-restricted receptor play? Lastly, if unsorted Tregs (T_s) exert ‘a tonic suppression on immune responses in general’ then their depletion would be expected to result dominantly in immunopathology and, to a minor extent, autoimmunity.

Immunopathology is driven by an anti-non-self response. As the animal is constantly under a non-self immunogenic load, the release of feedback regulation would result in an overshooting of anti-non-self activity with concomitant immunopathology. Autoimmunity is driven by an anti-self response. The source of the anti-self cells would be from the autoimmune boundary, the cells on the pathway to inactivation. Even if one postulates that Tregs anti-self maintain the level of these cells acceptably low, the release of this feedback inhibition by depletion would require a mechanism to induce them to effectors. The consequence would be generalized autoimmunity as well as immunopathology visualized experimentally as the biodestruction of the most sensitive tissue. This could well be mistakenly interpreted as a specific autoimmune target.

Finally, given that Tregs (T_s) only recognize epitopes (peptides), if they are to specifically regulate responsiveness to antigens (collections of linked epitopes), they must function in ARA. This requires that they be sorted to be either anti-self or anti-non-self. If Tregs play their principle role in ‘tonic suppression’ of global responsiveness then its TCR need not be engaged. In this case, the Treg cannot contribute to the breaking of tolerance to a given self-target. If they, in addition, ‘regulate individual discrete immune responses’, then in order to be involved in breaking tolerance they must be sorted to be anti-self and the question is raised, how is the Treg anti-a-given-self-inactivated so as to give rise to specific autoimmunity.

Funding

National Center for Research Resources (RR07716).

Glossary

Abbreviations

APC

antigen-presenting cell

ARA

associative recognition of antigen

BCR

B-cell antigen receptor

eT_h

effector T helper

ns/NS

non-self

P

peptide

Ps

anti-self P

Pns

anti-non-self P

R

MHC-encoded restricting element

RI

Class I MHC that restricts T_c

RII

Class II MHC that restricts T_h

T_c

cytotoxic T cells

T_s

T suppressors

s/S

self