The Adaptation Model of Immunity: Signal IV Matters Most in Determining the Functional Outcomes of Immune Responses Free

The self–nonself (SNS) theoretical model has dominated the immunology field since its conception in the 1940s (1). The SNS model was initially focused on explaining B cell activation, and later on it was expended to explain T cell activation. To address the limitations of the SNS model, three theoretical models of immunity have been proposed, i.e., the danger model in 1994 (2), the discontinuity model in 2013 (3), and the adaptation model in 2014 (4). Nevertheless, the SNS model continues to dominate the field and fully influence decision making for research funding and peer-reviewed publications in high-impact journals. As a result, no major breakthroughs have been made when it comes to a curative treatment of autoimmune diseases, allergy, solid organ transplantation, and cancer beyond alleviating the symptoms and prolonging patients’ survival. The SNS and danger models are designed to explain how an immune response is induced (5), with the discontinuity model attempting to complement and unify these models (3). In contrast, the adaptation model is designed to explain the decision-making process by activated T cells to predict whether an immune response may succeed or fail in destroying its target cell/tissue (6). In this review, I discuss the impacts and limitations of the theoretical models of immunity and expand on the adaptation model to fill current gaps in our understanding of the immune system. Throughout this review, signal I refers to Ag recognition by lymphocytes; signal II refers to costimulatory signals between lymphocytes and APCs, both of which are required for the activation of naive lymphocytes; signal III refers to cytokine signaling for the differentiation of activated T cells; and signal IV refers to the adaptation molecules, i.e., adaptation receptors (AdRs) and their nominal adaptation coreceptors or ligands (AdLs) that are expressed on T cells on receiving signals I/II to ensure that activated T cells can tolerate their cytotoxic products without committing suicide, and also relay survival signals in target cells during participation in tissue homeostasis (7, 8). In fact, the decision making of activated T cells for protecting or destroying their target cells during infection is orchestrated through the AdRs/AdLs.

In 1940s, the SNS model of immunity was proposed by Sir Frank Macfarlane Burnet, who shared the1960 Nobel Prize in Physiology or Medicine with Peter Medawar for the discovery of acquired immunological tolerance (9). According to this model, lymphocytes are educated to tolerate self-antigens and respond to nonself/foreign Ags. Definition of self/nonself was initially based on signal I or lymphocytes recognizing Ags, which are defined in each individual and across species early in development based on genetic variations (10). Later on, the “two-signal” model was proposed by Bretscher and Cohn (11). In their model, Ag recognition is considered signal I causing tolerance, and help from activated CD4⁺ T cells recognizing the same Ag is considered signal II for the activation of B cells (12). Because activation of CD4⁺ T cells needed to be explained, Lafferty et al (13, 14). proposed that signal II is in fact a costimulatory signal provided by dendritic cells (DCs) along with signal I to activate T cells. With the discovery of pattern recognition receptors interacting with foreign microbial pathogens or pathogen‐associated molecular pattern, it was proposed that activation of signal II is also determined by nonself or foreign pathogens (15–17). Then, Bretscher (12, 18) proposed a “two-step, two-signal” model postulating that signals I (peptide–MHC class II) and II (costimulatory signal) provided by DCs can only prime T cells (step 1), whereas other signals I and II provided by B cells presenting the same Ag are required for the activation of primed T cells (step 2) (Fig. 1A). Recently, Al-Yassin and Bretscher (19) introduced the “quorum” hypothesis, which is a transformation of the SNS model without criticizing the classic SNS model. The quorum hypothesis suggests that a single T cell remains inactive even by receiving two signals from DCs, unless they meet a quorum for activation. In other words, activation of a responder T cell would require Ag-mediated cooperation of lymphocytes. This hypothesis is supported by the observation that a quorum of two to five T cells is needed to recognize an Ag on DCs and get activated (20, 21). Unfortunately, transformation of the SNS model from relying on nonself entity of Ags to two-signal, two-step and quorum hypotheses has not been translated to clinical practice because of, first, insufficient interests for and overlooking the impact of theoretical models of immunity in clinical and translational research, which resulted in settling for the classic SNS model without any critical re-evaluation, or lack of interests in evaluating its new SNS versions in the clinic; in fact, there is more attention to generating new data rather than re-evaluating the existing data outside the box of the SNS model. Second, it is technically challenging to measure T cell activation at a single-cell level unless a certain number of T cells are induced. Single-cell sequencing of TcRαβ could characterize oligoclonality of TcR at a single-cell level for the evaluation of this hypothesis.

Many questions are yet to be addressed by the SNS model. For instance, self-reactive immune responses against myelin basic protein (MBP) (22, 23) or self-DNA (24) are present in healthy individuals without causing multiple sclerosis (MS) or lupus; foreign Ags in the food do not always trigger inflammatory bowel disease; self-ligands, such as heat shock proteins and high mobility group box 1, can also induce the expression of costimulatory molecules, indicating that the pattern recognition receptors are not solely the sensors of foreign entities; and CD34⁺ humanized NSG mice harboring human T cells and mouse DCs tolerate human tumor cell lines carrying HLA mismatch (25–28).

In 1994, Polly Matzinger (2) proposed the danger model of immunity. According to the danger model, the immune system does not care about the entity of an Ag being self or nonself, rather, it cares about danger or damage. In fact, this model relies on signal II or costimulation expressed by DCs to explain the induction of the immune response. Any damage caused by pathogens, necrotic cell death, or distress would release alarming signals or damage-associated molecular patterns (DAMPs) detected by the DAMP receptors on DCs to induce signal II (29, 30). In other words, the danger model relies solely on DCs (Fig. 1B) for the induction of tolerance by expressing signal I only or induction of an immune response by expressing signals I and II (31). When it comes to the understanding of central tolerance, the danger model suggests that the goal of negative selection in the thymus is not to induce tolerance to self-antigens, rather, it is to induce tolerance to self-DCs (32). To explain rejection of solid organ transplants, this model proposes that graft rejection is not because of alloreactivity, rather, it is because of damage in the organ as an alarm signal to upregulate signal II for the induction of the immune response against the graft. That is why unmatched living donor kidneys survive better than matched cadaver kidneys because living donor kidney transplant does need to be perfused to get reperfusion injury. Therefore, it is suggested to sustain signal I and block signal II for inducing tolerance to the graft; this could be accomplished by not giving cyclosporin A to the recipient, which blocks signal I, and instead giving anti-CD40 (33), which blocks signal II (5), as well as to minimize the damage during transplantation by giving superoxide dismutase (5). In autoimmune diseases, the reason for MS and lupus being more common in women than in men is because of menstrual cycles in women regularly experiencing cell death such that any mutation in genes that regulate a normal death process could release DAMPs during the menstrual cycle (5). In cancer, it is suggested to continue vaccination until tumors are cleared as reported by continuous vaccination prolonging survival of patients with follicular lymphoma (34).

Concentrating on signal II as an inducer of the immune response raises some questions yet to be addressed by the danger model. If it is the damage that induces maturation of DCs by upregulating signal II, why are mature DCs present in the thymic medulla in the absence of any damage during normal central tolerance? Also, expression of signals I and II by thymic DCs does not activate thymic T cells such that egressing T cells remain naive (35). If the damage-prone areas such as the skin, gut, and lungs are the triggers of graft-versus-host disease (GVHD), why does it happen mainly during allogeneic, but not autologous, stem cell transplantation? Similarly, identical twins show successful renal transplantation despite similar exposure to surgical damage (36).

The discontinuity model of immunity was first proposed in 2013 to explain innate immune responses (3). According to this model, an immune response is triggered on sudden changes in antigenic stimulation, and it is rendered tolerant by slow or continuous/chronic stimulation (3, 37). For instance, macrophages sense sudden changes in the environment but adapt to these changes when they persist. Also, NK cells that developed in an MHC class I–sufficient environment can eliminate cells that suddenly lose MHC class I, whereas those that developed in an MHC class I–deficient environment cannot do so (38, 39). Therefore, any discontinuity in the molecular motifs with which the NK cell receptors interact would activate NK cells, whereas continuously expressed motifs would tolerize them (3) (Fig. 1C). A discontinuity could be in self-motifs, such as tumors, or in nonself-motifs, such as pathogens. Such discontinuity is characterized based on the quantity of an Ag with respect to time. Accordingly, antigenic motifs such as viruses and tumors that change very rapidly fail to induce effective immune responses because the rate of change is faster than the time required for the induction of an immune response. In other words, an unusual motif will initially trigger an effector immune response, but when the antigenic motif is not cleared and becomes chronically present, it would induce tolerance. Also, a slow appearance of an unusual motif results in a weak response and eventually induces tolerance. To this end, failure of IFA in cancer vaccines is because of promoting Ag persistence at the site of immunization, thereby inducing tolerance (40). Also, repeated oral administration of Ags can induce tolerance for the treatment of children with egg allergy (41), and chronic infections can induce tolerance in both adaptive and innate immune cells (42–44). The discontinuity model seeks to unify the SNS model and the danger model by suggesting that discontinuity in self is due to malignancy or cellular stress would induce an immune response, whereas continuity in nonself, such as commensal bacteria, would induce regulatory mechanisms or tolerance. However, immune responses that continuously detect persistent Ags such as Mycobacterium tuberculosis (45) or cancer dormancy (46), thereby keeping them dormant, need to be explained by this model.

The SNS, danger, and discontinuity models are based on the assumption that immune response is induced only for protecting the host from distress or invaders that threaten the host rather than considering that the immune response could be an ongoing process, even in the absence of any threat, for participating in tissue homeostasis. In fact, T cells are positively selected to be self-reactive to get activated toward self-antigens in the absence of any threat. To this end, human microbiome, which is not restricted to the gut and can be found in all organs (47), plays an important role in the induction of the immune response in the absence of any infection or danger. For instance, the microbiome induces inflammatory T cell responses that produce TNF-α, IFN-γ, and IL-17, thereby participating in the intestinal stromal and epithelial cell activation in healthy individuals (48). Also, around 150,000 T cells reside in the cerebrospinal fluid of healthy individuals (49) to collaborate with the CNS-resident immune cells for regulating normal physiological function of the CNS. In mice, T cells that are specific for the self-antigen MBP show enhanced hippocampal neurogenesis and improved spatial learning compared with mice lacking MBP-specific T cells (50, 51). In the brain, IFN-γ–producing meningeal T cells prevent aberrant hyperexcitability in the CNS (52). In fact, primary function of the immune response is to participate in normal tissue homeostasis, and signal IV is to ensure that self-reactive T cells perform their primary function without committing suicide or causing any harm to their target tissue. This phenomenon has also been confirmed by the discovery of tissue-resident T memory (T_RM) cells being involved in tissue homeostasis without causing any harm (7, 8, 48). Secondary function of the immune response is to protect the host from pathogens, damage, cancer, and any invasion. To this end, signal IV determines the decision making of T cells to protect or destroy their target cells, depending on the presence or absence of AdRs on target cells and/or presence or absence of the nominal AdLs on T effector cells to get engaged with AdRs on target cells. A decade ago, Polly Matzinger (5) stated: “How does the immune system determine what kind of response it is going to make?…Nobody has yet made a general model of immunity that deals with this question…it could also be the tissue that determines the class of immunity.” That was a significant shift toward appreciating an active role of target tissues/cells during immune responses, which inspired me to develop the adaptation model of immunity to explain how an inflammatory immune response can manifest dual functions in protecting cell survival and exacerbating cell death during cancer, solid organ transplantation, GVHD, and autoimmune diseases (4, 6, 53). Signal IV is orchestrated through the expression of AdRs to induce antiapoptotic pathways on engagement with their nominal AdLs or coreceptors expressed on T cells. The antiapoptotic pathways could be Bcl2 family, such as Bcl-xL and Mcl-1, or serpins, such as SPI6/PI9, which block granzymes (54), as well as survival pathways, such as PI3K/Akt. Any alterations in the expression of the AdRs could result in autoimmune diseases, whereas upregulation or modulation of the AdRs promotes tumor growth in the presence of the tumor-reactive immune response. An AdL could exist as one directional ligand sending signal to its AdR, such as TNF-α/TNFR, or as a bidirectional coreceptor sending and receiving signal from its AdR, such as B7-H1/PD-1. They could also be in single- or dual-receptor forms (Fig. 1D). An example of single receptor is a putative AdR, B7-H1. It is mischaracterized as PD-L1 ligand, but it is a bidirectional receptor that relays survival signal both as trans, when engaged with PD-1 on T cells (Fig. 1D, lower left panel), and as cis, when both B7-H1 and PD-1 or B7.1 are expressed on DCs or target cells (55–58). In fact, target cells could express AdRs and AdLs in cis to protect themselves from activated T cells without requiring to get engaged with their nominal AdLs on T cells (59). The AdRs could also exist as a dual-receptor system (Fig. 1D, right panel) consisting of those linked to antiapoptotic pathways for protecting the target cell from an inflammatory immune response and countered by those linked to proapoptotic pathways facilitating the cell death. The outcome of T cell response would depend on the presence/absence of AdR or balance between the AdR and its counterreceptor, i.e., TNFR2 and TNFR1 (Fig. 1D, right panel).

The database on which the medullary negative selection was conceptualized needs to be carefully revisited to avoid overinterpretation of the results, as well as re-evaluation of the experimental models outside the SNS model. Many experiments using TcR transgenes are questionable, and many data that would challenge the negative selection concept have been overlooked. This has been discussed in a recent article about central tolerance (35). For instance, introduction of TcR transgenes results in artificial overexpression of the TcRs and the early expression of mature TcR at the double-negative stage (60). This could reduce the efficiency of positive selection and affect repertoire skewing in transgenic mice (61). Despite the central role of Aire in negative selection, Aire-deficient and wild-type mice showed no differences in the TcR Vβ repertoire (62), neither was there any major autoimmunity in Aire-deficient mice except for mild autoimmune-like dry eyes (62). Recently, there has been a retreat from the concept of medullary selection by suggesting that central tolerance prunes, but does not eliminate, autoreactive T cells (63). This is because frequency of CD8⁺ T cells specific for self- and nonself-Ags is similar in healthy adults without causing autoimmunity (64).

The adaptation model of immunity proposes that the purpose of cortical and medullary positive selections is to select self-reactive T cells that are functional in signal I (TcR and CD4/CD8) and signal II (costimulation), respectively, and eliminate suicidal T cells that are defective in signal I (TcR and/or CD4/CD8) or signal II (CD28) (Fig. 2A). Double-positive T cells in the cortex pass through three steps on receiving signal I (Fig. 2A). Around 90% of T cells die by Ag neglect because of their inability to recognize self-antigens for ZAP-70 signaling and NF-κB activation for the expression of antiapoptotic Bcl-2 (61, 65) (defective TcR signaling). This step determines autoreactivity of T cells. Around 5% of T cells that recognize self-antigens would die not because of the cortical negative selection but because of failure in MHC restriction (MHC neglect) or their inability to mount survival signals on engaging with CD4 or CD8 coreceptor and LcK activation (66–68) (defective LcK signaling). Finally, 5% of T cells are positively selected because of mounting survival signals after recognizing self-antigens and CD4 or CD8 coreceptor (functional signal I) (61). In the medulla, priming of single positive T cells through signals I/II results in PKC signaling and endoplasmic reticulum (ER) stress, which could kill defective T cells that are unable to mount ER stress response through upregulation of the expression of the ER chaperone GRP78 (69), elevation of antiapoptotic Bcl-xL (70, 71), and phosphorylation and inhibition of proapoptotic BAD by releasing it from Bcl-2 (72); defective T cells that fail to mount GRP78 or antiapoptotic Bcl-2, Bcl-xL, or Mcl-1 (73–75) and instead express proapoptotic Bim and Bax (73) will be eliminated (defective CD28 signaling) (Fig. 2A). In fact, these two tiers of positive selection are to secure functional T cells that survive the stress of signals I/II. In the medulla, autoreactive T cells are also selected for functional signal IV, and those that are able to express AdRs/AdLs on priming by signals I/II will survive (Fig. 2B) so as to not commit suicide or cause any harm to target cells/tissues on activation and participation in tissue homeostasis (Fig. 2B). In fact, there is no negative selection of medullary T cells expressing a high-affinity TcR for self-antigens, because these T cells become CD4⁺ Tregs (76–78) or CD8^αα+ Tregs (79–81). The affinity model of negative selection is based on surface plasmon resonance that is unable to resolve the contribution of the coreceptor, which is crucial for T cell selection (82–84). In addition, dominance of the SNS model resulted in misinterpretation and misunderstanding of the data. For instance, Palmer and colleagues (85) identified the OVA peptides with different range of affinity, and only those that fit within the SNS model were characterized as negative selector, whereas other peptides of similar affinity that did not fit within the SNS model were overlooked. Overall, signal I provided by cortical thymic epithelial cells determines MHC restriction of T cells and maturation of CD4⁺CD8⁺ double-positive T cells into CD4⁺ or CD8⁺ single-positive T cells. Signal II is mainly provided by thymic DCs or medullary thymic epithelial cells, which are involved in the elimination of double-positive T cells in the cortex (86, 87) and defective single-positive T cells in the medulla. Signal IV, the AdRs/AdLs pathway, is involved in positive selection of functional T cells in the cortex (88–90) and in the medulla (73) by rescuing T cells from apoptosis on the engagement of signals I/II (35) (Fig. 2A). Therefore, any escape from negative selection, or as the adaptation model characterizes it as the medullary positive selection, could result in the presence of circulating T cells with impaired function or massive apoptosis of defective T cells on activation in the periphery rather than causing autoimmunity (4, 6, 73). This is the case in patients with sepsis where thymic atrophy could lead to the escape of defective T cells that undergo apoptosis during Ag recognition or costimulation in the periphery leading to lymphopenia (91, 92). Although Al-Yassin and Bretscher (19) proposed the quorum hypothesis to explain peripheral tolerance, it could also explain central tolerance in the medulla where two signals by thymic DCs without a quorum cannot activate positively selected T cells. In the elderly, age-related thymic atrophy that alters central tolerance would not necessarily cause autoimmunity; rather, it results in a decline in naive T cells (93), more defective naive T cells with a decline in TcR repertoire diversity (94), as well as defective T cells that are inefficient in priming (95). Also, lymphopenia increases with age because of the reduction of naive T cells (96). Positive selection of T cells in the thymic cortex and medulla allows T cells to interact with and respond to the host’s cells. This could be because the primary role of T cells is to participate in normal cellular homeostasis and function as discussed earlier, while also protecting the host from pathogens, serving as a national guard. Although T cells are educated to recognize self-antigens through thymic positive selection, they could also recognize new or nonself Ags, even without antigenic mimicry (97), because of the highly flexible CDR3 region of TcRs. With the existence of such TcR flexibility, there would be no distinction between self and nonself for T cells, because similar frequency of T cells reacting with self and nonself Ags has been reported (64).

The term “autoimmunity” was conceived based on how the SNS model understands the immune system by proposing that breakage of immune tolerance to self-antigens results in autoimmune diseases. This definition disregards the fact that T cells are educated during thymic positive selection to be self-reactive. In fact, homeostatic autoreactive T cells and B cells have been detected in the absence of any infection or danger in healthy individuals (98, 99). Such autoreactive lymphocytes participate in normal physiological functions. For instance, gut microbiome induces inflammatory T cell responses that participate in the intestinal stromal and epithelial cell activation in the gut of healthy individuals without any harm to the host (48). Also, autoreactive Th17 cells are present in the gastrointestinal tract of healthy individuals, regulating mucosal barrier function, as well as in patients with Crohn’s disease (100). Target tissue can regulate T cell plasticity and adaptation such that T cells residing in the lamina propria will lose Foxp3 and convert to intraepithelial T cells on migration to the intestinal epithelium (101). In the brain, IFN-γ–producing meningeal T cells have been shown to prevent aberrant hyperexcitability in the CNS, linking meningeal immunity and neural circuits recruited for social behavior (52). In fact, activated T cells release bioactive brain-derived neurotrophic factor that supports neuronal survival (102). Also, an inflammatory IL-17A response in the brain can shape the anxiety state of an organism (103). T cells that are specific for the self-antigen MBP are detected in healthy hosts without harming myelin sheaths, but participating in enhanced hippocampal neurogenesis and improved spatial learning (50). More detailed discussion on the application of the adaptation model of immunity in understanding the etiology and treatment of MS appears in M.H. Manjili, under revision. In fact, it is not merely the breakage of immune tolerance to cause damage to the host, rather, it is the target tissue that may or may not tolerate ongoing immune responses. According to the adaptation model, downregulation, loss, or altered pattern of the expression of the AdRs in target cells/tissues would render them vulnerable to ongoing immune responses and, in turn, results in damage-induced acceleration of inflammation (Fig. 1D). For instance, in patients with Crohn’s disease, the entire intestine is exposed to an inflammatory immune response, but only some areas remain vulnerable to inflammation, resulting in a skip pattern of lesions while adjacent tissues remain unharmed. The reason for a skip pattern of lesions in Crohn’s disease is a skip loss of the AdRs in the intestine rendering some areas susceptible to Th1/Th17 inflammatory responses, whereas healthy areas that retain the AdRs tolerate the inflammation. Also, alternating periods of flare-ups and remission in Crohn’s disease could be caused by the downregulation and restoration of the AdRs, respectively. Therefore, the term “immune intolerance” is more appropriate for autoimmune diseases.

The adaptation model of immunity allows a new understanding of the previous observations, which would have otherwise been ignored or underappreciated. For instance, this model explains how inflammatory TNF-α can facilitate remission or exacerbate the intestinal lesions based on the expression of TNFR1 versus TNFR2 on the intestinal epithelial cells. Whereas the engagement of TNFR1 inhibits the survival of intestinal epithelial cells and induces apoptosis (104), activation of TNFR2 leads to increased cell survival (105) (Fig. 1D, right panel). TNFR2 lacks a death domain and activates prosurvival pathways through the recruitment of TRAF2 and subsequent activation of NF-κB involving PI3K and PKB/Akt (105, 106). Therefore, TNFR2 could be a potential AdR relaying survival signal on the engagement with the AdL, TNF-α. An imbalance between the AdR TNFR2 and TNFR1, as well as the pattern of the AdL TNF-α being soluble or membrane bound, could determine the beneficial or detrimental effect of this inflammatory immune response in the gut (Fig. 1D, lower right panel). A higher level of TNFR1 than TNFR2 could also dominate apoptosis in target cells (Fig. 1D, upper right panel).

Alopecia is a delayed-type hypersensitivity reaction where T cells attack and destroy hair follicles (107). Although steroids are used for dampening the immune response as a treatment for alopecia, many patients do not respond to such therapies (108); yet, they respond to diphencyprone (DCP) (109). DCP is a hapten that induces an immune response by attaching to proteins, and it is successfully used for the treatment of warts (110) and alopecia (111). In alopecia, DCP induces a delayed-type hypersensitivity by increasing the number of CD8⁺ T cells up to 600% in the epithelium and 250% around hair follicle bulbs (112). These apparently paradoxical observations cannot be explained by the SNS model proposing immune suppressant, and not DCP immune inducer, for the treatment of alopecia. It is likely that scarce AdLs on CD8⁺ T cells bind DCP hapten and in turn relay strong survival signals through the AdRs expressed on hair follicle bulbs. This is yet to be confirmed.

Rejection of allografts and acceptance of grafts between identical twins appear to be associated with allogeneity or nonself entity of cells, tissues, and organs, as suggested by the SNS model. Nevertheless, such guilt by association does not necessarily follow a cause–effect direction. The danger model argues that damage to the organ is the main cause of transplant rejection such that reducing the damage by means of superoxide dismutase could improve engraftment (5). This might also explain why less matched living donor organs do better than matched cadaver organs after transplantation. However, better engraftment of an organ between twins than allografts cannot be explained by the danger model. After all, both the SNS model and the danger model are designed to explain how an immune response is induced rather than predicting the outcomes of the immune response against the graft (35). To this end, the adaptation model suggests that graft rejection or tolerance is determined by the expression of the AdRs on the graft to tolerate the recipient’s alloreactive immune response induced by surgical procedure during organ transplantation. Present data suggest that AdRs such as B7-H1 are coded in the genome and should not be different within the same species, although the pattern of the expression in each organ or each individual could vary. Such variations could be regulated by distinct microbial composition in each individual. That could be why allografts from germ-free donors and recipients do better (113).

Individuals seem to differ in the pattern, i.e., types and levels, of the expression of the AdRs/AdLs, although the AdR/AdL pattern is adaptable in a new environment when enough time is given. If an immune response is induced before target cells are adapted to express the AdRs for the nominal AdLs or co-AdRs expressed on T cells, allografts will be rejected. If an allograft is given time to modulate the AdRs for communication with the host’s AdLs or co-AdRs before an immune response is induced, the allograft will tolerate alloreactive T cells. To this end, rats that tolerated the first skin allograft being introduced soon after birth would reject the second skin allograft, whereas the first graft remained viable in the presence of an ongoing antigraft immune response (114). In fact, the alloreactive T cell response did not change, but the organ’s adaptation changed. This phenomenon was termed by Woodruff (114, 115) as “adaptation.” Other studies confirmed such tissue adaptation. For instance, an allogenic skin graft is usually rejected within 2 wk after transplantation; however, administration of immune suppressants during the first 5 d of transplantation prolonged allograft survival to 1 mo because of delaying the immune response to allow the graft to modulate the AdRs pathway or signal IV (116). Once an allograft is adapted, it would tolerate the immune response, whereas a second and fresh allograft from the same donor would be rejected (116). Such tissue adaptation to an antigraft immune response was also demonstrated by whole-body irradiation before grafting (117), as well as in retransplantation studies showing that second fresh skin allograft applied to rats that already carried a well-established adapted skin allograft of the same donor was rejected, whereas the first allograft remained viable in the presence of all-reactive T cells (115). Although the concept of allograft adaptation was further expanded by Koene (118), lack of a theoretical model prevented further investigation into the discovery of the AdRs, as well as their co-AdRs or AdLs. Such tissue adaptation can explain the tolerance of tumors expressing neoantigens or tolerance of the fetus expressing male Ags during gestation. Living-related allografts show higher overall survival than cadaver allografts (119, 120), because living donor organs usually adapt more strongly and more rapidly than cadaver organs do.

The adaptation model not only explains tissue adaptation but also provides new insights into the understanding of immune privilege sites. Sir Peter Medawar (121), who observed that rabbit skin allografts survived in the anterior chamber (AC) of allogeneic rabbit hosts, coined the term “immune privilege” to explain his observation. In fact, corneal transplantation is the most successful allogeneic solid tissue transplantation. Decades later, it was found that AC is not an immune privilege site because it expresses lymphatic drainage (122) and harbors lymphocytes and APCs (123, 124). Therefore, tolerance of allograft in the AC is not because of the AC being an immune privilege site. For instance, although allogeneic thyroid tissue was grafted successfully in the AC of the eyes of guinea pigs before the induction of the immune response, the same tissue was rejected when animals were sensitized s.c. with the same thyroid tissue simultaneously for the induction of the immune response (115). However, s.c. tissue sensitization after allograft adaptation failed to reject the allograft in the AC, whereas the s.c. graft was rejected because of the immune response (115). Such tissue adaptation has also been reported for NK cell function. NK cells that developed in an MHC class I–sufficient environment can eliminate cells that suddenly lose MHC class I, whereas those that developed in an MHC class I–deficient environment cannot do so (38, 39). These reports suggest that tolerance or rejection of allografts in the AC of the eyes is not related to allogeneity or alloreactivity. To this end, transplantation procedures rather than alloantigens are involved in graft rejection. For instance, “circular incisions” in the cornea epithelium of the right eye resulted in the rejection of corneal allograft in the left eye, whereas “X-shaped incisions” were associated with the survival of corneal allograft (125). This was because circular incisions produced a rapid dissipation of corneal nerves and subsequent release of the neuropeptide substance P from the injured nerves interacting with the substance P receptor (NK1-R) on DCs and inducing inflammatory immune responses (126, 127). This was confirmed by a successful corneal allograft when the NK1-R was blocked by the administration of Spantide II during transplantation (125). Constitutive expression of B7-H1 in the cornea and retina protects them from rejection after corneal allograft, despite infiltration of CD4²⁺ T cells; however, blockade of B7-H1 accelerates allograft rejection (128). Also, passenger donor T cells in the organ could cause GVHD after solid organ transplantation with mortality rates of 75–85% in liver transplants, 100% in lung transplants, and 30% in other solid organ transplants (129, 130). This could be because alloreactive T cells did not have enough time to adapt/modulate expression of the AdLs or co-AdRs for the recipient’s tissue AdRs. Autologous T cells already express the AdLs or co-AdRs for the host’s AdRs; thus, they are less likely to cause GVHD. The adaptation model can also explain the incidence of severe GVHD after autologous stem cell transplantation (131) by proposing that alterations in the expression of the AdRs in target tissues and/or AdLs in autoreactive T cells could induce GVHD. This postulation is also supported by demonstrating that T_RM cells that are involved in local tissue homeostasis persist after allogeneic stem cell transplantation, and are involved in GVHD (132), whereas alloreactive T_EM cells do not cause GVHD (133). Interestingly, restoration of the putative AdR, B7-H1, expression by parenchymal tissues could prevent or ameliorate GVHD in humans (134, 135). The application of the adaptation model in GVHD is further discussed in my and Toor’s previous publication (53).

Tumor cells are genetically unstable carrying continuously changing mutations that are the basis for neoantigen vaccines. Pre-existing immune responses against tumor-associated self-antigens and against semi-nonself neoantigens have been reported in cancer patients (136, 137). These data are against the SNS model that assumes tolerogenic tumor cells expressing self-antigens are responsible for the failure of tumor immune surveillance. Tumor cells also produce inflammatory signals and damage their niche, which induce antitumor immune responses without protecting the host from cancer. Continuous administration of cancer vaccine, as suggested by the danger model, has also been tested without offering a cure for cancer. After all, human vaccines against infectious diseases have been successful only in a preventing setting. The adaptation model proposes that expression of many AdRs on tumor cells, one of which is reported to be B7-H1, is responsible for the tolerance of antitumor immune responses (Fig. 1D). To this end, antitumor efficacy of immune checkpoint inhibitors is mainly because of the blockade of survival pathway in tumor cells downstream of B7-H1 (138, 139). Limited efficacy of these therapies is because several AdRs may be expressed by tumor cells that compensate for the blockade of B7-H1. In fact, adverse events associated with immune checkpoint inhibitors suggest the presence of ongoing immune responses in healthy organs relaying survival signals to their target through the AdR B7-H1 such that blockade of this survival pathway renders healthy target cells vulnerable to ongoing immune responses. The existence of AdR/AdL or AdRs/co-AdRs in both cis and trans forms may be an advantage to overcome tumor immune-suppressive microenvironment by blocking cis pathways of tumor cell survival and eliminating the tumor in a T cell–independent fashion. Application of the adaptation model in cancer is further discussed in my previous publication (6).

The SNS model, including the quorum hypothesis, the danger model, and the discontinuity model explain how an immune response is induced without addressing the ongoing homeostatic immune responses in the organs. The adaptation model explains how an immune response functions to support or destroy its target cells. This model introduces signal IV, AdRs/AdLs or co-AdRs, through which activated T cells interact with target cells to participate in tissue homeostasis without causing any harm. In Crohn’s disease, where intestinal patchy lesions are evident, single-cell RNA sequencing (scRNAseq) of the paired biopsies or autopsies (lesions and adjacent normal tissues) from each patient should be analyzed to identify the AdRs that are present in normal tissues but are absent or downregulated in the lesions. The nominal AdLs could also be identified. The potential AdRs would then be knocked in or knocked out of an epithelial cell line cocultured with inflammatory T effector cells expressing the nominal AdL to confirm the role of signal IV in protecting the target cell from an inflammatory immune response. Similar experiments could be performed in patients with MS and rheumatoid arteritis. In cancer patients, biopsies from the tumor and adjacent normal cells can be subjected to scRNAseq to identify AdRs/AdLs. In GVHD, biopsies from the skin rash and adjacent unaffected skin should be subjected to scRNAseq to discover the AdLs expressed on T_RM cells and their nominal AdRs expressed in normal skin but lost or downregulated in the GVHD affected areas. Also, adoptive T cell therapy by means of TcR transgenic ablated of AdLs or co-AdRs could also be tested to antitumor T cells from relaying survival signals to tumor cells. Therefore, a curative therapy for immune intolerance (autoimmune) diseases, allergy, solid organ transplantation, and GVHD is to discover and modulate the tissue-specific AdRs rather than matching donor–recipients for signal I or using immune suppressants. Also, a curative immunotherapy for cancer is to downregulate/ablate the expression of the AdLs or co-AdRs on adoptively transferred T cells such that they could eliminate the tumor without transmitting survival signals, or use personalized cancer vaccines during cancer dormancy for preventing distant recurrence of the disease (140–142).

The authors have no financial conflicts of interest.

The figures were created with BioRender.com.