Co-inhibitory molecules

Abstract

Nearly forty years ago the concept was proposed that lymphocytes are negatively regulated by what are now called co-inhibitory signals. Nevertheless, it is only the more recent identification of numerous co-inhibitors and their critical functions that has brought co-inhibition to the forefront of immunologic research. Although co-inhibitory signals have been considered to directly regulate conventional T cells, more recent data has indicated a convergence between co-inhibitory signals and the other major negative control mechanism in the periphery that is mediated by regulatory T cells. Furthermore, it is now clear that lymphocytes are not the sole domain of co-inhibitory signals, as cells of the innate immune system, themselves controllers of immunity, are regulated by co-inhibitors they express. Thus, in order to better understand negative regulation in the periphery and apply this knowledge to the treatment of disease, a major focus for the future should be the definition of the conditions where co-inhibition controls effector cells intrinsically versus extrinsically (via regulatory or innate cells).

Introduction

The question of how the adaptive immune system prevents self-reactivity continues to be at or near the top of the hierarchy of important questions in immunology, with the favored solution changing from one decade to the next. Recently regulatory T cells (T_reg) have been the focus (again) of research on this question. Increasingly, however, negative regulation by receptors that work together with lymphocyte antigen-receptors to deliver ‘co-inhibitory’ signals have also taken center stage. While the rapidly increasing detailed description of co-inhibitory receptors and their intracellular signaling pathways has been reviewed elsewhere,^1–7 we focus here on the relationship between co-inhibitory receptors and their functions in terms of minimal models of immune regulation (solutions to self/nonself discrimination), with a particular focus on recent studies that suggest a convergence between co-inhibitory signals and the cells that regulate the immune response (regulatory T cells and innate immune system cells).

A Brief History of Efforts to Tackle Self/Nonself Discrimination

A unique feature of the immune system is the ability to discriminate self from nonself antigens, with strong responses against many foreign antigens and tolerance to self-antigens. Many theories have been proposed to solve the problem of self/nonself discrimination.

Timing of antigen exposure.

Burnet and Fenner proposed that there is a tolerogenic window early in the ontogeny of organisms.⁸ Despite the elegant studies conducted by Billingham, Brent and Medawar⁹ that supported Burnet's theory, numerous studies also provided evidence against this view (reviewed in refs. ¹⁰ and ¹¹). In addition, the fact that lymphocytes are generated throughout life also indicated the Burnet-Fenner theory was either incorrect or incomplete. If tolerance occurs only early in life, how do lymphocytes newly generated in an adult animal become self-tolerant? Lederberg formulated a one-signal model of lymphocyte activation¹² that resolved this problem in the Burnet-Fenner theory on tolerance. He proposed that there is a tolerogenic window early in the ontogeny of each lymphocyte rather than in the organism as a whole, allowing each lymphocyte to go through self-tolerance education whether the lymphocyte was born in a neonatal or adult animal. Lederberg's 1959 model proposing that antigen exposure in immature lymphocytes is tolerogenic, is not, as recently described,¹³ an extension of Burnet and Fenner's idea, but instead overturned their incorrect theory that postulated tolerance was a property uniquely of the fetal or neonatal period. The emergence of the central tolerance mechanisms, primarily deletion of autoreactive T cells in the thymus,^14,15 supported the Lederberg explanation that was proposed half a century ago. However, the one-signal model did not consider a need for tolerance in mature lymphocytes, that is a peripheral tolerance, a tolerance that would seem to be demanded by the presence of particular self antigens only outside the central lymphoid organs and by the capacity of lymphocytes to mutate leading to self-reactivity (e.g., somatic hypermutation).

To counter the problem of continuous lymphocyte generation and mutation in the life of lymphocytes and consequently rescue the Burnet-Fenner tolerogenic window in ontogeny, Bretscher and Cohn proposed the two-signal model of lymphocyte activation.¹⁶ According to this model, the optimal activation of T or B lymphocytes requires two signals in which the first signal arises from the engagement of the antigen with specific receptors of lymphocytes and the second signal is from the antigen specific T-helper cells (T_h), which are required to complete activation of the immune response. Based on this model, absence of self-reactive helpers can enforce tolerance throughout life due to a lack of help for newly generated helpers. This latter concept opened the chicken-egg dilemma by raising the question of which cell helped the first T_h cells? It also suffered from the same problem as Burnet's hypothesis, the experiments showing that there is no tolerance window defined uniquely in the fetal/neonatal period. Thus, while Lederberg explained much of self/nonself discrimination through a central tolerance mechanism, tolerance of the ‘peripheral self’ remained unresolved.

Co-stimulation, PAMPS and DAMPS.

An effort by Lafferty and Cunningham to solve the puzzle of T cell allo vs. xeno reactivity was a major step towards resolving peripheral tolerance, as it led to a revised two-signal model for lymphocyte activation.¹⁷ In this revised model, signal 2 (positive signal) or “co-stimulation” originates from antigen presenting cells (APCs) instead of T_h. In both of the two-signal theories, absence of signal 2 in lymphocytes will lead to tolerance (deletion or inactivation). However, it remained unclear how co-stimulation could help discriminate self from nonself; how co-stimulation could be present with foreign but not self-antigens. Charles Janeway suggested a solution to the deficiency in Lafferty and Cunningham's theory by proposing that pattern recognition receptors of the innate immune cells influenced the expression of co-stimulation. Janeway translated from lymphocytes to the APC, the Coutinho and Moller concept of mitogen receptors binding microbial products,¹⁸ as the primary stimulus for immune responses. According to Janeway's model, the interaction of pattern recognition receptors of APC with their ligands (pathogen associated molecular patterns or PAMPS) of microbes, induced APC activation and expression of co-stimulatory molecules.¹⁹ The identification of Toll like receptors (TLR), a few years later, supported this concept.²⁰ Presently, more than ten TLRs have been identified in mammals with their respective PAMP ligands. Despite the clear role of TLRs in regulating immune responses, Janeway's theory failed to easily explain transplant rejection and anti-viral immunity or the ability to harbor normal flora. Polly Matzinger introduced a new theory in the “danger model” to solve the issues in Janeway's proposal and left the idea of a self/nonself discrimination behind in favor of a danger no danger discrimination.²¹ This theory allowed for a peaceful co-existence between the immune system and our normal flora,²² unlike the self/nonself models. In the danger model, co-stimulation is induced by endogenous danger signals that arise from host cell damage. The danger model is, in a number of respects, more diverse in offering explanations for tumor immune responses, autoimmunity and transplant rejection and there is an increasing amount of evidence supporting this model.^23–29 Molecules that signal danger are also now called alarmins or danger associated molecular patterns (DAMPS).

Tolerance mediated by co-inhibition.

All of these minimal models of immune discrimination, be they self vs. nonself or danger vs. non-danger, give the job of tolerance inducing signals to the antigen receptor of lymphocytes. However, there is increasing evidence that tolerogenic signals are not derived from antigen receptor signals alone, and even before the concept of co-stimulation was proposed, the idea that there are co-receptors that provide inhibitory signals had been put forward. While examining the mechanisms of feedback suppression by antibody, Sinclair and Chan developed a model that explained the importance of the Fc portion of the antibody in suppression of the B cell response. Figure 1 shows the ‘Tripartite Inactivation model’ from their 1971 publication.³⁰ Tripartite referred to the three components, antigen, antibody (a co-inhibitory ligand) and the immunologically competent cell. This, the first proposal of a receptor (in this case an Fc receptor) that works together with an activating receptor (when co-aggregated) to mediate inactivation/tolerance, was further substantiated by the identification of Fc receptors on B cells and their negative co-signaling capacity, including the identification of a critical immuno-tyrosine based inhibitory motif (ITIM) in its intracytoplasmic domain. Sinclair later proposed that these negative signals are required for tolerance in T cells as well as B cells,³¹ and coined the term co-inhibition for this process in further postulating that the fundamental control of self/nonself discrimination in the periphery is determined by the balance between multiple co-stimulators and co-inhibitors.³² While co-stimulation contributes to immune discrimination because it is present only with DAMPS or PAMPS (and not in ‘healthy’ self tissues), we proposed that co-inhibition contributes, at least in part, by being upregulated during prolonged antigen exposure (chronic antigen receptor signaling).³² Numerous lines of evidence now support the role of co-inhibitory molecules in self-tolerance^33–35 and in control of responses during chronic antigen exposure.^36–43

Figure 1.

The origins of the concept of co-inhibition. The tripartite inactivation model³⁰ proposed that B cells are inactivated by antibody bound to antigen via the co-aggregation of the B cell antigen receptors with a receptor for the Fc portion of antibody. The model predicted the presence of negative signaling Fc receptors on B cells and that B cells are tolerized not by antigen receptor signals but instead by the co-operative signaling of antigen and Fc receptors. Reprinted with kind permission of Springer Science and Business Media. From page 611 in: Lindahl-Kiessling K, Aim G, Hanna MG, (eds.,). Morphological and Fundamental Aspects of Immunity, pp. 609–15. New York: Plenum Press 1971.

Tolerance and regulatory T cells.

The presence of autoreactive T cells in the periphery from healthy individuals^44,45 underscored the importance of peripheral tolerance, especially to control the low affinity autoreactive T cells that escape from the thymus.⁴⁶ The potential outcomes of peripheral tolerance are diverse, and include clonal anergy or unresponsiveness,^47–51 clonal deletion,^52–54 ignorance,^55–57 downregulation of T cell receptors or co-receptors^58,59 and suppression by T_reg cells.⁶⁰ Among peripheral tolerance mechanisms, T_reg cells have become of great interest due to their potential therapeutic applications in controlling autoimmunity and transplant rejection.^61–64 T_reg exhibit dominant peripheral tolerance mechanisms by suppressing self-reactive T cells.⁶⁰ The suppressive function of T_reg is mediated by negative signals to other T cells and APCs through cell contact or cytokines such as TGFβ and IL-10. The concept of T_reg or suppressor T cells originated in 1970 from studies of Gershon and Konda.⁶⁵ Research on these cells flourished until the discovery that there was no I–J region in major histocompatability gene complex (MHC), which had been expected to be the locus controlling T_reg. Moreover, other studies^14,15,48 suggested that deletion or inactivation/anergy of lymphocytes were the relevant mechanisms of immunological tolerance, which further dwindled enthusiasm for the T_reg field. However, studies by Sakaguchi⁶⁶ demonstrated the importance of CD4⁺ CD25⁺ suppressor cells in controlling autoimmunity, which rejuvenated enthusiasm for the potential importance of T_reg in immunological tolerance. Despite the popularity of T_reg studies, there have been very few efforts to incorporate them into a model of self/nonself discrimination.^67–69 What are the rules that allow T_reg to suppress self but not appropriate foreign antigen specific responses? The rules are far from clear at this point despite the immense amount of data exploring these cells. While T_reg mediate a dominant form of tolerance where effectors cells are regulated in a cell extrinsic fashion, co-inhibition has mostly been considered a cell intrinsic recessive form of tolerance. However, recent data that we will discuss is challenging this mutually exclusive viewpoint, and suggesting that many co-inhibitory receptors are involved in both recessive and dominant tolerance.

Co-Inhibitors in Recessive and Dominant Tolerance Mechanisms

For the purposes of this discussion we will consider that dominant tolerance is an antigen specific tolerance that is dominant when lymphocytes from the tolerant animal are mixed with lymphocytes from naïve animals (the mixture acts like the tolerant cells). Conversely, recessive tolerance is manifested by a lack of tolerance when the lymphocytes are mixed. However, it should be noted that there is at least the potential for an additional dominant tolerance mechanism that would not pass the ‘mixing’ test: The upregulation of co-inhibitory ligands within tissues leading to a local dominant tolerance that is not transferable with ‘tolerant’ lymphocytes to naïve recipients.

Summarized in Figure 2 and Table 1, is a minimal model of the currently described cellular interactions, either cell intrinsic (recessive) or extrinsic (dominant/regulatory), in which co-inhibitory pathways are known or thought to be involved. Multiple co-inhibitory receptor ligand pairs are likely to be involved in each of the five pathways illustrated, each serving substantially or slightly different roles in the problem of self/nonself discrimination. Mechanism number 1 in Figure 2 and Table 1 represents the most well documented co-inhibition scenario, where co-inhibitory ligands, which can be soluble (e.g., antibody) or expressed on the surface of cells (non-T cells; APC or other tissue cells), interact with co-inhibitory receptors and generate tolerance. This tolerance is generally considered to be a recessive form of tolerance and involves recruitment of phosphatases to ITIM motifs. However, it cannot be excluded that in some cases these interactions may turn the T cell into a T_reg, in which case dominant tolerance would ensue. In fact recent studies indicate that expression of PD-L1 on APC promotes generation of iT_reg in a population of naïve T cells.^70,71 It remains unclear how such a mechanism could function in vivo without causing a state of generalized immunosuppression. Another example of co-inhibition via mechanism 1 is the ability of HVEM on radioresistant cells to prevent T cell activation by its interaction with BTLA and/or CD160 on the responding T cells.⁷² PD-L1 is also expressed in non-hematopoeitic cells^73–77 and may bind with PD-1 on conventional T cells (T_con) to maintain recessive tolerance within tissues and tumors.⁷⁸ Although mechanism 1 is a recessive tolerance (not mediated by T_reg) acting directly on responding T cells, the inhibition of proliferation/activation of the responding T cell population could be expected to alter the ratio of effector to T_reg cells, favoring the T_reg. The concept that mechanism 1 is recessive is contingent upon the determining factor in responsiveness being regulation of co-inhibitor expression on the responding lymphocyte. That is, co-inhibitor levels changes while co-inhibitory ligands are a constant (not inducible) and thus do not ‘decide’ the outcome. However, the picture may become even more complex for mechanism 1 if co-inhibitory ligands are themselves also inducibly expressed in tissues, as has been seen in the setting of inflammatory cytokines and autoimmunity.^76,79 In this latter case, mechanism 1 would itself seem to be an effort to establish a form of dominant tolerance locally within the tissue, a dominant tolerance that is not mediated by T_reg but may nevertheless be useful.

Figure 2.

Mechanisms of co-inhibitory signaling involving dominant (T_reg) versus recessive mechanims. Abbreviations, include Ci (co-inhibitor), Ci-L (co-inhibitory ligand), Cs (co-stimulator), Cs-L (co-stimulatory ligand), T_con (conventional T cell), _Treg (regulatory T cell). See also Table 1 for a description of each type of mechanism shown in 1–5 in the figure.

Table 1.

Distinct mechanisms by which co-inhibitory receptors/ligands block conventional T cell (T_con) responses depends on the cells expressing co-inhibitors vs. co-inhibitor ligands, and may even switch their function from inhibition to stimulation

	Co-inhibitor	aCo-inhibitor ligand	Outcome	Examples; cBinding
1	Tcon	APC or tissue	Inhibitory signals to Tcon; recessive tolerance	PD-1/PD-L1, Fas; trans
2	Tcon	Treg	Inhibitory signals to Tcon; dominant tolerance	PD-1/PD-L1, BTLA/HVEM; trans
3	Treg	APC	Reduced co-stimulation to Tcon; dominant tolerance	CTLA-4/CD80-CD86; trans
4	Tcon	bTcon	Survival signals to Tcon; prolonged responses	BTLA/CD160/HVEM; cis
5	Tcon	APC	Reduced co-stimulation to Tcon; recessive tolerance	CTLA-4/CD80-CD86; trans

^aIn some cases, specifically #3 and 5, ligand for the co-inhibitor is also a co-stimulatory ligand.

^bThe co-inhibitory ligand (HVEM) is also expressed on T_reg, and at high levels, although low BTLA levels on these cells likely preclude significant cis interactions.

^cInteractions between receptors and ligands on the same cell (cis) versus different cells (trans).

The role of co-inhibitors in the dominant tolerance mediated by T_reg has only recently emerged. T_reg play a key role in the immunological tolerance against self-antigens as well as foreign antigens. T_reg can be divided into natural T_reg (nT_reg) and induced Tregs (i_Treg). The development of nT_reg is different from induced T_reg as the former develop in the thymus whereas iT_reg are induced in the periphery.⁸⁰ The mechanism of suppression by T_reg can be contact dependent or through cytokine dependent mechanisms.^81,82 It has recently emerged that the suppressive function of T_reg is mediated by co-inhibitory receptors. For example, T_reg lacking PD-L1 or CTLA-4 are not good suppressors.^83,84 It has been shown recently that the CTLA-4 in T_reg downregulates co-stimulatory molecules CD80 and CD86,^84,85 on APCs to maintain tolerance (mechanism 3 in Fig. 2 and Table 1). In contrast to CTLA-4, ligands of PD-1 and BTLA are more highly expressed by T_reg than T_con such that co-inhibitory ligands of T_reg bind with their receptors on T_con (mechanism 2 in Fig. 2 and Table 1). However, some studies have suggested an alternate possibility, that PD-1 on T_reg could negatively regulate immune responses by binding with its ligand, PD-L1, on other cells.⁸⁶ The mechanisms involved in this latter possibility are not clear. It was recently shown that PD-L1 was not only required for T_reg functions, but also required for the development and maintenance of iT_reg.⁷¹ CTLA-4 and PD-1 are not the only co-inhibitory pathways key to T_reg function. T_reg that lack HVEM have reduced capacity to suppress naïve wild type (WT) T cells.⁸⁷ Conversely, WT T_reg could not efficiently suppress BTLA^-/- T_con, which implied that T_reg utilized HVEM to inhibit the effectors through BTLA⁸⁷ and possibly CD160. Increasing the complexity even further, receptors involved in co-inhibition apparently can also have a positive impact on immune responses. BTLA expression and function in T cells is associated with increased T cell survival in both graft versus host disease and colitis models.^72,88–90 How BTLA functions to increase survival is not yet clear. However, a recent study indicates that BTLA and HVEM can interact in cis on T cells (see mechanism 4 in Fig. 2 and Table 1) and that the cis interaction promotes survival.⁹⁰ Surprisingly, it promotes survival even though the cis interaction blocks trans interaction of HVEM ligands (BTLA, CD160) with HVEM on adjacent cells, preventing HVEM signals (NFϰB activation).

T cell immunoglobulin (Ig) domain and mucin domain-3 (Tim-3), is a co-inhibitory molecule expressed by terminally differentiated T_h1 T cells. The binding of Tim-3 with its ligand galectin-9 induced apoptosis of T_h1 cells.⁹¹ A recent study reported that galectin-9 was expressed by T_reg and proposed that it could inhibit T_h1 cells by binding with Tim-3 on those cells.⁹² Consistent with their speculation, blocking antibodies to Tim-3 reduced the suppressive function of T_reg in vitro and in vivo. Although blocking Tim-3 pathway partially restored T_con proliferation in vitro, there was no evidence that it directly reduced the suppressive function of T_reg. A previous study from the same group reported that the ligand of Tim-3 can negatively regulate alloreactive CD8⁺ T cells.⁹³ Based on these findings, the interpretation that blocking Tim-3 pathway in vivo negated the suppressive function of T_reg is complicated by the possibility that the treatment could have directly enhanced alloreactive CD8⁺ T cell responses subsequently resulting in allograft rejection.

Given the above evidence that co-inhibitors are critical in T_reg function it raises the question of whether co-inhibitors actually have a critical role in recessive tolerance mechanisms. Conditional deletion of CTLA-4 only in T_reg showed delayed onset of the rapid lymphoproliferative disorder and autoimmunity that occurs when there is global deletion of CTLA-4, suggesting that CTLA-4 may also regulate the T_con intrinsically.⁸⁴ More recently, this concept was supported by elegant experiments by Ise et al.⁹⁴ and Jain et al.⁹⁵ that demonstrated the requirement for CTLA-4 in controlling T_con to prevent autoimmunity. Hence, the expression of CTLA-4 in T cells has a dual role. The expression of CTLA-4 in T_reg serves to control aberrant activation of T_con extrinsically, whereas CTLA-4 has intrinsic effect on T_con to maintain tolerance. Furthermore, numerous lines of evidence showed the involvement of co-inhibitory molecules in recessive tolerance mechanisms such as deletion and anergy of T cells.^96–99 Interestingly, even the well-known ability of B cell antigen presentation to tolerize naïve T cells^100,101 has been found to be dependent on the co-inhibitors PD-1 and CTLA-4.¹⁰² In another recent study, the adoptive transfer of CD25⁻CD4⁺CD45RB^high naive T cells into syngeneic Rag^-/- recipients that induces colitis was shown to be accelerated in HVEM^-/- Rag^-/- recipients. HVEM expression on radioresistant cells reduced the disease via interactions with BTLA and/or CD160,⁷² indicating a non-T_reg mediated tolerance through co-inhibition. Interestingly, BTLA was also required in non-T cells to reduce the disease.

Feto-maternal tolerance.

The mechanisms of feto-maternal tolerance in humans and mice have been discussed in detail elsewhere.¹⁰³ Here we will focus on the role of co-inhibitory molecules in the maintenance of maternal tolerance. Aluvihare et al.¹⁰⁴ reported the expansion of T_reg in allogeneic pregnancy in mice when compared to syngeneic pregnancy. Consistent with the mouse studies, it has been demonstrated that there is also expansion of T_reg in human pregnancies.¹⁰⁵ Furthermore, adoptive transfer of T_reg in an abortion prone mouse model¹⁰⁶ prevented fetal resorption, which suggested the importance of T_reg in allogeneic mating. PD-L1 is expressed by mouse⁷⁵ and human placenta,¹⁰⁷ which may serve to inhibit paternal antigen reactive T cells. Consistent with this possibility, paternal antigen specific T cells upregulated PD-1 upon encounter of cognate fetal antigen in pregnancy¹⁰⁸ and blockade of PD-L1 pathway induced fetal resorption and reduced litter sizes.⁷⁵ In contrast to this recessive tolerance action of PD-1 in pregnancy, adoptive transfer of purified T_reg from WT mice but not from PD-L1^-/- mice was shown to reduce the semi-allogeneic fetal resorption in PD-L1^-/- mice.⁸³ However, litter sizes were small when compared to WT females, suggesting the requirement of PD-L1 in other immune cells or tissues. Another study showed that PD-L1^-/- mice had an increased percentage of antigen presenting cells, which expressed a higher level of co-stimulatory molecules,¹⁰⁹ raising the possibility that this mechanism might have enhanced the allo-immune responses against semi-allogeneic fetuses. It therefore remains an open question as to whether co-inhibition contributes to fetal tolerance primarily via recessive or dominant⁸³ tolerance mechanisms. The role of co-inhibitory molecules in the maintenance of maternal tolerance may involve protecting suppressive functions of T_reg,^103,110 induction of apoptosis in paternal antigen specific T cells,¹⁰⁸ and a balancing of Th1/Th2 responses.^75,103

Exhaustion.

T cell adaptation or ‘exhaustion’ is a property that occurs in T cells due to persistent systemic antigen exposure^{43,58,111–114} and chronic viral infections, respectively.^36,115 Previous studies reported that exhausted anti-viral T cells expressed high levels of multiple co-inhibitory receptors,³⁸ including CTLA-4, PD-1 and LAG-3, which leads to T cells dysfunction^36,116,117 and persistent viremia. Furthermore, blocking co-inhibitory molecules induced strong immune responses by reversing the state of adaptation or exhaustion of T cells.^43,117–119 Reversal of exhausted T cells by blocking co-inhibitory pathways has become an important area due to its therapeutic applications in chronic viral infections such as HIV, and blocking multiple co-inhibitors is synergistic in reversing exhaustion.^38,41 While a number of studies implicate T_reg in the reduced responses in chronic viral infection, there are not yet many studies addressing the question of whether co-inhibition's contribution to ‘exhaustion’ is a recessive tolerance or via T_reg. Current data favors a non-T_reg contribution of co-inhibition.⁴¹

While the T_reg literature may have to some degree promoted a descriptive biology approach to immunology, the exhaustion literature may also have suffered this inertia. Exhaustion studies seem to be an example where evaluation of concepts is lost in the rush to generate descriptions of mechanism of an immunologic phenomenon. In providing the description of what molecules are involved in controlling exhaustion, the fact that these descriptions actually overturn (disprove) the concept of exhaustion seems to have been overlooked. The word and concept of exhaustion means to consume or tire completely, that is, the entities or resources used for positive action have been depleted. The literature showing a key role for co-inhibitors in putative “exhaustion” show the phenomenon is in fact not exhaustion, all the resources for positive action are present; it is instead an upregulation of negative regulatory pathways. As shown in Figure 3, relieving the cells of these co-inhibitory signals reveals that the cells are not exhausted and have all the resources to respond. Like the term “negative-costimulation”, an oxymoron often used to describe what is really co-inhibition, use of the term exhaustion when discussing tolerance through chronic antigen exposure misconstrues the essence of the phenomenon. While it is exciting to discover molecules that underlie the tolerance during chronic antigen exposure, as this provides new avenues for clinical treatments,¹²⁰ it is not clear why there would not also be excitement in (or even recognition of) the advance that it provides for a fundamental understanding of how the immune system works; that such tolerance works not through exhausting T cells (i.e., too many positive signals exhaust resources) but through a decision to shut down T cells by employing co-inhibitory (negative) signals when positive signals become chronic. As we have argued previously,¹²¹ if chronic antigen/virus were truly exhausting T cells, then additional positive signals to T cells should have no effect or deepen the exhaustion if the exhaustion was not already complete. Instead, “exhaustion” can be rescued by providing to T cells what can only be considered additional positive (exhausting) signals.¹²² Given the ability of co-inhibition blockade to restore responses, the only way to maintain the concept that chronic antigen exposure leads to exhaustion would be to postulate that co-inhibition is itself acting as additional positive signals responsible for depletion of resources. Based on existing data we favor the model that co-inhibitory signals inhibit the elaboration of effector functions but do not deplete the resources needed for effector function. Chronic antigen exposure can also lead to deletion of some of the responding repertoire of T cells, and co-inhibition is also likely to be central to this process. There is no evidence that this deletion is a result of exhausting resources.

Figure 3.

Relief from co-inhibition reveals that chronic antigen exposure does not lead to exhaustion of T cells. Chronic antigen exposure leads to long-term expression of co-inhibitory receptors in conventional T cells. Two alternative outcomes of chronic antigen exposure are depicted. The conventional view is shown on top, where chronic antigen exposure (e.g., chronic LCMV infection) leads to exhaustion of T cells. Exhaustion is a loss of resources needed for differentiation to effector function. The resources (R) that are putatively depleted have not been defined but could include signaling elements, transcription factors, cytokines, ATP etc. The second possible outcome is shown at the bottom, where the T cell is not exhausted, resources within the cell are maintained but not deployed because they are held in check by co-inhibitory signals. Blocking co-inhibitory signals differentiates between these two possibilities, as co-inhibitory blockade is predicted to restore effector function in the second model (bottom) but not if chronic antigen leads to exhaustion (top). Abbreviations are as described in Figure 2.

Tumor evasion mechanisms.

Tumor cells, as a mechanism of immune evasion, have exploited the property of co-inhibitory molecules that regulates immune responses against self-antigens. The anti-tumor T cell responses are limited due to the expression of co-inhibitory molecule by T cells, and co-inhibitory ligands by antigen presenting cells as well as by the tumor microenvironment. For example PD-L1 expressed by tumor cells induced T cell dysfunction, by binding with PD-1 expressed by tumor specific cytotoxic T cells.^78,123 A good prognosis for cancer patients was inversely proportional to the expression of PD-L1 in tumor cells.¹²⁴ Interestingly, a recent study suggests that PD-L1 sends signals directly to the tumor cells to trigger their resistance to killing, rather than PD-L1 sending co-inhibitory signals to T cells.¹²⁵ Another recent study in humans demonstrated the relationship of BTLA expression in anti-tumor effector T cells and inhibition of their function.¹²⁶ T_reg and APCs can also block the anti-tumor T cells by direct effects and also by the production of the indoleamine 2,3-dioxygenase (IDO) enzyme. The negative association of T_reg with tumor immune responses has been shown in various studies.¹²⁷ Depletion of T_reg using anti-CD25,¹²⁸ induced tumor immune responses and other strategies that were meant to attenuate T_reg function by anti-CTLA-4, anti-GITR treatment also induced strong tumor immune responses and rejection of tumors.^129,130 Although the latter studies demonstrated the induction of tumor immune responses, they did not demonstrate that the treatments affected T_reg directly. Indirect effects could occur through expression of the targeted molecules on other cells of the immune system. The role of CTLA-4 in T_reg mediated tumor immune suppression was demonstrated by the development of a T_reg-specific CTLA-4 knockout, lacking CTLA-4 only in Tregs.⁸⁴ In this mouse model the tumor immune responses were enhanced. The involvement of multiple co-inhibitory pathways opens up a possibility to develop an innovative tumor immunotherapy.

Role of co-inhibitory molecules in transplantation.

Induction of transplantation tolerance to foreign antigens remains the Holy Grail for transplantation immunology. The involvement of co-inhibitory molecules in the mechanisms of peripheral tolerance has allowed immunologists to develop new strategies that promote tolerance to allogeneic tissues. Long-term acceptance of allografts was achieved in various allograft models by using CTLA-4-Ig^131,132 (although this works by blocking CD28 co-stimulation), PD-L1-Ig,^133,134 and anti-BTLA treatments^132,135 alone or in combination with other therapies. On the other hand, blocking co-inhibitory pathways accelerated allograft rejection.^136–138 It has been demonstrated that intratracheal delivery of alloantigen prolonged the survival of cardiac allografts by allowing the development of donor specific T_reg.¹³⁹ Blockade of the PD-1/PD-L1 pathway during the administration of alloantigen, by using either anti-PD-1 or anti-PD-L1, accelerated rejection.¹⁴⁰ The conclusion was that PD-L1 blockade prevented the induction of T_reg. However, there was no direct evidence that PD-1 or PD-L1 blockade pre vented the induction of T_reg in this setting, as the adoptive transfer studies employed whole splenocytes rather than purified T_reg. Tolerance to various allografts achieved by treating the animals with several regimens could be prevented by using blocking antibodies to Tim-3,¹⁴¹ or PD-L1,¹³¹ or CTLA-4.¹⁴² The effects could be due to enhanced proliferation and cytokine responses. Blockade of co-inhibitory molecules induced strong immune responses by favoring T_h1 responses and expansion of cytotoxic T cells that lead to accelerated rejection. In terms of strategies to induce transplantation tolerance, there are at least two major approaches that could prove useful. One is the generation of biologics that act as agonists for co-inhibitory signals, with few such agents having been developed at this point. A second approach is to overexpress ligands within tissues to create an immune privileged environment for the transplant.^143–146

Co-inhibition, a controller of homeostasis, antigen specific responses or both?

The engagement of a co-inhibitory receptor with its ligand could influence the homeostasis of T cells. Blockade or absence of co-inhibitory molecules induced expansion of antigen-specific reactive T cells.^136,147 CTLA-4^-/- mice die by 3 w of age due to a lymphoproliferative disorder, which implied the importance of CTLA-4 in T cell homeostasis. However, in an important recent study the hyperproliferative response in CTLA-4^-/- T cells appear to be autoantigen driven to a large extent, and for the first time it was shown that CTLA-4 is critical in controlling T cells specific to natural autoantigens.⁹⁴ It has been demonstrated that antigen independent homeostatic expansion of T cells could be negatively regulated by BTLA.¹⁴⁸ In addition, the loss of BTLA in naïve T cells enhanced the generation of CD8⁺ memory T cells. Using an elegant model, Welsh and colleagues showed that PD-1 also plays a key role in controlling lymphopenia induced homeostatic proliferation of established anti-viral T cells.¹⁴⁹ However, our recent studies have shown that PD-1 and BTLA are mainly required to control lymphopenia induced homeostatic proliferation and effector function of recent thymic emigrants (Thangavelu G, et al. unpublished). Recent thymic emigrants are a T cell population with distinct properties and are particularly important early in immune system generation and during immune reconstitution post lymphocyte depletion that occurs in some viral infections and in conditioning used for bone marrow transplantation. Interestingly, syngeneic bone marrow transplantation induced autoimmunity in sub-lethally irradiated immunodeficient animals, but not in lethally irradiated immunocompetent mice.¹⁵⁰ It was suggested that the presence of radio-resistant T_reg cells prevented the onset of disease in lethally irradiated immunocompetent mice. The onset of the disease was affected by gut flora and could be prevented by co-transfer of T_reg along with the syngeneic bone marrow transplant. In addition, T_reg have also been shown to play a key role in controlling lymphopenia induced homeostatic proliferation of T cells.¹⁵¹ Whether T_reg require co-inhibitory molecules to prevent lymphopenia induced homeostatic proliferation of T cells has yet to be determined.

Interpreting experiments using antibodies targeting co-inhibitors.

The blockade of co-inhibitory pathways with monoclonal antibodies (mAb) has been an important strategy in various experimental models to test co-inhibitor function and generally has been found to increase immune responses, although in some cases the antibodies appear to be agonistic. The induction of strong immune responses could occur through releasing effector cells from co-inhibitory signals, by altering the ratio of T_reg/T_con, reducing T_reg function, or favoring a particular class of response (T_h1/T_h2/T_h17). Conversely, putative agonistic antibodies are assumed to inhibit responses by providing negative co-inhibitory signals to the cells they bind. However, in most cases there is very limited data to support the contention that the antibodies simply act by blocking or stimulating the co-inhibitory receptor. Often the evidence that a particular mAb blocks or activates a co-inhibitor is derived solely in vitro and then assumed to function similarly in vivo. However, mAb have the potential to do things in vivo that do not readily occur in vitro, such as opsonize cells leading to their destruction via various mechanisms. A recent example of this is an interesting study showing the importance of HVEM on radioresistant cells interacting with BTLA on T cells in the prevention of colitis.⁷² The mAb 6F7 specific to BTLA was used to show that an agonist mAb (conclusion derived from in vitro data) inhibits colitis. However, when we studied the effects of 6F7 in vivo, we found that this antibody physically depletes BTLA expressing cells.¹³⁵ Although it is possible the depletion is due to induction of apoptosis triggered by BTLA signaling, this seems unlikely given that loss of BTLA expressing cells occurs in a large fraction of these cells, while only a small fraction are likely to be engaging cognate antigen (a requirement for a co-inhibitory signal). Whether an antibody is blocking, depleting or agonistic when bound to a co-inhibitory receptor (or ligand) has important implications for its use therapeutically. An agonistic anti-co-inhibitor mAb may only temporarily inactivate the relevant antigen specific cells, while depletion would be a permanent elimination of relevant clones that could only be countered by recruitment of new thymic emigrants and newly generated B cells into the peripheral repertoire.

Controlling the Controllers, Innate Immunity

While T_reg cells and B cells are, a priori, the only cells with the potential to naturally generate an antigen specific dominant tolerance, they are not the only cells that can negatively control immune responses. The innate immune system can negatively regulate immune responses in an antigen nonspecific fashion, and perhaps also in a location specific fashion (e.g., tissue localized tolerogenic DCs). The function of co-inhibitory molecules in the innate immune system and their subsequent effect on tolerance vs. immunity in the adaptive immune system remains uncertain. It is important to decipher the function of co-inhibitory molecules on innate cells and how they may affect immune dysfunction. Multiple co-inhibitory receptors and receptors involved in inhibition along with their ligands such as PD-1:PD-L1/PD-L2, BTLA:HVEM, B7-H4 (B7S1), Pir-B:MHC I, and Siglec-10 (Siglec-G in mouse):CD24 are expressed or inducible on innate immune cells. These receptors and ligands are important in inhibition of the adaptive immune response by T and B cells and appear to be important for inhibition of the innate immune response as well.

Until recently it was not known whether PD-1 could be expressed on innate immune cells; however, recent reports have found PD-1 to be expressed on dendritic cells (DCs), NKT cells and macrophages.^6,152,153 Yao and colleagues found that PD-1 is inducible on splenic DCs, and upregulation of PD-1 on DCs inhibits the release of IL-12p70 and TNFα by T cells. PD-1 was also found to be inducible by lipoteichoic acid, Poly I:C, lipopolysaccharide (LPS) and peptidoglycan, but could be inhibited by IL-4 and CpG. Lack of PD-1 conferred a better innate immune response presumably by permitting the release of proinflammatory cytokines during Listeria monocytogenes (LM) infection.¹⁵² Furthermore, PD-1 on macrophages can play a role in the innate immune response to bacteria during sepsis. Blood monocytes from septic mice and patients, along with peritoneal macrophages in mice, express increased levels of PD-1, and this increase is associated with cellular dysfunction and characteristic morphological changes in these cells.¹⁵³ However, PD-1^-/- mice are protected from sepsis. It has been established that PD-L1 is a molecule that triggers a negative signal to T cells and is expressed on a wide range of cells including hematopoietic and nonhematopoietic cells.⁶ Negative regulation of T-cell proliferation may either be through interaction with PD-1 or B7-1,^154–157 and PD-L1^-/- mice have been shown to have enhanced CD4 and CD8 T-cell proliferation.¹⁵⁴ While the requirement of PD-L1 to regulate T-cell responses has been established, PD-L1 has also been reported to be necessary on T cells for proper DC maturation, which in turn appeared necessary for proper T cell responses.¹⁵⁸ Together these data paint a seemingly contradictory function of the PD-1 pathway in innate cells, inhibiting or enhancing their function, and will require further studies to elucidate the specific conditions that determine the outcome.

PD-L2 is a second ligand for PD-1, however, its expression is limited to DCs, macrophages and B1 B cells.^6,159 Recently, a naturally occurring IgM antibody in humans was found to be capable of binding and potentially cross-linking PD-L2. Cross-linking of PD-L2 on immature DCs increases antigen uptake and presentation of MHC/peptide complexes and increases their ability to stimulate T-cell responses.^160,161 Survival of DCs is enhanced when PD-L2 is cross-linked along with increased IL-12p70 production suggesting a T_h1 polarized response.^160,162 Release of cytokines such as IFNγ, TNFα and IL-10 in addition to IL-12p70 has been reported from PD-L2 cross-linking.^161,163 In vivo adoptive transfer of DCs treated with the PD-L2 cross-linking antibody in a mouse model of inflammatory airway disease can prevent disease when compared to untreated DCs.¹⁶³ Signaling of PD-L2 in DCs appears to be possible as PD-L2 cross-linking alters the cytokines produced by DCs, although the potential signaling pathway is not known, and given the very short intracellular domain in PD-L2, associated signaling adaptors may be involved. PD-L2 has clearly been demonstrated to have important in vivo functions in the setting of oral tolerance and airway hypersensitivity.^164,165 Immature DCs are known to be poor stimulators of T cells and the expression of PD-L1 and PD-L2 may contribute to immature DCs favoring inhibition of T-cell responses.¹⁵⁷ In addition, since PD-L1 can be induced on macrophages by LPS and IFNγ and PD-L2 can be induced by IL-4, the expression of these molecules by DCs may be influenced by T_h1 and T_h2 cells, respectively.

B and T lymphocyte attenuator (BTLA) and its ligand, HVEM, is another co-inhibitory pathway and both receptor and ligand are expressed on myeloid cells. In addition to its interaction with BTLA, HVEM interacts with another receptor named LIGHT.¹ Innate cells from BTLA^-/- and HVEM^-/- mice secrete increased amounts of proinflammatory cytokines and are more resistant to Listeriosis.¹⁶⁶ In contrast, LIGHT does not appear contribute to this resistance. Differences in bacterial clearance are seen as early as the first day post-infection with Listeria monocytogenes (LM), suggesting that the innate immune system is involved. Signaling from HVEM to BTLA potentially suppresses the innate immune response to prevent septic shock and cytokine storms; however, it is not clear whether the BTLA signaling occurs on innate cells or other cells/tissues.¹⁶⁶ Consistent with this possibility, Kim et al. found that the presence of T cells may be necessary to negatively regulate the innate immune system and prevent cytokine storms,¹⁶⁷ although the role of co-inhibitors were not examined. As well, the BTLA-HVEM pathway may also contribute to DC homeostasis within the lymphoid tissue by acting as an inhibitory checkpoint and contributing to the restriction of DC proliferation and accumulation. Both HVEM and BTLA-deficient mice have an increase in DCs within the spleen, particularly the CD8α⁻ subsets. Mice that are LIGHT-deficient have normal DC subsets suggesting that the HVEM and BTLA is the pathway regulating DC proliferation.¹⁶⁸ Since all conventional DCs express HVEM and BTLA it is possible that these cells are capable of both delivering and receiving an inhibitory signal.¹⁶⁸

B7-H4 is another co-inhibitory molecule that regulates T-cell activation. B7-H4 was found to be expressed on a subset of tumor macrophages that can suppress T-cell responses.¹⁶⁹ In addition to its role in macrophage function, B7-H4^-/- mice display augmented neutrophil responses. These mice are resistant to LM infection and have increased numbers of neutrophils.¹⁷⁰ Another receptor involved in innate cell inhibition is the Pir-B system, balancing the activating Pir-A receptor. Pir-B is an immunoglobulin-like receptor that provides a negative signal upon interaction with its ligand, MHC I. The Pir-B receptor is widely expressed on B cells, mast cells, dendritic cells, macrophages and neutrophils.¹⁷¹ This pathway, like the B7-H4 pathway, affects neutrophils as well as macrophages. When these cells are deficient in Pir-B there is an inability to inhibit integrin signaling and activation involved in adhesion. Pir-B^-/- neutrophils exhibit enhanced respiratory burst and secondary granule release along with hyperadhesion, and Pir-B^-/- macrophages are hyperadhesive and undergo rapid spreading due to an inability to inhibit activation.¹⁷² The recognition of MHC I by innate immune cells expressing Pir-B seems to be essential in preventing or dampening activation, however, DC maturation is impaired in Pir-B^-/- mice, even with the addition of anti-CD40, which normally causes maturation of the DC along with increased levels of MHC II, CD80 and CD86.¹⁷³ Whether these differences in function of Pir-B are controlled by the cell type or other factors has not been determined. There is little to no increase of these markers in Pir-B-deficient DCs and IL-12 production is impaired, which leads to defective regulation of T-cell activation and skews towards a T_h2 response. Although it is not known why Pir-B is required for DC maturation it does not appear to alter the function of immature DCs since antigen uptake is comparable to that of WT DCs.¹⁷³

Another potentially important co-inhibitory pathway in innate immunity is found in the Siglec family of receptors that bind sialic acid. These receptors contain an ITIM, but their role in the immune system is not completely understood.¹⁷⁴ It has recently been discovered that the interaction of CD24 and Siglec-10 can detect DAMPs and inhibit an immune response, but do not respond to pathogen-associated molecular patterns PAMPs.²⁴ This pathway can detect high mobility group box 1, heat shock protein 70 and heat shock protein 90 which are released following tissue damage and correspond to both cytoplasmic and nuclear DAMPs. While these data support the concept of a mechanism by which the innate immune system can distinguish between pathogens and tissue damage to direct the necessary immune response,²⁴ it is not yet clear why it is necessary to have differential control of responses to DAMPS versus PAMPS.

Conclusions

The data we have reviewed here present a strong case that co-inhibitory receptor ligand pathways are central to both recessive and dominant tolerance mechanisms, and that their control of innate immunity is a promising area for future research. The complexity of their interactions, including the cell types that express the receptors/ligands and other issues related to the context in which these signals are perceived can alter the outcome. Receptors predominantly contributing negative co-inhibitory signals may under some conditions positively regulate responses. It therefore becomes difficult to neatly categorize receptors based simply on their structural relationships or predominant functions. The precise role of co-inhibition in recessive vs. dominant tolerance needs to be more fully defined. Most experiments investigating co-inhibitors in dominant T_reg function suffer from the flaws common to experimental systems evaluating T_reg. The T_reg studied usually do not have a defined antigen specificity.^94,175 In systems where tolerance depends on a particular receptor/pathway (e.g., a co-inhibitor) the loss of tolerance by depleting T_reg does not by itself prove the receptor/pathway works via T_reg. In addition, dominant tolerance is tested by studying the response to an antigen of naïve cells that have been mixed with cells that are putatively tolerant via a T_reg mechanism. The control for such experiments has almost universally been the addition of non-T_reg (e.g., CD25⁻ cells) to the na≤ve population to show that the non-T_reg do not inhibit the response. However, if one is to demonstrate a true dominant tolerance preventing immune responses specifically to self, the required control is instead the addition of a control population of T cells that is tolerant via a recessive mechanism (e.g., tolerant by deletion of T cells with the appropriate specificity). Without this control, the dominant T_reg tolerance demonstrated might simply be a non-specific cellular competition that raises the threshold for naïve T-cell activation. Nevertheless, such a non-specific suppression could be important for inhibiting low affinity anti-self cells, buffering against homeostatic activation, and allowing recessive tolerance mechanisms to take hold.^98,175 Defining the role of co-inhibitors in these processes should provide important insights into the evolutionary solution for self/nonself discrimination and new avenues of immune intervention in disease.