Innate and adaptive immunity, NK cells and immune privilege

Immunity is best understood as a layered defence system. The innate layer is fast, germline-encoded and aimed at classes of threat rather than individual strains. The adaptive layer is slower to start, uses rearranged receptors and improves with exposure. Natural killer cells sit at the junction: they are innate lymphocytes that can kill abnormal cells without prior antigen-specific priming, and they also shape adaptive responses through cytokines and contact. Immune privilege is a third organising idea: some tissues actively restrain inflammation so that bystander damage does not destroy irreplaceable cells. It is not simply an absence of immunity; it is an actively maintained local compromise.

The innate response to viruses illustrates the logic. Innate cells detect infection largely through germline-encoded pattern recognition receptors located on the cell surface or within distinct intracellular compartments. These receptors include Toll-like receptors, RIG-I-like receptors, NOD-like receptors and cytosolic DNA sensors. Although some viral proteins can trigger these receptors, the predominant viral activators are nucleic acids. Placing receptors in several compartments allows innate cells to recognise viruses that replicate in different parts of the cell, which is a structural solution to a biological problem: a virus cannot easily hide its genome from every compartment at once.

Innate immunity has two functions: it directly kills pathogens and it determines the initiation and type of adaptive immune responses. In many settings the second function is indirect. Innate cells induce dendritic cell maturation with pro-inflammatory cytokines such as IL-1 and TNF-α, and mature dendritic cells then activate antigen-specific naive T cells and may dictate whether the response becomes T helper type 1 or T helper type 2. This division of labour explains why an innate defect can present as an adaptive failure. If the dendritic cell is not properly licensed, the T cell may never receive the signals it needs. ^[3]

Interferon-α/β-producing cells provide a clear example of the bridge between layers. Viral stimulation triggers these cells to produce large amounts of antiviral interferon-α/β and also induces them to differentiate into dendritic cells. Interferon-α/β and TNF-α produced by virus-activated interferon-producing cells act as autocrine survival and dendritic cell differentiation factors, respectively. The virus-induced dendritic cells stimulate naive CD4+ T cells to produce interferon-γ and interleukin-10, in contrast to IL-3-induced dendritic cells, which drive a T helper type 2 pattern. Interferon-producing cells therefore play two master roles in antiviral immunity: they inhibit viral replication directly and they trigger adaptive T cell immunity by becoming dendritic cells. ^[3]

Immune privilege extends this theme of local control. Classical immune-privileged sites include the eye, brain, placenta and testes, and the kynurenine pathway is a shared mechanism across them. The pathway contributes to privilege in conjunction with the adenosine/purinergic pathway and with immune checkpoints such as CTLA-4 and PD-1. Indoleamine 2,3-dioxygenase sits at the immune synapse that bridges innate and adaptive responses, involving antigen-presenting cells such as dendritic cells and macrophages and lymphocytes such as effector T cells, natural killer cells and regulatory T cells. Privilege is therefore an active metabolic and cellular programme, not a wall.

The concept has widened. Genome-wide transcriptome profiling of 23,843 genes identified three key immune mechanisms involved in testicular immune privilege: an immunosuppressive cytokine milieu, proteins regulating leukocyte trafficking through controlled cell junctions, and inhibition of complement activation. The same concept has been extended to chronically infected tissues and to cancers that have escaped immune control, where privilege contributes to disease progression. That extension matters clinically: a mechanism that protects the eye from inflammatory blindness can, in a tumour, protect malignant cells from immune attack. ^[4]

The cleanest contrast is between recognition strategies. Innate cells use germline-encoded receptors that are fixed in the genome and recognise broad molecular patterns. Adaptive B and T cells use clonally distributed receptors generated by rearrangement, so each clone recognises a distinct epitope. This gives adaptive immunity specificity and memory, but it also imposes a delay: rare antigen-specific clones must be found, activated and expanded. Innate immunity buys speed by recognising classes of molecule that are shared by many pathogens and by acting through pre-formed effector mechanisms. Adaptive immunity buys precision by waiting for the right clone, then producing long-lived memory.

Pattern recognition is not confined to classical pathogens. Pathogen-associated molecular patterns were once thought to be moieties found only on pathogens, but the concept now includes nucleic acids as well as classical molecules such as lipopolysaccharide. Nucleic acid sensing has become a major component of the antimicrobial arsenal, and a diverse range of pathogens is sensed through recognition of their genomes or nucleic acids that accumulate during replication. This is especially important for viruses, which may present few classical bacterial patterns but cannot avoid producing nucleic acid during replication.

The innate layer also determines which adaptive response is induced. Innate immunity not only triggers natural interferon-α/β-producing cells to produce antiviral interferon but also induces these cells to differentiate into dendritic cells. The resulting dendritic cells stimulate naive CD4+ T cells to produce interferon-γ and interleukin-10, whereas IL-3-induced dendritic cells stimulate a T helper type 2 pattern. The finding that IL-3-induced and virus-induced dendritic cells drive distinct T cell differentiation illustrates the critical role of innate immunity in determining the type of adaptive response, depending on the nature of the pathogen. This is why adjuvants matter: they are not passive carriers but instructors of the adaptive response.

Adaptive features are equally distinctive. Dendritic cells are the most potent antigen-presenting cells, and their activation prompts priming of immunity mediators based on B and T cells. T cell receptors recognise only short peptides bound to MHC molecules, whereas B cell receptors can recognise intact antigen. CD4+ T cells are called helper T cells because they help both B cells and CD8+ T cells. When B cells need T cell help, the response is T-cell dependent; when B cells produce antibody without T cells, it is T-cell independent. These categories are not merely textbook labels: they predict which patients lose antibody responses when T cell help is absent.

Immune privilege shows that the innate/adaptive distinction is not absolute in every tissue. In the testis, attenuated innate and adaptive responses are associated with reduced T cell numbers, except for IDO-myeloid-induced regulatory T cells whose numbers are increased, and with enhanced production of anti-inflammatory cytokines including IL-10 and TGF-β. The local environment therefore changes both layers at once. This is a useful general principle: privilege is not a single switch but a coordinated adjustment of innate sensing, leukocyte entry, antigen presentation and effector function.

Pattern recognition receptors are the sensory repertoire of innate cells. They are germline-encoded and present either on the cell surface or within distinct intracellular compartments. The main families include Toll-like receptors, RIG-I-like receptors, NOD-like receptors and cytosolic DNA sensors. Their ligands are pathogen-associated molecular patterns, a concept that now includes nucleic acids as well as classical bacterial molecules such as lipopolysaccharide. The receptor families are not redundant decoration: they cover different locations and different molecular structures, so an innate cell can detect a bacterium outside the cell, a virus in an endosome, or a genome in the cytosol.

Toll-like receptor expression differs between innate cell types. Human macrophages express high levels of TLR2 and TLR4, while plasmacytoid dendritic cells mainly express TLR7 and TLR9. Expression patterns also vary between species: TLR9 is restricted to a few cell types in humans but is widely distributed in mice. Downstream signalling molecules also differ. Plasmacytoid dendritic cells constitutively express the transcription factor IRF7, allowing them to produce high levels of type I interferons quickly after viral infection, whereas macrophages may respond in a more delayed manner. This is a clear example of cell-specific wiring: the same family of receptors produces different kinetics depending on which signalling proteins are pre-positioned. ^[1]

TLR2 illustrates dimerisation and ligand breadth. Functional TLR2 exists as a heterodimer with either TLR1 or TLR6 on the plasma membrane of innate and adaptive immune cells. It can be activated by lipoteichoic acid, a common component of Gram-positive bacteria, as well as by GPI anchors of parasitic protozoa such as Plasmodium falciparum. The TLR2/TLR6 heterodimer has also been implicated in the innate response to respiratory syncytial virus. Macrophages from mice deficient in TLR2 or TLR6 responded to respiratory syncytial virus with lower levels of inflammatory mediators than wild-type macrophages, and knockout mice challenged intranasally had elevated peak viral titres and lower numbers of neutrophils and activated dendritic cells in the lungs. The receptor therefore contributes to both cell recruitment and viral clearance in vivo. ^[1]

Cytosolic sensors add another layer. AIM2, NLRP3 and IPAF function redundantly in the recognition of Listeria monocytogenes. Redundancy is a design feature rather than a waste: if one sensor is evaded or disabled, another can still trigger inflammation. Fcγ receptors are also part of the innate recognition toolkit, although they recognise antibody rather than microbial patterns directly. FcγR is expressed on the surface of monocytes, dendritic cells, macrophages and natural killer cells, among others, and its main function is to bind the constant region of the heavy chain of immunoglobulins. Monomeric IgG binding to FcγRI activates an internalisation-recycling pathway, whereas cross-linking of FcγRs on monocytes activates phagocytosis and cytokine release. Antibodies lacking glycosylation bind poorly to FcγR, which shows that glycosylation is important for this recognition.

Pattern recognition also shapes dendritic cell maturation. Virus-like particles bind pattern recognition receptors on the surface of dendritic cells, are internalised, and then prompt maturation and presentation of peptides on MHC class I or class II molecules for priming CD8+ or CD4+ T cell responses, respectively. Because virus-like particles contain multimeric epitopes, they can cross-link B cell receptors strongly enough to prime B cells and induce antibody production even without CD4+ T cell help. This is not a contradiction of T-dependent immunity; it is an example of pattern strength and receptor geometry substituting for classical help.

The clinical corollary is that innate sensing defects can be narrow or broad. A defect in a single Toll-like receptor pathway may impair responses to particular organisms while leaving others intact, whereas a defect in a shared signalling molecule can affect several families at once. This is why the genetic lesion, not just the infection, matters when immunodeficiency is investigated. It also explains why adjuvants that target pattern recognition receptors can improve vaccine responses: they are deliberately engaging the instruction system that tells adaptive immunity what kind of response to make.

Phagocytes are the workhorses of innate clearance. Macrophages and dendritic cells sample tissues, internalise particulate antigen and either destroy it or present fragments to T cells. Fcγ receptors make this process antibody-directed: cross-linking of FcγRs on monocytes activates phagocytosis and the release of cytokines. Dendritic cells are the most potent antigen-presenting cells, and their activation prompts priming of immunity mediators based on B and T cells. This division of labour is important: a macrophage is primarily a cleaner and effector cell, whereas a dendritic cell is primarily a sampler and instructor, although the roles overlap in inflamed tissue.

Natural killer cells are innate lymphocytes that kill without clonal antigen-specific priming. A useful first-principles model is balance: activating signals from stress molecules and inhibitory signals from MHC class I determine whether the NK cell kills. When MHC class I is reduced, as often happens in virus-infected or transformed cells, inhibition is lost and killing is permitted. The passages do not give a numerical threshold for this decision, so it is best taught as a qualitative principle. What the passages do show is that NK cells are not isolated killers: interferon-α/β has pleiotropic effects on macrophages, NK cells and T cells. NK cells therefore receive cytokine instructions from innate sensors and can in turn shape the adaptive response.

Immune privilege uses several mechanisms to restrain NK and T cell killing. In the murine eye, NK cells and γδ T cells contribute to immune privilege, as does reduced expression of MHC molecules. This seems paradoxical at first: reduced MHC class I would normally release NK cells from inhibition. The resolution is that the privileged site also expresses inhibitory ligands and cytokines that keep local effectors quiet. In the pregnant uterus, immune privilege involves inhibitory PD-L1 and CTLA-4 on regulatory T cells and trophoblast giant cells, macrophage migration inhibitory factor from trophoblasts, inhibition of NK cell cytotoxicity, and upregulation of Fas ligand in the endometrium. Privilege is therefore a combination of altered antigen presentation, altered trafficking and active suppression. ^[1]

The kynurenine pathway is central to this suppression. The pathway contributes to immune privilege by four mechanisms: depletion of tryptophan via the stress response kinase GCN2 and suppression of the mTOR1 pathway, which inhibits effector T cell function and growth; direct effects of kynurenine on the aryl hydrocarbon receptor, which induces dendritic cell and macrophage differentiation and later favours regulatory T cells; promotion of CD4 T cell differentiation into regulatory T cells expressing CTLA-4; and kynurenine-mediated inhibition of IL-2 signalling, which impairs memory CD4 T cell survival. Upon IDO activation, antigen-presenting cells that can produce inflammatory cytokines such as IL-12 switch to producing inhibitory cytokines such as TGF-β and IL-10. IDO can also be triggered in the presence of apoptotic cells to induce self-tolerance or cancer persistence. The local antigen-presenting cell is therefore converted from an immunogenic instructor into a tolerogenic one. ^[4]

Tumours exploit the same machinery. IDO can be expressed by tumour cells and by myeloid cells surrounding the tumour, including antigen-presenting cells, tumour-associated macrophages and myeloid-derived suppressor cells, which mediate acquired immune tolerance toward the tumour. The tumour microenvironment accumulates immature and tolerogenic antigen-presenting cells expressing low levels of costimulatory proteins such as CD80 and CD86, so these cells lack the ability to stimulate naive T cells. When they migrate to draining lymph nodes and present tumour antigens, they fail to activate T cells. In mice, myeloid-derived suppressor cells infiltrate the tumour microenvironment and promote T cell dysfunction and regulatory T cell generation through IDO and arginase-1. Targeting these cells with monoclonal antibodies has restored the tumoricidal function of tumour-infiltrating T lymphocytes in mice.

Phagocyte and NK function therefore meet at several points. Antibody opsonisation directs phagocytosis through Fcγ receptors. Cytokines from innate sensors activate NK cells. Tumour-associated myeloid cells suppress both innate and adaptive effectors through IDO and arginase-1. And privileged tissues suppress local killing through checkpoint molecules, cytokines and metabolic enzymes. The clinical lesson is that “immune evasion” is rarely a single mechanism; it is usually a local programme that changes antigen presentation, effector access and metabolic support at the same time.

B cells and T cells are the two arms of adaptive recognition. B cells recognise intact antigen through the B cell receptor, internalise it, process it and present peptides on MHC class II to helper T cells. T cells recognise only short peptides bound to MHC molecules through the T cell receptor. CD4+ T cells are helper cells that help both B cells and CD8+ T cells, whereas CD8+ T cells become cytotoxic effectors. This division explains why MHC class I presents endogenous peptides to CD8+ T cells and MHC class II presents exogenous peptides to CD4+ T cells, and why dendritic cell maturation determines whether a response becomes humoral, cytotoxic or tolerogenic.

Antigen presentation is the meeting point. Virus-like particles are internalised by dendritic cells and processed for presentation through MHC class I and MHC class II molecules. CD4+ T cells help B cells to produce antibodies and enhance CD8+ T cell cytotoxicity. When B cells need help from T helper cells, the response is T-cell dependent; when B cells produce antibodies without T cells, the response is T-cell independent. This is not an all-or-none rule. Some virus-like particles can induce strong dendritic cell maturation and cytokine production, resulting in T and B cell immune responses without T helper cells, and this atypical phenomenon has also been observed with self-aggregating peptides used as antigen delivery systems. Under certain circumstances, T helper cells are not necessary for inducing cytotoxic T lymphocytes or antibody responses.

B cells can also act as antigen-presenting cells. Virus-like particles can directly cross-link B cell receptors, favouring capture and transport through the blood to the spleen, where antigen is delivered to follicular dendritic cells. Cross-linking of B cell receptors also prompts endocytosis of virus-like particles, B cell proliferation and increased production of MHC class II and costimulatory molecules. Activated B cells interact with T helper cells, allowing immunoglobulin class switching toward IgG and the generation of memory B cells. In conditions of strong B cell stimulation, virus-like particles can elicit a T-cell independent humoral response based on IgM and IgG. This illustrates a broader principle: B cells are not merely antibody factories; they are antigen-presenting cells that can influence T cell priming.

T cell differentiation is shaped by the innate context. Virus-induced dendritic cells stimulate naive CD4+ T cells to produce interferon-γ and interleukin-10, in contrast to IL-3-induced dendritic cells, which stimulate a T helper type 2 pattern. The finding that these two dendritic cell populations induce distinct T cell differentiation illustrates the critical role of innate immunity in determining the type of adaptive response, depending on the pathogen. Virus-induced dendritic cells induce naive CD4+ T cells to produce high levels of interferon-γ and interleukin-10 and a low level of IL-4, and intracellular cytokine staining shows three subpopulations with respect to interferon-γ and interleukin-10 expression. This heterogeneity is important: a single infection can generate more than one functional T cell phenotype.

Effector function differs by pathogen. For influenza virus, the protective immune response is based mainly on neutralising antibodies, but CD8+ T cells reduce the severity of symptoms and promote elimination of infected cells even though they do not prevent infection. Virus-like particles containing influenza proteins induce anti-haemagglutinin neutralising antibodies and CD8+ T cell responses that produce interferon-γ and lyse infected cells. This is a useful clinical model: antibody prevents attachment and entry, whereas cytotoxic T cells clean up cells that are already infected. A vaccine that induces only one arm may therefore reduce severe disease without fully blocking infection.

Immune privilege modifies B and T cell behaviour. In the pregnant uterus, inhibitory PD-L1 and CTLA-4 are expressed by regulatory T cells and trophoblast giant cells, macrophage migration inhibitory factor is expressed by trophoblasts, NK cytotoxicity is inhibited, and Fas ligand is upregulated in the endometrium. In the testis, attenuated innate and adaptive responses are associated with reduced T cell numbers, except for IDO-myeloid-induced regulatory T cells whose numbers are increased, and with enhanced production of anti-inflammatory cytokines including IL-10 and TGF-β. In the hair follicle, immune privilege-like mechanisms include IDO expression by tolerogenic antigen-presenting cells, low-level MHC class I expression and high-level expression of TGF-β and IL-10. Loss of privilege in the hair follicle causes alopecia areata, an autoimmune disease characterised by T cell cytotoxicity toward the follicle and transient, nonscarring hair loss.

The adaptive response also feeds back on innate cells. Fcγ receptors bind the constant region of immunoglobulins and enable interaction between adaptive molecules such as IgG and innate cells such as dendritic cells and macrophages, leading to specific recognition of pathogens. This means that antibody is not only a neutralising molecule; it is a targeting signal for phagocytes and NK cells. The same principle explains why glycosylation matters: antibodies lacking glycosylation bind poorly to FcγR. Adaptive immunity therefore does not replace innate immunity; it re-tunes it, focusing innate effector mechanisms onto particular targets.

In cancer, the balance between effector and regulatory arms determines outcome. IDO expressed by tumour cells and myeloid cells mediates acquired immune tolerance, and the tumour microenvironment accumulates tolerogenic antigen-presenting cells that fail to activate T cells. The kynurenine pathway promotes regulatory T cell differentiation and inhibits effector T cell function. This is why immune checkpoint blockade and IDO inhibition are conceptually attractive: they aim to release effector T cells from local suppression. The same logic applies in transplantation and autoimmunity, where the goal may be to preserve privilege or to break it selectively. B and T cell function is therefore always context-dependent: the same helper or cytotoxic programme can be protective in infection and harmful in autoimmunity or rejection.

Immune privilege is the capacity of certain tissues to restrain immune-mediated inflammation so that bystander damage does not destroy cells that cannot be replaced. The idea began with an observation rather than a theory. Immune privilege was first observed by van Dooremaal, a Dutch ophthalmologist who, in 1873, transplanted murine skin into the eyes of rabbits and dogs in an attempt to study the formation of cataracts. In 1948, Medawar reassessed the prolonged survival of foreign tissue grafted into the anterior chamber of the eye and coined the term immune privilege, proposing that the apparent absence of lymphatic drainage protected the graft from rejection. In 1975, Kaplan and colleagues showed that antigen introduced into the anterior chamber of the eye was able to induce a systemic downregulation of the antigen-specific cellular response, a process he called immune deviation. That third observation changed the field: privilege was not simply a wall that excluded immune cells, but an active process that could alter the systemic response.

The anatomical logic is straightforward. The eye and the central nervous system are formed from the neuroepithelium during embryonic development, and each has a nonexpandable anatomic barrier; because these barriers do not permit swelling in the case of inflammation and these sites have limited cellular regenerative capacity, immune privilege is a necessary protective feature. Privileged sites can also be categorised by their cellular regenerative capacity, with the eyes and brain exhibiting limited capacity, while other sites contain stem cells that can repair damaged tissue. This explains why the eye tolerates a foreign graft better than the skin: a brisk inflammatory response in the anterior chamber would destroy the visual axis long before it eliminated the graft.

Classical privileged sites are the eye, brain, placenta and testes, but the list has widened. Immunoregulatory processes also operate within the hair follicle and the colon, and these organs were identified as immune-privileged sites despite lacking anatomic barriers. Furthermore, the concept has been extended to chronically infected tissues and to tissues in which cancer has escaped immune control; in both situations, privilege contributes to disease progression. Recognising this extension is clinically important, because a mechanism that protects the retina from inflammatory blindness can, inside a tumour, protect malignant cells from immune attack.

The eye illustrates layered defence. The eye is anatomically divided into three layers: the outer fibrous layer, the middle vascular layer and the inner sensorineural layer containing the blood-retina barrier, and the blood-retina barrier contributes to immune privilege. Indoleamine 2,3-dioxygenase expressed by human corneal cells lining the anterior chamber contributes to ocular immune privilege, and it is expressed in many ocular tissues and contributes to the immune privilege of corneal allografts. T cells cocultured with iris and ciliary body cells acquire regulatory activity induced by IDO and TGF-beta expression. Other ocular tissues such as the uvea, conjunctiva and periocular fascia contain IDO-expressing macrophages and tolerogenic dendritic cells secreting immunosuppressive cytokines. Additional mechanisms include PD-L1, macrophage migration inhibitory factor, TRAIL and FasL, all of which inhibit T cell proliferation and survival and increase regulatory T cell numbers. In the murine eye, natural killer cells and gamma-delta T cells contribute to privilege, as does reduced expression of MHC molecules.

The testis shows how a barrier and a metabolic programme cooperate. Testicular immune privilege was initially attributed to the blood-testis barrier, which limits the passage of antibodies into the testes and limits the access of germ cell antigens to testicular interstitial immune cells. In 1980, Yoshida and colleagues reported that rodent epididymal and testicular extracts had high IDO activity, and later work showed that, unlike the inflammation-induced expression of IDO seen in nearly all mammalian tissues, expression in the testes is constitutive. Healthy testes expressed significantly higher levels of IDO compared with blood. Sertoli cells provide local immune tolerance and nourishment to developing germ cells, and inhibiting IDO activity abrogates their ability to protect pancreatic islet allografts, an effect involving IDO- and TGF-beta-dependent emergence of autoantigen-specific regulatory T cells. Testicular privilege also involves immunosuppressive IL-10 and TGF-beta secreted by macrophages and by peritubular, Sertoli and Leydig cells. Attenuated innate and adaptive responses in the testis are associated with reduced T cell numbers, except for IDO-myeloid-induced regulatory T cells whose numbers are increased.

The pregnant uterus is a special case because the fetus is a semihaplotypic allograft tolerated by the maternal immune system, and placental immune tolerance protects the fetus from maternal responses directed toward paternally inherited antigens. Factors include inhibitory PD-L1 and CTLA-4 on regulatory T cells and trophoblast giant cells, macrophage migration inhibitory factor from trophoblasts, inhibition of NK cell cytotoxicity, and upregulated Fas ligand in the endometrium. The hair follicle shows privilege-like mechanisms including IDO expression by tolerogenic antigen-presenting cells, low-level MHC class I expression and high-level expression of TGF-beta and IL-10. The intestinal tract adds FasL, TRAIL, thymic stromal lymphopoietin, prostaglandin E2, retinoic acid, TGF-beta and IL-10, which contribute to induction of tolerogenic dendritic cells.

The kynurenine pathway is the shared metabolic backbone. Upon activation, IDO1 catabolises the essential amino acid tryptophan into metabolites collectively known as kynurenines, and IDO-expressing myeloid cells locally exert a profound inhibitory effect on T cells, which in turn suppress proinflammatory processes occurring in response to tissue damage, infection and cancer. The pathway involves three distinct rate-limiting enzymes, IDO1, IDO2 and tryptophan 2,3-dioxygenase, the last being constitutively expressed in the liver and dominant in health. IDO is expressed by monocytes, macrophages, dendritic cells, endothelial cells, fibroblasts and certain cancer cells, and is upregulated by inflammatory molecules and cytokines including IL-1, interferon-gamma and TNF-alpha. The pathway contributes to immune privilege by four mechanisms: tryptophan depletion via GCN2 and mTOR1 suppression, kynurenine effects on the aryl hydrocarbon receptor, promotion of regulatory T cell differentiation, and inhibition of IL-2 signalling that impairs memory CD4 T cell survival. Upon IDO activation, antigen-presenting cells that produce inflammatory cytokines such as IL-12 switch to producing TGF-beta and IL-10, converting local T cell function from immunogenic to tolerogenic. Genome-wide transcriptome profiling of 23,843 genes identified three key immune mechanisms involved in testicular privilege: an immunosuppressive cytokine milieu, proteins regulating leukocyte trafficking through controlled cell junctions, and inhibition of complement activation.

The first pitfall is to imagine privilege as an absolute wall. It is not. The integrity of anatomic barriers within some immune-privileged sites is not absolute, as shown by the existence of an epithelial gate beyond the endothelial vascular barrier in the testes; this gate allows selective cellular trafficking in response to inflammation. Privilege is therefore a modulated set of rules rather than a sealed compartment. A graft in the anterior chamber may survive longer than a skin graft, but it is not invisible to the immune system, and privilege can be broken by infection, trauma or inflammation.

The second pitfall is to attribute privilege to a single barrier. Testicular privilege was initially attributed to the blood-testis barrier, yet the constitutive expression of IDO in the testis shows that metabolic control is part of the explanation. Similarly, ocular privilege is not explained by the blood-retina barrier alone: corneal IDO, tolerogenic dendritic cells, PD-L1, FasL and reduced MHC expression all contribute. Privilege is a programme with redundant layers, which is why blocking one mechanism often produces only a partial loss of protection.

The third pitfall is to treat IDO as the only relevant enzyme. The kynurenine pathway involves three distinct rate-limiting enzymes, IDO1, IDO2 and tryptophan 2,3-dioxygenase, and the last is constitutively expressed in the liver and plays a dominant role in health. IDO1 can be dramatically enhanced during inflammation, making it the most important of the three during inflammatory disease, whereas IDO2 is present at lower levels and is not redundant with IDO1. Conflating these enzymes leads to incorrect predictions about which tissues will be protected and which drugs will work.

The fourth pitfall concerns self versus non-self discrimination. It remains unclear how some nucleic acid sensors discriminate self from non-self, and a mechanism presumably exists allowing pattern recognition receptors such as IFI16 to distinguish virally derived from host DNA. A related unresolved question is how viral RNA and DNA are made accessible to receptors, since it is not well understood how nucleic acids are presented to cytosolic sensors when viral DNA is shielded by a capsid in the cytoplasm and replicates within the nucleus. Teaching privilege or innate sensing as a solved problem creates false confidence.

The fifth pitfall is to assume that a sensor must be protective. Toll-like receptor 4 signalling appears to benefit mouse mammary tumour virus: it activates quiescent B cells, encouraging division necessary for viral genome integration, and promotes IL-10 secretion, an immunosuppressive cytokine that helps the virus persist. Measles virus haemagglutinin triggers TLR2, and the resulting CD150 expression may benefit the virus by conferring susceptibility. In lymphocytic choriomeningitis virus infection, where much of the central nervous system damage is caused by the immune response itself, it remains to be determined whether TLR2 signalling is protective or pathological.

The sixth pitfall is to extrapolate from in vitro findings to the whole animal. Activation of TLR3 by the double-stranded RNA analogue poly(I:C) contributed to type I interferon and cytokine production in macrophages, yet a later study found no difference in survival, viral titres or pathology of TLR3-deficient mice following reovirus challenge. Similarly, species differences matter: TLR9 is restricted to a few cell types in humans but is widely distributed in mice. A pathway that looks essential in one system may be redundant in another.

The seventh pitfall is to confuse dendritic cells with macrophages. Unlike macrophages, the main function of dendritic cells is not to clear infection but to prepare the immune system, recruiting inflammatory cells and migrating to lymph nodes to activate T cells, and dendritic cells are more efficient than macrophages at processing antigens and presenting them through MHC class I. Lysosomes of macrophages contain very few MHC class II molecules and antigen is digested into amino acids there, whereas dendritic cells produce high quantities of MHC class II molecules that load antigen-derived peptides. Mistaking one for the other leads to the wrong prediction about whether antigen will be presented or destroyed.

Immune privilege is used at the bedside in transplantation, in autoimmunity, in pregnancy care, in cancer immunotherapy and in vaccine design. In corneal transplantation, the privileged status of the eye is a clinical asset: IDO expressed by human corneal cells lining the anterior chamber contributes to the immune privilege of corneal allografts, and T cells cocultured with iris and ciliary body cells acquire regulatory activity induced by IDO and TGF-beta. This is why corneal grafts can survive with less aggressive immunosuppression than other solid organ grafts, although privilege is not absolute and rejection still occurs when the local programme is disturbed.

Sertoli cells illustrate how privilege can be borrowed. Sertoli cells were used as a therapeutic strategy in autoimmune diabetes to self-protect and coprotect allogeneic and xenogeneic pancreatic grafts from immune destruction in different experimental settings, and inhibiting IDO activity abrogates their ability to protect pancreatic islet allografts. Sertoli cells provide myeloid and lymphoid immune tolerance to protect germ cells in the testes and cotransplanted cells at ectopic sites. The wider principle is that a privileged niche can be packaged: if the metabolic and cytokine programme travels with the cells, the protection travels with them.

Pregnancy is the commonest clinical setting in which privilege matters. The fetus is a semihaplotypic allograft tolerated by the maternal immune system, and in rare cases spontaneous abortions occur due to a break in immune tolerance induced by placental inflammation. IDO-deficient mice and mice treated with IDO inhibitors fail to tolerate the fetus during allogeneic pregnancy, experience colitis due to failure of mucosal tolerance of the intestinal microbiota, and lose the ability to clear apoptotic cells. Blocking or ablating IDO worsens inflammation in models of graft-versus-host disease, autoimmunity and chronic conditions such as chronic granulomatous disease and diabetes. These observations establish IDO as a key controller of inflammation rather than a curiosity of the eye.

Autoimmunity provides the mirror image, where the therapeutic aim is to restore or to break privilege depending on the disease. Alopecia areata results from loss of immune privilege within the hair follicle and is characterised by T cell cytotoxicity toward the follicles, causing transient, nonscarring hair loss; the JAK2 inhibitor ruxolitinib blocks CD8 T cell cytotoxic function and leads to improvement. In experimental autoimmune encephalomyelitis, a model resulting from a breach in immune tolerance to central nervous system antigens that resembles multiple sclerosis, administration of the IDO inhibitor 1-methyl-d-tryptophan exacerbated the condition. In inflammatory bowel disease models, CTLA-4-Ig (abatacept) induces IL-4- and IL-10-producing regulatory T cells in an IDO-dependent manner and improves colitis, whereas anti-CTLA-4 monoclonal antibodies such as ipilimumab block this pathway and frequently exacerbate or cause immune-mediated colitis.

Cancer is where privilege is most dangerous. IDO can be expressed by tumour cells and by myeloid cells surrounding the tumour, including antigen-presenting cells, tumour-associated macrophages and myeloid-derived suppressor cells, which mediate acquired immune tolerance toward the tumour. The tumour microenvironment accumulates immature and tolerogenic antigen-presenting cells expressing low levels of CD80 and CD86, so these cells lack the ability to stimulate naive T cells, and when they reach draining lymph nodes they fail to activate tumour-specific T cells. In mice, myeloid-derived suppressor cells promote T cell dysfunction and regulatory T cell generation through IDO and arginase-1, and targeting them with monoclonal antibodies restored the tumoricidal function of tumour-infiltrating T lymphocytes. Because IDO acts upstream on the antigen-presenting milieu, it is nonredundant with more distal T cell checkpoints, so blocking IDO concurrently with CTLA-4 or PD-1 might confer additive benefit.

Vaccine design uses the same logic in reverse. The first point for developing a smart virus-like particle vaccine is selecting a candidate according to the type of immune response required: if the pathogen can be blocked by antibodies, choose a particle that induces antibodies, and if the pathogen is intracellular, choose particles targeting dendritic cells that cross-present efficiently. Fusion of epitopes to particle proteins was more efficient than chemical conjugation for generating cytotoxicity. Exogenous epitopes preferably range from 20 to 30 amino acids, avoiding positively charged or highly hydrophobic residues and sequences prone to beta-strand formation that could interfere with self-assembly. Particle size matters: nanoparticles up to 200 nm directly reach the lymph nodes some hours after injection, moving through lymphatic vessels without help of any antigen-presenting cell, and persist in germinal centres on the surface of follicular dendritic cells for several weeks. Human papillomavirus L1 expressed alone or with L2 generates 40-nm virus-like particles. For Japanese encephalitis, a single immunisation with virus-like particles without adjuvant protected 50 per cent of challenged mice, and protection can reach up to 100 per cent with two intraperitoneal doses. ^[2]

Finally, privilege and immune escape shape oncology practice in sanctuary sites. Primary testicular lymphoma is treated according to Lugano staging, with radical orchiectomy followed by 6 to 8 cycles of R-CHOP and prophylactic radiotherapy (25 to 30 Gy) to the contralateral scrotum, and given the high propensity for central nervous system relapse, central nervous system prophylaxis is recommended, including intrathecal methotrexate with or without cytarabine and two cycles of high-dose methotrexate. In central nervous system lymphoma, bispecific antibodies bridge CD3 on T cells and tumour antigens and act independently of MHC, so they may retain cytotoxicity where tumours lack MHC and cause immune escape; glofitamab partially penetrates the blood-brain barrier, with an average cerebrospinal fluid concentration only 0.1 to 0.4 per cent of that in peripheral blood, yet sufficient to induce responses, and in one series patients received glofitamab with a median of 6.5 cycles. Understanding privilege therefore changes drug choice, dose intensity and surveillance, not just textbook descriptions. ^[5]