Antigen Recognition By Autoreactive Cd4+ Thymocytes Drives Homeostasis Of The Thymic Medulla

The thymic medulla is dedicated for purging the T-cell receptor (TCR) repertoire of self-reactive specificities. Medullary thymic epithelial cells (mTECs) play a pivotal role in this process because they express numerous peripheral tissue-restricted self-antigens. Although it is well known that medulla formation depends on the development of single-positive (SP) thymocytes, the mechanisms underlying this requirement are incompletely understood. We demonstrate here that conventional SP CD4+ thymocytes bearing autoreactive TCRs drive a homeostatic process that fine-tunes medullary plasticity in adult mice by governing the expansion and patterning of the medulla. This process exhibits strict dependence on TCR-reactivity with self-antigens expressed by mTECs, as well as engagement of the CD28-CD80/CD86 costimulatory axis. These interactions induce the expression of lymphotoxin α in autoreactive CD4+ thymocytes and RANK in mTECs. Lymphotoxin in turn drives mTEC development in synergy with RANKL and CD40L. Our results show that Ag-dependent interactions between autoreactive CD4+ thymocytes and mTECs fine-tune homeostasis of the medulla by completing the signaling axes implicated in mTEC expansion and medullary organization.


Introduction
The thymus ensures the generation of a self-tolerant T cell receptor (TCR) repertoire. Tolerance to self-antigens (Ags) is established in the medulla, a specialized microenvironment mainly composed of medullary thymic epithelial cells (mTECs) and dendritic cells (DCs). mTECs are critical for inducing selftolerance because they constitute a thymic reservoir of numerous peripheral tissue-restricted self-Ags (TRAs) [1,2]. Defective TRA expression results in autoimmunity [3]. mTECs can present TRAs to autoreactive CD8 + and CD4 + SP thymocytes to promote their deletion or the generation of natural regulatory T cells [4,5,6]. TRAs expressed by mTECs are also captured by thymic DCs, which help to purge the thymocyte repertoire of autoreactive TCR specificities [4]. Negative selection is further reinforced by circulating DCs displaying self-Ags captured in the periphery [7]. The medulla thus ensures the establishment of T-cell tolerance via tight collaboration between mTECs and DCs [8].
Mice presenting a block in thymocyte development at the double-positive (DP) stage exhibit small scattered medullary islets [9,10,11], indicating that medulla formation requires SP thymo-cytes [12]. SP thymocytes and mTECs thus engage in a bidirectional ''crosstalk'' that controls formation and organization of the medulla [12,13] as well as thymocyte deletion. We recently discovered that autoreactive CD4 + thymocytes play a privileged role in governing the development of the mTEC subset that displays a fully mature phenotype, which constitutes approximately 20% of the total mTEC population [14]. However, the nature of the cellular interactions and molecular mechanisms that controls development and architectural organization of the entire medullary compartment remains poorly understood.
Three members of the tumor necrosis factor receptor superfamily and their ligands have been implicated in mTEC development [13,14,15,16,17,18]. Cooperation between the engagement of receptor activator of nuclear factor Kappa B (RANK) and CD40 on mTECs by RANKL and CD40L expressed by SP thymocytes determines mature mTEC cellularity in adult mice [16] while lymphotoxin b receptor (LTbR) and LTa1b2 (LT) influence medulla organization [19,20,21]. However, the respective roles of these three signaling axes in medulla development, and the sequence in which they operate, remain elusive.
We demonstrate here that conventional autoreactive CD4 + thymocytes are essential and sufficient for fostering total mTEC expansion and determining the 3-dimensional (3D) organization of the entire medullary compartment. This critical function of autoreactive CD4 + thymocytes is mediated by Ag-dependent interactions with mTECs displaying cognate Ag-MHCII complexes and engagement of the CD28-CD80/86 costimulatory axis. This crosstalk induces LTa expression in the CD4 + thymocytes and RANK in mTECs, thereby completing two of the three signaling axes regulating medulla formation and organization. Finally, we show that the crosstalk between autoreactive CD4 + thymocytes and mTECs is a dynamic homeostatic process that governs the remarkable plasticity of the postnatal medulla, allowing it to adapt its size and organization to the output of autoreactive thymocytes.

RT-PCR
Real-time RT-PCR was performed as described [14]. GAPDH mRNA was used for normalization. Primer sequences are available upon request.
These results demonstrate that CD4 + thymocytes are required and sufficient for sustaining medulla formation by controlling total mTEC cellularity. This selective dependence on CD4 + thymocytes is specific of mTECs, as a block in CD4 + thymocyte development has no adverse effects on DCs.
The role of autoreactive CD4 + thymocytes in driving mTEC cellularity was further assessed by generating mixed BM chimeras in which irradiated WT males were reconstituted with variable mixtures of BM from WT and Marilyn:Rag2 2/2 males ( Figure 2C).
Reconstitution with Marilyn:Rag2 2/2 BM increased mTEC cellularity to greater than WT levels in a dose dependent manner. This was associated with a significant increase in proliferating Ki67 + mTECs ( Figure 2D), suggesting that autoreactive CD4 + thymocytes amplify mTEC cellularity directly by stimulating their proliferation.
We next studied medulla formation in OTII:Rag2 2/2 mice, which express an MHCII-restricted TCR specific for ovalbumin (OVA), and OTII:Rag2 2/2 mice carrying a RIP-mOVA transgene driving the synthesis of membrane-bound OVA in mTECs [40]. Medullary areas were increased only slightly in OTII:Rag2 2/2 mice relative to Rag2 2/2 controls. They were however expanded markedly in RIP-mOVA:OTII:Rag2 2/2 mice ( Figure 3A). This was accompanied by a strong increase in total mTEC numbers ( Figure 3B). CD4 + thymocytes in both OTII:Rag2 2/2 and RIP-mOVA:OTII:Rag2 2/2 mice upregulated expression of the chemokine receptor CCR7, enabling their migration into the medulla, where its ligand CCL21 is produced (data not shown). Positive selection also induced similar upregulation in TCR and CD3 expression in the two strains of mice (data not shown). Differential medulla expansion in OTII:Rag2 2/2 and RIP-mOVA:OTII:Rag2 2/2 mice could thus not be explained by differences in migration of CD4 + thymocytes into the medulla or in TCR upregulation upon positive selection.
We previously established that the development of mature Aire + mTECs could be driven by WT CD4 + thymocytes in reaggregated thymic organ culture (RTOC) experiments [14]. We therefore used RTOC to determine whether Ag-specific interactions with CD4 + thymocytes were required for promoting mTEC development in vitro. 2-dGUO-treated fetal thymic stroma was reaggregated with OTII:Rag2 2/2 thymocytes in the presence or absence of OVA peptide (OVAp) ( Figure 3C-D). mTEC development remained at baseline levels in RTOCs performed with OTII thymocytes in the absence of OVAp. In contrast, the addition of OVAp induced strong increases in the frequencies and numbers of total and mature mTECs, reaching levels equivalent to those induced by WT thymocytes.
To determine whether medullary defects in TCR-transgenic mice can be corrected in vivo by providing the cognate Ag, we injected OVAp into adult OTII:Rag2 2/2 mice. After 5-7 days, medullary areas were strongly enlarged compared to PBS-injected controls ( Figure 4A and S3A). This was accompanied by marked increases in total mTEC numbers ( Figure 4B) and frequencies of proliferating Ki67 + mTECs ( Figure 4C). OVAp injection also induced strong increases in mature mTECs, with enhanced Aire and Aire-dependent TRA expression ( Figure S3B). 3D reconstructions of thymic lobes confirmed that medullary volume was increased in the OVAp-injected mice (1.4%) compared to PBSinjected controls (0.7%) ( Figure S4A, movies S1, S2). In the OVAp-injected mice, individual medullary islets were reduced in number and had a larger size distribution ( Figure S4B). OVAp injection thus led to marked remodeling of the 3D organization of the medulla. Similar results were obtained by injecting the H-Y Dby peptide (H-Yp) into Marilyn:Rag2 2/2 females ( Figure 4D, Figure S4C-D, movies S3, S4).
Taken together, these results establish that autoreactive CD4 + thymocytes control development and organization of the thymic medulla in an Ag-dependent manner. This dynamic homeostatic process operates in adult mice and controls mTEC proliferation, medullary growth and 3D patterning of the medulla.

LTbR-signaling controls mTEC cellularity in synergy with RANK and CD40
Although it has been established that RANKL-RANK, CD40L-CD40 and LT-LTbR signaling contribute to mTEC development [14,15,16,17,19], the respective role of these three signaling axes remain unclear. To investigate this question, 2-dGUO-treated thymic lobes were cultured with agonistic anti-LTbR antibodies, CD40L and/or RANKL ( Figure 5). Anti-LTbR antibodies induced a significant increase in total mTEC numbers. Little increase was induced by RANKL and CD40L, alone or in combination. Synergistic increases in mTEC numbers were observed when anti-LTbR antibodies were combined with RANKL, CD40L or both. Combining all 3 stimuli induced total mTEC numbers equivalent to those induced by thymocytes. A different division of labor emerged when mature mTECs were quantified ( Figure 5A-B). RANKL and CD40L each induced a modest increase in the proportion of mature mTECs. A synergistic increase was obtained when RANKL and CD40L were added together, attaining control levels induced by thymocytes. In contrast, anti-LTbR antibodies had little or no influence on mature mTEC frequencies, either when added alone or in combination with RANKL, CD40L or both. Signaling via LTbR, CD40 and RANK thus make distinct contributions to mTEC cellularity and maturation. LTbR-signaling has a dominant role in determining total mTEC cellularity. Conversely, CD40 and RANK are critical for driving the development of mature mTEC.

LT expression is regulated in CD4 + thymocytes by Agspecific interactions with mTECs
To determine whether Ag-specific interactions with autoreactive CD4 + thymocytes might play a dominant role in inducing signals governing mTEC development, RANKL, CD40L, LTa and LTb mRNAs were quantified by qRT-PCR in DP and CD4 + thymocytes from OTII:Rag2 2/2 and Rip-mOVA:OTII:Rag2 2/2 mice. In both strains, RANKL and CD40L mRNAs were upregulated strongly in CD4 + thymocytes relative to DP thymocytes. LTb mRNA was also upregulated, albeit only ,2fold, in CD4 + thymocytes from both strains ( Figure 6A). RANKL, CD40L and LTb mRNAs were thus induced in positively-selected OTII thymocytes independently of OVA expression. In contrast, LTa mRNA expression was upregulated only in CD4 + thymocytes of Rip-mOVA:OTII:Rag2 mice, suggesting that its induction occurs upon Ag-specific interactions with mTECs.

Discussion
We investigated the respective contributions of CD4 + and CD8 + thymocytes to formation and patterning of the thymic medulla. Mice lacking CD8 + thymocytes exhibited normally sized medullary islets and WT mTEC numbers. In contrast, mice lacking CD4 + thymocytes exhibited poorly developed medullary islets and strongly reduced mTEC cellularity. CD4 + thymocytes are thus essential and sufficient for driving proper medulla growth, whereas CD8 + thymocytes are dispensable for sustaining this process. Despite this privileged role of CD4 + thymocytes, medulla formation was severely impaired in MHCII-restricted TCRtransgenic mice in which positive selection of CD4 + thymocytes occurs normally while there is no negative selection since the cognate Ag is absent. RTOC experiments also demonstrated that OTII thymocytes were unable to drive mTEC development in the absence of OVAp. Positively-selected CD4 + thymocytes are thus not sufficient per se for sustaining medulla formation. Instead, several lines of evidence indicate that medullary growth requires Ag-specific interactions between autoreactive CD4 + thymocytes and mTECs, similar to those that induce negative selection. First, OTII thymocytes can promote mTEC development in RTOCs as efficiently as WT thymocytes when OVAp is provided. Second, whereas mTEC cellularity was strongly decreased in OTII:Rag2 2/2 mice and Marilyn:Rag2 2/2 females, this defect was corrected in RIP-mOVA:OTII:Rag2 2/2 mice and Marilyn:-Rag2 2/2 males, which express the cognate Ags. Third, medullary defects in OTII:Rag2 2/2 mice and Marilyn:Rag2 2/2 females could be corrected by injecting OVAp and H-Yp, respectively. The latter finding shows that defective medulla formation in these mice is not simply due to an intrinsic developmental block. It also emphasizes the remarkable plasticity of the medulla and indicates that the control of its size by autoreactive CD4 + thymocytes is a dynamic homeostatic process operating in adult mice. This is consistent with studies indicating that medullary defects in adult SCID mice can be corrected by injecting mature T cells [43,44]. Fourth, rare CD4 + thymocytes with non-transgenic TCR specificities in OTII:Rag +/+ mice suffice to sustain medulla formation, suggesting that this process is sensitive to low numbers of autoreactive CD4 + thymocytes. This is consistent with the fact that SP thymocytes reside in the medulla for 4-5 days and are highly mobile, favoring numerous encounters with mTECs [45,46].
We exploited OTII:Rag2 2/2 and RIP-mOVA:OTII:Rag2 2/2 mice to determine whether upregulation of LT, RANKL and CD40L expression in SP CD4 + thymocytes might be driven by Ag-specific interactions with mTECs. LTb, RANKL and CD40L mRNAs were upregulated independently of OVA expression in both OTII:Rag2 2/2 and RIP-mOVA:OTII:Rag2 2/2 CD4 + thymocytes. Conversely, increased LTa mRNA expression was dependent on OVA expression by mTECs, as it was only observed in RIP-mOVA:OTII:Rag2 2/2 mice. Upregulation of LT expression by TCR stimulation was confirmed in vitro by activating OTII thymocytes with anti-CD3/CD28 antibodies or by co-culture with OVAp-loaded mTECs. These results suggest that expression of RANKL and CD40L is induced by positive selection in the cortex whereas that of LT is activated by subsequent Ag-specific interactions with mTECs ( Figure 6H). Activation of LT expression is thus uncoupled physically and temporally from the induction of RANKL and CD40L expression.
Cooperation between RANKL and CD40L signaling has been shown to be critical for mTEC development in the postnatal thymus [16,17]. However, RANKL and CD40L expression by CD4 + thymocytes is not sufficient because mTEC development is severely impaired in OTII:Rag2 2/2 mice even though their thymocytes express both ligands. This could be reconciled by three non-mutually-exclusive mechanisms. First, efficient delivery of RANKL and CD40L signals to mTECs may require stable Agdriven contacts with CD4 + thymocytes. Second, effective control of mTEC development by RANKL and CD40L requires collaboration with LT expression, which is induced in CD4 + thymocytes by Ag-dependent interactions with mTECs. Finally, LT produced by autoreactive CD4 + thymocytes enhances the responsiveness of mTECs to RANKL signals by inducing RANK expression in mTECs ( Figure 6H).
Our FTOC experiments indicated that LT, RANKL and CD40L signals make differential contributions to mTEC expansion and maturation. LTbR signaling was critical for fostering an increase in mTEC cellularity but had little effect on mTEC maturation. A key role of LTbR signaling in mTEC expansion is consistent with the observation that LTbR 2/2 and LTa 2/2 mice have small medullas [19,20,21] whereas LT over-expression in T cells leads to drastic medulla enlargement [47]. In contrast to LTbR signaling, synergy between RANKL and CD40L was essential for driving mTEC maturation rather than increasing mTEC numbers. This is consistent with studies showing that mTEC maturation requires cooperation between RANKL and CD40L and that mTEC cellularity is decreased only modestly in RANKL 2/2 and CD40 2/2 mice [16,17]. In conclusion, at least four parameters -medullary size, mTEC numbers, 3D organization of the medulla and mTEC maturationare modulated in adult mice by Ag-specific and costimulatorymolecule-dependent interactions between autoreactive CD4 + thymocytes and mTECs displaying auto-Ag-MHCII complexes (Fig. 6H). These interactions induce LT expression by autoreactive CD4 + thymocytes and RANK expression by mTECs, thereby completing key signaling axes required for medulla formation (Fig. 6H). This unique crosstalk between autoreactive CD4 + thymocytes and mTECs regulates a homeostatic fine-tuning process that controls mTEC cellularity and 3D organization of the postnatal medulla, thereby maintaining a medullary microenvironment that is optimal for ensuring central T-cell tolerance. Figure S1 mTECs but not DCs are impaired in mice lacking CD4 + thymocytes. (A) Representative FACS profiles for the expression of Aire and CD80 by CD45 2 EpCAM + Ly51 2/lo mTECs from WT, b2m 2/2 , H2-Aa 2/2 and CIIta IV-/IVmice: numbers represent the percentages of cells within the indicated gates. Graphs show number of Aire + , CD80 hi , CD80 int and CD80 lo mTECs: the means and SD were derived from three measurements, each with three mice per genotype; statistical significance relative to WT. (B) Representative FACS profiles for the expression of CD11c and PDCA1 by CD45 + hematopoietic cells (top profiles), and the expression of Sirpa and CD8a by CD45 + CD11c hi cDCs (bottom profiles), are shown for thymi from WT, b2m 2/2 , H2-Aa 2/2 and CIIta IV-/IVmice. Graphs show numbers per thymus of CD11c int PDCA1 + pDCs, CD11c hi CD8alo Sirpa + cDCs and CD11c hi CD8a hi Sirpa 2 cDCs for the indicated genotypes: the means and SEM are derived from two experiments, each with four to seven mice per group. Numbers of mice is indicated. (C) Thymic sections from WT, b2m 2/2 , H2-Aa 2/2 and CIIta IV-/IVmice were stained for the CD11c marker: m denotes the medulla. Results are representative of two experiments, each with two mice per group. (TIFF) Figure S2 Positively-selected CD4 + thymocytes in 3BBM74:Rag2 2/2 , B3K508:Rag1 2/2 and OTII:Rag2 2/2 are inefficient at inducing medulla expansion whereas low numbers of autoreactive CD4 + thymocytes are sufficient. (A) Thymic sections from 3BBM74:Rag2 2/2 , B3K508:Rag1 2/2 , OTII:Rag2 2/2 and WT mice were stained with antibodies against K8 and K14: m denotes the medulla; c denotes the cortex. Graphs show quantifications of the K14 + areas (left graph) and medullary areas (right graph) for each genotype: symbols represent individual confocal images; horizontal lines represent medians; data is pooled from three independent experiments, each with two to three mice per group. (B) Graphs show quantifications of the medullary areas, K14 + and MTS10 + areas in thymic sections from OTII:Rag2 2/2 , OTII:Rag2 +/+ and WT mice: symbols represent individual confocal images; horizontal lines represent medians; data was pooled from three independent experiments, each with three mice per group. (C) Representative FACS profiles are shown for the expression of Ly51 by CD45 2 EpCAM + Ly51 2/lo mTECs from OTII:Rag2 2/2 , OTII:Rag2 +/+ and WT mice: numbers indicate percentages of cells within the indicated gates.  . LT expression is induced by Ag-specific activation of CD4 + thymocytes. (A) RANKL, CD40L, LTa and LTb mRNAs were quantified in DP and CD4 + thymocytes from OTII:Rag2 2/2 and Rip-mOVA:OTII:Rag2 2/2 mice: means and SEM are from 3 experiments, each with 2 mice per group. (B) LTa mRNA and cell surface LT were assessed for unstimulated and anti-CD3/CD28-activated CD4 + thymocytes from OTII:Rag2 2/2 or Marilyn:Rag2 2/2 mice: data representative of 3 experiments. (C) LTa mRNA was quantified in CD4 + thymocytes from OTII:Rag2 2/2 mice co-cultured with unloaded (none) or OVAp-loaded mTECs: data representative of 2 experiments. (D) LTa mRNA was quantified in CD4 + thymocytes from OTII:Rag2 2/2 mice isolated 1.5 days after injection of PBS or OVAp: data representative of 3 experiments. (E) b-casein, CRP and RANK mRNAs were quantified in mTECs from WT, LTa 2/2 mice and OTII:Rag2 2/2 mice 5 days after injection of PBS or OVAp. (F) LTa mRNA was quantified in DP and CD4 + thymocytes from CD80/86 2/2 mice: means and SEM are derived from 2 experiments, each with 2 mice per group. (G) Graphs show distributions of medullary areas (mm 2 ) in WT, CD80/86 2/2 and CD28 2/2 thymi (left), and thymi from DT-treated WT and Foxp3-DTR mice (right): significance relative to WT. (H) Positive selection induces CD40L and RANKL expression in thymocytes. After migrating into the medulla, CD4 + thymocytes scan the surface of mTECs for the presence of auto-Ag-MHCII complexes. Ag-specific and CD28-CD80/86 dependent interactions between CD4 + thymocytes and mTECs induce the expression of LT in CD4 + thymocytes and RANK in mTECs, thereby completing the signaling axes required for promoting mTEC expansion and maturation. doi:10.1371/journal.pone.0052591.g006 OTII:Rag2 2/2 mice injected with PBS or OVA 323-339 . Horizontal lines represent medians and SEM. Numbers of isolated medullary islets in the thymic lobes are indicated below. The 3D reconstructions can be visualized in movies S1, S2. (C) Female Marilyn:Rag2 2/2 mice were injected i.v with PBS or H-Y Dby peptide. 5 days later, 3D reconstructions of thymic lobes were then generated from serial sections stained with antibodies against K14 and DAPI. 3D reconstructions depicting the thymic lobes (light blue, DAPI) and medulla (red, K14) are shown: axes are in mm. Volumes and percentages of the thymic medulla are indicated. (D) The graph depicts the volumes (mm 3 ) of individual medullary islets in thymic lobes from Marilyn:Rag2 2/2 mice injected with PBS or Dby H-Y peptide. Horizontal lines represent medians and SEM. Numbers of isolated medullary islets in the thymic lobes are indicated below. The 3D reconstructions can be visualized in movies S3, S4. (TIFF)

Supporting Information
Movie S1 3D reconstruction of an entire thymic lobe from OTII:Rag2 2/2 mouse injected with PBS. Thymic and medullary volumes were reconstructed using a whole thymic lobe. 20 mm thick serial sections were stained with DAPI (light blue) and antibodies against K14 (red). Volume rendering was performed from epifluorescence images, as described in methods. All axes are graduated in mm.

(AVI)
Movie S2 3D reconstruction of an entire thymic lobe from OTII:Rag2 2/2 mouse injected with the OVA 323-339 peptide. Volumes were determined as described for Movie S1. (AVI) Movie S3 3D reconstruction of an entire thymic lobe from female Marilyn:Rag2 2/2 mouse injected with PBS. Volumes were determined as described for Movie S1. (AVI) Movie S4 3D reconstruction of an entire thymic lobe from female Marilyn:Rag2 2/2 mouse injected with the H-Y Dby peptide. Volumes were determined as described for Movie S1. (AVI)