Lymphotoxin α fine-tunes T cell clonal deletion by regulating thymic entry of antigen-presenting cells

Medullary thymic epithelial cells (mTEC) purge the T cell repertoire of autoreactive thymocytes. Although dendritic cells (DC) reinforce this process by transporting innocuous peripheral self-antigens, the mechanisms that control their thymic entry remain unclear. Here we show that mTEC-CD4+ thymocyte crosstalk regulates the thymus homing of SHPS-1+ conventional DCs (cDC), plasmacytoid DCs (pDC) and macrophages. This homing process is controlled by lymphotoxin α (LTα), which negatively regulates CCL2, CCL8 and CCL12 chemokines in mTECs. Consequently, Ltα-deficient mice have increased expression of these chemokines that correlates with augmented classical NF-κB subunits and increased thymic recruitment of cDCs, pDCs and macrophages. This enhanced migration depends mainly on the chemokine receptor CCR2, and increases thymic clonal deletion. Altogether, this study identifies a fine-tuning mechanism of T cell repertoire selection and paves the way for therapeutic interventions to treat autoimmune disorders.

T hymic clonal deletion, called negative selection, prevents the generation of autoreactive T cells that could induce autoimmunity. The thymus is subdivided into a medulla surrounded by a cortex, both compartments dedicated to specific selection processes 1,2 . The medulla has a key function in purging the T cell repertoire of self-reactive specificities, through the large diversity of self-antigens (Ag) expressed by medullary thymic epithelial cells (mTEC) [3][4][5][6] . The cortex also promotes the deletion of autoreactive T cells 7,8 . It has been estimated that~75% of negatively selected cells are deleted at the double-positive (DP) stage in the cortex, and that~25% are deleted at the singlepositive (SP) stage in the medulla 7 . Dendritic cells (DC) are involved in this process since their constitutive ablation results in impaired clonal deletion and fatal autoimmunity 9 . DCs are involved in the deletion of both DP thymocytes in the cortex and SP thymocytes in the medulla 10 . They constitute a heterogeneous population comprising three distinct subsets: CD11c int BST-2 hi plasmacytoid DCs (pDC) and two CD11c hi conventional DC (cDC) subsets-CD11c hi CD8α hi SHPS-1 − resident and CD11c hi CD8α lo SHPS-1 + migratory cDCs 11,12 . Intrathymically derived resident cDCs, located in close proximity to mTECs, possess the ability to cross-present self-Ags, expressed by mTECs, to thymocytes [13][14][15] . Although mTECs express a large array of self-Ags that critically contributes to negative selection, they cannot cover the whole spectrum of self-Ags expressed in peripheral tissues 5,16 . Migratory cDCs and pDCs reinforce the deletion of autoreactive thymocytes by continuously migrating from the blood to the thymus, where they display peripheral self-Ags that would be otherwise not presented to thymocytes [17][18][19][20][21] .
While the migration of cDCs and pDCs in the thymus was described to depend on CCR2 and CCR9, respectively 18,19 , the implication of the thymic microenvironment and more specifically that of mTEC-thymocyte crosstalk in this process remains unknown. Furthermore, although thymic macrophages constitute another type of Ag-presenting cells (APC) that has long been associated with the clearance of apoptotic thymocytes 22,23 , the mechanisms that sustain their thymic entry as well as their respective contribution in clonal deletion remain elusive.
Here, we show that Ag-specific interactions between mTECs and CD4 + thymocytes regulate the thymic entry of peripheral SHPS-1 + cDCs, pDCs and F4/80 + CD11b + macrophages. This phenomenon is tightly controlled by the tumour necrosis factor (TNF) member, lymphotoxin α (LTα), induced in CD4 + thymocytes upon crosstalk, which represses CCL2, CCL8 and CCL12 expression in CD80 lo mTECs. We observed that increased expression of these chemokines in CD80 lo mTECs from Ltα −/− mice (Ltα −/− mTEC lo ) correlates with an upregulation of c-Rel and p65 classical nuclear factor-kappa B (NF-κB) subunits, previously described to regulate CCL2 and CCL8 [24][25][26] . Enhanced thymic recruitment of peripheral DCs and macrophages in Ltα −/− mice is drastically reduced in the absence of the chemokine receptor CCR2. We also show that CCR1 and CCR5 control the thymic pool of these cell types, but at a lesser extent than CCR2. Importantly, we demonstrate that LTα-regulated DC and macrophage thymus homing fine tunes the deletion of autoreactive thymocytes in both the cortex and medulla. Finally, we show that migratory cDCs and macrophages are more competent in vivo for deleting autoreactive thymocytes than pDCs, a phenomenon accentuated on a Ltα-deficient background. Altogether, this study reveals an unexpected role for mTEC-thymocyte crosstalk in controlling the thymic entry of peripheral DCs and macrophages. This process, tightly regulated by LTα, which in turn represses CCL2, CCL8 and CCL12 expression, substantially impinges clonal deletion.
LTα limits APC thymic entry through CCR2 ligands in mTECs. We investigated the underlying mechanism(s) by which mTEC-CD4 + thymocyte crosstalk regulates the thymic recruitment of peripheral DCs and macrophages. Since three TNF members, RANKL, CD40L and LTα1β2, are implicated in mTECthymocyte crosstalk 12,28-31 , we examined whether they could be regulated upon Ag-specific interactions with mTECs. In contrast to Tnfsf11 (RANKL) and Cd40lg (CD40L), we found that only Lta (LTα) was upregulated in OTII CD4 + thymocytes co-cultured with OVA 323-339 -loaded mTECs (Fig. 2a), suggesting that LTα could be involved in the thymic entry of peripheral APCs mediated by mTEC-CD4 + thymocyte crosstalk. To investigate this hypothesis, we first evaluated on thymic sections the area occupied by CD11c + cells in the cortex and medulla and found that DC enrichment in the medulla was increased in Ltα −/− mice ( Supplementary Fig. 6a). Consistently, flow cytometry analyses revealed that Ltα −/− mice had increased frequencies and numbers of thymic cDCs and pDCs compared with wild-type (WT) mice (Fig. 2b). Furthermore, although numbers of CD8α hi SHPS-1 − resident cDCs were unaltered, numbers of CD8α lo SHPS-1 + migratory cDCs were increased in these mice (Fig. 2c). Frequencies and numbers of F4/80 + CD11b + macrophages were also higher in Ltα −/− thymi than in WT thymi (Fig. 2d), with a distribution largely increased in the medulla ( Supplementary Fig. 6b). In line with a previous study 32 , we observed that CD8α lo SHPS-1 + cDCs and pDCs were more proliferative than CD8α hi SHPS-1 − cDCs ( Supplementary Fig. 6c). Similar frequencies of proliferating Ki-67 + SHPS-1 + cDCs, pDCs and macrophages in WT and Ltα −/− mice indicate that the elevated numbers of these cell types in Ltα −/− mice were not due to To demonstrate this hypothesis, sublethally irradiated WT and Ltα −/− recipients were adoptively transferred with CD45.1 nucleated blood cells (Fig. 2e). Three days later, we found increased frequencies and numbers of CD45.1 donor cells in Ltα −/− thymi ( Fig. 2f) that contained increased numbers of cDCs, pDCs and macrophages (Fig. 2g, h). Similar results were obtained by adoptively transferring CD45.1 splenic DC and macrophageenriched cells into nonirradiated WT and Ltα −/− recipients ( Supplementary Fig. 7). Accordingly, a substantial reduction in numbers of circulating SHPS-1 + cDC, pDC and macrophage was observed in the blood of Ltα −/− mice (p < 0.001 for SHPS-1 + cDCs, p < 0.01 for pDCs and p < 0.05 for macrophages by unpaired Student's t-test, Supplementary Fig. 8a-c). Consistently with previous observations 33 , these cell types were also reduced in the spleen of these mice (Supplementary Fig. 8d-f). This demonstrates that increased numbers of DCs and macrophages are limited to the thymus and result from an enhanced recruitment, indicating that LTα constitutes a feedback repressor of APC thymic entry. We next hypothesised that LTα could regulate the expression of key chemokines involved in the thymus homing of DCs and macrophages. Because CCR9 has been implicated in the thymic entry of peripheral pDCs 18 , we first assessed the expression of its ligand Ccl25, which was unaltered in the total thymus and purified mTECs from Ltα −/− mice (Fig. 3a). Since we and others previously showed that CCL2 overexpression in the thymus leads to increased thymic pDC cellularity and that its receptor CCR2 is involved in thymic SHPS-1 + cDC and pDC homeostasis 19,34,35 , we hypothesised that LTα could modulate CCL2 expression. Strikingly, Ccl2 expression was substantially higher in both total thymus (p < 0.05 by a two-tailed Mann-Whitney test) and purified mTECs (p < 0.05 by a two-tailed Mann-Whitney test) in Ltα −/− mice than in WT mice (Fig. 3b). CCL2 is a major ligand for CCR2, which has other potential ligands, such as CCL7, and Ccl12 mRNAs were measured by qPCR in purified mTECs from WT (n = 5) and Ltα −/− (n = 5) mice. d Ccl2, Ccl8 and Ccl12 mRNAs were measured by qPCR in purified mTECs from OTII:OTII (n = 6), OTII:RipmOVA (n = 6) and OTIIxLtα −/− :RipmOVA (n = 6) chimeras. e Ccl2, Ccl8 and Ccl12 mRNAs were measured by qPCR in purified mTECs loaded (n = 6) or not (n = 6) with OVA 323-339 peptide co-cultured with OTII CD4 + thymocytes in the presence or not of recombinant LTβR-Fc chimera. f Ccl2, Ccl8 and Ccl12 mRNAs were measured by qPCR in purified mTECs loaded with OVA 323-339 peptide and cocultured with CD4 + thymocytes from OTII-Rag2 −/− (n = 9) or OTII-Rag2 −/− xLtα −/− (n = 9) mice. g Lta mRNA was measured by qPCR in purified OTII CD4 + thymocytes co-cultured with mTECs (n = 7), thymic SHPS-1 + cDCs (n = 8), SHPS-1 -cDCs (n = 8) and pDCs (n = 8) loaded or not with OVA 323-339 peptide. a-g Data are representative of two independent experiments. Error bars show mean ± SEM, *p < 0.05, **p < 0.01, ***p < 0.001 using two-tailed Mann-Whitney test for b-d, and unpaired Student's t-test for e-g NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-03619-9 ARTICLE NATURE COMMUNICATIONS | (2018) 9:1262 | DOI: 10.1038/s41467-018-03619-9 | www.nature.com/naturecommunications CCL8 and CCL12 36,37 . Although CCL8 has been detected in the thymus 19 , CCL7 and CCL12 have not been yet described to be expressed in this tissue. In contrast to Ccl7, we found that Ccl8 and Ccl12 expression was also increased in Ltα −/− mTECs (Fig. 3c). Considering that LTα is induced upon crosstalk ( Fig. 2a), we next examined whether Ccl2, Ccl8 and Ccl12 expression in mTECs could be regulated by crosstalk with OTII CD4 + thymocytes. The expression of these three ligands was increased in mTECs from OTII:RipmOVA mice compared with OTII:OTII mice (Fig. 3d), which was even more pronounced in OTII:RipmOVA mice backcrossed on a Ltα −/− background, indicating that LTα represses mTEC ability to express these chemokines upon crosstalk. To further determine the role of direct Ag-specific interactions with CD4 + thymocytes in this chemokine expression, WT mTECs loaded or not with OVA 323-339 peptide were co-cultured with OTII CD4 + thymocytes. Ccl2, Ccl8 and Ccl12 were upregulated in OVA 323-339loaded mTECs compared with unloaded mTECs (Fig. 3e).
Moreover, the addition of a soluble LTβR-Fc chimera, which blocks LTα1β2/LTβR interactions, resulted in a more pronounced upregulation of these chemokines, indicating that LTα1β2/LTβR axis acts as a negative regulator of these chemokines upon mTEC-CD4 + thymocyte crosstalk. We also found higher levels of Ccl2, Ccl8 and Ccl12 in mTECs co-cultured with CD4 + thymocytes from OTIIxLta −/− compared with those from OTII mice, suggesting that LTα, specifically in CD4 + thymocytes, controls the expression of these chemokines (Fig. 3f). Moreover, although DCs can cross-present Ags expressed by mTECs 13,14 , the chemokine upregulation observed in co-cultures with OTII CD4 + thymocytes, in absence of DCs (Fig. 3e), indicates that interactions with CD4 + thymocytes are sufficient to induce CCL2, CCL8 and CCL12 expression in mTECs. Using an antibody allowing the detection of MCP1-4 (i.e., CCL2, CCL7, CCL8 and CCL13, the latter being not expressed in mice), we found that in contrast to mTECs, thymic DC subsets did not express detectable levels of these chemokines in OTII-Rag2 −/− and RipmOVAxOTII-Rag2 −/− mice as well as in WT and Ltα −/− mice ( Supplementary Fig. 9). This suggests that DCs do not possess the ability to attract peripheral APCs through the production of these chemokines. Finally, in contrast to OVA 323-339 -loaded SHPS-1 + cDCs, SHPS-1 − cDCs and pDCs, only OVA 323-339 -loaded mTECs were able to induce Lta expression in CD4 + thymocytes, excluding a potential implication of DCs in the regulation of these chemokines through LTα induction (Fig. 3g). Altogether, these data show that LTα represses CCL2, CCL8 and CCL12 expression induced in mTECs upon crosstalk with CD4 + thymocytes.
LTα-regulated thymic entry of APCs depends on CCR2. CCL2, CCL8 and CCL12 chemokines are known ligands for CCR2 36 , and CCL8 is also a ligand for CCR1 and CCR5 36,37,38,39 . To investigate a potential involvement of these chemokine receptors in the thymic entry of peripheral DCs and macrophages, we first examined their expression in blood-derived SHPS-1 + cDCs, pDCs and macrophages in WT and Ltα −/− mice. All these cell types significantly expressed CCR2 in these mice (p < 0.05 by twotailed Mann-Whitney test, Supplementary Fig. 10), indicating that they possess the ability to migrate in a CCR2-dependent manner into the thymus. CCR1 and CCR5 were also weakly detectable in these three cell types, although macrophages had high levels of CCR1 in both mice. To determine CCR2, CCR1 and CCR5 contributions in regulating the thymic pool of SHPS-1 + cDCs, pDCs and macrophages, we generated mixed BM chimeras, in which lethally irradiated CD45.1xCD45.2 WT recipients were reconstituted with CD45.1 WT BM cells together with either CD45.2 WT, Ccr2 RFP/RFP , Ccr1 −/− or Ccr5 −/− BM cells (ratio 1:1) (Fig. 4a). Six weeks later, thymic SHPS-1 + cDCs, pDCs and macrophages were analysed ( Supplementary Fig. 11). Strikingly, we found strongly reduced frequencies of these three cell types derived from CD45.2 Ccr2 RFP/RFP BM cells compared with those derived from CD45.2 WT BM cells (Fig. 4b). While macrophages were unaffected, we also found a slight reduction in SHPS-1 + cDCs and pDCs derived from CD45.2 Ccr1 −/− BM and in these three cell types derived from CD45.2 Ccr5 −/− BM cells.
These results indicate that compared with CCR1 and CCR5, CCR2 is a key regulator of the thymic pool of SHPS-1 + cDCs, pDCs and macrophages. Finally, to firmly demonstrate that enhanced thymic entry of DCs and macrophages in Ltα −/− mice was mediated by CCR2, blood nucleated cells from CD45.1 WT mice and Ccr2 RFP/RFPdeficient mice (ratio 1:1) were co-transferred into sublethally irradiated WT and Ltα −/− recipients (Fig. 4c). Consistently with our adoptive transfer (AT) experiments in Ltα −/− mice (Fig. 2e, f), frequencies and numbers of CD45.2 − RFP − cells corresponding to CD45.1 donor cells ( Supplementary Fig. 12a) were more elevated in Ltα −/− than in WT recipients (Fig. 4d). Among total donor cells, we found increased frequencies and numbers of cDCs, pDCs and macrophages ( Supplementary Fig. 12b, c). In contrast, increased thymus homing of total donor cells and of these three cell types in Ltα −/− mice was strongly impaired when donor cells were of Ccr2 RFP/RFP -deficient origin (Fig. 4d, e). Altogether, these data reveal that LTα controls the thymus homing of APCs in a CCR2-dependent manner.
Induction of CCR2 ligands and NF-κB subunits in Ltα −/− mice. We next investigated by which mechanisms CCL2, CCL8 and CCL12 chemokines are overexpressed in Ltα −/− mTECs. This cell type can be subdivided into two main subsets based on CD80 level 40 (Fig. 5a). Consistently with previous studies 41, 42 , Ltα −/− mice have normal frequencies and numbers of CD80 lo (mTEC lo ) and CD80 hi (mTEC hi ) mTECs, suggesting that increased expression in these chemokines (Fig. 3b, c) was not due to increased mTEC numbers. We found, by qPCR and flow cytometry, that CCR2 ligands were specifically upregulated in Ltα −/− mTEC lo (Fig. 5b, c). CCL2 and CCL8 are known to be regulated by the classical NF-κB pathway in different cell types [24][25][26] . Notably, p65 binding to the mouse Ccl2 promoter is involved in CCL2 expression 43,44 . We identified two putative NF-κB binding sites for c-Rel and p65, by in silico analysis, in the Ccl12 promoter region (Supplementary Table 1), suggesting that this gene could be also regulated by the classical NF-κB pathway. The level of p65 phosphorylation at serine 536 (ser536), which is associated with the upregulation of CCL2 45,46 , was unaltered in Ltα −/− mTEC lo (Fig. 5d). We next assessed whether Ltα −/− mTEC lo have a differential usage in the classical and non-classical NF-κB pathways, the latter known to be preferentially induced by LTα1β2/LTβR axis 6,47 . We found at mRNA and protein levels that the non-classical NF-κB subunit Relb (RelB) was decreased whereas classical NF-κB subunits Rel (cRel) and Rela (p65) were enhanced in Ltα −/− mTEC lo (Fig. 5e-g).
We next analysed the effect of LTα1β2/LTβR axis upon Agspecific interactions with CD4 + thymocytes in the regulation of NF-κB subunits in mTEC lo that express the LTβR receptor (Fig. 5h). Interestingly, mTECs co-cultured with OTIIxLtα −/− CD4 + thymocytes had reduced levels of Relb compared with mTECs co-cultured with OTII CD4 + thymocytes (Fig. 5i). In contrast, increased expression of Rel and Rela correlates with CCL2, CCL8 and CCL12 overexpression in these cells (Fig. 3f,  Fig. 5i). Thus, the disruption of the LTα1β2/LTβR axis in the context of Ag-specific interactions with CD4 + thymocytes leads to the upregulation of cRel and p65 classical NF-κB subunits and CCL2, CCL8 and CCL12 chemokines, suggesting that the chemokine upregulation in Ltα −/− mTECs is controlled by the overexpression of classical NF-κB subunits.
LTα-regulated APC thymic entry fine-tunes clonal deletion. Since thymus homing of peripheral DCs and macrophages was enhanced in Ltα −/− mice (Fig. 2e-h), we next investigated its impact on clonal deletion. Interestingly, numbers of DP (CD4 + CD8 + ), CD4 lo CD8 lo and CD4 + SP (CD4 + CD8 − ) cells were significantly reduced in Ltα −/− mice compared to WT mice (Fig. 6a, b). We also observed reduced numbers of CCR7 − cortical and CCR7 + medullary CD4 + SP (Fig. 6c). This cannot be explained by defective thymus homing of early thymic progenitors, since we previously observed that their numbers were normal in Ltα −/− thymi 42 . We thus analysed strongly autoreactive thymocytes, based on the expression of the Ikaros family transcription factor Helios and PD-1, as previously   (Fig. 6a). We found reduced numbers of autoreactive Helios + PD-1 − cells in CCR7 − and CCR7 + CD4 + SP cells in Ltα −/− mice, suggesting that clonal deletion was enhanced both in the cortex and medulla in these mice (Fig. 6d, e). Furthermore, we found reduced numbers of cortical Helios + PD-1 + CD4 lo CD8 lo post-positively selected cells (Fig. 6f). Similar results were observed when Ltα −/− BM cells were transplanted into CD45.1 WT recipients ( Supplementary Fig. 13), suggesting that LTα expression in haematopoietic cells controls the deletion of cortical and medullary thymocytes. An enhanced clonal deletion in Ltα −/− mice is supported by reduced frequencies and numbers of CD69 − CD62L + mature CD4 + SP in these mice (Fig. 6g). Therefore, these data suggest that enhanced thymus homing of DCs and macrophages in Ltα −/− mice leads to increased clonal deletion.
High ability of cDCs and macrophages for clonal deletion. We next assessed the tissue distribution of SHPS-1 + cDCs, pDCs and macrophages by transferring cells sorted from CCR2 RFP/+ heterozygous mice (Fig. 8a), allowing us to track their thymic entry with the red fluorescent protein (RFP) reporter gene. For better detection, cell-sorted peripheral DCs and macrophages from CCR2 RFP/+ donor mice were adoptively transferred into Ltα −/− recipients, in which thymus homing of these cells is increased (Fig. 2e-h, Supplementary Fig. 14a, b). Consistent with our AT experiments of CD45.1 donor cells (Figs. 1e-h, 2e-h, 4c-e, 7d, e), we found that these three cell types efficiently homed into the Ltα −/− thymus and retained their phenotypic hallmarks, as revealed by CD11c, BST-2 and F4/ 80 staining (Fig. 8b). Interestingly, RFP + cDCs and pDCs were preferentially located in the cortex, whereas RFP + F4/80 + macrophages were similarly distributed in the cortex and medulla.
To investigate the respective role of these three cell types in clonal deletion, WT BM-derived cDCs, pDCs and macrophages, expressing CCR2 and MHCII molecules, were generated (Fig. 8c,  Supplementary Fig. 14c-e). This indicates that these cell types should be competent to migrate in a CCR2-dependent manner and present Ags via MHCII molecules. The same number of OVA 323-339 -loaded BM-derived cDCs, pDCs or macrophages was first adoptively transferred into OTII-Rag2 −/− mice (Fig. 8d). OVA 323-339 -loaded BM-derived cDCs and pDCs were able to delete autoreactive thymocytes in OTII-Rag2 −/− mice compared to non-injected OTII-Rag2 −/− controls (Fig. 8e). Whereas the role of thymic macrophages in clonal deletion remains largely elusive, we found that OVA 323-339 -loaded BM-derived macrophages were able to delete in vivo both DP and Vα2 + Vβ5 + CD4 + SP thymocytes (Fig. 8e). This is consistent with the observation that adoptively transferred macrophages were localised in both the cortex and the medulla (Fig. 8b), where DP and SP thymocytes are respectively eliminated 7 . Importantly, BMderived cDCs and macrophages were more efficient than BMderived pDCs in deleting total thymocytes, including DP and Vα2 + Vβ5 + CD4 + SP cells. Consistently, BM-derived cDCs and macrophages expressed higher levels of MHCII than BMderived pDCs (Fig. 8c). Furthermore, thymic SHPS-1 + cDCs express higher levels of MHCII than pDCs, and are thus more prone to clonal deletion 20 . OVA 323-339 -loaded BM-derived cDCs, pDCs and macrophages deleted more efficiently total thymocytes, DP and Vα2 + Vβ5 + CD4 + SP cells in OTII-Rag2 −/− x Ltα −/− than in OTII-Rag2 −/− mice (Fig. 8e), which is consistent with the superior ability of these three cell types to home into the thymus on a Ltα −/− background ( Fig. 2e-h, Fig. 7d, e). Altogether these data indicate that migratory cDCs and macrophages have a higher capacity to delete autoreactive thymocytes than pDCs, a phenomenon accentuated in Ltα −/− mice.

Discussion
Thymic clonal deletion crucially prevents the generation of hazardous autoreactive T cells that could induce autoimmunity. mTECs are essential in this process through their ability to widely express self-Ags that can be cross-presented by resident cDCs [13][14][15]48,49 . Peripheral DCs, by continuously migrating into the thymus, also contribute to the deletion of autoreactive thymocytes by sampling peripheral self-Ags [17][18][19] . Although migratory DCs are involved in this tolerogenic process, the implication of mTECthymocyte crosstalk in regulating their thymic recruitment remained unknown so far.
We provide strong evidence that Ag-specific interactions between mTECs and CD4 + thymocytes regulate the thymic entry of peripheral DCs and macrophages. We found increased numbers of SHPS-1 + cDCs, pDCs and macrophages in the thymus of RipmOVAxOTII-Rag2 −/− and OTII:RipmOVA mice compared with OTII-Rag2 −/− and OTII:OTII mice, respectively. Furthermore, thymus homing of these cell types was enhanced in RipmOVAxOTII-Rag2 −/− recipients upon AT of donor cells. Although RANKL and CD40L are implicated in mTECthymocyte crosstalk 12 , these two TNF members were expressed at similar extents in CD4 + thymocytes from OTII-Rag2 −/− and RipmOVAxOTII-Rag2 −/− mice, indicating that they are unlikely responsible for the increased thymic entry of peripheral APCs observed in RipmOVAxOTII-Rag2 −/− mice. Nevertheless, we cannot exclude a potential role of these TNF members in other aspects of thymic DC biology. Future investigations are expected to clarify this issue. We show that this regulatory mechanism of peripheral APC recruitment is tightly controlled by LTα, which is specifically induced in autoreactive CD4 + thymocytes upon crosstalk with mTECs 50 . Of note, LTα was shown to be expressed ARTICLE in SP thymocytes as a membrane anchored LTα1β2 heterocomplex, which binds to LTβR 30,47 . Our data show that LTα negatively regulates the expression of CCL2, CCL8 and CCL12 chemokines in mTECs, which attenuates thymus homing of APCs, mainly in a CCR2-dependent manner. Consequently, migratory cDCs, pDCs and macrophages were increased in Ltα −/− compared to WT thymi. Contrarily to Ltα −/− mice, Ltbr −/− mice have reduced numbers of thymic SHPS-1 − resident cDCs and pDCs 51 . It remains nevertheless unclear whether LTβR controls the thymic pool of macrophages and peripheral APC entry. These differences are not surprising, since these mice have distinct defects in medulla organisation, mTEC subsets and autoimmunity 12 . Here, we demonstrate by AT of donor cells both in sublethally irradiated and unmanipulated mice, that LTα negatively regulates the thymus homing not only of peripheral cDCs and pDCs but also of macrophages. Our results, based on OTII:OTII and OTII:RipmOVA mTECs and mTEC-CD4 + thymocyte co-cultures, strongly indicate that Ccl2, Ccl8 and Ccl12 were upregulated in mTECs upon Agspecific interactions with CD4 + thymocytes. This upregulation was negatively controlled by LTα, specifically in CD4 + thymocytes, since it was exacerbated in absence of LTα or when LTα1β2/LTβR interactions were blocked. Furthermore, CCL2, CCL8 and CCL12 were specifically upregulated in Ltα −/− mTEC lo . Although mTEC lo were initially described to contain precursors giving rise to functional mTEC hi [12], they were reported to contain mTECs, expressing the chemokine CCL21 and the transcription factor FEZF2 implicated in the attraction of thymocytes into the medulla and in the expression of self-Ags, respectively 52,53 . Our study revealed that mTEC lo also express chemokines involved in the thymic entry of peripheral APCs, further highlighting a functional role of this subset in T cell selection. We further show that LTα controls mTEC lo properties by regulating CCL2, CCL8 and CCL12. This chemokine upregulation was likely due to a differential usage in the classical and non-classical NF-κB pathways, which are known to function as an interconnected signalling network rather than two independent pathways, even in mTECs 6,54,55 . Whereas in WT conditions, the lymphotoxin signalling activates preferentially the non-classical NF-κB pathway 6,47 , our data indicate a preferential usage of the classical NF-κB pathway, characterised by an upregulation of cRel and p65 when the lymphotoxin signalling is disrupted. Since CCL2 and CCL8 are regulated by classical NF-κB members [24][25][26] , which are overexpressed in Ltα −/− mice, this chemokine upregulation observed in Ltα −/− mTEC lo is likely mediated by the classical NF-κB pathway. Chromatin immunoprecipitation assays are expected to prove a direct regulation of these chemokines by classical NF-κB members. However, the mTEC lo subset represents a rare cell type (~2 × 10 5 cells/thymus) rendering technically difficult such an approach. Furthermore, although DCs can crosspresent Ags expressed by mTECs 13,14 , it is unlikely that they participate in CCL2, CCL8 and CCL12 expression since cocultures of CD4 + thymocytes with mTECs alone was sufficient to induce these chemokines. Furthermore, DCs did not express CCR2 ligands in the context of mTEC-CD4 + thymocyte crosstalk or upon LTα-mediated regulation. Another hypothesis would be that DCs could be indirectly involved in this regulation by inducing LTα in CD4 + thymocytes. We showed that in contrast to mTECs, neither thymic SHPS-1 + , SHPS-1 − cDCs nor pDCs were able to induce LTα in CD4 + thymocytes, excluding a potential contribution of DCs in the expression of these chemokines through LTα upregulation. Nevertheless, we cannot definitively rule out a role of DC-mediated indirect antigen presentation in the thymic recruitment of peripheral APCs. Experiments based on the use of DC-depleted mice in the context of thymic crosstalk are expected to clarify this issue in the next future.
Mixed BM chimera and AT experiments demonstrated that CCR2 plays a major role in the thymus homing of SHPS-1 + cDCs, pDCs and macrophages. Although a drastic reduction (~80-95%) of these three cell types was observed in the Ccr2 RFP/RFP -deficient donor group, a moderate but nonnegligible decrease of thymic SHPS-1 + cDCs, pDCs and macrophages was observed in Ccr1 −/− and Ccr5 −/− groups. While pDCs have been reported to migrate through CCR9 18 , our data thus show that they also use CCR2. Moreover, co-AT of WT and Ccr2 RFP/RFP -deficient donor cells indicates that although the thymic entry of SHPS-1 + cDCs is impaired in the absence of CCR2, this process is not fully abrogated, suggesting the implication of other chemokine receptor(s). Potential candidates could be CCR1 and CCR5, since thymic SHPS-1 + cDCs were moderately disturbed in Ccr1 −/− and Ccr5 −/− donor groups. Future experiments, based on the analysis of Ccr2 and Ccr5 as well as Ccr2 and Ccr1 double-deficient mice, are expected to clarify this issue.
Interestingly, since negatively selected thymocytes do not directly die, but instead remain viable for few hours in the medulla 56 , it is likely that autoreactive thymocytes have sufficient time to provide instructive signals to mTECs, that would regulate the thymic recruitment of peripheral DCs and macrophages. Interestingly, we demonstrate that this regulation loop controls the clonal deletion of autoreactive T cells ( Supplementary  Fig. 15). Autoreactive thymocytes were highly deleted at the DP, CD4 lo CD8 lo and CD4 + SP stages in Ltα −/− mice, indicating that enhanced clonal deletion occurs in both the cortex and medulla.
By AT experiments, we demonstrated that cDCs, pDCs and macrophages were all able to delete efficiently DP and SP cells, a process accentuated on a Ltα −/− background. According to the type of negatively selected cells, SHPS-1 + cDCs and pDCs were preferentially localised in the cortex, whereas macrophages were distributed throughout the thymus. Macrophages have been associated with the clearance of apoptotic bodies in the thymus 22,23 . Although they share common hallmarks with DCs, by expressing MHCII and CD80/CD86 molecules, implicated in T cell selection, their role in clonal deletion and the mechanisms that sustain their thymic entry remain unknown. So far, only one group has reported that F4/80 + CD11b + macrophages are able to delete autoreactive thymocytes in vitro by using reaggregated thymic organ cultures 57 . Here, we unravel that peripheral macrophages migrate in a CCR2-dependent manner into the thymus and that they may play an unsuspected role in clonal deletion.
Finally, from a therapeutic perspective, because Ltα deficiency increases DC and macrophage thymic entry, it would be interesting to determine whether LTα loss can protect and treat from autoimmunity. Generating an inducible transgenic mouse model allowing LTα deletion at a specific time point, i.e., before or after the development of autoimmune signs, would be useful to define the potential of LTα as a new target to prevent or treat autoimmunity associated with defective T cell selection.
In sum, this complex cellular interplay between mTECs, CD4 + thymocytes, peripheral DCs and macrophages constitutes a fine- Fig. 8 Migratory cDCs and macrophages are more efficient than pDCs for clonal deletion, a process accentuated on a Ltα −/− background. a Experimental setup: Ltα −/− recipients were injected with cell-sorted cDCs, pDCs or macrophages from Ccr2 RFP/+ heterozygous mice. The thymic localisation of these cell types was analysed three days later on thymic sections. b Thymic sections were stained with antibodies against the medulla specific marker K14 (magenta) and CD11c (green), BST-2 (green) or F4/80 (green). Adoptively transferred RFP + cells were detected in red. m and c denote the medulla and the cortex, respectively. The graph shows the ratio of medullary vs. cortical density of adoptively transferred cells. Twenty-five sections derived from two mice for each genotype were quantified for each condition. Scale bar: 100 µm. c CCR2 and MHCII in BM-derived cDCs, pDCs and macrophages from WT mice were analysed by flow cytometry. Histograms show the MFI values of CCR2 and MHCII expression normalised to the FMO value of each population analysed. Data are representative of two independent experiments (n = 3 mice per group and per experiment). FMO fluorescence minus one, MFI mean fluorescence intensity. d Experimental setup: AT of purified OVA 323-339 -loaded BM-derived cDCs, pDCs or macrophages into OTII-Rag2 −/− or OTII-Rag2 −/− xLtα −/− recipients. Clonal deletion was analysed 3-5 days after AT. e Numbers of total thymic cells, DP and Vβ5 + Vα2 + CD4 + SP cells were analysed for each condition. UT untreated OTII-Rag2 −/− mice. Data are representative of three independent experiments (n = 4 mice per group and per experiment). MΦ macrophage. Error bars show mean ± SEM, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001 using unpaired Student's t-test for b, twotailed Mann-Whitney test for c and one-tailed Mann-Whitney test for e tuning mechanism that allows the thymus to adapt its capacity of deleting autoreactive T cells to physiological and pathological fluctuations. This study should open new therapeutic perspectives for autoimmune disorders, based on the deletion of hazardous T cells via the manipulation of thymic entry of peripheral DCs and macrophages.
Confocal image analysis. Confocal microscopy images were segmented using homemade Matlab scripts. Twenty-five images for Ltα −/− thymic sections with AT of purified DCs and macrophages from Ccr2 RFP/+ mice and six to eleven images for WT and Ltα −/− thymic sections were analysed. Briefly, images were sequentially treated by: (i) median filtering over 2 pixels, (ii) thresholding, using the automatic threshold value (according to Otsu method, as implemented in the Matlab function graythresh), (iii) smoothing (dilatation followed by erosion) over 3 pixels, and (iv) background subtraction, with the background estimated by median filtering of the colocalization image over 32 pixels. For cells expressing RFP, colocalization between the green (endogenous DC or macrophage staining) and red channels (RFP) was estimated as the minimum intensity of the images of both channels. This ensured validating signals arising only from high staining for both colours before proceeding to cell detection. In all cases, cells were detected from the resulting binary images. The medulla was detected from the K14 staining by median filtering over 7 pixels, automatically thresholding, filling holes and smoothing over 30 pixels. Defining a binary mask corresponding to the medulla and to the cortex allowed automated counting of cells in each compartment. Cell density was computed as cell number over surface (d = N/S), for medulla and cortex, allowing computing the ratio of medulla versus cortex density (ratio = d medulla /d cortex ).
Quantitative RT-PCR. Total RNA was prepared with TRIzol reagent (Invitrogen). cDNA was synthesised with oligo dT primers (Life Technologies) and Superscript II reverse transcriptase (Invitrogen). PCR was performed using SYBR Premix Ex Taq (Takara) with the ABI Prism 7500 Fast PCR System (Applied Biosystems). Actin mRNA was used for normalisation. Primers are listed in Supplementary  Table 3.
Analysis for NF-κB binding sites in the Ccl12 promoter. In silico analysis of the mouse Ccl12 promoter was performed using the software MatInspector (https:// www.genomatix.de). The promoter region from −1000 to +100 bp of the transcription start site was searched for NF-κB binding sites using the Matrix family library version 10.0.
Statistics. All data are presented as means ± standard error of mean (SEM). Statistical analysis was performed with GraphPad Prism 7.03 software using unpaired Student's t-test for normal distribution or Mann-Whitney test. ****p < 0.0001; ***p < 0.001; **p < 0.01; *p < 0.05. Normal distribution of the data was assessed using d'Agostino-Pearson omnibus normality test.
Data availability. The authors declare that the data supporting the findings of this study are available within the article and its supplementary information files, or are available upon reasonable requests to the authors.