What would Treg-cell biology look like when viewed from a rationalized perspective?

Tolerance versus unresponsiveness

The experimental manipulation of an animal so that it becomes specifically unresponsive to a given antigen, will be referred to as a state of “unresponsiveness.” The extrapolation of that observation to a theory of the S-NS discrimination will be referred to as a state of “tolerance.” Unresponsiveness is observation; tolerance is theory. Stated differently, “tolerance” is defined by a theory of the S-NS discrimination.

S versus NS

The terms S and NS have become a semantic juggler's act 1, yet if one wishes to communicate there is no escaping their use. I tried to introduce the discrimination not-to-be-ridded (NTBR)–to-be-ridded (TBR), but without success, leaving no choice but to define the terms, S and NS, as unambiguously as possible.

S and NS are defined by the immune system, not by the immunologist. For the germline selected (innate) immune system, NS is distinguished from the self-of-the-species; for the somatically selected (adaptive) immune system, NS is distinguished from the self-of-the-individual. Here, we are only concerned with the “adaptive”–-and therefore somatically selected—system.

In order to anticipate the unexpected in the antigenic universe, the adaptive immune system generates a repertoire that is large and random with respect to the property, S or NS. This repertoire contains specificities (anti-S), which if expressed would debilitate the individual, as well as specificities (anti-NS), which if not expressed would leave the individual unable to cope with harm. The adaptive immune system of each individual learns the distinction between S and NS during a developmental time window when all S is present and no NS, and the system is tolerizable—only due to the absence of effector T-helper (eTh) cells. The state of tolerance to S is maintained as long as S persists. The repertoire must be purged of anti-S (NTBR) leaving anti-NS (TBR) to protect the individual. S and NS are defined by this sorting process.

Not all autogenously generated antigen appears as S to the adaptive immune system. There is a class of autogenously generated antigen that is NS to the immune system. These are largely waste products from various sources, necrosis of cells, wound débridement, denatured, and effete protein, etc. Ridding them is beneficial to health (TBR) and these antigens are defined by the immune system as NS. Ridding any other autogenously generated antigen is detrimental and defined by the immune system as S (NTBR). Immune ridding of the NS targets (TBR) of housekeeping is referred to as “autoreactivity,” whereas the ridding of the S targets (NTBR) results in debilitation referred to as “autoimmunity.” In sum, autoreactivity is beneficial and selected for, whereas autoimmunity is detrimental and selected against. Any theory of the S-NS discrimination must explain how the immune system learns to distinguish autoreactivity from autoimmunity, which is an aside in this Commentary. There are residual ambiguities in this formulation that involve antigens in sequestered sites, such as brain, eye, gut, fetus, etc., but these need not be of concern here. The above definitions as well as a detailed model accounting for this somatic learning process have been discussed in detail 2-4 and only directly relevant aspects will be dealt with.

Treg cells are T-suppressor cells

All biological processes have inhibitory feedback mechanisms that regulate the magnitude of their outputs 5. In the case of the adaptive immune system, that is the postulated role of Treg cells. However, the term “Treg cell” is very poorly chosen because T-helper (Th) cells are just as regulatory as Treg cells, the phenotype of which, namely suppression, is obscured. I will therefore use the symbol “Tsu” instead of what is referred to in the literature as “Treg cells” in order to emphasize their function.

Autoimmunity versus innocent bystander immunopathology

Autoimmunity implies the breaking of tolerance by induction of a self-generating level of eTh-cell anti-S activity. The specificity of the attack on S is determined by the specificities of the TCR/BCR. Innocent bystander immunopathology is due to the unspecific spilling over of the effector functions that attack bystander tissues. Distinguishing the two processes experimentally might sometimes be difficult because the differential sensitivity of tissues to an effector attack can imitate specificity, but conceptually they are worlds apart. In any case, the interpretation of any given experiment requires making this distinction. The overshooting of an anti-NS response results in innocent bystander immunopathology, the regulation of which is the proposed role of T-suppressor (Tsu) cells (Treg cells).

Précis of the theoretical framework

T cells are born in an enclave, the thymus, where they undergo a positive selection that establishes their MHC-restriction specificity and phenotype (cytotoxic Tc or helper Th cell), as well as a negative selection that purges a significant proportion of their anti-S repertoire. Thymic-S consists not only of endogenous S (endoS) but also of a subset of peripheral S that is ectopically expressed in thymus (ectoS) under the control of the transcription factor, Aire. The repertoire of Th/c cells that leave the thymus is largely anti-NS, with a small proportion, which are anti-peripheral S (perS). The perS antigens are not expressed in thymus, only in the periphery. Thus there are two steady states established by this efflux of Th/c cells, one anti-NS, the immune boundary, the other anti-perS, the autoimmune boundary. The Th/c cells in the autoimmune boundary encounter their perS ligands and undergo negative selection, whereas the Th/c cells in the immune boundary await an NS-ligand in order to respond, which if not encountered, results in their turnover. Thus there is a two-stage purging of anti-S from the Th/c repertoire, mostly in the thymus, the remainder in the periphery. Whether Tsu (Treg) cells follow this pathway is what is under debate.

The cells that leave the thymus will be referred to as initial state or iT cells. Here, we will be concerned with T-helper (Th) cells and Tsu cells, both of which are class II MHC-restricted. When the TCRs of iTh or iTsu cells encounter a ligand (Signal 1), the cell is signaled to undergo differentiation to an anticipatory state, symbolized as aTh or aTsu, where a decision is made between inactivation and activation, which is the first step on the pathway to effectors, eTh or eTsu cells. Activation requires a second signal (Signal 2) to tell the aT cell which pathway to take. It might be stressed that both aTh and aTsu cells require Signal 2 delivered by an eTh cell to be activated 6. The consequence for Tsu (Treg) cells is that its repertoire must be sorted like any other T cell by a negative selection purging anti-S, leaving a functional Tsu anti-NS repertoire. This two stage–two signal process is the product of pure logic, requiring eventually that the two stages and two signals be mapped onto cells and mechanism. In sum, Signal 2 delivered by an anti-NS eTh cell to an a-cell puts it on the activation pathway. Signal 1 alone results in inactivation (i.e. negative selection). Signals 1+2 result in activation. The requirement for two signals in order to activate guarantees that no cell can be activated that in principle could not have been inactivated, an important safeguard against autoimmunity.

Stated succinctly, the recognition of S is purged epitope by epitope; the induction and regulation of the effector response to NS is mediated antigen by antigen. Antigens are combinatorials of linked epitopes, the recognition of which requires knowledge of the combinatorial. Now we are in a position to ask, “What is the role of effector T-suppressor (eTsu) cells?”

Tsu (Treg) cells cannot determine tolerance; they can only establish unresponsiveness

As a consequence of a central and peripheral negative selection process, the functional peripheral T-cell repertoires of both the cytotoxic (Tc) and helper (Th) T cells are anti-NS. In order for the Tsu cells to function by purging the anti-S from the repertoires of iTc cells, iTh cells, and iB cells, they must be postulated to be selected as being solely anti-S. This is contrary to fact as there are numerous examples of the inhibition of responsiveness to NS (e.g. bacteria) by eTsu cells (e.g. 7). If the Tsu repertoire were left unsorted, a contradiction in function would arise. Tolerance requires almost complete shutdown of the response to S. An unsorted Tsu repertoire cannot distinguish S from NS. Consequently, a shutdown of the response to NS would be lethal. Therefore, the key question to settle is whether the Tsu repertoire is, in fact, sorted to be anti-S or anti-NS or left unsorted. The most reasonable answer is that the Tsu repertoire is somatically negatively selected in thymus to be anti-NS like the repertoires of all other classes of i-cells. Given this assumption, how does this assumption deal with all of the studies showing that Tsu cells can alleviate the symptoms of a variety of autoimmune diseases, the most frequently used argument permitting the extrapolation of the observation of unresponsiveness to tolerance?

Autoimmune disease arises as a consequence of the breaking of tolerance. A normally tolerant animal is one in which anti-S eTh cells are lacking. Tolerance is broken by inducing anti-S eTh cells to a level that is autogenerative, driven by the given S-antigen. The breaking of tolerance at the level of eTh cells also breaks tolerance at the level of the other classes of T and B cells, which in turn contribute to the effector attack described as autoimmunity. The activation of a-cells is eTh-cell-dependent. The effector response of the immune system treats an S-antigen to which tolerance has been broken as being indistinguishable from an NS-antigen. If the repertoire of Tsu cells is selected to be anti-NS, then the anti-S Tsu cell that has been derived experimentally, and which has been used to create various levels of unresponsiveness during an S-antigen-specific autoimmune episode, must be derived from an anti-S Tsu cell in the autoimmune boundary. In other words, tolerance must be broken at the level of the Tsu cell exactly as it is at the level of the Th cell. If Tsu cells are somatically selected to be anti-NS, the protection against autoimmunity by Tsu (Treg) cells cannot be used as an argument that the evolutionary (germline) selected role of Tsu cells is to establish and maintain tolerance. An S-antigen under autoimmune attack is indistinguishable from an NS-antigen. The breaking of tolerance in the Tsu and Th repertoires involves families of antigens, the epitopes of which, if specificity is to be respected, would be seen in linked recognition or its equivalent by Tsu cells and Th cells. The experimental manipulation of a Tsu cell so that it renders an expected immune response unresponsive, whether that response is autoimmune or not, cannot be extrapolated to argue that the Tsu cell determines “tolerance.”

There is another aspect worth considering. The TCR is polyreactive 8, meaning that it signals the cell on binding chemically distinguishable peptide ligands that are random with respect to the property S or NS. The MHC-encoded restricting element (R) presents the TCR with a peptide (P) roughly 10 amino acids in length, of which five are buried in the anchoring groove and five provide potential epitopes for the TCR. This means that the maximum epitopic repertoire recognizable by the TCR is capped at 20⁵ (3.2e6). Individual members of the TCR family can reasonably be viewed as responding to 1 or 2 or 3 or 4 or 5 of the exposed amino acids. Using the extremes for illustration, a TCR that recognizes as signaling, one amino acid in a fixed position in P, will be able to bind 20⁴ distinguishable peptides, whereas a TCR that recognizes five amino acids in the peptide will be able to bind only one peptide. From the point of view of the TCR, it is unireactive; from the point of view of the immune system, the TCR is polyreactive ranging from 20⁴ to one. Now, if one defines a specificity index (si) as the probability that a random epitope is S, and if si = 0.01 is a reasonable value, then roughly 60% of naïve thymic T cells will be anti-S (i.e. bind at least one S peptide) and 40% anti-NS (i.e. bind no S peptide) 9. Negative selection of anti-S will skew the peripheral paratopic repertoire toward low polyreactivity of the residual anti-NS 1. Clearly, a TCR that signals on binding 1 amino acid has a much higher probability of finding it in an S-peptide than one that binds 5 amino acids.

For completeness, if suppression by Tsu (Treg) cells were to determine tolerance, individuals with adaptive systems would accept, not reject grafts from other members of the same species. I recall that individuals without adaptive immune systems accept such grafts. As a more general way to view this point, the individual would be unresponsive to NS-antigens that share epitopes with S. Given that roughly 10% of NS-antigens are cross-reactive with S, this would be lethal, not to mention contrary to fact. Further, if the Tsu repertoire included recognition of S-epitopes presented on the antigen-responsive i-cells themselves, no immune response would be possible. Lastly, iTsu cells themselves cannot be suppressible by eTsu cells. Sorting of the iTsu repertoire can only be accomplished by S-driven negative selection like that operating on all other T cells, helper or cytotoxic. If all antigen responsive cells require negative selection to make an S-NS discrimination then postulating a role for suppression in determining tolerance becomes, at best, gratuitous.

Stated succinctly, Tsu (Treg) cells regulate the response to NS, not to S. Autoimmunity is a condition where S is converted into NS. The mechanism of this conversion is not the one used to establish and maintain tolerance. Tolerance engages epitopes; conversion engages antigens.

Challenging the validity of above arguments

I will analyze two thought-provoking investigations, one of which appears to pose formidable contradictions to the above framework, making it necessary to deal with it in nitty-gritty detail.

Guerau-de-Arellano et al. 10, by using a technique that allowed them to shut off Aire function in a temporally controlled manner, were able to demonstrate that the failure to express Aire in the perinatal period resulted in multiorgan autoimmunity, whereas the failure to express Aire somewhat later resulted in few deleterious consequences. They conclude that a “previously tolerized T-cell pool can buffer newly generated autoreactive T cells that might arise.” As the mechanism for this buffering is, to say the least, obscure, I refer back to an earlier suggestion that the evolutionary selection pressures for the ectopic expression of peripheral S (ectoS) in thymus are twofold: (1) to deal with a peripheral subset of S that is delayed in expression as a peptide class II MHC complex 11, and (2) to prevent the swamping of peripheral negative selection mechanisms 6 (see above).

Peripheral S-antigens, the expression of which is delayed in development until after the immune system is functional, cannot be distinguished from NS-antigens, inevitably resulting in autoimmunity. As S-antigens are selected to function in the physiology of the organism, not to escape the immune system, the existence of delayed expression S would be no surprise. Ectopic thymic expression in the period preceding the appearance of a delayed peripheral S would establish tolerance to it. Once established, the appearance of the delayed S in the periphery would take over to maintain the state of tolerance independently of ectopic expression in thymus (i.e. Aire independently).

To return now to the reasoning of Guerau-de-Arellano et al., they ask three central questions, (1) What is so special about the neonatal period; (2) Why do not the anti-S T cells leaving an adult thymus that lacks Aire expression trigger autoimmunity; (3) Why is only a restricted set of tissues involved?

To answer to the first question in the framework of the two stage–two signal model outlined above, during the neonatal period in mouse, the immune system switches from a tolerizable-only state to one that is both tolerizable and inducible due to the appearance of anti-NS eTh cells 12. Tolerance can be maintained only as long as peripheral S persists; it cannot be established for peripheral S that appears de novo after the system is responsive. The T cells leaving the embryonic/neonatal thymus that are anti-peripheral S are negatively selected by interaction with peripheral S because the neonatal periphery lacks anti-peripheral S eTh cells. This deficit lasts as long as Signal 1-driven negative selection by peripheral S is maintained. As for the second question, Aire-determined expression of ectopic S (delayed expression peripheral S) in thymus initiates the establishment of tolerance by negative selection. Why, once established, can this particular Aire function be bypassed? The perinate lacks expression of delayed peripheral S and depends on negative selection by ectopic S in thymus to maintain tolerance to it, until it appears. In the adult, the expression of delayed S-antigen in the periphery takes over to maintain the state of tolerance making negative selection in thymus by ectopic S expression a redundancy. This brings us to the third question. Why is the autoimmune disease characteristic of Aire-KO mice restricted to such a limited set of tissues? The major targets of the autoimmunity would be expected to be among those that were ectopically expressed in Aire-WT thymus. It is suggested that these S-antigens are germline-selected to be ectopically expressed in thymus (where there is an absence of eTh cells) because they are delayed in expression in the periphery. This is the only way to avoid autoimmunity to them. All of the findings of Guerau-de-Arellano et al. are reasonably explained by this model.

“Striking similarities” between the autoimmune disease resulting from neonatal thymectomy and the neonatal shutting off of Aire expression are stressed by these authors 10. Neonatal thymectomy terminates all efflux of iT cells from thymus. The resultant autoimmunity must derive from cells that were peripheralized prior to thymectomy. These cells responsible for the autoimmunity are anti-NS/perS iTh cells. Neonatal Aire inactivation differs from neonatal thymectomy in that, for the former, there is continued efflux of cells, a high proportion of which are anti-NS/ectoS/perS iTh cells, whereas for the latter there is not only a lymphopenia but the peripheralized cells express a repertoire that has a small proportion of anti-S that is directed against the peripheral S that is not expressed in thymus (perS). Lymphopenia has a tendency to activate a-cells, independently of eTh cells. Therefore, neonatal Aire inactivation, when examined closely, would be expected to result in an autoimmunity to the set of delayed expression targets constant between individuals, whereas thymectomy would favor innocent bystander pathology as well as a set of autoimmune targets variable between individuals. Innocent bystander pathology can appear to have specificity due to the variable sensitivity of tissues to different effector mechanisms. The pathologies and targets of perinatal thymectomized and Aire inactivated animals would be expected to be distinguishable. It is not the “striking similarities” but the predicted striking differences that need to be searched for here.

The above investigation was extended by Yang et al. 13 who introduce evidence that they interpret to demonstrate a role for Tsu (Treg) cells in establishing and maintaining peripheral S-tolerance. As this assumption is viewed as ruled out here, it becomes crucial to analyze their findings. In order to do this, the key parameters to keep in mind are the selected specificities of the Tsu and Th repertoires, as well as the relevant S targets. These targets are defined by the anti-S eTh cells that initiate the autoimmunity. Under any theory of the S-NS discrimination based on the assumption that the eTsu function to inactivate anti-S (i.e. to inhibit autoimmunity), they must, as a minimum, regulate the magnitude of the anti-S eTh-cell response to an ineffectual level, and this process requires that thymic selection be for anti-S Tsu cells, essentially what Yang et al. imply.

In an initial experiment, Yang et al. confirm the findings of Guerau-de-Arellano et al.: that late inactivation of Aire does not result in an autoimmunity characteristic of early inactivation of Aire. They then show (1) that perinatal but not adult depletion of Tsu cells in Aire-WT mice results in an autoimmunity resembling that of Aire-KO mice, and (2) that perinatal, but not adult Tsu cells from Aire-WT mice can protect against the autoimmunity characterizing Aire-KO mice. These two observations form the basis for their argument that Tsu (Treg) cells function to maintain peripheral S-tolerance, and by extension, that they are thymically selected to be anti-S.

Any interpretation of these two observations requires:

Now we can examine the first observation, why does depletion of Tsu cells from Aire-WT perinates, but not from adults, result in autoimmunity? As autoimmunity is antigen-specific, the Tsu and the Th cells must see the same S-antigen. The iTh cells exiting the Aire-WT thymus, perinatal or adult, are anti-NS/perS. If the Tsu cells were positively selected by Aire-WT thymic-S, they would be anti-endoS/ectoS and, therefore, would be incapable of specifically suppressing an anti-perS eTh-cell response potentially responsible for mediating autoimmunity. If the Tsu cells are negatively selected like all other T cells, they would be anti-NS/perS and, therefore, capable of suppressing an anti-perS eTh-cell response but that would be a redundancy. Positively selected Tsu cells acting as suppressors of autoimmunity, cannot explain why autoimmunity directed at perS to which they are blind, is not detected in Aire-WT mice. Negative selection of anti-perS iTh cells by perS in the periphery, obviates the need for anti-perS Tsu cells in establishing or maintaining tolerance. They would be viewed simply as an unavoidable byproduct. The key here is that consideration of the specificity of thymic selection does not predict that depletion of perinatal Tsu cells in the Aire-WT would result in autoimmunity. What about the failure of depletion of adult Tsu cells to provoke autoimmunity? Why do perinatal and adult depletion of Tsu cells result in different outcomes? As there is little reason to suspect that the two thymi, perinatal and adult, present different categories of S, the difference must reside in the periphery and, if so, would have to be due to the postulated appearance of delayed expression S. I will develop that argument after a discussion of the second observation, as they are related.

Why do perinatal, but not adult Tsu cells from Aire-WT mice, protect against the autoimmunity expressed by the Aire-KO mouse? If the Tsu cells are positively selected in Aire-WT perinate thymus, they would be anti-endoS/ectoS. If the Tsu cells are negatively selected, they would be anti-NS/perS. The iTh cells responsible for the delayed autoimmunity, exit the Aire-KO thymus as anti-NS/ectoS/perS. Therefore, only positively selected anti-ectoS Tsu cells can, in principle, block the autoimmunity due to these Th cells. This implies a contradiction between the first observation that required negative selection of Tsu cells, and this finding which requires positive selection of Tsu cells. Further, as there is no difference between adult and perinate with respect to these parameters, once again no difference between the behavior of perinate and adult Tsu cells is expected.

How should we deal with these contradictions between argumentation based on specificity of thymic selection and the Yang et al. interpretation of their observations? What might a reinterpretation of these experiments look like?

The finding of a difference between depletion of Aire-WT perinatal and adult Tsu cells suggests that depletion of Tsu cells has an effect that deceptively mimics the Aire-KO pathology and is unrelated to any putative role of Tsu (Treg) cells as regulators of tolerance. The perinate and the adult differ in that ectoS (i.e. the delayed expression S) is expressed in the periphery of the adult, not in that of the perinate. The perinatal peripheralized anti-NS iTh cells that do not encounter antigen accumulate to a steady-state level. There they undergo receptor revision creating anti-ectoS iTh cells that accumulates due to the absence of ectoS (i.e. they become intolerant). The experimental depletion of all perinatal Tsu cells removes the feedback regulation of all effector responses. The steady-state anti-NS/ectoS eTh response cascades essentially out of control, creating two problems: (i) innocent bystander immunopathology as well as (ii) autoimmunity aggravated in a polyreactive system by the additional unavoidable coupling by NS of anti-NS eTh cells in the immune boundary to the anti-S aTh cells in the autoimmune boundary 9. The perinatal peripherally derived polyreactive anti-ectoS iTh cell can be induced both by delayed expression S (ectoS) and by NS itself. As this multiplicity of effects manifest themselves as a multitissue, multimechanism pathology, the antigen-specific components required to interpret the findings of Yang et al. are obscured. The pathology will include, but cannot be identical to, the defined autoimmunity manifested in Aire-KO mice. Assays such as weight loss or histology are too far removed from the specificity elements needed to interpret their results at the level of tolerance. In the Aire-WT adult, where ectoS is peripherally expressed, negative selection deleting the peripherally derived anti-ectoS iTh cells as they appear, masks to a limited extent the indirect effect of global Tsu-cell depletion, explaining why in adults Tsu-cell depletion is less dramatic when compared to perinatal Tsu-cell depletion. A role for Tsu (Treg) cells in the S-NS discrimination by the Aire-WT animal that establishes or maintains tolerance in the Th population is in no way indicated.

What about the second observation that there is a difference between Aire-WT perinatal and adult Tsu cells in that perinatal but not adult Tsu cells protect the Aire-KO mouse from autoimmunity? The negatively selected iTh cells exiting the Aire-KO thymus are anti-NS/ectoS/perS. For Tsu cells to protect against the autoimmunity to delayed S (ectoS) they must be anti-ectoS. This finding would be consistent with the existence of a positively selected Tsu cells from the Aire-WT thymus, perinatal or adult, as they would be anti-ectoS. The exited perinatal anti-ectoS Tsu cells do not encounter delayed S (ectoS) and accumulate to a steady state level ready to respond when the delayed S (ectoS) appears. The adult Aire-WT anti-ectoS iTsu cells encounter ectoS in the periphery and are Signal 1 inactivated (tolerized) by it as they exit. Consequently, they cannot protect the Aire-KO mouse. Unfortunately, this straightforward explanation is not entirely satisfactory as it is questionable whether Tsu cells are positively selected to be anti-ectoS in thymus. Further, the required assumption that Tsu cells are negatively sorted in the periphery is antithetical to their playing a suppressive role as sole mediators of tolerance. If Tsu cells are negatively sorted in Aire-WT thymus then they would exit as anti-NS/perS and be unable to regulate the anti-ectoS eTh-cell autoimmune response of the Aire-KO mouse. However, the peripheralized perinatal Tsu cells from the Aire-WT mouse do not encounter the delayed expression S (ectoS). During the window when these ectoS-antigens are not expressed peripherally in the Aire-KO mouse, they gain recognition of them, probably as a consequence of receptor revision 9. When the delayed S (ectoS) appears and induces an anti-ectoS eTh-cell response, the peripherally derived perinatal anti-ectoS iTsu cells can respond to inhibit it. From the point of view of these perinatal Tsu cells, they are regulating a response to NS. The adult anti-NS iTsu cells encounter the delayed S (ectoS) in the periphery and any derivative anti-ectoS Tsu cells due to receptor editing are inactivated as the arise. Consequently, adult Tsu cells from Aire-WT mice are blind to ectoS and cannot inhibit the anti-ectoS eTh-cell autoimmune response of the Aire-KO mouse. Most importantly, negative selection of anti-S Tsu (Treg) cells in the periphery is predicted.

It might now be asked, why do not the Tsu (Treg) cells from the Aire-KO mouse itself protect it from the autoimmunity driven by anti-ectoS eTh cells? Positively selected Tsu cells would be anti-endoS and irrelevant. Negatively selected Tsu would be anti-NS/ectoS/perS and, in principle, capable of controlling an anti-ectoS eTh-cell autoimmune response. The anti-ectoS Tsu cells exiting the thymus before the peripheral expression of delayed S (ectoS) would simply turnover. After the expression of delayed S (ectoS) the anti-ectoS Tsu cells would be negatively selected as they exit and, therefore, ineffectual. In addition, the early exited or perinatal Tsu cells would be diluted out by the late exited or adult Tsu cells, leaving their suppressive function minimal.

In essence then, all of the observed differences in responsiveness between perinatal and adult Tsu cells can be attributed indirectly to negative selective processes in the periphery by the delayed expression subset of peripheral S. These observations are not extrapolatable to “tolerance” as defined by the S-NS discrimination.

In essence then, in the two stage–two signal framework, Tsu cells function normally to regulate the magnitude of the effector response to NS. The immunologist can manipulate the immune system so that an S-component behaves as an NS-component, and anti-S Tsu cells can be fished out of the autoimmune boundary and presented at an artificially high concentration to inhibit a given autoimmune response to below visible pathology. In other words, although Tsu cells can be experimentally manipulated to ameliorate an autoimmune response, as a contribution to “understanding,” the extrapolation of this observation to “tolerance” as defined by the S-NS discrimination, if possible, has yet to be properly expounded.

It is not clear to me what is the take home lesson (or extrapolation to tolerance) that Yang et al. want us to consider. In the framework of Tsu cells determining tolerance, Tanaka and Sakaguchi 14 in their perspective on the Yang et al. study, assume that Tsu (Treg) cells “are specialized for suppressing peripheral activation and expansion of those self-reactive T cells that have escaped elimination in the thymus.” This assumption needs a clarification of the term “escaped elimination.” If the self-reactive T cells escape elimination because the negative selective process has limitations then the pattern of autoimmune targets would be expected to be random between individuals. If the self-reactive T cells escape elimination because a given peripheral self-ligand is not expressed in thymus and delayed in expression in the periphery, then a defined pattern of autoimmunity seen in all individuals would emerge. In order to interpret the Yang et al. findings, we are clearly dealing with the latter and an analysis of the type carried out above is necessitated. Tanaka and Sakaguchi interpret this study as having shown that “a specific population of Treg cells produced particularly early in life is highly efficient in preventing autoimmune disease and sustaining stable self-tolerance 14.” Further, in their view “Yang et al. have shown a new link between Treg-cell development and the function of Aire during perinatal life.” The authors do not deal with the assumptions needed to interpret this description in terms of lineage, selection or specificity, nor for that matter, with what the adult Treg cells see and do.

In this same framework, we might explore whether a meaningful model can be generated by assuming that there exists two classes or lineages or phenotypes of Tsu cells, one a perinatal anti-S population positively selected in both thymus and periphery, completely inhibitory of a response to any S (Tsu1), the other a peripherally derived anti-NS population acting as a regulatory feedback mechanism permitting an optimal, but not run away effector response to NS (Tsu2). Tsu1 cells responsible for tolerance would have to be positively selected by all S, thymic, and peripheral, to be adequately anti-S, whereas the Tsu2 cells would have to be negatively selected by all S to be adequately anti-NS. If Tsu2 cells are derived in the periphery from the CD4⁺ Th lineage, they would have the same negatively selected repertoire, anti-NS, and pose no problem as regulatory elements. However, Tsu1 cells pose a contradiction. They leave the thymus as anti-endoS/ectoS iTsu1 cells. They do not recognize perS and therefore cannot determine tolerance to it. If it is argued that they can be peripherally positively selected by perS, then they must have a mechanism to make an S-NS discrimination as the periphery includes the normal immunogenic NS load and the shutting down of a response to it would be lethal. If it is argued that there is no need for Tsu1 cells to see perS (i.e. selection ceases when they exit) because the Th cells potentially responsible for autoimmunity to perS have been negatively selected to ignore it, then any proposed role for Tsu1cells in establishing tolerance becomes gratuitous. Further, the Tsu1/Tsu2 model leaves us with no role for the T-helper cell. It obviates the need for an eTh-cell-driven Signal 2 in making the S-NS discrimination. Signal 1 would be activating and Signals 1 + 2 delivered by eTsu cells would be inactivating. Such a model would require reinterpreting a mountain of data to the contrary, namely that T-helper cells play an indispensible inductive role. Lastly, the Tsu1/Tsu2 model as formulated above does not predict a difference between perinatal and adult Tsu cells without adding ad hoc tweaks. A two lineage Tsu1/Tsu2 model under which the Tsu1 cell determines tolerance and Tsu2 feedback regulates magnitude appears as untenable.

In sum then, we are left with findings, which are in need of an explanatory theory that competes with that provided by the two stage–two signal model. Given this, and all things considered, I believe it fair to conclude that Tsu (Treg) cells are negatively somatically selected, as are all other T and B cells, to be anti-NS, and contribute to regulate the magnitude of the effector response, not the S-NS discrimination (i.e. tolerance).