Dendritic Cells in the Immune System—History, Lineages, Tissues, Tolerance, and Immunity

All living things, both plants and animals, must defend themselves against infection. The types of infectious agent that would infect and inhabit larger eukaryotic animals are the viruses; the prokaryotic bacteria; and the eukaryotic groups of fungi, single-celled protozoa, and multicellular helminths. To infect, they must cross the defensive barriers provided by host epithelia, including the external surfaces of the skin and the internal (but topologically external) mucosal surfaces. These preexisting barriers to infection have their own forms of constitutive defense, provided, for example, by antiviral and antimicrobial proteins such as defensins and cathelicidins. Having gained access to the internal milieu of the host, in order to survive and replicate, some infectious agents invade intracellular niches (typically viruses, intracellular bacteria, and protozoa), while others adopt an extracellular habit (commonly pyogenic bacteria and other protozoa). Other organisms survive and replicate either within or outside of the epithelial barriers, as exemplified by fungi colonizing the skin or helminths cohabiting within the mucosal tracts of the intestine. Hence, to survive, any multicellular animal must be able to eliminate or at least control pathogens from any class of infectious agent. In the case of humans, these range in size over 8 or more orders of magnitude from the smallest viruses to longest tapeworms, can be located both within and outside of host cells, and may have enormously complex and transformative life cycles (the malarial parasite, for example).

All plants can defend themselves against infection, although they do not possess an immune system as such. The innate immune system emerged with the dawning of the animal kingdom and particularly with the evolution of the first multicellular organisms. Innate immunity provides the earliest molecular and cellular, and often system-wide, mechanisms of defense that are rapidly triggered by infection. For example, phagocytes such as macrophages can capture and internalize many bacteria, which are then degraded. The soluble complement system can also rapidly be activated to coat microbes and help macrophages do this. Fundamentally, innate responses have several crucial functions. First, they may rapidly and efficiently eliminate infectious agents. In general, immunological mechanisms that effect, or bring about, the elimination or control of infection are termed effector mechanisms. Second, early innate responses typically cause local inflammation. This enables yet more molecular and cellular effectors to be recruited from the blood into local sites of infection, further amplifying their efficacy. Third, innate responses can lead to later system-wide modulations of organ function. These result, for example, in fever and “sickness behavior” through actions on the brain, and the so-called acute-phase response produced largely by the liver. In the latter case, extraordinarily high quantities of new molecules are pumped into the bloodstream. Many of these seem to be opsonins, which coat infectious agents, bind to complementary receptors on cells such as macrophages, and promote uptake and clearance. Finally, innate responses are absolutely essential for the subsequent initiation and regulation of adaptive immune responses (see below).

So-called pattern recognition receptors (PRRs) are essential for innate immunity. These cellular receptors enable different types of infectious agents to be sensed with a high degree of accuracy. This is because PRRs can recognize highly conserved molecular motifs on which infectious agents depend for their integrity, replication, and survival. These components, which cannot be synthesized by the host, are generally known as pathogen-associated molecular patterns (PAMPs). The PRRs also facilitate precise discrimination between classes of infectious agents, for example, by recognition of lipopolysaccharide as a component of Gram-negative bacterial cell walls, absent in Gram-positive species. Similar molecules also exist in soluble form and are sometimes termed pattern recognition molecules. They include molecules such as C1q and mannose-binding lectin of the complement system. These can bind directly to bacterial surfaces or to opsonic molecules attached to them, and trigger the so-called classical and lectin pathways, which both result in complement activation. Arguably, PRRs enable the innate immune system to discriminate perfectly between infectious nonself and healthy self.

Some PRRs, such as scavenger receptors and members of the C-type lectin receptor family, can promote the uptake and internalization of infectious agents through receptor-mediated endocytosis or phagocytosis, by macrophages, for example. Other PRRs, however, trigger rapid secretion of proinflammatory cytokines and/or type I interferons (IFN-I), which contribute to rapid initiation of inflammatory and antiviral responses, respectively. These PRRs include the Toll-like receptors (TLRs), associated with the plasma membrane and endocytic compartments of cells such as macrophages, and the RIG-I (retinoic acid-inducible gene I)-like and NOD (nucleotide oligomerization domain)-like receptors in the cytosol. PRRs are also known to sense cell or tissue damage and death, whether or not it results from infection. This is through their capacity to recognize so-called damage-associated molecular patterns (DAMPs). One example is extracellular ATP, which in fact acts as a DAMP in animals as well as plants. Hence, through such mechanisms, innate immunity can rapidly sense and discriminate between different types of “danger” that may be presented to the host and trigger the responses needed for defense and restoration of homeostasis.

The rapid initiation of innate responses critically depends on the different types of cells that are present at sites of infection, all of which express PRRs. These include epithelial cells of tissues such as skin and the mucosae (the gut, lung and airways, and urogenital tract). Under physiological conditions, most tissues also contain resident populations of macrophages and mast cells and, in many, recently described innate lymphoid cells (ILCs). Collectively, it may be convenient to think of these as “alarm cells” that sense danger and rapidly sound “alarm signals” that are provided by molecular alarmins. The latter include the proinflammatory cytokines and lipid mediators that induce local inflammation. Subsequently, large numbers of reinforcements are recruited from the blood to help fight the danger and repair the damage. These include the granulocytes—neutrophils, or basophils and eosinophils—and later monocytes, as well as complement components and “natural” antibodies.

The abundant cellular expression of multiple types of different PRRs that are “hard-wired” to induce rapid cellular responses, as well as soluble forms that can be quickly activated, ensures that innate responses to infection are extremely fast, generally large, and often (though not always) highly effective. DCs are also present in tissues. They too can detect infection, tissue damage, and cell death through their own expression of PRRs and by sensing of alarmins, and also contribute to innate responses. However, they also have pivotal functions in the adaptive immune system, in the relatively few species that have one.

The vast majority of animal species on the planet survive and continue to evolve with the innate immune system as their only means of defense against infection. And yet, around the time of the Cambrian explosion some 540 million years ago, the adaptive immune system evolved within a remarkably short period of geological time (5). This provides immunity that is based on two main types of lymphocytes, T cells and antibody-producing B cells. Each has its own specialized type of receptor that enables it to recognize components of infectious agents or antigens: T-cell receptors (TCRs), and B-cell receptors (BCRs) which can be secreted as antibodies. This new type of immunity, as recognized in humans, probably first evolved in jawed fish. Through the millennia, with the emergence of all other vertebrates, adaptive immunity has coopted the mechanisms of innate immunity, with which it coexists, for its own purposes. Recent evidence indicates that an alternative form of adaptive immunity originated even earlier in jawless fish (agnathans), such as lampreys and hagfish, but is apparently an evolutionary dead end. In this review, the term “adaptive immunity” will be used only for the type that exists in jawed vertebrates (gnathostomes).

Precisely why the adaptive immune system “suddenly” appeared remains an almost complete mystery. However, according to the 2R hypothesis, it may have been facilitated by two entire genome replications before or around the time of the Cambrian explosion; even a third round may have occurred in some bony fish. If so, it would have enabled the different classes of antigen receptors (BCRs and TCRs) to evolve on paralogous chromosomes from an ancestral gene locus. This form of immunity probably also depended on the incorporation into the genome of a transposon, perhaps from a virus, that encodes the recombinase-activating gene (RAG). (There are in fact two RAG genes in gnathostomes, but only one in agnathans.) RAG enables a remarkable genetic mechanism known as somatic recombination to operate in developing T cells and B cells (but in no other cell type). This generates the extremely diverse sets of TCRs and BCRs. These lymphocyte antigen receptors are clonally distributed such that, at any one time, all those expressed by any T cell or B cell are essentially the same. They are in fact encoded by germ line gene segments that can be joined by somatic recombination to form functional genes. Together with additional mechanisms of diversification, a quite astonishing total number of receptors could be generated from extremely small portions of the genome that would otherwise require billions or trillions more DNA.

T cells and B cells, and their respective receptors, have very different but complementary functions in adaptive immunity. BCRs recognize small portions (epitopes) of extracellular molecules that typically retain their native, three-dimensional conformations. Hence, antibodies can bind directly to antigens such as viruses and bacteria before they infect cells. In contrast, so-called conventional T cells express αβ TCRs. In general, αβ TCRs can recognize intracellular antigens such as viruses and bacteria that have infected or been internalized by host cells. In actual fact, they recognize peptides derived from antigens inside cells that have been bound to classical major histocompatibility complex (MHC) molecules, which enable their transport from intracellular locations up to the cell surface. Thus, αβ TCRs recognize peptide-MHC complexes but not free peptides alone, and are often said to be MHC restricted. Classical MHC molecules are expressed by host cells so that they can be continuously surveyed by T cells and recognized if they have become infected by infectious agents or otherwise contain them. Conventionally, host cells are said to present their antigens to T cells and are hence termed antigen-presenting cells (APCs); the process as a whole is known as antigen presentation. In this review, the terms “conventional” and “classical” (see above) will normally be omitted, but they are used here since, for example, some “unconventional” T cells express γδ TCRs instead and some of these can recognize other, “nonclassical” MHC molecules.

Overall, antigen recognition by lymphocytes is anticipatory in nature, since vast repertoires of antigen receptors are generated in any individual even before any infection has occurred. It is also highly discriminatory, since even the slightest molecular change in an epitope may abolish recognition. It is, however, highly promiscuous, in that any given receptor has the potential to cross-react with perhaps millions of different epitopes from unrelated antigens that just happen to have the same shape (6). (The same is actually also true of MHC molecules binding peptides.) Functionally, what this means is that there is an extremely high chance that the adaptive immune system will be able to recognize an antigen from any given infectious agent it may encounter in the future, but the frequency of lymphocytes that can recognize any given epitope is extremely low. Hence, in any adaptive immune response, only these extremely rare clones of lymphocytes that recognize antigen can participate. Originating from a hypothesis put forward decades ago, this is known as clonal selection (by antigen). These few clones first need to be activated, since naive (antigen-inexperienced) lymphocytes are resting and impotent. They also need to be greatly expanded in numbers through multiple rounds of proliferation (or clonal expansion) before they can reach a critical mass to function effectively. In addition, lymphocytes are in fact highly plastic cells that can generate different types of effector responses (see below) but need to be “instructed” as to which type they should produce, as well as where they should go. All this takes time. For these reasons, the first, primary adaptive response produced against any infectious agent appears to be extremely slow. As measured, for example, by the frequency of antigen-specific T cells and antibodies, it may take days to reach a peak, in contrast to innate responses measured in minutes or hours. These responses are, however, much faster if the same infection occurs again, because the adaptive immune system can generate an immunological memory of what it has previously encountered (see below). DCs have essential roles in all stages of adaptive immune responses, and we now know they also enable these responses to be initiated much more quickly than it might seem.

The original description (7) of a Steinman-Cohn DC was of a morphologically distinct, trace cell type that was noted among adherent cells cultured from mouse SLOs, such as spleen, lymph nodes, and Peyer’s patches of the intestine. These cells, as originally defined, continually extended and contracted spiny cell extensions in culture that were termed “dendrites.” They did not phagocytose particulate tracers, in contrast to the majority population of relatively sessile macrophages, which were avid phagocytes. Subsequently, DCs were discovered to be the most potent stimulators of T cell and T-cell-mediated immune responses, compared to other cell types tested (9–13). They seemed to have a unique capacity to initiate T-cell activation, thus acting as accessory cells. Moreover, through elaboration of cytokines such as interleukin-12 (IL-12), it seemed likely that they could also regulate the subsequent type of immunity that was induced (13). Hence, the original definition of a DC was a distinct cell type that could be isolated from SLOs and which was nonphagocytic but highly immunostimulatory.

Soon after their description, it was suggested that isolated DCs might be counterparts of the interdigitating cells. These had previously been identified within sections of lymphoid tissues, and particularly in the T-cell-rich zones of spleen and lymph nodes (14–16). It was only later realized that these cells were mostly liberated from lymphoid tissues after enzymatic digestion, for example, with collagenase. Hence, interdigitating cells appeared to be relatively fixed within the T-cell areas, in contrast to the apparently more free DCs first studied. These likely represented the nests of DCs that were later noted in more peripheral regions of tissues such as the marginal zones of spleen (17).

The development of monoclonal antibodies (MAbs) was published just 2 years after the initial description of DCs, and also rewarded with a Nobel Prize (18). This was crucial for further exploration of the phenotype and function of DCs in culture, and of their possible relationship to other cell types. It was soon appreciated that both isolated DCs and interdigitating cells expressed MHC-II molecules, and they were later found to express costimulatory molecules, thus helping to explain their capacity to initiate T-cell responses. Newly developed MAbs also showed that both cell types also expressed CD11c (part of the complement receptor type 4). However, they could be distinguished phenotypically since the majority of DCs that were readily liberated expressed 33D1 (the first MAb generated against DCs, later shown to recognize DCIR2 [DC inhibitory receptor 2]), whereas interdigitating cells expressed CD205 (a C-type lectin receptor) (17, 19).

The more general term “lymphoid DC” was often adopted for presumptive DCs isolated from or localized within SLOs. These also included cells within the medulla of the thymus, which, after isolation, were also found to be nonphagocytic but to have immunostimulatory function in culture (20, 21). However, early evidence suggested that these thymic DCs might be lymphoid in origin, deriving from the common lymphoid progenitor (CLP) rather than common myeloid progenitor (CMP) downstream of the hematopoietic stem cell. Matters were further complicated by the later finding that some DCs expressed the lymphoid markers CD4 (a subset of DCIR2⁺ cells) or CD8 (generally interdigitating cells). Hence, the term “lymphoid” was used rather uneasily to reflect either or both a lymphoid tissue and/or a hematopoietic lymphoid origin.

Attention then turned to presumptive DCs in NLTs (this abbreviation also including “nonlymphoid organs”). The first detailed studies were of Langerhans cells (LCs) isolated from the epidermis of mouse skin (22, 23). These cells have a dendritic morphology within the tissue and form a dense network of cells in the suprabasal region of the epidermis. They also contain a characteristic organelle known as the Birbeck granule. After isolation from the skin, LCs were found to be weak or inactive as stimulators of T-cell responses in vitro. They were also shown to be phagocytic (24). During culture, however, the cells acquired potent immunostimulatory activity accompanied by profound phenotypic and functional changes, including increased expression of MHC-II molecules but loss of Birbeck granules and a markedly reduced capacity for phagocytosis. They thus came to resemble Steinman-Cohn DCs.

Working around the same time, other investigators had been studying a trace cell type that expressed MHC-II molecules in NLTs such as heart and kidney (25, 26). These cells, which became known as interstitial DCs, were identified in the connective tissue of all NLTs studied, except for the central nervous system. The impetus for such studies was the suggestion, made as early as 1957, that transplant rejection was initiated by a small number of “passenger leukocytes” that were carried over in the graft, while the bulk of the transplanted material was in fact nonimmunogenic. Consistent with this idea, when interstitial cells were isolated from vascularized organs such as heart and kidney, it was found that their phenotypes and functions changed during culture similarly to what had been observed for LCs (27).

The term “myeloid DC” was introduced by some workers to discriminate between the presumptive DCs that were isolated from or identified within NLTs and the lymphoid DCs and interdigitating cells noted above (28). The use of this term probably also originated from observations that DCs in NLTs, as well as myeloid cells such as monocytes and macrophages, expressed Fc receptors and CD11b. These cells did not, however, express CD4 or CD8, and later evidence indicated that they probably originated from myeloid progenitors downstream of the CMP. Unfortunately, the focus on shared myeloid markers to characterize DC populations has bedeviled the field and caused further confusion that, to some extent, has still not been fully resolved.

Because of the developments above, the concept of a DC needed to encompass both a phagocytic, nonimmunostimulatory cell in NLTs as well as the Steinman-Cohn DC isolated from lymphoid tissues. Moreover, it appeared likely that one could develop into the other. This transition was termed “maturation,” since cells in NLTs were considered to be relatively immature compared to mature cells in lymphoid tissues that had immunostimulatory activity. The concept of a DC also needed to be revised to encompass a migratory stage, as the cells apparently moved from peripheral tissues into SLOs, for example, via afferent lymph. In larger animals, it is possible to cannulate afferent lymphatics and to study migrating cells directly, but this is extremely difficult in smaller animals such as rodents. However, by surgically removing regional lymph nodes and allowing time for the lymphatics to reanastomose, one can obtain cells that would otherwise travel in afferent lymph after cannulation of the thoracic duct; this source is termed “pseudoafferent lymph” (29).

In culture, cells obtained from lymph or pseudoafferent lymph were found to have characteristic lamellipodiae, or “veils,” and became known as veiled cells (16). These cells were found to be traveling in lymph derived from tissues such as gut, liver, and skin in a continuous manner (30, 31). In the latter case they resembled LCs since they contained Birbeck granules, and similar cells were also detected in regional lymph nodes (32, 33). (An early pioneer, Brigitte Balfour, who was related to a former British prime minister, cannulated the lymphatics of her own leg in order to obtain veiled cells, and made vast numbers of cinematographic recordings of their movements in culture that now fill a back room somewhere in London [S. C. Knight, personal communication]). Veiled cells were also found to stimulate T-cell responses in culture (30, 34). The concept that veiled cells represent an intermediate stage of DC trafficking from NLTs into SLOs was reinforced by other studies. For example, studies of skin explants and transplants enabled the direct visualization of presumptive LCs leaving the epidermis and migrating through dermal lymphatics out of the skin (35, 36). Moreover, while the numbers of such cells in lymph were generally quite low, the flux of these cells increased massively after infection or inflammation in the site from which they originated. This suggested that they may have important roles in generating immune responses within SLOs.

DCs had also been identified in human blood, although it was not clear if these might be en route to or migrating from NLTs (37). In early studies, the capacity of DCs to migrate via the blood into the spleen, where they homed to areas rich in T cells, was demonstrated after adoptive transfer of labeled DCs in mice (38, 39). Moreover, DCs were observed to disappear from cardiac tissue after mouse heart transplantation, and donor cells were then detected in recipient spleen, where they were visualized often in close association with recipient T cells (40). Stimuli promoting DC maturation and migration also started to be identified. These included proinflammatory cytokines such as IL-1 and tumor necrosis factor α (TNF-α) (41, 42) and microbial products such as bacterial lipopolysaccharide, which also induced potent immunostimulatory activity in vitro (43). Hence, the general concept evolved that, in vivo, infection and inflammation in NLTs stimulated the maturation and migration of DCs into SLOs to stimulate TD immunity.

DCs were clearly shown to have a hematopoietic origin and to originate from bone marrow of adult animals. It was discovered that cells resembling DCs could also be generated in culture from mouse and rat bone marrow progenitors, particularly after supplementation of cultures with granulocyte-macrophage colony-stimulating factor (GM-CSF; CSF-2) (44, 45). There were perhaps two main drivers for these studies. First, as a trace cell type, the isolation of “primary” DCs from tissues was laborious, time-consuming, and often costly (not just expensive) in terms of the large number of animals that were needed for even a single experiment. Second, the appreciation that DCs might be central to the initiation and regulation of T-cell-mediated immunity almost immediately suggested that they may have enormous therapeutic potential. It was subsequently shown that DCs could be generated from macaque and human bone marrow CD34⁺ progenitors (46–48). At around the same time it was also discovered that cells resembling DCs could also be generated from human monocytes in culture (49, 50), although their differentiation into veiled cells in culture had in fact been reported some years earlier (51).

The discovery that it was possible to generate monocyte-derived DCs (moDCs) suggested that, in vivo, monocytes are bipotential cells that can differentiate into either macrophages or DCs with immunostimulatory activity. It was therefore proposed that monocytes might be recruited to peripheral sites of inflammation and infection, where they differentiated into moDCs before migrating to SLOs to stimulate T-cell-mediated immunity. Later work further reinforced this view, particularly after injection of particulates into the skin or the blood (see “Lineages”). But, from a therapeutic standpoint, these findings were critical. At last it became possible to initiate human clinical trials in which cancer patients received infusions of their own “ex vivo-generated” DCs that had been pulsed with tumor antigens and matured in one form or another. Thousands of patients have now been treated in a large number of clinical trials, which continue with varying degrees of success (52–54). (In some centers, perhaps a quarter of patients with end-stage [IV] melanoma who came off trial but continued to be treated with their DCs on an ad hoc basis have lived for many years afterwards, compared sadly to survival rates more in terms of months [C. Figdor and G. Schuler, personal communication].)

As early as the 1950s, pathologists had noted an unusual cell type within SLOs, particularly in disease settings, which was originally known as a plasmacytoid monocyte or plasmacytoid T cell. In 1997, it was discovered that if these cells were isolated from inflamed human tonsils and cultured with IL-3 and CD40 ligand (normally expressed by activated T cells in vivo), they developed immunostimulatory activity (55). They also appeared to be nonphagocytic and thus, functionally, came to resemble Steinman-Cohn DCs. In many other respects, however, they were very different: they appeared to circulate primarily in the blood (but not in lymph), had not been identified in many other tissues under physiological conditions, had a very different phenotype, were later shown to have the specialized capacity to secrete extremely high levels of IFN-I during viral infections, and additionally showed evidence of a possible lymphoid origin. Whether or not these cells should be considered as DCs is still debated by some investigators.

Many early studies focused particularly on the immunostimulatory properties of DCs. Parallel developments in the field, however, clearly demonstrated that at least some, if not all, of these cells might also contribute to tolerance. For example, increasing evidence showed that interdigitating DCs of the thymus were probably involved in the induction of central tolerance during T-cell development (despite their immunostimulatory function in vitro). Moreover, much later studies showed that veiled cells migrated constitutively from the gut in a relatively immature state. Some of them contained apoptotic cell contents, presumably derived from the tissue (56). It was therefore proposed that these cells might contribute to peripheral tolerance in regional lymphoid tissues. This paradigm was subsequently extended to DCs that migrated constitutively from all other NLTs in which they had been identified. Hence, DCs in different types of tissue appeared to have different cellular functions (e.g., phagocytosis versus immunostimulation); could contribute to tolerance and/or immunity depending on circumstances; and while some were clearly related (e.g., those migrating from one site to another), others (e.g., plasmacytoid DCs [pDCs]) seemed very different.

Another cell with dendritic morphology had been described within SLOs in 1965 (57, 58). Unlike interdigitating cells, however, these cells were identified within the germinal centers that develop from B-cell follicles during the course of adaptive immune responses. They were originally known as “antigen-retaining reticular cells.” However, they were termed “follicular dendritic cells” by Steinman in 1978 (59), since at that time he believed they might be related to the Steinman-Cohn DCs he first described. Despite this early confusion, it is now absolutely clear that the follicular DC is a completely distinct cell type, in terms of both its stromal cell origin and functions. These cells retain native antigens on their cell surface and are involved in selection of B cells during the so-called germinal center reaction (60). Hence, in this review, the term “dendritic cell” specifically excludes follicular DCs.

By the turn of the century, at least eight types of DCs or dendritic-like cells had been identified: LCs and interstitial DCs within peripheral NLTs (commonly termed myeloid DCs); afferent lymph veiled cells and blood DCs in the circulation; Steinman-Cohn DCs and interdigitating cells within the lymphoid tissues (sometimes termed lymphoid DCs); moDCs in peripheral and possibly lymphoid tissues; and pDCs. Nevertheless, there appeared to be a commonality between them in that each cell type was increasingly shown to contribute to both T-cell tolerance and T-cell-mediated immune responses. Even so, into the fifth decade after Steinman’s first publication, the field of DC immunobiology appears to have become very complicated and often confused. It seems that an ever increasing number of “new DCs” are being described, particularly from phenotypic studies of cells from different tissues, some of which appear to have varying immunological functions depending on the experimental system that is tested. There is therefore understandable debate about if and how it might ever be possible to define a DC, and continuing controversy over the relationship between these and other cell types, and particularly macrophages, that seems well beyond semantics. This review focuses specifically on just a few types of DCs that seem quite distinctive, though others will be mentioned. This is to try to distinguish clearly between “the wood and the trees.” Whether or not it leads only to a clearing in a forest remains to be seen; it may be the edge of the wood.

The two types of classical DCs are crucially important for initiation of primary T-cell responses and make a substantial contribution to the regulation of adaptive immunity. These two subsets are as distinct from each other as are T cells and B cells (81). Classical DCs have been identified in most tissues studied closely, and all contain both types. However, these cells are absent from the parenchyma of the brain and the eye, though they do populate sites such as meninges and outer cornea. They are also absent from stratified squamous epithelia such as the epidermis of skin, which contains LCs instead. Both types of classical DCs have been identified in all warm-blooded species studied, including the mouse, rat, sheep, cow, pig, and human; the subsets in mouse and human seem extremely similar to each other at a transcriptional level (82, 83). Cells resembling classical DCs have also been identified in fish and amphibians (84–86), though further study is required to determine whether or not these are classical DCs and, if so, whether there are two subsets. Nevertheless, the evolutionary conservation of two distinct subsets of classical DCs in warm-blooded species at least suggests that they play fundamentally different roles in the immune system.

In mice, pre-DCs migrate from the bone marrow via the blood into the tissues, where, locally, they differentiate into mature cells. Both the pre-DC and its presumptive upstream CDP and MDP progenitors express the CX3CR1 chemokine (fractalkine) receptor, but expression is extinguished as they differentiate into the mature cells (87). Classical DCs express the transcription factor Zbtb46 (zDC; Btbd4), which helps to discriminate between these cells and moDCs, pDCs, and LCs (64). However, this is not a stable or definitive marker, as it is downregulated on DC maturation, can be induced in monocytes, and is or can be expressed in other cell types. The homeostatic proliferation of pre-DCs and/or the development of classical DCs also depend on the Flt3L growth factor and its receptor, CD135. In general, classical DCs have a short half-life of just a few days in tissues, and this is tightly regulated: experimentally decreasing their life span has detrimental effects on the induction of immunity, while increasing it can result in autoimmune manifestations (88).

In this review, for simplicity, the two types of classical DCs will be designated as DCI and DCII subsets, and these are intended as neutral terms. The pre-DC generates two resident populations in all SLOs, and probably the thymus. These will be termed cDCI and cDCII cells, an abbreviation (cDC) that is otherwise used for classical or conventional DCs. The pre-DC also generates two populations in NLTs, which, however, are only transiently resident because they can subsequently migrate to SLOs. To distinguish these cells from the resident populations in SLOs, they will be termed migratory mDCI and mDCII cells, an abbreviation (mDC) that is otherwise used for myeloid DCs. Although detailed studies have not been performed in all cases, the respective populations of cDCI and mDCI, and cDCII and mDCII, appear to be extremely similar in many respects. However, their phenotypes do vary depending on their tissue of localization (see below), presumably reflecting tissue-specific specializations (89).

In general, classical DCs in NLTs express relatively low levels of MHC-II molecules compared to some of their counterparts in SLOs. A common view is that DCs in NLTs can endocytose or phagocytose antigens and generate intracellular peptide–MHC-II complexes. (They may also retain intracellular antigenic epitopes for considerable periods of time, even after antigen has been eliminated [90].) In response to stimuli such as TLR agonists and proinflammatory mediators, they decrease further uptake, begin to express high levels of peptide–MHC-II molecules at the cell surface, and increase their expression of costimulatory molecules such as CD86. They also express CD40, which enables cross talk between DCs and activated T cells via CD40L. Together, these regulate complex, bidirectional cascades of costimulatory (and conversely, coinhibitory) molecular interactions. The related phenotypic and functional changes are commonly referred to as DC maturation. The cells then migrate from the tissue into SLOs (91). Here, due to their new and increased expression of CCR7, they home to T-cell areas where the corresponding chemokines (CCL19 and CCL21) are produced (92, 93). Here these migratory DCs can activate antigen-specific T cells and trigger subsequent adaptive responses. An extreme view might hold that just one population of classical DCs could do the whole job itself. And yet there are two distinct subsets of migratory DCs in NLTs, and it is unlikely that the two resident subsets in SLOs are mere spectators. This strongly implies a marked division of labor between the two main types of classical DCs (DCI and DCII), and very close cooperation between all four in adaptive immune responses as a whole.

pDCs seem specialized to induce antiviral resistance and may regulate CD4⁺ T-cell responses and perhaps contribute to immunological memory of CD8⁺ T cells. Despite the original demonstration that pDCs can develop into immunostimulatory cells (55), subsequent studies have found that this capacity may be relatively limited compared to that of classical DCs. There seems to be little or no persuasive evidence that pDCs initiate T-cell responses in vivo. Unlike differentiation of classical DCs, the development of pDCs is dependent on the transcription factor E2-2 although it also requires Flt3L (65, 114, 115). While pDCs may normally be generated from the CMP under homeostatic conditions, they can also be generated from the CLP. Interestingly, and irrespective of myeloid or lymphoid origin, human pDCs express pre-Tα, mouse pDCs have rearranged IgH DJ segments, and both express RAG (116). These respective molecules are normally essential for formation of the pre-TCR and expression of the pre-BCR during intermediate stages of T- and B-cell differentiation, and for somatic recombination of both TCRs and BCRs. It has been proposed that this reflects the evolutionary origins of pDCs (116), a point that is further considered below.

In general, pDCs circulate in the blood and populate SLOs, which they enter via high endothelial venules (HEVs), and may also home to the thymus (see “Tolerance”). Small numbers of pDCs also populate mucosal tissues, within the lamina propria of the gut and lung and airways, for example, but appear to be excluded from other peripheral tissues under homeostatic conditions. Hence, migration of pDCs to the mucosal tissues could represent homing to lymphoid tissue within the gut wall, for example, or more generally migration in response to homeostatic inflammation (see “Tissues”). Such homing is dependent on expression of CCR9, although entry to SLOs is not. However, in pathological settings, pDCs can accumulate in large numbers in sites such as inflamed skin and within melanomas and ovarian and mammary carcinomas. Their roles in these sites are unclear.

Perhaps the best-known function of pDCs is the expression of exceptionally high levels of IFN-I in response to viral infections, and particularly of IFN-α. Both human and mouse pDCs selectively express TLR7 and TLR9, which enable them to sense genomic single-stranded RNA and DNA in endosomes. In fact, they may also sense host nucleic acids liberated during tissue damage and, in turn, facilitate wound healing through ancient (innate) mechanisms (117, 118). Nevertheless, IFN-I secretion is likely to be tightly regulated through a number of inhibitory receptors, such as Siglec-H (sialic acid-binding immunoglobulin-type lectin H) in mouse as well as CD303 (Clec4c; BDCA-2) and immunoglobulin-like transcript 7 (ILT7) in human, all of which regulate TLR responses (95, 119). Mouse pDCs also express tetherin (BST-2 [bone marrow stromal antigen 2]), which can physically bind live viruses to infected cells. In response to viral infection, the cells undergo autophagy, which may route endogenous viral antigens into the peptide–MHC-II pathway.

Through their capacity to secrete high levels of IFN-I, pDCs may play central roles in promoting antiviral resistance as well as regulating innate and adaptive responses. In the latter case, for example, at a molecular level, IFN-I can modulate functions of classical DCs by promoting extended antigen sampling of endosomal compartments for peptide–MHC-II loading (120); enhance cross-presentation and peptide–MHC-I loading; and regulate IL-12 expression, either to promote Th1 polarization or conversely to facilitate activation of CTLs (and innate-like NK cells). At the cellular level, IFN-I stimulates the differentiation of classical DCs from progenitors (121) and increases the maturation and migration of classical DCs from NLTs into SLOs. Furthermore, IFN-I decreases the generation of Tregs; increases the expression by CD8⁺ T cells of perforin and granzymes, which are both important for CTL killing; and promotes the survival of memory CD8⁺ T cells. Nevertheless, some of these proposed functions are inferences from the described roles of IFN-I in other settings, and, particularly since IFN-I is a large family of different classes, further study is required to confirm or refute a direct role of pDCs in some of them. Overall, however, pDCs may play regulatory roles in adaptive immunity either directly or by modulating the functions of other DC populations (122). This apparent synergy has, for example, been shown to be crucial for clearance of viral infections (123).

moDCs may play particularly important roles in the effector phase of T-cell-mediated responses, and perhaps particularly those of CD4⁺ T cells. It now seems clear that monocytes are bipotential cells that can home to the tissues and differentiate into moDCs or macrophages, though monocytes, as such, have also been identified within some tissues (124). Other populations of probable monocyte-derived cells, described as “Tip-DCs” and “LysoDCs,” have also been reported (below, and see “Tissues”). Discrimination between moDCs, classical DCs (and particularly DCII cells), as well as monocyte-derived and other macrophages has proven highly problematic. In large part, this is because of the shared expression of molecules that can be further modulated during differentiation. For example, most or all express CD11c, and moDCs and mDCII cells both express MHC-II and CD11b. Nevertheless, a particularly useful marker of monocyte-derived macrophages in tissues is CX3CR1, since its expression increases to high levels during the differentiation of these cells from monocytes; it is also expressed by resident macrophage progenitors (125). In contrast, expression is lost during differentiation of pre-DCs into classical DCs. However, CX3CR1 is probably expressed at intermediate levels during development of moDCs in tissues, so this is not definitive. This molecule is the fractalkine receptor, which is presumed to facilitate cellular interactions with epithelial cells in tissues, though its precise functions remain unclear (126). In the mouse, many macrophage populations, including monocyte-derived populations as well as LCs, but not classical DCs, express the F4/80 molecule (127, 128). Importantly, CD64 (FcγR) also seems to be selectively expressed by moDCs but not classical DCs, thus helping to distinguish between them (108, 129).

Different monocyte subsets have been identified in the blood. In humans, those that express high levels of CD14, with or without CD16, are the likely equivalents of those that express high levels of Ly6C (Gr-1) in the mouse. Most evidence indicates that these cells can be recruited to inflammatory sites, where they have the potential to develop into moDCs or macrophages. Hence, this population has been termed inflammatory monocytes. In contrast, in physiological conditions, another human subset with high-level expression of CD16 but low CD14, which is the likely equivalent of those expressing low levels of Ly6C in mouse, are associated with the luminal surfaces of endothelial cells. Hence, they have been termed patrolling monocytes. It seems probable that both types of monocytes can be recruited to tissues under different circumstances and, depending on which and the particular microenvironment in which they find themselves, may “choose” alternative fates. Hence, they may develop into either moDCs or macrophages (130, 131). However, these tissue populations probably represent a continuum of different phenotypes and functions ranging from moDCs, LysoDCs, and Tip-DCs to macrophages (which are further subdivided by some into M1 or M2 types), and quite a lot more in between (132). Nevertheless, moDCs and macrophages, as commonly described, are probably very different in function.

There is much evidence that monocytes and/or monocyte-derived cells such as moDCs play important roles in the resolution of infections by bacteria and fungi; examples include infection with Citrobacter in the intestine and Aspergillus in the lungs (133–135). They have also been associated with the polarization of Th1 responses due at least in part to their secretion of IL-12, but also with Th2 responses in other settings such as allergic asthma (136). In addition, moDCs seem able to elicit “recall” responses by activating memory T cells, though their relative importance in this respect, compared to classical DCs, is still not clear (137). In fact, both moDCs and classical DCs most likely play complementary and synergistic roles during adaptive immune responses in general (132) (see “Immunity”). In addition, populations of Tip-DCs have been identified within the spleen during infections with intracellular pathogens such as Listeria monocytogenes or Trypanosoma cruzi. These cells can secrete TNF-α and express inducible nitric oxide synthase, which are required for synthesis of reactive nitrogen intermediates, but also stimulate primary T-cell responses in culture. They are most likely monocyte-derived cells but functionally appear to represent a transitional stage between moDCs and macrophages. This fate choice appears to depend in part on which TLR agonists might be sensed (138).

A particularly potent stimulus for the differentiation of monocytes into moDCs may be phagocytosis of particles at endothelial cell surfaces. For example, in a transmigration system in vitro, CD14⁺ and/or CD16⁺ monocytes were able to traverse endothelial cell monolayers into an artificial matrix where they internalized particles, after which they crossed back and developed into moDCs (139, 140). Furthermore, injection of particulates into mouse skin was observed to recruit monocytes into the site, which subsequently migrated into the regional lymph nodes and developed into particle-laden moDCs in the T-cell-rich paracortical regions (141). Earlier work had also demonstrated that injection of particulates into the bloodstream of rats led to their uptake by cells within the marginating pool in the sinusoids of the liver, where Kupffer cells reside (142–144). Subsequently, particle-laden cells were detected within the hepatic lymph, presumably en route to the regional lymph nodes. These findings raise the possibility that Kupffer cells initially phagocytosed the particles and perhaps then elaborated chemokines to recruit monocytes from the bloodstream, which, after phagocytosis, began to develop into moDCs. The cells that were subsequently detected in lymph nodes indeed had little phagocytic activity and could stimulate primary T-cell responses in culture, but additionally expressed CD103. Whether or not such expression was due to their microenvironment is not clear. However, taken together with the fact that the two main subsets of monocytes can migrate into tissues (above), it is possible that each could separately generate a distinct population of moDCs. While there seems to be no evidence to support this as yet, potential parallels with the classical DCI and DCII subsets should be obvious and worthy of closer scrutiny.

At least three different roles for moDCs in immunity may be envisaged. First, at some stages of differentiation, they may acquire effector functions, which contribute to the elimination of infectious agents in peripheral tissues, as seems likely for cells such as Tip-DCs. Second, moDCs may acquire peripheral antigens and migrate into SLOs to initiate primary T-cell responses. Evidence has been presented that, while the entry of monocytes into the tissues is CCR2 dependent, subsequent migration of moDCs into SLOs depends on expression of CCR7 and CCR8 (145). In this respect they could be viewed as a rapidly recruited source of immunostimulatory cells that assist the induction of adaptive immune responses, perhaps after classical DCs have migrated from the tissue and before repopulation from the pre-DC. Nevertheless, at present there seems to be little available evidence to support a major role of moDCs in the activation of primary T-cell responses. Finally, there is good evidence that moDCs may instead play essential roles in the effector phases of T-cell-mediated immune responses in NLTs (see “Immunity”).

Epidermal LCs were first identified in 1868, but even today their roles in immunological responses have not been clearly elucidated. The dermis of skin, like most other tissues, contains the two subsets of classical DCs. In contrast, the epidermis contains only LCs. Related cells have been identified in all other stratified squamous epithelia, such as those of the oropharynx, vagina, ectocervix, and anus. LCs are situated in a suprabasal position deep within the epidermis, where they express MHC-II and CD11c and, in the mouse, F4/80. Epidermal LCs and dermal DCI cells also express CD207 (Langerin). This is involved in the development of Birbeck granules, which are specialized endosomal compartments of unclear function that have, however, only been identified in LCs (see “Tissues”). Unlike the development of other DC subsets, that of mouse LCs requires transcription factors such as ID2 and RUNX3 and is also dependent on transforming growth factor β (TGF-β) (which can be used to generate LC-like cells from progenitors in culture [70]). In addition, LC development is dependent on the CSF-1 receptor (M-CSFR; CD115). However, LCs persist in the skin of CD115-deficient mice, most probably because they require IL-34, which is an alternative ligand for this receptor to which it binds with a higher affinity than CSF-1.

LCs may have important functions related to monitoring the integrity of the epidermal barrier and contributing to repair. For example, in mice, skin oils apparently diffusing through the epidermis after damage to the outermost stratum corneum may be presented by LCs to a subset of specialized, innate-like NKT cells (146). These in turn may elaborate cytokines such as IL-22 to induce keratinocyte proliferation, and hence effect repair. Though their cell bodies are situated deep within the epidermis, LCs can extend cytoplasmic processes upwards between the keratinocytes and appear able to sample antigens beneath the stratum corneum (147). There seems to be little available evidence supporting the idea that LCs play any major role in initiating T-cell responses. In contrast, they may induce or maintain tolerance within the skin, even in the presence of TLR agonists (148). In part this seems due to their capacity to maintain populations of neonatal Tregs that home to the skin after birth (149) (see “Tolerance”). LCs also appear to contribute to polarized CD4⁺ T-cell responses of different types (Th17, Th1, and/or Th2) against antigens that may breach the uppermost layers of the skin or penetrate the epidermis. Other evidence tends to suggest that their contributions to CD8⁺ T-cell responses may be indirect (150).

It is, however, important to note that there are very distinct differences between the skin of different species of jawed vertebrates, which all possess epidermal LCs. For example, obviously the skin of mammals has an outer keratinized layer that is coated with skin oils and generally surrounded by air (except, of course, for aquatic mammals such as whales). In fish and amphibian tadpoles, however, the skin is a mucosal tissue surrounded by an aqueous environment, similar to that of the mouse or human embryo developing in the amniotic sac. There are even major differences between mouse and human skin. Both contain T cells in the epidermis, but those in mouse are mostly unconventional γδ T cells (dendritic epidermal T cells), whereas most in human are conventional αβ memory T cells. Possibly, therefore, the epidermal LCs of different species may have evolved some different functions, and there may be immunological differences between LCs of mouse and human.

Unlike all other DC lineages discussed here, the epidermal LCs of skin first arise very early during embryonic development. In the mouse, primitive myelopoiesis begins in the yolk sac at around embryonic day 7 (E7) and primitive myelopoietic (nonmonocytic) progenitors seed the epidermis at around E8.5. Here they generate LCs, which form a self-renewing population (151, 152); they also generate the resident populations of macrophages in the brain and liver, the microglial and Kupffer cells (153). A second wave of LCs then populates the epidermis, which develop from fetal liver-derived monocytes at E16.5 after the commencement of definitive hematopoiesis. (The same progenitors may also generate a second population of Kupffer cells in the liver, populate the lamina propria of the lung with cells that develop into alveolar macrophages, and probably give rise to resident macrophages of many other tissues [154].) The fetal-derived epidermal LCs then undergo rapid proliferation after birth and come to represent the predominant, self-renewing population in adult skin under homeostatic conditions (151). Under inflammatory conditions, at least, epidermal LCs can be further supplemented by bone marrow-derived monocytes that home to the skin epidermis and develop into apparently indistinguishable cells. By contrast, corresponding populations in the oral mucosa appear to originate from pre-DCs, with a contribution from monocyte-derived cells, though they too express a very similar transcriptomic signature to those in skin (155).

While the LCs of adult mouse and human skin have been associated with tolerogenic and immune functions (see above), the functions of the yolk sac-derived LCs within the embryo are unknown. These cells populate the epidermis even before the endothelium develops at around E9 and later differentiates into lymphatic and venous components at E11.5; hence, these LCs are present in the skin before the blood and lymph circulatory systems have developed. Potentially they may have specialized homeostatic functions within this specialized “mucosal-like” environment (cf. fish and amphibian tadpoles above). Curiously, however, the establishment of this population coincides with development of the umbilical link between fetus and mother. The only immunological cells of relevance at this stage would seem to be maternally transferred lymphocytes. In principle at least, maternal T cells could conceivably gain access to fetal skin perhaps in response to the endogenous chemokine CXCL12 (stromal cell-derived factor 1 [SDF-1]). This plays important roles in the movement of embryonic cells during morphogenesis and is a ligand for the CXCR4 receptor on mature T cells. Fetal engraftment by maternal T cells has in fact been demonstrated in some cases of SCID, for example (156, 157). Such maternally transferred T cells could be detrimental to the fetus (through recognition of paternal antigens) or beneficial if they were Tregs. Hence, it is an open question whether embryonic epidermal LCs might be involved in the induction or maintenance of tolerance, through local interactions or even after migration from fetus to mother (158, 159).

Numerous publications have postulated the existence of other presumptive DC subsets in addition to the five types discussed here. Perhaps the most persuasive (or at least suggestive) evidence is for a subdivision of mouse DCII cells in SLOs into two further subsets. This is based on their apparent dependence on the transcription factors Notch2 or klf4 for development (79). These putative subsets have also been associated with Th17 and Th2 responses, respectively. Moreover, there is undoubted phenotypic heterogeneity of DCII cells within mouse SLOs based on differential expression of molecules such as Clec4a4 (CDIR2; 33D1; see above), CD4, and ESAM. As another example, subsets of “double-negative” DCs that do not express “selected” DCI or DCII markers have been reported in various tissues. These include the dermis of skin (XCR1^–CD11b^–) (160), spleen (CD4^–CD8^–) (161), Peyer’s patches (CD8^–CD11b^–) (162, 163), and mesenteric lymph nodes (MLNs) (CD103^–CD11b^–) (164). These double-negative cells have variably been reported to be IRF4 dependent, to secrete IL-12, and/or to express CD207 (Langerin) or CX3CR1. In addition, a double-positive population has been reported in gut (CD103⁺CX3CR1⁺) (164), and there are undoubtedly other examples. However, the expression of different transcription factors can be induced or extinguished because of intrinsic developmental programs or extrinsic microenvironmental influences, and can also be modulated during cellular responses. It is therefore possible that at least some of the additional DC subsets that have been proposed, such as those noted above, might represent early stages during the development of progenitors, such as pre-DCs, or of monocytes adopting different cell fates within tissues (compare Tip-DCs and LysoDCs above). Given the general preoccupation with three main types of polarized CD4⁺ T-cell responses (Th17, Th1, and Th2), it is perhaps not surprising that some might expect the existence of three dedicated subsets of classical DCs (see “Immunity”). There are almost certainly two subsets of classical DCs; there may be three; many more seem like fashion accessories.

The term “steady state” is often used to describe conditions in tissues under physiological conditions in the absence of overt infection. However, the presence of commensal organisms on external stratified epithelia such as the epidermis of skin, and at luminal surfaces within mucosal tracts such as that of the gut, is continuously sensed by the immune system. Indeed, the microbiota also shape adaptive immunity, and a state of mutualism is established and maintained between them and their hosts (165–169). The microbiota induce a state of “homeostatic inflammation,” also termed “tonic stimulation,” which is physiological and tightly regulated and controlled (170). However, depending on the relative loads of commensals, the levels of homeostatic inflammation may vary, perhaps being highest in gut and lowest in organs and tissues such as heart and skeletal muscle, for example. This should be borne in mind when considering the relative complexity of DCs and other populations in different tissues. It is also a general, though not entirely accurate, rule that naive T cells are excluded from NLTs. However, populations of effector and/or memory T cells may or may not constitutively traffic into different tissues, which adds a further layer of complexity to the physiological conditions within each. Hence, in this review the term “homeostatic conditions” is often used for those in normal tissues in general, while the term “inflammation” relates to conditions that prevail after infection or tissue damage.

Lymph from large regions of the gastrointestinal (GI) tract probably drains to the MLNs, though some areas drain into other regional lymph nodes (see below). Throughout the small and large intestine there are also distinct types of gut-associated lymphoid tissues (GALTs) that play specialized roles in tolerance and immunity (204). Descending from the duodenum, the jejunum and ileum contain variable numbers of Peyer’s patches, which, in humans, reach a peak in the third decade of life but then decline in number (205). These are embedded within the mucosal propria and submucosal tissue. They have specialized, sieve-like M cells within a follicle-associated epithelium, juxtaposed between the intestinal epithelial cells. Descending further into the cecum is the appendix, which may have specialized functions yet to be discovered. There are also increasing numbers of “solitary isolated lymphoid tissues,” such as isolated lymphoid follicles, descending from the ileum to the colon. These most probably represent tertiary lymphoid structures. They have much less well-defined structures than SLOs, lacking defined follicles and T-cell zones, for example, and are not developmentally controlled but are locally induced by a subset of ILCs (206). DCs populate the lamina propria and submucosal tissue of the gut wall and GALTs.

Distinct subsets of DCs also populate the lamina propria of all other mucosal tissues and their respective MALTs. Here, the focus will be on the lung and airways and their associated MALTs, though the urogenital tract will be mentioned briefly.

Early studies demonstrated that DCs could be isolated from the blood, where they appeared to circulate in both relatively immature and mature forms (37). It was believed that the former represented DCs en route to the tissues, while the latter represented those migrating from these sites. However, this interpretation needs to be revisited in the light of our current understanding that classical DCs are generated locally in tissues from the pre-DCs.

Some common themes emerge. The increasing number of reports of antigen-sampling cells in NLTs is provocative (see above). Epidermal LCs in skin sample antigens from under the stratum corneum impregnated with skin oils and surrounded by air (or water). Within mucosal tissues, CX3CR1⁺ “macrophages” of the gut sample the lumen under the mucus, while perhaps related cells of the lung sample the airspaces under mucus or airway lubricants. Perhaps they exist in the urogenital tract. Within nonmucosal and fully vascularized tissues, similar cells in contact with blood appear to do likewise. Perhaps these are also present in other tissues, such as kidney, where they may sample filtrates. It might be speculated that the function of these macrophage-like populations may be to monitor “danger.” If none is detected, they may regulate the populations of classical DCs that populate the dermis of skin, lamina propria of the mucosal tissues, and connective tissue of other organs and tissues. Hence, antigen-sampling LCs and macrophages may control the homeostatic maturation of classical DCs and/or induce regulatory functions that promote tolerance when the latter migrate into SLOs.

If danger is sensed by the antigen-sampling macrophage-like cells, they may recruit classical DCs, such as DCI cells, to sample its source for themselves. These then transport the antigen into SLOs to initiate protective immunity. Conceivably, the CDII cells are coopted if the danger spreads into the tissues. Hence, innocuous antigens such as from food or aerosols may generally result in tolerance, or perhaps “ignorance.” Commensals may be contained by barrier defenses and homeostatic “background immunity,” which includes T-independent production of IgA (these responses generally lack memory, to enable the microbiota to evolve). However, pathogens lead to the initiation of protective immunity, which may include TD IgA responses against commensals that invade the lamina propria. At the tissue and organ level, firewalls also exist. The MLNs and the liver, respectively, provide firewall protection for the lymph and blood derived from the gut. Likewise, the pLNs and the spleen may play comparable roles for the lymph and blood derived from all other peripheral tissues and organs. Presumably, related mechanisms may also be involved in protection of the brain, though much of this tissue seems very different (see below).

It is generally accepted that DCs are absent from the parenchyma of the brain. Some CD11c⁺ populations have been identified in isolated areas such as the pituitary (259), although these could well represent monocyte-derived cells rather than classical DCs. Moreover, there is no evidence that the abundant microglial cells that are present throughout are any more than relatively sessile resident macrophages that remain within the tissues. However, cells that closely resemble classical DCs have been identified in sites such as the meninges and the choroid plexus in contact with the cerebrospinal fluid. Early studies demonstrated that small tracers could be transported from the brain into the deep cervical lymph nodes, and were believed to do so via drainage through the cribriform plate. However, the more recent finding of the existence of a specialized dural lymphatic system (sometimes termed “glymphatics”) that drains to these lymph nodes may provide an alternative explanation for these observations (260, 261). It may also provide a potential route by which DCs may migrate to SLOs, potentially to induce immunity against pathogens in the central nervous system, though this has yet to be studied in depth.

There are very close links between the central nervous system and the immune system. For example, it is quite clear that inflammation in peripheral tissues can be regulated via the hypothalamic-pituitary-adrenal axis (262). Furthermore, gut immunity can be regulated by the enteric nervous system and is controlled in part by the vagus nerve (263). Even the development of SLOs may be initiated by nerves that infiltrate tissues (222). Hence, the nervous system may regulate DC responses at least indirectly. However, both nociceptors (264) and neurotransmitters can directly influence DCs (265). Perhaps more remarkable, however, is the recent report that DCs (and macrophages) may be directly innervated in lymph nodes “by a mesh of filamentous neurofilament positive structures originating from single nerve fibers and covering each single APC similar to a glass fishing float” (266). There are also other highly provocative findings suggesting, for example, that nerve endings in spleen can form synapses with a subset of memory T cells that produce acetylcholine (262). Hence, investigation of the bipotential links between the central nervous system and the immune system, and of potential roles of DCs in them, may be a most fruitful field of study for the future.

The thymus is an encapsulated and lobulated organ. Within the lobules, the tissue is broadly divided into an outer cortex and inner medulla by the corticomedullary junction, which is rich in blood vessels and afferent lymphatics. A conduit system has also been identified, akin to that described in SLOs (273). The conduits appear to originate at blood vessels, traverse the medulla, and may terminate at Hassall’s corpuscles, which develop from medullary thymic epithelial cells (mTECs) after they lose Aire expression (see below); alternatively, they may originate at the latter and extend to the blood vessels. The early thymic precursors migrate to the thymus from the blood via the corticomedullary junction and home to the outer cortex. As these progenitors develop into mature T cells, they are believed to travel through the cortex into the medulla. Finally, mature T cells leave the thymus, enter the blood, and begin their lymphocyte recirculation between different SLOs.

The thymic cortex and medulla contain dense networks of highly specialized epithelial cells that nurture thymocyte development and play largely distinct roles in thymocyte selection. The cortical thymic epithelial cells (cTECs) contain a distinct form of the proteasome termed the thymoproteasome, which generates peptides for MHC-I loading. They also express specialized lysosomal proteases, cathepsin L and thymus-specific serine protease (TSSP), which generate peptides for MHC-II loading (274). Hence, cTECs express a unique spectrum of self peptide–MHC-I and peptide–MHC-II complexes that are not likely to be generated in other tissues. These positively select CD8⁺ and CD4⁺ thymocytes, respectively. Mice lacking thymoproteasomes or the specialized lysosomal proteases have deficiencies of CD8⁺ and CD4⁺ T cells, respectively.

In contrast, mTECs have the remarkable capacity to express an enormously diverse spectrum of otherwise tissue-specific antigens (TSAs) that are normally expressed only in peripheral (extrathymic) tissues. Such ectopic expression within the thymus is regulated, in whole or in part, by a transcriptional regulator termed the autoimmune regulator (Aire) and the transcription factor Fezf2, which induce the expression of distinct subsets of TSAs (275–282). Consequently, mTECs can generate tissue-specific peptide–MHC-I complexes that contribute to negative selection of CD8⁺ thymocytes. In addition, mTECs may undergo macroautophagy and generate tissue-specific peptide–MHC-II complexes that may contribute to negative selection of CD4⁺ thymocytes. Humans with mutations in Aire suffer from the devastating autoimmune disease autoimmune polyendocrinopathy-candidiasis ectodermal dystrophy (APECED), and different hypomorphic mutations may underlie many other autoimmune manifestations. Likewise, mice deficient in Aire manifest with autoimmune disease, although the condition is generally less severe than that in humans. In contrast, mice deficient in FezF2 also present with autoimmune disease, but typically affecting a different spectrum of tissues. Hence, Aire and Fezf2 control the expression of distinct subsets of TSAs.

Peripheral tissues contain different populations of Tregs (269, 297). The centrally generated tTregs home to these sites from the thymus and are supplemented by pTregs that are generated within the periphery; conversely, it seems likely that pTregs, as well as small numbers of naive T cells (298), may home back to the thymus, for reasons largely unknown. While tTregs develop from thymocytes during their differentiation, pTregs can develop from mature, naive T cells that have these regulatory functions “imposed” upon them. In both settings, the master transcription factor FoxP3 regulates their development and functions (299). Mice that lack expression of FoxP3 develop lymphoproliferative autoimmune disease but can be rescued if normal expression is restored. Rare humans lacking FoxP3 expression develop IPEX (immune dysregulation, polyendocrinopathy, enteropathy, X-linked). Typically, they suffer from severe gut and skin inflammation, as well as autoimmune-mediated damage to other tissues and organs such as the thyroid. (It is interesting that many autoimmune conditions resulting from defects in central tolerance involve the endocrine tissues; see below.)

The thymus is critically important particularly for generating tTregs around the perinatal period, which migrate to peripheral tissues. Those that were induced by thymic DCs may, in turn, be selected, expanded, and maintained by peripheral DCs that express TSAs (300). Neonatally thymectomized but not adult mice, as well as rare humans with thymic hypoplasia (DiGeorge syndrome), suffer from severe autoimmune disease caused by the early, or aberrant, populations of T cells they respectively develop. A major stimulus for recruitment of tTregs into some tissues is likely the early colonization with commensals. For example, a wave of Tregs is recruited into neonatal skin of mice in response to establishment of its microbiota (301). Peripheral tolerance in the intestine and lung is associated with an influx of tTregs, perhaps also in response to their distinct and diverse populations of commensals (302). Hence, it appears that tTregs may be particularly important in maintaining tolerance to the microbiota at such sites. However, other evidence indicates that the majority of pTregs in the intestine are generated against food antigens (303). Nevertheless, from TCR repertoire analysis it is quite clear that the populations of Tregs vary enormously between different tissues (thousands of different specificities have been detected in peripheral sites, compared to just hundreds in SLOs) (297). The largely Aire-dependent, and hence TSA-specific, population of tTregs might control the induction of autoimmune T-cell responses against TSAs that might be liberated, for example, during tissue damage caused by invading commensals. But if, as seems likely, classical DCs are crucial for T-cell activation, then this would need to be regulated at the level of DCs. Alternatively, or in addition, a hypothetical population of DCA-specific tTregs (above) might also regulate the responses of DCs that captured and expressed commensal or other antigens. (This is not, however, intended to imply that Tregs only act on DCs and may not also regulate the functions of other cell types.)

In contrast to skin and gut (see above), tTregs may not be recruited to tissues lacking commensals, such as the exocrine pancreas, to which conventional T cells may normally be “ignorant” (302). If so, this may pertain to other nonmucosal or fully vascularized organs and tissues, such as heart and skeletal muscle. Perhaps this necessitates additional mechanisms for the maintenance of tolerance under other circumstances, such as might be provided by pTregs in such sites. It is currently believed that the continuous, homeostatic maturation and migration of DCs from peripheral tissues into SLOs plays a central role in generating pTregs (see below). In addition, there is accumulating evidence that expression of FoxP3 and regulatory function can be induced or imposed on populations of effector and memory T cells (304, 305).

Aire can also be expressed in the periphery and may control expression of TSAs in a cell context-dependent manner, different from those generated by mTECs. So-called extrathymic Aire-expressing cells have been identified in mouse SLOs (277, 306). These are bone marrow-derived cells but seem to be distinct from DCs, although Aire expression has also been reported in splenic marginal-zone DCs (probably the cDCII subset) (307), as well as other cells such as CD14⁺ monocytes (308). There is evidence that extrathymic Aire-expressing cells can directly inactivate, or perhaps anergize, CD4⁺ T cells even under inflammatory conditions in the lymph node (rather than inducing deletional tolerance or Tregs) (277). Another transcription factor, deformed epidermal autoregulatory factor 1 (Deaf-1), also controls further ectopic TSA expression in the periphery (309). Deaf-1 is expressed by lymph node stromal cells (LNSCs) and/or FRCs and can also be expressed in the pancreas (310). Hence, it appears that Aire and Deaf-1 may regulate ectopic expression of distinct subsets of TSAs in the periphery (i.e., within both SLOs and at least some NLTs). Moreover, FRCs in lymph nodes may themselves delete autoreactive CD8⁺ T cells that recognize the TSAs they express (309). Whether or not such antigens can also be acquired by peripheral DCs, perhaps for generation of pTregs, awaits further investigation. However, a converse pathway has also been demonstrated, in which LNSCs acquire peptide-MHC complexes from DCs and subsequently delete CD4⁺ T cells (311). Whether they also strip other membrane molecules from DCs (or “cross-dress” in them [312]) remains to be seen, but could explain observations that LNSCs may stimulate CD4⁺ T-cell responses (313).

Any tissue monitors itself, or is monitored, to sense infection and damage. The PRRs of epithelial cells in mucosal tissues are generally localized on the tissue-facing, abluminal surfaces. Hence, they preferentially recognize invaders of the lamina propria rather than commensals contained within the lumen. In addition to specialized populations of macrophages, which may play homeostatic rather than primary defense roles, most NLTs contain resident cells such as mast cells within the connective tissue or mucosae. Tissues such as the dermis of skin, the lamina propria of the gut and lung, and the liver additionally harbor different populations of ILCs (327–329). They are also present in SLOs such as the spleen, MLNs, and tonsils (330). ILCs are lymphoid cells in origin, deriving from the CLP, and they closely resemble T cells but produce innate-like responses and lack expression of TCRs. They include three different types that are classified as ILC1, ILC2, and ILC3 cells. In response to infection or tissue damage, these secrete polarized patterns of cytokines very similar to those produced by Th1, Th2, and Th17 cells, respectively. They also express the corresponding master transcription factors T-bet, GATA-3, and ROR-γ, respectively. Innate-like NK cells have also been assigned to the general class of ILCs and closely resemble CTLs in their cellular cytotoxicity and cytokine secretion profiles. ILCs may represent very ancient lineages that evolved before the emergence of conventional T cells and presumably classical MHC molecules. Their existence reinforces the common view that the immune system focuses on sensing general classes of infectious agents and generating characteristic immunological outputs to deal with each.

Within peripheral tissues, DCs coexist with diverse types of tissue-specific and resident cells and may be bathed in complement components (etc.) produced by resident macrophages; the latter indeed may also regulate DC functions (see “Tissues”). Collectively, however, these may constitute local immunological networks that monitor their environment for infection and damage, immediately mount and amplify innate effector responses, and rapidly trigger inflammation to recruit new effectors. The composition of the DC populations itself also varies between tissues: classical DCs, LCs, and/or pDCs under homeostatic conditions in some, with pDCs and/or moDCs during and after inflammation in these or others (the same will be true for regulatory, effector, and/or memory T cells). Any or all DCs may contribute to innate responses, but it may be the classical DCs that are also essential for initiating subsequent adaptive responses. First, each classical DC may sense alarmins, PAMPs, and/or DAMPs within its own tissue microenvironment. This type of information from many varied inputs is then likely to be integrated, and in turn, each cell may generate a finely tuned molecular and cellular response as an output. Second, any DC may internalize antigens (through pinocytosis, receptor-mediated endocytosis, phagocytosis, and/or macropinocytosis) and generate representative peptide-MHC complexes from them. Third, DCs then migrate from NLTs into SLOs, where they can select antigen-specific T cells, deliver costimulatory signals to activate them, and modulate the functions of activated T cells at least in part according to the information they received in their peripheral microenvironment.

Primary T-cell responses are initiated within SLOs. There is abundant evidence that each subset of classical DCs has specialized capacities to activate naive, antigen-specific T cells (and hence TD responses), and this is essential for generating protective responses (see “Lineages” and “Tissues”). Other DC subsets can express MHC-II and costimulatory molecules, as may other lineages, such as B cells. It is, however, important to stress that expression of these molecules is also essential for the induction of Tregs (see “Tolerance”) and may regulate the differentiation of effector and/or memory T cells. Moreover, the exuberant expression of MHC or costimulatory molecules that can be driven by various stimuli in vitro does not necessarily translate to the in vivo setting, where it may be more tightly regulated and controlled. The evidence that DCs can initiate primary T-cell responses in vivo seems strongest for classical DCs, perhaps less so for moDCs, and least convincing for LCs and pDCs.

Some of the complex choreographies that are executed by T cells and DCs within different microenvironments of SLOs during priming have been noted (see “Tissues”). It is quite clear that both activated CD4⁺ and CD8⁺ T cells can make further sequential interactions with other cognate DCs in the tissue. During these, they presumably start to express their effector programs, which may be serially modulated by the different sets of information they receive from successive DCs (for example, depending on the quantity and quality of costimulatory and/or coinhibitory signals each can deliver). The frequency of these interactions is further enhanced by the increased dwell times of activated T cells within the SLOs compared to their naive counterparts, which are controlled in part by IFN-I (such as might be secreted by pDCs) (344). T cells proliferate at an exceptionally fast rate compared to other eukaryotic cells, each cell cycle being completed within just 6 to 8 h. This leads to a rapid expansion of the antigen-specific clones, each of which may interact with migratory and/or resident DCs. In turn, the effector programs of each T cell may be further modulated depending on the particular microenvironments that migratory DCs experienced in the periphery and/or that resident DCs experience within the SLOs. Potentially, the responses of any DC itself may be modulated by small molecules delivered via conduits or mediators produced by monocytes (181), moDCs, or “swarms” of neutrophils (345) that enter SLOs (see “Tissues”). This may drive further diversification of T-cell responses, producing even more heterogeneity in early effector-T-cell populations.

Some T cells execute their effector functions within the SLOs. These include CD4⁺ T cells that activate CD8⁺ T cells and potentially drive the generation of CTLs locally or later within NLTs. They also include CD4⁺ T cells that migrate to the borders of T-cell zones and the follicles where they activate antigen-specific B cells. Some CD4⁺ T cells may alternatively differentiate into follicular helper T cells, which are involved in the germinal center reaction; precisely where this latter fate is decided is not entirely clear, though it may be controlled at least in part by moDCs (346). Other CD4⁺ effector T cells may instead develop into memory T cells (see below). However, some CD4⁺ T cells migrate from the SLOs and execute their effector functions instead within NLTs (see below). There is good evidence that homing of effector T cells to different NLTs can be dependent on where their antigen originated, and this is controlled at least in part by DCs; hence, effector T cells may selectively migrate to distinct sites such as the skin, gut, or lung (347–349).

Memory T cells can persist for the lifetime of any mouse or human, though they decline in numbers with time. While much still remains to be discovered, it seems clear that CD4⁺ memory T cells can persist in the apparent absence of the antigen to which they were induced, and that CD4⁺ T cells are also essential for the generation and maintenance of CD8⁺ memory T cells. However, the latter are better understood primarily because of the relative ease of studying their function in in vitro assays. Nevertheless, based on their phenotype and function, CD4⁺ memory T cells have been divided into central memory (Tcm) and effector memory (Tem) subsets, while others that appear to reside in peripheral tissues are termed resident memory (Trm) cells (355). The former, Tcm cells, recirculate between SLOs and can rapidly reexpress molecules such as CD40L on restimulation; within SLOs they may be positioned most closely to potential sites of pathogen entry (356). In contrast, Tem cells tend to populate peripheral tissues and can rapidly reacquire effector cell functions such as polarized patterns of cytokine secretion. It has been suggested that Tcm cells can develop into Tem cells after restimulation.

There is still much to be learned about the factors that regulate the generation of CD4⁺ and CD8⁺ memory T cells. However, pDCs have been implicated particularly in this process. As noted earlier, these cells may play specialized roles in viral infections and perhaps particularly in the regulation of CD8⁺ T-cell and CTL responses (see “Lineages”). As was also noted, pDCs can secrete high levels of IFN-I. Importantly, at the tissue level, it is known that IFN-I can prevent the egress of lymphocytes from SLOs (344). This latter example may be most relevant to the phenomenon of lymph node shutdown, which is observed soon after infection and can be induced in regional lymph nodes after injection of IFN-I into the skin (G. G. Macpherson, personal communication). The markedly increased cellularity of SLOs during this period may increase the frequency and duration of interaction between T cells and their cognate DCs, and possibly enhance the diversification of effector cells. Because CD8⁺ T cells generally have a higher activation threshold than CD4⁺ T cells (200), it is possible that this could also apply to their development into memory T cells. Hence, in principle, the capacity of pDCs to contribute to lymph node shutdown and extend intercellular communications could directly or indirectly contribute to robust CD8 memory-T-cell generation. Whether or not there might be segregation between moDCs and pDCs for generation of CD4⁺ and CD8⁺ memory T cells remains to be seen. The relative contributions of the different subsets to reactivation of memory cells during secondary responses is unclear. However, there is good evidence at least that classical DCs can elicit such recall responses, presumably in SLOs and also in bone marrow and NLTs (104, 357, 358).

From an evolutionary standpoint, and if ontogeny indeed recapitulates phylogeny, it is tempting to speculate about the origins of DCs. The embryonic, yolk sac-derived population of LCs may be the most ancient of all. All animals have hematopoietic tissues that can generate cells, such as the wandering amoebocytes observed by Metchnikoff in starfish larvae or, later in evolution, blood-circulating cells such as monocytes (5). It is possible to envisage that cells such as amoebocytes might first have evolved into resident macrophages within the tissues, including LCs in the epidermis of skin. Potentially the earliest LCs, generated by a form of primitive myelopoiesis well before lymphoid cells evolved, may have played homeostatic roles in the skin of ancient fish, but might have used phagocytosis in defense. From extant species we know that LCs are present in all jawed fish that have been studied, but this does not seem known for agnathans. During the earliest stages of adaptive evolution in jawed fish, LCs may then have evolved into essential immunostimulatory cells for immune responses within skin-associated lymphoid tissues (359) (unless they could migrate to the spleen). The later LC populations successively derived from fetal liver- and bone marrow-derived monocytes, as observed in mice, may conceivably represent examples of evolutionary convergence. Certainly, as a population, all LCs seem homogeneous in skin at a transcriptional level (including the oral mucosal LCs that are generated from pre-DCs; see “Tissues”).

Blood-circulating monocytes may have evolved before adaptive immunity. Monocytic cells can even be found in agnathans (360). They were presumably early inflammatory cells since their capacity to circulate meant that they could be rapidly recruited to sites of infection and damage. Here, their greater numbers might have helped to reinforce the defense and repair functions of the more ancient and sessile (amoebocyte-derived) macrophages and LCs. With the evolution of adaptive immunity, their functions may have also diverged to generate populations of moDCs as immunostimulatory cells. Again, their circulatory capacity may have enabled them now to migrate from the tissues into the newly evolved spleen. Hence, these may have represented early migratory DCs. In the natterjack toad (Bufo calamita), for example, migratory monocytes in the red pulp of spleen appear to differentiate into giant, dendritic-like cells after immunization (86).

Potentially, the phagocytic activity of early monocyte-derived cells may have enabled them to contribute particularly in early immunity against bacteria, which perhaps have the greatest diversity of PAMPs for any eukaryotic host. However, most viruses are too small to be phagocytosed and may have fewer PAMPs, in part because they can envelop themselves within host cell-derived membranes. This may have provided the evolutionary pressure for the emergence of pDCs. These cells can be produced from both myeloid and lymphoid lineages and have certain features of both. In particular, they contain RAG and partial TCR or BCR transcripts (116). This could suggest that pDCs evolved during the diversification of the myeloid and lymphoid lineages. It would therefore be interesting to investigate whether pDCs are also present in fish and amphibians, which this hypothesis would predict, in addition to warm-blooded animals such as rodents, pigs, and humans. The earliest functions of pDCs may have been for inducing antiviral resistance. However, these cells also had the capacity to circulate within the blood and therefore could travel to the spleen. Perhaps therefore they acquired specialized functions to facilitate CD8⁺ T-cell responses against viruses. In contrast, the moDCs were presumably more specialized for CD4⁺ T-cell responses against bacteria, and most likely protozoa and fungal spores since these can be readily phagocytosed. Potentially, therefore, moDCs and pDCs could have represented the earliest DC-like cells with diversified but synergistic functions that drove early CD4⁺ and CD8⁺ T-cell responses.

Classical DCs can apparently also be generated from both myeloid and lymphoid lineages in vivo. Those in lymphoid tissues can also express either CD4 or CD8, which are otherwise generally associated with conventional or specialized αβ T-cell populations. However, they do not contain RAG or the abortive transcripts of pDCs noted above. It is therefore possible that they evolved even later than pDCs, as is also suggested by the apparently late appearance of the pre-DCs in simple linear models of differentiation (see “Lineages”). These two subsets of CDI and CDII cells might therefore be viewed as being “superimposed” upon the pDC and moDC populations that already existed, with LCs perhaps confined to local defense of the skin. Instead of focusing on particular types of infectious agents, the classical DC may have evolved primarily to sense PAMPs and DAMPs derived from intracellular and extracellular sources in general. This may have enabled a fine-tuning of protective T-cell responses against the respective infectious agents noted above. These DCs may also have facilitated more-robust responses to be generated against infectious agents such as helminths, which are generally difficult to eliminate and often cohabit with humans. Helminths are eukaryotes, and some can mask themselves in host proteins. Hence, they may express many fewer PAMPs than other types of infectious agents but can cause barrier damage that generates DAMPs. Sensing of these by classical DCs may have enhanced the sensing of “danger” and the capacity to induce T-cell responses in general. Interdigitating cells resembling classical DCs have also been identified in the splenic white pulp of different fish, including sharks (85), trout (84), and zebrafish (361). It is not, however, clear if these are classical DCs or monocyte-derived cells. Transcriptional comparisons with those of more recently evolved species may therefore provide insights into whether one subset of classical pre-DCs evolved before another, or whether they arose even later in the evolution of jawed vertebrates.

There are many ways to build an immune system (362–364). All jawed animals have skin and gut (and fish have gills, too), so it is important to endow them with local forms of defense at the very least (359, 365). The essential specialized tissues for adaptive immunity seem to be a thymus and spleen, but thereafter additional primary lymphoid tissues and SLOs can evolve in different species. The essential cellular building blocks are T cells, B cells to produce antibodies, and almost certainly DCs. The critical molecular building blocks include lymphocytes with highly diversified antigen receptor repertoires and MHC molecules. However, the genetic mechanisms for diversification can differ between species, including, for example, somatic recombination in humans and mice, gene conversion in chickens, and somatic hypermutation in sharks. And at the cellular and molecular level, some startling differences are also apparent. The cod, for example, lacks CD4⁺ T cells and all components associated with the peptide–MHC-II pathway (366). This was presumably not an act of wanton vandalism, but due to evolutionary pressures that are still not understood (though a possible third round of genome duplication in some bony fish may be relevant [367]). Studies of DCs in this species may prove particularly interesting.

Two cornerstones in our understanding of adaptive immunity were laid decades ago. The first was the clonal selection theory proposed separately by Burnet and Talmage. This envisaged the existence of a vast number of clonally distributed lymphocyte receptors that could potentially recognize almost any antigen in the universe. The second, originating from the later work of Tonegawa, was the mechanistic insight into precisely how this remarkable generation of diversity could be accomplished, through rearrangement of germ line gene segments accompanied by additional mechanisms of diversification. Since then, our thinking about both processes has been dominated by the role of antigen, in terms of both the diversification of antigen receptors and the subsequent selection of lymphocyte clones by antigen in immunological responses.

There are, in fact, two well-recognized systems for receptor diversification and clonal selection in adaptive immunity that have also been studied for decades. The first was as originally conceived, and occurs within primary lymphoid tissues that have evolved in all jawed vertebrates. Through whichever genetic mechanisms are used in the species, enormous repertoires of BCRs and TCRs can be created. Functional clones of naive B cells and T cells are selected in the thymus through their capacities to recognize self-antigen, in its native form for B cells and as self peptide-MHC complexes for T cells. This generates the primary repertoires of mature T cells and B cells that can subsequently be selected by foreign antigen for protective immunity. During T-cell development in the thymus, DCs may facilitate deletion of autoreactive thymocytes. Crucially, however, they may also select and shape the precise repertoires of tTregs that are subsequently required to regulate their functions in peripheral tissues. During this process it seems likely that they select tTregs with diverse, though moderate, affinities. In turn, the affinity of TCRs may regulate the precise effector functions of these cells (272). Hence, DCs may clonally select and diversify these tTreg populations. Moreover, because it is not possible to ensure that all representations of self are available within the thymus, DCs in normal peripheral tissues generate further repertoires of pTregs that regulate their own functions (and perhaps of course those of other APCs) to ensure the maintenance of tolerance. Again, the same considerations apply. Therefore, at these levels, DCs may be responsible for both clonal selection and generation of diversity of Tregs in tolerance.

The second well-recognized system for clonal diversification and receptor selection occurs within germinal centers of SLOs and acts exclusively on B cells (368). In this case, the generation of diversity occurs through somatic hypermutation, a mechanism that exists in all jawed vertebrates, although germinal centers only evolved later in birds and mammals. During this process, random point mutations are introduced into the variable regions of BCRs in clones of B cells that were selected by foreign antigen and activated in primary responses. This process massively increases diversification of their BCRs, and these diversified clones then undergo repeated rounds of iterative selection against foreign antigen. This is retained for considerable periods of time and displayed on the follicular DC network within the germinal center. This results in the selection of B-cell clones with high-affinity BCRs (different from those they initially expressed) and from which memory B cells can be subsequently generated; all other clones are eliminated. It has been estimated that in some cases the germinal center reaction may select for survival as few as one B-cell clone out of the vast numbers that were originally generated. For B cells, antigen recognition is inseparable from their primary effector functions. These are mediated by soluble forms of their BCRs, antibodies, that are secreted after differentiation into plasma cells. Hence, the germinal center reaction effectively both diversifies and clonally selects for enhanced B-cell effector functions.

For T cells, the situation is entirely different. There is no known equivalent of the germinal center reaction for T cells and no mechanism for clonal diversification of TCRs (since T cells are not subject to somatic hypermutation in birds and mammals). Moreover, the effector functions of T cells are completely unrelated to their antigen specificities, even though TCRs directionally target these toward other cell types (for example, during formation of immunological synapses between helper T cells and B cells, or between CTLs and infected cells). Instead, the effector functions of T cells are mediated by the specialized membrane molecules and soluble mediators they can express or secrete, and which modulate the functions of other cell types, to help, suppress, or otherwise regulate them or to induce apoptosis to kill. Hence, the only way in which it possible to generate diversity of T-cell responses is to select antigen-specific clones and differentially modify their effector responses (or to recruit others into the response for the same purposes). This diversification may originate during priming itself within the T-cell zones of SLOs. Even at this stage, DCs are likely to select clones of T cells with TCRs of different affinities that directly regulate their effector functions (334). Subsequently, the activated T cells execute their complex choreographies as they serially engage with DCs in the T-cell zone, just as B cells do with follicular DCs within their separate follicular compartment (though they may dance together at the borders).

Distinct populations of DCs may be centrally involved in clonal selection and generation of diversity during the different stages of T-cell or TD responses. The classical DCs (which may themselves be highly diversified) may select heterogeneous populations of antigen-specific T cells to be activated within SLOs. Consequently, they generate diverse populations of effector T cells within them. Within NLTs, moDCs (and perhaps LCs) may select from these clones and further diversify their functions during clonal expansion. Meanwhile, pDCs may modulate the functions of classical DCs in SLOs, and possibly moDCs in NLTs where they are present, in order to help drive further diversification of effector T cells (and perhaps recruit additional clones within SLOs). These processes may continue until protective T-cell clones are generated, at which point antigen can be eliminated. Selection and diversification of effector-T-cell clones by all DCs then inevitably ceases (although potentially DCs might subsequently drive other responses that help dampen the response). However, many or all of the early effector-T-cell clones that are generated in response to infection have the potential to develop into memory-T-cell clones. Some of these become long-lived cells that can provide immunological memory for a lifetime. Hence, if the same antigen is encountered again, these memory T cells can be rapidly reactivated by classical (and perhaps other) DCs to provide almost immediate protection. Hence, DCs may drive the evolution of adaptive responses that are central to immunity and tolerance (see above).

There is continuing debate about whether it is possible to define “dendritic cells,” whether these are in fact “just another type of macrophage,” or whether they should be included in the mononuclear phagocyte system. From the viewpoints expressed in this review, the five main lineages of DCs that are discussed appear to comprise, at least in part, an integrated DC network. If more are identified, perhaps these too may be accommodated within the general framework of ideas that has been presented. The individual functions of each DC subset seem diverse and specialized, as are those of macrophage populations within different tissues. Collectively they seem to be synergistic; any DC therefore may be essential for adaptive immunity or tolerance, but none by itself is sufficient. Whether or not DCs should be included in the mononuclear phagocyte system seems mainly a semantic argument, for example, depending on whether a cutoff point is taken above or below monocytes during development or if all are simply grouped together. It matters not. All such categorizations are anthropomorphic constructs designed to help our understanding of systems whose complexity may be beyond mere human comprehension. It is with this in mind that this review ends.