T-cell development, differentiation and regulatory T cells

T cells begin life as thymic progenitor thymocytes and leave the thymus as mature, self-restricted, self-tolerant lymphocytes. Between those two points lies a sequence of developmental decisions that can be understood from first principles. A progenitor must first commit to the T lineage, then rearrange its T cell receptor genes, then prove that its receptor can recognise self-MHC molecules at all, then prove that it is not dangerously reactive to self-peptide, and finally choose between the major functional fates available to a mature T cell. The thymus provides the stromal cells, the self-antigens and the cytokines that make these decisions possible. When the process fails, the consequences are immunodeficiency, autoimmunity or both.

The mature T cell compartment is not a single population. T cells consist of various subsets, that is, Th1, Th2, Th17, T follicular helper, cytotoxic T lymphocytes and regulatory T cells, and these originate from thymic progenitor thymocytes^[1]. Each subset has a characteristic cytokine profile and a characteristic transcriptional programme. The classical division is into CD4+ helper cells, which recognise peptide on MHC class II and orchestrate other cells, and CD8+ cytotoxic cells, which recognise peptide on MHC class I and kill infected or transformed cells. Within the CD4+ compartment, further specialisation allows the immune system to tailor its response to the type of threat: intracellular microbes, helminths, extracellular bacteria and fungi, or the need for restraint.

Two forms of tolerance are generated in the thymus. Self-tolerance is induced in thymus either by negative selection or by natural regulatory T cell development^[1]. These are not alternative names for the same process. Negative selection deletes strongly self-reactive thymocytes, whereas natural regulatory T cell development converts some self-reactive thymocytes into cells whose job is to suppress others. Most of the natural regulatory T cells are derived from CD4+ single-positive thymocytes residing in the medullary compartment of the thymus^[1]. This anatomical detail matters: the medulla is where thymocytes encounter the widest range of self-antigens, including tissue-restricted antigens expressed by medullary thymic epithelial cells, and it is therefore the logical place to make the final tolerance decisions.

The specificity of the decision depends on which self-antigen is seen and where. It is hypothesised that tolerance of uncommon self-antigens such as myosin, which is usually presented after muscle injury, is preferentially recognised by the T cell receptor and mediated by natural regulatory T cells^[1]. By contrast, cells that are involved in chronic engagement of T cell receptor and CD28 signalling by recognising ubiquitous antigen, for example albumin, the fifth component of complement or insulin, are negatively selected^[1]. Decreased presentation of cognate antigens on medullary thymic epithelial cells or dendritic cells can induce natural regulatory T cell development^[1]. Distinct antigen-presenting cell subsets may preserve different T cell receptor specificities and their ability to mediate negative selection^[1].

Once T cells leave the thymus they are still not fixed. Naive CD4+ T cells are programmed by various cytokines and other factors from dendritic cells to produce transcription factors^[1]. The environment at the time of priming therefore determines which programme is installed. This is why the same antigen can drive different outcomes depending on the adjuvant, the dose and the route. It also explains why innate signals are so influential: they change the cytokines made by dendritic cells, and those cytokines in turn change the transcription factors expressed by the responding T cell.

Metabolism is part of the story. Although it has been well appreciated that T cell activation is accompanied by a metabolic switch to glycolysis, how glycolytic activity is regulated during T cell differentiation is poorly understood^[1]. When naive T cells were activated under polarising conditions in vitro, Th1, Th2 and Th17 cells contained much higher glycolytic activity than regulatory T cells, indicating strong up-regulation of glucose metabolism in effector T cells^[1]. This observation is a useful organising principle: effector lineages are metabolically demanding and depend on glucose, whereas regulatory lineages can use different fuels. It also opens therapeutic possibilities, because metabolic inhibitors can shift the balance between effector and regulatory fates.

Regulatory T cells deserve early definition because they appear throughout the pathway. Foxp3 is the regulatory T cell master regulator, and it has been shown to antagonise ROR-gamma-t function and Th17 differentiation^[1]. Regulatory T cells are themselves heterogeneous. All major types of regulatory T cells are involved in oral tolerance, including thymic-derived natural regulatory T cells, mucosally induced regulatory T cells, IL-10 secreting CD4+CD25lowCD45RBlow type 1 regulatory T cells, TGF-beta-dependent latency-associated peptide-positive Th3 type regulatory T cells and CD8+ regulatory T cells^[1]. This heterogeneity means that a single marker such as CD25 will never capture the whole regulatory compartment.

Positive selection and negative selection are the two thymic checkpoints that together produce a useful and safe T cell repertoire. The logic is easiest to grasp as a pair of questions. Positive selection asks: can this T cell receptor recognise self-MHC at all? Negative selection asks: does it recognise self-MHC plus self-peptide too strongly? A thymocyte must answer yes to the first and no to the second. If it fails the first, it dies by neglect. If it fails the second, it is deleted, or in some cases diverted into the regulatory lineage.

Negative selection is not a single event but a family of outcomes shaped by antigen presentation. Cells that engage their T cell receptor and CD28 chronically, recognising ubiquitous antigen such as albumin, the fifth component of complement or insulin, are negatively selected^[1]. The word chronically is important. A brief, low-avidity encounter may allow survival, whereas sustained signalling drives deletion. This is why the availability of self-antigen in the thymus matters so much: a self-antigen that is never presented cannot delete the clones that recognise it.

Natural regulatory T cell development is the third outcome. Decreased presentation of cognate antigens on medullary thymic epithelial cells or dendritic cells can induce natural regulatory T cell development^[1]. This suggests that the regulatory fate is favoured when the selecting antigen is present at low abundance or on a particular antigen-presenting cell subset. The functional consequence is a population of CD4+ cells that recognises self-antigen but suppresses rather than attacks. Most natural regulatory T cells are derived from CD4+ single-positive thymocytes residing in the medullary compartment^[1], which places the decision at the final stage of thymic maturation.

The antigen-presenting cell matters as much as the antigen. Distinct antigen-presenting cell subsets may preserve different T cell receptor specificities and their ability to mediate negative selection^[1]. Dendritic cells and medullary thymic epithelial cells differ in their capacity to process and present antigens, in their expression of co-stimulatory molecules and in their cytokine output. A thymocyte that encounters its antigen on one cell type may be deleted, while the same thymocyte encountering the same antigen on another cell type may become regulatory. This is a good example of how cellular context converts a single molecular interaction into divergent fates.

The clinical importance of these checkpoints is considerable. Defects in negative selection allow self-reactive T cells to escape, which predisposes to autoimmune disease. Defects in natural regulatory T cell development cause a severe multisystem autoimmune syndrome, because the peripheral restraint is missing. Defects in positive selection or in T cell receptor signalling cause immunodeficiency, because too few T cells reach maturity. The same pathway can therefore produce opposite clinical pictures depending on which step fails.

Beyond the thymus, tolerance is reinforced peripherally. Regulatory T cells are central to this. In various autoimmune syndromes, quantitative or qualitative defects in CD4+CD25+ regulatory T cells have been reported^[1]. In contrast, in various forms of cancer, an increase in the number of CD4+CD25+ regulatory T cells was noted and related to downregulation of antitumour immune responses and poor prognosis^[1]. A reduction in the number of these cells could be beneficial in that setting, as demonstrated by a study in which CD4+CD25+ regulatory T cell depletion enhanced the antitumour immune response in renal cell cancer^[1]. This is the mirror-image problem of autoimmunity: too little regulation causes self-damage, whereas too much regulation allows tumours to escape.

T helper type 1 cells are the principal helpers for cell-mediated immunity. Th1 cells produce interferon-gamma and are mainly devoted to protection against intracellular microbes. This makes teleological sense. An intracellular organism such as Mycobacterium tuberculosis or Salmonella is hidden inside macrophages, where antibody cannot reach it. The immune system therefore needs a way to activate the infected macrophage so that it kills the organism in its own phagosome. Interferon-gamma is that signal. It upregulates macrophage antimicrobial mechanisms, enhances antigen presentation and recruits further effector cells to the site.

The transcriptional programme of Th1 cells centres on T-bet. T box expressed in T cells is a major factor for Th1 cell differentiation and interferon-gamma production^[1]. It can induce chromatin remodelling of interferon-gamma alleles and IL-12 receptor beta-2 expression, and this promotes interferon-gamma production as well as Th1 cell expansion induced by IL-12^[1]. In mature Th1 cells, however, reiteration of interferon-gamma expression and stable chromatin remodelling are relatively independent of T-bet activity^[1]. This is a subtle but important point: T-bet is required to establish the programme, but the established programme can run without it. Students often over-simplify this as T-bet equals Th1, when the reality is that T-bet initiates and then the epigenetic state maintains.

Signal transducer and activator of transcription proteins are also involved. STAT4 and STAT1 are involved in Th1 cell differentiation, and STAT4 is activated by IL-12 leading to Th1 and Th17 cell differentiation^[1]. Interferon-gamma production also occurs with nuclear factor kappa-B, with multiple cis elements being involved^[1]. The redundancy here is instructive: the cell integrates several signalling inputs, and no single knockout completely abolishes the response in every context.

The polarising conditions for Th1 are well established in principle. Interleukin-12 from dendritic cells activates STAT4, which induces T-bet, which in turn upregulates the IL-12 receptor beta-2 chain, creating a positive feedback loop that stabilises the Th1 fate. Interferon-gamma produced by the responding cell acts in an autocrine loop through STAT1, reinforcing T-bet expression. This mutual reinforcement explains why Th1 commitment is relatively stable once established.

Metabolically, Th1 cells behave as effector cells. Th1, Th2 and Th17 cells contained much higher glycolytic activity than regulatory T cells, indicating strong up-regulation of glucose metabolism in effector T cells^[1]. Treatment with 2-deoxyglucose, which inhibits glycolytic enzyme activity directly, was capable of diminishing Th1 and Th2 differentiation^[1]. Interestingly, HIF1-alpha was not strongly induced under Th1- and Th2-polarising conditions and was not involved in their functional differentiation^[1]. The glycolytic pathway therefore contributes to Th1 differentiation in a HIF1-alpha-independent manner^[1]. This is a good reminder that two cells can both be glycolytic while using different molecular machinery to get there.

The innate signals that shape Th1 development are also worth noting. A lower dose of lipopolysaccharide (0.1 micrograms), through TLR4 signalling, induced a Th2 response to inhaled antigens in a murine allergic sensitisation model, whereas high doses of lipopolysaccharide (100 micrograms) with antigen resulted in a Th1 response^[1]. This dose-dependence is clinically relevant because it shows that the quantity of an innate stimulus, not just its identity, can determine the adaptive outcome. It also explains why environmental endotoxin exposure may protect against allergy while low-level exposure may favour it.

T helper type 2 cells are the principal helpers for humoral and anti-helminth immunity. Th2 cells produce IL-4, IL-5, IL-9 and IL-13 and are involved in protection against gastrointestinal nematodes, but are also responsible for allergic disorders. The same cytokines that expel a worm can drive bronchial hyperreactivity, mucus production and eosinophilia. This duality is the central clinical fact about Th2 immunity: it is protective against one class of pathogen and pathogenic in the context of allergy.

The master regulator of Th2 is GATA3, a member of the GATA transcription factor family capable of binding to the DNA sequence GATA^[1]. Without GATA3, Th2 cell differentiation is completely abolished both in vivo and in vitro^[1]. This is one of the strongest statements of necessity in T helper biology. GATA3 can bind to 1279 genes in Th2 cells and to 17 genes among 26 highly Th2-specific STAT6-dependent inducible genes^[1]. Among the 26 Th2-specific genes, 10 showed GATA3-dependent transcription while the remaining 16 genes were STAT6 dependent^[1]. Production of Th2 cytokines is also promoted by GATA3 binding to promoters of IL-5 and IL-13 and to enhancers of IL-4^[1]. GATA3 has the ability to instruct Th2 commitment, promote Th2 cell expansion and suppress Th1 cell differentiation, thus facilitating Th2 differentiation^[1].

STAT6 is also involved in the expression of Th2-specific cytokines. For example, IL-24 is mediated by the coordinate action of STAT6 and c-Jun transcription factors at the transcriptional level^[1]. Recently it was reported that STAT3 cooperates with STAT6 in promoting Th2 cell development^[1]. A strong STAT5 signalling, correlated with higher expression of CD25, is required for Th2 and induced regulatory T cell differentiation, whereas weak STAT5 signalling causes cell proliferation and survival of Th1 and Th17 cells^[1]. In vivo, promiscuous expression of an activated form of STAT5 suppresses the production of both Th1 and Th17 cytokines and promotes the development of Th2 lineage cells^[1]. This demonstrates that the strength of a shared cytokine signal can be the deciding variable between lineages.

The polarising conditions for Th2 are also well defined. When cognate antigen stimulation is present, IL-4 and IL-2 are required by the naive CD4+ T cells to differentiate into IL-4-producing Th2 cells^[1]. Interleukin-4 activates STAT6, which induces GATA3, which in turn promotes IL-4 production, creating another positive feedback loop. Interleukin-2 activates STAT5, which reinforces the programme. The requirement for both signals explains why Th2 responses are particularly dependent on the local cytokine milieu at the time of priming.

Antigen dose is an additional determinant. A lower concentration of ovalbumin peptide (1 and 10 nanograms per millilitre) induced Th2 commitment, while higher concentrations (1 microgram per millilitre and 100 nanograms per millilitre) failed to elicit Th2 development^[1]. This is a striking observation: the same peptide at different doses can drive different lineages. Stimulation of CD4+ T cells with dendritic cells along with TLR2 or TLR9 agonists in the presence of 10 nanograms per millilitre of ovalbumin peptide, the optimal antigen concentration for Th2 development, resulted in suppression of IL-4 production and Th2 development^[1]. This suggests that TLR-activated dendritic cells can block Th2 lineage commitment independent of antigen dosage^[1].

Metabolically, Th2 cells resemble Th1 cells more than they resemble regulatory T cells. Th1, Th2 and Th17 cells contained much higher glycolytic activity than regulatory T cells^[1]. 2-Deoxyglucose treatment was capable of diminishing Th2 differentiation^[1]. Yet HIF1-alpha was not strongly induced under Th2-polarising conditions and was not involved in Th2 functional differentiation^[1]. So Th2 cells are glycolytic but not HIF1-alpha-dependent. This kind of dissociation is useful clinically, because it means that metabolic drugs may have selective effects on different effector lineages.

T helper 17 cells are a distinct subset of CD4+ T cells that are protective against extracellular microbes but are responsible for autoimmune disorders in mice. Their properties in humans are only partially known, but the broad outline is clear. Th17 cells produce IL-17 and are important for recruiting neutrophils and for defending epithelial surfaces against extracellular bacteria and fungi. When the response is misdirected or excessive, it drives pathology in psoriasis, psoriatic arthritis, ankylosing spondylitis and inflammatory bowel disease.

Th17 cells were recognised as a lineage distinct from Th1, Th2 and Th0. A novel subset of Th, distinct from Th1, Th2 and Th0 cells and called Th17, has been described. Analysis of transcription-factor expression by IL-17-producing CD4+ T cells showed that Th17 effectors lacked expression of T-bet and Hlx in comparison with Th1-polarised cells, and of GATA-3 compared with Th2-polarised populations. This supported and extended previous findings that identified Th17 as a product of an effector lineage distinct from Th1 and Th2. The orphan nuclear receptor ROR-gamma-t directs the differentiation programme of mouse Th17.

The cytokine requirements for Th17 differentiation are instructive. Transforming growth factor-beta stimulates naive CD4+ T cells either to differentiate into Th17 cells in the presence of IL-6, or alternatively to differentiate into induced regulatory T cells in the presence of IL-2 (or IL-1 beta in human)^[1]. In the absence of IL-6 and in the presence of TGF-beta and IL-21, Th17 cells can also be induced^[1]. Moreover, IL-1 beta and TNF-alpha were found to amplify the Th17 response induced by TGF-beta and IL-6, but could not substitute for either of these cytokines. The picture is therefore one of a shared precursor: TGF-beta alone favours regulatory T cells, whereas TGF-beta plus IL-6 diverts the cell towards Th17.

STAT3 is involved in Th17 cell differentiation, expansion and maintenance^[1]. Stimulation of the common precursor cell of regulatory T cell and Th17 by IL-6 activates STAT3 signalling and induces IL-21 expression^[1]. IL-21 induces Th17 differentiation, suppresses Foxp3 expression and maintains a sustained STAT3 activation in a self-service autocrine pattern, that is, Th17 cells secrete IL-21, which in turn causes Th17 cells to induce cell differentiation^[1]. STAT3 can also be activated by IL-23 and is responsible for the induction of ROR-gamma-t and IL-23 receptor, allowing the persistence of Th17 cells^[1]. This two-stage model is important: TGF-beta plus IL-6 initiates the lineage, and IL-23 maintains it.

Metabolically, Th17 cells are distinctive. HIF1-alpha was strongly induced at the protein level in T cells undergoing Th17 differentiation, whereas the lowest levels were found in regulatory T cells, and mRNA levels of HIF1-alpha were also highly induced in T cells undergoing Th17 differentiation^[1]. This identifies a selective up-regulation of HIF1-alpha under Th17-polarising conditions^[1]. Naive T cells from wild-type and HIF1-alpha-deficient mice were differentiated under Th17 conditions, and HIF1-alpha-deficient cells exhibited a reduction of IL-17-positive cells^[1]. CFSE dilution assays revealed a similar degree of proliferation between wild-type and HIF1-alpha-deficient cells, indicating that the defective IL-17 production was largely independent of cell expansion^[1].

The downstream consequences of HIF1-alpha deficiency are specific. Upon T cell receptor restimulation, effector Th17 cells lacking HIF1-alpha secreted a lower amount of IL-17 but a comparable level of IL-2 relative to wild-type cells^[1]. Real-time PCR analyses revealed decreased expression of the Th17 family cytokines IL-17, IL-17F, IL-21 and IL-22^[1]. Thus, HIF1-alpha deficiency results in impairment in the Th17 differentiation programme^[1]. In contrast, in vitro differentiation of Th1 and Th2 effector cells was largely independent of HIF1-alpha^[1]. Expression of ROR-alpha, ROR-gamma-t, AHR, IRF4, IL-6 receptor, IL-21 receptor and phosphorylation of Stat3 were comparable between differentiating wild-type and HIF1-alpha-deficient cells, whereas expression of IL-23 receptor was significantly down-regulated and Foxp3 was up-regulated in HIF1-alpha-deficient cells under Th17-polarising conditions^[1].

Human Th17 biology adds further complexity. Th17 cells, some of which produce both IL-17 and interferon-gamma (Th17/Th1), are present in the gut of patients with Crohn's disease. Both Th17 and Th17/Th1 clones showed selective expression of IL-23 receptor, CCR6 and the transcription factor ROR-gamma-t, and they exhibited similar functional features, such as the ability to help B cells, low cytotoxicity and poor susceptibility to regulation by autologous regulatory T cells. These subsets also expressed the Th1 transcription factor T-bet, and stimulation in the presence of IL-12 down-regulated ROR-gamma-t and IL-17 production but induced interferon-gamma. The identification of a subset sharing features of both Th1 and Th17 may raise new issues concerning developmental and functional relationships between the two lineages.

The therapeutic relevance follows directly. Because Th17 cells depend on IL-23 for persistence and on ROR-gamma-t for their programme, these molecules are attractive drug targets in psoriasis and inflammatory bowel disease. Because HIF1-alpha is selectively required for Th17 differentiation, metabolic manipulation could in principle tilt the balance away from Th17. And because Th17 cells are less susceptible than Th1 or Th2 cells to suppression by autologous regulatory T cells, they may be harder to control once established.

Cytotoxic T lymphocytes and regulatory T cells sit at opposite ends of the effector spectrum. One exists to kill, the other to restrain killing. Both arise from the same thymic output, and both are essential: a host without cytotoxic T cells cannot clear viruses or tumours, whereas a host without regulatory T cells destroys its own tissues. Understanding them as a pair is more useful than learning them as separate chapters.

When exposed to foreign antigens, peripheral naive CD8+ T cells differentiate into two reciprocal subsets: short-lived effector T cells, that is, cytotoxic T lymphocytes, and long-lived memory T cells^[1]. Memory T cells can be subdivided into central memory and effector memory cells. Central memory cells express high levels of CCR7 and CD62L and lack immediate effector function, but they efficiently stimulate dendritic cells in secondary lymphoid organs and induce a new wave of effector cells when a secondary challenge occurs^[1]. Effector memory cells express low levels of CCR7 and CD62L, migrate to the infection site and produce cytokines and cytolytic molecules^[1]. Effector memory cells possess most features of cytotoxic T lymphocytes, yet they persist after the elimination of the invading pathogen^[1]. A newer subset with stem-cell-like properties, the memory stem T cell, has been identified in humans and mice^[1].

The transcriptional control of CD8+ differentiation is a study in balance. T-bet, eomesodermin, Bcl-6 and Blimp-1 are involved in CD8+ T cell differentiation^[1]. T-bet is the master regulator of CD8+ T cells; its expression is responsible for interferon-gamma production and it participates in the activation of cytolytic genes such as granzyme B and perforin^[1]. Eomes is another member of the T-box family and a key transcriptional factor for CD8+ T cell differentiation, and T-bet and Eomes cooperate redundantly to induce effector CD8+ differentiation while also acting reciprocally to induce memory development^[1]. The balance is not fixed: T-bet promotes differentiation of short-lived effector CD8+ T cells at the expense of central memory cells, whereas Eomes expression favours memory CD8+ differentiation^[1]. Differences in T-bet quantity between lineages may arise from asymmetric degradation, since proteasomes are unequally distributed during asymmetric cell division^[1].

Bcl-6 and Blimp-1 are transcriptional repressors that act in opposition. Blimp-1 expression is required for terminal differentiation of effector CD8+ T cells, that is, the short-lived cytotoxic lymphocytes^[1]. Bcl-6 probably works as a reciprocal regulator of Blimp-1 during CD8+ differentiation^[1]. In general, lymphocytes with higher Bcl-6 expression exhibit greater proliferative capacity, less secretory capacity and promote memory development, whereas lymphocytes with higher Blimp-1 expression exhibit lower proliferative capacity and greater secretory capacity and are more conducive to cytotoxic lymphocyte development^[1]. Blimp-1 is also highly expressed in exhausted CD8+ T cells, and T-bet can induce Blimp-1 transcription via enhanced IL-2 receptor signalling^[1]. STAT5 is required for maintaining the phenotype of effector CD8+ T cells and for inducing the anti-apoptotic molecule Bcl-2 in response to IL-7 and IL-15^[1].

Regulatory T cells are defined by FOXP3. Treg development is controlled by the transcription factor Foxp3, and mutation of the Foxp3 gene results in fatal autoimmune disorders in humans, such as IPEX, or in mice, such as lymphoproliferative disorder^[1]. Stable expression of Foxp3 is essential for immune homeostasis, and Foxp3 is required for Treg lineage commitment, differentiation, expansion and function^[1]. Sustained Foxp3 expression in the mature Treg cell is essential to maintain the existing phenotype and to execute immunosuppressive function, and reduced or abolished Foxp3 production results in acquisition of effector T cell properties with production of inflammatory cytokines^[1].

The phenotype is well characterised. Immunosuppressive CD4+CD25+ regulatory T cells constitute a small subset, 5 to 10 per cent, of CD4+ T helper cells that develop in the thymus^[1]. They are characterised by expression of the IL-2 receptor alpha chain CD25, cytotoxic T lymphocyte-associated antigen 4 and glucocorticoid-induced TNF receptor^[1]. The most specific molecular marker, functionally involved in immune suppression, is the FOXP3 protein, and in its absence CD4+CD25+ regulatory T cells do not develop^[1]. FOXP3 codes a 48-kilodalton protein of 431 amino acids from the forkhead/winged-helix transcription factor family, and the forkhead box itself consists of 80 to 100 amino acids forming a DNA-binding motif^[1].

Suppression is not a single mechanism. Several studies suggest immunosuppression to be cell contact-dependent, while others demonstrate that suppression can also be cell contact-independent, for example via production of the immunosuppressive cytokines IL-10 and TGF-beta, preferential IL-2 consumption by CD4+CD25+ regulatory T cells, or direct lysis of T cells via perforin and granzymes^[1]. Initiation of immunosuppression requires antigen exposure, but suppressive effects can extend through bystander immunosuppression, and these cells suppress autoimmune, antimicrobial and antitumour immune responses^[1]. Treg cells suppress Th1 function through inhibition of interferon-gamma transcription during Th1 priming without disrupting T-bet expression and Th1 programming, and this suppression is either IL-10 dependent or independent depending on the target T cell stage of activation and its tissue location^[1].

The interaction between antigen-presenting cells and lymphocytes is the point where innate sensing becomes adaptive instruction. Macrophages and dendritic cells both present antigen, but they differ in emphasis: dendritic cells are specialised for migration and T cell priming, whereas macrophages are specialised for phagocytosis, killing and cytokine production. Both can act as antigen-presenting cells, and the rules that govern the encounter are the same.

Antigen-presenting cells take up antigen and digest it in the cytosol to process the epitope, which is then presented together with MHC molecules to the T cell receptor on the T cell surface^[1]. Simultaneously, antigen-presenting cells secrete co-stimulatory molecules such as CD80 and CD86 that bind the co-stimulatory receptor CD28 on T cells^[1]. Thus all three elements are required for T cell activation: epitope, MHC molecule and co-stimulation signal^[1]. This three-signal model is the single most useful framework for understanding macrophage-lymphocyte interaction, because it explains why antigen alone is not enough and why adjuvants matter.

The conversation is bidirectional. Upon T cell receptor activation, T cells produce CD154, also called CD40 ligand, to bind CD40 on the cell surface of antigen-presenting cells and further activate them^[1]. This is the mechanism by which a T cell licenses a macrophage or dendritic cell to upregulate co-stimulatory molecules and produce inflammatory cytokines. The lineage commitment of T helper cells is determined by the cytokine milieu, transcription factors and co-stimulatory molecules such as CD28 and CD154, and the transcription factors involved are activated by T cell receptor signalling^[1]. IL-12 and interferon-gamma are essential for induction of Th1 cells^[1], which is the cytokine loop by which a macrophage that has ingested an intracellular organism recruits T cell help.

Dendritic cell maturation follows the same logic. Upon recognition of foreign antigen through Toll-like receptor and pathogen-associated molecular pattern interaction, immature dendritic cells resident in tissues mature into professional antigen-presenting cells to induce effector and memory T cell responses in lymphoid organs, and dendritic cells are also capable of inducing antigen-specific T cell tolerance and immunosuppression^[1]. The same cell type can therefore prime or tolerise depending on the signals it receives, which is why the context of antigen encounter matters more than the antigen itself.

Peripheral tolerance is maintained by several mechanisms that involve macrophage and lymphocyte interactions. Peripheral deletion of the autoreactive T cell lineage is mediated by apoptosis involving activation of the Fas receptor by Fas ligand and inactivation of the survival protein Bcl-2 by its antagonist Bim^[1]. A non-apoptotic mechanism has been identified in which autoreactive CD8+ T cells actively invade hepatocytes and are degraded in the endosome/lysosome of the hepatocytes, a process known as emperipolesis^[1]. The invasion depends on T cell activation, filamentous actin reorganisation, myosin light chain kinase and other kinases such as PI3K^[1]. Inhibition of this suicide emperipolesis by wortmannin is associated with accumulation of autoreactive CD8+ T cells in the liver, and breach of tolerance results in autoimmune hepatitis^[1].

Co-inhibitory receptors provide a second layer of control. By interrupting co-stimulation, functional tolerance of T cells, that is anergy, can be developed and maintained by counter-regulatory receptors such as cytotoxic T lymphocyte-associated antigen 4, which shares structural similarity with CD28 and binds CD80 and CD86, blocking CD28 co-stimulation^[1]. Programmed cell death-1 is also crucial for maintenance of peripheral tolerance^[1]. These molecules are the physiological brakes that prevent an antigen-presenting cell from over-stimulating a T cell, and they are the targets of modern checkpoint therapy.

Lymph node stromal cells add a further layer. Lymph nodes are a primary location where peripheral tolerance takes place, and lymph node stromal cells, similar to medullary thymic epithelial cells, are able to express a variety of tissue-specific antigens to induce immune tolerance of T cells^[1]. Peripheral expression of these antigens is either AIRE dependent or independent, and another regulating factor, deformed epidermal autoregulatory factor 1, is also involved^[1]. Deaf1 variant isoforms inhibit the transcriptional activity of canonical Deaf1 and suppress peripheral tissue antigen expression, and Deaf1 transcript has been detected in every subset of lymph node stromal cells^[1]. All subsets of lymph node stromal cells can express tissue-specific antigens and present them to activate antigen-specific CD8+ T cells under both steady-state and inflammatory conditions^[1].

T cell intrinsic signalling also determines susceptibility to regulation. T cell intrinsic TRAF6 is essential in the maintenance of peripheral tolerance, and deletion of TRAF6 in T cells leads to hyperactivation of the PI3K-Akt pathway and increased resistance of effector T cells to suppression by CD4+CD25+ regulatory T cells, finally resulting in multiorgan inflammatory disease^[1]. Because TRAF6 is an important adaptor in Toll-like receptor signalling, activation of Toll-like receptors expressed in T cells may be involved in maintaining T cell susceptibility to regulatory T cells^[1]. The macrophage-lymphocyte relationship is therefore not only about activation but also about keeping activation within safe limits.

IPEX is the clinical proof that regulatory T cells are not optional. The acronym stands for immune dysregulation, polyendocrinopathy, enteropathy, X-linked, and the condition is caused by loss of functional FOXP3. It is an X-linked recessive disorder caused by mutations in the FOXP3 gene located on the X chromosome at Xq11.23 to Xq13.3, consisting of 11 coding exons^[1]. After the initial discovery of the gene defect in scurfy mice, which develop a disease highly homologous to that of patients with IPEX, mutations were demonstrated in the orthologous human gene^[1].

The genetics are more heterogeneous than the name suggests. IPEX might be genetically more heterogeneous than initially presumed, as a few patients with IPEX were found not to carry mutations in FOXP3, and in one family that included an affected female, IPEX was linked to an autosomal locus^[1]. In these cases, IPEX could be the result of regulatory or conditional mutations outside FOXP3 coding regions, and a mutation in a region following the final coding exon has been identified that results in a reduction in FOXP3 mRNA expression^[1]. The inheritance pattern is X-linked recessive in the case of FOXP3 mutations, with no reported female patients and asymptomatic female carriers of the gene mutation, but the disease might be more heterogeneous than initially presumed^[1].

The incidence of IPEX is not known, but it is likely to be rare, although unfamiliarity with the disease might be a contributing factor^[1]. Symptoms usually start in the perinatal period or in early childhood, but they can also first arise in adulthood, and first symptoms usually include type 1 diabetes and secretory diarrhoea or ileus^[1]. This age range is important clinically: a young infant with intractable diarrhoea, diabetes and eczema should prompt consideration of IPEX, but so should an adult with a compatible constellation and no other explanation.

The cellular basis is loss of regulatory T cell function. In the absence of FOXP3, CD4+CD25+ regulatory T cells do not develop^[1]. Reduced or abolished FOXP3 production in Treg cells results in acquisition of effector T cell properties with production of inflammatory cytokines^[1]. This explains why the syndrome is not simply a failure of suppression but an active inflammatory disorder: the cells that should restrain effector responses either fail to develop or convert into cells that produce inflammatory cytokines.

IPEX also illustrates the central regulatory role of regulatory T cells in the immune system more broadly. Apart from the severe defects in CD4+CD25+ regulatory T cells that result in IPEX, other, perhaps more subtle, changes in the CD4+CD25+ regulatory T cell population can contribute to the pathogenesis of disease^[1]. In various autoimmune syndromes, quantitative or qualitative defects in CD4+CD25+ regulatory T cells have been reported^[1]. This spectrum is useful clinically, because it suggests that partial Treg dysfunction may underlie common autoimmune diseases that do not meet criteria for IPEX.

The mirror image occurs in cancer. In various forms of cancer, an increase in the number of CD4+CD25+ regulatory T cells was noted and related to downregulation of antitumour immune responses and poor prognosis^[1]. In this situation, a reduction in the number of CD4+CD25+ regulatory T cells could be beneficial, as demonstrated by a study in which CD4+CD25+ regulatory T cell depletion enhanced the antitumour immune response in renal cell cancer^[1]. The same cell population that prevents autoimmunity can prevent tumour rejection, which is why therapeutic manipulation requires precision rather than simple depletion.

The developmental biology of the thymic regulatory lineage is also relevant to IPEX. It has been suggested that FOXP3-negative natural regulatory T cell precursors, induced by T cell receptor signalling, can use IL-2, IL-15 or IL-7 to activate Foxp3 expression without the need for additional T cell receptor signals^[1]. Natural regulatory T cell development is believed to begin early at the double-positive stage in the paediatric thymus, and Foxp3-positive double-positive thymocytes with a functional IL-7 receptor and upregulated Bcl-2 protect themselves from negative selection^[1]. Hassall's corpuscles, groups of epithelial cells in the thymic medulla, may serve as specialised small niches required to support nascent natural regulatory T cell development^[1]. STAT5 binds the promoter and the first intron of the Foxp3 gene to promote regulatory T cell development, and loss of STAT5 activation abolishes regulatory T cell differentiation, although Foxp3 can be induced in the absence of STAT5 in developing thymocytes and maintenance of Foxp3 in mature regulatory T cells is STAT5 independent^[1].

The first pitfall is to assume that every CD4+CD25+FOXP3+ cell is a regulatory T cell. Not all the CD4+CD25+Foxp3+ cells from peripheral blood activated by heat shock protein 60 are regulatory T cells, because activated CD4+ effector T cells can also transiently express Foxp3, and only cells with a CD4+CD25+Foxp3+CD30+ phenotype possess suppressive function^[1]. A marker combination that is adequate in the thymus may be misleading in inflamed peripheral tissue.

The second pitfall is to assume that Toll-like receptor expression on T cells is settled. Toll-like receptor expression in T cells is controversial, and the discrepancy may be attributed in part to different protocols used for T cell purification and different ligands used for Toll-like receptor activation^[1]. When polyinosinic-polycytidylic acid and CpG DNA were added to murine CD4+ T cell cultures that were T cell receptor activated by anti-CD3 antibody, TLR3 and TLR9 expression was upregulated with enhanced survival, whereas levels of TLR2 and TLR4 were undetectable when peptidoglycan and lipopolysaccharide were used^[1]. Human CD4+CD25+CD127-negative regulatory T cells isolated from peripheral blood mononuclear cells do not express TLR2^[1]. Claims that a given Toll-like receptor is or is not present on human regulatory T cells should therefore be read with attention to the method.

The third pitfall is to generalise from one ligand to the whole family. Activation of all the identified Toll-like receptors except TLR3 results in signalling through the MyD88-NF-kappa-B pathway, yet it is not known why activation of Toll-like receptors by different ligands results in different outcomes although they act via a common pathway^[1]. The direct effect of an individual Toll-like receptor ligand on regulatory T cells is completely different from that of another, even though almost all share a common pathway^[1]. Flagellin augments suppressive capacity and increases Foxp3, whereas triggering TLR8 reverses suppressive function^[1].

The fourth pitfall is contamination. The purity of CD4+CD25+ regulatory T cell preparations used in some studies was above 90 per cent, which implies possible contamination by other cell types, and contaminated antigen-presenting cells within the regulatory T cell population may promote suppressive function through TLR4 signalling triggered in the antigen-presenting cell rather than through TLR2 signalling in the regulatory T cell^[1]. Potential contamination of commercial lipopolysaccharide preparations with TLR2 ligands, or impurities in the cell preparations, may create discrepant results^[1]. A finding that depends on an impure reagent or an impure cell population is not a finding about the cell of interest.

The fifth pitfall is species extrapolation. Effective antitumour immunity elicited by CpG DNA in mouse is not seen in humans^[1]. Mice and humans differ in Toll-like receptor expression patterns, in the distribution of TLR9 and in the responsiveness of regulatory T cell subsets. A mouse result is a hypothesis about human biology, not a proof.

The sixth pitfall is to treat regulatory T cells as fixed. Regulatory T cell phenotypic plasticity is seen in their expression of proinflammatory cytokines such as IL-17, interferon-gamma or IL-2 under certain conditions and in their reprogramming into T helper-like cells^[1]. Mice systemically administered high doses of CpG oligodeoxynucleotide at 50 to 100 micrograms per mouse show activation of naive regulatory T cells in the spleen to acquire potent suppressor activity, mediated by the immunoregulatory enzyme IDO in plasmacytoid dendritic cells^[1]. When IDO was blocked, CpG treatment stimulated plasmacytoid dendritic cells to express IL-6, which in turn reprogrammed Foxp3 lineage regulatory T cells to express IL-17 and become Th17-like effector T cells^[1]. CD4+Foxp3+ regulatory T cells can also be reprogrammed into the T follicular helper lineage in mouse Peyer's patches under interaction with B cells and loss of Foxp3 expression^[1].

The seventh pitfall is to forget that suppression can be measured wrongly. Suppression depends on the ratio of regulatory to effector cells, on the activation state of both populations and on the assay used. Flagellin increased the suppressive capacity of CD4+CD25+ regulatory T cells such that a suppressive ratio of 1 to 81 improved to 1 to 243^[1]. A single ratio tested in a single assay may miss a real biological effect or invent a spurious one.

The clinical uses of this material fall into four groups: diagnosis of immune dysregulation, manipulation of regulatory T cells in cancer, use of Toll-like receptor ligands as adjuvants, and management of autoimmune disease through checkpoint and metabolic pathways.

In diagnosis, IPEX is the paradigm. A male infant with intractable secretory diarrhoea, type 1 diabetes and eczema should be investigated for FOXP3 mutations, remembering that symptoms usually start in the perinatal period or early childhood but can first arise in adulthood, and that the first symptoms usually include type 1 diabetes and secretory diarrhoea or ileus^[1]. Because a few patients with IPEX do not carry FOXP3 coding mutations, and because one family with an affected female was linked to an autosomal locus, genetic testing should include regulatory and intronic regions^[1]. The diagnosis matters because it changes management towards aggressive immunological control and, in severe cases, transplantation.

In oncology, regulatory T cells are a barrier to effective immunity. An increase in the number of CD4+CD25+ regulatory T cells has been related to downregulation of antitumour immune responses and poor prognosis, and depletion of these cells enhanced the antitumour immune response in renal cell cancer^[1]. Pam3CSK4, a TLR1/TLR2 ligand, can induce tumour remission in severe combined immunodeficiency mice by diminishing the suppressive function of Foxp3-positive regulatory T cells and enhancing the cytotoxicity of tumour-specific cytotoxic T lymphocytes, and adoptive transfer of cytotoxic T lymphocytes and regulatory T cells pretreated with Pam3CSK4 from wild-type mice into tumour-bearing severe combined immunodeficiency mice can restore antitumour immunity by reciprocal downregulation of regulatory T cells and upregulation of cytotoxic T lymphocyte function^[1]. The principle is that transiently reducing regulatory function can release effector function, but the same manoeuvre risks autoimmunity.

In vaccinology, the reprogramming of regulatory T cells is exploited rather than reversed. The reprogramming of regulatory T cells has been recognised to play a critical role in the initiation of certain innate immune responses by vaccination with a Toll-like receptor agonist adjuvant, that is, CpG oligodeoxynucleotide^[1]. When IDO was blocked, CpG treatment stimulated plasmacytoid dendritic cells to express IL-6, which reprogrammed Foxp3 lineage regulatory T cells to express IL-17 and become Th17-like effector T cells, and the converted regulatory T cells played a helper role essential for initial priming of CD8+ T cells to a new cross-presented antigen, in a CD40 ligand-dependent manner^[1]. This process, unlike help from conventional non-regulatory CD4+ cells, did not require preactivation or prior exposure to antigen^[1]. Adjuvant design therefore has to consider not only dendritic cells but also the regulatory compartment.

In autoimmunity and transplantation, the aim is usually to strengthen regulation or to block the pathways that overcome it. Cytotoxic T lymphocyte-associated antigen 4 and programmed cell death-1 maintain peripheral tolerance, and interrupting co-stimulation through these receptors can produce anergy^[1]. Intrinsic IL-12 is capable of converting Foxp3-positive regulatory T cells into interferon-gamma-positive Th1, IL-17-positive Th17, or transitional cells, and the transitional Foxp3-positive interferon-gamma-positive cells further differentiate into interferon-gamma-positive Th1 cells but not Foxp3-positive regulatory T cells, although they still retain regulatory functions at that stage^[1]. This has practical implications for cytokine therapy: delivering IL-12 to a tumour or its draining lymph node may overcome immunosuppression, but the same cytokine can destabilise regulatory populations elsewhere.

Metabolic targeting is an emerging strategy. In the Th17-polarised transfer model of experimental autoimmune encephalomyelitis, adding 2-deoxyglucose during in vitro expansion of draining lymph node cells resulted in reduced IL-17-positive cells and a simultaneous increase of Foxp3-positive regulatory T cells, and when these cells were transferred into recipients, the 2-deoxyglucose-treated cells exhibited markedly reduced ability to cause disease, with diminished leukocyte infiltration and spinal cord inflammation^[1]. Blocking glycolysis alters Th17 and regulatory T cell differentiation and protects mice from experimental autoimmune encephalomyelitis^[1]. The broader principle, that the HIF1-alpha-dependent glycolytic pathway serves as a metabolic checkpoint distinguishing these two closely related lineages, suggests that metabolic drugs could be used to tilt the balance between effector and regulatory fates^[1].

Finally, the interaction between antigen-presenting cells and lymphocytes is the target of checkpoint blockade. Because cytotoxic T lymphocyte-associated antigen 4 binds CD80 and CD86 and blocks CD28 co-stimulation, and because programmed cell death-1 is crucial for maintenance of peripheral tolerance, antibodies against these molecules release effector T cells from inhibition^[1]. The cost is predictable: releasing the brakes can produce immune-related adverse events that resemble the very autoimmune syndromes that regulatory T cells exist to prevent. Understanding the macrophage-lymphocyte dialogue, the three-signal model and the biology of FOXP3-positive regulatory T cells is therefore not academic; it is the basis for predicting both the benefit and the harm of modern immunotherapy.