Revisiting Current Concepts on the Tolerogenicity of Steady-State Dendritic Cell Subsets and Their Maturation Stages

The introduction of the term dendritic cell (DC) “maturation” has been confusing because it suggested that immature DCs are not fully developed, as the term “maturation” is used for other immune cell types (such as T and B cells and macrophages) to indicate the programming toward fully differentiated and functional cells. Furthermore, the dogma evolved that functionally immature DCs are the only tolerogenic DCs (tolDCs), whereas mature DCs are always immunogenic. Surprisingly, proof for a tolerogenic function of immature DC under steady-state conditions in vivo is limited, despite the successful therapeutic application of immature DC in animal models of transplantation, rheumatoid arthritis, and autoimmunity as well as promising perspectives in human clinical trials (1–5). For the discussion on tolerogenic plasmacytoid DC (pDC), we refer to excellent recent review articles (6, 7). In this review, we propose that functional tolerogenicity of DCs for self-antigens in vivo is connected to activation, different between DC subsets, and mainly directed by the release of cytokines and soluble mediators (Fig. 1). We propose that tolerogenic and immunogenic DC function represent two generally distinct pathways of DC activation, characterized by qualitative and quantitative markers (Fig. 1).

DC “maturation” has been historically introduced as a term to distinguish mature Ag-presenting from immature nonpresenting DC, suggesting that non–Ag-presenting DCs were incompletely developed progenitors (8–10). Only later it became clear that immature DCs are fully differentiated DCs functioning as resident sensors (resDCs) for apoptotic cells, pathogens, and inflammation (11, 12). A key characteristic in DC biology is their migration. DC take up antigenic material and migrate via the lymphatics into the draining lymph node (LN) to present the transported Ags to T cells (migratory DCs [migDCs]). At this “mature” stage, DC exhibit upregulated MHC class I (MHC I), MHC class II (MHC II), and costimulatory molecules at their surface for efficient Ag presentation and naive T cell activation (13). Later, it became clear that, in mice, migDC can transport not only pathogen-derived Ags but also self-antigens and, thereby, can be subdivided into tolerogenic migDC (tol-migDC) and immunogenic or inflammatory migDC (inf-migDC) [(14) and references therein].

The question remains, what are the critical factors or functions that distinguish the tol-migDC from inf-migDC? In this review, we present a selection of markers, including surface molecules, secreted mediators of cytokines and transcription factors that enable this distinction. Some of these markers are expressed in both cell types but at different levels (quantitative) and others only in one of the two types (qualitative) (Fig. 1). The concept that mature DCs can still act tolerogenic and how to distinguish the two types of migDC emerged from data obtained with TNF-matured bone marrow–derived DC (TNF-BMDC). When TNF-BMDC were repetitively injected into mice, surprisingly, IL-10⁺ Tr1 cells were induced that prevented autoimmunity in contrast to injections of immunogenic LPS/CD40-matured DC (14, 15). These in vitro generated tolerogenic “semimature” and CCR7⁺ migratory TNF-BMDC expressed a partially mature phenotype for surface MHC II and costimulatory molecules, lacked proinflammatory cytokine production and transcriptional profile (15, 16). Thereby they resembled Tr1-inducing mature tol-mig-cDC, shown to migrate to lung draining LN after noninflammatory intranasal OVA application with the ability to block asthma in mice (17). Thus, the decisive criteria to distinguish between tolerogenic and immunogenic DC is not exclusively “maturation yes or no” but the degree of maturation together with more recently identified qualitative markers (Fig. 1, and see below).

Anergy is a well-documented mechanism of T cell tolerance and refers to a long-lasting state of unresponsiveness to TCR stimulation without the acquisition of regulatory functions. Foxp3⁺ regulatory T cells (Tregs) have also sometimes been termed “anergic,” which may merely reflect suboptimal stimulatory conditions that can be overcome by mature DC (18, 19). Initial in vitro experiments suggested that providing only Ag-specific signal 1 via MHC II → TCR in the absence of costimulatory signal 2 via CD80/CD86 → CD28 induces T cell anergy, a phenomenon that can be mimicked using anti-CD3 Abs (20, 21). Consequently, MHC II^low CD80/CD86^neg/low immature DC have been shown to induce T cell anergy in vitro (22, 23). In the presence of TGF-β, induction of Foxp3⁺ Treg rather than anergy will occur (24). Restimulation of murine anergic T cells in vitro by immature DC converted them into IL-10⁺ CD25⁺ Foxp3^– Tr1 (25).

In contrast, the role of DC in CD4⁺ T cell anergy induction in vivo is unclear. i.v. injections of high doses of peptide Ag (26) or single or repeated low doses of DEC205 Ab-targeted Ag (27) can induce T cell anergy in the spleen (SP) or LN. However, continuous application of low doses such as applied via osmotic minipumps (28) or i.p. injection of DEC-205 Ab-targeted Ag (29, 30) induced Foxp3⁺ Treg. Also, in healthy unmanipulated mice, self-antigen–specific anergic T cells have been identified, which, however, depended on the presence of Treg (31, 32). Thus, it still needs to be clarified whether these examples of self-antigen–specific anergic T cells at steady state reflect an anergy-like state that is induced via extrinsic Ag application or extrinsic suppression by Treg rather than intrinsic anergy induction by suboptimal endogenous Ag presentation. Moreover, in all of these cases, it remains to be elucidated whether DC or other APCs are involved.

In murine LN, resDCs can acquire s.c. injected Ags from the conduit system and process these Ags in acidic compartments, although anergy induction was not tested (33). Similarly, the murine SP contains primarily resDC in the steady state that almost exclusively exhibit immature surface profiles (see also below) (34) (Fig. 1). However, so far, there is no clear evidence for induction of CD4 T cell anergy by resDC in LN or SP presenting self-antigens under unmanipulated steady-state conditions. Although splenic immature resDC do express low levels of MHC II molecules, these may be predominantly empty (i.e., not loaded with any peptides during the steady state) in contrast to B cells and macrophages (35). Similarly, empty MHC molecules have been identified for human HLA-DR1 and various MHC I molecules (36). Moreover, the turnover of surface peptide–MHC II complexes on immature DC, promoted by the E3 ubiquitin ligase MARCH1, is rapid with a half-life of only 4 h before they are internalized and degraded (37). Under these conditions of empty or rapidly recycling MHC II molecules on immature DC, self-antigen presentation for tolerance induction cannot occur or is very limited. Thus, the classical way of Ag uptake, processing, and presentation may not be an effective tolerance mechanism of immature resDC, although suggested earlier (38, 39). DC activation abrogates MARCH1 activity and stabilizes peptide–MHC II complexes on their surface (37). An alternative role for resDC in peripheral tolerance induction is suggested by experimental evidence from LN in which migDC transfer their self–peptide-MHC II complexes to resDC (40). It remains elusive whether this process is accompanied by DC activation to stabilize and, thereby, prolong tolerogenic Ag presentation (see also below and Fig. 1). At this point, it is tempting to speculate whether other ligands for MHC molecules than the TCR, CD4, and CD8 contribute to immature DC tolerogenicity, as indicated by tolerogenic T cell signals delivered via LAG3 (41).

Under pathophysiological conditions, anergic T cells have been found in mice after certain chronic infections such as with Helicobacter pylori (42) or Staphylococcus aureus (43). For the latter, chronic release of Staphylococcus enterotoxin B is believed to be the responsible mechanism because its repetitive injection induced CD4⁺ T cell anergy by binding to Vβ8⁺ TCR–expressing T cells (44, 45). The anatomical site and the type of APC responsible for anergy induction in vivo are not well investigated but splenic macrophages using PD-L1– or CTLA-4 dependent mechanisms have been identified (42, 46). In humans, Staphylococcus enterotoxin B was found to bind to Vβ3, 14, and 17 in patients suffering from staphylococcal toxic shock (47). These data suggest that T cell anergy may be a mechanism used by macrophages to limit immunopathology during chronic infections, rather than a mechanism of preserving self-tolerance by DC. Hence, the role of immature DC in peripheral tolerance induction remains elusive (Fig. 2).

Within the thymus, high-affinity self-reactive thymocytes at high Ag doses are deleted, whereas intermediate TCR affinity CD4⁺ T cells or cells recognizing only low peptide levels are developing into so-called natural or thymic CD4⁺ Foxp3⁺ Treg (tTreg) (48–51). Whereas the majority of Treg in the immune system are thymus-derived tTreg, naive CD4⁺ T cells outside the thymus can develop into peripheral Treg (pTreg) as well, and these pTreg function similarly to tTreg to prevent T cell–mediated tissue damage in vivo (52). Under steady-state conditions, central tolerance represents a major T cell tolerance mechanism as indicated in mice in which neonatal thymectomy resulted in autoimmunity (53, 54). Although central tolerance is prone to be imperfect to allow the generation of a broad circulating T cell repertoire (38, 48), the impact of peripheral tolerance for self-antigen tolerance is less clear. Whereas the loss of tol-migDC and thereby peripheral self-antigen transport in CCR7^–/– mice did not lead to obvious autoimmune symptoms (55, 56), the deletion of IKKβ in CD11c⁺ cells, which specifically affects tol-migDC, resulted in spontaneous autoimmunity (57).

Thymic DCs are composed of three major subtypes, including resident cDC1, but also migratory cDC2 and pDC subsets as revealed from parabiotic mice (48, 58). cDC1 are the predominant subset in the thymus and originate from intrathymic bone marrow (BM)–derived precursors (58) and are therefore considered as resDC. Thymic cDC1 are mainly located in the medulla, and because clonal deletion occurs mainly in this medulla, it was suggested that thymic cDC1 are involved in the negative selection of thymocytes (59). Indeed, cDC1 regulate clonal deletion; however, it turned out that predominantly cortical cDC1 mediate this process (60). In contrast to thymic cDC1, cDC2 continually migrate as mature cells from peripheral tissues into the thymus (58). In mice, these migDC retain their endocytic capacity and locate inside perivascular regions and around small vessels in the thymus, where they capture blood-borne Ag (61, 62) or transport peripheral Ags to the thymus (63).

The existence of distinct thymic DCs suggests functional specialization of the individual subtypes in the induction and regulation of central tolerance. However, so far, no conclusive data are available identifying a particular thymic DC subset that is explicitly involved in either CD4⁺ Foxp3⁺ tTreg induction or clonal deletion. In mice, migratory cDC (predominantly cDC2) that enter the thymus from the periphery are involved in the deletion of CD4⁺ T cells (63, 64), although these studies may not provide evidence of the physiological role of thymic migDC in negative selection. Intrathymic cross-presentation has been proposed to be essential for CD8⁺ T cell tolerance induction (65), but also in this review, the relative contribution of the individual thymic DC subsets is unclear. Among thymic DC, cDC1 are particularly efficient in cross-presentation, suggesting that cDC1 rather than cDC2 do play a major role in the negative selection of CD8⁺ T cells (66).

tTreg, in contrast, are most efficiently induced by thymic cDC2 (64). Consistent with their migratory phenotype, steady-state thymic cDC2 upregulate MHC II and costimulatory molecules including CD80 and CD86 and are more mature than other thymic DC subsets (58, 66). This semimature state of thymic cDC2 could explain their preferential tTreg instruction capacity. Other studies, however, do not exclude a role of thymic cDC1 in tTreg induction, as the reduced numbers of tTreg in XCL1-deficient mice indirectly suggest that thymic (XCR1⁺) cDC1 are competent to induce tTreg in vivo (67).

In conclusion, mostly murine data indicate that negative selection of both CD4⁺ and CD8⁺ T cells relies on direct Ag presentation by thymic medullary epithelial cells and their Ag transfer to cross-presenting res-cDC1, which mediates CD8⁺ T cell deletion, whereas thymic mig-DC (mig-cDC2 or pDC) preferentially capture thymic and peripheral Ags for MHC II processing and presentation that primarily mediates the deletion of CD4⁺ T cells (48, 68–70).

One of the pathways promoting a “tolerogenic activation” phenotype in DC involves β-catenin, the central component and transcriptional coactivator of canonical Wnt signaling (71). Originally, β-catenin was found to be activated in murine BM DCs by mechanical disruption of E-cadherin/β-catenin cell-cell contacts, which resulted in a sudden release and nuclear localization of β-catenin (72). This activation of β-catenin signaling promoted a tolerogenic BM DC phenotype characterized by the upregulation of surface markers (MHC II, costimulatory molecules, and CCR7) without the production of proinflammatory cytokines, leading to the induction of IL-10–producing Treg in vitro. In agreement, mice with a DC-specific deletion of β-catenin exhibit increased Th1/Th17 and reduced Foxp3⁺ Treg responses and exacerbated intestinal and CNS but not joint inflammation (73–75). The latter indicates that the capacity of β-catenin to regulate the tolerogenic or immunogenic function of DC may be context and/or organ specific. In contrast, although the ability of E-cadherin downregulation and stabilized β-catenin to convey, respectively, a tolerogenic Langerhans cell (LC) and DC phenotype in vivo remains elusive (76, 77), β-catenin has been shown to directly activate a number of pathways that stimulate the development of Treg. These include the induction of the immunoregulatory cytokine IL-10, the vitamin A metabolite retinoic acid (RA), IDO, and vascular endothelial growth factor (78–82).

Steady-state migration of DC into the LN critically depends on their CCR7 expression (56). However, the signals that induce steady-state DC migration are not fully resolved. Clearly, E-cadherin/β-catenin signaling and uptake of apoptotic cells as a source of self-antigen induces upregulation of CCR7 (72, 83) and thus enables DC migration and Treg induction. Oral tolerance is mediated by migDC in mice (84). Intracellular signaling via MyD88 partially promotes tol-migDC traveling to mesenteric LN but independently of microbial recognition or TNF signals (85), indicating that commensal or inflammatory signals do not contribute to this intestinal LN DC migration. Recently, self-antigen–specific activation of skin-resident T cells has been reported to drive steady-state migration of cDC1 to the LN (86). Although CD40-CD40L (CD154)–dependent interactions between DCs and T cells support DC migration under inflammatory conditions (87), it remains to be investigated whether this crosstalk also mobilizes tol-migDC, potentially by a modified CD40 signal (88).

The expression of MHC II, CD80, CD86, and CCR7 molecules is genetically linked via the IRF4 and IRF8 signaling pathways in murine BM DCs and possibly other DC subsets (89). Therefore, CCR7 upregulation to guide DC migration into the LN cannot occur without concomitant upregulation of MHC II and costimulatory molecules. Of note, these phenotypically mature tol-migDC expressing elevated levels of costimulation do not induce immunity, indicating that even substantial amounts of both signal 1 and 2 do not preclude from T cell tolerance induction. On the contrary, this indicates that MHC plus costimulation are required for both, but additional molecules such as cytokines or other secreted products may be decisive to determine the tolerogenicity or immunogenicity of DC.

Some activation markers differ quantitatively between tol-migDC and inf-migDC (Fig. 1). Their differential production of tolerogenic cytokines, as opposed to proinflammatory cytokines, represents qualitative differences linked to the de novo induction of the tolerogenic transcription factor RelB or the immunogenic transcription factors NF-κBp50 and c-Rel (Fig. 1) (14, 90).

In steady-state LN tol-migDC arrive at a mature state as compared with resDC with respect to their surface MHC II, CD86, and CD40 expression. In contrast, tol-migDC exhibit a less mature phenotype when compared with inf-migDC (91). Therefore, the tol-migDC surface phenotype was termed “semimature” (14), indicating quantitative differences for these markers between tol-migDC and inf-migDC (Fig. 1). Thus, the quantitative terms “semimature” or “phenotypically mature” may still be useful, although more recent data indicate that distinct transcription factor and cytokine profiles determine the tolerogenic or immunogenic function of migDC, suggesting the use of a qualitative term like “tolerogenic activation” for such DC (Fig. 1).

In the steady-state apoptotic cells represent a suitable source of self-antigen for tolerance induction. The basically observed response of DC after recognition of apoptotic cells is the release of tolerogenic factors such as TGF-β, IDO, IL-10, and RA (83) and the cDC1 subset seems to be specialized in this (92). Instead, the cytokine profile of murine inf-migDC includes proinflammatory and T-helper cell–polarizing cytokines such as TNF, IL-6, IL-1β, IL-12p70, and IL-23 (93, 94). Thus, the release of cytokines and other soluble mediators represents a qualitative measure to distinguish tolerogenic and immunogenic DC activation (Fig. 1).

The tol-migDC of skin draining LN do not only express intermediate levels of surface MHC II, CD86, or CD40 but also induce RelB and IL-12p40 in the absence of NF-κB (RelA/p50) and c-Rel (57, 90, 91, 95). Of note, RelB and IL-12p40 are not exclusive markers for tol-migDC because they are further upregulated in inf-migDC (Fig. 1) contributing to their proinflammatory functions. In line, RelB deficiency in DC protects from autoimmunity (96), whereas a tolerogenic function of IL-12p40 may rely on antagonizing IL-12p70 for Th1 induction (97, 98). Thus, RelB and IL-12p40 show quantitative differences between tol-migDC and inf-migDC (Fig. 1).

When the genetic signature of the XCR1⁺ cDC1 subset in LN was analyzed, inf-mig-cDC1 and tol-mig-cDC1 appeared both as Relb⁺ and IL12b(IL-12p40)⁺ mature DC at the mRNA level but with a distinct cytokine signature. Although inf-mig-cDC1 exhibit a proinflammatory cytokine profile, the tol-mig-cDC1 displayed a TGF-β signature (90). In mice, tol-migDC require αVβ8 integrin (99) and its release from latency-associated peptide (LAP) binding (91) to convert latent TGF-β into its active form and to enable the conversion of naive T cells into Foxp3⁺ pTreg. Recent evidence suggests that αVβ8 integrin is also required to activate and present TGF-β to naive CD8⁺ T cells for their development into tissue-resident memory T cells (100). In addition, IL-10 secretion by DC is associated with their tolerogenic function, which has been reviewed before (101). Of note, so far IL-10 did not appear to be upregulated in mRNA profiles of murine and human steady-state DC (90, 102, 103), which requires further investigation of its role in the steady state and regulation by posttranscriptional mechanisms. Together, the reciprocal release of pro- and anti-inflammatory cytokines or the bystander activation of resDC by inf-migDC to release proinflammatory cytokines may be the decisive qualitative markers to distinguish tolerogenic from immunogenic DC function (Fig. 1).

One of the mechanisms facilitating Treg induction in vivo is the production of the immunomodulatory enzyme IDO by DC and the subsequent tryptophan catabolism, which induces the proliferation of Treg and steers DC into a tolerogenic state with sustained TGF-β production. IDO expression can be induced by aryl hydrocarbon receptor signaling, which by itself also results in the induction of noninflammatory DC characterized by increased TGF-β and IL-10 but decreased proinflammatory cytokine secretion (104). IDO is mainly produced by intestinal CD103⁺ cDC1 (105); however, in orally tolerized mice, the main IDO-expressing DC subset in the SP are CD11b⁺ cDC2 (106).

CD8⁺ OT-I deletion by tol-migDC has been demonstrated in K5-mOVA mice, but the responsible DC type had not been identified (107). Ab-targeting studies revealed that LCs may be responsible for this mechanism of T cell tolerance (Fig. 2) (108).

Although all tol-migDC subsets have been described to induce Treg by measuring Treg frequencies, the precise mechanisms of induction and the Treg subsets have not been addressed in depth.

Earlier work clearly contributed to understanding tolDC biology, but most reports were either not specifically addressing the role of DC activation states (i.e., whether resDC or migDC were involved or whether thymic or peripheral lymphoid organ [LN/SP] DC were responsible for the observed effects). Using the common CD11c promoter for steady-state expression of neo–self-antigens led to CD8⁺ T cell tolerance (109) and, conversely, steady-state CD11c⁺ DC deletion to peripheral induction of autoaggressive CD4⁺ T cells (110). Others demonstrated a tolerogenic role of the cDC1 subset leading to CD8⁺ T cell hyporesponsiveness and deletion (111), without distinguishing resident from migratory cells. Altogether, these studies do not allow to define DC subsets nor their activation/maturation or migratory state.

Accumulating evidence suggests a context-dependent and/or organ-specific functional specialization of tol-migDC subsets. In the absence of TGF-β and the optional presence of IL-10 the conversion of naive CD4⁺ T cells into CD4⁺ IL-10⁺ Foxp3^– Tr1 cells has been thoroughly demonstrated (101). For example, intranasally applied OVA-induced IL-10⁺ Tr1 cells via OVA transport by tol-mig-cDC1 to the lung draining LN (17). In contrast, the conversion of naive CD4⁺ T cells into Foxp3⁺ pTreg (also called induced Treg) represents a major mechanism of peripheral tolerance induction (112). Transgenic OVA expression has been widely used to study T cell tolerance to this neo–self-antigen but mostly without addressing the role of DC subsets in the tolerization process (113). Using K5-mOVA–transgenic mice as a model of neo–self-antigen expression in the skin, we could establish that naive T cells are converted to Foxp3⁺ pTreg by dermal RelB⁺ CD103⁺ LAP⁺ tol-mig-cDC1s (91). Similar results were obtained by injection of cDC1-specific anti-CD205/myelin oligodendrocyte glycoprotein–self-antigen fusion proteins that led to induction of Foxp3⁺ pTreg in peripheral LN of mice. This was abrogated in Ccr7^–/– mice indicating that only migDC are able to mediate pTreg conversion. The use of langerin/CD207 targeting against the same Ag in irradiated mouse chimeras established that both migratory cDC1 and LC can mediate pTreg generation in skin draining LN (29, 30).

Finally, beyond de novo conversion of naive T cells into pTreg, tol-migDC can also restimulate thymus-generated CD4⁺ Foxp3⁺ tTreg. The analysis of tol-migDC subsets in skin draining LN of CD11c-specific RelB knockout mice indicated that tol-mig-cDC2 are unable to generate pTreg but instead restimulated tTreg with self-antigen and help from self-antigen–specific low-affinity TCR memory-like CD4⁺ IL-2⁺ T cells (114). These data are conflicting with those from others showing in vitro that tol-mig-cDC2 express RA and can induce Treg (115), but these in vitro findings may not reflect the in vivo function of RA in cDC2. These examples illustrate that the type of DC involved in Treg induction strongly depends on the type of Ag, tissue, and microenvironmental cues (Fig. 2).

Interestingly, there is increasing evidence that DC-DC cooperation may occur between migDC transporting peripheral Ags or whole peptide/MHC II complexes to LN and transferring them to resDC for indirect Ag presentation (40). Recently, we found that s.c. injected LPS-matured BM DC converted res-cDC1 in the LN for Th1 induction. This conversion includes a downregulation of transcripts for MHC II, TGF-β, and autophagy, which would argue for a loss of tolerogenic Ag presentation (95). It is, however, still unclear whether the recipient resDC respond to this tolerogenic Ag transfer by at least some degree of activation (semimaturation) or remain at a bona-fide immature state. Clarification of this question will substantially further our understanding of resident versus migratory tolDC subset function.

Monocyte-derived DC do not appear as Ag-transporting cells in steady-state LN because they are mainly induced under inflammatory conditions (116, 117). This may relate to findings that monocytes preferentially differentiate into monocyte-derived DC for migration into peripheral tissues, for example the skin, where they replenish emigrating LCs and DC and contribute to inflammation (118), and not necessarily LN (119).

In vitro data revealed that murine DC exposed to inflammatory mediators or infection do not only upregulate MHC and costimulatory molecules but also release numerous proinflammatory (TNF, IL-6, IL-1β) and CD4⁺ T cell–polarizing cytokines (IL-12p70, IL-23) (16). In contrast, in vivo evidence, that inf-migDC arriving in the LN actually produce these cytokines is lacking. In fact, for vaccine BM DC it has been shown that they not only transfer Ag to the res-cDC1 subset in the LN but also stimulate their release of IL-12p70 required for Th1 polarization, thus converting them into mature bystander DC (95). Cooperation for tolerance induction between migDC and resDC subsets has been discussed before (40), but whether tol-migDC can induce also tolerogenic cytokine production in bystander resDC remains to be investigated.

Cross-presentation under steady-state conditions by XCR1⁺ cDC1, and specifically the Langerin⁺ subset located in the splenic marginal zone of mice, is critical of the maintenance of self-tolerance (120). Whereas splenic Langerin^– cDC1 are rather immature regarding their low levels of MHC II and costimulatory molecules, these Langerin⁺ cDC1, that make up to 40% among cDC1, exhibit higher steady-state expression of the activation markers including CD80 and CD86 (120, 121). Langerin⁺ cDC1 express the dead-cell receptor CD36 and preferentially phagocytose circulating apoptotic cells. Dead-cell phagocytosis results in the upregulation of CD80 and 4-1BBL on Langerin⁺ cDC1, followed by migration into the T cell areas within the white pulp (120). The mechanisms by which Langerin⁺ cDC1 initiate tolerance remains unknown. In contrast to Langerin⁺ tol-migDC in the LN, splenic Langerin⁺ tol-mig-cDC1 are inefficient in the induction of pTreg in vivo (30).

Ex vivo isolated SP cDC1 have been shown to convert naive CD4⁺ T cells into Treg with the help of endogenous TGF-β (29), a phenomenon that has also been observed in vivo and which was potentiated by BTLA expression on these cells (29, 122). However, isolation of splenic DC without inducing their maturation/activation is not possible without adding suppressive reagents. Analyzing pooled SP and LN cells (thus not distinguishing between resDC and migDC) after injection of OVA-Ab fusion proteins demonstrated that anti–DEC205-OVA targeting to cDC1 leads to de novo conversion of Foxp3⁺ Treg (123, 124). In contrast, anti–DCIR2-OVA targeting of cDC2 expanded pre-existing Foxp3⁺ Treg (29), similar to what we found in LN for the respective migDC subsets by using mice expressing a transgenic neo–self-antigen in the skin (91, 114). Ag targeting of DCIR-2 or Treml4 molecules expressed on resDC (and macrophages) but not migDC could not suppress EAE to a higher extent as nontargeted Ag injection (30). Thus, Ag presentation by the resDC is not excluded by injection of DCIR-2– or Treml4-targeted or nontargeted Ag, whereas the data do not provide evidence in favor of this or whether macrophages are involved. Together, it remains unclear whether Treg conversion can result from immature DC Ag presentation in the SP.

In the steady state, the majority of splenic cDC1 and cDC2 are located in the marginal zone and bridging channels, respectively (120, 125–127), indicating that these DC need to relocate to the T cell zones within the white pulp to initiate immune responses or to maintain self-tolerance. In contrast to LN homing, the role of CCR7 for steady-state DC migration within the SP is still debated, although several reports support the importance of CCR7-directed intrasplenic DC migration in response to various stimuli (127, 128). Thus, also the SP contains a fraction of cDC1 that appears to fit into the tol-migDC concept that we propose in this review.

In this review, we questioned several generally accepted features of tolDC under steady-state conditions. The role of immature DC for anergy induction in vivo is unclear. Instead, all tolDC subsets, even in the SP, appear as CCR7⁺ migratory cells (tol-migDC) that have undergone some degree of phenotypic “maturation,” or as a better term, “activation” as used for all other immune cells. The level of activation markers on the surface of tol-mig-DC is lower than on inflammation- or infection-exposed migDC (inf-migDC), suggesting their use only as quantitative markers for discrimination. A qualitative distinction between tol-migDC and inf-migDC can be made by specific transcription factors and secreted products (Fig. 1). Individual tol-migDC subsets use different mechanisms to induce, convert, or restimulate Treg/Tr1 cells, in some cases requiring help from other T cells (Fig. 2) and potentially also bystander resDC.

We thank all laboratory members for support and advice. R.A.B. and B.E.C. are members of the Research Center for Immunotherapy (Forschungszentrum Immuntherapie) Mainz.

The authors have no financial conflicts of interest.