Langerhans cells are precommitted to immune tolerance induction

Abstract

Antigen-dependent interactions between T lymphocytes and dendritic cells (DCs) can produce two distinct outcomes: tolerance and immunity. It is generally considered that all DC subsets are capable of supporting both tolerogenic and immunogenic responses, depending on their exposure to activating signals. Here, we tested whether epidermal Langerhans cells (LCs) can support immunogenic responses in vivo in the absence of antigen presentation by other DC subsets. CD4 T cells responding to antigen presentation by activated LCs initially proliferated but then failed to differentiate into effector/memory cells or to survive long term. The tolerogenic function of LCs was maintained after exposure to potent adjuvants and occurred despite up-regulation of the costimulatory molecules CD80, CD86, and IL-12, but was consistent with their failure to translocate the NF-κB family member RelB from the cytoplasm to the nucleus. Commitment of LCs to tolerogenic function may explain why commensal microorganisms expressing Toll-like receptor (TLR) ligands but confined to the skin epithelium are tolerated, whereas invading pathogens that breach the epithelial basement membrane and activate dermal DCs stimulate a strong immune response.

Dendritic cells (DCs) initiate adaptive immune responses by priming antigen-specific T cells in secondary lymphoid organs. After sampling antigens in peripheral tissues, DCs migrate to lymph nodes (LN), where they present antigenic peptides bound to major histocompatibility (MHC) molecules (1). Epidermal Langerhans cells (LCs) have long been regarded as prototypic DCs, highly active in antigen uptake and rapidly acquiring potent costimulatory capacity after in vitro culture (2). Recently, the immunogenicity of LCs has been questioned on the basis of findings in several in vivo experimental models. During herpes viral infection of the skin, migrated LCs isolated from draining LN (dLN) were unable to induce proliferation of virus-specific CD8 T cells in vitro (3). In LC ablation models, positive (4, 5), negative (6–8), and redundant (9) contributions of LCs to contact hypersensitivity responses were reported. The current lack of consensus regarding LC function may relate, at least in part, to the difficulties in determining the contribution of a relatively small number of LCs to responses driven primarily by non-LC DC subsets in cutaneous LN (cLN).

Here we directly tested the in vivo function of LCs, using a previously described bone marrow (BM) chimeric mouse model in which only LCs can present specific antigen to CD4 T cells (10). In this model, all DC subsets express MHC class II IA molecules but only LCs express MHC class II IE, which is absolutely required to present moth cytochrome C peptide (pMCC) to 5C.C7 T-cell receptor (TCR) transgenic T cells (11, 12). The response of adoptively transferred 5C.C7 CD4 T cells can thus be used as a readout for LC function. We compared 5C.C7 T-cell responses to LCs with those in chimeras expressing IE on nonepidermal DCs or all DC subsets, immunizing with peptide or protein antigens delivered via multiple routes and with diverse adjuvants. Our results show that LCs displayed tolerogenic function under all conditions examined and maintained a tolerogenic NF-κB signature by failing to translocate RelB to the nucleus (13) even when highly activated.

Results

Restriction of MHCII-IE Expression to LCs.

BM chimeras in which IE expression is confined to LCs have been extensively characterized previously (10). The chimeras were engineered using two lines of IEα^d-transgenic mice on the C57BL/6 (MHCII-IA⁺IE⁻) background: 107-1 (here termed IE⁺), expressing IE with WT distribution, and 36-2 (here termed IE⁻), expressing IE only on thymic epithelium and thereby mediating IE-dependent positive selection and Treg development, as well as tolerance to IE (14). Unlike other DC subsets, LCs are radioresistant (15), such that in IE⁻→IE⁺ chimeras (here termed LC chimeras), only skin LCs and migratory LCs (m-LCs) in cLN expressed IE, whereas the remaining DCs, B cells, and radioresistant stromal cells were IE-negative (Fig. 1 A and B and Fig. S1 A and B) (10). We confirmed that migratory dermal DCs (m-DDCs), conventional DCs (cDCs), B cells, and stromal cells from LC chimeras could not present the MCC_87–103 epitope to IE-restricted 5C.C7 T cells using in vitro stimulation with hen egg lysozyme-moth cytochrome C protein (HELMCC; a protein antigen containing the MCC_87–103 epitope) (10) (Fig. S1C).

Fig. 1.

Characterization of chimeric mouse models with expression of IE restricted to either LCs or nonepidermal DCs. (A) Schematic representation of LC chimeras. (B) Expression of IE in the skin and skin-draining LNs of chimeric mice. Representative flow cytometric plots are gated to show the frequency of IE⁺ DCs as a percentage of total DCs. (C) Schematic representation of control chimeras.

As controls for the IE⁻→IE⁺ LC chimeras, we generated IE⁺→IE⁻ chimeras (Fig. 1C) (10). Because all B cells in LC chimeras were IE⁻, the IE⁺ BM for control chimeras was obtained from RAG^−/− donors and was mixed with autologous IE⁻ BM to generate an equivalent IE⁻ B-cell compartment. The proportion of IE⁺ RAG^−/− BM was adjusted to 25% so that frequency of skin-derived IE⁺ migratory DCs (m-DCs) (Fig. 1B) was matched in cLN of LC and 25% control chimeras, to control for potential differences in cognate MHCII exposure and peptide presentation. Equivalent IE-restricted peptide presentation was confirmed by measuring recruitment of 5C.C7 T cells into division (Fig. S2 A and B).

5C.C7 T cells survived long term in both LC and control chimeras but rapidly disappeared in IE⁻ mice (Fig. S2C), confirming that expression of IE by m-LCs in cLN was sufficient for survival of naïve IE-restricted T cells. Intravenous injection of pMCC induced equivalent rates of deletion in both chimeras (Fig. S2D), excluding long-term radiation effects as a possible cause of differential responses in the two chimeras (16).

CD4 T Cells Activated by m-LCs Fail to Differentiate into Effector/Memory Cells and Do Not Survive Long Term.

We tested the ability of LCs to sustain an immunogenic CD4 T-cell response in vivo by transferring carboxyfluorescein diacetate succinimidyl ester (CFSE)-labeled 5C.C7 cells into LC or control chimeras and immunizing s.c. with pMCC emulsified in complete Freund’s adjuvant (CFA) (Fig. 2). Despite similar recruitment of T cells into division (Fig. S2B), an 8.6-fold higher peak in T-cell numbers was observed in control compared with LC chimeras, and cells survived long term only in control chimeras (Fig. 2 A and B). By day 10, most donor T cells in control but not LC chimeras had acquired a CD44^highCD62L^low effector memory (Tem) phenotype (17) (Fig. 2C). To test for effector function, cytokine production by 5C.C7 cells was measured following in vitro restimulation with peptide plus IE⁺ splenic DCs. Abundant production of interleukin (IL)-17 and IFNγ was seen in control but not LC chimeras (Fig. 2D). The difference in peak effector numbers was 240-fold for IFNγ and 25-fold for IL-17 (Fig. 2E). LC chimeras retained some IL-2-producing CD4 T cells (5.5-fold decreased), accounting for their initial proliferative response. IL-4, IL-5, and IL-10 were never detected, nor was induction of foxp3 expression. Similar results were obtained using HELMCC protein in CFA as the immunogen, except that cytokine production during the effector phase was biased toward IL-17 rather than IFNγ (Fig. S3 A–D).

Fig. 2.

In vivo response of naïve CD4⁺ T cells to antigen presented by m-LCs or nonepidermal DCs. Fully reconstituted (>3 mo) LC or control chimeras were adoptively transferred with 2 × 10⁵ CFSE-labeled 5C.C7 T cells and s.c. immunized with 10 μg MCC peptide/CFA. (A) Response of donor 5C.C7 T cells in dLNs. Representative flow cytometric plots of CD4 T cells are gated to show the frequency of donor 5C.C7 cells (Left) and their CFSE-dilution profiles (Right). (B) Absolute number of donor 5C.C7 T cells in dLN and spleen. Data are from one representative experiment out of three (3–5 animals per group). (C) Acquisition of effector memory phenotype by donor 5C.C7 T cells. Donor 5C.C7 T cells in dLN were gated for undivided CFSE^high 5C.C7 cells (gate I) and fully divided CD62L⁻CFSE⁻ 5C.C7 cells (gate II). (Right) Mean absolute numbers of cells within the two gates. Values are for one representative experiment. (D and E) Representative flow cytometric plots (D) and absolute number (E) of cytokine-producing donor 5C.C7 cells in dLN. Numbers in D indicate the frequency of cells in each of the four quadrants. (F and G) Antigen presentation by LCs does not support differentiation of CD4⁺ memory cells. Memory response to day 60 s.c. challenge with peptide/IFA (F) or to day 80 intradermal challenge with peptide-pulsed IE⁺ splenic DCs (G). Draining LNs were collected 16 h (F) or 3 d after challenge (G). (F Left) Frequency of 5C.C7 cells expressed as a percentage of total CD4 T cells in unchallenged versus challenged mice. (F Center) Expression of CD44 and CD62L. (F Right) Cytokine production after challenge. (G) Frequency of donor 5C.C7 cells (expressed as percent of total CD4 T cells) in draining LNs and ear skin after intradermal ear challenge. One representative experiment out of three is shown.

The difference between T-cell survival in LC and control chimeras could not be explained by the disappearance of IE⁺ DCs in LC chimeras, because the number of IE⁺ m-LCs and m-DDCs in dLN of LC and control chimeras, respectively, were similar over the course of the response (Fig. S4), consistent with local maintenance of LC homeostasis (15).

To test for antigen-specific memory 60 d (Fig. 2F) and 90 d (Fig. S5A) post immunization, chimeras were challenged at a different site with peptide emulsified in incomplete Freund’s adjuvant (IFA). Sixteen hours after challenge, the frequency of 5C.C7 cells in dLN of control chimeras increased by 15-fold (Fig. 2F, Left and Fig. S5A, Left). The rapid increase in 5C.C7 cell numbers was largely due to redistribution to dLN (Fig. S5D). Responding cells underwent blast transformation (Fig. S5C) and down-regulated CD62L (Fig. 2F, Center and Fig. S5A, Center). Donor 5C.C7 cells produced IFNγ after in vitro restimulation (Fig. 2F, Right and Fig. S5A, Right). These responses were not seen in LC chimeras. Interestingly, 5C.C7 cells in dLN of both chimeras produced IL-2, indicating that the surviving cells in LC chimeras were still capable of responding to TCR stimulation. Similar results were obtained after in vivo challenge of mice primed with HELMCC protein rather than pMCC (Fig. S3E).

It remained possible that CD4 T cells in LC chimeras failed to mount a memory response to rechallenge because of a specific defect in m-LC antigen presentation. We therefore challenged primed chimeras with intradermal injection of antigen-pulsed IE⁺ splenic DCs. Three days after DC injection, the frequency of 5C.C7 cells in the dLN of control chimeras had increased fivefold (Fig. 2G) and the cells had become CD62L^low (Fig. S5B, Left). These changes did not occur in LC chimeras. Cytokine production (mainly IL-2) was higher in control compared with LC chimeras (Fig. S5B, Right). Furthermore, donor T cells were found at the site of skin challenge in control but not LC chimeras (Fig. 2G), excluding the possibility that sequestration in the skin could account for the disappearance of antigen-specific T cells in LC chimeras. Taken together, these experiments indicated that LCs exposed to s.c. antigen recruited CD4 T cells into an abortive proliferative response that resulted in tolerance rather than generation of effector/memory function.

Effect of Activation Status on LC Function.

Migrating LCs retained their previously documented CD80/86^low phenotype (10) in response to s.c. immunization, whereas CD80 and CD86 expression on migrating DDCs increased within 4 h and continued to increase until day 4 postimmunization (Fig. S6 A and B). However, LCs expressed more CD40 than DDCs (10) and further up-regulated their CD40 expression from day 2 onward (Fig. S6 A and B). We therefore tested whether ligation of CD40 could convert LCs to an immunogenic phenotype, as had been described previously for other tolerogenic DC subsets (18). LC chimeras were treated with agonistic anti-CD40 antibodies on days 0 and 2 after s.c. immunization. However, even the combined pMCC/CFA/anti-CD40 treatment did not support the generation of CD4 T-cell memory in LC chimeras (Fig. S6C).

Considering that the failure of m-LCs to up-regulate CD80/86 expression after s.c. injection may have indicated inadequate exposure to adjuvant, we switched to an epicutaneous immunization approach in which LCs were directly exposed to protein antigen/adjuvant via topical application in aqueous cream (19). We did not use tape stripping, which may disturb the integrity of the epidermis (20). Even without the addition of adjuvants, application of cream under an occlusive bandage caused m-LCs to up-regulate both CD80 and CD86, producing a bimodal CD80/86 profile as m-LCs first reached the dLN 48 h after immunization (Fig. 3 A and B). The tempo of CD69 up-regulation by antigen-specific T cells showed a 1–2 d delay after the arrival of m-LCs from the immunization site (Fig. 3C), suggesting that the T-cell response was driven by migrating antigen-bearing LCs rather than free antigen presented by m-LCs already present in the LN at the time of immunization. Addition of adjuvants to the epicutaneous cream caused further activation of m-LCs, with 2.2- to 3.1-fold increases in CD80 and CD86 expression in response to CFA-derived particulate material (heat-killed Mycobacterium tuberculosis H37Ra), TLR1/2 ligand Pam3Cys-Ser-(Lys)4 (Pam3CSK), or the TLR3 ligand polyinosinic acid:polycytidylic acid (poly I:C) (Fig. 3D). Epicutaneous immunization with cream containing CFA particulates also induced over 20% of m-LCs in dLN of LC chimeras to express IL-12, generating three- to fourfold more IL-12-producing IE⁺ m-DCs than the same treatment in control chimeras (Fig. 3E). However, despite their activated phenotype and IL-12 production, LCs responding to epicutaneous immunization with a combination of HELMCC and CFA particulates still failed to support the generation of CD4 T-cell memory, as indicated by the lack of response to in vivo challenge with peptide/IFA (Fig. 3F).

Fig. 3.

Epicutaneous immunization activates LCs but does not support development of CD4 memory cells. (A–C) LC or control chimeras were immunized with HELMCC in cream applied onto hairless abdominal skin. Representative flow profiles (A) and kinetics of CD80 and CD86 expression (B) by m-LCs and m-DDCs in dLN (mean of three per group ± SEM). (C) Activation of 5C.C7 T cells (three mice per group). MFI, mean fluorescence intensity. (D) Expression of CD40, CD80, and CD86 on day 4 after immunization with cream containing either CFA particles, imiquimod, Pam3CSK, poly I:C, lipopolysaccharide (LPS), or curdlan, as indicated. (E) Intracellular IL-12p40/p70 expression by IE⁺ m-DDCs (control chimeras) and IE⁺ m-LCs (LC chimeras) in dLN 6 d after epicutaneous immunization with cream containing HELMCC and CFA particles. (Upper) Representative dot plots showing the frequency of IL-12-positive cells (gated) among IE⁺ m-DCs. (Lower) Absolute number of IL-12⁺ m-DDCs (circles) and IL-12⁺ m-LCs (triangles) in dLNs. (F) Frequency of donor 5C.C7 T cells in dLN after memory recall with peptide/IFA on day 90 after epicutaneous immunization of LC chimeras with HELMCC/cream/CFA particulates. (G and H) Response of adoptively transferred 5C.C7 cells in B10.BR mice immunized either epicutaneously for 5 d with HELMCC/cream containing a mixture of CFA particulates, Pam3CSK, poly I:C, imiquimod, and curdlan (triangles) or s.c. with HELMCC/CFA (circles). Absolute numbers of donor 5C.C7 cells (G) and cytokine-producing donor 5C.C7 cells (H) are shown. (I) LCs migrating to draining LNs after immunization fail to translocate the NF-κB subunit RelB to the nucleus. Migratory LCs or DDCs were flow-sorted from dLN of chimeric mice after s.c. (Left) or epicutaneous (Right) immunization, and RelB translocation to the nucleus was analyzed by confocal microscopy. Graphs show the mean percentage (±SEM) of RelB translocations per visual field for 6–8 fields containing >200 DCs per sample.

LC-Driven Responses in WT Mice.

The experiments described above indicated that immunization of LC chimeras rendered them tolerant to specific antigen. To test whether LCs also induced tolerance in unmanipulated animals, we compared responses to epicutaneous and s.c. immunization in WT mice, reasoning that if epicutaneous antigen were presented mainly by LCs, then epicutaneous responses should recapitulate the tolerogenic responses we had documented in LC chimeras. B10.BR mice adoptively transferred with 5C.C7 cells were either immunized s.c. with HELMCC/CFA or epicutaneously with HELMCC in cream containing a mixture of potent adjuvants (Fig. 3 G and H). Fewer 5C.C7 cells were recovered 6 d after epicutaneous immunization (down 5.2-fold in dLN and 3.9-fold in spleen compared with s.c. immunization), and no donor T cells could be recovered by day 70 (Fig. 3G). The number of effector cells was also markedly reduced (down 61-fold for IL-17-, 5.9-fold for IFNγ-, and 7.7-fold for IL-2-producing cells;) (Fig. 3H). In a second experiment comparing epicutaneous immunization of WT hosts versus LC chimeras, the day 7 response of 5C.C7 cells in LC chimeras was over 80% of that in WT hosts, indicating that presentation of free antigen by resident LN DCs in WT mice is unlikely to account for more than a small proportion of the response. Thus, the effect of epicutaneous immunization in WT mice mirrored that seen in LC chimeras, confirming that LCs subserve a tolerogenic function in normal animals.

Activated LCs Fail to Translocate RelB to the Nucleus.

The surprising lack of correlation between costimulatory molecule expression and LC function in vivo led us to test LCs for further correlates of DC tolerogenicity. Activation of the NF-κB transcription factor RelB, as indicated by translocation to the nucleus, is one of the best-established markers of DC immunogenicity in vivo (13, 21). Whereas a proportion of m-DDCs showed clear evidence of nuclear translocation of RelB after skin painting with a contact sensitizer, s.c. immunization with CFA, and epicutaneous immunization with cream/CFA particulates, RelB translocation was never seen in m-LCs (Fig. 3I and Fig. S7). Thus, the activation and nuclear translocation of RelB appeared to be a reliable correlate of DC immunogenicity in vivo.

Visualizing Activation and Migration of LCs.

In addition to differential activation of RelB, LCs and DDCs show consistent differences in their migratory behavior, with migrating LCs slower in reaching dLN than DDCs (9, 10). We used intravital microscopy to monitor the behavior of LCs over the first 4 d of the epicutaneous response (Fig. S8). In the steady state, LCs were sessile (mean velocity <1 μm/min), with their dendrites remaining almost completely immobile as described previously (9, 22). Ninety-six hours after application of cream onto ear skin, LCs appeared as round cells with retracted dendrites, deeply embedded into underlying collagen; these changes were particularly apparent with cream/CFA. However, actual crossing of the basement membrane and entry into dermis were only infrequently observed, consistent with the delayed kinetics of migration.

Migratory LCs Inhibit the Effector Phase of the Immune Response.

Although T-cell activation in LC chimeras correlated with the arrival of m-LCs from the immunization site (Fig. 3), it remained possible that they arrived too late to rescue a default tolerogenic response stimulated by steady-state m-LCs already in the node. To test whether migrating LCs could actively participate in ongoing responses initiated by rapidly migrating m-DDCs, we created combined radiation chimeras in which both LCs and DDCs expressed IE (Fig. 4A). In these chimeras, IE⁺ m-DCs in cLN comprised a 1:1 mixture of m-LCs and m-DDCs, compared with a 1:3 mixture in WT mice (10). The number of donor T cells in the first 10 d post immunization was similar in combined and control chimeras (Fig. 4B), but the number of effector cells in combined chimeras was significantly reduced (down 8.6-fold for IFNγ and 5-fold for IL-17) (Fig. 4C). Memory cell numbers were relatively preserved (Fig. 4B), as was memory function (Fig. 4D). These results indicate that LCs potently regulate the effector phase of the immune response by limiting T-cell effector function when the ratio of m-LCs to m-DDCs is sufficiently high. This limiting of T-cell effector function appears to be a direct LC-mediated effect, because antigen-specific foxp3⁺ regulatory T cells did not emerge at any time post immunization.

Fig. 4.

LCs inhibit CD4 T-cell effector responses initiated and maintained by nonepidermal DC subsets. (A) Schematic representation of combined chimeras. (B–D) Combined, control, and LC chimeras were adoptively transferred with 2 × 10⁵ CFSE-labeled 5C.C7 T cells and s.c. immunized with pMCC/CFA. (B) Absolute number of donor 5C.C7 T cells in dLN. Data are from one representative experiment of two, with three or four animals per group. (C) Representative flow cytometric plots (Left) and absolute number (Right) of cytokine-producing donor 5C.C7 cells 10 d postimmunization. Each symbol represents an individual mouse. (D) Memory response of combined chimeras to intradermal challenge with peptide-loaded IE⁺ DCs 80 d after priming. (Left) Frequency of 5C.C7 T cells in draining LN and skin of challenged versus unchallenged mice. (Right) Expression of CD44 and CD62L.

A second possibility is that early presentation of free antigen by steady-state antigen-presenting m-LCs renders CD4 T cells unable to respond productively to a subsequent exposure to activated m-LCs. To test this, we delayed the transfer of 5C.C7 T cells for 3 d after LC chimera immunization to allow migration of activated m-LCs (Fig. S9). T cells transferred into hosts preimmunized with cream/adjuvant/antigen underwent only low-level CD69 up-regulation and proliferation, suggesting significant competition from the endogenous T-cell response (Fig. S9A). When the hosts were treated with cream/adjuvant but administration of antigen was delayed until the day after 5C.C7 T-cell transfer, significantly more proliferation was observed but no effector cytokines were detected (Fig. S9B). Thus, primary antigen presentation by preactivated m-LCs still failed to drive effector/memory differentiation in naïve CD4 T cells.

Discussion

Precommitment of DC subsets to specialized functions has gained acceptance with the demonstration that the ability to cross-present is restricted to CD8⁺ cDCs and CD103⁺ DCs (23, 24). However, the existence of DC subsets that are precommitted to tolerance induction remains controversial. To test defined DC subsets for tolerogenicity, we have developed a mouse model using transgenic expression of MHCII-IE to target specific antigen presentation to individual DC subsets, enabling direct functional measurement in vivo. Using this approach, we show here that LCs maintain tolerogenic function under a range of conditions that are commonly believed to induce immunogenicity in all DC subsets.

Function of IE⁺ LCs was measured by comparing IE-restricted CD4 T-cell responses under three different conditions: when LCs were the only DC subtype capable of processing and presenting specific antigen (LC chimeras) (Fig. 1A); when all DC subsets with the exception of LCs could present antigen (control chimeras) (Fig. 1C); and when both LCs and non-LC DCs could present antigen (combined chimeras) (Fig. 4A). The full complement of MHCII-IA-expressing DCs was present in all three models, the only differences being due to DC subset-specific expression of the additional MHCII-IE allele required for specific antigen presentation. We chose this approach to avoid the difficulties inherent in interpreting the data from MHCII knockout mice and chimeras, in which adoptively transferred CD4 T cells are rapidly desensitized due to lack of baseline TCR engagement (25, 26).

Our results indicate that naïve CD4 T cells initially proliferated strongly in response to antigen presented by LCs but then gradually disappeared without effector/memory cell differentiation, rendering the animal tolerant to subsequent challenge with specific antigen. This response was independent of whether peptide or protein antigens were used, whether they were delivered subcutaneously or epicutaneously, and whether potent adjuvants including CFA, agonistic anti-CD40 mAb, and TLR ligands were included in the immunization. Thus, LCs appear to possess an inherent commitment to tolerogenic function, even when displaying a CD80/86^high phenotype associated with immunogenicity in other DC subtypes.

Although this finding may be considered surprising in the light of the currently accepted two-signal model of T-cell activation (27), it is consistent with the well-established phenomenon of a strong CD28-dependent proliferative burst preceding i.v. peptide-mediated tolerance induction in vivo (12). Our results indicate that costimulatory molecule expression by DCs may be necessary but not sufficient for immunogenicity in vivo. One of the additional biochemical requirements for immunogenicity is believed to be activation of the NF-κB subunit RelB (13). DCs derived from RelB^−/− mice or RelB^−/− chimeras, or treated with an NF-κB inhibitor (RelB^low DCs), can induce antigen-specific tolerance (13) and suppress inflammatory arthritis (28). Our finding that LCs fail both to activate RelB (Fig. 3I and Fig. S7) and to generate an effector/memory CD4 T-cell response adds support to the notion that RelB may serve as a master regulator of DC function.

The ability of LCs to drive proliferation of naïve 5C.C7 CD4 T cells in vivo is consistent with the potent ability of LCs to drive in vitro responses (2), but differs from published results obtained with OTII CD4 T cells in MHCII knockout chimeras (26). The difference may be due to the relatively low affinity of OTII cells for specific antigen-MHC and/or to the MHCII^−/− DC milieu, which would have led to TCR desensitization via TCRζ chain dephosphorylation (25, 29). Indeed, we established that in vivo T-cell responses proceeded under essentially physiological conditions in our models. Thus, the IE-expressing LCs and m-LCs in the chimeras fully supported survival of naïve T cells (Fig. S2C), which have the most stringent requirements for cognate MHCII contact (30). We also demonstrated identical kinetics of deletional tolerance in fully reconstituted LC and control chimeras (Fig. S2D), to exclude quantitative differences in antigen presentation as a cause of differential cell fate in our chimeras. Stromal effects were excluded by showing that LN stroma could not present IE-restricted antigen to CD4 T cells (Fig. S1C). This is in sharp contrast to recently reported results for CD8 T cells, which can survive by means of contact with MHCI expressed by either hematopoietic or stromal compartments (31), and can be rendered tolerant by specific antigen presented by radioresistant LN stroma (32).

LC-dependent presentation of antigen could potently suppress generation of IL-17- and IFNγ-secreting effector cells in combined chimeras in which the ratio of antigen-presenting m-LCs to m-DDCs was made artificially high to provide an unequivocal result (Fig. 4). In unmanipulated mice in which the ratio of m-LCs to m-DDCs in cLN is 1:3 rather than 1:1, the effect of LCs would be smaller, which may explain the lack of effect in some (9, 33, 34), but not all (7, 8, 35), models of contact sensitivity. The ability of LCs to suppress the response to antigen presentation by other DC subsets argues against the possibility that tolerance in our models is a default response to presentation of free antigen without active involvement of migrated skin DCs.

Understanding the in vivo function of LCs may provide clues as to how DCs can mediate tolerance to TLR-expressing commensal organisms colonizing epithelial surfaces such as skin and bowel, whilst retaining the ability to prime a strong immune response to pathogens. We propose that LCs mediate tolerance to skin commensals under steady-state conditions when the structural integrity of the basement membrane that usually provides an epidermal/dermal barrier is intact. In contrast, invading pathogens that breach the barrier will generate a strong response overwhelmingly mediated by rapidly migrating DDCs, whereas minor disturbances will be subject to a combination of immunogenic DDC signals and LC modulation of effector function but not memory generation.

Finally, our findings provide direct evidence of a DC subset committed to tolerance induction while responding to immunogenic signals and displaying what is currently considered to be an immunogenic surface phenotype. The four recently described skin DC subsets (10) thus include those specialized for negative regulation of CD4 T cells in addition to those specialized for cross-presentation to CD8 T cells (23). On the basis of these findings, we predict that DC subsets precommitted to induction of tolerance or immunity in CD4 T cells will coexist with cross-presenting DCs in many organs, allowing the full range of differential T-cell responses to be generated as CD4 T cells integrate a range of tolerogenic and immunogenic signals from DCs and, in turn, regulate tolerance and immunity within the CD8 T-cell compartment.

Materials and Methods

Mice.

IEα^d transgenic mouse lines 107-1 and 36-2 and 5C.C7 RAG1^−/− TCR transgenic mice are described in ref. 10. CD11c-YFP transgenic mice (36) were obtained from M. Nussenzweig (The Rockefeller University, New York, NY). More details in SI Materials and Methods. Approval for all animal experimentation was obtained from the Animal Ethics Committees at the University of Sydney and the Wistar Institute.

BM Chimeras.

LC chimeras and control chimeras are described in ref. 10. More details in SI Materials and Methods.

Adoptive Transfer of T Cells and Immunizations.

T-cell adoptive transfer and s.c immunization were performed essentially as described in ref. 10. For epicutaneous immunization, 10 μg HELMCC was mixed with adjuvants in 150 mg aqueous cream (Sorbolene; Kenkay) applied onto hairless skin and secured with an occlusive bandage. More details in SI Materials and Methods.

Flow Cytometry.

The analysis and antibodies used are described in detail in SI Materials and Methods.

T-Cell Effector and Memory Assays.

For effector restimulation, lymph node and spleen cell suspensions were cultured with 10 μM pMCC for 10 h (effectors) or 16 h (memory cells) in the presence of magnetically isolated (Miltenyi Biotech) IE⁺ splenic DCs and Brefeldin A. After culture, cells were stained as for flow cytometry, fixed, permeabilized, and stained using anti-IFNγ, anti-IL2, and anti-IL17 antibodies. For memory recall, mice were challenged s.c. into front footpads with 10 μg pMCC in IFA or intradermally into the ear pinna with MCC-pulsed IE⁺ splenic DCs. Culture and staining for cytokine detection were as described for effector cells. More details are available in SI Materials and Methods.

RelB Staining.

Chimeric mice were skin-painted with fluorescein isothiocyanate as described (10) or immunized s.c. or epicutaneously. m-LCs and m-DDCs were isolated from draining LNs by flow sorting, cytospun onto glass slides, fixed, and stained for RelB and nuclear DNA and analyzed by confocal microscopy. Details of sorting and staining procedures are in SI Materials and Methods.

Two-Photon Intravital Microscopy.

Two-photon intravital microscopy of LCs and DDCs was performed on ear skin of anesthetized CD11c-YFP mice. Details of imaging and image analysis are described in SI Materials and Methods.