The thymic medulla is required for Foxp3⁺ regulatory but not conventional CD4⁺ thymocyte development

As positive selection involves changes in the chemokine receptors CCR7 and CCR9 (Choi et al., 2008), we used these and other (HSA/CD69/CD62L/Qa2) markers to define distinct subsets of SP4 thymocytes in Rag2GFP reporter mice, in which GFP levels are a measure of maturation status (Boursalian et al., 2004; McCaughtry et al., 2007). Fig. 1 A shows that CD69⁺αβ TCR^hi SP4 thymocytes contain CCR7⁺CCR9⁻ and CCR7^−/loCCR9⁺ subsets, whereas more mature CD69⁻ cells are CCR7⁺CCR9⁻. Analysis of CD62L/Qa2 showed the CD69⁺CCR7^−/loCCR9⁺ subset to be CD62L^loQa2^lo and to express the highest levels of Rag2GFP among SP4 thymocytes, whereas CD69⁻CCR7⁺CCR9⁻ cells were CD62L^hiQa2^hi with the lowest levels of Rag2GFP (Fig. 1 B). CD69⁺CCR7⁺CCR9⁻ and CD69⁻CCR7⁺CCR9⁻ but not CD69⁺CCR7^−/loCCR9⁺ SP4 thymocytes (Fig. 2 A) expressed mRNA for Foxo1 and Klf2, known regulators of thymocyte egress (Carlson et al., 2006; Gubbels Bupp et al., 2009), whereas mRNA encoding S1pr1, another regulator of thymocyte emigration (Matloubian et al., 2004), was limited to the CD69⁻CCR7⁺CCR9⁻ subset (Fig. 2 A). Importantly, when reaggregate thymus organ cultures (Rossi et al., 2007) were initiated with purified CD69⁺CCR7^−/loCCR9⁺ SP4 thymocytes, their progeny had up-regulated CCR7 and down-regulated both CCR9 (Fig. 2 B) and CD69 (not depicted), providing evidence for a maturation sequence in which the CD69⁺CCR7^−/loCCR9⁺ subset represents newly generated SP4 thymocytes after positive selection, followed by CD69⁺CCR7⁺CCR9⁻ and then CD69⁻CCR7⁺CCR9⁻ cells (Fig. 2 C).

In relation to nT_reg cell emergence, Foxp3 mRNA was detectable in CD69⁺CCR7⁺CCR9⁻ SP4 thymocytes, with higher levels noted in the most mature CD69⁻CCR7⁺CCR9⁻ subset (Fig. 2 A), indicating that Foxp3⁺ nT_reg cell development is first detectable within the CD69⁺ stage of SP4 thymocyte maturation but after induction of CCR7 and loss of CCR9. To further relate changes in CCR7/CCR9 to distinct stages in nT_reg cell development, we analyzed CD25 and Foxp3 by flow cytometry. Consistent with quantitative PCR (qPCR) data, the earliest CD69⁺CCR7^−/loCCR9⁺ cells were Foxp3⁻, and Foxp3⁻CD25⁺ nT_reg cell precursors were barely detectable in this population (Fig. 2 D). In contrast, CD69⁺CCR7⁺CCR9⁻ SP4 thymocytes contained Foxp3⁻CD25⁺ nT_reg cell precursors as well as their more mature CD25⁺Foxp3⁺ progeny (Fig. 2 D), whereas the most mature CD69⁻CCR7⁺CCR9⁻ SP4 thymocytes contained Foxp3⁺ nT_reg cells but lacked Foxp3⁻CD25⁺ precursors (Fig. 2 D). Thus, Foxp3⁻CD25⁺ nT_reg cell precursor appearance maps to the transition between CCR7^−/loCCR9⁺ and CCR7⁺CCR9⁻ stages in the CD69⁺ phase of SP4 thymocyte development (Fig. 2 C).

A recent study has suggested a correlation between the thymic medulla and SP4 thymocyte development, most notably the absence of CD69⁻Qa2⁺ SP4 thymocytes in Relb^−/− mice displaying a severe block in mTEC development (Fig. 3 A; Li et al., 2007). However, detailed analysis of the role of mTECs in Relb^−/− mice is confounded by their complex phenotype that includes DC deficiencies and multiorgan autoimmunity (Weih et al., 1995; Wu et al., 1998). We therefore investigated the mTEC requirements of both conventional and nT_reg SP4 thymocytes by grafting alymphoid 2-deoxyguanosine (dGuo)–treated fetal thymus organ cultures (Rossi et al., 2007), from either Relb^−/− or WT embryos, into unmanipulated WT mice. Importantly, any absence of Qa2⁺ cells in our experiments is not caused by the genetic background of the Relb^−/− mice used here, as Relb^−/− fetal liver chimaeras with WT hosts generated Qa2⁺ SP4 thymocytes (not depicted). Confocal analysis of WT and Relb^−/− TEC grafts confirmed an mTEC defect in the latter (Fig. 3 B), consistent with a cell-intrinsic role for RelB in mTEC development. Importantly, although grafting of Relb^−/− TECs into nude hosts resulted in autoimmunity (Fig. 4, A and B), WT hosts grafted with Relb^−/− TECs showed no signs of disease (not depicted), presumably as a result of peripheral tolerance mechanisms involving T_reg cells generated in the host WT thymus. Thus, grafting of unmanipulated WT hosts with thymuses harboring a cell-intrinsic RelB-dependent mTEC deficiency provides a model to study the role of mTECs in SP4 thymocyte development in the absence of autoimmunity.

Analysis of SP4 thymocytes in Relb^−/− TEC grafts within WT hosts revealed both early CD69⁺Qa2⁻ and late CD69⁻Qa2⁺ subsets at proportions and numbers comparable with WT grafts (Fig. 3, C and D). Importantly, SP4 thymocytes in WT and Relb^−/− grafts were predominantly CD44^loHSA^int (Fig. 3 E) and expressed lower levels of Qa2 compared with peripheral CD4 T cells (Fig. 3 F), indicating that they were not peripheral T cells circulating back to the thymus (McCaughtry et al., 2007; Hale and Fink, 2009). Our findings do not support previous suggestions from experiments involving in vitro TEC lines (Li et al., 2007), that mTECs provide essential support for SP4 thymocyte maturation. The reason for this difference is unknown, although it may relate to whether in vitro systems reflect the functional capacities of mTECs in vivo. Instead, our data show that conventional SP4 thymocyte development in vivo occurs in the absence of RelB-dependent mTECs. Interestingly, absence of Qa2⁺CD69⁻ SP4 thymocytes was also reported in Aire^−/− mice (Li et al., 2007). Given that Aire^−/− mice show an increased mTEC compartment (Anderson et al., 2002), the impact made by Aire on thymocyte development is unclear.

We next analyzed the requirements for thymic support after positive selection by investigating the ability of CD69⁺CCR7^−/loCCR9⁺ SP4 thymocytes to mature extrathymically. Although recent thymic emigrants undergo maturation outside the thymus (Boursalian et al., 2004), it is unclear how far back in development this window of thymic independence extends, particularly in relation to the post–positive selection stages described here. Initially, CD69⁺CCR7^−/loCCR9⁺ SP4 thymocytes from adult CD45.2⁺ WT mice, with an HSA^hiCD62L^loQa2⁻CD69⁺ phenotype (Fig. 4 C), were i.v. injected into congenic CD45.1⁺ WT mice. Analysis after 7 (Fig. 4 C) and 14 d (not depicted) showed that injected cells had acquired an HSA^loCD62L^hiQa2^hiCD69^lo phenotype comparable with host naive CD4⁺ T cells. To provide a comparative analysis of extrathymic maturation, CD69⁺CCR7^−/loCCR9⁺ thymocytes from CD45.2⁺ mice were coinjected into CD45.1⁺CD45.2⁺ hosts at a 1:1 ratio with either less mature CD45.1⁺CD4⁺8⁺69⁺ or more mature CD45.1⁺CD69⁻CCR7⁺CCR9⁻ SP4 thymocytes. Analysis of spleen 7 d later revealed a ratio of 10:1 after coinjection of CD69⁺CCR7^−/loCCR9⁺ SP4 and less mature CD4⁺8⁺69⁺ thymocytes and a 0.2:1 ratio after coinjection of CD69⁺CCR7^−/loCCR9⁺ SP4 and more mature CD69⁻CCR7⁺CCR9⁻ SP4 thymocytes (Fig. 4 D). Thus, CD69⁺CCR7^−/loCCR9⁺ SP4 thymocytes show an emerging capacity for thymic independence, strengthening the notion that conventional SP4 thymocyte development can occur in the absence of thymic medullary support from the earliest SP4 thymocyte stages. Although mice receiving CD69⁺CCR7^−/loCCR9⁺ SP4 thymocytes showed no signs of autoimmunity (not depicted), it is unclear whether this reflects the timing of negative selection in relation to this subset or the control of autoreactivity by T_reg cells generated in the WT host thymus.

We next examined the role of RelB-dependent mTECs in the development of Foxp3⁺ nT_reg cells and Foxp3⁻CD25⁺ precursors. CD69⁺Foxp3⁻CD25⁺ and CD69⁺CCR7^−/loCCR9⁺ subsets of SP4 thymocytes, the former reflecting nT_reg cell precursors (Lio and Hsieh, 2008) and the latter capable of extrathymic generation of conventional CD4 T cells (Fig. 4), were purified from CD45.2⁺ Foxp3GFP reporter mice and transferred i.v. into WT CD45.1⁺ hosts. CD69⁺Foxp3⁻CD25⁺ nT_reg cell precursors, shown here to be at a CD69⁺CCR7⁺CCR9⁻ intermediate stage (Fig. 2 D), generated Foxp3⁺CD25⁺ nT_reg cells (Fig. 5 A). In contrast to their capacity to generate conventional CD4 T cells extrathymically (Fig. 4), CD69⁺CCR7^−/loCCR9⁺ SP4 thymocytes failed to give rise to Foxp3⁺CD25⁺ nT_reg cells outside the thymus (Fig. 5 A). Thus, the CD69⁺CCR7^−/loCCR9⁺ stage after positive selection marks a point where continued maturation of conventional but not Foxp3⁺ regulatory SP4 thymocytes can occur independently of thymic support.

Importantly, and unlike conventional SP4 thymocyte development, Relb^−/− TEC grafts showed a significant reduction in Foxp3⁺ SP4 thymocytes compared with WT grafts (Fig. 5, B–D). Furthermore, nT_reg cell precursors were also significantly reduced in mTEC-deficient Relb^−/− TEC grafts (Fig. 5, B–D). Although the few nT_reg cells in Relb^−/− grafts do not rule out inefficient generation via residual RelB-independent mTECs, the large majority of nT_reg cell development appears to be controlled by RelB-dependent mTECs. Moreover, although the presence of host-derived CD11c⁺ DCs in both WT and Relb^−/− grafts (Fig. 5 E) argues against the defect in nT_reg cell production being solely caused by the absence of DCs, we cannot exclude impaired DC function in the absence of RelB-dependent mTECs.

In conclusion, we show that conventional and Foxp3⁺ nT_reg CD4 T cells demonstrate a differential requirement for RelB-dependent medullary thymic microenvironments during their development. In particular, mTECs support the generation of Foxp3⁻CD25⁺ nT_reg cell precursors and their Foxp3⁺CD25⁺ nT_reg cell progeny from within the pool of medullary CCR7⁺CCR9⁻ SP4 thymocytes. The dependency of Foxp3⁻CD25⁺ nT_reg cell precursor generation on both CD28–CD80/CD86 and TCR–MHC interactions (Lio and Hsieh, 2008; Lio et al., 2010; Vang et al., 2010; Hinterberger et al., 2011) fits well with the expression of these co-stimulatory molecules and MHC class I/II by mTECs (Rossi et al., 2007) and suggests that provision of these molecules by mTECs is linked to their ability to support nT_reg cell generation as shown here. That mTECs provide TCR ligands for nT_reg cell development fits well with the generation of antigen-specific TCR transgenic T_reg cells after the targeting of model antigens to mTECs (Aschenbrenner et al., 2007; Hinterberger et al., 2010) and the normal numbers of nT_reg cells generated when hematopoietic cells are selectively MHC class II deficient (Aschenbrenner et al., 2007; Liston et al., 2008). In contrast, the ability of conventional SP4 thymocytes to continue their maturation, either in the absence of RelB-dependent mTECs or extrathymically, reveals differences in the maturational requirements for nT_reg cells and conventional T cells and warrants a rethinking of the role of the medulla in T cell development. Thus, rather than representing a microenvironment fostering late-stage αβ T cell development per se, the primary role of the medulla is in the generation of self-tolerance via negative selection and the generation of Foxp3⁺ CD4 nT_reg cells, a scenario compatible with the T cell–mediated autoimmunity in Ccr7^−/− mice that occurs after inefficient thymocyte access to the medulla (Kurobe et al., 2006).

WT CD45.2⁺ C57BL/6, congenic CD45.1⁺ C57BL/6 (BoyJ), CD45.1⁺CD45.2⁺ C57BL/6, nude C57BL/6, C57BL/6 Relb^−/− (Weih et al., 1995) mice, FVB/N RAG2GFP (Yu et al., 1999), and C57BL/6 Foxp3GFP reporter mice (gift from T. Strom, Beth Israel Deaconess Medical Center, Boston, MA; Bettelli et al., 2006) were bred at the University of Birmingham in accordance with Home Office regulations. For timed matings, the day of vaginal plug detection was designated as day 0. All animal experiments were performed in accordance with University of Birmingham (Local Ethical Review Panel) and national UK Home Office regulations.

Thymocyte, splenocyte, and LN suspensions were stained with the following antibodies: PECy7/Alexa Fluor 700/PE anti-CD4 (clone GK1.5; eBioscience) or PerCP-Cy5.5/APC eFluor 780/V500 anti-CD4 (clone RM4-5; eBioscience/BD), eFluor 450/FITC/V500 anti-CD8 (clone 53-6.7; eBioscience/BD) or biotinylated anti-CD8 clone (YTS156.7.7; BioLegend), APC eFluor 780 anti-TCRβ (clone H57-597; eBioscience), FITC/APC/PerCP-Cy5.5 anti-CD69 (clone H1.2F3; eBioscience), APC anti-CD62L (clone MEL-14; BioLegend), FITC/Alexa Fluor 700 anti-CD44 (clone IM7; eBioscience), PE anti-CD3ε (clone 145-2C11; eBioscience), APC eFluor 780/PE anti-HSA/CD24 (clone M1/69; BD/eBioscience), biotinylated/FITC anti-Qa2 (clone 695H1-9.9; BioLegend/eBioscience), eFluor 780/eFluor 450 anti-CD45.1 (clone A20; eBioscience), PE/Alexa Fluor 700 anti-CD45.2 (clone 104; eBioscience), PE anti-CCR9 (clone eBio CW-1.2; eBioscience), APC/PE anti-CD25 (clone PC61/PC61.5; BioLegend/eBioscience), and APC anti–IFN-γ (clone XMG1.2; BD). For surface CCR7 expression, thymocytes were incubated in recombinant CCL19-Ig (eBioscience), followed by biotinylated goat anti–human Ig (eBioscience). All biotinylated antibodies were picked up with PECy7-conjugated streptavidin (eBioscience). For intracellular staining of Foxp3, cells were fixed and permeabilized using the Foxp3/Transcription Factor Staining Buffer Set (eBioscience) according to the manufacturer’s protocol and stained with PE anti-Foxp3 (clone FJK-16s; eBioscience). Flow cytometry was performed on a Fortessa analyzer using FACSDiva6.2 software (BD), with data subsequently analyzed with FlowJo 8.7 software (Tree Star). Purified populations of thymocytes were sorted from adult thymus suspensions using a MoFlo XDP cell sorter (Beckman Coulter).

Embryonic day (E) 15 thymus lobes, cultured for 7 d in 1.35 mM dGuo were transplanted under the kidney capsule of 8-wk-old recipient mice as described previously (Rossi et al., 2007), with grafts in WT mice and nude mice harvested after 8–12 wk and 6 wk, respectively.

CD69⁺CCR7^−/loCCR9⁺Foxp3⁻ and CD69⁺Foxp3⁻CD25⁺ subsets of SP4 αβ TCR^hi thymocytes were MoFlo sorted from CD45.2⁺ FoxP3GFP adult reporter mice and transferred i.v. in a 300-µl volume of PBS into CD45.1⁺ BoyJ mice. LN and spleen were harvested after 7 d and analyzed by flow cytometry for CD45.1/CD45.2 together with maturation markers to study the persistence of donor-derived cells. For coinjection experiments, purified CD69⁺CCR7^−/loCCR9⁺ SP4 from CD45.2⁺ C57BL/6 mice and either CD4⁺8⁺69⁺ or CD69⁻CCR7⁺CCR9⁻ SP4 from CD45.1⁺ BoyJ mice were mixed at a 1:1 ratio and injected i.v. into CD45.1⁺CD45.2⁺ hosts as described above. After 7 d, injected cells were identified in LN and spleen preparations on the basis of either a CD45.1⁺CD45.2⁻ or CD45.1⁻CD45.2⁺ phenotype by flow cytometry and expressed as a ratio.

Thymus grafts were recovered from host mice, and 7-µm sections were cut, then fixed in acetone, and stained with the following antibodies: the mTEC marker ERTR5 (gift of W. van Ewijk, Riken Yokohama Institute, Yokohama City, Kanagawa, 230-0045), detected with Alexa Fluor 594 goat anti–rat IgM (Invitrogen), the cortical TEC marker rabbit anti-β5t (MBL), detected with donkey anti–rabbit 488 (Invitrogen), FITC-conjugated anti-CD11c (clone N418; eBioscience) followed by rabbit anti-FITC (Invitrogen), and donkey anti–rabbit 488 (Invitrogen). Images were obtained using an LSM 780 microscope and analyzed using LSM software (Carl Zeiss).

C57BL/6 nude mice, grafted with either WT or Relb^−/− dGuo-treated thymus lobes, were sacrificed 6 wk after transplantation, and organs and sera were collected. LN cells from the indicated grafted mice were analyzed for intracellular IFN-γ cytokine production and CD44/CD62L expression as described previously (Gaspal et al., 2011). To detect autoantibodies, sera (1:40 dilution) was added to sections of composite tissue blocks containing the indicated tissues and detected with goat anti–mouse IgG FITC (SouthernBiotech), counterstained with DAPI. Lymphocytic infiltrates were detected in paraffin wax–embedded sections of liver as described previously (Rossi et al., 2007). Images were acquired with a microscope (DM6000; Leica) using a 20× objective.

qPCR analysis of freshly sorted thymocyte populations was performed exactly as described previously (Roberts et al., 2012). Primer used are as follows: β-actin QuantiTect Mm Actb 1SG Primer Assay (QIAGEN QT00095242); Foxo1 forward, 5′-TGTCAGGCTAAGAGTTAGTGAGCA-3′; and reverse, 5′-GGGTGAAGGGCATCTTTG-3′; Klf2 forward, 5′-CTCAGCGAGCCTATCTTGCC-3′; and reverse, 5′-CACGTTGTTTAGGTCCTCATCC-3′; S1pr1 forward, 5′-AAATGCCCCAACGGAGACT-3′; and reverse, 5′-CTGATTTGCTGCGGCTAAATTC-3′; Foxp3 forward, 5′-CCCAGGAAAGACAGCAACCTT-3′; and reverse, 5′-TTCTCACAACCAGGCCACTTG-3′; Ccr9 forward, 5′-ACCATGATGCCCACAGAACT-3′; and reverse, 5′-GGGAAGAGTGGCAAGAAAGA-3′; and Ccr7 forward, 5′-CTAGCTGGAGAGAGACAAGA-3′; and reverse, 5′-TATCCGTCATGGTCTTGAGC-3′.

We thank Dr. Terry Strom for Foxp3GFP mice, provided by Dr. Nick Jones.

This study was supported by a Medical Research Council (MRC) Programme Grant. J.E. Cowan is supported by a PhD Studentship from the MRC Centre for Immune Regulation.

The authors have no competing financial interests.