Main

Dendritic cells (DCs) are generally described as following a life cycle that was modelled on the conclusions of the seminal studies carried out in the 1970s and 1980s by Steinman and colleagues on DCs purified from lymphoid organs, and on the eponymous skin cells discovered by Paul Langerhans 100 years earlier; this model is usually referred to as the 'Langerhans cell paradigm' (reviewed in Refs 1–3). According to this paradigm, DCs are present in peripheral tissues in an immature state that is specialized for sampling the environment using various endocytic mechanisms, but is characterized by low levels of expression of MHC molecules and T-cell co-stimulatory molecules.

Immature DCs are well equipped with a series of receptors for pathogen-associated molecular patterns and for secondary inflammatory compounds, such as Toll-like receptors (TLRs)4,5, nucleotide-binding oligomerization domain (NOD) proteins6, RIG-I-like receptors6, C-type lectin receptors7,8, cytokine receptors9 and chemokine receptors10. Signalling through these receptors triggers DC migration towards the secondary lymphoid organs. On reaching these organs, DCs develop into a mature state, which is characterized by high levels of expression of MHC and T-cell co-stimulatory molecules, and the ability to present antigen captured in the periphery to T cells3. According to this pathway, DCs would provide the necessary link between the probable points of pathogen entry and the lymph nodes, bringing in and presenting antigens that T cells would otherwise not be able to detect. In this context, the phenomenon of maturation was defined as the series of phenotypic changes that enabled DCs to initiate immunity. These changes were later also observed in DCs that induced tolerance rather than immunity11,12, so the term maturation is currently used to refer to these phenotypic changes rather than to any particular functional capability (reviewed in Refs 13,14) (Box 1).

Since the enunciation of the Langerhans cell paradigm as a general model of DC function, analyses of the DCs found in the thymus, spleen and lymph nodes has revealed considerable heterogeneity, such that several DC types that emerge from distinct life cycle are now well defined (reviewed in Ref. 15). Some types follow the life cycle described by the Langerhans cell paradigm, but others do not (reviewed in Ref. 13). Further, different DC types have distinct roles in the initiation of immunity to specific pathogens. In this Review, we summarize the recent studies on the life cycles and antigen-presenting functions of the subsets that constitute this DC network.

Heterogeneity of the DC network

There are two main categories of DCs: plasmacytoid DCs and conventional DCs. Many studies, especially in the human system, use the terms lymphoid and myeloid to refer to these two DC categories, but these terms are now considered obsolete and should be avoided15. Plasmacytoid DCs circulate through the blood and lymphoid tissues and only acquire the typical DC morphology after activation, which is accompanied by the release of type I interferons (IFNs). The role of plasmacytoid DCs in antigen presentation and T-cell priming is unclear, as in fact is their categorization as DCs; this is why they are sometimes referred to as 'IFN-producing cells'16. We refer the reader to other excellent reviews that have specifically dealt with plasmacytoid DCs17,18; in this Review, we focus on conventional DCs and will from here use the term DC to refer only to this type of DC.

Lymphoid-organ DC subsets in the steady state. The DCs present in the thymus, spleen and lymph nodes can be subdivided into five populations that are usually distinguished by the pattern of expression of the markers listed in Table 1. These populations can be grouped into two main categories that are distinguished by the paths they follow to access the lymphoid organs (reviewed in Ref. 13) (Fig. 1).

Table 1 Mouse dendritic-cell subsets
Full size table
Figure 1: The developmental pathway of migratory and lymphoid-organ-resident DCs.
figure 1

Migratory dendritic cells (DCs) have an immature phenotype in peripheral tissues, where they are dedicated to surveying their environment and endocytosing extracellular material. These DCs migrate through the afferent lymph to the local lymph nodes. Here they acquire a mature phenotype. The migration and maturation of these DCs occur even in germ-free animals or in mice deficient in Toll-like receptor (TLR) signalling, indicating that this process can be triggered by an inherent programme of DC differentiation or by the constant release of inflammatory compounds in the tissues. The lymph nodes also contain resident DCs. Resident DCs are also found in the spleen — the only DC type present here. Resident DCs develop within the lymphoid organs themselves, where they spend their entire lifespan in an immature state unless they encounter pathogen products or inflammatory signals, which cause their maturation in situ. Therefore, in the steady state (in the absence of overt infections), almost all of the splenic DCs, and half of the lymph-node DCs (the resident types) have an immature phenotype, whereas the migratory DCs have a mature one.

Full size image

The first category corresponds to the migratory DCs, which develop from earlier precursors in peripheral tissues and travel through the afferent lymphatics to reach the local draining lymph nodes19, where they constitute approximately 50% of all lymph-node DCs. This group of DCs is largely absent from the spleen and thymus because these organs do not receive afferent lymph. There are two subtypes of migratory DCs. The first subtype is found in all lymph nodes and corresponds to the interstitial DCs (Table 1). The interstitial DCs contained in subcutaneous lymph nodes migrate from the dermis and are often termed dermal DCs. The subcutaneous lymph nodes also contain a second population of migratory DCs, namely the Langerhans cells, which migrate from the skin epidermis. Some lymph nodes and the Peyer's patches contain another migratory DC population that is usually termed the 'triple-negative' DCs20 (Table 1). It is unclear whether the triple-negative DCs are functionally distinct from the interstitial DCs, so here both subsets are treated as a single DC group.

Migratory DCs follow the life cycle described by the Langerhans cell paradigm: they traffic from peripheral tissues to the lymph nodes, where they exhibit a mature phenotype13. The lymph nodes of mice or rats that are maintained in germ-free conditions21,22, or are deficient in both of the TLR signalling molecules MyD88 (myeloid differentiation primary-response gene 88) and TRIF (TIR-domain-containing adaptor protein inducing IFNβ; also known as TICAM1) contain a similar proportion of migratory DCs to normal mice (N.S. Wilson et al., unpublished observations). Furthermore, in these mice the migratory DCs appear mature, as in normal mice. This implies that migration and maturation of these DC types proceeds constitutively23 and independently of pathogens and TLR signalling, perhaps being triggered by inflammatory compounds that are constitutively released in low amounts by peripheral tissues. The paucity of DCs in the efferent lymph has led to the idea that DCs die in the lymph nodes (reviewed in Ref. 19), but new evidence suggests that migratory DCs may traffic further to the bone marrow24 and thymus25. Further investigation is required to establish the immunological significance of these observations.

The second major category of lymphoid-organ DCs are the blood-derived or resident DCs, which constitute the second half of lymph-node DCs and all the splenic and thymic DCs13 (Fig. 1). They can be subdivided into three types that are distinguished by their expression of CD4 and CD8: CD4+ DCs, CD8+ DCs and CD4−CD8− (double negative) DCs (Table 1). Throughout this Review, we refer to the CD4+ and CD4−CD8− DCs as a single CD8− DC type. The lymphoid-organ-resident DCs do not conform to the Langerhans cell paradigm; they develop from bone-marrow precursors within the lymphoid organs without previously trafficking through peripheral tissues15,26,27,28. Furthermore, in the absence of infection, the resident DCs maintain an immature phenotype throughout their entire lifespan29, so they can be distinguished from migratory DCs in the lymph nodes by their lower cell-surface expression of MHC class II and T-cell co-stimulatory molecules29,30. Therefore, almost all splenic DCs and approximately half of the lymph-node DCs are immature under steady-state conditions. This is true not only in mice but also in rats31, cows32, pigs33 and humans34 that are free of overt infections or inflammation. During in vitro culture, lymphoid-organ-resident DCs spontaneously acquire a mature phenotype that is similar to that of the migratory DC populations that have reached the lymph nodes29,35. Such maturation is also induced in situ in mice or humans in response to pathogen-associated stimuli that enter the lymphoid organs29,34,36,37,38,39, or in individuals who suffer multiple traumas34; in the case of traumas, maturation probably results from resident DCs responding to endogenous inflammatory molecules that are released from damaged tissues.

Monocyte-derived DCs: precursors of migratory DCs or 'emergency DCs'? Monocytes can be differentiated into DCs in vitro (reviewed in Ref. 15) (Box 2), and this is the most common DC type used in studies of mouse and human DC biology, and for immunotherapy (for a summary, see the table of the results of clinical trials using DC vaccines that has been compiled by D. Hart and colleagues at the Mater Medical Research Institute). However, it is unclear whether the monocyte-derived DCs correspond to any of the DC subsets contained in the lymphoid organs in the steady state (Table 1). It has been suggested that in these conditions monocytes are precursors of migratory DCs, but this is still a controversial notion (reviewed in Ref. 40).

The contribution of monocyte-derived DCs to immune responses becomes evident during inflammatory processes41,42,43,44. For example, in the skin of mice infected with Leishmania major, large numbers of monocyte-derived DCs accumulate44, and one form of leprosy induces the accumulation of monocyte-derived DCs in the skin of human patients45. If monocyte-derived DCs are in fact the equivalents of migratory DCs, their accumulation during these skin infections could be seen as the result of boosting their normal programme of recruitment and differentiation. On the other hand, infection with Listeria monocytogenes or systemic inflammation induces the accumulation of monocyte-derived DCs in the spleen, an organ that is devoid of migratory DCs in the steady state27,46. In this situation, monocytes seem to represent an 'emergency' source of a distinct DC type that emerges only at sites of inflammation. Of course these two possibilities are not mutually exclusive, and distinct populations of monocytes may be precursors of migratory DCs in the steady state and a source of new DCs in any inflamed tissue, respectively15,40.

The important messages to take from these studies are that monocyte-derived DCs represent only one of the several types of DC that constitute the DC network in vivo, and that the conclusions of functional studies that are based on this DC subset may not be applicable to the other DC types.

Antigen-presenting functions of DC subtypes

If there is one functional feature that defines DCs, it is their high capacity to capture, process and present antigens — a prerequisite for T-cell priming. However, not all DCs have equivalent antigen-presenting roles in vivo. Differences at this level may be determined by the intrinsic abilities of each DC type to capture and process antigens, and to load the resulting antigenic peptides onto their MHC molecules. In vivo, the role of each DC subset can be further influenced by other factors that supersede their intrinsic abilities; for example, the anatomical location of the DC, the accessibility of the antigen to that location and the effect that pathogens may exert on the DC.

Intrinsic antigen-presentation capacities of DC subtypes. On a first approximation, all DCs efficiently present peptide antigens on their MHC class I and II molecules because peptide-ligand binding is, with exceptions, a requirement for cell-surface expression of these molecules3, and all DC subsets express high levels of MHC class I and II molecules, especially when mature. Therefore, the question is not whether the DC types differ in their capacity to generate peptide–MHC complexes, but whether they differ in their ability to incorporate peptides that are derived from a given antigen into their presentation pathways. To be presented, the antigens have to access the compartments where proteolytic degradation generates peptide ligands for MHC class I or II molecules. The peptides presented by MHC class I molecules are derived from proteins degraded mainly in the cytosol by the proteasome, whereas MHC class II molecules present peptides that are derived from proteins degraded in endosomal compartments by the cathepsins and other hydrolytic enzymes (reviewed in Ref. 3).

Antigens can be categorized as endogenous (that is, synthesized by the antigen-presenting cell itself) or exogenous (that is, synthesized by other cells). Any endogenous polypeptide can occur in the cytosol as a functional protein or as a defective ribosomal product47, so DCs continually present peptides that are derived from endogenous proteins on MHC class I molecules, including those derived from proteins that do not normally function in the cytosol, such as secreted or membrane proteins (Fig. 2). Similarly, endogenous proteins that access the endosomal compartments of DCs are efficiently presented on their own MHC class II molecules. Such proteins include components of the endocytic pathway, membrane proteins (which are turned over by endosomal degradation) and even cytosolic proteins that are transferred into endosomes by autophagy or direct translocation3,48 (Fig. 2). Therefore, all DCs constitutively present peptides that are derived from their own components on MHC class I and II molecules49,50,51.

Figure 2: The antigen-presentation pathways in dendritic cells.
figure 2

All dendritic cells (DCs) have functional MHC class I and MHC class II presentation pathways. MHC class I molecules present peptides that are derived from proteins degraded mainly in the cytosol, which in most DC types comprise almost exclusively endogenous proteins (synthesized by the cell itself). MHC class II molecules acquire peptide cargo that is generated by proteolytic degradation in endosomal compartments. The precursor proteins of these peptides include exogenous material that is endocytosed from the extracellular environment, and also endogenous components, such as plasma membrane proteins, components of the endocytic pathway and cytosolic proteins that access the endosomes by autophagy. CD8+ DCs have a unique ability to deliver exogenous antigens to the MHC class I (cross-presentation) pathway, although the mechanisms involved in this pathway are still poorly understood. The bifurcated arrow indicates that the MHC class II and the MHC class I cross-presentation pathways may 'compete' for exogenous antigens in CD8+ DCs, or that the endocytic mechanism involved in internalization of a given antigen may determine whether it is preferentially delivered to the MHC class II pathway or the MHC class I cross-presentation pathway. TAP, transporter associated with antigen processing.

Full size image

When cells become infected with a virus, the endogenous viral antigens are incorporated into their antigen-presentation pathways. DCs are no exception52,53, and all tested DC types present endogenous viral antigens with high efficiency. For example, both the CD8+ and CD8− DCs from the spleen present viral antigens on their MHC class I and II molecules if they are infected with herpes simplex virus (HSV)38 or influenza virus (L.J. Young et al., unpublished observations), and skin DCs present antigens that are encoded by infecting lentiviral vectors54. However, many viruses have mechanisms that interfere with the antigen-presentation pathways of the cells they infect55, so the participation of DCs in antiviral responses can be compromised if they become infected. These conclusions are relevant for the evaluation of the studies of DC involvement during viral infections in vivo that we describe below.

The presentation of exogenous antigens is more complicated than the presentation of endogenous antigens because it relies on the ability of cells to deliver the antigens to the correct processing compartments. These antigens must first be endocytosed by pinocytosis, phagocytosis or receptor-mediated endocytosis3. The internalized antigens thus become readily accessible to endosomal proteases and so can be presented by MHC class II molecules. In addition, some cells can present these antigens via MHC class I molecules, a process known as cross-presentation. This pathway is of particular relevance in DCs because they appear to be the main cell population that can cross-present antigens in vivo56, and this enables them to play important roles in tolerance induction and in antiviral and antitumour immunity57. As most cells express MHC class I molecules but only a few can cross-present, this pathway must rely on specific machinery to handle endocytosed antigens, but how this machinery works remains obscure58 (Fig. 2).

Differences in the mechanisms that are available to capture exogenous antigens offer opportunities for functional specializations among DC subsets. For instance, among the lymphoid-organ-resident DCs, the CD8+ DCs are the most efficient at phagocytosing dead cells and, consequently, at MHC class II presentation and MHC class I cross-presentation of cellular antigens59,60,61,62,63,64 (Table 2). This is probably due to the differential expression of as-yet-unidentified receptors for dead cells rather than to differences in phagocytic capacity because all DCs phagocytose latex beads or bacteria60,61,64,65. Indeed, multiple putative antigen receptors are differentially expressed among DC types; for example, the C-type lectin receptors CD205 (also known as Ly75), CD207 (also known as langerin), CIRE, 33D1 and the mannose receptor66,67,68,69,70,71, as well as some TLRs72,73,74, which are normally associated with pathogen recognition but can capture antigens as well (Table 2). Differential expression of these receptors can provide each DC subset with distinct capacities to capture and initiate responses against specific pathogens. For instance, the expression of TLR11 by CD8+ DCs enables this subset to play the major part in MHC class II presentation of the Toxoplasma gondii antigen profilin75. The third important mechanism of endocytosis in DCs is pinocytosis, which probably has an important role in the presentation of soluble proteins76, but no significant differences have been described in pinocytic activity among DC subsets64,77,78.

Table 2 Uptake, MHC class I and MHC class II presentation of several model antigens by CD8+ and CD8− dendritic cells
Full size table

Are there further specializations among DC types downstream of antigen capture? Numerous studies indicate that this is the case. Among the lymphoid-organ-resident DCs, CD8+ DCs are by far the most efficient at cross-presenting cellular59, soluble78 or latex-bead-associated antigens64, or antigens captured by C-type lectin receptors68,69 (Table 2). Comparisons of Langerhans cells with CD8+ DCs also showed that although Langerhans cells can cross-present, they are approximately tenfold less efficient than CD8+ DCs79. In the case of cellular antigens, or antigens taken up by CD205, the contrast between CD8+ and CD8− DCs can be explained by differential uptake59,60,61,62,63,64,68,69,80. However, CD8− DCs are also inefficient at cross-presenting antigens that are equally captured by the two subsets (such as pinocytosed soluble antigens or antigens associated with latex beads)64 or that only they can capture (such as by DC inhibitory receptor 2 (DCIR2) or dectin-1)68,69 (Table 2). By contrast, CD8− DCs seem to be more efficient than CD8+ DCs at presenting exogenous antigens by MHC class II molecules (Table 2). This is particularly true for phagocytosed antigens64 and for antigens captured by C-type lectin receptors68,69,80, although less so for pinocytosed soluble antigens64,77,78,81 (Table 2).

One potential explanation for this dichotomy is that the MHC class I and II pathways are in general more efficient in CD8+ and CD8− DCs, respectively. In support of this suggestion, the CD8+ DCs express larger amounts of the components of the MHC class I machinery, whereas CD8− DCs express larger amounts of the components of the MHC class II machinery68. This is an attractive hypothesis, although it has some caveats. First, it is unlikely that lack of cross-presentation in CD8− DCs is due to insufficient expression of the components of the MHC class I machinery because these DCs can present endogenous antigens on their MHC class I molecules as efficiently as CD8+ DCs (see above). Second, it is questionable that the reported differences in expression of several of the components of the MHC class II machinery can have a major impact on antigen presentation overall, because even their complete absence has a small effect (such as lack of asparaginyl endopeptidase82, cystatin C81, HLA-DO83,84 or IFNγ-inducible thiol reductase85,86). Although it cannot be disregarded that the total effect of the combined differences may have an impact on the presentation pathways overall. In any case, it is clear that CD8+ DCs can have important roles in MHC class II presentation because they are the predominant, if not the only, DC subset that carries out this function when the antigen is cell associated62 or profilin75. Furthermore, previous comparative studies have shown that CD8+ and CD8− DCs synthesize similar amounts of MHC class II molecules, and load them with similar antigenic peptides29,50,81,87; that presentation of endogenous self or viral antigens by MHC class I and II is carried out equivalently by CD8+ and CD8− DCs (see above)38,49,51; and that exogenous soluble antigens are efficiently presented by MHC class II by both subsets29,64,77,78,81.

We suggest an alternative explanation for the differential abilities of CD8+ and CD8− DCs that is based on three premises (Fig. 2). The first is that the MHC class I and II presentation pathways are fully operational in both DC subsets. The second is that cross-presentation requires specialized machinery that is present in CD8+ DCs but largely absent in CD8− DCs64, so CD8− DCs can present exogenous antigens only through the MHC class II pathway (Fig. 2). The third premise is that CD8+ DCs can deliver exogenous antigens to either the MHC class II or the MHC class I cross-presentation pathways, but how much of a given antigen is delivered to each pathway depends on the mechanism involved in its uptake70. Therefore, pinocytosed soluble antigens efficiently access both pathways, antigens captured by phagocytosis or CD205 are delivered preferentially to the cross-presentation pathway, and antigens captured by the mannose receptor hardly access the MHC class II pathway at all70.

This hypothesis implies a 'branched' view of the endocytic route of DCs as opposed to a 'linear' one (Fig. 2). It has been generally assumed that endocytosed antigens inevitably end up in compartments for MHC class II presentation that are located somewhere along a linear endocytic pathway3. However, increasing evidence indicates that the endocytic pathway of DCs (and possibly other cells as well) is more 'multidirectional' than is usually appreciated, so that the destination and fate of an antigen may vary depending on the mechanism used for its internalization. There is experimental support for this hypothesis.

First, DCs must constantly process the large amounts of extracellular proteins that they constitutively endocytose76, and many of their membrane proteins — notoriously, MHC molecules — have short half-lives50,87; however, some antigens can be remarkably long-lived88,89, accumulate in 'storage' compartments90 and can even be 'regurgitated' for transfer to B cells91. Second, the range of endosomal proteases encountered by internalized antigens in DCs is a subset of the total repertoire92,93. Third, DCs preferentially use phagosomes that contain TLR ligands as a source of peptide–MHC class II complexes over phagosomes that are devoid of such ligands94. Fourth, in DCs, the same peptide–MHC class II complex can be present in two different conformations (type A and type B conformers), whereas other cell types can only generate type A conformers, suggesting that the unique compartments where type B conformers originate95 are found only in DCs96. Fifth, analysis of intracellular trafficking of two parasite antigens captured by the same DCs showed that the two antigens accessed non-overlapping compartments97. Last, the endosomal compartments that are reached by soluble proteins taken up by pinocytosis are different to those reached if the proteins are internalized by the mannose receptor70,98,99.

On the basis of these observations, it is plausible that cross-presentation requires the delivery of antigen to a specialized compartment (or secondary pathway with multiple compartments) that is found only in CD8+ DCs70 (Fig. 2). It remains unclear which are the specific features of this putative compartment that enable cross-presentation58. They might include unique protease activities100; differential acidification89; a mechanism of antigen transfer to the cytosol directly101,102 or after passage through the endoplasmic reticulum103; the ability to fuse with the endoplasmic reticulum to generate hybrid compartments104,105,106; or the destination of a unique MHC class I trafficking pathway107. Future investigation should establish whether this is a valid model, identify which of these potential mechanisms confer CD8+ DCs their capacity to cross-present, and verify whether such mechanisms occur or can be upregulated108 in other DCs.

DC subsets in tolerance and immunity in vivo

The role of DC subsets in the presentation of self and pathogen antigens in vivo. In the discussion above, we have summarized the studies that have addressed the intrinsic antigen-presentation abilities of DC populations. The types of antigen used in these studies were mainly purified or recombinant proteins in soluble form, or that were associated with cells or artificial particles. In the discussion below, we review the studies that have analysed the role of the DC subtypes in the presentation of real self or pathogen antigens in vivo.

In the thymus, the main role for DCs is to induce central tolerance by presenting self antigens to developing thymocytes. Such antigens are endogenous proteins that are expressed by DCs, and tissue-specific antigens that are ectopically expressed by thymic epithelial cells109. The ability of the thymic DCs to present exogenous antigens on MHC class II molecules, and to cross-present them on MHC class I molecules, allows thymic DCs to mediate negative selection of both CD4+ and CD8+ thymocytes110,111. This task may be assisted by DCs that enter the thymus from peripheral tissues25. Despite this process of thymic selection, autoreactive T cells can escape thymic selection and enter the periphery, and these must be held in check by mechanisms of peripheral tolerance that are elicited primarily by DCs in the spleen and lymph nodes112.

The role of the DC subsets that are contained in the secondary lymphoid organs is determined by whether they have access only to antigens carried in the blood, only to those contained in peripheral tissues, or both. Unlike the skin, the gut or the respiratory mucosa, the blood is rarely referred to as a peripheral tissue that is subject to DC surveillance, but the blood represents a major pathway for the dissemination of self proteins, which might be captured by lymphoid-organ-resident DCs to induce tolerance. More importantly, a plethora of major pathogens readily access the blood upon inoculation in the skin by arthropods, such as the causal agents of yellow fever (a virus), Lyme disease (Borrelia spp., a bacterium) or malaria (Plasmodium spp., a protozoan). Other pathogens enter the blood after colonizing the gut — for example, Salmonella spp. or Toxoplasma spp. The DCs that reside in the lymph nodes and the spleen are ideally located to monitor the blood, detect these infections and undergo maturation in situ to initiate immunity1. It is therefore important to establish the antigen-presenting function of this DC group.

Analyses of the presentation of intracellular pathogens that are inoculated by the intravenous route have confirmed functional specializations among the lymphoid-organ-resident DC subsets. Measurements of MHC class I presentation of antigens from herpes simplex virus type 1 (HSV1), influenza virus, vaccinia virus or the cytosolic bacterium L. monocytogenes identified CD8+ DCs as the major subset involved113,114. This could reflect that these pathogens only infect CD8+ DCs115, which would then present the antigens using the endogenous route, or that MHC class I presentation of these antigens occurs in vivo by cross-presentation, the pathway found predominantly in CD8+ DCs. In support of the second alternative, impairment of cross-presentation in vivo, caused by the induction of systemic DC maturation with TLR ligands or malaria infection, led to the inhibition of cytotoxic T lymphocyte (CTL) priming against subsequent infections with HSV1 or influenza virus38. The role of DC subsets in MHC class II presentation of these viruses has not been directly assessed, but the fact that CTL priming requires the presentation of viral antigens on MHC class I and II molecules by the same DCs116 implies that at least CD8+ DCs are also involved in this activity.

The importance of MHC class II presentation by CD8+ DCs has been clearly established in a T. gondii infection model. Profilin is an immunodominant antigen of this parasite and is preferentially captured by CD8+ DCs through TLR11 (Table 2), so this is the main DC type that drives a CD4+ T-cell response against T. gondii75. It is likely that this is accompanied by cross-presentation of profilin-derived epitopes, but this has not been established yet.

Studies of the presentation of antigens expressed by the malaria parasite (Plasmodium spp.) have provided a more complete picture of the function of lymphoid-organ-resident DCs against blood infections. The main cross-presenting DCs in this model of infection are also the CD8+ DCs, whereas MHC class II presentation is carried out more efficiently by CD8− DCs (Ref. 39; R.J. Lundie et al., unpublished observations). These observations suggest that both DC subsets can capture malaria parasites and present the parasite antigens according to the intrinsic ability of each DC type. The importance of this activity has been highlighted by kinetic studies of DC function during the course of malaria infection. At early time points, the DCs efficiently present parasite antigens, but this presentation declines gradually as parasitaemia increases (Ref. 39; R.J. Lundie et al., unpublished observations). This state of 'DC paralysis' has been attributed to several causes, the relative importance of which may vary depending on the mouse strain, parasite species or time post-infection117. Such effects not only impair the immune response against the malaria parasite, but also against secondary infections38, and may contribute to poor vaccination outcomes or poor control of viral infections, in areas where malaria is endemic118,119.

Similar effects have been described in pathologies that are associated with systemic induction of DC activation, such as during sepsis38,120. Although the causes for the immunosuppression associated with these pathologies are probably complex, restoration of DC function can reverse some of their deleterious effects121. The implication of these studies is that antigen presentation by lymphoid-organ-resident DCs has a major role in immunity to blood-borne pathogens and that impairment of this function can cause severe immunodeficiency.

Presentation of peripheral-tissue antigens. It might be expected that the DCs that present antigens which are expressed in peripheral tissues should be the migratory subsets, but the real picture is more complicated. In several studies, the migratory DCs were found to carry tissue antigens captured in the gut or skin to their respective lymph nodes22,122. This finding implies that migratory DCs can capture cellular material, so this activity is not restricted solely to the CD8+ DC subset. As expected, the migratory DCs present tissue antigens on MHC class II molecules123. More surprisingly, however, the resident CD8+ DCs also present gut antigens on MHC class II molecules123, and they are the main, if not the only, DC subset that cross-presents pancreatic antigen on MHC class I molecules124. This suggests that migratory DCs transfer antigen to the resident CD8+ DCs upon entry into the T-cell areas of the lymph nodes125. Alternatively, it might be that the CD8+ DCs capture antigens that are ectopically expressed by epithelial cells within the lymph node126. In either scenario, the high cross-presenting capacity of CD8+ DCs might account for their dominant role in MHC class I presentation of self antigens. This does not exclude a contribution by the migratory DCs because, even though they are less capable of cross-presentation on a per-cell basis79, this capability may be significant when the antigen is available to a large number of migratory DCs, or when the antigen is expressed at high levels127. The conclusion of these studies is that self antigens are presented in the steady state on MHC class I and II molecules by both immature resident CD8+ DCs and mature migratory DCs, with the former being biased towards MHC class I cross-presentation and the latter towards MHC class II presentation. This activity is probably important in the induction of peripheral tolerance, although it remains controversial whether any, only one, or both migratory and resident CD8+ DCs are tolerogenic in the steady state57,126,128.

Studies of the presentation of antigens that are derived from pathogens that infect the skin, lungs or gut by DC subsets have similarly yielded some surprising and also some controversial findings. A common conclusion of numerous studies is the lack of involvement of Langerhans cells in the presentation of pathogens that infect the skin, such as L. major129,130, and influenza virus, vaccinia virus or HSV131,132. By contrast, interstitial (dermal) DCs and CD8+ DCs seem to have crucial roles in the presentation of these antigens38,54,131,132,133. These two populations are also important in the presentation of viruses that infect the lungs or the gut134,135. However, the specific function of each population is unclear. It is generally accepted that migratory DCs have a fundamental role in carrying antigens that are acquired in the peripheral tissues to the lymph nodes, and that the antigens can somehow be transferred to the CD8+ DCs, but there is no agreement about the relative contribution of each population to MHC class II presentation and MHC class I cross-presentation. This might be due to differences in the effect of viruses on the DCs, to the number of migratory DCs that have access to the virus and to the amount of antigen available to each individual DC type. A problem that hampers interpretation of the results is that most studies have measured presentation by MHC class I or II, but not both. Nevertheless, our current understanding of migratory properties and the intrinsic antigen-presentation abilities of distinct DC subsets allow us to draw a working hypothesis, modified from an original proposal by Carbone and colleagues125, for DC-subset specializations in different types of infection.

A model for antigen-presenting functions of migratory and resident DC populations. A first scenario is one in which the migratory DCs are themselves infected by a virus that has no deleterious effect on the DCs — for example, a lentivirus vector54 (Fig. 3a). In this situation, the migratory DCs are the main subset that presents endogenously produced antigens on MHC class I molecules54 and, most likely, on MHC class II molecules, in the draining lymph node.

Figure 3: The antigen-presenting functions of migratory and resident dendritic-cell populations in three hypothetical scenarios of tissue infection.
figure 3

a | A virus infects dendritic cells (DCs) in tissues but does not inactivate their capacity to migrate and present antigen. The migratory DCs can by themselves present endogenous viral proteins on their own MHC class I and II molecules and stimulate CD4+ and CD8+ T cells. b | A virus infects a small portion of the tissue that is drained by the lymph node, avoiding infecting DCs or inactivating infected cells. Only a small proportion of the DCs migrating to the lymph node carry viral antigens. Some of these immigrants die, transferring the antigens to CD8+ DCs. Both the migratory DCs that survive plus the resident CD8+ DCs now contain exogenous viral antigens, which they present using, by preference, their MHC class II and MHC class I cross-presentation pathways, respectively. Migratory DCs therefore present viral antigens mainly to CD4+ T cells and resident CD8+ DCs present viral antigens to CD8+ T cells. c | As in panel b, but the infection is more widespread, so a larger proportion of migratory DCs that reach the lymph node contain antigen that is captured in the tissue. Although the efficiency of presentation of the viral antigens by migratory and resident CD8+ DCs remains the same on a per-cell basis, the relative contribution of the migratory DCs to cross-presentation in the lymph node is higher.

Full size image

The second scenario is one in which the migratory DCs are not infected, or if they are infected they cannot present antigens because they die (whether 'voluntarily' or induced by the virus) or because the virus expresses immunoevasins that disable antigen presentation55 (Fig. 3b,c). This may occur during infection with vaccinia virus, influenza virus, reovirus or HSV52,54,113,131,132,135. In this setting, the migratory DCs are still required to carry viral antigens, whether endogenously expressed or captured from infected cells, to the lymph nodes133, and transfer the antigens to the resident DCs, perhaps in the form of endosomal vesicles or apoptotic bodies. The main resident DC population that acquires the antigen would be the CD8+ DCs, owing to their special ability to capture dead cells or cell fragments. The migratory DCs that survive and have not been inactivated by viral immunoevasins, as well as the resident CD8+ DCs, can then present the antigen according to their intrinsic abilities: migratory DCs preferentially through the MHC class II presentation pathway and resident CD8+ DCs preferentially through the MHC class I cross-presentation pathway (Fig. 3b). In situations in which few of the migratory DCs contained in the lymph nodes have immigrated from the site of infection — for example, during HSV infection of a small portion of skin — many CD8+ DCs will obtain antigen provided by a few immigrated dermal DCs; therefore the subset that cross-presents antigen on MHC class I will be mainly the CD8+ DCs131, whereas those that present antigen on MHC class II will be mainly the immigrated DCs132,135 (Fig. 3b).

When the infection affects a much larger portion of the tissue that is drained by the lymph node, as is the case during influenza virus infection of the lung, many interstitial DCs from the tissue will carry the viral antigen. In this scenario, the relative antigen-presentation efficiency of the two DC subsets may still be the same on a per-cell basis, but the relative number of migratory DCs that contain antigen is much higher and, correspondingly, their relative contribution to cross-presentation within the lymph node increases134 (Fig. 3c). This hypothetical model may be further complicated by the recruitment of monocytes to sites of inflammation, followed by their conversion into DCs, because the monocyte-derived DCs may then take over some of the antigen-presenting functions that are initially carried out by the migratory and lymphoid-organ-resident DC subsets42,44. We believe this general model can explain most of the results reported so far, but more work needs to be carried out to confirm, modify or discard it.

Conclusions and future directions

The definition of methods to generate mouse and human DCs from monocyte precursors in vitro represented a turning point for the DC field (Box 2), as this led to an explosion of studies of the biological properties and therapeutic potential of these cells. However, this was a mixed blessing because it contributed to a set view of DCs as a widely distributed but largely homogeneous network of cells, the life cycle of which generally followed the Langerhans cell paradigm. It is now clear that this network is heterogeneous, and that distinct DC populations emerge from independent developmental branches that follow different life cycles and have non-overlapping functions. Indeed, only now are we starting to learn the role that monocyte-derived DCs have in vivo136.

The heterogeneity of the DC network only became evident when phenotypically distinct populations of DCs were purified from lymphoid organs and systematically compared. These studies have been carried out mainly in the mouse system. Unfortunately, our knowledge of human DC heterogeneity is poor by comparison; we hardly know which are the human equivalents of the mouse lymphoid-organ DCs, but we do know that most are not monocyte-derived DCs136. A priority for future studies should be to close this knowledge gap. Defining the cell-surface phenotype and function of the human lymphoid-organ DCs should help to develop targeting strategies for vaccination, whether the goal is to induce immunity or tolerance. It may also help to define methods to generate them in vivo, with the consequent benefits for basic biological studies and immunotherapeutic applications. To achieve this goal, it will also be important to expand the studies of DC-subtype specialization in healthy and diseased states in the mouse system. More experimental models that can measure simultaneously the antigen uptake and presentation by DC subsets in vivo will be required.

Studies of DCs have revealed novel mechanisms of antigen handling, processing and presentation58,87. More work will be required to understand how different DC types regulate the preferential delivery of exogenous antigens to the MHC class II or MHC class I (cross-presentation) pathways. The mechanisms of cross-presentation in particular are highly controversial58. One of the problems with studies of this pathway is that even though some phenomena appear to correlate with the ability to cross-present, it can be difficult to establish whether the phenomenon is a requirement for cross-presentation, a 'red herring' without direct function in this pathway or even a technical artefact. Comparative analyses of closely related DC types that differ in antigen handling, MHC class II presentation and MHC class I cross-presentation, such as the lymphoid-organ-resident CD8+ and CD8− DCs, should shed light on these issues. To conclude, we anticipate a dramatic increase in the number of studies of DCs that will be based on populations that are extracted from mouse and human lymphoid organs, or their in vitro equivalents73. We believe this will lead to important discoveries from both the immunological and cell biological points of view, and will open the way for novel therapeutic developments.