Variable Requirement of Dendritic Cells for Recruitment of NK and T Cells to Different TLR Agonists1

TLRs initiate the host immune response to microbial pathogens by activating cells of the innate immune system. Dendritic cells (DCs) can be categorized into two major groups, conventional DCs (including CD8+ and CD8− DCs) and plasmacytoid DCs. In mice, these subsets of DCs express a variety of TLRs, with conventional DCs responding in vitro to predominantly TLR3, TLR4, TLR5, and TLR9 ligands, and plasmacytoid DCs responding mainly to TLR7 and TLR9 ligands. However, the in vivo requirement of DCs to initiate immune responses to specific TLR agonists is not fully known. Using mice depleted of >90% of CD11c+ MHC class II+ DCs, we demonstrate that cellular recruitment, including CD4+ T cell and CX5+DX5+ NK cell recruitment to draining lymph nodes following the footpad administration of TLR4 and TLR5 agonists, is dramatically decreased upon reduction of DC numbers, but type I IFN production can partially substitute for DCs in response to TLR3 and TLR7 agonists. Interestingly, TLR ligands can activate T cells and NK cells in the draining lymph nodes, even with reduced DC numbers. The findings reveal considerable plasticity in the response to TLR agonists, with TLR4 and TLR5 agonists sharing the requirement of DCs for subsequent lymph node recruitment of NK and T cells.

D endritic cells (DCs) 4 are APCs present in virtually every tissue, and are potent inducers of innate and adaptive immunity (1). Importantly, various subsets of DCs exist and drive specific immunological responses. Conventional DC subsets, as well as Ly6C ϩ B220 ϩ PDCA-1 ϩ CD11c low plasmacytoid DCs, use a variety of pattern recognition molecules, such as TLRs, to recognize pathogen-associated molecular patterns and distinguish self from nonself (2). These microbial recognition molecules, which include peptidoglycan, dsRNA, LPS, flagellin, ssRNA/nucleoside analogs, and unmethylated CpG DNA are recognized predominantly by TLR2, TLR3, TLR4, TLR5, TLR7, and TLR9, respectively (3). Recognition of TLR ligands by DCs induces maturation and migration to secondary lymphoid tissue where they recruit, interact with, and activate various cell populations including T cells and NK cells. Recently, it has been shown that individual TLR agonists may induce differential responses through activation of distinct signaling cascades, driving Th1 or Th2 responses in CD4 ϩ T cells (4). Even more recently, TLRactivated DCs have been shown to stimulate NK activity, which in turn can produce IFN-␥ and provide an additional Th1 priming signal (5,6). Although much is known about TLR-dependent activation of T cells and NK cells, whether DCs play an obligatory role in this complex process in response to different TLR agonists is still unknown. For example, TLR7 (TLR7 is active in mice, whereas TLR8 is active in humans) and TLR9 ligands can activate plasmacytoid DCs and B cells (7,8), whereas dsRNA can be recognized by various cells including DCs, fibroblasts, macrophages, and endothelial cells (9 -11). All of these TLR agonists induce type I IFN production (12), which can induce chemokine-mediated cellular recruitment and substitute for costimulatory signals for CD4 ϩ T cell activation (13).
In this study, we investigated the in vivo requirement for DCs in TLR-induced cellular recruitment to draining lymph nodes. Using a transgenic mouse model that can be selectively depleted of conventional CD11c ϩ MHC class II ϩ DCs (up to 90%), we show that cellular recruitment, including CX5 ϩ DX5 ϩ NK and CD4 ϩ T cell subsets, to draining lymph nodes of locally administered TLR4 and TLR5 agonists has an absolute requirement for conventional DCs. Interestingly, TLR3, TLR7, and TLR9 agonist-induced cellular recruitment, and activation of T cells and NK cells by any TLR agonist do not have the same absolute requirements for conventional DCs. Blocking type I IFN in TLR3 ligand-treated, DC-depleted mice inhibited the increase in total cellularity and recruitment of NK cells, but not CD4 ϩ T cells, whereas blocking type I IFN in TLR7 ligand treated mice inhibited the increase of all cell types examined. In contrast, blocking type I IFN in TLR9 ligandtreated, DC-depleted mice had no effect on total cellularity and recruitment, demonstrating that redundant and differing pathways exist for activation and recruitment of innate and adaptive immune system components by different TLR agonists.

Mice
Transgenic B6.FVB-Tg .Itgax-DTR/EGFP.57 Lan/J mice backcrossed to a BALB/c background that can be depleted of DCs have been referred to as CD11c-DTR mice (14). Mice were maintained in a breeding colony at the University of Florida College of Medicine Animal Facility under specific, pathogen-free conditions. Mice were genotyped from tail DNA using published primer sequences. Age-and sex-matched littermates between 8 and 12 wk of age not expressing the transgene were used as controls.

Generation of bone marrow-derived DCs
DCs derived from murine bone marrow DCs were generated with murine recombinant GM-CSF (PeproTech), as previously described (17).

Depletion of DCs
To deplete CD11c ϩ DCs, transgenic CD11c-DTR and wild-type littermates were treated with an i.p. injection of 4 ng per gram of body weight of diphtheria toxin (DT; Sigma-Aldrich), as previously described (14,18). At the time of TLR administration (24 h), depletion of DCs and DC subsets in popliteal and inguinal lymph nodes was similar to depletion of splenic DCs observed in previous reports (18) (Fig. 1), with a 96% depletion of CD8 ϩ DCs, over 90% depletion of total DCs and CD8 Ϫ DCs, and ϳ75% depletion of plasmacytoid DCs. For plasmacytoid DCs, only PDCA-1 high B220 high cells were considered, as a PDCA-1 low B220 high cell population appeared following TLR treatment in ipsilateral lymph nodes, indicating increased expression of PDCA-1 on B cells. When examined, all of the PDCA-1 high B220 high cells were also CD11c low in both ipsilateral and contralateral lymph nodes (data not shown).

Footpad injection and lymph node single-cell suspensions
Twenty-four hours after DT administration, mice were anesthetized and injected in the right hind footpad with 1 g of flagellin or LPS, 0.1 g of poly(I:C), 0.1 g of resiquimod (R848), or 10 g of CpG in 50 l of PBS. Doses of TLR agonists were carefully titrated in preliminary studies to induce local immune response in the ipsilateral popliteal lymph node without inducing a response in the contralateral lymph node or spleen ((15) and data not shown). When noted, 10 6 wild-type bone marrow-derived DCs were injected with or without LPS into the footpad, or 1 h before TLR ligand injection, 200 l of type I IFN blocking or control antiserum was injected i.p. or i.v (16). Twenty-four hours later, mice were sacrificed and both popliteal lymph nodes were harvested. The lymph nodes were dissected and processed to a single-cell suspension as previously described (15).
All samples were analyzed on a FACSCalibur (BD Biosciences) and a specialized software package (CellQuest; BD Biosciences). Isotype controls (BD Pharmingen) were used for all analysis. Dead cells were removed from analysis through 7-aminoactinomycin D (BD Pharmingen) staining.

Statistical analysis
Data were analyzed using the statistical software program StatView 5.0 statistical software package (Abacus Concepts) and are reported as the mean Ϯ SEM. Differences were considered significant at p Ͻ 0.05, by either the paired Student's t test or ANOVA.

Injection of TLR agonists activates conventional and plasmacytoid DCs in draining lymph nodes
To explore whether TLR agonists induce DC recruitment to the draining lymph nodes and DC maturation is seen in vivo, we examined the draining and contralateral popliteal lymph nodes, following footpad injection of the TLR agonists. Adjuvants, such as TLR agonists, injected locally are known to induce DC maturation and migration to local lymph nodes where they create an environment conducive to creating immune responsiveness (15,19). We first investigated whether footpad injection of TLR agonists increased the number and maturation state of conventional (CD11c ϩ MHC class II ϩ ) DCs in the ipsilateral popliteal lymph node, compared with the contralateral popliteal lymph node.
We have previously shown that conventional CD8␣ Ϫ DCs and plasmacytoid DCs are recruited to draining lymph nodes following the footpad injection of recombinant flagellin, a potent TLR5 agonist (15). In this study, all of the TLR agents induced a recruitment of both total CD11c ϩ DCs and plasmacytoid (PDCA-1 ϩ , B220 ϩ ) DCs into the draining (ipsilateral) popliteal lymph nodes, and increased their activation state (Fig. 2). Because these TLR agonists do not all directly activate plasmacytoid DCs, these data suggest that some activation occurs secondary to cytokine production by other cells. As shown in Fig. 1, we then used a novel in vivo model to transiently deplete mice of CD11c and MHC class II coexpressing DCs to determine their obligate requirement in the recruitment and activation of other cell type to draining lymph nodes of TLR agonist-treated mice.

DCs are required for maximal increases in lymph node cellularity in response to LPS or flagellin but not poly(I:C), resiquimod (R848), or CpG treatment
Because local administration of adjuvants increases total cellularity in draining lymph nodes, we tested whether DCs regulate the attraction of cells to the draining lymph nodes 24 h after the injection of different TLR agonists. In wild-type mice, injection of any TLR agonist into the footpad caused at least a 2-fold increase in total lymph node cellularity, and specific recruitment of CD4 ϩ T cells and CX5 ϩ DX5 ϩ NK cells in the ipsilateral vs contralateral lymph node ( p Ͻ 0.05) (Fig. 3, left panels). However, in CD11c-DTR mice pretreated with DT, the injection of LPS and flagellin did not increase total cellularity (n Ն 4 mice for each group) ( p ϭ 0.4495 for LPS; p ϭ 0.9917 for flagellin), or specific CD4 ϩ T cell ( p ϭ 0.9347 for LPS; p ϭ 0.8943 for flagellin) or NK cell recruitment ( p ϭ 0.089 for LPS; p ϭ 0.9623 for flagellin), whereas poly(I:C), resiquimod (R848), and CpG still were capable of increasing cellularity, even with the extensive reduction of CD11c ϩ DCs (Fig. 3, right panels).
We next tested whether the cellular recruitment into the draining popliteal lymph node in response to LPS could be rescued in the CD11c-DTR mice by administering wild-type bone marrow-derived DCs at the time of DT administration. We used GM-CSFcultured DCs because DCs generated using this protocol most resembled the immature conventional DC subtype being depleted in our model. This population of DCs contained Ͻ0.1% contaminating NK cells (data not shown). At 24 h after DC administration, these mice were either treated with LPS or untreated, and cellularity was examined by flow cytometry. Interestingly, LPS-treated FIGURE 2. Total conventional (CD11c ϩ , MHC class II ϩ ) and plasmacytoid DC numbers and activation status in draining popliteal lymph node following footpad injection with specific TLR agonists. Wild-type mice received the footpad injection of the specific TLR agonists, and the draining ipsilateral (f) and contralateral (Ⅺ) lymph nodes were harvested 24 h later. Single cell suspensions were obtained, and the total number of conventional CD11c ϩ MHC class II ϩ cells (A), and PDCA-1 ϩ B220 ϩ (C) cells, and their activation status (CD86 ϩ ) (B and D) were determined. Values represent an n ϭ 4 mice. ‫,ء‬ p Ͻ 0.05 by paired Student's t test in the ipsilateral lymph node, compared with the contralateral lymph node. Resiquimod (R848) treatment is indicated.

FIGURE 3.
Lymph node total, CD4 ϩ T cell and NK cell recruitment to specific TLR agonists in wild-type and CD11c-DTR mice. Wild-type (B6) and CD11c-DTR mice received the i.p. injection of 4 ng per gram of body weight of DT followed 24 h later with the footpad injection of different TLR agonists, and the cellularity in the ipsilateral (f) and contralateral (Ⅺ) popliteal lymph nodes was determined 24 h later. A, Total cellular number in the ipsilateral and contralateral lymph nodes from wild-type and CD11c-DTR mice. B, Total CD4 ϩ T cell number in the ipsilateral and contralateral lymph nodes. C, CX5 ϩ DX5 ϩ NK cell number in the ipsilateral and contralateral lymph nodes. Each value is obtained from two pooled lymph nodes in n ϭ 4 mice. ‫,ء‬ p Ͻ 0.05 by paired Student's t test in the ipsilateral lymph node, compared with the contralateral lymph node. Resiquimod (R848) treatment is indicated.
CD11c-DTR mice pretreated with wild-type DCs demonstrated a greater than 2-fold increase in total cellularity, whereas the CD11c-DTR mice that were treated with DCs, but not treated with LPS, demonstrated an insignificant, albeit modest, increase in total cellular and CD4 cellular recruitment ( p ϭ 0.19 total cellularity; p ϭ 0.10 for CD4 cells) (Fig. 4A). Similarly, repletion with wildtype DCs restored the increase in CD4 ϩ T cell and NK cell recruitment to draining lymph nodes of CD11c-DTR mice treated with LPS (Fig. 4, B and C). Surprisingly, injection of wild-type DCs into CD11c-DTR mice without LPS injection was sufficient to induce some NK cell recruitment to the draining lymph nodes, suggesting that small numbers of activated DCs present in the GM-CSF generated DCs may have been sufficient to induce NK cell recruitment to the lymph node. These data demonstrate that the presence of the full complement of conventional DCs is absolutely required for maximal in vivo cellular recruitment to draining lymph nodes following TLR4 ligand treatment (and likely TLR5 ligand), but redundant mechanisms exist that do not require DCs in response to the other TLR agonists. Alternatively, the residual DC compartment following DT-mediated depletion, composed mainly of plasmacytoid DCs and more immature CD11c low DCs may be sufficient for the recruitment of cells to locally administer TLR3, TLR7, and TLR9 agonists.

DCs are not required for in vivo T cell or NK cell activation by TLR agonists
To test whether DCs are necessary for TLR-induced T cell and NK cell activation, we examined the expression of CD25 and CD69 on CD4 ϩ and CX5 ϩ DX5 ϩ cells. Preliminary data showed that CD69 expression was a more sensitive marker for T cell and NK cell activation 24 h following TLR ligand administration, and was used in subsequent experiments. In wild-type mice, we found that flagellin, LPS, resiquimod (R848), and CpG all induced significant activation of CD4 ϩ T cells in the draining popliteal lymph node, whereas poly(I:C) did not (Fig. 5A). Unexpectedly, TLR agonist administration to CD11c-DTR mice also yielded significant activation of CD4 ϩ T cells in the draining lymph nodes, as measured by CD69 expression (Fig. 5A). Furthermore, CD69 expression was increased on NK cells in both wild-type and CD11c-DTR mice in response to TLR agonists (Fig. 5B), demonstrating that DC depletion does not significantly alter TLR agonist-induced T cell or NK cell activation in vivo.
We also examined very briefly whether these TLR agonists polarized the CD4 ϩ T cells to a Th1 or a Th2 phenotype. CD4 ϩenriched T cells obtained from the ipsilateral and contralateral lymph nodes of wild-type and CD11c-DTR mice pretreated with the footpad injection of TLR agonists were incubated on anti-CD3coated plates with GolgiStop, and Th1 and Th2 cytokines determined by flow cytometry. Footpad injections of resiquimod (TLR7  agonist) increased the number of both IL-4-and IFN-␥-secreting CD4 ϩ T cells in both the ipsilateral and contralateral lymph nodes (data not shown).
Taken together, these data reveal that DCs regulate TLR4-and TLR5-mediated NK and T cell recruitment, but are dispensable for TLR3-, TLR7-, and TLR9-mediated NK and T cell recruitment. In addition, in vivo T cell and NK cell activation in response to TLR agonists do not require DCs, although we could not determine whether Th polarization of the CD4 ϩ T cells varied with the different TLR agonists or the depletion of conventional DCs.

Type I IFN partially substitutes for DCs in TLR3-and TLR7-induced cellular recruitment
Because TLR3, TLR7, and TLR9 ligands all caused cellular recruitment in DC-depleted mice, we wished to examine alternative mechanisms. A potential mechanism for several TLR-induced immune responses, including DC maturation, depends on the production of type I IFN (20). In fact, IFN-induced chemokines, such as IP-10 and RANTES, are important recruiters of CD4 ϩ T cells and NK cells (13). In preliminary studies, local footpad injection of IFN-␣ increased lymph node cellularity (data not shown) and specific recruitment of T cells and NK cells to the draining popliteal lymph nodes (Fig. 6). Furthermore, pretreatment with an antiserum that neutralizes type I IFNs, but not a control antiserum, completely inhibited IFN-induced T cell and NK cell recruitment in wild-type mice (Fig. 6). Interestingly, pretreatment of DC-depleted mice with anti-IFN antiserum partially inhibited the recruitment of total cells, and completely inhibited the recruitment of NK cells to the draining lymph nodes of locally administered TLR3 ligand, but had no effect on CD4 ϩ T cell recruitment in DC-depleted mice (Fig. 7, A-C). Somewhat similar results were seen with the TLR7 agonist, where blockade of IFN-␣ prevented the recruitment of both NK and T cells to the draining lymph nodes (Fig. 7, D-F). In contrast, pretreatment of DC-depleted mice with anti-IFN antiserum had no effect on the recruitment of NK or CD4 ϩ T cells in response to the TLR9 agonist, CpG (Fig. 7, G-I). These results demonstrate that type I IFN production may partially substitute for DCs in TLR3-and TLR7-induced (but not TLR9-induced) in vivo cellular responses, in this case cellular recruitment to draining lymph nodes.
To examine whether type I IFNs also contribute to the recruitment of cells to the lymph node in response to bacterial LPS in mice that have a full complement of DCs, wild-type mice were pretreated with anti-IFN antisera before the footpad injection of LPS. As shown in Fig. 8, passive immunization with anti-IFN antisera partially attenuated the recruitment of total cells and NK cells, suggesting that although there is an absolute dependency on DCs for this response to LPS, endogenous type I IFN production also contributes to the cellular recruitment by these DCs.

Discussion
In this study, we have shown considerable plasticity in the DC response to different TLR agonists, in vivo, with conventional DCs being absolutely required for only TLR4-and TLR5-mediated recruitment to draining lymph nodes following local administration. In a previous study, we observed that flagellin activates DCs in TLR4 mutant mice, and flagellin signaling was completely inhibited by mutating an amino acid in the TLR5 binding domain (15). We used the same flagellin preparation in the present studies, which was essentially endotoxin-free, as determined by Limulus assay. Those findings suggest that the LPS and flagellin responses were specific to binding to their respective TLR, and not due to cross-reactivity.
Expectedly, local treatment with TLR agonists induced DC maturation in vivo, leading to increased cellular recruitment to the draining lymph nodes. All TLR agonists were similarly effective in this regard. These findings are therefore consistent with the observations of Kamath et al. (21) who showed that T cell activation is likely a common response to infection and can be mediated by the triggering of several different TLRs. DCs are known to interact with CD4 ϩ T cells and recent reports have shown the importance of DC-NK cell interactions in modulating immune responses (5,6,22). Interestingly, when conventional DCs were depleted from transgenic mice, only TLR4 and TLR5 agonists failed to induce significant immune cell recruitment to the draining lymph nodes, showing similarities in DC signaling through TLR4 and TLR5. The fact that repletion with wild-type DCs restored LPS-induced cellular recruitment in these DC depleted mice demonstrates the necessity of DCs for these in vivo responses through TLR4 sig-naling. Although many other cells including macrophages/monocytes, endothelial cells, epithelial cells and/or neutrophils express TLR4 and/or TLR5, it appears that only conventional CD11c ϩ MHC class II ϩ DCs can produce the chemokines and/or inflammatory signals necessary to induce the full recruitment of major effector cell populations of the innate and adaptive immunity to the draining lymph nodes. Because activation of both NK cells and T cells occurred in response to multiple TLR ligands when the vast majority of conventional DCs were depleted, production of these cytokines/chemokines or other inflammatory mediators by other cells present in the lymph node environment may substitute for DC activation signals, allowing cells to generate immune responses without DCs. In fact, LPS-induced cellular recruitment to the popliteal lymph node was attenuated in wild-type mice by the administration of anti-IFN antiserum, suggesting that type I IFN production by conventional DCs may be an important priming signal in lymph node cellular recruitment.
TLR3, TLR7, and TLR9 signaling pathways are known to have potent adjuvant properties in mice (12). One mechanism by which these TLRs can induce adjuvant-like effects is through the induction of type I IFN. TLR3 agonists, for example, can activate type I IFN production (3), and the activation of IFN-inducible genes in many cell types, whereas TLR7 and TLR9 agonists can activate type I IFN production mainly in plasmacytoid DCs (23,24). Because plasmacytoid DCs were only ϳ75% depleted in DT-treated CD11c-DTR mice, it is possible that these IFN-producing cells were capable of producing sufficient IFN to elicit the observed response. In confirmation of this suggestion, we have shown that TLR3, TLR7, and TLR9 ligands possess adjuvant effects in the absence of DCs, as they caused the recruitment and activation of T cells (with the exception of poly(I:C)) and NK cells in the absence of DCs. To confirm that type I IFNs participated in the DC-independent TLR-induced responses, we blocked type I IFN with a neutralizing Ab before a footpad injection of a TLR3, TLR7, or TLR9 agonist. For the TLR7 agonist, neutralization of type I IFN inhibited total cellular T cell and NK cell recruitment to the draining lymph node in DC-depleted mice, indicating that type I IFN can partially substitute for DCs in our system. Similar results were seen with the TLR3 agonist, although the recruitment of T cells was not completely prevented, whereas NK cells and total cellularity were significantly blocked. In contrast, blocking type I IFNs had no effect on the recruitment of cells in response to a TLR9 agonist in DC-depleted mice. The differences in the cellular responses to TLR3, TLR7, and TLR9 agonists in the presence of type I IFN blockade are not fully understood but may be mediated, in part, by the different cell types that respond to the respective TLR ligand.
These results demonstrate, however, that different TLR agonists activate DCs in a ligand-specific manner and can activate innate and adaptive immunity in vivo dependently and independently of DCs. Although responses dependent on DCs appear more robust, other pathways exist in the absence of DCs to preserve some immune functions. There are a number of important clinical implications for these studies. Because TLR agonists are currently being explored as adjuvants for vaccine development, the current studies highlight the potential differences that may well be observed. For example, vaccination with TLR3, TLR7, TLR8, or TLR9 ligands can potentially serve as adjuvants, even in patients with decreased DC function and/or number. Similarly, one can envision different antimicrobial activities of specific TLR agonists based on their ability to activate different effector cell populations. In general, these findings reconfirm the in vivo redundancy in the recognition and signaling of microbial pathogens by the host.