Human Dendritic Cells: Potent Antigen-Presenting Cells at the Crossroads of Innate and Adaptive Immunity¹

Paul Langerhans first described dendritic cells (DCs)³ in human skin but thought these were cutaneous nerve cells (1). Steinman and Cohn (2) discovered these cells almost a century later in mouse spleen and applied the term “dendritic cells” based on their unique morphology. For almost two decades, however, proving the specialized properties of these cells for initiating immunity required depletion of other cells like monocytes, macrophages, and B cells that also presented Ag. Potent immunostimulatory activity was enriched in the remaining DCs. The paradigm emerged that Ag capture segregated to immature, peripherally distributed DCs, with Ag presentation and lymphocyte stimulation being an acquired property of mature DCs in secondary lymphoid organs. The essential and pivotal role of DCs in the onset of immunity was thus established.

Progress in the study of DC biology exploded in the 1990s. Investigators developed cytokine-driven methods for expanding and differentiating DCs ex vivo in both mouse and human systems (3, 4, 5, 6, 7, 8, 9, 10), and further refinements continue to emerge (11, 12, 13, 14, 15). For the first time, sufficient numbers of DCs became accessible for large-scale study and applications.

DCs are the single most central player in all immune responses, both innate and adaptive. DCs are exceptionally potent immunogens under inflammatory conditions, yet are also critical to the induction and maintenance of self-tolerance in the steady state (16, 17, 18). The heterogeneity of DCs and their activation states afford investigators more opportunities to define and manipulate the immune response using these specialized leukocytes.

Human DCs are all bone marrow-derived leukocytes (19). They are distinct from follicular DCs, which are not leukocytes and of stromal origin. Follicular DCs interact with B cells in the germinal center, but they have no role in T cell, NK cell, or NKT cell immunity as do conventional or plasmacytoid DCs (20).

Human DCs comprise at least four types defined under cytokine-driven conditions in vitro (Fig. 1). These include conventional or “myeloid” DCs (5, 6, 8, 10, 12, 14, 15, 21, 22): 1) CD14⁺ blood monocyte-derived DCs (moDCs); 2) dermal DCs or interstitial DCs (DDC-IDCs); 3) Langerhans cells (LCs); and 4) plasmacytoid DCs (23). A trace population of DCs also circulates in human blood. Those DCs are lineage negative, CD11c⁺, CD86⁺, and HLA-DR^bright and express CD83 after activation by brief overnight culture (24). Although not identical, they share many phenotypic and immunostimulatory features with cytokine-generated conventional DCs in vitro.

DCs are also a major component of lymphoid tissues, particularly the T cell areas. In the case of lymph nodes, DCs access these areas via afferent lymphatics. Investigators now know that most DCs in lymphoid organs are immature or semimature in the steady state and efficiently process self-Ags to induce and maintain tolerance (16, 18, 25). Under inflammatory conditions, however, DCs undergo a complex maturation process. The specifics of maturation vary with the stimulus, but the result is stimulation of innate and adaptive immunity. DCs in lymphoid tissues include populations termed “conventional” or myeloid, as well as “plasmacytoid”, the latter named because of their morphological resemblance to plasma cells (23). This review will focus on human DCs generated in vitro with cytokines, because these have yielded ample numbers and purity for experimental and clinical evaluations.

The precursors of DC progeny generated in vitro with cytokines should approximate resident populations of DCs that exist in vivo under steady-state conditions. The cytokine-driven DC progeny are induced populations and hence would not be found in vivo except under inflammatory conditions. Manipulation of immunity using DCs generated in vitro should therefore exploit the less mature and nonactivated forms to promote tolerance and the activated and mature forms to break tolerance and promote immunity.

The most accessible DC precursor is the CD14⁺ monocyte in peripheral blood, which differentiates into CD14^neg, CD11c⁺, CD83⁺⁺, HLA-DR^bright, moDCs under the influence of GM-CSF and IL-4 (5, 6, 12, 26) (Fig. 1 and Table I). Blood monocytes also exhibit a sensitivity to ambient cytokines that can drive their differentiation to DCs resembling either activated LCs or DDC-IDCs (27, 28, 29, 30, 31). This may provide biologic flexibility to monocyte precursors and moDCs as first responders to pathogens, allergens, and other causes of inflammation.

Plasmacytoid DC precursors comprise another population identifiable in the circulation as lineage negative, HLA-DR^bright, BDCA-2⁺, BDCA-4⁺, and CD123^bright cells (23, 32) (Fig. 1 and Table I). Human plasmacytoid DCs lack myeloid markers CD11c and CD33. In contrast, the mouse counterpart expresses low levels of CD11c along with B220 and Gr1 and expresses CD123 only after fms-like tyrosine kinase 3 ligand (Flt3-L) treatment (33, 34, 35). Conventional DCs in humans also express the CD123 receptor for IL-3, but less than plasmacytoid DCs (36). IL-3 supports plasmacytoid DC viability, and it is partially interchangeable with GM-CSF for conventional DCs because of a common β-chain shared by receptors for both cytokines (37).

G-CSF mobilizes increased proportions of plasmacytoid DCs from the bone marrow (38, 39), but Flt3-L is the main cytokine for development of plasmacytoid DCs from hemopoietic stem cells (40, 41). Thrombopoietin in combination with Flt-3L supports the generation of human plasmacytoid DCs in vitro from CD34⁺ hemopoietic progenitor cells (HPCs) (42), but it has proven more difficult to achieve large yields in comparison to those of conventional DCs generated in vitro with cytokines (3, 4, 7, 8, 9, 10, 12, 13, 14, 15).

Freshly isolated plasmacytoid DCs express much lower levels of MHC and costimulatory molecules than their conventional DC counterparts (23, 43). They also capture, process, and load Ags onto MHC molecules less effectively. These nonactivated plasmacytoid DCs are therefore poor stimulators of T lymphocytes. IL-3 in combination with CD40 ligand (CD40L) or microbial products leads to full plasmacytoid DC activation and more potent lymphocyte stimulation (44, 45, 46, 47, 48, 49).

In addition to circulating blood DCs and moDCs, the conventional or nonplasmacytoid DCs include LCs and DDC-IDCs. These two resident populations of DCs are distributed throughout the body where the immune system first encounters Ag. Thus, LCs localize to epithelial surfaces of skin and mucosa, and the DDC-IDCs localize to the subepithelial tissues of the dermis in skin and the interstitia of solid organs. That LCs are bone marrow-derived leukocytes was first established in 1979 (19), and that LCs are cutaneous DCs was determined several years later (50).

These DCs develop in vitro from CD34⁺ HPCs under the aegis of defined cytokines that include GM-CSF and TNF-α (4, 7, 9) (Fig. 1 and Table I). c-kit-ligand (7, 9) and Flt-3-L (41) are also critical to expansion of the clonogenic DC progenitors but do not affect DC differentiation. TGF-β, especially under serum-free conditions, supports LC differentiation (10, 14, 15, 51). These LCs have Birbeck granules and express Langerin and e-cadherin, which are down-regulated with activation (15, 52, 53).

Human LCs do not express CD11b, an epitope found on moDCs and DDC-IDCs (15). DDC-IDCs develop through a CD14⁺ intermediate (8, 21, 22), for which the alternative macrophage differentiation pathway is suppressed by IL-4 (26), just as with moDCs. LCs do not develop through a CD14⁺ intermediate (21), although some investigators have reported differentiation of Langerhans-like DCs from CD14⁺ blood monocytes (28, 29). All conventional DCs express CD1a, which is not limited to LCs. DDC-IDCs and moDCs, but not LCs, express the other CD1 isoforms, including CD1d (15), which presents glycolipid Ags (see Table I).

All DCs require some form of terminal maturation to become fully immunogenic. We now know that there are many different environmental stimuli that can mature DCs in different ways. All of the various lymphocyte responses require activation by mature DCs, so that DC maturation is a pivotal event in the control of innate and adaptive immunity. Maturation also ensures that DCs do not revert to a less mature and less immunogenic form. Microbial products constitute a physiologic activation stimulus via TLRs on both plasmacytoid and conventional DCs (47, 48). CD40L (CD154), either expressed by activated T cells or as a multimeric recombinant protein, can also mature DCs (54, 55). A combination of inflammatory cytokines that includes IL-1-β, TNF-α, IL-6, and PGE₂ (11) is often used to mature DCs for study in vitro and for use in clinical vaccine trials.

Terminal activation down-regulates certain chemokine receptors like CCR1 and CCR5, adhesion molecules like e-cadherin on LCs, and Ag uptake. Maturation up-regulates numerous other properties of immunogenic DCs, many expressed with distinct kinetics, including inducible costimulatory molecules, cytokine secretion, and Ag presentation. Mature DCs also increase CCR7 expression, which mediates trafficking into lymph nodes after vaccination (56, 57, 58). A theoretical disadvantage is the lower IL-12p70 secretion after exposure to PGE₂ (59) compared with CD40L. Data in vitro (15) and in vivo (60), however, demonstrate ample Th1 and CTL responses after stimulation by DCs matured in the presence of PGE₂, regardless of any effect on IL-12p70.

moDCs and plasmacytoid DCs have been, respectively, labeled DC1 and DC2 because of their propensity to stimulate Th1- vs Th2-type responses. This oversimplification, however, neglects stimulation of more varied T cell responses. The label DC1 also fails to incorporate any consideration of LCs or DDC-IDCs as conventional DCs. Skewing toward Th2 responses by plasmacytoid DCs in the presence of IL-3 led to their being termed DC2. This has no bearing, however, on their role as tolerogenic DCs in inducing CD4⁺ and CD8⁺ regulatory T cells (61, 62). It also overlooks the major physiologic role of plasmacytoid DCs as the most abundant source of type I IFNs after activation by viruses (44, 45, 46, 47, 48, 49, 63). DDC-IDCs and moDCs have been considered homologous, because both develop from a postproliferative, CD14⁺ precursor. More recent phenotypic and functional data, however, indicate that these two types of DCs are distinct (15, 64, 65). The specific descriptive term for each type of DC is therefore more useful than nomenclature like DC1 vs DC2, or myeloid vs lymphoid DCs.

DCs use phagocytosis, endocytosis, pinocytosis, and specific receptors to capture microbial pathogens, dead or dying cells, immune complexes, and other Ags for immune presentation. DCs share most features of Ag processing and presentation on class I and II MHC molecules with other APCs. Ags acquired from the extracellular environment are typically processed onto class II MHC molecules, whereas class I MHC molecules bear Ags synthesized in the cytosolic compartment. DCs are endowed, however, with the capacity to cross-present exogenous Ags on class I MHC complexes (66, 67, 68, 69, 70, 71, 72).

Cross-presentation entails processing and presentation of exogenous Ag on the DC’s own MHC molecules, including class I, to autologous MHC-restricted T cells, regardless of the MHC alleles expressed by the Ag source. Investigators have emphasized moDCs in studies of cross-presentation because of their ready availability in vitro and high phagocytic activity. Phagocytosis is only one means of Ag uptake, however, and the alternative fate of phagocytosed Ag is sequestration and degradation rather than processing and MHC-restricted presentation. In fact, cytokine-induced, CD34⁺ HPC-derived LCs are much less phagocytic than moDCs, yet elicit more potent T cell responses by cross-presentation (15). The specific receptors involved on these induced LCs have not yet been defined. LCs in situ, however, do express low levels of DEC-205 (CD205), one of the C-type lectin receptors, and even up-regulate this receptor with maturation in skin explant cultures (73).

Much has also been made of the distinction between apoptotic and necrotic cell death as a source of cross-presented Ag (74). Whether Ag remains intact or denatured during apoptosis or necrosis, as well as any association with additional danger signals, are the greater determinants of effective cross-presentation and a tolerant or immune outcome (75).

The early phases of infection elucidate how DCs first encounter, respond, and limit an inflammatory insult to the steady state. DCs express pattern recognition receptors that bind pathogen-associated molecular patterns expressed by microbial products or molecules from damaged host tissues. These pattern recognition receptors include TLRs, most of which signal through MyD88.

Examples of TLR-ligand interactions include peptidoglycan binding of TLR2, viral dsRNA binding of TLR3 (mimicked by synthetic dsRNA, e.g., poly(I:C)), LPS binding of TLR4, viral ssRNA binding of TLR7, and unmethylated bacterial CpG DNA motif binding of TLR9. Conventional DCs express several of TLRs1–6 and 8, depending on the subset and activation state (14, 48, 76). Plasmacytoid DCs are the only DCs that express TLR7 and 9, although blood monocytes express TLR7 and B cells express TLR9 (48, 76).

Ligand binding of TLRs up-regulates CD83, costimulatory molecules, and CCR7, which drives DC migration to T cell areas of draining lymph nodes (46, 48). TLR activation also leads to the secretion of huge amounts of IFN-α by plasmacytoid DCs (44, 45, 48, 49, 76) or IL-6, IL-10, TNF, and/or IL-12 by conventional DCs, depending on the subset and particular stimulus (48, 76).

These early activated DCs thus play an important role in the bystander activation of other DCs and the recruitment of NK, NK T, and CD8⁺ T cells, which then secrete IFN-γ and other inflammatory cytokines that support the ensuing adaptive immune response (49, 63) (Fig. 2). Appropriate stimulation of TLRs on DCs by their respective ligands can thus initiate the entire spectrum of innate and, in turn, acquired immunity. The inflammatory cytokine combinations (11) used in vitro for terminal DC maturation and activation mimic the sequelae of physiologic TLR ligand binding.

Another group of pattern recognition receptors expressed by DCs are the C-type lectin receptors, which bind the carbohydrate moieties of glycoprotein self-Ags and pathogens for processing and presentation on MHC molecules. Resident or nonactivated forms of DCs express these receptors, which are specialized for Ag capture and processing. Expression by immature moDCs includes among others, the macrophage mannose receptor (MMR: CD206), DEC-205 (CD205), DC-specific intercellular adhesion molecule-grabbing nonintegrin (DC-SIGN) (CD209), BDCA-2, and dectin-1. LCs express DEC-205, and DDC-IDCs express DEC-205 and DC-SIGN (73). LCs also express Langerin (CD207), which is an endocytic receptor that induces the formation of Birbeck granules (53). With the exception of DEC-205 (73, 77), DC activation and maturation down-regulate expression of C-type lectin receptors, as DC function changes from Ag uptake to Ag presentation.

TLRs and C-type lectin receptors work in concert to balance immune tolerance with activation. DCs use C-type lectin receptors to sample and present self-Ags and harmless environmental Ags in the steady state, thereby maintaining peripheral tolerance (16, 18). Some pathogens also exploit these receptors to evade immune activation, e.g., HIV and DC-SIGN (78, 79, 80). Perturbation of the steady state by concomitant exposure to an activating stimulus like TLR-binding ligands or CD40L can override any tolerizing function of C-type lectin receptors and lead to immune activation (77, 81).

Resident populations of mouse and human DCs and their precursors express some combination of activating (CD16 (mouse; human monocyte precursor subset), CD32a (human only), and CD64 (mouse and human)) and inhibitory (CD32b (mouse and human)) Fcγ receptors (19, 82, 83, 84, 85, 86). Among cultured and cytokine-induced human DCs, only circulating blood DCs and moDCs maintain expression of Fcγ receptors (84, 87). These are not passive uptake receptors but signal the cell using immunoreceptor tyrosine-based inhibitory and immunoreceptor tyrosine-based activation motifs (82). The balance of activating and inhibitory Fcγ receptors; the binding avidity of allelic receptor isoforms; the species, subclass, and density of the opsonizing IgG Ab; and the size of the immune complex all determine the outcome of these receptor-ligand interactions (87, 88, 89, 90, 91). Targeting tumor Ags to Fcγ receptors on DCs increases the efficiency of Ag cross-presentation (91, 92, 93). Activating Fcγ receptors also promote DC maturation and activation (87), which may be unrelated to effects on Ag processing (91). CD32b opposes signaling through activating Fcγ receptors and may exert a separate inhibitory effect as well (87, 88, 92).

Investigators have found distinct cytokine profiles, chemokine responsiveness, and functional segregation of LC and DDC-IDC émigrés to discrete T cell or B cell predominant areas of secondary lymphoid organs (94, 95, 96). Putative counterparts of cytokine-induced moDCs have also been identified in vivo (97, 98).

Direct comparisons of these three subtypes of conventional DCs in vitro have shown that all express comparable maturation phenotypes and stimulatory function in allogeneic MLRs. Cytokine-induced, CD34⁺ HPC-derived LCs, however, are superior stimulators of CTL, at least against a recall viral Ag (15). LCs do this in the complete absence of bioactive IL-12p70, even after an optimal stimulus delivered by recombinant human CD40L-trimer (15, 55, 99). Mature LC and DDC émigrés from human skin also do not secrete IL-12p70 (99).

IL-12 has long been considered essential for generating Th1 and CTL responses, but perhaps overemphasized due to inadvertent measurement of inactive IL-12p40 homodimers (100). IL-12p70 supports stimulation of Th1 and CTL, but it may do so by first stimulating NK cells to secrete IFN-γ, which conditions DCs and Th1 responses (101, 102). Both LCs and DDC-IDCs secrete ample IL-12p40, suggesting that other heterodimeric cytokines with activity similar to IL-12 and incorporating the p40 chain, e.g., IL-23, merit further consideration (100, 103).

Recent data have established an important role for DCs in innate immune responses by NK and (65, 104, 105, 106) and NKT cells (107, 108, 109, 110). DCs enhance the reactivity of resting NK cells, which otherwise respond to the aggregate of activating and inhibitory signals on their targets. Among conventional DCs, human moDCs are the most potent stimulators of NK cell proliferation and cytotoxicity (65). LCs lack sufficient IL-12p70 secretion to induce NK cell activation but provide unidentified factors that sustain NK cell proliferation and survival after activation by moDCs (65).

NKT cells are a trace population of T cells sharing some features of NK cells, which express an invariant TCR α-chain and respond to glycolipid Ags presented on CD1d (111, 112). Autologous tissues provide self-glycolipid ligands such that NKT cells are in a chronic state of low level activation with suppressor function mediated by IL-4. Tumors like melanoma provide glycolipid ligands like GD3 (113), and the drug α-galactosylceramide is an artificial ligand. The presentation of glycolipid ligand by CD1d-expressing DCs (107, 108, 110), however, or the simultaneous delivery of a microbial stimulus to DCs resulting in high IL-12p70 secretion (114), leads in either case to robust activation of NKT cells and IFN-γ secretion. The critical operative is the need for DCs to convert NKT cells from suppressor to effector function. Among conventional DCs, only moDCs and DDC-IDCs express CD1d and only moDCs secrete IL-12p70 (15). Therefore, it would seem that only moDCs could mediate NKT cell activation.

In either setting, both DC-stimulated NK and NKT cell innate immune effectors become potent sources of IFN-γ and other inflammatory cytokines. This can support the innate maturation of resident populations of DCs through a reciprocal activating interaction and expand the ensuing adaptive Th1 response mediated by cytotoxic effectors (101, 106, 115) (Fig. 2).

Constitutive CD4⁺CD25⁺Foxp3⁺ Tregs are not as phenotypically discrete in humans as in the mouse. Immature DCs, interacting with constitutive Tregs in a contact-dependent manner, secrete immunosuppressive IL-10. This induces regulatory or suppressor T cells, Tr1 and Th3, which exert their effects in a contact-independent manner by the respective secretion of IL-10 or TGF-β (116, 117, 118). CD4⁺ (61) and CD8⁺ (62) regulatory T cells can develop from primary stimulation by plasmacytoid DCs.

IDO is an immunosuppressive enzyme because it limits T cell proliferation by degrading the rare but essential amino acid tryptophan, which also generates kynurenine and other potentially toxic metabolites (119, 120, 121, 122, 123, 124, 125, 126). Constitutive Tregs express CTLA4, which can condition DCs to become regulatory with an associated increase in IDO activity. Activated T cells also increase expression of CTLA4, which turns off the adaptive T cell immune response and may work through IDO.

Investigators have proposed a subset of regulatory CD123(IL-3R)⁺ DCs that functions through increased IDO activity and tryptophan depletion (120). All human DCs express CD123, but plasmacytoid DCs express much higher levels. It is not clear whether IDO halts T cell responsiveness by tryptophan starvation only or if it can also induce regulatory T cells like Tr1 or Th3. IDO is not active in immature DCs, except in LCs (M. Rossi, C. Antczack, D. H. Munn, and J. W. Young, unpublished data). Therefore, immature DCs should not stimulate inducible Tregs by this mechanism as well as mature DCs. Mature, activated moDCs, LCs, and DDC-IDCs all exert IDO enzymatic activity (M. Rossi, C. Antczack, D. H. Munn, and J. W. Young, unpublished data), which therefore merits investigation as a negative feedback mechanism against otherwise unchecked immune responses.

There is great interest in altering the cytokine milieu that drives DC immunogenicity (127) or in using DCs to expand Tregs for the control of autoimmunity (128, 129). Most current clinical studies, however, use DCs for active immunotherapy trials in cancer. Most tumor Ags are poor immunogens because they are self-Ags or self-differentiation Ags, to which there is considerable tolerance. DCs provide a potential solution to this challenge by coupling tumor Ag with all of the requisite costimulatory ligands, cytokines, and chemokine-directed migration to secondary lymphoid organs. There they can stimulate incoming T cells to exit via efferent lymph into the periphery as cytolytic and helper T cell effectors.

Challenges to designing the optimal DC vaccine include the choice of DC subset or combination of subsets. For example, whether the functional distinctions between conventional DC subsets in vitro (15) have physiologic relevance in vivo is the subject of an ongoing vaccine trial in melanoma at our institution. The presumptive advantage of LCs has been the rationale for other investigators to include CD34⁺ HPC-derived DCs, which comprise LCs among the progeny, in vaccine preparations (130). The malleability of moDC precursors under certain cytokine conditions might also yield moDC progeny that function more like LCs (28, 29). Other unknowns include optimal Ag-loading strategies like peptide pulsing, overlapping polypeptide pulsing, cross-presentation of dying tumor cells, fused tumor-DC heterokaryons, DNA or RNA transfection with or without a vector construct, frequency and route of immunization, and cell dose. One issue that seems finally laid to rest is the need for activated and terminally mature DCs to avoid any reversion to immature DCs that may be inactive or even generate suppressive Tregs (131, 132).

The first human DC vaccine trial used the rare circulating DCs isolated ex vivo from steady-state pheresis products and loaded with tumor-specific idiotypes to treat patients with follicular lymphoma (133). This approach is not selective for any one of several DC subsets in blood, and the yields are low. The advent of the cytokine-generated DC era has supported large-scale clinical evaluations, and a number of trials have followed suit (Refs.130 and 134 , among others). In the aggregate, these studies have shown that DC vaccinations are safe and that tumor-specific T cell responses can be generated by DC vaccination using standard immunologic assays in vitro. Although patients eligible for these early phase clinical trials have advanced disease, clinical responses have been achieved in some instances (130, 133, 134, 135). Major challenges remain in terms of harnessing the capacity of DCs for simultaneous presentation of multiple tumor Ags tailored to their own MHC molecules, rather than presentation of only a few peptides with defined MHC restrictions. Migration of DCs to draining lymph nodes also requires optimization after vaccination.

We propose that the division of labor among distinct human DC subtypes achieves the most effective balance between steady-state tolerance and the induction of innate and adaptive immunity against pathogens, tumors, and other harmful insults. Maintenance of tolerance in the steady state is an active process involving resting or semimature plasmacytoid and conventional DCs. Breakdowns in this homeostasis can result in autoimmunity. Perturbation of the steady state by harmful Ags should first lead to the onset of innate immunity mediated by rapid responders in the form of plasmacytoid and moDC stimulators and NK and NKT cell effectors (Fig. 2). These innate effectors in turn provide abundant inflammatory cytokines, notably IFN-γ, which along with other inflammatory cytokines support the activation and maturation of resident and circulating populations of DCs (Fig. 2). These are critical to the onset and expansion of adaptive immunity. IDO activity exerted by mature DCs may dampen T cell responses by mechanisms to be defined and provide a clue to the unsolved riddle of how DCs turn off otherwise unchecked immune stimulation. Interference with such regulatory mechanisms should be useful in sustaining desirable immune responses. The use of DC subtypes as vaccines to stimulate both innate and adaptive immunity, either in combination or in a prime-boost sequence, may prove most useful clinically by harnessing two valuable effector cell compartments.

We appreciate the contributions of all present and former members of our laboratory, as well as many important collaborators and colleagues whose work has helped to advance the field of DC immunology. We thank Drs. Adam Boruchov, Christian Munz, and Ralph Steinman for helpful review of this manuscript.