Thymocytes trigger self-antigen-controlling pathways in immature medullary thymic epithelial stages

Interactions of developing T cells with Aire+ medullary thymic epithelial cells expressing high levels of MHCII molecules (mTEChi) are critical for the induction of central tolerance in the thymus. In turn, thymocytes regulate the cellularity of Aire+ mTEChi. However, it remains unknown whether thymocytes control the precursors of Aire+ mTEChi that are contained in mTEClo cells or other mTEClo subsets that have recently been delineated by single-cell transcriptomic analyses. Here, using three distinct transgenic mouse models, in which antigen presentation between mTECs and CD4+ thymocytes is perturbed, we show by high-throughput RNA-seq that self-reactive CD4+ thymocytes induce key transcriptional regulators in mTEClo and control the composition of mTEClo subsets, including Aire+ mTEChi precursors, post-Aire and tuft-like mTECs. Furthermore, these interactions upregulate the expression of tissue-restricted self-antigens, cytokines, chemokines, and adhesion molecules important for T-cell development. This gene activation program induced in mTEClo is combined with a global increase of the active H3K4me3 histone mark. Finally, we demonstrate that these self-reactive interactions between CD4+ thymocytes and mTECs critically prevent multiorgan autoimmunity. Our genome-wide study thus reveals that self-reactive CD4+ thymocytes control multiple unsuspected facets from immature stages of mTECs, which determines their heterogeneity.


Introduction
The thymic medulla ensures the generation of a self-tolerant T-cell repertoire (Klein et al., 2014;Lopes et al., 2015).By their unique ability to express tissue-restricted self-antigens (TRAs) (Derbinski et al., 2001;Sansom et al., 2014), medullary thymic epithelial cells (mTECs) promote the development of Foxp3 + regulatory T cells and the deletion by apoptosis of self-reactive thymocytes capable of inducing autoimmunity (Klein et al., 2019).The expression of TRAs that mirrors body's self-antigens is controlled by Aire (Autoimmune regulator) and Fezf2 (Fez family zinc finger 2) transcription factors (Anderson et al., 2002;Takaba et al., 2015).Aire-dependent TRAs are generally characterized by a repressive chromatin state enriched in the trimethylation of lysine-27 of histone H3 (H3K27me3) histone mark (Handel et al., 2018;Org et al., 2009;Sansom et al., 2014).In accordance with their essential role in regulating the expression of TRAs, Aire -/-and Fezf2 -/-mice show defective clonal deletion of autoreactive thymocytes and develop signs of autoimmunity in several peripheral tissues (Anderson et al., 2002;Takaba et al., 2015).
Based on the level of the co-expressed MHC class II and CD80 molecules, mTECs were initially subdivided into mTEC lo (MHCII lo CD80 lo ) and mTEC hi (MHCII hi CD80 hi ) (Gray et al., 2006).The relationship between these two subsets has been established with reaggregate thymus organ cultures in which mTEC lo give rise to mature Aire + mTEC hi (Gäbler et al., 2007;Gray et al., 2007).Although mTEC hi express a highly diverse array of TRAs under Aire's action that releases stalled RNA polymerase and modulates chromatin accessibility, mTEC lo already express a substantial amount of TRAs (Derbinski et al., 2005;Giraud et al., 2012;Koh et al., 2018;Kyewski and Klein, 2006;Sansom et al., 2014).Recent single-cell transcriptomic analyses indicate that the heterogeneity of mTECs, especially in the mTEC lo compartment, is more complex than previously thought (Irla, 2020;Kadouri et al., 2020).mTEC lo with low or no expression of CD80 have been shown to be divided into three main subsets: CCL21 + mTECs, implicated in the attraction of positively selected thymocytes in the medulla (Lkhagvasuren et al., 2013), involucrin + TPA hi post-Aire mTECs corresponding to the ultimate mTEC differentiation stage (Metzger et al., 2013;Michel et al., 2017;Nishikawa et al., 2010), and the newly reported tuft-like mTECs that show properties of gut chemosensory epithelial tuft cells expressing the doublecortin-like kinase 1 (DCLK1) marker (Bornstein et al., 2018;Miller et al., 2018).Based on single-cell transcriptomic analyses, mTECs were then classified into four major groups encompassing mTEC I:CCL21 + mTECs, mTEC II:Aire + mTECs, mTEC III:post-Aire mTECs, and mTEC IV:tuft-like mTECs (Bornstein et al., 2018).Furthermore, mTEC lo with intermediate levels of CD80 and MHCII lie into mTEC single-cell clusters that are defined as proliferating and maturational, expressing Fezf2 and preceding the Aire + mTEC hi stage (Baran-Gale et al., 2020;Dhalla et al., 2020).These transit-amplifying cells were recently referred as to as TAC-TECs (Wells et al., 2020).
In the postnatal thymus, while mTECs control the selection of thymocytes, conversely CD4 + thymocytes control the cellularity of Aire + mTEC hi by activating RANK and CD40-induced NF-κb signaling pathways (Akiyama et al., 2008;Hikosaka et al., 2008;Irla, 2020;Irla et al., 2008).These bidirectional interactions between mTECs and thymocytes are commonly referred to as thymic crosstalk (Lopes et al., 2015;van Ewijk et al., 1994).However, it remains unknown whether CD4 + thymocytes act exclusively on mature Aire + mTEC hi or upstream on their TAC-TEC precursors contained in mTEC lo and whether the development of the newly identified Fezf2 + , post-Aire, and tuft-like subsets is regulated or not by CD4 + thymocytes.
In this study, using high-throughput RNA-sequencing (RNA-seq), we show that self-reactive CD4 + thymocytes induce in mTEC lo pivotal transcriptional regulators for their differentiation and function.Accordingly, self-reactive CD4 + thymocytes control the composition of the mTEC lo compartment, that is the precursors of Aire + mTEC hi , post-Aire cells, and tuft-like mTECs.Our data also reveal that self-reactive CD4 + thymocytes upregulate in mTEC lo the expression of TRAs, chemokines, cytokines, and adhesion molecules involved in T-cell development.This gene activation program correlates with increased levels of the active trimethylation of lysine-4 of histone 3 (H3K4me3) mark, including the loci of Fezf2-dependent and Aire/Fezf2-independent TRAs, indicative of an epigenetic regulation for their expression.Finally, we demonstrate that disrupted MHCII/TCR interactions between mTECs and CD4 + thymocytes lead to the generation of mature T cells containing self-specificities capable of inducing multiorgan autoimmunity.Altogether, our genome-wide study reveals that self-reactive CD4 + thymocytes control the developmental transcriptional programs of mTEC lo , which conditions their differentiation and function as inducers of T-cell tolerance.

CD4 + thymocytes induce key transcriptional programs in mTEC lo cells
Several NF-κb members are involved in Aire + mTEC hi development (Burkly et al., 1995;Lomada et al., 2007;Riemann et al., 2017;Shen et al., 2019;Zhang et al., 2006).However, it remains unclear whether the NF-κb or other signaling pathways are activated by CD4 + thymocytes specifically in mTEC lo cells.To investigate the effects of CD4 + thymocytes in mTEC lo , we used mice deficient in CD4 + thymocytes (ΔCD4 mice) because they lack the promoter IV of the class II transactivator (Ciita) gene that controls MHCII expression in cortical TECs (cTECs) (Waldburger et al., 2003).We first analyzed by flow cytometry the total and phosphorylated forms of IKKα, p65, and RelB NF-κb members and p38 and Erk1/2 MAPK proteins in mTEC lo from ΔCD4 mice according to the gating strategy shown in Figure 1-figure supplement 1A.Interestingly, the phosphorylation level of IKKα and p38 MAPK was substantially reduced in ΔCD4 mice (Figure 1A and B, Figure 1-figure supplement 2), indicating that CD4 + thymocytes may have an impact in mTEC lo by activating the IKKα intermediate of the nonclassical NF-κB pathway and the p38 MAPK pathway.
To gain insights into the effects of CD4 + thymocytes in mTEC lo , we analyzed by high-throughput RNA-seq the gene expression profiles of mTEC lo purified from WT and ΔCD4 mice (Figure 1-figure supplement 1B).We found that CD4 + thymocytes upregulated 989 genes (fold change [FC] >2) reaching significance for 248 of them (Cuffdiff p<0.05) (Figure 1C).957 genes were also downregulated (FC < 0.5) with 178 genes reaching significance (Cuffdiff p<0.05).We analyzed whether the genes significantly up-or downregulated by CD4 + thymocytes corresponded to TRAs, as defined by an expression restricted to 1-5 of peripheral tissues (Sansom et al., 2014).Interestingly, the genes upregulated by CD4 + thymocytes exhibited approximately fourfold more of TRAs over non-TRAs (Figure 1D, left panel).The comparison of the proportion of TRAs among the upregulated genes with those of the genome revealed a strong statistical TRA overrepresentation (p=5.2 × 10 -10 ) (Figure 1D, right panel).Most of the TRAs upregulated by CD4 + thymocytes were sensitive to the action of Aire (Aire-dependent TRAs) or controlled by Aire and Fezf2-independent mechanisms (Aire/ Fezf2-independent TRAs) (Figure 1E, Supplementary file 1).The upregulation of some of these TRAs by CD4 + thymocytes was confirmed by qPCR in mTEC lo purified from ΔCD4 mice (Figure 1F).The same results were observed with mTEC lo purified from MHCII -/-mice, also lacking CD4 + thymocytes, excluding any potential indirect effect of CIITA in the phenotype observed in ΔCD4 mice (Figure 1 Remarkably, among the non-TRAs upregulated by CD4 + thymocytes in mTEC lo , 37 corresponded to 50 mTEC-specific transcription factors that are induced by the histone deacetylase 3 (HDAC3) (Goldfarb et al., 2016;Figure 1G).Some of them, such as the interferon regulatory factor 4 (Irf4), Irf7, and the Ets transcription factor member, Spib, are known to regulate mTEC differentiation and function (Akiyama et al., 2014;Haljasorg et al., 2017;Otero et al., 2013).We also identified other transcription factors such as Nfkb2, Trp53, and Relb implicated in mTEC differentiation (Riemann et al., 2017;Rodrigues et al., 2017;Zhang et al., 2006).Finally, we found that CD4 + thymocytes upregulate in mTEC lo the expression of some cytokines and cell adhesion molecules such as integrins and cadherins (Figure 1H, Figure 1-figure supplement 3B).Given that mTEC lo are heterogeneous (Irla, 2020;Kadouri et al., 2020), we then analyzed whether the cytokines and adhesion molecules, which are upregulated by CD4 + thymocytes, are specific to a particular subset of mTEC lo .To this end, we reanalyzed single-cell RNA-seq data performed on total CD45 -EpCAM + TECs (Wells et al., 2020).Single cells were projected into a UMAP reduced-dimensional space and, using the 15 first principal components, six clusters were obtained, as in Wells et al., 2020 (Figure 1-figure supplement 4A).Wellestablished markers were used to distinguish the different TEC subsets such as Psmb11 and Prss16 for cTECs, Ccl21a and Krt5 for CCL21 + mTECs (also called mTEC I), Stmn1, Ska1, Fezf2 and Aire for TAC-TECs, Aire and Fezf2 for Aire + mTECs (also called mTEC II), Pigr and Cldn3 for post-Aire mTECs (also called mTEC III), and Avil and Pou2f3 for tuft-like mTECs (also called mTEC IV) (Figure 1-figure supplement 4B).In contrast to CCL21 + mTECs, some genes upregulated by CD4 + thymocytes were expressed by tuft-like mTECs (Figure 1I).Interestingly, many genes encoding for cytokines and cell adhesion molecules were associated with Aire + mTECs and post-Aire cells with some of them already expressed in TAC-TECs, suggesting that CD4 + thymocytes may act upstream of Aire + mTEC hi .These results thus provide the first evidence that CD4 + thymocytes are able to induce in mTEC lo essential transcriptional regulators for mTEC differentiation and function as well as TRAs, adhesion molecules, and cytokines.

CD4 + thymocytes regulate maturational programs in mTEC lo through MHCII/TCR interactions
We next investigated by which mechanism CD4 + thymocytes regulate the transcriptional programs of mTEC lo .Given that MHCII/TCR interactions with mTECs are critical for CD4 + T-cell selection (Klein et al., 2019), we hypothesized that these interactions could play an important role in initiating   transcriptional programs that govern the functional and developmental properties of mTEC lo .To this end, we used a unique transgenic mouse model in which MHCII expression is selectively abrogated in mTECs (mTEC ΔMHCII mice) (Irla et al., 2008).In contrast to their WT counterparts, we found that OVA 323-339 -loaded mTECs from mTEC ΔMHCII mice were ineffective at activating OTII-specific CD4 + T cells, demonstrating that the capacity of antigen presentation of mTECs to CD4 + T cells is impaired in these mice (Figure 2A).The comparison of the gene expression profiles of mTEC lo purified from WT and mTEC ΔMHCII mice (Figure 1-figure supplement 1B) revealed that MHCII/TCR interactions with CD4 + thymocytes resulted in the upregulation of 1300 genes (FC > 2), 449 of them reaching statistical significance (Cuffdiff p<0.05).846 genes were also downregulated (FC < 0.5) with 340 reaching significance (Cuffdiff p<0.05) (Figure 2B).Similarly to the comparison of WT versus ΔCD4 mice (Figure 1D), the genes significantly upregulated by MHCII/TCR interactions in mTEC lo corresponded preferentially to TRAs (p=4.5 × 10 -13 ) that are mainly Aire-dependent and Aire/Fezf2-independent (Figure 2C-E, Supplementary file 2).In line with the recent discovery of Aire expression in mTECs expressing intermediate levels of CD80 identified in the proliferating and maturational stage mTEC single-cell clusters (Dhalla et al., 2020), we found a strong correlation (p=2 × 10 -16 ) between gene upregulation induced by MHCII/TCR interactions and the responsiveness of genes to Aire's action obtained from the comparison between WT and Aire -/-mTEC hi (Figure 2F).These data are in agreement with the identification of a list of activation factors including Aire among the non-TRA genes induced by MHCII/TCR interactions with CD4 + thymocytes in mTEC lo (Figure 2G).mTEC lo from mTEC ΔMHCII mice expressed ~4.5-fold less Aire than WT mTEC lo , with substantial levels of 15.8 and 73.7 fragments per kilobase of transcript per million mapped reads (FPKM), respectively.For comparison, Aire expression level in WT mTEC hi was 448.9 FPKM.mTEC lo from mTEC ΔMHCII mice also expressed ~1.5-fold less Fezf2 than WT mTEC lo (90.2 versus 134.5 FPKM, respectively).This reduction in Aire and Fezf2 expression in mTEC ΔMHCII mice was also confirmed by qPCR (Figure 2H).These results highlight the importance of MHCII/ TCR interactions with CD4 + thymocytes in upregulating Aire and Fezf2 mRNAs and some of their associated TRAs in mTEC lo .Interestingly, 17 HDAC3-regulated transcription factors as well as Nfkb2, Trp53, and Relb transcription factors were induced by MHCII/TCR interactions with CD4 + thymocytes (Figure 2I).Moreover, the expression of several cytokines, chemokines, and cell adhesion molecules was also upregulated (Figure 2J, Figure 2-figure supplement 1A).Using single-cell RNA-seq data (Figure 1-figure supplement 4), we found that these genes were poorly associated with CCL21 + and tuft-like mTEC lo (Figure 2K).Consistently with the altered cellularity of Aire + mTECs in mTEC ΔMHCII plot of gene expression levels (fragments per kilobase of transcript per million mapped reads [FPKM]) of mTEC lo from WT versus ΔCD4 mice.Genes with fold difference ≥2 and p-adj<0.05 were considered as upregulated or downregulated genes (red and blue dots, respectively).RNA-seq was performed on two independent biological replicates with mTEC lo derived from 3 to 5 mice.(D) Numbers of tissue-restricted self-antigens (TRAs) and non-TRAs in genes up-and downregulated (left panel) and the proportion of upregulated TRAs compared to those in the all genome (right panel).ND, not determined.(E) Numbers of induced Aire-dependent, Fezf2-dependent, Aire/Fezf2-dependent, and Aire/Fezf2-independent TRAs.(F) The expression of Aire-dependent (Meig1, Nov), Fezf2-dependent (Fcer2a, Kcnj5), Aire/Fezf2-dependent (Krt1, Reig1), and Aire/Fezf2-independent (Crp, Rsph1) TRAs measured by qPCR in WT (n = 3-4) and ΔCD4 (n = 3-4) mTEC lo .(G) Expression fold change in HDAC3induced transcriptional regulators and other transcription factors significantly upregulated in WT versus ΔCD4 mTEC lo .The color code represents gene expression level.(H) Heatmaps of genes encoding for cell adhesion molecules and cytokines that were significantly downregulated in mTEC lo from ΔCD4 mice.(I) Hierarchical clustering and heatmap of mean expression of these cell adhesion molecules and cytokines in mTEC subsets identified by scRNA-seq.Error bars show mean ± SEM, *p<0.05,**p<0.01 using two-tailed Mann-Whitney test for (A), (B) and (F) and chi-squared test for (D).
The online version of this article includes the following figure supplement(s) for figure 1:           mice (Irla et al., 2008), some of these genes were associated with post-Aire cells.Strikingly, many genes upregulated by CD4 + thymocytes in mTEC lo were highly expressed by Aire + mTECs.Interestingly, several of these genes were already expressed by TAC-TECs, including Aire and Fezf2, strongly suggesting that an enhanced transcriptional activity promoted by MHCII/TCR interactions with CD4 + thymocytes accompanies the transition from TAC-TECs to Aire + mTECs.Altogether, these data show that CD4 + thymocytes, through MHCII/TCR interactions, control the functional properties of mTEC lo and activate key transcriptional programs governing their differentiation and function.
TCR/MHCII interactions with CD4 + thymocytes regulate the development of Fezf2 + pre-Aire, post-Aire, and tuft-like mTEC subsets Since key transcription factors implicated in mTEC differentiation were upregulated in mTEC lo by MHCII/TCR-mediated interactions with CD4 + thymocytes (Figures 1G and 2I), we next analyzed the composition for the newly identified mTEC subsets in ΔCD4 and mTEC ΔMHCII mice.In agreement with our previous study (Irla et al., 2008), we first observed a substantial reduction in the frequencies and numbers of mTEC hi in both mice (Figure 3A).Furthermore, numbers of mTEC lo were also substantially reduced.Consequently, ΔCD4 and mTEC ΔMHCII mice have a globally reduced cellularity in total mTECs.An Aire/Fezf2 co-staining both by histology and flow cytometry then revealed a substantial reduction in Aire -Fezf2 + and Aire + Fezf2 + cells (Figure 3B and C).We further analyzed by flow cytometry Aire and Fezf2 expression specifically in mTEC lo and mTEC hi , according to the gating strategy shown in Figure 1-figure supplement 1A.In agreement with the detection of Aire in the proliferating and maturational single-cell clusters in mTEC lo (Baran-Gale et al., 2020;Dhalla et al., 2020;Wells et al., 2020), we found that Aire protein was expressed in a small fraction of mTEC lo compared to mTEC hi in WT, ΔCD4, and mTEC ΔMHCII mice (Figure 3C).Aire -Fezf2 + and Aire + Fezf2 + mTECs were reduced in mTEC lo of ΔCD4 and mTEC ΔMHCII mice with a more marked effect in mTEC hi .This decrease was not due to impaired proliferation since normal frequencies of Ki-67 + proliferating cells were observed in ΔCD4 and mTEC ΔMHCII mice (Figure 3-figure supplement 1).Furthermore, numbers of involucrin + TPA + Aire -post-Aire cells were reduced in the medulla of ΔCD4 and mTEC ΔMHCII mice (Figure 3-figure supplement 2A), consistently with the decrease of Aire + mTEC hi (Figure 3B and  C).In contrast, the frequencies of CCL21 + cells among mTEC lo were not altered in ΔCD4 and mTEC ΔM- HCII mice (Figure 3D).This is in line with the observation that few genes upregulated by TCR/MHCII interactions with CD4 + thymocytes were associated with CCL21 + mTECs (Figures 1I and 2K).We also analyzed tuft-like mTECs since the expression of the transcription factor Pou2f3, known to control the development of this cell type (Bornstein et al., 2018;Miller et al., 2018), was decreased in mTEC lo of ΔCD4 and mTEC ΔMHCII mice (Figures 1G and 2I).We found that numbers of tuft-like mTECs identified by flow cytometry using the DCLK1 marker were reduced in both mice (Figure 3E, Figure 3figure supplement 2B), indicating that their development is promoted by MHCII/TCR interactions with CD4 + thymocytes.Importantly, Aire -Fezf2 + and Aire + Fezf2 + mTEC lo and mTEC hi as well as CCL21 + and DCLK1 + tuft-like mTEC lo were similarly reduced in MHCII -/-mice, further confirming that CD4 + thymocytes control the cellularity of these novel mTEC subsets (Figure 3-figure supplement 3).Altogether, these data reveal that MHCII/TCR-mediated interactions with CD4 + thymocytes have a broad impact on mTEC composition by controlling the cellularity of not only Aire + Fezf2 + mTECs but also Fezf2 + pre-Aire + mTECs, post-Aire, and tuft-like cells.
To define the genome-wide effects of highly self-reactive CD4 + thymocytes in mTEC lo , we compared the gene expression profiles of mTEC lo from RipmOVAxOTII-Rag2 -/-versus OTII-Rag2 -/- mice (Figure 1-figure supplement 1B) and found an upregulation of 1438 genes (FC > 2) reaching statistical significance for 522 of them (Cuffdiff p<0.05).620 genes were also downregulated (FC < 0.5) with 136 reaching significance (Cuffdiff p<0.05) (Figure 4C).The genes upregulated exhibited an approximately fourfold more of TRA over non-TRA genes (p=4.7 × 10 -23 ), which corresponded mainly to Aire-dependent and Aire/Fezf2-independent TRAs (Figure 4D-F, Supplementary file 3).Similarly to the WT versus mTEC ΔMHCII comparison, we found a strong correlation (p=6.2 × 10 -7 ) between the genes upregulated by self-reactive CD4 + thymocytes and the responsiveness of genes to Aire's action obtained from the comparison between WT and Aire -/-mTEC hi (Figure 4G).These results support an impact of antigen-specific interactions in the expression of TRAs in mTEC lo , notably on Aire-dependent TRAs.Importantly, these results are in agreement with the induction of a list of activation factors including Aire and Fezf2 among the non-TRA genes (Figure 4H).Similarly to the comparisons of the WT versus ΔCD4 or mTEC ΔMHCII mice, numerous HDAC3-induced regulators as well as Sirt1, Nfkb2, Relb, and Trp53 transcription factors were upregulated in mTEC lo of RipmOVAx-OTII-Rag2 -/-mice compared to OTII-Rag2 -/-mice (Figure 4I).Interestingly, 21 out of 30 top targets of the Foxn1 transcription factor, implicated in TEC differentiation and growth (Žuklys et al., 2016), as well as cytokines, chemokines, and cell adhesion molecules, were also upregulated (Figure 4J, K Figure 4-figure supplement 2B).We found that few of these genes were associated with CCL21 + and tuft-like mTECs (Figure 4L).In contrast, many genes encoding for activation factors, cytokines, chemokines, and cell adhesion molecules were associated with Aire + and post-Aire mTECs, consistently with the fact that antigen-specific interactions with CD4 + thymocytes control the cellularity of Aire + mTECs.Moreover, most of these genes, including Aire and Fezf2, were already expressed by solid arrowheads indicate Aire + Fezf2 -, Aire -Fezf2 + , and Aire + Fezf2 + cells, respectively.The histogram shows the density of Aire + Fezf2 -, Aire -Fezf2 + , and Aire + Fezf2 + cells.(C-E) Flow cytometry profiles, frequencies, and numbers of Aire -Fezf2 -, Aire -Fezf2 + , and Aire + Fezf2 + cells in total mTECs, mTEC lo , and mTEC hi (C), of CCL21 + cells in mTEC lo (D) and of DCKL1 + cells in Aire -mTEC lo (E) from WT, ΔCD4, and mTEC ΔMHCII mice.II Abs: secondary antibodies.Data are representative of 2-3 independent experiments (n = 2-5 mice per group and experiment).Error bars show mean ± SEM, *p<0.05,**p<0.01,***p<0.001,****p<0.0001using unpaired Student's t-test for (B) and two-tailed Mann-Whitney test for (A) and (C-E).
The online version of this article includes the following figure supplement(s) for figure 3:       TAC-TECs, further highlighting that CD4 + thymocytes act upstream of Aire + mTEC hi .Altogether, these data reveal that highly self-reactive CD4 + thymocytes control in mTEC lo not only key transcription factors driven by their differentiation but also key molecules for T-cell development and selection such as TRAs, cytokines, chemokines, and adhesion molecules.
The online version of this article includes the following figure supplement(s) for figure 4:   The online version of this article includes the following figure supplement(s) for figure 5: Self-reactive CD4 + thymocytes enhance the level of active H3K4me3 mark in mTEC lo Since histone modifications constitute important regulatory mechanisms that control the open and closed states of mTEC chromatin (Ucar and Rattay, 2015), we investigated whether self-reactive CD4 + thymocytes induce histone modifications in mTECs.We first analyzed in WT mTEC lo the repressive H3K27me3 and the active H3K4me3 marks using chromatin immunoprecipitation (ChIP) followed by high-throughput sequencing (ChIP-seq).As expected, metagene analyses showed that Airedependent TRAs had higher levels of H3K27me3 in their genes than in all genes of the genome, confirming that they are in a repressive state (Figure 6A).In contrast, Fezf2-dependent TRAs had a significant enrichment of H3K4me3 in their transcriptional start site (TSS) (Figure 6B).Similarly, Aire/ Fezf2-independent TRAs were associated with low levels of H3K27me3 in their genes and high levels of H3K4me3 in their TSS.Thus, in contrast to Aire-dependent TRAs that are associated with the repressive H3K27me3 histone mark, Fezf2-dependent and Aire/Fezf2-independent TRAs are associated with the active H3K4me3 mark, indicating that these distinct TRAs are subjected to a specific epigenetic regulation.We next assessed whether highly self-reactive CD4 + thymocytes control the H3K27me3 and H3K4me3 chromatin landscape in mTEC lo .In contrast to H3K27me3, we found an increased global level of H3K4me3 in RipmOVAxOTII-Rag2 -/-compared to OTII-Rag2 -/-mice by flow cytometry (Figure 6C).We further analyzed by nano-ChIP-seq whether self-reactive CD4 + thymocytes regulate in mTEC lo the level of these two histone marks in Aire-dependent, Fezf2-dependent, and Aire/Fezf2independent TRA genes.H3K27me3 levels in Aire-dependent TRA genes were comparable in mTEC lo from RipmOVAxOTII-Rag2 -/-and OTII-Rag2 -/-mice (Figure 6D, left panel).Although lower, H3K27me3 levels in Fezf2-dependent and Aire/Fezf2-independent TRAs as well as in all genes were similar in both mice, indicating that the interactions with self-reactive CD4 + thymocytes do not regulate this repressive mark in TRA genes (Figure 6D, left panel).In contrast, H3K4me3 global level was increased in the TSS of all TRAs in RipmOVAxOTII-Rag2 -/-compared to OTII-Rag2 -/-mice as well as in all genes (Figure 6D, right panel).For representation, whereas the Aire/Fezf2-independent TRA, E2F transcription factor 2 (E2f2) induced by these interactions, was barely devoid of H3K27me3 in both mice, it was marked by H3K4me3 in its TSS specifically in RipmOVAxOTII-Rag2 -/-mice (Figure 6E).These results thus show that self-reactive CD4 + thymocytes enhance the global level of the active H3K4me3 histone mark in mTEC lo and in particular in the TSS of Fezf2-dependent and Aire/Fezf2-independent TRAs, indicative of an epigenetic regulation for their expression.

MHCII/TCR interactions between mTECs and CD4 + thymocytes prevent the development of autoimmunity
We next evaluated the impact of mTEC-CD4 + thymocyte interactions on the generation of selftolerant T cells by taking advantage that CD4 + and CD8 + T cells develop in mTEC ΔMHCII mice, in which MHCII/TCR interactions between mTECs and CD4 + thymocytes are abrogated.Interestingly, since TRAs induced by MHCII/TCR interactions showed a diverse peripheral tissue distribution in mTEC lo (Figure 7A, Supplementary file 4), we analyzed the TCRVβ usage in mTEC ΔMHCII mice by flow cytometry.TCRVβ usage was more altered in CD69 -mature CD4 + thymocytes than in CD8 + thymocytes (Figure 7B).Some TCRVβ were also altered in splenic CD4 + and CD8 + T cells.To determine whether these T cells contained self-reactive specificities, we adoptively transferred splenocytes from mTEC ΔM- HCII or WT mice into lymphopenic Rag2 -/-recipients (Figure 7C).Mice that received splenocytes derived from mTEC ΔMHCII mice lost significantly more weight than mice transferred with WT splenocytes (Figure 7D).They also exhibited splenomegaly with increased follicle areas and T-cell numbers showing a CD62L lo CD44 hi effector and CD62L hi CD44 hi central memory phenotype (Figure 7E-G).Immune infiltrates in lungs and salivary glands were observed by histology and flow cytometry in 75    and 41% of mice, respectively (Figure 7H and I).These two tissues contained increased numbers of central memory as well as CD44 + CD69 + and CD44 + CD69 -activated CD4 + and CD8 + T cells (Figure 7J).T-cell infiltrates were also observed in other tissues such as kidney, liver, and colon in agreement with the defective TRA expression associated with these tissues (Figure 7A and K).Altogether, these data show that in the absence of MHCII/TCR interactions between mTECs and CD4 + thymocytes, T cells

Discussion
Since mTECs play a crucial role in immunological tolerance by their exclusive expression of TRAs, it is essential to deepen our knowledge of the mechanisms that sustain their differentiation.Here using three distinct transgenic models, we found that self-reactive CD4 + thymocytes control the developmental transcriptional programs from the mTEC lo stage, including TAC-TECs that precede Aire + mTECs.CD4 + thymocytes increase in mTEC lo the phosphorylation of p38 MAPK and IKKα, the latter implicated in mTEC development (Lomada et al., 2007;Shen et al., 2019).Moreover, self-reactive CD4 + thymocytes increase RelB phosphorylation level.Interestingly, this nonclassical NF-κB subunit is crucial for mTEC differentiation and Aire-dependent and -independent TRA expression (Riemann et al., 2017).These data thus suggest that CD4 + thymocytes activate intracellular pathways from the mTEC lo stage, although alterations in mTEC lo subset composition could also contribute to the differences observed.Nevertheless, the substantial and homogeneous reduction in the levels of phospho-IKKα, -p38, and -RelB argues instead for impaired activation of IKKα, p38, and RelB signaling in the absence of self-reactive CD4 + thymocytes.Analysis of the mTEC lo transcriptional landscape by high-throughput RNA-seq revealed that self-reactive CD4 + thymocytes upregulate Nfkb2 (p52), known to form an heterocomplex with RelB in the nucleus upon activation (Irla et al., 2010).p52 is important for mTEC development, Aire, and TRA expression (Zhang et al., 2006;Zhu et al., 2006).Consequently, ΔCD4, mTEC ΔMHCII , and OTII-Rag2 -/-mice in which MHCII/TCR interactions between mTECs and CD4 + thymocytes are disrupted have altered Relb and Nfkb2 expression, and reduced Aire + mTEC numbers and Aire-dependent TRA representation.Our results are in agreement with the fact that RANK-induced NF-κB signaling is activated by membrane-bound RANKL and not soluble RANKL and thereby in the context of physical interactions between mTECs and CD4 + thymocytes (Asano et al., 2019).These interactions also upregulate Trp53 (p53) that controls the mTEC niche (Rodrigues et al., 2017) and Irf4 and Irf7 transcription factors that regulate key chemokines implicated in thymocyte medullary localization and mTEC differentiation (Haljasorg et al., 2017;Otero et al., 2013).Furthermore, the deacetylase Sirtuin-1 (Sirt1), which regulates Aire activity (Chuprin et al., 2015), and Spib, which limits mTEC differentiation (Akiyama et al., 2014), were also upregulated.Self-reactive CD4 + thymocytes thus induce key transcription factors that both positively and negatively control mTEC differentiation.Remarkably, our three different transgenic revealed that CD4 + thymocytes induce HDAC3-dependent mTEC-specific transcription factors (Goldfarb et al., 2016).Among them, Pou2f3 is involved in tuft-like mTEC development (Bornstein et al., 2018;Miller et al., 2018), which is consistent with our results showing that self-reactive CD4 + thymocytes control the cellularity of these cells.Our data thus identify that CD4 + thymocytes control the expression of master transcriptional regulators of mTEC differentiation and function.
In line with these data, we found that self-reactive CD4 + thymocytes regulate TEC development from a progenitor stage since they increase numbers of TEPC-enriched cells that express nonnegligible MHCII levels.Interestingly, we provide the first evidence that self-reactive CD4 + thymocytes to the initial weight.(E, F) Representative spleen pictures and their weights (E) and hematoxylin/eosin counterstained splenic sections (F).Scale bar, 1 mm.The histogram shows follicle areas.(G) Numbers of splenic CD3 + , CD4 + , and CD8 + T cells and of naive (CD44 lo CD62L hi ), effector memory (EM; CD44 hi CD62L lo ) and central memory (CM; CD44 hi CD62L hi ) phenotype.(H) Lung and salivary gland (SG) immune infiltrates detected by hematoxylin/ eosin counterstaining.Scale bar, 1 mm.(I, J) Numbers of T cells (I) and of naive, effector and central memory phenotype as well as CD44 + CD69 + and CD44 + CD69 -T cells (J) in lungs and SG.(K) Schematic of T-cell infiltrates in mice transferred with mTEC ΔMHCII T cells relative to those transferred with WT T cells.Each circle and black triangles represent an individual mouse and T-cell infiltration in a specific tissue, respectively.Data are representative of two independent experiments (n = 5-7 mice per group and experiment).Error bars show mean ± SEM, ****p<0.0001using two-way ANOVA for (D) and unpaired Student's t-test for (B) and (E-J).*p<0.05,**p<0.01,***p<0.001.
The online version of this article includes the following figure supplement(s) for figure 7:  control the cellularity of Fezf2 + mTECs.Accordingly, the expression of Fezf2 and its respective TRAs was enhanced by CD4 + thymocytes.Moreover, self-reactive CD4 + thymocytes regulate the cellularity of CCL21 + , post-Aire, and tuft-like cells in mTEC lo .These results are in full agreement with our previous findings that self-reactive thymocytes drive medulla expansion and increase the overall cellularity of the mTEC compartment (Irla et al., 2012).Because the heterogeneous composition in mTEC lo could influence the expression of the upregulated genes by self-reactive CD4 + thymocytes, we reanalyzed single-cell RNA-seq data in order to define their respective expression pattern in mTEC subsets.In accordance with the moderately altered frequencies of CCL21 + cells among mTEC lo observed by flow cytometry, few genes upregulated by self-reactive CD4 + thymocytes were associated with this mTEC subset.In contrast, we found that antigen-specific interactions with CD4 + thymocytes strongly upregulate genes associated with TAC-TECs, Aire + mTECs, and post-Aire cells.These findings indicate that self-reactive CD4 + thymocytes act from the precursors of Aire + mTEC hi (i.e., in TAC-TECs) to the post-Aire stage.It is interesting to note that although strongly altered the development of mTEC hi is not completely abrogated in the absence of CD4 + thymocytes or MHCII/TCR-mediated interactions with CD4 + thymocytes.This could be explained by the fact that invariant NKT have been proposed to participate in mTEC differentiation by expressing RANKL (White et al., 2014).Overall, our results thus reveal that antigen-specific interactions with CD4 + thymocytes have an unsuspected broad impact on mTEC composition by driving their development from an early progenitor to a late post-Aire stage.Interestingly, high-throughput RNA-seq showed that MHCII/TCR interactions with CD4 + thymocytes upregulate the expression of chemokines in mTEC lo .Among them, CCL19 (CCR7 ligand) is implicated in the medullary localization of thymocytes and the emigration of newly generated T cells (Ueno et al., 2004); and CCL22 (CCR4 ligand) implicated in medullary entry and thymocyte/dendritic cell interactions (Hu et al., 2015).Self-reactive CD4 + thymocytes also enhance CCL2 (CCR2 ligand) and CCL20 (CCR6 ligand) that promote the entry of peripheral dendritic cells and Foxp3 + regulatory T cells into the thymus (Baba et al., 2009;Cédile et al., 2014;Cowan et al., 2018;Lopes et al., 2018;Borelli and Irla, 2021).mTEC-CD4 + thymocyte interactions thus induce key chemokines that regulate the trafficking of thymocytes and dendritic cells that participate in tolerance induction.Moreover, cytokines such as Il15 and Fgf21 implicated in invariant NKT development and TEC protection against senescence, as well as adhesion molecules involved in mTEC-thymocyte interactions, were also induced (Pezzi et al., 2016;White et al., 2014;Youm et al., 2016).Altogether, our data show that self-reactive CD4 + thymocytes regulate functional properties of mTECs by inducing chemokines, cytokines, and adhesion molecules that are critical for T-cell development.
The expression of TRAs is regulated by Aire and to a lesser extent by Fezf2 (Anderson et al., 2002;Takaba et al., 2015).In agreement with other studies (Gray et al., 2007;Takaba et al., 2015), we found Fezf2 in both mTEC lo and mTEC hi , whereas Aire protein is mainly expressed in mTEC hi .Nevertheless and in line with recent single-cell transcriptomic analyses (Baran-Gale et al., 2020;Dhalla et al., 2020;Wells et al., 2020), we detected Aire by flow cytometry, qPCR, and RNA-seq in a small subset (~1.5%) of mTEC lo .CD4 + thymocyte interactions upregulate Aire and Fezf2 and some of their respective TRAs in these cells.Interestingly, in contrast to Aire-dependent TRAs that are characterized by high levels of H3K27me3 (Handel et al., 2018;Sansom et al., 2014), we found that Fezf2-dependent TRAs show high levels of H3K4me3.This highlights that Aire and Fezf2 use distinct epigenetic modes in regulating TRA expression.Remarkably, these interactions also induce in mTEC lo numerous Aire/ Fezf2-independent TRAs, whose regulation remains unknown.Similarly to Fezf2-dependent TRAs, they had high levels of H3K4me3 in their TSS, suggesting that Aire/Fezf2-independent TRAs are not subjected to the same regulatory transcriptional mechanisms than Aire-dependent TRAs.Our results are consistent with a previous study indicating that the Aire-independent TRA, Gad1, shows active epigenetic marks (Tykocinski et al., 2010).Remarkably, self-reactive CD4 + thymocytes increase H3K4me3 level in the TSS of all TRA categories, thus providing a novel epigenetic mechanistic insight into how they regulate the mTEC gene expression profile.In line with TRA regulation and the development of distinct mTEC subsets, the repertoire of mature T cells contains autoreactive cells when MHCII/TCR interactions were abrogated between mTECs and CD4 + thymocytes.Accordingly, the adoptive transfer of splenocytes from mTEC ΔMHCII mice is capable of inducing signs of autoimmunity, illustrating the fact that mTEC-CD4 + thymocyte interactions are critical for the generation of a selftolerant T-cell repertoire.Future investigations based on TCR sequencing analysis are expected to define to which extent the TCR repertoire is altered in mTEC ΔMHCII mice.In summary, our genome-wide scale study reveals that self-reactive CD4 + thymocytes activate transcriptional programs from the TAC-TEC stage that sustains the differentiation into Aire + Fezf2 + and post-Aire mTECs (Figure 7-figure supplement 1).These interactions also upregulate the expression of TRAs, cytokines, chemokines, and adhesion molecules that are all implicated in mTEC function.Thus, CD4 + thymocytes control several unsuspected aspects of mTEC lo required for the establishment of T-cell tolerance.mTEC purification mTECs were isolated by enzymatic digestion with 50 μg/ml of Liberase TM (Roche) and 100 μg/ml of DNase I (Roche) in HBSS medium, as previously described (Lopes et al., 2017).CD45 + hematopoietic cells were depleted using anti-CD45 magnetic beads by autoMACS with the depleteS program (Miltenyi Biotec).Total mTECs (EpCAM + UEA-1 + Ly51 lo ), mTEC lo (EpCAM + UEA-1 + Ly51 lo CD80 lo/int ), and mTEC hi (EpCAM + UEA-1 + Ly51 lo CD80 hi ) were sorted with a FACSAriaIII cell sorter (BD).The purity of sorted mTEC lo was >98%.Flow cytometry gating strategies are shown in Figure 1-figure supplement 1.

Quantitative RT-PCR
Total RNA was prepared with TRIzol (Invitrogen).cDNAs was synthesized with oligo(dT) using Superscript II reverse transcriptase (Invitrogen).qPCR was performed with the ABI 7500 fast real-time PCR system (Applied Biosystems) and SYBR Premix Ex Taq master mix (Takara).Primers are listed in the Key resources table.
In vivo transfer of splenocytes into Rag2 -/-recipients 3.10 6 splenocytes purified from the spleen of WT and mTEC ΔMHCII mice of 8 weeks of age were intravenously injected into Rag2 -/-female recipients.CD3 + , CD4 + , and CD8 + T-cell infiltrates were analyzed 6 weeks after transfer by histology and flow cytometry in different peripheral tissues.

RNA-seq experiments
Total RNA purified from mTEC lo (Figure 1-figure supplement 1) was extracted with miRNeasy Micro Kit (QIAGEN), and RNA quality was assessed on an Agilent 2100 BioAnalyzer (Agilent Technologies).RNA Integrity Number values over 8 were obtained.RNA-seq libraries were generated using the SMART-Seq-v4-Ultra Low Input RNA Kit (Clontech) combined to the Nextera library preparation kit (Illumina) following the manufacturer's instructions.Libraries were sequenced with the Illumina NextSeq 500 machine to generate datasets of single-end 75 bp reads.Two independent biological replicates were used per each condition.RNA-seq data have been deposited with Gene Expression Omnibus (GEO) under the accession number GSE144650.

RNA-seq analysis
The sequencing reads were mapped to the Mus musculus (mm10) reference genome using the TopHat 2 (version 2.0.12)aligner (Kim et al., 2013).The reads mapping to the annotated genes (igenome UCSC mm10 GTF: https://support.illumina.com/sequencing/sequencing_software/igenome.html)were counted, normalized, and compared using Cuffdiff2 (version 2.2.1;Trapnell et al., 2013) between two conditions.Cuffdiff2 generated expression levels as FPKM, FCs, and p-values to assess the statistical significance of the FPKM difference of each gene between the tested two conditions.Genes showing a significant variation in gene expression between WT and ΔCD4, or WT and mTEC ΔMHCII , or RIPmOVAxOTII-Rag2 -/-and OTII-Rag2 -/-mice (p-value≤0.05,FC difference ≥ 2 or ≤ 0.5) were considered as up-or downregulated.The TRA and non-TRA gene assignments were obtained from Sansom et al., 2014.In this report, the identification of the specificity of expression for each gene in the genome was carried out by analyzing the microarray expression profiles of a large number of different mouse tissues.Aire-dependent, Fezf2-dependent, Aire/Fezf2-dependent, and Aire/Fezf2-independent TRAs were identified using Aire -/-mTEC hi RNA-seq datasets and Fezf2 -/- total mTEC microarray datasets, obtained from the NCBI GEO database (GSE87133 and GSE69105,respectively).
Correlation between the variation of gene expression in mTEC lo from WT versus mTEC ΔMHCII or RIPmOVAxOTII-Rag2 -/-versus OTII-Rag2 -/-mice, and of the same genes in mTEC hi from WT versus Aire -/-mice was performed doing a locally regression (loess) with the R software (http://www.rproject.org/).Differential gene expression in WT versus Aire -/-mTEC hi was obtained by processing using TopHat2 and Cuffdiff2, the sequencing reads corresponding to WT (Chuprin et al., 2015) and Aire -/- (Danan-Gotthold et al., 2016) mTEC hi RNA-seq datasets, which were obtained from the NCBI GEO database (GSE68190 and GSE87133, respectively).HDAC3-dependent mTEC-specific transcription factors regulated by mTEC-thymocyte crosstalk were identified by comparing the top 50 transcriptional regulators that are induced by HDAC3 (Goldfarb et al., 2016) with genes upregulated in the different mouse models.TRAs differentially expressed between mTEC lo from WT and mTEC ΔMHCII mice were classified according to their tissue distribution using the mouse ENCODE transcriptome database (Yue, 2014).Only tissues that showed the highest expression were taken into account.

Single-cell RNA-seq analysis
Single-cell RNA-seq count matrix from Wells et al., 2020 (accession number GSE137699) was reanalyzed with the Seurat package (Hao et al., 2021).QC analysis was performed by filtering out cells with a number of feature counts under 200 or over 4000, and a proportion of mitochondrial counts over 4%.Sample integration was performed as described in the Seurat vignette.After PCA for dimension reduction, 15 first dimensions were conserved.Cells were clustered and visualized with UMAP.Cluster annotation was performed by identifying sets of specific markers to each cluster using a differential expression test (FindMarkers function, test = 'roc').Heatmaps were generated using the pheatmap R package.Nano-ChIP-seq experiments Nano-ChIP-seq was performed as previously described (Adli and Bernstein, 2011) on 5.10 4 purified mTEC lo (Figure 1-figure supplement 1).ChIP-seq libraries were prepared with TruSeq ChIP Sample Preparation Kit (Illumina), and 2 × 75 bp paired-end reads were sequenced on an Illumina HiSeq.ChIP-seq data have been deposited with GEO under the accession number GSE144680.

ChIP-seq analysis
Reads were aligned on the mouse genome (mm10) using Bowtie 2 and default parameters (Langmead and Salzberg, 2012).Properly paired alignments were selected using Samtools view with the 0x2flag (-f option).Nonuniquely mapped reads-pairs were filtered out by removing reads with the 'XS' tag set by Bowtie 2. Normalized bedgraphs for ChIP and input samples were generated using MACS2 (Zhang et al., 2008) with the callpeak command in BAMPE mode with the --SPMR option.For the diffused H3K27me3 histone mark, the -broad option was used.ChIP enrichment was calculated parsing the ChIP and input normalized bedgraphs with MACS2 and the bdgcmp command (-m FE option).The obtained bedgraphs were converted to wig using the bedGraphToWig.pl script with the --step 10 parameter.MACS2-generated peak calling files were converted to BED files using the cut -f 1-6 command.The obtained Wig and BED files were parsed by CEAS (Shin et al., 2009) to generate metagene profile plots corresponding to the average enrichment of H3K4me3 in 3 kb TSS windows or H3K27me3 at gene loci.H3K4me3 and H3K27me3 CEAS-dumped files were parsed to compute ratios of ChIP/input in 1 kb TSS windows and gene loci, respectively.Statistical significance between ChIP enrichment data was tested using the nonparametric Mann-Whitney test.Data were visualized using the Integrative Genomics Viewer (IGV) (Robinson et al., 2011).

Statistics
Data are presented as means ± standard error of mean (SEM

Figure 1 .
Figure 1.The transcriptional profile and IKKα and p38 MAPK signaling pathways are impaired in mTEC lo of ΔCD4 mice.(A, B) Total IKKα, p38 MAPK, phospho-IKKα(Ser180)/IKKβ(Ser181), and p38 MAPK (Thr180/Tyr182) (A) and the ratio of phospho/total proteins (B) analyzed by flow cytometry in mTEC lo from WT and ΔCD4 mice.Data are representative of two independent experiments (n = 3-4 mice per group and experiment).(C) Scatter Figure 1 continued on next page

Figure supplement 1 .
Figure supplement 1. Gating strategy used to purify mTEC lo cells.

Figure supplement 4 .
Figure supplement 4. Identification of thymic epithelial cell (TEC) subsets by single-cell RNA-seq.

Figure 2 .
Figure 2. The transcriptional and functional properties of mTEC lo are impaired in mTEC ΔMHCII mice.(A) Percentages of CD69 + OTII CD4 + T cells cultured or not with variable numbers of OVA 323-339 -loaded WT or mTEC ΔMHCII mTECs derived from two independent experiments (n = 2-3 mice per group and experiment).(B) Scatter plot of gene expression levels (fragments per kilobase of transcript per million mapped reads [FPKM]) of mTEC lo from WT versus mTEC ΔMHCII mice.Genes with fold difference ≥2 and p-adj<0.05 were considered as upregulated or downregulated genes (red and blue dots,

Figure 2
Figure 2 continued on next page Figure supplement 1. Altered expression of some cytokines, cell adhesion molecules, and chemokines in mTEC lo from mTEC ΔMHCII mice.

Figure 5
Figure 5 continued on next page

Figure 5 continued
Figure 5 continued

Figure supplement 1 .
Figure supplement 1. CD4 + thymocytes through MHCII/TCR-mediated interactions control transcriptional programs of mTEC lo that drive their differentiation and function.

Figure 7 continued
Figure 7 continued Lopes et al. eLife 2022;11:e69982.DOI: https://doi.org/10.7554/eLife.6998218 of 31 the animal facilities of the CIML (Marseille, France).Standard food and water were given ad libitum.Males and females were used at the age of 5-6 weeks.All experiments were done in accordance with national and European laws for laboratory animal welfare (EEC Council Directive 2010/63/UE), and were approved by the Marseille Ethical Committee for Animal Experimentation (Comité National de Réflexion Ethique sur l'Expérimentation Animale no.14).