Does T Cell Activation Require a Quorum of Lymphocytes?

The stochastic nature of the processes involved in the generation of diversity of Ag-specific receptors (1, 2) inevitably leads to lymphocytes with receptors that recognize self-antigens. Indeed, such autoreactive, mature lymphocytes have been detected in healthy individuals (3, 4), and their activation in patients can result in the pathological state known as autoimmunity. How, then, does the immune system avoid autoimmunity in healthy individuals? This topic has been the subject of intense investigation over the last 60 years. Much has been learned. Older reports inevitably involved input/output studies, in which Ag was given to intact animals and immune responses were measured, or more sophisticated studies of a similar kind. These reports inevitably reflected investigations at the level of the system because so little was known, and envisaged, at the cellular and molecular levels. Contemporary studies, in contrast, tend to be much more analytical, studying various phenomena at the cellular and molecular levels and employing very powerful and often recently devised technology. The former and older type of study is valuable in being less obtrusive than the latter type but leads to more abstract conclusions. It is the sign of pivotal progress when different approaches, and different kinds of analysis, lead to a coherent picture. We suggest that this end can be facilitated by revisiting and reviewing different studies, reported in different decades and primarily exploiting different levels of analysis, in the attempt to achieve such a coherent picture.

We were stimulated to write this article by recent reports (5, 6) suggesting the possibility of a novel mechanism of peripheral T cell tolerance. These reports suggest that if a quorum (i.e., a minimum number) of T cells is needed to achieve T cell activation, and only a few T cells specific for a peripheral self-antigen exist, the self-antigen will rarely, if ever, activate its corresponding T cells. This proposal could therefore contribute to an understanding of how peripheral tolerance is realized.

This suggestion will strike many as interesting and novel in the context of predominant, contemporary views, which hold that a single T cell can be activated by an APC under appropriate circumstances. We make the case in this article that this “Quorum Hypothesis” provides a more comprehensive and coherent framework for understanding how T lymphocytes interact differently with Ag to result in their activation and inactivation. We describe and review some important observations from classical and contemporary literature to make this case.

In primary lymphoid organs, negative selection removes the majority of lymphocytes with reactivity toward self-antigens expressed in these organs, resulting in a state of central tolerance (7, 8). However, not all self-antigens are expressed in primary lymphoid organs at levels sufficient for reliable negative selection, resulting in some autoreactive lymphocytes emigrating to the periphery. Therefore, a mechanism for peripheral tolerance is required to prevent the activation of autoreactive lymphocytes that can lead to autoimmunity. B cells are inactivated by Ag in the absence of helper CD4 T cells (9). Reports, in contrast, have been conflicting regarding CD8 T cell activation; in certain contexts, CD4 T cells appear necessary for CD8 T cell activation (10–12), whereas in others, they do not (13–15). Most immunologists envisage that CD4 T cells are required when CD8 T cells are activated under noninflammatory conditions, but under inflammatory conditions the APCs are envisaged to express sufficient costimulatory molecules to activate the CD8 T cell in the absence of CD4 Th cells (16, 17). This proposal can be seen as an extension of Matzinger’s Danger Model, the most commonly held view for how Ag activates CD4 T cells. The Danger Model, formulated in 1994 (18), was an elaboration on Janeway’s earlier proposal (19). Dangerous entities, such as pathogen-associated molecular patterns (PAMPs) or alarmins, produced under conditions of stress (as occurs with tissue damage or in infections), are sensed by APCs, causing them to upregulate their expression of costimulatory molecules necessary for the activation of an Ag-specific CD4 T cell. In the absence of danger signals, the Ag would inactivate its corresponding CD4 T cell (18, 20, 21). Thus, according to the Danger Model, the inactivation/activation by Ag of peripheral CD4 T cells is determined by whether the circumstances are nondangerous/dangerous rather than whether the CD4 T cells are specific for a peripheral self-antigen (pS) or a foreign Ag (F).

An alternative and contemporary model, consistent with the quorum concept, is the Two Step, Two Signal Model of CD4 T cell Activation, proposed by one of us in 1999 (22, 23). This model will be briefly considered in a later section of this paper. We now outline why we consider the Quorum Hypothesis to be more plausible and coherent than the Danger Model.

This recent hypothesis, that a quorum of Ag-specific T lymphocytes is required for the activation of CD8 and CD4 T lymphocytes, arose in the context of the contemporary view that single CD8 and single CD4 T lymphocytes can be activated under dangerous circumstances. We shall review below the ability of the Quorum Hypothesis and of the Danger Model to account for diverse observations on the activation and inactivation of T cells. However, before embarking on this analysis, we feel it important to acknowledge that the Quorum Hypothesis is not quite as novel as often proposed. The Two Signal Model of lymphocyte activation, proposed by Bretscher and Cohn (24) in 1970, postulated that the activation of all lymphocytes requires the Ag-mediated cooperation of lymphocytes and that Ag can inactivate single lymphocytes. Thus, the 1970 model postulated, put in contemporary terms, that a quorum of lymphocytes is required for the Ag-dependent activation of lymphocytes, whereas Ag can inactivate single lymphocytes. The essential differences between the Danger Model and the Quorum Hypothesis are depicted in Fig. 1.

Do the Quorum Hypothesis and the Two Signal Model of lymphocyte activation account for peripheral tolerance? The Two Signal Model was proposed in part because it offered such an accounting. Most foreign Ags impact the immune system after birth. Consider one such foreign Ag, “F.” Lymphocytes specific for F will accumulate in the body before birth. When F impinges upon the immune system after birth, F-specific lymphocytes will most often be activated, as they are present in sufficient numbers to establish quorum. This is not the case for pS. We anticipate pS is present early in development and continues to be present throughout life. When the first pS-specific lymphocyte is generated during development, it will be inactivated by pS because a quorum has not been established at this time. Because pS is continually present throughout life, pS-specific lymphocytes will be deleted as generated, ensuring the accumulation of only a few pS-specific lymphocytes. The main limitation of this elimination mechanism is the level of the presence of pS. Obviously, if pS is present at low levels, such that the inactivation of pS-specific lymphocytes takes a considerable time or does not significantly occur, pS-specific lymphocytes will accumulate. This accumulation represents a threat.

We now consider some classical and contemporary studies, most not well acknowledged in the contemporary literature, that we believe are central in understanding the basis of peripheral tolerance.

It was known in the early 1960s that neonatal administration of high doses of some foreign proteins to animals rendered them unresponsive for an Ab response to a normally immunogenic challenge administered later in life (25–27), most likely by depletion of lymphocytes specific for these proteins. William Weigle took this analysis a few steps further. He reported a series of observations (28, 29) that, we believe, are important for understanding peripheral tolerance. We first list the most pertinent observations before considering them in the context of the Quorum Hypothesis:

1. Weigle induced unresponsiveness to BSA in neonatal rabbits by administering a series of high doses of this Ag, which was evident by the lack of Ab production upon a challenge with the Ag at an age >3 mo. This unresponsiveness could be broken by the administration of cross-reacting serum albumins, such as human serum albumin (HSA), as assessed by the ability of the rabbits to produce Abs that bind to both BSA and HSA (28). These albumins were administered without the deliberate use of adjuvant. We employ the word “deliberate” because these albumin preparations most likely contained PAMPs, such as endotoxin.

2. Unresponsive rabbits, immunized a few times with HSA, again without the deliberate use of adjuvant, regained their ability to directly respond to a BSA challenge at a time when they would otherwise still be unresponsive to BSA (29).

3. The greater the degree of cross-reaction between the foreign serum albumin and BSA, as assessed with rabbit anti-BSA Ab, the lower the ability of the foreign serum albumin to break the unresponsive state (28).

These observations were initially puzzling and inexplicable. They only made sense once the possibility was proposed that the activation of the “responder lymphocyte” (i.e., the Ab precursor cell) required the presence of other lymphocytes, which we designate as “helper lymphocytes,” and that these helper lymphocytes must be first induced to express their helper activity (24, 30). It seems we must conclude that BSA-specific Ab precursor cells exist in 3-mo-old unresponsive rabbits, as immunization with HSA results in the production of anti-BSA Abs. These lymphocytes cannot be induced upon immunization with BSA but are by immunization with HSA. This is readily explicable, in general terms, if a quorum of lymphocytes is required to generate an Ab response: there is no quorum of BSA-specific lymphocytes to generate an Ab response upon immunization with BSA because of the lack of regeneration of sufficient BSA-specific lymphocytes, but there is a quorum of HSA-specific lymphocytes, which can help activate Ab precursor cells with receptors that bind to both BSA and HSA, because of the cross-reaction between the two proteins in rabbits.

How, then, did the BSA-unresponsive rabbits, after repeated immunizations with HSA, recover their ability to respond directly to a BSA challenge? We suggest this observation is only understandable if there are helper lymphocytes specific for shared epitopes of both BSA and HSA present in BSA-unresponsive rabbits, which are activated upon administering HSA but not BSA. This again makes sense if a quorum of lymphocytes is required for Ag to activate helper lymphocytes.

Weigle’s third set of observations, summarized in Table I, further supports our interpretation. In considering these observations, we assume the cross-reaction at the Ab level roughly correlates with the degree of cross-reaction at the level of the helper lymphocytes and with the degree of depletion of albumin-specific lymphocytes following the neonatal administration of BSA. With this in mind, we turn our attention to the observations in Table I. Because of the neonatal administration of BSA to the rabbits, it seems most probable that there were fewer BSA-specific lymphocytes in 3-mo-old unresponsive rabbits than there were for other serum albumins. There likely are a few more lymphocytes specific for sheep serum albumin (SSA), as this albumin cross-reacts with BSA to the tune of 75%; immunization with SSA did not break the unresponsive state of any of the seven rabbits challenged. There likely is a larger number of lymphocytes specific for pig serum albumin (PSA), as this albumin only cross-reacts 32% with BSA, and a challenge with PSA broke the unresponsive state in two of the five rabbits challenged. HSA shows 15% cross-reaction with BSA, and a challenge with this Ag broke the unresponsive state in 12 of 14 rabbits. These observations are most readily understood if 1) a quorum of Ag-specific lymphocytes is needed to initiate an Ab response and 2) a quorum is also needed for the activation of helper lymphocytes. The immunizations of rabbits with HSA, PSA, SSA, and BSA were carried out in an identical manner without any deliberate use of adjuvant. Our explanation of these observations involves quantitative considerations and is supportive of the Quorum Hypothesis. The degree of danger associated with the different immunizations should be comparable, and so the diverse observations made are difficult to reconcile with the Danger Model.

Bretscher and Cohn (24) proposed in their 1970 article that the activation of Th cells required T cell cooperation. A number of subsequent studies support this idea (23, 31, 32). In particular, it was shown that T cells specific for the nominal Ag Q could be facilitated by T cells specific for the nominal Ag R, chosen not to cross-react with Q, in the presence of the conjugate Q–R, but not in the presence of Q and R as separate molecules. In other words, the interaction between the T cells required the operational recognition of linked epitopes by the interacting T cells. We discuss later how this interaction is envisaged to occur. We briefly outline one of these studies, showing an in vivo requirement for CD4 T cell collaboration for CD4 T cell activation, to provide context.

We determined the peptide specificity of all the cytokine-producing CD4 T cells generated upon immunizing BALB/c mice with hen egg lysozyme (HEL) (33) using an improved ELISPOT assay (34). The repertoire of CD4 T cells in BALB/c mice was shown to be dominated by those specific for the peptide HEL_105–120, designated as the “major peptide,” whereas the other HEL peptides, recognized by HEL-specific CD4 T cells, were called “minor peptides,” with HEL_11–25 being the “major minor peptide” (33). We ablated the CD4 T cells specific for the major peptide by a protocol known to remove, in an Ag-specific fashion, the majority of CD4 T cells specific for this peptide (35, 36). In accord with the Quorum Hypothesis, we found that the ablation of CD4 T cells specific for the major peptide dramatically impaired the generation of cytokine-producing CD4 T cells specific for the minor peptides normally generated upon immunization with HEL. Thus, CD4 T cells specific for the major peptide appear to be required for the activation of CD4 T cells specific for minor peptides (32). These results are summarized in Table II. Further supporting the Quorum Hypothesis, we found that the ability to generate minor peptide–specific, but not major peptide–specific, cytokine-producing CD4 T cells was restored if the mice were challenged with HEL coupled to OVA instead of HEL alone. These observations show that the unresponsiveness observed is due to the ablation of the CD4 T cells specific for the major HEL peptide and not to the generation of major HEL peptide–specific inhibitory T cells (32). The nature of the challenge with HEL was identical in the normal mice and in mice depleted of HEL_105–120-specific CD4 T cells, and CD4 T cells specific for minor peptides exist in both groups of mice. The activation of CD4 T cells specific for minor HEL peptides in the former but not the latter mice is therefore difficult to reconcile with the Danger Model (32).

As mentioned above, an analysis of the role of CD4 T cells in the priming of naive CD8 T cells has led to different conclusions, namely that the activation of CD8 T cells is sometimes dependent and sometimes independent of CD4 T cells. We argue that the observations currently available support the possibility of a broadly coherent view based upon the Quorum Hypothesis. We explore a more detailed form of the Quorum Hypothesis according to which effector CD4 T cells are required for CD8 T cell activation if there is an insufficient number of CD8 T cells to establish a quorum. However, if a quorum of CD8 T cells is present, their activation would be CD4 T cell independent. This particular idea is made somewhat plausible by the fact that some activated CD8 T cells have some of the properties associated with activated CD4 T cells and are involved in their helper function, such as expression of CD40L (37–39) and the production of IL-2 (40). We refer to this particular form of the Quorum Hypothesis as the Quorum Hypothesis for CD8 T Lymphocyte Activation. This hypothesis is supported by a number of studies.

Wang et al. (15) generated dendritic cells (DCs) in vitro that for genetic reasons lack expression of MHC class II (MHC II) molecules and pulsed them with the H2-D^b–restricted peptide gp_33–41 from the lymphocytic choriomeningitis virus. The administration of such loaded DCs failed to generate gp_33–41-specific effector CD8 T cells in vivo. The DCs employed were generated in medium containing FCS. Priming could be achieved if the mice are given any of the following: 1) mice were given peptide-pulsed DCs that express MHC II, assumed to present foreign FCS peptides; 2) mice were given MHC II–deficient DCs pulsed with two different MHC class I–binding peptides (gp_33–41 and N_52–59 from vesicular stomatitis virus); and 3) a large number of CD8 TCR-transgenic T cells, specific for gp_33–41 in the context of H2-D^b, were challenged in vivo with MHC II–deficient DCs loaded with gp_33–41. Interestingly, transferring lower numbers of such TCR-transgenic CD8 T cells failed to support the in vivo priming of the host’s gp_33–41-specific CD8 T cells (15). All these observations are consistent with the predictions of the Quorum Hypothesis for CD8 T Cell Activation. In addition, these observations are not readily accounted for by the Danger Model.

Another more recent study also appears to support this Quorum Hypothesis. Palmer and colleagues (5) tested whether a quorum of OVA-specific transgenic CD8 T cells could cause autoimmune diabetes in OVA-transgenic mice whose β-islet cells express OVA. Because OVA is a self-antigen, these mice are not expected to have CD8 T cells specific for OVA. Palmer and colleagues determined that a quorum of ∼2–5 OVA-specific transgenic CD8 T cells in one lymph node could, upon immunization with peptide-loaded DCs in a manner that only the CD8 T cells in this lymph node were activated, cause diabetes. Lower numbers of the transgenic CD8 T cells were inefficient in causing diabetes, but this inefficiency could be partially overcome by providing OVA-specific transgenic CD4 T cells. Again, these observations are more readily understood in the context of the Quorum Hypothesis than the Danger Model.

In 2007, Chan et al. (41) used RAG knockout (RAG-KO) mice to study rejection and tolerance of foreign grafts, such as heart and islet tissues, transplanted under the kidney capsule. These RAG-KO mice do not have B cells or T cells and are consequently immunologically inert. Female RAG-KO mice were given different sterile grafts under the kidney capsule, and the grafts were allowed to heal in, so the tissue was expected not to express alarmins. These mice were then reconstituted with wild-type, but otherwise syngeneic, stem cells. The reconstituted mice rejected grafts that differed by multiple minor Ags but not grafts differing by only one minor Ag. Moreover, the latter grafts induced unresponsiveness to the Ag. The authors concluded that the degree of antigenic disparity, among other factors, governs the ability of transplants to induce immunity and that, in the absence of sufficient disparity, tolerance ensues (41). These observations are consistent with the Quorum Hypothesis, even though it is unclear whether rejection requires a quorum of CD8, a quorum of CD4 T cells, or a quorum of a combination of the two types of T cells (42). The observations reported are not readily accounted for by the Danger Model, as this model posits that whether Ag activates/inactivates T cells depends upon whether danger is present/absent rather than the degree of foreignness of the Ag (42).

An important point in our understanding of the mechanism by which CD4 T cells help in the activation of B cells is that this interaction requires the operational recognition of linked epitopes (43, 44). Thus, the activation of a hapten (ha)-specific B cell can generally only be helped by Q-specific CD4 Th cells in the presence of the conjugate ha–Q, but not in the presence of Q and ha–R, in which R is a carrier that does not cross-react with Q. This requirement for linked recognition is explained by the MHC-restricted model of the B cell/Th cell interaction (45). The physiological importance of linked recognition is clear: it limits the help delivered by a Th cell to a B cell with which it forms a synapse. Thus, the Th cell is highly focused in its delivery of help.

One feature of the 1999 Two Step, Two Signal Model for the activation of CD4 T cells (23), a model already referred to, is related to how CD4 T cells are envisaged to interact in a focused manner. It was known that the activation of a Q-specific responder CD4 T cell can only be helped by R-specific CD4 T helper cells, in which Q and R are chosen not to cross-react, in the presence of the conjugate Q–R and not in the presence of the separate molecules Q and R. This requirement for the recognition of linked epitopes, in the cooperation between CD4 T cells that leads to CD4 T cell activation, was a prediction of the 1970 model of lymphocyte activation (24) and is supported by various studies (22, 31, 32, 46, 47). The 1999 model for CD4 T cell activation postulated that the Ag-specific cooperation between Q-specific and R-specific CD4 T cells would be mediated by a B cell specific for either Q or R, able to take up the conjugate Ag via its Ag-specific receptors. Our studies support such a role for B cells in mediating CD4 T cell cooperation (48, 49).

In 1991, Janeway and colleagues (50) demonstrated in mice the importance of B cell activation in breaking T cell tolerance to the self-antigen mouse cytochrome c (MCC). They reported that 1) injection of MCC in CFA into C3H mice failed to generate MCC-specific B cell or T cell responses and that 2) immunization in exactly the same way with MCC in CFA resulted in the activation of MCC-specific CD4 T cells if the immunized mice had also received MCC-specific activated B cells. These activated B cells were raised by immunizing syngeneic mice with human cytochrome C in CFA. The activation of these MCC-specific B cells in the donor mice is likely CD4 T cell dependent, so the findings are consistent with the activation of CD4 T cells being CD4 T cell cooperation dependent, mediated by Ag-specific B cells. Note that these observations are paradoxical within the context of the Danger Model. If MCC-specific CD4 T cells exist, it would be anticipated these CD4 T cells would be induced upon immunizing C3H mice with MCC in CFA.

The role of B cells in autoimmunity is being increasingly recognized, even in those diseases in which autoantibodies are not the major cause of immunopathogenesis (51, 52). We conclude that the involvement of B cells in autoimmunity goes beyond their ability to produce Abs. Therefore, the efficacy of B cell depletion to treat a variety of autoimmune and chronic inflammatory diseases (53, 54) is understandable within the context of their central role as APC.

We have discussed above ideas and evidence concerning different forms of the Quorum Hypothesis as they might apply to the activation of the three main classes of lymphocytes: B cells, CD4 T cells, and CD8 T cells. However, this review of ideas and observations raises paradoxes concerning the role of B cells as APCs. We briefly highlight these paradoxes and put forward a speculative idea on how some of these paradoxes might be resolved.

The 1991 study of Janeway and colleagues (50) and one of our studies (49) support the idea that CD4 T cell cooperation is mediated by B cells, in contrast to other studies outlined above. Thus, the 2001 study of Wang et al. (15) and the 2017 study of Palmer and colleagues (5) both employed peptide-loaded DCs as the source of Ag for the in vivo activation of CD8 T cells. The straightforward interpretation of these studies is obviously inconsistent with an obligatory role of B cells as APCs mediating T cell interactions. We favor this possibility for the reasons outlined above. We thus tried to imagine whether the studies of Wang and Palmer and their colleagues, which support the generality of the Quorum Hypothesis, might be reconciled with B cells having a central role as APC. In considering this question, we realized that the researchers had to give a number of peptide-loaded DCs, and that the DCs had to be loaded with sufficient peptide, in order to achieve their objective, namely the priming of CD8 T cells. Suppose there is a requirement for B cells as APC. In this case, the researchers would have had to give peptide-loaded DCs in sufficient numbers, and sufficiently load them with peptide, for the DCs to provide B cells with the peptide. We suggest that exosomes, produced by these loaded DCs, can fuse with membranes of some B cells, resulting in B cells able to mediate interactions between T cells. This admittedly speculative possibility would, if successfully tested, allow one to maintain the idea that B cells have a central role as APCs. In addition, this idea is also significant in providing an explanation for the efficacy of treatment of cell-mediated autoimmunity by depletion of B cells (51–54). The efficacy of this treatment can be understood if a quorum of lymphocytes is required for Ag to activate T cells and if B cells have a central role as APCs in mediating the T cell interactions involved.

We would like to emphasize that our considered view is not that danger signals do not exist or are unimportant. It seems conceptually inevitable and evident from observation that such signals have been selected by evolution to modulate immune responses. However, this conclusion is different from the view incorporated into the Danger Model. The Danger Model postulates that danger/alarmin signals are required to activate CD4 T cells. Our problem with this view is that such signals are not Ag-specific and so would be expected to often lead to the activation of CD4 T cells specific for peripheral self-antigens. In contrast, the Quorum Hypothesis, which we suggest is supported by several studies outlined in this study, would appear to provide a much more secure mechanism of limiting the activation of such CD4 T cells. However, this conclusion does not mean that other factors cannot influence whether a quorum of lymphocytes is reached. For example, the roles of CTLA-4/B7 and PD1/PDL1 interactions, in limiting lymphocyte multiplication and so limiting the occurrence of autoimmunity (55, 56), are readily understood in the context of the Quorum Hypothesis. Similarly, PAMPs and danger signals (19, 21) may, under some circumstances, influence the occurrence of autoimmunity through a quorum mechanism. Addressing the question of whether danger or a quorum of lymphocytes determines whether Ag can initiate an immune response is not a matter of splitting hairs. It is in the context of the answer to this question that strategies to prevent and treat autoimmunity can be developed (57). For example, our preferred perspective explains why B cell depletion can be an effective form of treating cell-mediated autoimmunity. Our perspective also explains why Ags that cross-react with peripheral self-antigens induce not only autoreactive Abs (58) but also autoreactive CD4 T cells (59).

The authors have no financial conflicts of interest.