Identification of embryonic precursor cells that differentiate into thymic epithelial cells expressing autoimmune regulator

RANK expression may identify the mTEC lineage because of its critical role in mTEC differentiation. To elucidate the mTEC differentiation program, we used mutant mice in which an internal ribosome entry site (IRES) linking the coding sequence for an enhanced GFP (EGFP)-Cre recombinase fusion protein was inserted into the final exon of the RANK gene (creating RANK-EGFP mice; Maeda et al., 2012). In these mice, mTEC development was intact, and EGFP expression was detected in adult mTECs in which RANK mRNA expression was confirmed by quantitative PCR (qPCR) analysis (unpublished data).

EGFP-expressing cells in the thymic stromal fraction (CD45^–TER119^–) were identified on embryonic days E13.5, E14.5, E15.5, and E17.5 by flow cytometry. The frequency of EGFP-expressing cells gradually increased in the TEC marker EpCAM⁺ fraction during embryonic development (Fig. 1 A). EpCAM⁺ cells were further divided according to the expression of an mTEC marker, UEA-1 ligand. EGFP expression was detected in UEA-1⁺ TECs, but not in UEA-1^– TECs (Fig. 1 B). Immunohistochemical analysis also revealed EGFP expression in cells expressing the mTEC marker keratin-5 in the thymus at E13.5, E14.5, and E15.5 (Fig. 1 C). In the thymus at E17.5, the majority of EGFP⁺ cells expressed Aire and UEA-1 ligand (Fig. 1 D, white arrows), and a few EGFP⁺ UEA-1⁺ cells were negative for Aire expression (Fig. 1 D, yellow arrow). Overall, these data supported the selective expression of RANK in UEA-1⁺ TECs. Interestingly, the expression level of EGFP in UEA-1⁺ TECs was gradually increased during embryonic development (Fig. 1 B) and was positively correlated with the expression level of MHCII, a maturation marker of TECs (Fig. 1 E). Thus, the expression level of RANK may be correlated with mTEC differentiation.

We next aimed to identify precursor cells of Aire⁺ mTECs that received RANKL signaling. Inhibition of RANKL signaling may block mTEC differentiation in RANK-expressing precursor cells of Aire⁺ mTECs. Indeed, UEA-1⁺ TECs expressing EGFP were found in the thymus of Rankl^−/− RANK-EGFP embryos and RANK-EGFP embryos treated with an anti-RANKL antibody (RANKL-Ab; Fig. 2 A), which abolished the RANKL–RANK interaction (Furuya et al., 2011; Sugiyama et al., 2012; Khan et al., 2014). qPCR analysis of Aire expression confirmed that the RANKL-Ab treatment efficiently blocked mTEC differentiation (Fig. 2 B). The expression level of EGFP in UEA-1⁺ TECs was slightly lower in these mice than that in control RANK-EGFP mice (Fig. 2 A). Moreover, UEA-1⁺ TECs in embryonic mice deficient in RANKL signaling showed a moderate level of MHCII expression (MHCII^mid) and a low level of CD80 expression (CD80^–; Fig. 2 C). Consequently, we speculated that RANK⁺UEA-1⁺MHCII^midCD80^– TECs in embryonic mice deficient in RANKL signaling could be candidate precursor cells of Aire⁺ mTECs (hereafter referred to as pMECs).

pMECs were sorted from the thymus of E17.5 RANK-EGFP mice treated with RANKL-Ab (Fig. S1). qPCR analysis indicated that expression of Aire and TSAs was practically absent in pMECs (Fig. 2 D). Analysis of the CpG methylation status using combined bisulfite restriction analysis (COBRA) indicated that tissue-specific differentially methylated regions (T-DMRs) in the Aire gene (Yagi et al., 2008; Wu et al., 2012) were significantly hypermethylated in pMECs as compared with that in the CD80⁺ mature mTEC fraction containing Aire-expressing cells (Fig. 2 E). Thus, Aire gene expression may be silenced at the epigenetic level in pMECs.

Surprisingly, the expression of the cTEC marker Ly51 was high in pMECs (Fig. 2 F). Moreover, qPCR analysis indicated that pMECs expressed high levels of the cTEC markers β5t and keratin-8 in addition to keratin-5 (Fig. 2 G). Hence, pMECs were different from the previously reported Ly51^–CD80^– immature mTEC fraction (Gäbler et al., 2007; Gray et al., 2007). Overall, pMECs expressed low levels of mature mTEC markers and high levels of cTEC molecules.

We next examined whether pMECs were capable of differentiating into Aire-expressing mature mTECs by using RTOC. We used an RTOC derived from an aly/aly fetus that lacked Aire⁺ mTECs because of a dysfunctional mutation in NIK (Kajiura et al., 2004). pMECs were reaggregated with thymic cells of E15.5 aly/aly mice. When RTOC was analyzed after 5 d of culture, UEA-1⁺CD80⁺ mature mTECs developed in the RTOC were mixed with pMECs (Fig. 3 A). These UEA-1⁺CD80⁺ mature mTECs expressed EGFP (Fig. 3 A), confirming their development from the introduced pMECs. qPCR analysis revealed induction of Aire mRNA in RTOC mixed with pMECs (Fig. 3 B). Moreover, immunohistochemical staining indicated generation of Aire⁺ mTECs in the RTOC mixed with pMECs (Fig. 3 C). These data strongly suggested that pMECs were capable of differentiating into Aire-expressing mature mTECs.

We further investigated the differentiation potential of pMECs in vivo. pMECs derived from C57BL/6 mice (H-2Kb⁺) were reaggregated with thymic cells of E15.5 aly/aly mice on the BALB/cA background (H-2kb^–; Shinzawa et al., 2011). Aly/aly RTOCs mixed with pMECs then were transplanted on kidney capsules of athymic mice and analyzed 4 wk after grafting. Flow cytometric analysis indicated that H-2Kb⁺ TECs expressed UEA-1-ligand, RANK, and CD80 (Fig. 3 D), suggesting that pMECs differentiated into mature mTECs. In contrast, Ly51⁺UEA-1^– cTECs were nearly absent in the H-2Kb⁺ TEC fraction (Fig. 3 D). These data strongly suggested that pMECs could differentiate into mTECs and would lack differentiation potential into cTECs.

Next, we tested whether pMECs could differentiate into self-tolerance–inducing mTECs by RTOC transplantation in kidneys of athymic nude mice. Aly/aly mice exhibited autoimmunity owing to developmental defects in mature mTECs. We therefore examined whether introducing pMECs suppressed the onset of autoimmunity by the aly/aly mutation. In this experiment, aly/aly mice in the BALB/cA background (Shinzawa et al., 2011), RANK-EGFP mice in the BALB/cA background, and BALB/cA nude mice were used for MHC-matched congenic transplantation. Aly/aly RTOC mixed with pMECs was grafted on kidney capsules of athymic mice and analyzed 8 wk after grafting. Immunohistochemical staining of the thymus generated from the transplanted RTOC supported the development of medullar regions containing Aire⁺ mTECs when pMECs were mixed with the aly/aly RTOC (Fig. 3 E). This suggested that the introduction of pMECs in the thymus allowed for long-term maintenance of Aire⁺ mTECs. The number of generated medullas in each grafted thymus may be lower than that in wild-type grafted thymus (unpublished data), likely owing to the requirement of ligand-expressing niches (Rossi et al., 2007; Roberts et al., 2012) for the proliferation of pMECs. We further analyzed autoimmune responses in nude mice receiving aly/aly RTOCs. As expected, nude mice receiving aly/aly RTOCs exhibited substantial inflammatory cell infiltration in the liver, salivary glands, and pancreas of recipient nude mice. Strikingly, introduction of pMECs in the aly/aly RTOC inhibited cell infiltration (Fig. 3 F). Moreover, whereas sera from nude mice receiving aly/aly RTOC and aly/aly RTOC with EpCAM^– cells contained autoantibodies against these organs, such autoantibody production was suppressed in nude mice receiving aly/aly RTOC with pMECs (Fig. 3 G). These data strongly suggested that introduction of pMECs in the thymus was sufficient for inhibiting autoimmunity provoked by mTEC dysfunction. Hence, we concluded that pMECs could differentiate into self-tolerance–inducing mTECs. Collectively, our data supported that pMECs were, indeed, precursor cells of Aire⁺ mTECs.

Because RANK signaling initiates pMEC differentiation by activating intracellular signaling pathways, we next aimed to examine the signaling pathways downstream of RANK that drive pMEC differentiation. We then examined whether differentiation of pMECs into Aire⁺ mTECs required TRAF6 and RelB.

Flow cytometric analysis revealed that there was a dramatic reduction in MHCII^hiUEA-1⁺ TECs of the Traf6^−/− thymus at E17.5 (Fig. 4 A). The expression level of UEA-1 ligand in TECs was also reduced in the Traf6^−/− thymus, as previously reported (Akiyama et al., 2005), implying that TRAF6-dependent NF-κB activation may control UEA-1 ligand expression in TECs. Moreover, analysis of Traf6^−/− RANK-EGFP mice suggested that MHCII^midUEA-1⁺ TECs accumulated in the Traf6^−/− embryonic thymus expressing RANK and high levels of Ly51 (Fig. 4 B and 4C). These data suggested that the absence of TRAF6 arrested mTEC differentiation at pMECs. Thus, differentiation of pMECs into Aire⁺ mTECs induced by RANK signaling was likely to be dependent on TRAF6.

Unexpectedly, Relb^−/− and Relb^−/− RANK-EGFP mice showed dramatically different profiles of UEA-1⁺ TEC differentiation at E17.5. The expression of UEA-1 ligand in Relb^−/− TECs was reduced to a similar extent as that in the Traf6^−/− thymus, which also suggested the contribution of NF-κB activation to UEA-1 ligand expression in TECs. Importantly UEA-1⁺ TECs in the embryonic Relb^−/− thymus expressed significantly lower levels of MHCII than that in the Traf6^−/− thymus (Fig. 4 A). Moreover, RANK expression levels were significantly lower in UEA-1⁺ TECs in the embryonic Relb^−/− thymus than that in the Traf6^−/− thymus (Fig. 4 B). Interestingly, UEA-1⁺ TECs in the Relb^−/− thymus expressed significantly higher levels of CD24, which has been reported to function as a stem cell marker in other tissues (Rietze et al., 2001; Shackleton et al., 2006), as compared with that in the wild-type and Traf6^−/− thymus (Fig. 4 D). Given that mTEC differentiation was arrested at pMECs in the Traf6^−/− thymus, the absence of RelB arrested mTEC differentiation at different precursor cells from pMECs. Despite such differences in MHCII and RANK expression levels, Ly51 expression in UEA-1⁺ TECs in the Relb^−/− thymus was as high as that in pMECs, implying that UEA-1⁺MHCII^loCD24^hi TECs in the Relb^−/− thymus and pMECs may have the same origin (Fig. 4 C). During early differentiation, UEA-1⁺ TECs expressed low levels of RANK and MHCII (Fig. 1, B and D). Moreover, CD24 expression in UEA-1⁺ TECs was down-regulated during embryonic thymus development (Fig. 4 E). Accordingly, UEA-1⁺RANK^loMHCII^loCD24^hi TECs accumulated in the Relb^−/− embryonic thymus were likely to be in an earlier stage of differentiation than pMECs. We therefore hypothesized that UEA-1⁺RANK^loMHCII^loCD24^hi TECs may be candidate progenitors of pMECs (hereafter referred to as pro-pMECs).

Given that RelB activation promotes the differentiation of pro-pMECs into pMECs, we next examined the upstream receptors promoting the differentiation of pro-pMECs through RelB activation. Because LtβR is known to activate the nonclassical NF-κB pathway, causing nuclear translocation of RelB, we first investigated embryonic TEC development in LtβR-deficient mice (Ltbr^−/−). Interestingly, all MHCII^hi, MHCII^mid, and MHCII^lo UEA-1⁺ TEC fractions were detected in the LtβR-deficient thymus (Fig. 5 A). Thus, LtβR signaling may be upstream of RelB in the differentiation of pro-pMECs, and other signaling pathways could cooperate with LtβR signaling. Because RANK expression was detected in MHCII^loUEA-1⁺ TECs in Relb^−/− embryos (Fig. 4 B), we hypothesized that RANK signaling may be involved in the differentiation of pro-pMECs. Indeed, double deficiency of RANK and LtβR signaling (Rank^−/−Ltbr^−/− and Ltbr^−/− treated with RANKL-Ab) caused accumulation of the UEA-1⁺MHCII^lo TEC fraction in E17.5 embryos (Fig. 5 A), similar to that observed in Relb^−/− embryos (Fig. 4 A). UEA-1⁺ TECs in the thymus of Rank^−/−Ltbr^−/− embryos and Ltbr^−/− embryos treated with RANKL-Ab exhibited high CD24 expression at E17.5 (Fig. 5 A). Moreover, analysis of Rankl^−/−Ltbr^−/− RANK-EGFP embryos and Ltbr^−/− RANK-EGFP embryos treated with RANKL-Ab indicated that RANK expression was low in UEA-1⁺ TECs (Fig. 5 B). These data suggested that UEA-1⁺ TECs in embryos deficient in RANK and LtβR signaling were identical to pro-pMECs detected in the Relb^−/− embryonic thymus. Thus, we hypothesized that LtβR and RANK signaling may cooperatively promote the differentiation of pro-pMECs into pMECs.

We next investigated whether pro-pMECs were capable of differentiating into Aire⁺ mTECs. Pro-pMECs were sorted from the thymus of Ltbr^−/− RANK-EGFP mice at E15.5 treated with RANKL-Ab (Fig. S2). Sorted pro-pMECs showed higher levels of CD24 expression and lower levels of EGFP expression as compared with pMECs (Fig. 6 A) and were reaggregated with the aly/aly embryonic thymus. Flow cytometric analysis of the RTOC after 5 d of culture indicated the generation of EGFP⁺UEA-1⁺CD80⁺ mTECs in RTOCs mixed with pro-pMECs (Fig. 6 A). Further RTOC experiments suggested that pro-pMECs (H2-Kb⁺) were converted into RANK⁺UEA-1⁺TECs, whereas fewer pro-pMECs were converted into UEA-1^– TECs (Fig. 6 B). In addition to CD80⁺RANK⁺UEA-1⁺ TECs, the H2-Kb⁺ fraction contained CD80^–UEA-1⁺ TECs that expressed relatively high levels of EGFP as compared with that in pro-pMECs (Fig. 6 B). This implied that pro-pMECs may be converted into pMECs in addition to mature mTECs. Moreover, immunohistochemical staining suggested the development of EGFP⁺Aire⁺UEA-1⁺ mTECs in RTOCs with pro-pMECs (Fig. 6 C). Overall, these data supported the differentiation of pro-pMECs into Aire⁺ mTECs in RTOC. Conversion of pro-pMECs into Aire⁺ mTECs was less efficient than that of pMECs, which may be explained by the absence of LtβR in addition to the limited niches expressing RANKL.

To further confirm our hypothesis, we directly injected limited numbers of pro-pMECs and pMECs (10 cells) into the aly/aly thymus (Fig. 6 D), which would minimize the possibility of contamination of other cell types. Pro-pMECs and pMECs were then sorted from RANK-EGFP mice treated with RANKL-Ab at E14.5 (Fig. S3). Immunohistochemical staining indicated that areas consisting of UEA-1⁺ cells containing Aire⁺ mTECs were detected in the grafted thymus injected with pro-pMECs or pMECs (Fig. 6 D). On the other hand, UEA-1⁺ areas were not found in thymi injected with EpCAM^– cells, with the exception of one sample. These data further supported the idea that pro-pMECs and pMECs were capable of differentiating into Aire⁺ mTECs.

In vivo data indicated the requirements of RANK/LtβR and RelB for pMEC induction and RANK and TRAF6 for pMEC differentiation into Aire⁺ mTECs, respectively. Finally, we confirmed that these receptors and downstream molecules were indeed linked.

Stimulation of fetal thymic stroma organ culture (2DG-FTOC) with recombinant RANKL induces Aire⁺ mTECs (Rossi et al., 2007; Akiyama et al., 2008, 2014). Furthermore, stimulation with agonistic LtβR-Ab enhances RANK mRNA expression in 2DG-FTOC without inducing Aire⁺ mTECs (Mouri et al., 2011). However, cells in which RANK expression is up-regulated by LtβR signaling have not been examined. Stimulation of 2DG-FTOC from RANK-EGFP embryos treated with an agonistic LtβR antibody induced EGFP⁺UEA-1⁺MHCII^mid TECs (Fig. 7 A). Such induction was completely abolished in Relb^−/− 2DG-FTOC. These data suggested that LtβR signaling induced pMECs in a RelB-dependent manner, consistent with in vivo genetic data.

In contrast to LtβR signaling, stimulation with RANK ligand induced both EGFP⁺MHCII^hi mTECs and EGFP⁺UEA-1⁺MHCII^mid TECs in 2DG-FTOC (Fig. 7 A). The absence of RelB completely inhibited RANKL-dependent induction of these cell types (Fig. 7 A). Thus, RANKL induced mature mTECs and pMECs in a RelB-dependent manner. RelB expression should be required in RANKL-dependent mature mTECs because mature mTECs are derived from pMECs induced by RANK signaling. In contrast to Relb^−/− 2DG-FTOC, EGFP⁺UEA-1⁺MHCII^mid TECs were induced by RANKL stimulation in the absence of TRAF6 (Fig. 7 B). RANKL-dependent induction of MHCII^hi mTECs did not occur in Traf6^−/− 2DG-FTOC (Fig. 7 B). Thus, although RANK signaling required TRAF6 for differentiation of pMECs into mature mTECs, TRAF6 was dispensable for RANK signaling-dependent pMEC induction in the fetal thymic stroma. Consequently, these data strongly suggested that RANK signaling promoted two different stages in mTEC differentiation through two distinct intracellular signaling pathways (i.e., TRAF6 and RelB dependent) for each differentiation stage.

The requirement for RelB in RANK and LtβR signaling-dependent pMEC induction suggested that activation of the nonclassical NF-κB pathway may promote pMEC induction. To further verify this mechanism by gain-of-function experiments, we used MV-1, a synthetic inhibitor of cellular inhibitors of apoptosis (cIAPs) 1 and 2 (Varfolomeev et al., 2007). Because cIAPs function as negative regulators of the nonclassical NF-κB pathway by inducing degradation of NIK in resting cells (Akiyama et al., 2012), inhibition of cIAPs by MV-1 activates the nonclassical NF-κB pathway through accumulation and activation of NIK without ligand stimulation (Varfolomeev et al., 2007). MV-1 treatment efficiently induced RANK⁺UEA-1⁺MHCII^mid TECs (Fig. 7 C), despite unexpected increases in UEA-1 ligand expression in the cTEC fraction (Fig. 7 C). Thus, MV-1 treatment was sufficient for inducing pMECs in 2DG-FTOC. As expected, the MV-1–dependent induction of pMECs was abolished by the absence of RelB (Fig. 7 C), confirming that pMEC induction by MV-1 was mediated by the nonclassical NF-κB pathway, leading to RelB activation. Furthermore, treatment of 2DG-FTOC with a combination of MV-1 and RANKL demonstrated that MV-1 significantly enhanced the RANKL-dependent induction of MHCII^hi mature mTECs (Fig. 7 D). Moreover, qPCR analysis indicated that MV-1 enhanced the RANKL-dependent up-regulation of Aire and TSA, whereas treatment with MV-1 alone did not (Fig. 7 E). These data suggested that MV-1 induced large numbers of pMECs by efficient activation of the nonclassical NF-κB pathway; in turn, RANKL induced an increase in the number of mature Aire⁺ mTECs.

From these results, we propose a mechanism of embryonic Aire⁺ mTEC differentiation as follows. First, RANK and LtβR signaling activate RelB through the nonclassical NF-κB pathway in pro-pMECs expressing low levels of RANK, thereby initiating differentiation of pro-pMECs into pMECs expressing high levels of RANK. Subsequently, TRAF6-dependent RANK signaling drives the differentiation of pMECs into Aire⁺ mTECs (Fig. 8).

Ly51^– mTECs expressing low levels of MHCII and CD80 are thought to be immature mTECs that differentiate into Aire⁺ mTECs (Gäbler et al., 2007; Gray et al., 2007). In contrast, pMECs and pro-pMECs show Ly51 expression, which is down-regulated when pMECs differentiate into mature mTECs. Thus, pMECs should represent an earlier differentiation stage than Ly51^– immature mTECs. Immature Ly51^– mTECs may represent a transitional stage between pMECs and Aire⁺ mTECs and are converted in Aire⁺ mTECs without additional signaling.

Recent studies have proposed that common precursor cells for mTECs and cTECs express the cTEC molecules β5t and CD205 (Baik et al., 2013; Ohigashi et al., 2013, 2015; Ribeiro et al., 2013; Mayer et al., 2016). Our data suggested that pMECs express high levels of the cTEC molecules β5t, Ly51, and keratin-8. Thus, the cTEC gene expression profile of common precursor cells may be retained in pro-pMECs and pMECs during their differentiation. A recent study suggested that claudin3/4 (Clds)⁺ SSEA-1⁺ TECs show properties of mTEC stem cells (Sekai et al., 2014). Importantly, this cell fraction does not express UEA-1 ligand and RANK (Sekai et al., 2014; Baik et al., 2016). Furthermore, a recent study indicated that differentiation of Clds⁺SSEA-1⁺ TECs did not require RelB (Baik et al., 2016). Therefore, pro-pMECs and pMECs may be differentiated from Clds⁺SSEA-1⁺ TECs. The mechanisms linking pro-pMECs and pMECs with SSEA-1⁺Clds⁺ TECs should be clarified in the future.

In vitro RTOC experiments suggested that pMECs could be converted into Aire⁺ mTECs after 5 d of culture. Intriguingly, Aire⁺ mTECs were also observed in the thymus derived from RTOCs mixed with pMECs at 8 wk after kidney transplantation. Because Aire⁺ mTECs turn over rapidly (Gray et al., 2007), these data suggested that pMECs may give rise to Aire⁺ mTECs from 5 d to 8 wk after transplantation. Moreover, inoculation of the transplanted fetal thymus with 10 pMECs resulted in a medullar region containing Aire⁺ mTECs at 4 wk after transplantation, supporting the proliferative nature of pMECs or their progeny. Further studies are needed to elucidate the proliferative and stem cell-like properties of pMECs.

Recent studies have suggested the high heterogeneity of mTECs in the adult thymus (Lkhagvasuren et al., 2013; Danzl et al., 2014). LtβR signaling preferentially promotes the development of a subset of mTECs and expression of some chemokines (Seach et al., 2008; White et al., 2010; Lkhagvasuren et al., 2013). Our data suggested that both RANK and LtβR signaling induced pMECs. However, pMECs induced by RANK signaling may be inherently different from pMECs induced by LtβR signaling. Such differences in pMEC stages could indirectly influence the properties of mature mTECs, even if RANK signaling promotes the differentiation of both types of pMECs into mature mTECs. Additionally, our findings may clarify the role of lymphotoxin signaling in Aire⁺ mTECs. A previous study showed that excess lymphotoxin signaling increases Aire⁺ mTECs in vivo (Chin et al., 2003); however, other studies have reported that lymphotoxin is not directly associated with differentiation of Aire⁺ mTECs (Venanzi et al., 2007) and is not essential for Aire⁺ mTEC differentiation (Boehm et al., 2003). Our model explains the sufficiency of excess lymphotoxin signaling in Aire⁺ mTEC induction, even though this pathway may not be essential for Aire⁺ mTEC differentiation.

Our finding suggested distinct roles of the TRAF6-dependent pathway and nonclassical NF-κB pathway induced by RANK signaling. To the best of our knowledge, this is the first example of the physiological function of RANK signaling independent of TRAF6. Interestingly, TRAF6-independent RANK signaling up-regulated RANK expression. High surface expression of RANK may be a prerequisite for RANK signaling controlled by a higher-order TRAF6-oligomerization complex (Yin et al., 2009). The involvement of this auto-amplification mechanism for RANK expression should be verified in other RANK signaling-mediated cell differentiation events.

In summary, our data suggested that introduction of pMECs into the thymus efficiently suppressed autoimmunity induced by mTEC dysfunction. Given that autoimmune polyendocrinopathy-candidiasis-ectodermal dystrophy and/or other autoimmune diseases are caused by mTEC dysfunction as a result of Aire mutations (Mathis and Benoist, 2009; Akiyama et al., 2013), pMECs may be used to prevent the onset of such autoimmune diseases. Further characterization of pMECs is needed to develop such cell-based therapies.

C57BL/6, BALB/cA, BALB/cA nu/nu, and aly/aly mice were obtained from CLEA. Aly/aly mice on the BALB/cA background (backcrossed 10 times) were described previously (Shinzawa et al., 2011). RANK-EGFP mice on the C57BL/6 background were also described previously (Maeda et al., 2012). RANK-EGFP mice on a BALB/cA background were prepared by backcrossing with BALB/cA mice for 10 generations. Traf6^−/−, Rank^−/−, Ltbr^−/−, and Rankl^−/− mice were described previously (Akiyama et al., 2005, 2008; Mouri et al., 2011). Relb^−/− mice were purchased from the Jackson Laboratory. Littermate mice were used as controls. All mice were maintained under specific pathogen–free conditions and were handled in accordance with the Guidelines for Animal Experiments of the Institute of Medical Science (The University of Tokyo, Tokyo, Japan). The morning of finding a vaginal plug was designated as E0.5.

The following fluorescent anti–mouse antibodies were used for flow cytometry: anti-CD16/32, APCCy7-anti-CD45 (clone 30 F-11), APCCy7- or PECy7-anti-TER119 (clone TER-119), PE- or PECy7-anti-EpCAM (CD326, clone G8.8), FITC- or PECy7-anti-MHCII (I-A/I-E, clone M5/114.15.2), APC-anti-Ly51 (clone 6C3; BioLegend), PE-anti-CD80 (clone 16-10A1), PE-anti-CD24 (clone M1/69; eBioscience), and biotinylated UEA-1 (Vector Laboratories) antibodies. An agonistic anti-LtbR antibody was purchased from Alexis Biochemicals. Recombinant RANKL was a generous gift from Oriental Yeast Co. Anti-RANKL antibodies were prepared from a subclone of hybridoma obtained by fusing mouse myeloma cells with B cells.

Embryonic thymi were digested in RPMI-1640 medium containing Collagenase/Dispase (Roche) at 0.125% wt/vol plus DNase I (Sigma-Aldrich) at 0.01% wt/vol by incubation at 37°C for 20 min (twice), with pipetting after each incubation. Reactions were stopped by addition of 10 mM EDTA and washing with FACS buffer. Single-cell suspensions were stained with anti–mouse antibodies as described above. Dead cells were excluded with 7-aminoactinomycin D staining. Cells were sorted using a cell sorter (FACSAria; BD). The sorting efficiency was routinely >98%. To compare MFI values in different experiments, external positive controls and/or internal positive and negative controls were used to adjust staining and measurement conditions. Otherwise, MFI values were compared in same day experiments.

For isolation of pMECs, 100 µg anti-RANKL-Ab was intraperitoneally injected into RANK-EGFP mice at 13 d after vaginal plug detection. pMECs (3 × 10⁴) from E17.5 RANK-EGFP mice treated with anti-RANKL-Ab were reaggregated together with trypsin-digested thymic cells (5 × 10⁵) from E15.5 or E16.5 aly/aly mice on the BALB/cA background. Anti-RANKL-Ab (100 µg) was subcutaneously injected into LtbR^−/− RANK-EGFP mice at 9 d after vaginal plug detection when pro-pMECs were sorted. Pro-pMECs (3 × 10⁴) from E15.5 LtbR^−/− RANK-EGFP mice treated with RANKL-Ab were reaggregated with digested thymic cells (5 × 10⁵) from E15.5 or E16.5 aly/aly mice. RTOCs were cultured on Nucleopore filters (Whatman) placed in R10 medium containing RPMI-1640 (Invitrogen) supplemented with 10% FBS, 2 mM l-glutamine (Wako), 1× nonessential amino acids (NEAAs; Sigma-Aldrich), 100 U/ml penicillin (Ban-yu), 100 µg/ml streptomycin (Meiji), and 50 µM 2-mercaptoethanol (Wako) for 5 d. RTOCs were analyzed by FACS, RT-PCR, and immunohistochemical staining analysis. For implantation experiments, RTOCs were grafted under the kidney capsule of 6-wk-old female nude mice.

Thymic lobes were isolated from E15.5 mice and cultured for 4 d on Nucleopore filters (Whatman) placed on R10 medium containing 1.35 mM 2′-deoxyguanosine (2DG; Sigma-Aldrich). Fetal thymic stroma (2DG-FTOC) was subsequently cultured in R10 with recombinant RANKL (1 µg/ml; Oriental Yeast Co.), agonistic anti-LtbR monoclonal antibodies (1 µg/ml), or MV1 (5 µM). 5 d after ligation, 2DG-FTOC was used for the analysis.

For immunohistochemical staining, samples were embedded in OCT compound frozen on dry ice and sectioned into 5-µm slices. Cryosections were fixed for 20 min in 4% paraformaldehyde in PBS on ice and washed with PBS containing 10 mM glycine. After treatment with 0.1% Triton X in PBS for 5 min on ice, sections were blocked with 10% goat serum in PBS. RANK expression in RANK-EGFP mice was detected using a polyclonal chicken anti-GFP antibody (1:300; Abcam) followed by a secondary goat anti–chicken Alexa Fluor 488–conjugated antibody (1:100; Invitrogen). Keratin-5 was detected using a polyclonal rabbit anti–mouse K5 antibody (1:300; Covance), followed by a secondary goat anti–rabbit Alexa Fluor 546–conjugated antibody (1:100; Invitrogen). Aire was detected using a monoclonal rat anti-Aire antibody (1:300; clone 5H12; eBioscience) and a secondary goat anti–rat Alexa Fluor 546–conjugated antibody (1:100; Invitrogen). Biotin UEA-1 (1:50; Vector Laboratories) and streptavidin Alexa Fluor 647 (1:100; Invitrogen) were used to detect UEA-1–binding protein.

Total RNA was extracted using an RNeasy Micro kit (QIAGEN). cDNA was obtained by random-primed RT using a PrimescriptII First Strand cDNA Synthesis kit (Takara Bio, Inc.). qPCR was performed using an ABI PRISM 7300 Sequence Detection System and SYBR Green Master Mix (Toyobo). Expression values for each sample were normalized to GAPDH or 36B4.

Organs from nude mice were stained with hematoxylin and eosin at 8 wk after implantation. The degrees of inflammatory cell infiltration were scored as follows: 0, no detectable infiltration; 1, a focus of perivascular infiltration; 2, several foci of perivascular infiltration; 3, cellular infiltration in >50% of the vasculature; 4, severe infiltration throughout the interstitial region. To detect autoantibodies in serum, tissue samples from Rag2^−/− mice were embedded in OCT compound, frozen on dry ice, and sectioned into 5-µm-thick slices. Cryosections were fixed with ice-cold acetone for 5 min and washed with PBS. Samples were blocked with 10% goat serum in PBS and incubated with serum (100× dilution) for 1 h at room temperature. The slides were subsequently incubated with secondary antibody (anti–mouse IgG-Alexa Fluor 488) and 1 µg/ml propidium iodide for 40 min at room temperature. Confocal color images were obtained using an FV1000D (Olympus).

100 µg anti–RANKL-Ab was injected subcutaneously into RANK-EGFP mice (B6) 9 d after vaginal plug detection. Thymic lobes were dissected from E14.5 embryos, and thymic stromal cells were prepared as previously described (Akiyama et al., 2014). Cells were sorted using a cell sorter (BD Aria; BD). Sorted cells were injected into E14.5 thymic lobes of aly/aly mice (10 cells/lobe). After culture for 1 d, thymic lobes were transplanted under the kidney capsule of 6-wk-old female nude mice. At 4 wk after transplantation, grafted thymi were subjected to immunohistochemical analysis.

DNA methylation analysis was performed as described previously (Wu et al., 2012). In brief, cells isolated from the embryonic thymus were subjected to bisulfite treatment using an EZ DNA Methylation-Direct kit (Zymo Research Cooperation). PCR of Aire was performed using bisulfite-converted genomic DNA with the following forward and reverse primers: 5′-TTTGATTTAAATATTTGTTGGATAAGGA-3′ and 5′-CCCATTTTAATACTCCCAATCTTCT-3′, respectively. COBRA was performed by evenly dividing the PCR products to incubate with HpyCH4IV for 5 h at 37°C, and samples were then analyzed with a microchip electrophoresis system (MultiNA; Shimadzu Biotech). Unmethylated DNA levels were calculated as the percentage of undigested fragment compared with the sum of the digested and undigested fragments.

N- and C-terminal protected Boc-MV-1 was synthesized using conventional solution-phase methods, in which 1-(3-dimethylaminopropyl)-3-ethylcarbodiimide (EDC) hydrochloride and 1-hydroxybenzotriazole (HOBt) hydrate were used as coupling reagents. The coupling conditions were as follows: N-terminal free amino acid or peptide (1 Eq.), Boc-protected amino acid (1.2 Eq.), EDC (1.5 Eq.), HOBt (1.5 Eq.), and N,N-diisopropylethylamine (DIPEA, 2 Eq.) in dichloromethane (CH₂Cl₂) at room temperature for 5 h. The deprotection conditions were as follows: Boc-protected peptide (1 Eq.), trifluoroacetic acid (TFA, 10 Eq.) in CH₂Cl₂ at room temperature for 30 min. Boc-MV-1: ¹H NMR (400 MHz, CDCl₃): δ 7.19–7.33 (m, 11H), 6.70 (br s, 1H), 5.37 (t, 1H, J = 8.4 Hz), 4.47 (m, 1H), 4.38 (d, 1H, J = 8.4 Hz), 3.62 (m, 1H), 3.49 (s, 3H), 3.38 (m, 1H), 2.79 (s, 3H), 2.07 (m, 1H), 1.61–1.82 (m, 10H), 1.49 (s, 9H), 1.29 (d, 3H, J = 10.8 Hz), 0.96–1.24 (m, 6H); [HR-ESI(+)]: m/z calculated for C₃₈H₅₃N₄O₇ [M+H]⁺: 677.3914, found: 677.3920. A solution of Boc-MV-1 (1.35 g, 2.0 mmol) in 4 M HCl in 1,4-dioxane (6 ml) was prepared at 0°C, and the mixture was stirred at room temperature for 4 h. Removal of the solvent afforded MV-1 hydrochloride salt (1.22 g, >99%), as a colorless crystal. MV-1: [HR-ESI(+)]: m/z calc’d for C₃₃H₄₅N₄O₅ [M+H]⁺: 577.3390, found: 577.3368.

P-values were calculated using Student’s t tests with a two-tailed distribution and two-sample equivalent variance parameters or Mann-Whitney U test.

Figs. S1–S3 include gating strategies for isolation of pMECs, pro-pMECs, and other TECs.

This work was supported by a Grant-in-Aid for Challenging Exploratory Research from Japanese Society for the Promotion of Science (26670234; T. Akiyama); a Grant-in-Aid for Scientific Research on Innovative Areas from Ministry of Education, Culture, Sports, Science, and Technology in Japan (15H01150 and 16H01631; T. Akiyama); Princess Takamatsu Cancer Research Fund (T. Akiyama); and a grant from the Specially Promoted Project of the Toyota Physical and Chemical Research Institute (T. Akiyama).

H. Yasuda belongs to Oriental Yeast Co. Ltd., which provided the recombinant RANKL and anti–RANKL-Ab used in this study. The authors declare no additional competing financial interests.