Aire-dependent production of XCL1 mediates medullary accumulation of thymic dendritic cells and contributes to regulatory T cell development

tDCs are predominantly accumulated in the medulla and sparsely detectable in the cortex (Barclay and Mayrhofer, 1981; Flotte et al., 1983; Kurobe et al., 2006). To identify chemokines that mediate the medullary accumulation of tDCs, we screened for the expression of chemokine receptors in isolated tDCs by RT-PCR analysis. Among the mouse chemokine receptors so far identified, Ccr2, Ccr4, Ccr5, Ccr6, Ccr7, Ccr8, Cxcr1, Cxcr3, Cxcr4, Xcr1, and Cx3cr1 were detected in CD11c⁺ cells isolated from the thymus (Fig. 1 A). We then examined chemokines that could bind to these transcript-detectable receptors for their ability to attract CD11c⁺ thymic cells. We found that CCL19 (CCR7 ligand), CCL21 (CCR7 ligand), CXCL12 (CXCR4 ligand), or XCL1 (XCR1 ligand) attracted CD11c⁺ thymic cells (Fig. 1 B). These results revealed the potential roles of CCR7, CXCR4, and XCR1 in the chemotactic regulation of tDCs. However, we detected no obvious defects in the medullary accumulation of CD11c⁺ DCs in the thymus of mice deficient for CCR7, CCR7 ligands, or CXCR4 (Fig. 1 C; also see Fig. 3 D), even though the medullary region in CCR7- or CCR7 ligand–deficient mice was smaller than that in control mice (Fig. 1 C; Kurobe et al., 2006; Nitta et al., 2009). Thus, instead of CCR7 and CXCR4, the XCL1–XCR1 chemokine axis may play a major role in regulating the medullary accumulation of tDCs.

We next examined the expression of Xcr1 and Xcl1 in various cell populations in the thymus. We found that Xcr1 transcripts in the thymus were detected exclusively in CD11c⁺ tDCs and not in CD45⁺ total thymocytes, CD11b⁺CD11c⁻ macrophages, or nonhematopoietic thymic stromal cells, including CD45⁻I-A⁺UEA1⁺ mTECs and CD45⁻I-A⁺Ly51⁺ cortical thymic epithelial cells (cTECs; Fig. 2 A). Among tDCs, Xcr1 was most highly expressed in I-A^highCD11c^highCD11b⁻ lymphoid DCs compared with I-A^highCD11c^highCD11b⁺ myeloid DCs or I-A^mediumCD11c^mediumB220⁺ plasmacytoid DCs (Fig. 2 B). Comparable amounts of Xcr1 were detected in DCs from the thymus, spleen, subcutaneous lymph nodes, and mesenteric lymph nodes (Fig. 2 C). In contrast, Xcl1 transcripts in the thymus were detected in CD45⁻I-A⁺UEA1⁺ mTECs rather than CD45⁺ thymocytes, CD11c⁺ tDCs, CD45⁻I-A⁻MTS15⁺ fibroblasts, or CD45⁻I-A⁺Ly51⁺ cTECs (Fig. 2 D). Among the mTECs, Xcl1 was prominently detected in CD45⁻I-A^highUEA1⁺ (MHC II^hi) mTECs rather than CD45⁻I-A^lowUEA1⁺ (MHC II^lo) mTECs (Fig. 2 E). In addition to mTECs, Xcl1 was also detected in a small subpopulation of CD45⁺ thymocytes expressing NK1.1, including NK1.1⁺CD3⁻ NK cells and NK1.1⁺CD3⁺ NKT cells (Kelner et al., 1994; Hedrick et al., 1997; unpublished data). Previous studies showed that Xcl1 is expressed in the secondary lymphoid organs by Th1-polarized CD4⁺ T cells and activated CD8⁺ T cells (Dorner et al., 2002, 2009). The expression of Xcl1 in mTECs was lower than that in activated CD8⁺ T cells (Fig. 2 F). However, in the thymus, XCL1 is prominently produced by a fraction of mTECs, whereas its receptor XCR1 is exclusively expressed by tDCs.

To examine whether XCL1 plays a role in the localization of tDCs, we generated Xcl1-deficient mice (Fig. S1). In the thymus of Xcl1-deficient mice, we found that the medullary region was not densely populated with CD11c⁺ tDCs (Fig. 3 A). Measurement of tDC density in various regions of the thymus sections (Fig. S2 A) indicated that the number of tDCs per unit area in the middle medullary region (M region) was significantly (P < 0.001) decreased in Xcl1-deficient mice as compared with control mice, whereas the number of tDCs per unit area in the deep cortical region close to the CMJ (CMJ-C region) was significantly (P < 0.01) elevated in Xcl1-deficient mice as compared with control mice (Fig. 3, A and B). In contrast, the densities of tDCs in the subcapsular zone (SCZ) region, the middle cortical region (C region), and the peripheral medullary region close to the CMJ (CMJ-M region) were not significantly (P > 0.05) altered by the Xcl1 deficiency (Fig. 3, A and B). Similarly to Xcl1-deficient mice of BALB/c background, Xcl1-deficient mice of C57BL/6 background exhibited altered distribution of tDCs, namely sparseness in the M region and density in the CMJ-C region (Fig. S2 B). Nonetheless, the absolute numbers of tDCs, including lymphoid DC, myeloid DC, and plasmacytoid DC subsets, were not affected in Xcl1-deficient mice (Fig. 3 C). These results indicate that XCL1 is essential for the accumulation of tDCs in the inner medullary region and that the deficiency of XCL1 causes the aberrant arrest of tDCs in the CMJ-C region.

The localization of tDCs was specifically altered in Xcl1-deficient mice but not in CCR7 ligand–deficient plt/plt mice or CXCR4-deficient mice (Fig. 1 C and Fig. 3 D). The defective accumulation of tDCs in the thymic medulla caused by the XCL1 deficiency was not further compromised by an additional lack of CCR7 ligands (Fig. 3 D). Thus, among the chemokines that exert chemotactic activity on tDCs (Fig. 1 B), XCL1 appears to be the major regulator of the tDC medullary accumulation.

Unlike tDCs, CD11b⁺ macrophages in the thymus were not enriched in the medulla of WT mice but were most densely distributed in the CMJ-C region (Fig. 3, E and F). This distribution was not affected in Xcl1-deficient mice (Fig. 3, E and F). The distribution and number of CD4⁺CD8⁺, CD4⁺CD8⁻, or CD4⁻CD8⁺ thymocytes were normal in Xcl1-deficient mice (unpublished data). These results indicate that in the thymus, XCL1 influences the distribution of tDCs specifically.

It was previously shown that tDCs contribute to the negative selection of self-reactive thymocytes (Jenkinson et al., 1992; Gallegos and Bevan, 2004). We therefore examined whether negative selection might be affected in the thymus of Xcl1-deficient mice. As tDCs have been implicated for their roles in negative selection by mammary tumor virus (Mtv)–encoded superantigens (Moore et al., 1994), we first analyzed the negative selection of Vβ3⁺, Vβ5⁺, and Vβ11⁺ T cells in Mtv-expressing BALB/c mice. We found that the deletion of Vβ3⁺ and Vβ5⁺ TCR^high CD4⁺CD8⁻ thymocytes in the Mtv⁺ BALB/c background, as compared with Mtv⁻ C57BL/6 background, was slightly but significantly (P < 0.05) disturbed in Xcl1-deficient mice (Fig. 4 A). Meanwhile, Vβ3⁺ and Vβ5⁺ CD4⁻CD8⁺ thymocytes, as well as Vβ11⁺ thymocytes in CD4⁺CD8⁻ and CD4⁻CD8⁺ subsets, were deleted comparably in BALB/c mice irrespective of the presence or absence of Xcl1 (Fig. 4 A). Moreover, the deletion of Vβ3⁺, Vβ5⁺, and Vβ11⁺ T cells detectable in the spleen was not defective in Xcl1-deficient BALB/c mice (Fig. S3). Thus, XCL1 plays a minor role in the negative selection of self-Mtv–reactive thymocytes.

We then examined whether negative selection of 2C-TCR–transgenic thymocytes reactive to systemically expressed self-antigens (Sha et al., 1988) might be affected by Xcl1 deficiency. We found that the deletion of 2C-TCR⁺CD4⁺CD8⁺ thymocytes in negatively selecting H-2^k/d mice appeared undisturbed in the absence of XCL1 (Fig. 4 B). We further analyzed the role of XCL1 in the negative selection of thymocytes reactive to tissue-restricted antigens that are promiscuously expressed by mTECs. tDCs are known to cooperate with mTECs in the negative selection of thymocytes reactive to tissue-restricted antigens (Gallegos and Bevan, 2004). The rat insulin promoter (RIP)–driven transgene of membrane-bound OVA (mOVA) is expressed in the thymus by mTECs and deletes OVA-reactive OT-I-TCR and OT-II-TCR–transgenic thymocytes (Kurts et al., 1996). We found in irradiation-induced bone marrow chimera experiments that both OT-I-TCR–transgenic CD4⁻CD8⁺ and OT-II-TCR–transgenic CD4⁺CD8⁻ thymocytes were deleted in the thymic microenvironments of RIP-mOVA–transgenic mice even in the absence of XCL1 (Fig. 4 C), indicating that XCL1 is not needed for the negative selection of thymocytes reactive to the tissue-restricted antigen. These results collectively suggest that the role of XCL1 in negative selection is minor and dispensable.

It was also shown that tDCs play a role in the thymic generation of nT reg cells (Watanabe et al., 2005; Proietto et al., 2008; Hanabuchi et al., 2010). We thus examined whether the generation of nT reg cells in the thymus might be affected in Xcl1-deficient mice. We found that CD4⁺CD8⁻CD25⁺Foxp3⁺ T reg cells in the thymus of Xcl1-deficient adult mice showed reductions in frequency, as well as in absolute number, when compared with those in the thymus of WT mice (Fig. 5, A and B). The absolute numbers of thymic nT reg cells were significantly (P < 0.05) reduced in Xcl1-deficient mice throughout ontogeny from the day of birth to 3 mo old (Fig. 5 B). In contrast, the absolute numbers of other populations of thymocytes, including CD4⁻CD8⁻ double-negative, CD4⁺CD8⁺ double-positive, CD4⁺CD8⁻ CD4-single-positive, and CD4^-CD8⁺ CD8-single-positive thymocytes, were comparable between Xcl1-deficient mice and WT mice (Fig. S4 A), indicating that the cell number reduction in the thymus of Xcl1-deficient mice was specific for thymic nT reg cells. A majority of Foxp3⁺ nT reg cells in the thymus were localized in the medullary region (Fontenot et al., 2005; also shown in Fig. 5, C and D). The numbers per unit area of Foxp3⁺ cells in the middle medullary M region, as well as the peripheral medullary CMJ-M region, were significantly (P < 0.001) smaller in Xcl1-deficient mice than in WT mice (Fig. 5, C and D). Among Foxp3⁺ cells in the M and CMJ-M regions, the frequency of CD11c⁺ DC-attached Foxp3⁺ cells was severely reduced in Xcl1-deficient mice (Fig. 5, E and F), suggesting the role of tDCs in the XCL1-mediated regulation of nT reg cell generation in the thymus. CD4⁺CD8⁻CD25⁺ cells isolated from the thymus of Xcl1-deficient mice and WT mice comparably suppressed the anti-CD3–stimulated proliferation of CD4⁺CD8⁻CD25⁻ conventional T cells (Fig. 5 G), indicating that nT reg cells generated in the thymus of Xcl1-deficient mice are functional on a per cell basis but are defective in cell number. These results indicate that XCL1 is needed for the optimal generation of thymic nT reg cells.

Unlike the reduced number of nT reg cells in the thymus, the number of total Foxp3⁺ cells in the periphery was not reduced in Xcl1-deficient mice throughout ontogeny (Fig. S4, B-D). However, the number of thymic nT reg cell–derived spleen nT reg cells, which were identified by the expression of the nuclear factor Helios along with Foxp3 (Sugimoto et al., 2006; Thornton et al., 2010), was reduced in Xcl1-deficient mice (Fig. 5 H and Fig. S4 E), suggesting that XCL1 deficiency causes the reduction of nT reg cells in the periphery.

No signs of tissue inflammation were detected in Xcl1-deficient mice at 6 wk old and 3 mo old (unpublished data). However, we detected inflammatory lesions and autoantibody deposits in several tissues, including heart, liver, stomach, salivary glands, and lacrimal glands, in Xcl1-deficient mice at 12–18 mo old (unpublished data). Importantly, the intravenous transfer of thymocytes from Xcl1-deficient BALB/c background mice into athymic BALB/c-nu/nu mice caused severe lymphocyte infiltration and tissue damage of lacrimal glands (Fig. 6, A and B). The tissue lesions were prominent in lacrimal glands, weakly detectable in salivary glands, and not detected in heart, liver, stomach, pancreas, or kidney. The administration of equal numbers of WT BALB/c thymocytes along with thymocytes from Xcl1-deficient mice significantly reduced the lesions in lacrimal glands (Fig. 6, A and B), suggesting that nT reg cells generated in the normal thymus are capable of suppressing the dacryoadenitis that nT reg cells generated in the thymus of Xcl1-deficient mice fail to control. These results indicate that nT reg cell development is impaired in the thymus of Xcl1-deficient mice and that the thymocytes from Xcl1-deficient mice are potent in causing, and fail to regulate, autoimmune dacryoadenitis.

It was previously shown that Aire⁺ mTECs play a role in the thymic generation of T reg cells (Aschenbrenner et al., 2007). Finally, we wished to examine whether the expression of Aire might be affected in the thymus of Xcl1-deficient mice and whether the expression of XCL1 might be affected in the thymus of Aire-deficient mice. The density of Aire⁺ cells in the thymic medulla was comparable between Xcl1-deficient mice and WT mice (Fig. S4, F and G). Quantitative PCR analysis showed that the expression of Aire and Aire-dependent tissue-restricted antigens, such as salivary protein 1, was not diminished in mTECs isolated from Xcl1-deficient mice (unpublished data), indicating that XCL1 is not needed for the expression and function of Aire in mTECs.

In contrast, we found that the expression of Xcl1 transcripts was severely defective in mTECs that were isolated from Aire-deficient mice (Fig. 7 A). As has been recently reported (Laan et al., 2009), the expression of other chemokines, such as Ccl19, Ccl21, and Ccl25, in mTECs was also modulated by the Aire deficiency (Fig. 7 A). However, the reduction of Xcl1 in Aire-deficient mTECs was remarkably severe (the expression of which reduced to ∼1% of that observed in normal mTECs), which is in contrast to the relatively mild alterations in the expression of the other chemokines in Aire-deficient mTECs (15–200% of the amount in normal mTECs; Fig. 7 A).

The density of tDCs in the middle medullary M region was significantly (P < 0.01) decreased in Aire-deficient mice, whereas that in the CMJ-C region was significantly (P < 0.05) elevated in Aire-deficient mice (Fig. 7, B and C). The density of tDCs in the other regions (SCZ, C, and CMJ-M regions) was not significantly (P > 0.05) altered by the Aire deficiency (Fig. 7, B and C). The absolute numbers of tDCs and their subsets were not affected in Aire-deficient mice (Fig. 7 D). These results indicate that as in Xcl1-deficient mice, Aire-deficient mice exhibit defective accumulation of tDCs in the medullary region.

The density of Foxp3⁺ cells in the thymic medulla (CMJ-M and M regions) was accordingly and significantly (P < 0.05) lower in Aire-deficient mice than in WT mice (Fig. 7, E and F). Moreover, the absolute number of CD4⁺CD8⁻CD25⁺Foxp3⁺ cells in the thymus was significantly (P < 0.05) reduced in Aire-deficient mice (Fig. 7 G). These results indicate that Aire-deficient mice exhibit defective generation of nT reg cells in the thymus.

It is well known that tDCs are detected abundantly in the medullary region and sparsely in the cortical region (Barclay and Mayrhofer, 1981; Flotte et al., 1983; Kurobe et al., 2006). However, the molecular mechanisms that contribute to this medullary accumulation of tDCs are unknown. Screening for the chemokine receptors expressed by tDCs and their chemotactic responses to candidate ligands has revealed that the chemokine ligands for CCR7, CXCR4, and XCR1 are capable of attracting tDCs. By analyzing mice deficient for these chemokine ligands or their receptors, we identify that the XCL1–XCR1 chemokine axis plays a potent role in the medullary accumulation of tDCs. In contrast, we find that CCR7 and CXCR4 play a minor, if any, role in the accumulation of tDCs in the thymic medulla.

XCL1, which is also known as lymphotactin, ATAC, and SCM1, is the only member of the C family of chemokines encoded in the mouse genome (Kelner et al., 1994). XCL1 is produced by activated CD8⁺ T cells, Th1-polarized CD4⁺ T cells, and NK cells (Kelner et al., 1994; Hedrick et al., 1997; Dorner et al., 2002, 2004, 2009). A recent study shows that the XCL1 receptor, XCR1, is expressed by CD8⁺ DCs in the spleen and that the XCL1-XCR1–mediated interaction between CD8⁺ T cells and CD8⁺ DCs is important for the development of cytotoxicity of CD8⁺ T cells to antigens cross-presented by CD8⁺ DCs (Dorner et al., 2009). However, the functions of the XCL1–XCR1 chemokine axis in the thymus have remained unknown. This study reveals the role of the XCL1–XCR1 chemokine axis in the localization of tDCs in the thymic medulla.

In the thymus, XCR1 is exclusively expressed by tDCs. Among the tDC subpopulations, XCR1 is most highly detectable in I-A^highCD11c^highCD11b⁻ lymphoid DCs, a population that largely overlaps with the CD8⁺Sirpα⁻ conventional DC subpopulation of tDCs (Wu and Shortman, 2005). In contrast, XCL1 in the thymus is expressed in the class II MHC^high subpopulation of mTECs in addition to NK and NKT cells. The Xcl1 transcripts expressed by mTECs are in smaller amounts than those expressed by activated CD8⁺ T cells in the spleen but are prominent in the thymus when compared with those expressed by other cells in the thymus, including cTECs, thymic fibroblasts, tDCs, or CD45⁺ thymocytes. In Xcl1-deficient mice, tDCs fail to localize in the medullary region and are instead accumulated in the deep cortex and CMJ regions, although the numbers and subpopulations of tDCs are not altered. The mislocalization of tDCs in the deep cortex and CMJ regions of the thymus was detected in irradiated Xcl1-deficient mice that were reconstituted with normal bone marrow cells but not in irradiated normal mice that were reconstituted with Xcl1-deficient bone marrow cells (unpublished data). These results suggest that XCL1 produced by irradiation-resistant thymic stromal cells, rather than irradiation-sensitive bone marrow–derived thymocytes, contributes to the attraction of XCR1-expressing tDCs into the inner medullary region of the thymus. We think that mTECs, rather than hematopoietic cells including NK cells and NKT cells, in the thymus play a major role in the XCL1-mediated medullary localization of tDCs.

The arrest of tDCs in the deep cortex and CMJ regions in Xcl1-deficient mice suggests that the tDCs that are accumulated in the medullary region of normal mice may be derived from tDCs that are either generated in the deep cortex and CMJ regions of the thymus or migrate into the thymus from the circulation at the deep cortex and CMJ regions. The XCL1-mediated attraction of XCR1-expressing tDCs to the medulla may facilitate efficient interactions between mTECs and tDCs. It is also possible that the distribution of tDCs may be further regulated by the balance between the XCL1-mediated attraction to the medullary region and the attraction to the deep cortex and CMJ regions by unknown factors, and the loss of XCL1 may result in the aberrant attraction of tDCs to the deep cortex and CMJ regions.

Our results also reveal that Aire is essential for the production of XCL1 by mTECs. Aire is a nuclear factor expressed by a subpopulation of mTECs and is implicated for its role in the promiscuous gene expression of tissue-restricted self-antigens in mTECs (Derbinski et al., 2005; Mathis and Benoist, 2009). It is also suggested that Aire regulates the development of mTECs, thereby indirectly contributing to the promiscuous gene expression in mTECs (Gillard et al., 2007; Yano et al., 2008). Our results show that the expression of Xcl1 transcripts in mTECs isolated from Aire-deficient mice is severely reduced to ∼1% of the amount observed in WT mTECs. Accordingly, as in Xcl1-deficient mice, the medullary accumulation of tDCs is severely impaired in Aire-deficient mice and the tDCs are aberrantly arrested in the deep cortex and CMJ regions. Aire may regulate the development of XCL1-producing mTECs. Alternatively, Aire may promiscuously regulate the expression of Xcl1 gene in mTECs.

Concomitant with the failure in the medullary accumulation of tDCs, mice deficient for either Xcl1 or Aire exhibit a reduction in the number of T reg cells generated in the thymus. Most T reg cells in the thymus are generated in the medullary region (Fontenot et al., 2005) and it is suggested that mTECs (Aschenbrenner et al., 2007; Spence and Green, 2008), tDCs (Proietto et al., 2008; Hanabuchi et al., 2010), and their cooperation (Watanabe et al., 2005) contribute to the generation of T reg cells in the thymus. Our results indicate that the deficiency of either Xcl1 or Aire causes the reduced generation of T reg cells in the thymus. The cellularity of thymic T reg cells in Xcl1-deficient mice and Aire-deficient mice is approximately half of that in WT mice and is significantly low when compared with that in WT mice. However, the reduced numbers of thymic T reg cells are not significantly (P > 0.05) different between Xcl1-deficient mice and Aire-deficient mice. Our results also indicate that Aire deficiency causes severe loss of XCL1 expression by mTECs, whereas Xcl1 deficiency does not reduce Aire expression by mTECs. We therefore think that Aire is essential for the mTEC expression of XCL1, which attracts XCR1-expressing tDCs to the medullary region and contributes to the optimal generation of T reg cells in the thymus.

Proximal interactions between tDCs and mTECs in the thymic medulla may promote optimal generation of nT reg cells, possibly via the production of γc cytokines, including IL-2, IL-7, and thymic stromal lymphopoietin (Watanabe et al., 2005; Ziegler and Liu, 2006; Mazzucchelli et al., 2008; Vang et al., 2008). Among the tDC subpopulations, lymphoid DCs most highly express XCR1 and, thus, may play a major role in the interaction with mTECs and the generation of nT reg cells. Unlike tDCs, CD4⁺CD8⁻CD25⁺ thymic T reg cells do not express detectable Xcr1 transcripts (unpublished data), ruling out the possibility of a direct effect of XCL1 on T reg cells. Thus, our results suggest that XCL1-producing mTECs attract tDCs into the inner medullary region, thereby promoting the generation of nT reg cells, and that optimal nT reg cell development in the thymus requires a specialized medullary microenvironment that is formed by the XCL1-mediated attraction of tDCs by Aire-dependent mTECs. Whether XCL1-mediated tDC accumulation in the medulla affects TCR repertoire of nT reg cells remains unclear.

A recent work using CD11c-Cre mice crossed with mice that express diphtheria toxin A under the control of a loxP-flanked neomycin resistance cassette from the ROSA26 locus has shown that these DC-depleted mice are not defective in the generation of nT reg cells in the thymus (Ohnmacht et al., 2009), potentially contradicting our results indicating the role of tDCs in nT reg cell generation. However, that work also described that the depletion of tDCs is incomplete in the DC-depleted mice (Ohnmacht et al., 2009) but did not describe the intrathymic localization of the remaining tDCs. Thus, the incompletely depleted tDCs in the DC-depleted mice may be enriched in the inner medullary region and be sufficient for the unreduced generation of nT reg cells.

Our results show that the number of nT reg cells in the thymus is reduced in Aire-deficient mice. However, several studies have described that the generation of T reg cells is not defective in Aire-deficient mice (Anderson et al., 2002, 2005; Liston et al., 2003; Kuroda et al., 2005; Hubert et al., 2009), particularly in the spleen and the lymph nodes (Anderson et al., 2005; Kuroda et al., 2005; Hubert et al., 2009). Indeed, our results show that the number of T reg cells in Aire-deficient mice is reduced in the thymus to approximately half of that in WT mice but is not reduced in the spleen and the lymph nodes of Aire-deficient mice. It should be noted that previous studies have indicated that the number of T reg cells in the thymus is slightly reduced in Aire-deficient mice, although those studies concluded no loss of T reg cell generation in the thymus (Anderson et al., 2005; Kuroda et al., 2005; Hubert et al., 2009). We think that Aire indeed contributes to the optimal generation of nT reg cells in the thymus and that the generation of induced T reg cells, particularly in the periphery (Piccirillo and Shevach, 2004; Curotto de Lafaille and Lafaille, 2009), has veiled the reduced cellularity of nT reg cells in Aire-deficient mice.

The present results show that thymocytes from Xcl1-deficient mice are potent in eliciting severe lymphocyte infiltration and tissue damage of lacrimal glands in athymic nu/nu mice, indicating that the thymocytes generated in Xcl1-deficient mice fail to establish self-tolerance. The mixture of thymocytes from normal mice reduces the dacryoadenitis caused by the thymocytes from Xcl1-deficient mice, supporting the possibility that the thymocytes from Xcl1-deficient mice are potent in triggering, and fail to regulate, the autoimmunity and that nT reg cell development is impaired in the thymus of Xcl1-deficient mice. In a similar manner, Aire-deficient mice tend to exhibit inflammatory failure of exocrine tissues, including lacrimal glands (Anderson et al., 2002, 2005; Kuroda et al., 2005; Hubert et al., 2009). The commonness of the target organs may reflect a similarity in the breakdown of self-tolerance in Xcl1-deficient mice and Aire-deficient mice. Nonetheless, our results do not rule out the possibility that the absence of XCL1 in peripheral T cells also contributes to the onset of the dacryoadenitis.

Finally, our results reveal that the XCL1-mediated medullary accumulation of tDCs plays only a minor role in the negative selection of self-reactive thymocytes. In contrast, tDCs are known for their roles in negative selection, particularly in the intrathymic presentation of peripheral antigens, by cooperating with promiscuous gene expression in mTECs (Gallegos and Bevan, 2004; Koble and Kyewski, 2009; Nitta et al., 2009) and by transport from the circulation (Bonasio et al., 2006). The tDCs that are mislocalized in the CMJ regions in the absence of XCL1 may be sufficient for the cross-presentation of mTEC-expressed tissue-restricted antigens and the negative selection of developing thymocytes reactive to those self-antigens. In addition, recent studies indicated that DCs in the thymic cortex are capable of inducing the deletion of negatively selected thymocytes (McCaughtry et al., 2008) and that CCR2 is involved in the accumulation of CD8⁻Sirpα⁺ tDC subpopulation in the thymic cortex and the intrathymic negative selection against blood-borne antigens (Baba et al., 2009). Thus, the XCL1-mediated medullary accumulation is not needed for the deletion of the majority of negatively selected thymocytes.

In conclusion, the present results indicate that the XCL1–XCR1 chemokine axis contributes to the medullary accumulation of tDCs and the thymic development of nT reg cells. The results also suggest that Aire expressed in mTECs regulates the XCL1-mediated medullary accumulation of tDCs and that XCL1-mediated proximal interaction between tDCs and mTECs in the thymic medulla contributes to the thymic development of nT reg cells. Our results imply a novel role of Aire in regulating autoimmunity via the XCL1-mediated medullary accumulation of tDCs and that the breakdown of self-tolerance in Aire deficiency may involve the failure to localize tDCs in the medulla in an XCL1-dependent manner.

Xcl1-deficient mice were generated at Merck Research Laboratories (Fig. S1, A–C). Aire-deficient mice were generated at the University of Basel (Fig. S5). Ccr7-deficient mice (Förster et al., 1999), Mx-Cre x Cxcr4^flox/flox mice (Sugiyama et al., 2006), and plt/plt mice (Nakano et al., 1998), as well as OT-I and OT-II TCR-transgenic and RIP-mOVA–transgenic mice (Kurts et al., 1996; Barnden et al., 1998), were described previously. Mx-Cre x Cxcr4^flox/flox mice were injected with poly I poly C, and only mice that exhibited a nearly complete loss of CXCR4 genomic sequence and the undetectable expression of CXCR4 gene in hematopoietic cells were further analyzed (Sugiyama et al., 2006). Mice were maintained under specific pathogen-free conditions in our animal facility, and experiments were performed under the approval of the Institutional Animal Care Committee of the University of Tokushima.

Bone marrow cells were magnetically depleted of T cells using biotin-conjugated antibodies specific for CD4, CD8, and Thy1.2 and streptavidin-conjugated magnetic beads (Miltenyi Biotec). Recipient mice were injected with T cell–depleted bone marrow cells (4 × 10⁷) 1 d after 9.25 Gy x-ray irradiation. The mice were analyzed 4–5 wk after the reconstitution.

5 × 10⁷ thymocytes isolated from WT or Xcl1-deficient (Xcl1-KO) mice of BALB/c background were intravenously transferred into BALB/c-nu/nu mice. Where indicated, equal numbers of thymocytes (5 × 10⁷ + 5 × 10⁷) were mixed before the transfer. Various organs at 8 wk after the transfer were fixed with 4% phosphate-buffered formaldehyde, pH 7.2. Paraffin-embedded sections were stained with hematoxylin and eosin. Two pathologists independently evaluated the histology without being informed of the conditions of individual mice. Inflammatory lesions of the tissues were scored as previously described (Ishimaru et al., 2008) and as follows: 0 = no inflammation, 1 = 1–5 foci composed of >20 mononuclear cells per focus, 2 = >5 such foci but without significant parenchymal destruction, 3 = degeneration of parenchymal tissue, and 4 = extensive infiltration with mononuclear cells and extensive parenchymal destruction.

Multicolor flow cytometry analysis and cell sorting were performed using FACSCalibur and FACSAria II (BD). Intracellular staining of Foxp3 and Helios was performed according to the manufacturer’s instructions (eBioscience). Thymic stromal cells were prepared by digesting thymic fragments with collagenase, dispase, and DNase I (Roche) and enriched by depleting CD45⁺ cells with a magnetic cell sorter (Miltenyi Biotec) before cell sorting, as described previously (Gray et al., 2002). For DC analysis, thymus cells were prepared by digesting thymic fragments with collagenase D and DNase I.

Frozen thymus tissues embedded in OCT compound (Sakura) were sliced into 5-µm-thick sections, fixed with acetone, and stained with the following antibodies: mTEC-specific monoclonal antibody ER-TR5 (a gift from W. van Ewijk, Erasmus University, Rotterdam, Netherlands) followed by Alexa Fluor 633–conjugated anti–rat IgG antibody (Invitrogen); biotinylated UEA1 (Vector Laboratories) followed by Alexa Fluor 633–conjugated streptavidin (Invitrogen); FITC-conjugated anti-Foxp3 monoclonal antibody (eBioscience); biotinylated anti-CD11c or anti-CD11b monoclonal antibody (eBioscience) followed by Alexa Fluor 488–conjugated or Alexa Fluor 546–conjugated streptavidin (Invitrogen); and anti-Aire antibody (Santa Cruz Biotechnology, Inc.) followed by FITC-conjugated anti–rabbit IgG antibody (Invitrogen). Images were analyzed with a TSC SP2 confocal laser-scanning microscope and Confocal software version 2.6 (Leica).

10⁶ collagenase-digested thymus cells were placed in a Transwell chamber (6.5-mm diameter, 5-µm pore; Corning) that was inserted into a 100-nM chemokine-containing culture well. Cells were incubated for 2 h, counted, stained for CD11c and I-A^b, and analyzed by flow cytometry. Chemotactic index is the ratio of the numbers of CD11c⁺I-A^b+ DCs that migrated to the bottom of culture wells in the presence and absence of chemokines.

According to the methods previously reported (Tai et al., 2005), 5 × 10⁴ CFSE-labeled CD4⁺CD25⁻ lymph node T cells isolated from B6-Ly5.1 mice were cultured with 5 × 10⁴ CD25⁺CD4⁺CD8⁻ or CD25⁻CD4⁺CD8⁻ thymocytes in the presence of 10⁵ 20 Gy-irradiated T cell–depleted spleen cells from B6 mice and 1 µg/ml anti-CD3 monoclonal antibody (clone 2C11). Cells were harvested at 72 h and analyzed by flow cytometry.

Total cellular RNA was reverse transcribed with oligo-dT primer and Superscript III reverse transcription (Invitrogen). cDNA was PCR amplified, electrophoresed, and visualized with ethidium bromide. For quantitative analysis, real-time RT-PCR was performed using SYBR Premix Ex Taq (Takara Bio Inc.) and Light Cycler DX400 (Roche). Amplified products were confirmed to be single bands by gel electrophoresis and were normalized to the amount of HPRT products. Primer sequences are listed in Table S1.

Statistical comparison was performed with the Student’s t test (two-tailed) using Excel software (Microsoft).

Fig. S1 shows genomic structure of Xcl1-deficient mice. Fig. S2 shows analysis of CD11c⁺ cells in thymus sections. Fig. S3 shows negative selection of T cells in the periphery of Xcl1-deficient mice. Fig. S4 shows thymocytes, splenocytes, and Aire⁺ mTECs in Xcl1-deficient mice. Fig. S5 shows targeting strategy to generate Aire-Cre-GFP knockin mice. Table S1 shows primer sequences used for RT-PCR analysis.

We would like to express our special appreciation to the late professor Cunlan Liu for initiating the present study. We also would like to thank Tomoo Ueno and Fumi Saito for technical support in the laboratory and Kensuke Takada for reading the manuscript.

This study was supported by a MEXT Grant-in-Aid for Scientific Research on Priority Area “Immunological Self”, the JST Strategic International Cooperative Program, the Swiss National Science Foundation, the European Framework Programme 6, and the Max Planck Society.

The authors have no competing financial interests.