Crucial Contribution of Thymic Sirpα⁺ Conventional Dendritic Cells to Central Tolerance against Blood-Borne Antigens in a CCR2-Dependent Manner

The thymus is vital for development of T cells. T progenitor cells in the thymus are subjected to positive and negative selection, and survivors become self-MHC-restricted and self-tolerant mature naive T cells. Negative selection induces clonal deletion of potentially pathogenic autoreactive T cells and consequently decreases the risk of the development of autoimmune disorders (1). Thus, negative selection has a major role in central tolerance. Medullary thymic epithelial cells (mTECs)² are major inducers of negative selection. mTECs express the autoimmune regulator (AIRE) gene, which induces the ectopic expression of a milieu of peripheral tissue-specific Ags in the thymus resulting in the clonal deletion of autoreactive T progenitors with specificity for these Ags (2, 3, 4). Another type of thymic APCs, in particularly dendritic cells (DCs), have also been shown to contribute to negative selection (5, 6, 7). However, the detailed molecular and cellular mechanisms by which thymic DCs mediate negative selection remain largely unknown.

Thymic DCs are heterogeneous, similar to DCs in peripheral lymphoid organs such as lymph nodes and spleen. In humans and mice, thymic DCs are classified into two distinct subsets, CD11c⁺B220⁺ plasmacytoid DCs (pDCs) and CD11c⁺B220⁻ conventional DCs (cDCs). cDCs are further divided into CD11c⁺CD11b⁻CD8α⁺Sirpα⁻ and CD11c⁺CD11b⁺CD8α⁻Sirpα⁺ subsets (8, 9). CD8α⁺Sirpα⁻ cDCs, the most abundant subset among these three thymic DC subsets, are clustered in the medulla (10, 11). These CD8α⁺Sirpα⁻ cDCs also express AIRE and can present endogenous self-Ags. In addition, they can cross-present tissue-specific Ags derived from the mTECs for negative selection (12, 13). In contrast, the intrathymic location and functions of another minor cDC, CD11c⁺CD11b⁺CD8α⁻Sirpα⁺, subset remain unclear, although this subset is presumed to migrate from the bloodstream (8). Proietto et al. (14) demonstrated that Sirpα⁺ cDCs can induce thymocytes to efficiently differentiate into regulatory T cells in vitro. However, the roles of Sirpα⁺ cDCs in central tolerance and regulatory T cell generation in vivo and the nature of the target autoantigens of central tolerance remain elusive.

Chemokines and their receptors have essential roles in controlling the homeostatic homing of immune cells including DCs and T cells (15, 16, 17). We examined the composition of thymic DC subsets in mice deficient in CCR1, CCR2, CCR5, or CX3CR1, the chemokine receptors which are expressed by DCs (18, 19). We observed that Sirpα⁺ cDCs, but not Sirpα⁻ cDCs or pDCs, were selectively decreased in the thymus of CCR2-deficient mice, but not in the other chemokine receptor gene-deficient mice. Interestingly, CCR2-deficient mice exhibited a modest impairment in intrathymic negative selection against i.v. injected Ags. Concomitantly, CCR2 deficiency allowed releasing more autoreactive T cells against serum Ags into periphery. These Sirpα⁺ cDCs migrated from bone marrow to thymus by the way of the peripheral blood and showed a unique intrathymic localization confined to perivascular and cortical areas. Moreover, Sirpα⁺ cDCs had a greater capacity to uptake blood-borne Ags than Sirpα⁻ cDCs, along with their unique intrathymic localization. Thus, our present study suggests that thymic Sirpα⁺ cDCs may function as a specialized APC for the development of central tolerance to blood-borne Ags.

Specific pathogen-free 6- to 7-wk-old male BALB/c mice were purchased from Charles River Japan and designated as wild-type (WT) mice. CCR1^−/− and CX3CR1^−/− mice were provided by Dr. P. M. Murphy (National Institute of Allergy and Infectious Diseases, National Institutes of Health, Bethesda, MD) (20, 21). CCR2^−/− (22) and CCR5^−/− mice (23) were provided by Dr. W. Kuziel (University of Texas San Antonio, San Antonio, TX) and Dr. Kouji Matsushima (University of Tokyo, Tokyo, Japan), respectively. All chemokine receptor-deficient mice were backcrossed to BALB/c mice for 8–10 generations. DO11.10 mice expressing a transgenic TCR that recognizes the OVA_323–339 peptide in the context of I-A^d were maintained as heterozygotes. DO11.10 mice were backcrossed to CCR2^−/− mice to generate DO11.10/CCR2^−/− mice. Genotyping for the CCR2 gene was done by direct PCR from whole blood samples using an Ampdirect Plus kit (Shimadzu) and the specific primers (sense, 5′-CACGAAGTATCCAAGAGCTTG-3′ and antisense, 5′-CCCAAGTGACTACACTTGTTA-3′). The mouse experiments were performed under specific pathogen-free conditions in accordance with the Guidelines for the Care and Use of Laboratory Animals of Kanazawa University.

Rat anti-mouse mAbs used were anti-CD3ε (145-2C11; Miltenyi Biotec), anti-CD4 (RM4-5; BD Pharmingen), anti-CD8 (53-6.7; BD Pharmingen), anti-CD25 (PC61; BD Pharmingen), anti-CD45R/B220 (RA3-6B2; BD Pharmingen), anti-CD172a/Sirpα (P84; BD Pharmingen), anti-DO11.10 clonotypic TCR (KJ1-26; BD Pharmingen), anti-F4/80 (A3-1; Serotec), and anti-Ly51 (6C3; BioLegend). Hamster anti-mouse CD11c (HL-3) and mouse anti-mouse I-A^d (AMS-32.1) mAbs were purchased from BD Pharmingen. Rabbit anti-mouse CCR2 mAb and anti-mouse type IV collagen (Col IV) polyclonal Ab were purchased from Epitomics and LSL, respectively. Goat anti-mouse MCP-2 polyclonal Ab was purchased from Santa Cruz Biotechnology. Isotype-matched control IgGs for each rat and hamster mAbs were purchased from BD Pharmingen. Mouse, rabbit, and goat IgG (Sigma-Aldrich) served as controls.

Thymus was digested in 0.6 mg/ml collagenase type IV (Sigma-Aldrich) and 25 Kunitz units/ml DNase Ι (Sigma-Aldrich) in RPMI 1640 (Sigma-Aldrich) at 37°C for 20 min. The low-density cells were further isolated from the resultant single-cell suspensions using Histopaque-1077 reagent (Sigma-Aldrich). PBMCs were isolated from whole blood using Histopaque-1083 reagent (Sigma-Aldrich). Bone marrow cells were washed out with cold RPMI 1640 medium from the femoral and tibial bones.

The low-density cells from thymus, PBMCs, and bone marrow cells were stained with various combinations of fluorescent dye-conjugated or nonconjugated specific Abs in PBS supplemented with 2 mM EDTA and 3% FBS. For nonconjugated Abs, fluorescent-conjugated secondary Abs were used. After washing in PBS, expression of cell surface molecular markers was analyzed using a FACSCalibur (BD Biosciences) with CellQuest Pro software (BD Biosciences).

Thymic tissues were frozen in OCT compound (Sakura) and 6-μm-thick cryostat sections were stained with H&E. For immunofluorescence analysis, 6-μm-thick cryostat sections were fixed with cold acetone for 3 min and incubated with Protein Block Reagent (DakoCytomation) to block nonspecific binding. Then fluorescent immunostaining was done by the standard method (for details, see the figure legends). After washing with 0.05% Tween 20-PBS, slides were mounted in fluorescent mounting medium (DakoCytomation). Immunofluorescence was detected in a setting that excluded the nonspecific signal of the isotype control using a fluorescence microscope (BX50; Olympus) or confocal laser-scanning microscope (LSM510; Zeiss). DP Controller software (Olympus) and Zen 2007 software (Zeiss) were used for image processing.

Total RNAs were extracted from tissues using a RNeasy Mini Kit (Qiagen) and then reverse-transcribed using SuperScript III First-Strand Synthesis System (Invitrogen). PCR was done using the cDNAs, 2.5 mM dNTP mix (Takara), TaqDNA polymerase (Takara), and the specific primer sets for the GAPDH gene (sense, 5′-CAC TGA GCA TCT CCC TCA CA-3′ and antisense, 5′-TGG GTG CAG CGA ACT TTA TT-3′), CD45 gene (sense, 5′-AAG ACA GAG TGC AAA GGA GAC-3′ and antisense, 5′-TGT AGG TGT TTG CCC TGT GAC AAA GAC-3′), keratin 8 gene (sense, 5′-ACG GTG AAC CAG AGC CTG T-3′ and antisense, 5′-CTC CAC TTG GTC TCC AGC AT-3′), MCP-1 gene (sense, 5′-CCC ACT CAC CTG CTG CTA CT-3′ and antisense, 5′-TCT GGA CCC ATT CCT TCT TG-3′), MCP-2 gene (sense, 5′-CAG TCA CCT GCT GCT TTC AT-3′ and antisense, 5′-ATA CCC TGC TTG GTC TGG AA-3′), and MCP-3 gene (sense, 5′-AAA CAA AAG ATC CCC AAG AGG-3′ and antisense, 5′-CAC AGA CTT CCA TGC CCT TC-3′) for 30 cycles of 95°C for 30 s, 55°C for 30 s, and 72°C for 30 s.

DO11.10-transgenic mice with or without CCR2 gene deficiency were administered 200 μg of OVA_323–339 peptide (ABGENT) in PBS through the tail vein. To induce thymocyte deletion independently of Ag presentation, mice were injected i.p. with 50 μg of anti-CD3ε mAb (24). Two days after injection, thymocytes were collected and stained with the following combinations of Abs: anti-CD4, anti-CD8, and anti-DO11.10 or anti-CD4, anti-CD25, and anti-DO11.10 Abs. To detect apoptotic cells, thymocytes were stained using an Annexin V-FITC Apoptosis Detection Kit (Merck). After being stained, the cells were analyzed by FCM.

Bone marrow cells were cultured in RPMI 1640 medium supplemented with 10% FBS and mouse GM-CSF (R&D Systems) at a concentration of 20 ng/ml. An equal volume of culture medium of the same content was added at 4 days, and one-half of the medium was replaced with fresh culture medium at 7 days after the plating. Most bone marrow cells were differentiated into immature DCs as judged by morphological appearances at 10 days after the initiation of the culture. The resultant immature DCs were stained with 1 μM 5-chloromethylfluorescein diacetate (CMFDA; Invitrogen) dye and 1 million cells were injected into the tibial bone marrow cavity of each mouse. After the injection, low-density cells were obtained from thymus, lymph nodes, or PBMCs to determine the presence of CMFDA-stained DCs by using FCM.

Alexa Fluor 488-conjugated OVA protein (OVA₄₈₈), Alexa Fluor 647-conjugated OVA protein (OVA₆₄₇) (Invitrogen), or mouse serum IgG (Sigma-Aldrich), which was conjugated with Alexa Fluor 647, using an Alexa Fluor 647 protein labeling kit (Invitrogen), was injected into the tail vein of mice. Thymic low-density cells and PBMCs were isolated at the indicated time points after OVA protein injection and were stained with anti-CD11c and anti-Sirpα Abs. Then the cells were analyzed by FCM. For the localization of the Ag uptake, cryostat sections of frozen thymic tissues were obtained from mice injected with OVA protein and were stained with anti-Sirpα, anti-CD11c, anti-I-A^d, anti-Ly51, or anti-Col IV Abs and were then observed by fluorescence microscope.

Low-density cells were isolated from the thymus and were incubated with 10 μg/ml OVA₆₄₇ in RPMI 1640 at 37°C for 20 min. As a negative control, incubation was conducted on ice. Endocytosis by each thymic DC subset was analyzed by FCM after being stained with anti-CD11c and anti-Sirpα Abs. In some experiments, low-density cells were preincubated with 10 μM cytochalasin D (Cyt D; Sigma-Aldrich), an actin inhibitor (25), 100 mM ammonium chloride (NH₄Cl) (Wako), an inhibitor of the clathrin-dependent pathway (26), or 0.5 mg/ml mannan (Sigma-Aldrich) at 37°C for 15 min before incubation with OVA₆₄₇ at 37°C for 20 min in the presence of fresh inhibitors.

Bone marrow cells were obtained from WT or CCR2^−/− mice and were stained with 2 μM CMFDA dye. Twenty million cells were injected into the tail vein of CCR2^−/− mice. OVA₆₄₇ was injected into the tail vein at 2 days after injection. Thymic low-density cells were isolated at 1 h after OVA protein injection, and the presence of donor-derived Sirpα⁺ cDCs and their capability of Ag uptake were analyzed by FCM.

Spleen mononuclear cells were isolated from WT or CCR2^−/− mice and were labeled with 25 μM CFSE using a CellTrace CFSE Cell Proliferation Kit (Invitrogen). Ten million prelabeled cells were injected into the tail vein of WT mice. One day after injection, mice were immunized with total mouse serum protein emulsified in CFA. PBS in CFA was immunized as a control. Two days after immunization, lymphocytes were harvested from draining and nondraining lymph nodes and stained with anti-CD4 mAb. The percentage of CFSE-diluted divided cells was analyzed by FCM.

Data are represented as mean ± SD. Statistical significance was determined by one-way ANOVA followed by the Tukey-Kramer test. A value of p < 0.05 was considered statistically significant.

Consistent with a previous report (8), three distinct populations of thymic CD11c⁺ DCs have been identified: B220⁺ pDC, B220⁻Sirpα⁻ cDC, and B220⁻Sirpα⁺ cDC subsets (Fig. 1^,A). cDC and pDC subsets were present mainly in the forward scatter (FSC^high), side scatter SSC^high, and SSC^low areas upon FCM, respectively (Fig. 1^,B). The pivotal role of chemokines in the trafficking of DCs prompted us to examine thymic DC subsets in mice deficient in chemokine receptor genes. Sirpα⁺ DCs were markedly decreased in CCR2^−/− mice, compared with WT mice, both in the relative (Fig. 1, C and D) and absolute number (Fig. 1^,E), whereas Sirpα⁻ DC (Fig. 1 C) and B220⁺ pDC numbers (data not shown) were not changed in CCR2^−/− mice. In contrast, no significant changes were observed on thymic cDC and pDC subsets in mice deficient in other chemokine receptors including CCR1, CCR5, and CX3CR1. Moreover, we did not observe any differences in thymic B220⁺ B cell and F4/80⁺ macrophage numbers between WT and CCR2^−/− mice (data not shown). Microscopic studies of the thymus failed to reveal any morphological differences between WT and CCR2^−/− mice in terms of the total cellularity, the distribution of thymocytes in each developmental stage, and the localization of Ly51⁺ cortical thymic epithelial cells and I-A^{d high} mTEC (supplemental Fig. S1³). Thus, CCR2^−/− mice exhibit a selective decrease in the Sirpα⁺ DC subset in thymus.

Sirpα⁺ DCs are presumed to have the capacity to carry peripheral tissue Ags into the thymus (14). We next investigated the roles of Sirpα⁺ DCs in thymus on taking in an i.v. administered Ag. PBS injection did not cause any changes in each developmental stage of thymocytes in DO11.10 and DO11.10/CCR2^−/− mice (data not shown). On the contrary, i.v. administration of OVA_323–339 peptide markedly reduced the proportion and absolute number of clonotypic CD4/CD8 double-positive (DP) thymocytes in DO11.10 mice. CCR2 gene ablation modestly attenuated this reduction (Fig. 2, A–C). OVA peptide injection consistently increased the proportion of annexin V⁺ apoptotic cells in DO11.10 mouse thymus compared with that in DO11.10/CCR2^−/− mice (Fig. 2,D). In contrast, OVA peptide induced a modest increase in the number of DO11.10⁺CD25⁺CD4⁺ regulatory T cell phenotype to similar extents in both DO11.10 and DO11.10/CCR2^−/− thymus (Fig. 2 E). Thus, decreased thymic Sirpα⁺ DCs in CCR2^−/− mice may be associated with a moderately impaired thymic negative selection. Moreover, following i.p. injection with anti-CD3 Ab (24), thymocytes were deleted to similar extents in DO11.10 and DO11.10/CCR2^−/− mice (supplemental Fig. S2), indicating the absence of intrinsic defects of thymocytes in the absence of CCR2. These results collectively suggest that thymic Sirpα⁺ DCs can contribute to intrathymic negative selection of a bloodstream-derived Ag without inducing regulatory T cells.

To elucidate the functions of thymic Sirpα⁺ DCs more in detail, we determined their intrathymic localization. In thymi of WT mice, Sirpα was mainly detected on CD11c⁺ DCs scattered in the thymic cortex (Fig. 3, A and B), but not on CD11c⁺ DCs clustered in medulla, the predominating site of thymic CD8α⁺Sirpα⁻ DCs. Moreover, most Sirpα⁺ DCs were localized in close proximity to small vessels with single Col IV⁺ basement membrane or inside perivascular regions (PVRs) separated by two Col IV⁺ basement membranes in the cortex (Fig. 3,C). The thymic DC population includes APCs crucially involved in the central tolerance system involving bloodstream C5 Ag (27). Furthermore, Sirpα⁺ DCs are selectively localized in PVRs or in close proximity to small vessels, both essential components of the blood-thymus barrier (28). Hence, we hypothesized that this DC subset might be involved in Ag uptake from the bloodstream. To address this possibility, we treated WT mice i.v. with OVA₆₄₇ and examined its uptake by thymic DCs. Intrathymic Sirpα⁺ DCs, but not Sirpα⁻ DCs, took up OVA protein in a dose-dependent manner (Fig. 4,A), maintaining a stable level from 1 to 4 h after the injection and decreasing thereafter (Fig. 4,B). Recently, it was reported that bloodstream DCs could efficiently capture and transport particulate bacteria into the spleen when particulate bacteria were i.v. injected (29). Indeed, bloodstream CD11c⁺ cells rapidly disappeared from the peripheral blood after capturing OVA protein (Fig. 4,C). By contrast, the uptake by intrathymic Sirpα⁺ DCs reached a peak level at 15 min, decreasing to the stable level thereafter. Thus, there may be a remote possibility that circulating DCs migrated into the thymus after capturing OVA protein inside the bloodstream. Furthermore, in addition to an exogenous protein, intrathymic Sirpα⁺ DCs also captured an endogenous serum protein, mouse IgG, which was conjugated with Alexa Fluor 647, when it was administered i.v. (supplemental Fig. S3). Thus, Sirpα⁺ DCs can effectively capture peripheral Ags from the bloodstream across the blood-thymus barrier. This notion was further supported by the observation that Sirpα⁺ DCs engulfed OVA protein with a higher efficiency than Sirpα⁻ DCs when cultured in vitro with OVA₆₄₇ (Fig. 4, D and E). Mannan from Saccharomyces cerevisiae, but not NH₄Cl or Cyt D from Zygosporium mansonii, markedly inhibited endocytosis of OVA protein by Sirpα⁻ DCs (Fig. 4,F, upper panel). On the contrary, uptake of OVA protein by Sirpα⁺ DCs was markedly attenuated by NH₄Cl and Cyt D, but not mannan (Fig. 4 F, lower panel). These observations suggest that thymic Sirpα⁺ DCs can endocytose soluble Ags more efficiently than Sirpα⁻ DCs, in a clathrin-dependent, but not mannose receptor-dependent manner.

We examined sequentially intrathymic localization of OVA-derived signals after i.v. injection of OVA₄₈₈. By 0.5 h, OVA₄₈₈-derived signals were detected in Sirpα⁺ cells (Fig. 5,A), CD11c⁺ DCs (Fig. 5,B) and inside PVRs or in close proximity to small vessels (Fig. 5,C). Although some signals remained nearby in small vessels, signals inside PVRs were obviously decreased at 6 h (Fig. 5,D), as judged by the Col IV immunostaining pattern. At 18 h after the injection, OVA₄₈₈-derived signals were mainly scattered in the Ly51⁺ cortical area but not in the I-A^{d high} medullary area (Fig. 5,E). Because OVA₄₈₈-derived signals were constantly detected in Sirpα⁺ DCs at every time point (data not shown), these observations suggest that Sirpα⁺ DCs initially capture bloodstream OVA protein inside PVRs or in nearby small vessels and then migrate into the cortical parenchyma. To examine the process of migration more in detail, OVA₆₄₇ (blue) and OVA₄₈₈ (green) were i.v. injected sequentially with an interval of either 6 or 18 h as shown in Fig. 5,F. When OVA₄₈₈ was injected 6 h after OVA₆₄₇, double-positive CD11c^high DCs were evidently detected (8.1%), while single-positive cells were sparse (Fig. 5,F, left upper panel). Even at 18 h after the injection, double-positive CD11c^high DCs were still present (3.2%) with substantial numbers of OVA₄₈₈-derived signal single-positive (3.9%) or OVA₆₄₇-derived signal single-positive cells (2.3%; Fig. 5,F, left lower panel). Thus, CD11c^high DCs with Sirpα expression can persistently be in close interaction with the bloodstream while they are migrating into cortical parenchyma (Fig. 5 G).

It is possible that a decreased intrathymic Sirpα⁺ DC number may account for the defect in their migration in CCR2^−/− mice, because the thymic Sirpα⁺ cDC subset is presumed to migrate from the bloodstream (14). Most CD11c⁺B220⁻ DCs in peripheral blood and bone marrow expressed abundantly Sirpα (supplemental Fig. S4), similarly as observed on thymic Sirpα⁺ DCs, and this population expressed CCR2 (supplemental Fig. S5). CCR2^−/− mice exhibited a moderate reduction in CD11c⁺B220⁻ DCs in peripheral blood, but not bone marrow (Fig. 6, A and B). This suggests a possible defect in the migration of CD11c⁺B220⁻ DCs from bone marrow in CCR2^−/− mice. To test this possibility, bone marrow cells were induced to differentiate to DCs with in vitro GM-CSF stimulation, labeled with CMFDA, and injected into bone marrow of WT mice (Fig. 6,C, upper illustration). Under these conditions, >80% of injected cells expressed CD11c, Sirpα, and CCR2, but not B220 (supplemental Figs. S4 and S5). WT-derived DCs appeared in peripheral blood rapidly within 2 h after the intra-bone marrow injection, whereas CCR2^−/− mouse-derived DCs migrated into peripheral blood less efficiently (Fig. 6, C and D). Interestingly, CD11c⁺B220⁻Sirpα⁺ DCs appeared in thymus by 6 h after intra-bone marrow injection (Fig. 6,E). These observations suggest that CCR2-mediated signals were critical of the migration of Sirpα⁺ DCs from bone marrow into the thymus. Moreover, Sirpα⁺ DCs were markedly decreased in PVRs of CCR2^−/− thymus compared with those of WT thymus (WT mice, 7.0 ± 2.6/site; CCR2^−/− mice, 2.3 ± 1.5/site; Fig. 3, D–F). Furthermore, the decrease was more evident in the region inside the PVRs compared with that outside the PVRs (Fig. 3,G). CCR2 was expressed also by a portion of intrathymic Sirpα⁺ DCs, but not Sirpα⁻ DCs (Fig. 3,H). Three mouse chemokines, MCP-1, MCP-2, and MCP-3, can bind to CCR2 (30). Among these chemokines, only MCP-2 mRNA was constitutively expressed in thymus, particularly keratin 8-positive thymic stroma, but not CD45-positive thymocytes (Fig. 7, A and B). Moreover, MCP-2 immunoreactivities were consistently detected inside the PVRs (Fig. 7,C, upper panels) and on Sirpα⁺ cells in the PVRs (Fig. 7 C, lower panels). Thus, it is probable that the CCR2-MCP-2 interaction can contribute to intrathymic localization of Sirpα⁺ DCs, particularly in the PVRs.

Because the PVR was proved to be a main location of the uptake of circulating Ags, we further examined the effects of CCR2 deficiency on the capability of Sirpα⁺ DCs to uptake Ags from the bloodstream. Indeed, when OVA₆₄₇ was injected i.v., CCR2^−/− mice exhibited a reduced proportion of intrathymic DCs capturing OVA protein compared with WT mice (Fig. 8, A and B). Moreover, after the OVA₆₄₇ injection, Sirpα⁺ DCs of WT mice contained a substantial proportion of OVA^high cells, which represent the cells with a higher uptake of OVA protein, and this population was markedly reduced in CCR2^−/− mice (Fig. 8, C and D). Moreover, among Sirpα⁺ DCs, the CCR2-expressing population was a main cell type which captured OVA protein (Fig. 8,E). CMFDA-labeled WT mouse-derived bone marrow cells appeared in thymus 2 days after the adoptive transfer to CCR2-deficient mice and a substantial proportion of these stained cells expressed CD11c and Sirpα simultaneously (Fig. 8,F). Sirpα⁺CD11c⁺ DCs appeared in thymus similarly when CMFDA-labeled CCR2-deficient mouse-derived bone marrow cells were adoptively transferred (data not shown). When OVA₆₄₇ was injected i.v. 2 days after the adoptive transfer, WT donor-derived Sirpα⁺CD11c⁺ DCs captured OVA protein more efficiently than CCR2-deficient DCs in the CCR2-deficient thymus (Fig. 8 G). Thus, CCR2-mediated signals may at least partially regulate the function of Sirpα⁺ DCs to uptake Ag from the bloodstream (supplemental Fig. S6).

We observed that CCR2^−/− mice did not exhibit any signs suggestive of autoimmune disorders until 1 year after the birth (our unpublished data). Hence, we examined whether autoreactive T cells against certain self-Ags in the bloodstream accumulated in the periphery of CCR2^−/− mice. We examined the accumulation of autoreactive T cells in the draining lymph nodes in WT mice that received CFSE-labeled WT or CCR2^−/− mouse-derived splenocytes and were subsequently immunized with mouse serum emulsified in CFA. Immunization with total serum protein increased the cell division of CCR2^−/− mouse-derived CD4⁺ T cells inside draining lymph nodes (10.6%) to a greater extent than immunization with PBS (4.3%; Fig. 9,A). Moreover, CD4⁺ T cell division was significantly increased in the recipients of CCR2^−/− mouse-derived splenocytes compared with the recipients of WT mouse-derived splenocytes (Fig. 9 B). Thus, the lack of CCR2 can result in enhanced accumulation of autoreactive T cells against serum self-Ags.

Mouse thymus CD11c⁺ cDCs can be classified into two populations, a major CD8α⁺ and a minor CD8α⁻ one (31). CD8α⁻ cDCs can pick up CD8αβ heterodimer from thymocytes and retain them on the cell surface, thus precluding the use of CD8α as a reliable marker to distinguish these two populations. Wu and Shortman (8) observed that CD8α⁻ but not CD8α⁺ cDCs simultaneously express the Sirpα molecule and proposed the use of Sirpα as a marker of this minor cDC population. Concomitantly, it was proposed that the interaction between thymocytes and DCs in thymic cortex can also have profound effects on positive selection (32). Likewise, McCaughtry et al. (33) observed that clonal deletion of autoreactive thymocytes requires the stimuli from rare CD11c⁺ cortical DCs. Given the unique localization of Sirpα⁺ DCs confined to the cortex, these observations suggest the potential involvement of Sirpα⁺ DCs in central tolerance, but their small number hinders the isolation for a detailed analysis of Sirpα⁺ DC function.

A partial but selective reduction in intrathymic Sirpα⁺ cDCs in CCR2^−/− mice prompted us to investigate the thymic selection process in WT and CCR2^−/− mice to elucidate the role of intrathymic Sirpα⁺ cDCs in the process. When DO11.10 TCR-transgenic mice were administered immunogenic OVA_323–339 peptide i.v., CCR2 gene ablation partially attenuated the clonal negative deletion by apoptosis of the DO11.10⁺ DP thymocyte population. Intraperitoneal injection of anti-CD3 Ab deleted thymocytes to similar extents in WT and CCR2^−/− mice, excluding the possibility that CCR2 deficiency impaired the apoptotic response of thymocytes. Negative selection can be exerted by various types of APCs including Sirpα⁻ cDCs, B cells, macrophages, cortical thymic epithelial cells, and mTEC in addition to Sirpα⁺ cDCs. We failed to detect any apparent differences in other APC populations than Sirpα⁺ cDCs between WT and CCR2^−/− mice. Thus, it is unlikely that reduced negative selection in CCR2^−/− mice can be ascribed to the changes in these cell populations. Furthermore, accumulating evidence implicates intrathymic CD4⁺CD25⁺ regulatory T cells as an essential cell component in central tolerance. Indeed, Proietto et al. (14) recently reported the capability of Sirpα⁺ cDCs to induce the differentiation of regulatory T cells in vitro. However, OVA peptide injection induced the differentiation of regulatory T cells to similar extents in both DO11.10 and DO11.10/CCR2^−/− thymus. Thus, it is probable that CCR2 deficiency reduced modestly intrathymic Sirpα⁺ DCs without affecting regulatory cell induction and partially attenuated negative selection in vivo.

It remains elusive on the trafficking modes of Sirpα⁺ DCs. In CCR2^−/− mice, Sirpα⁺ DCs were decreased moderately in peripheral blood and thymus, but were increased in bone marrow. Considering that CCR2 signaling can regulate the mobilization of monocytes from bone marrow to peripheral blood (34, 35), these observations raised the possibility of a defect in the trafficking of Sirpα⁺ DCs from bone marrow in CCR2^−/− mice. Indeed, WT mouse-derived Sirpα⁺ DCs, injected into bone marrow, appeared first in peripheral blood and then the thymus. On the contrary, CCR2^−/− mouse-derived Sirpα⁺ DCs exhibited impairment in the egress from bone marrow to peripheral blood. These observations suggest that bone marrow-derived Sirpα⁺ DCs migrated to peripheral blood in response to CCR2-mediated signals and subsequently traffic to the thymus.

In the thymus, Sirpα⁺ DCs were characteristically localized in close proximity to small blood vessels and inside the PVRs, sites which are compartmentalized by a vascular basement membrane and a border membrane separating them from the thymic parenchyma (36). It is of note that Sirpα⁺ cells in the PVRs were markedly decreased in CCR2^−/− mice to a greater extent than the decrease in total Sirpα⁺ cell number. Thus, intrathymic CCR2 signaling can regulate their unique localization. This notion was supported by the observation that MCP-2, a potential ligand for CCR2, was constitutively detected in the PVRs, where Sirpα⁺ DCs were present.

PVRs can provide a pathway for hematopoietic progenitor cells and mature T cells to traverse from the bloodstream to the thymic parenchyma (36) and are presumed to constitute the blood-thymus barrier, which can protect the thymic parenchyma from bloodstream-derived macromolecules (28). Thus, the unique localization of Sirpα⁺ cDCs in the thymus suggested their potential interactions with bloodstream-derived Ag. This assumption was strengthened by our present observation that intrathymic Sirpα⁺ cDCs rapidly and specifically captured OVA protein and serum IgG following i.v. injection. Moreover, injected Ags were initially detected inside PVRs or in nearby small vessels and were subsequently in the cortical parenchyme, and the injected Ag-derived signals were consistently colocalized with CD11c and Sirpα. Thus, after CD11c⁺Sirpα⁺ cDCs, located around the PVRs, capture the Ags, they presumably move to cortical parenchyme to educate T cells. Indeed, CCR2^−/− thymus-derived Sirpα⁺ DCs exhibited a reduced capacity to uptake OVA. The lack of CCR2 can hinder the proper intrathymic localization of Sirpα⁺ DCs and their distinctive function, Ag uptake from bloodstream, thereby reducing Ag presentation in the cortical parenchyma and subsequent negative selection against a blood-borne Ag. This hypothesis is supported by our observation in that CD4⁺ T cells reactive to certain serum self-Ags accumulated in the periphery of the recipients of CCR2^−/− mouse-derived splenocytes to a greater extent than the recipients of WT mouse-derived splenocytes.

DCs can uptake free soluble Ags, in three distinct manners, by clathrin-mediated endocytosis, nonclathrin/caveolae endocytosis, and macropinocytosis (25). Thymic Sirpα⁺ cDCs could endocytose OVA Ags more efficiently than thymic Sirpα⁻ cDCs when they were cultured in vitro with OVA Ags. Furthermore, NH₄Cl, an inhibitor of clathrin-mediated endocytosis (26), markedly inhibited OVA endocytosis by Sirpα⁺ cDCs, but not by Sirpα⁻ cDCs. On the contrary, OVA protein endocytosis by Sirpα⁻ DCs was partially inhibited by mannan, whereas mannan had few effects on OVA protein endocytosis by Sirpα⁺ DCs. These observations suggest that thymic Sirpα⁺ cDCs characteristically can efficiently endocytose Ags in a manner distinct from thymic Sirpα⁻ cDCs.

Balazs et al. (29) reported that bloodstream DCs could efficiently capture and transport particulate bacteria into the spleen when particulate bacteria were i.v. injected. We also observed that CD11c⁺ DCs rapidly disappeared from peripheral blood after uptake of i.v. injected OVA protein. Given the capacity of CD11c⁺ DCs to move rapidly from blood to thymus, blood CD11c⁺ DCs may migrate into thymus after capturing the i.v. injected Ag. However, Ag-capturing DCs appeared very rapidly in the thymus, reaching maximal levels before disappearance of Ag-capturing circulating DCs from the peripheral blood. Furthermore, when OVA protein was injected i.v. into mice that contained bloodstream DCs labeled with fluorescent-conjugated latex beads, latex-labeled DCs did not appear in the thymus (our unpublished data). Thus, it is remotely possible that bloodstream DCs captured OVA protein and subsequently migrated into thymus.

In this study, we identified the unique intrathymic localization and functions of thymic Sirpα⁺ DCs that are involved in negative selection, particularly against blood-borne Ags. Serum protein can also induce negative selection in thymus (27, 37) but the molecular and cellular mechanisms remain to be elucidated. Because Sirpα⁺ cDCs can uptake serum protein such as IgG, these cells may induce central tolerance to blood-borne-derived Ags, in addition to Ags presented by the well-characterized intrathymic AIRE-mediated pathway.

We have shown that CCR2-mediated signals can regulate various biological aspects of Sirpα⁺ DCs such as their appropriate intrathymic localization and Ag uptake capacity. It is widely held that CCR2 might be a potential therapeutic target for several autoimmune disorders. However, because CCR2-mediated signals may contribute to thymic negative selection against blood-borne Ags, CCR2 blockade may aggravate autoimmune disorders similar to the observation on the murine collagen-induced arthritis model (38). Moreover, Lauritzsen et al (39) reported that proteins secreted from tumor cells into peripheral blood were transported into the thymus to eventually cause clonal deletion of tumor Ag-specific T cell repertoires. Given the potential capacity of intrathymic Sirpα⁺ DCs to capture blood-borne Ags, they may have a role in the development of tumor tolerance. Because human thymus contains DCs with similar phenotypes and intrathymic localization as Sirpα⁺ cDCs (40), a more detailed elucidation of the functions of Sirpα⁺ cDCs may provide us with useful insights to develop a better therapeutic strategy for cancer and stem cell transplantation as well as autoimmune disorders.

We express our gratitude to Drs. Joost J. Oppenheim (National Cancer Institute-Frederick, Frederick, MD) and Nobuyuki Onai (Akita University, Akita, Japan), and Yi Zhang (University of Michigan, Ann Arbor, MI) for critical review of this manuscript. We thank Drs. William Kuziel, Kouji Matsushima, and Philip Murphy for providing us with CCR2-, CCR5-, and CCR1- and CX3CR1-deficient mice, respectively.

The authors have no financial conflict of interest.