Acquisition of self-tolerance in the thymus requires T cells to discriminate strong versus weak T cell receptor binding by self-peptide–MHC complexes. We find this discrimination is reported by expression of the transcription factor Helios, which is induced during negative selection but decreases during positive selection. Helios and the proapoptotic protein Bim were coinduced in 55% of nascent CCR7− CD4+ CD69+ thymocytes. These were short-lived cells that up-regulated PD-1 and down-regulated CD4 and CD8 during Bim-dependent apoptosis. Helios and Bim were also coinduced at the subsequent CCR7+ CD4+ CD69+ CD8− stage, and this second wave of Bim-dependent negative selection involved 20% of nascent cells. Unlike CCR7− counterparts, Helios+ CCR7+ CD4+ cells mount a concurrent Card11- and c-Rel–dependent activation response that opposes Bim-mediated apoptosis. This “hollow” activation response consists of many NF-κB target genes but lacks key growth mediators like IL-2 and Myc, and the thymocytes were not induced to proliferate. These findings identify Helios as the first marker known to diverge during positive and negative selection of thymocytes and reveal the extent, stage, and molecular nature of two distinct waves of clonal deletion in the normal thymus.
To understand how self-tolerance is acquired in T cells and how it breaks down as a result of inherited predisposition to autoimmune disease, it is necessary to resolve how thymocytes discriminate weak versus strong TCR engagement by self-peptide–MHC (self-pMHC) antigens. This discrimination is central to determining cell fate, as binding to self-pMHC induces positive selection of thymocytes if it is relatively weak but negative selection if it is stronger (von Boehmer, 1990; Gallegos and Bevan, 2006; Gascoigne and Palmer, 2011; Stritesky et al., 2012). Elucidating this selection process requires identification of molecular markers that qualitatively differentiate the thymocyte response to weak versus strong self-pMHC stimulation.
Qualitative differences in the intracellular location of phosphorylated ERK in response to positive versus negatively selecting pMHC have been revealed in vitro (Daniels et al., 2006), but it is difficult to extend this marker to analyses of the natural thymocyte repertoire in vivo. Differences in the magnitude of induction of specific genes and corresponding proteins, notably Bcl2l11 (Bim), Nr4a1 (Nur77), CD69, and PD-1, have been revealed in thymocytes undergoing positive versus negative selection in vivo (Liston et al., 2004b; Baldwin and Hogquist, 2007; Moran et al., 2011). However, because the magnitude of induction during positive selection grades continuously into negative selection, it is difficult to identify a threshold level of these markers that objectively resolves thymocytes making a positive versus negative selection response.
Understanding the molecular differentiation of positive versus negative selection responses also depends on determining whether these responses occur at the same stage of thymocyte maturation in TCR+ CCR7− cortical thymocytes (CCR7 is a chemokine receptor required for thymocyte migration into the medulla [Ueno et al., 2004; Nitta et al., 2009]) or whether negative selection occurs after positive selection in TCR+ CCR7+ medullary thymocytes. Several in vitro (Swat et al., 1991; Kishimoto and Sprent, 1997) and in vivo (Kappler et al., 1987; Pircher et al., 1989; Surh and Sprent, 1994; Pobezinsky et al., 2012) studies support both possibilities, although the need to perform experiments in tissue culture, in TCR transgenic (Tg) mice, or by tracking T cells reacting to superantigens means that the extent of negative selection in CCR7− or CCR7+ thymocytes in the normal repertoire has yet to be directly measured.
Here we show that the transcription factor Helios qualitatively differentiates CD4+ thymocytes making a positive versus negative selection response in vivo. Helios expression increases during negative selection but decreases during positive selection. Using this marker, we enumerate the fraction of thymocytes undergoing negative selection at different stages of helper T cell development, revealing two main waves of negative selection: one in CCR7− CD24+ CD4+/lo CD8+/lo thymocytes and one in CCR7+ CD4+ CD8− thymocytes. Although Bim is strongly induced and required for negative selection at both stages, it is accompanied by inhibitory versus stimulatory programs at the two stages. Card11 signaling concurrently induces many NF-κB–dependent T cell survival and growth genes during negative selection selectively at the CCR7+ stage, representing a “hollow” activation response comprising many but not all genes known to be induced by strong TCR agonists in mature T cells. This Bim-opposing response does not occur during negative selection at the CCR7− stage, and instead the inhibitory receptor PD-1 is strongly coinduced with Bim.
Helios expression differentiates thymocytes undergoing positive and negative selection
To distinguish CD4+ CD8− (single positive [SP]; CD4SP) thymocytes undergoing negative versus positive selection, we first used flow cytometry to compare protein expression in 3A9 TCR Tg thymocytes. In TCR3A9 single-Tg (STg) mice, weak binding of this TCR to self-peptide–I-Ak complexes promoted positive selection of TCRhi CCR7− CD4+ 8+ (double positive [DP]) cortical thymocytes into CCR7+ CD4SP medullary thymocytes (Fig. 1 A). In contrast, in TCR3A9 double-Tg (DTg) mice that also carry an insulin promoter–driven hen egg lysozyme (HEL [insHEL]) transgene, Aire-dependent expression in medullary thymic epithelial cells generated HEL peptide–I-Ak complexes that were bound strongly by TCR3A9 and triggered negative selection of the CCR7+ subset of SP and DP thymocytes that homed to the medulla or corticomedullary junction, a process of deletion previously shown to require CCR7 ligands (Fig. 1, A and B; Lesage et al., 2002; Liston et al., 2003; Nitta et al., 2009). Elimination of CD4SP thymocytes was severely blunted in Bim−/− DTg mice lacking the proapoptotic protein Bim (Fig. 1, A and B), consistent with previous evidence that Bim is essential for thymocyte apoptosis (Bouillet et al., 2002) and that Bim protein and mRNA are induced to higher levels during negative selection (Liston et al., 2004a, 2007; Baldwin and Hogquist, 2007). Bim protein also increased, albeit to a lesser extent, in CD4SP thymocytes undergoing positive selection in STg mice (Fig. 1 A). A similar graded induction of Nur77 (Fig. 1 A), GITR, CD69, and other proteins was observed (not depicted). However, the magnitudes of these differences were too small and the overlap in the distributions too great to resolve thymocytes undergoing positive and negative selection.
A survey of a large number of proteins revealed only one, the Ikaros family transcription factor Helios (Ikzf2), that qualitatively differentiated positive and negative selection responses. Helios was highly expressed in nearly all Foxp3+ T regulatory cells (T reg cells) in the thymus, consistent with published results (Thornton et al., 2010), but when Foxp3+ cells were excluded from the flow cytometric analysis, Helios was also selectively induced in Foxp3− TCR3A9+ CD4SP cells undergoing negative selection (Fig. 1 A). In contrast, Helios was expressed at low levels in DP cells and was down-regulated further in positively selected CD4SP cells from STg mice (Fig. 1 A). Helios was induced in TCR3A9+ CD4SP cells from Bim−/− DTg animals, although this induction was slightly blunted when compared with Bim+/+ DTg mice (Fig. 1, A and C). Helios induction increased markedly when Bim−/− HEL-specific thymocytes were present at low frequencies in mixed bone marrow chimeras (Fig. 1 D), suggesting that the slight blunting in Bim−/− DTg mice was caused by increased intraclonal competition for a limited number of HEL-presenting APCs. In Bim-deficient STg mice, positive selection into TCR3A9+ CD4SP cells occurred normally and was still accompanied by Helios down-regulation (Fig. 1 C). Thus, strongly self-reactive Helios+ Foxp3− TCR3A9+ CD4SP thymocytes increased in number 10-fold in Bim−/− compared with Bim+/+ DTg mice, whereas the number of weakly self-reactive Helios− Foxp3− TCR3A9+ CD4SP cells that were positively selected in STg mice was not altered by Bim deficiency (Fig. 1 C).
Helios induction occurs in a large fraction of CCR7− and CCR7+ CD4SP thymocytes
To track Helios expression in thymocytes from mice with natural TCR repertoires, we subdivided CD4SP thymocytes into maturation stages. DP cells have a CD24+ CCR7− phenotype, and these markers divide CD4SP cells into subsets that suggest CCR7 is induced and then CD24 is down-regulated (Fig. 2 A; Davalos-Misslitz et al., 2007). This proposed developmental sequence was corroborated by progressive increase in expression of the thymocyte maturation marker Qa-2 (McCaughtry et al., 2007), allowing resolution of three CD4SP subsets referred to here as SP1–SP3 (Fig. 2 A): CD24+ CCR7− (SP1), CD24+ CCR7+ (SP2), and CD24− CCR7+ (SP3).
To confirm the proposed developmental sequence above, we used these markers to extend published experiments (Lucas et al., 1993) tracing thymocyte maturation after in vivo pulse labeling with BrdU, which becomes incorporated into the DNA of proliferating DP thymocytes (Fig. 2 B). Each mouse received a single BrdU injection 1–7 d before sacrifice, so that all thymocyte samples were analyzed on the same day. As expected (Lucas et al., 1993), the BrdU+ thymocytes were predominantly at the DP stage 1–4 d after BrdU injection, whereas by days 6–7, most had a CD4SP phenotype. Subsetting the BrdU+ CD4SP cells showed that most were at the SP1 stage 1–2 d after BrdU injection, and there was a gradual transition to SP2 and SP3 from 3 to 7 d, so that by day 7 most of the BrdU+ CD4SP cells were at the SP3 stage (Fig. 2 B). These results are consistent with a DP–SP1–SP2–SP3 developmental progression during CD4SP maturation.
To extend the evidence above that Helios marks strongly self-reactive thymocytes undergoing negative selection, we analyzed Helios expression in thymocytes in the natural TCR repertoire after gating out Foxp3+ cells. Helios+ Foxp3− thymocytes were detectable in wild-type mice, and their number was dramatically increased when elimination of strongly self-reactive thymocytes was inhibited either in Bim−/− mice or in mice expressing human BCL-2 driven by the mouse Vav promoter (Tg(Vav-BCL2); Fig. 2 C; Bouillet et al., 2002). In contrast, there was little effect on the number of Helios− Foxp3− thymocytes (Fig. 2 C), consistent with the evidence above that Bim deficiency does not alter positive selection. In both Bim−/− and Vav-BCL2 Tg mice, Helios+ cells were strikingly increased among the DP, CD4lo CD8lo, and CD4SP populations, accounting for up to 10% of DP cells, 80% of CD4lo CD8lo cells, 65% of SP1 cells, and 37% of SP2 cells in Bim−/− mice, reflecting 10-fold increases in Helios+ SP1 and SP2 cell numbers (Fig. 2 C). High Helios expression was correlated with high Bim expression at the single cell level, and this was accentuated in thymocytes expressing Tg BCL-2 (Fig. 2 C), which enables Bim to accumulate to high levels in T cells (Jorgensen et al., 2007). Thus, Helios marks thymocytes that have been rescued from Bim-mediated negative selection.
Helios, PD-1, and coreceptor dulling mark T cells that bind strongly to self-antigens during the CCR7-negative stage
To study cells known to be strongly self-reactive in B10.BR mice, we focused on thymocytes using the Vβ5 or Vβ11 TCR V-regions that bind self-superantigens and I-Ek (Dyson et al., 1991; Woodland et al., 1991). In Bim+/+ thymi, there was a 10–20-fold decrease in the number of CD4SP cells expressing Vβ5 or Vβ11 relative to the corresponding numbers at the DP stage, so that only 2% of Vβ5+ and 9% of Vβ11+ cells were CD4SP and these were mostly immature CCR7− SP1 cells (Fig. 3, A [second and third columns] and B). In contrast, in both the TCRβ+ CD4SP population as a whole (equivalent to Fig. 2 C and not depicted) and in the subset expressing Vβ8 that does not bind self-superantigen, CD4SP thymocytes were distributed normally across the SP1–SP3 stages.
The dramatic drop in Vβ5+ and Vβ11+ cell numbers at the CD4SP stage was almost completely abolished in Bim−/− mice, whereas there was no effect of Bim deficiency on numbers at the DP stage (Fig. 3 B). The rescued Vβ5+ and Vβ11+ CD4SP cells nevertheless showed little evidence of maturing past the CCR7− SP1 stage (Fig. 3 A, third column), and many fell outside the CD4SP or DP gates exhibiting a CD4lo CD8lo “dulled” phenotype (Fig. 3 A, second column). This dulled phenotype matches that of DP thymocytes undergoing negative selection in vitro (Swat et al., 1991; Vasquez et al., 1992; Page et al., 1993) or in vivo (McCaughtry et al., 2008; Pobezinsky et al., 2012). 75% or more of the Vβ5+ or Vβ11+ CD4lo CD8lo and SP1 thymocytes in Bim−/− mice expressed high levels of Helios, and many of these also expressed the inhibitory receptor PD-1 (Fig. 3 A, fifth and sixth columns), which is induced during negative selection of MHC I–restricted DP thymocytes (Baldwin and Hogquist, 2007; McCaughtry et al., 2008). Many non-superantigen–reactive Vβ8+ thymocytes in Bim−/− mice were also Helios+, accounting for 13% of DP cells, 85% of CD4lo CD8lo, and 50% of SP1 thymocytes (Fig. 3 A, bottom row). These frequencies and the fraction of CD4lo CD8lo (DP dull) cells among Vβ8+ Bim−/− thymocytes were equivalent to those in polyclonal TCR-expressing cells shown in Fig. 2 C, which was similar to when all TCRβ+ thymocytes were examined (not depicted). Despite the prominent increase in superantigen-reactive thymocytes (Fig. 3 B), Bim deficiency resulted in only a fourfold increase in Vβ11+ Foxp3− CD4+ splenocytes and no significant increase in Vβ5+ Foxp3− CD4+ splenocytes (not depicted). These findings demonstrate that, as well as Bim-mediated apoptosis, Helios and PD-1 induction and coreceptor dulling characterize the response of strongly self-reactive CCR7− thymocytes to self-antigen.
The results above implied that more than half of the nascent T cells in the natural repertoire are strongly self-reactive and normally deleted by Bim, but steady-state analysis alone may not allow for the most accurate estimate if strongly and weakly self-reactive cells dwell for different periods of time in the thymus of Bim−/− mice. To address this, we first used in vivo BrdU labeling and flow cytometry to identify TCR-signaled cells within the pool of recently proliferating (BrdU+) nascent (CD24+ CCR7−) thymocytes. 3 d after BrdU injection, 96–97% of CD69− TCRβ− and CD69− TCRβ+ cells were DP cells (Fig. 4 A), whereas 32% of the CD69+ TCRβ+ population fell outside the DP gate, including 16% in the CD4SP gate (Fig. 4 A, top). By comparison, CD69 was not induced in BrdU+ thymocytes from Zap70mrt/mrt mice, which express a catalytically mutant form of Zap-70 at 25% of the wild-type quantity and fail to be positively selected past the DP stage (Fig. 4 B, top row; Siggs et al., 2007), confirming that CD69 is a specific marker of thymocytes that have received a pMHC-TCR signal (Hare et al., 1999). Of the newly formed CD4SP thymocytes, 85% were CD69+ TCRβ+ (Fig. 4 A, bottom), indicating that CD69 up-regulation is also a sensitive marker of thymocytes that had progressed to the CD4SP stage. Based on these data, we used CD69 up-regulation as a marker of TCR-signaled thymocytes.
In Bim−/− mice, CD69+ cells were more than twice as frequent among nascent BrdU+ cells 24 h after BrdU injection (mean ± SEM: 8.2 ± 0.3%, n = 4 Bim−/− mice; 3.6 ± 0.2%, n = 5 Bim+/+; Fig. 4 B), suggesting that in wild-type animals, 56% of the TCR-signaled nascent thymocytes had been deleted. The newly selected CD69+ population included a Helios+ subset that was markedly expanded in Bim−/− mice (Fig. 4, B and C), indicating many of the cells rescued from death by Bim deficiency were Helios+. Nearly all of these Helios+ cells fell within the CD69+ gate 24 h after BrdU, consistent with evidence that CD69 is rapidly induced during negative selection as well as during positive selection (McCaughtry et al., 2008). When the analysis was performed 3 d after BrdU injection, the difference in nascent CD69+ cell frequencies between Bim−/− mice (36.3 ± 2.7%, n = 5) and Bim+/+ mice (17.3 ± 0.6%, n = 7; Fig. 4 C) suggested that 52% of CD69+ TCR-signaled thymocytes had been deleted in the Bim+/+ animals. Many, though not all, of the Helios+ CD69+ cells were also PD-1+, and they expressed less CD4 and CD8 than their Helios− CD69+ counterparts (Fig. 4, B and C). In contrast with this response, ∼15% of Helios− CD69+ cells had differentiated into CD4SP cells by 3 d after BrdU injection, and this was unaffected by Bim deficiency (Fig. 4 C). These findings extend the evidence above that Helios discriminates thymocytes destined for negative versus positive selection, including in the polyclonal TCR repertoire, and suggest that the fraction of TCR-signaled thymocytes that die by Bim-dependent negative selection at the CCR7− stage is ∼55%. This is close to the 65.6% Helios+ cells observed among steady-state SP1 Foxp3− CD4SP cells in Bim−/− mice (Fig. 2 C).
A distinct hollow activation response accompanies Helios and Bim induction in strongly self-reactive thymocytes at the CCR7+ stage
As the results above showed that CCR7− thymocytes that were induced to express Helios also induced PD-1 and failed to progress to the CCR7+ stage, we asked whether a comparable response occurred within the Helios+ CCR7+ SP2 thymocytes that bound strongly to self-pMHC in the thymus medulla. We first analyzed SP1 and SP2 subsets in TCR3A9 × insHEL DTg mice, where most CD4SP cells bind strongly to self-pMHC produced selectively in Aire+ medullary thymic epithelial cells (Liston et al., 2003, 2004a). 10-fold fewer SP2 cells were present in Bim+/+ DTg mice compared with STg mice (Fig. 5 B), and those present appeared transitional between the SP1 and SP2 stages as they had yet to fully up-regulate CCR7 (Fig. 5, A and B). Induction of Helios and Bim occurred predominantly in these SP2 cells in Bim+/+ DTg mice (Fig. 5 A). In Bim−/− DTg mice, CCR7 up-regulation was restored (Fig. 5 A) so that Bim−/− STg and Bim−/− DTg mice had a similar number of SP2 cells, ∼10-fold greater than Bim+/+ DTg mice (Fig. 5 B). Unlike the response of CCR7− SP1 thymocytes, the CCR7+ self-reactive cells did not express PD-1 (Fig. 5 C) and were not developmentally arrested as CD4lo 8lo cells (Fig. 5 D) when rescued from deletion in Bim−/− DTg mice. Instead, many reached the periphery, resulting in a 38-fold increase in TCR3A9+ Foxp3− CD4+ splenocytes in Bim−/− DTg mice compared with Bim+/+ DTg mice (Fig. 5 E).
Although the inhibitory receptor PD-1 was not induced in strongly self-reactive CD4SP thymocytes at the CCR7+ SP2 stage of maturation, many other mature T cell activation genes were strongly induced. We analyzed DTg versus STg thymocyte microarray data from sorted CD69+ CD4SP cells representing a mixture of SP1 and SP2 cells (Liston et al., 2007) for the expression of 49 mRNAs known to be induced in primary mature T cells activated for 3 h (Fig. 6 A; Shaffer et al., 2001). Of these 49 mRNAs, 19 were preferentially induced in CD4+ CD8lo CD69+ TCR3A9+ thymocytes from DTg mice compared with their counterparts that bound self-pMHC weakly in STg mice. These included the mRNAs encoding CD25, OX40, GITR, and c-Rel but also other mediators of the mature T cell growth-activating response: the integrin ligand ICAM-1, the antiapoptosis proteins Bcl-2 and A1/Bfl1 (Bcl2a1a), the transcription factor partners NF-κB2 and Rel-B, the chemokines MIP-1β (Ccl4) and lymphotactin (Xcl1), and the signaling molecule Gadd45β. Another 14 mature T cell activation genes were induced to similar levels in CD4+ CD8lo CD69+ TCR3A9+ thymocytes that were strongly or weakly self-reactive from DTg and STg mice, respectively. Only 16 of the 49 mature T cell activation genes examined were not induced in CD4+ CD8lo CD69+ TCR3A9+ thymocytes from DTg mice, although it was striking that those missing included two critical factors for T cell proliferation, Il2 and Myc (Fig. 6 A). The absence of Il2 and Myc (Trumpp et al., 2001; Dose et al., 2009) implied that the T cell activation response induced by strong binding to self-pMHC in CCR7+ CD4SP thymocytes was hollow in that it lacked the capacity for autocrine cell growth and proliferation. Consistent with this, the proliferation-associated protein Ki-67 was highly expressed in double-negative and DP cells but down-regulated in CCR7+ CD4SP cells in DTg mice (Fig. 6 B), and the mRNA levels of eight genes comprising a mitosis subsignature (Shaffer et al., 2001) were significantly down-regulated relative to preselection thymocytes (Fig. 6 A).
To confirm that strong binding to self-pMHC induced a distinct response in CCR7− SP1 versus CCR7+ SP2 thymocytes, we compared expression of a subset of these T cell activation proteins on Helios− and Helios+ SP1 and SP2 cells in the normal repertoire by flow cytometry (Fig. 6 C). PD-1 was strongly induced on Helios+ SP1 cells but was hardly induced on Helios+ SP2 cells. In contrast, CD25, OX40, and GITR were strongly induced on Helios+ SP2 cells, consistent with the DTg microarray data, but only weakly or not at all on Helios+ SP1 cells. Other TCR-induced proteins such as CD5, CD69, and Bim were induced on Helios+ cells at both stages, although to slightly higher levels on SP2 cells. These data reveal that, in the natural repertoire, the phenotypes of Helios+ Foxp3− CD4SP cells vary with maturation stage, with PD-1 induction a feature at the SP1 stage, whereas the induction of GITR, OX40, and CD25 accompanies Helios and Bim induction at the SP2 stage.
Card11 and c-Rel mutations perturb Helios+ Foxp3− CD4SP thymocytes selectively at the SP2 and SP3 stages
Many of the T cell activation genes induced in strongly self-reactive CCR7+ (SP2 and SP3) thymocytes are transcriptionally induced by NF-κB in mature T cells (Pahl, 1999; Shaffer et al., 2001), but these genes were not induced in Helios+ cells at the CCR7− SP1 stage. This implied that NF-κB might have a distinct function in the activation response of strongly self-reactive CCR7+ CD4SP thymocytes. To test this, we used mice bearing a partial loss of function mutation in TCR signaling to NF-κB, Card11unmodulated (Card11unm), which reduces but does not abolish CD25 induction by anti-CD3/CD28 stimulation in mature CD4+ T cells (Jun et al., 2003). In parallel, we analyzed mice lacking one member of the NF-κB transcription factor family, c-Rel (Köntgen et al., 1995). The formation of SP1 Helios+ Foxp3− CD4SP thymocytes that strongly express PD-1 was intact in Card11unm/unm and Rel−/− mice (Fig. 7, A and C). In contrast, although Helios was still induced in Card11unm/unm and Rel−/− CCR7+ thymocytes, the frequency and number of Helios+ SP2 and SP3 Foxp3− thymocytes were significantly decreased by these mutations (Fig. 7, B and C). The number of Helios− CD4SP thymocytes at each maturation stage SP1-SP3 was not significantly affected by homozygous mutations in Card11 or Rel (not depicted). These data demonstrate that Card11 and c-Rel are dispensable for the formation of CCR7− SP1 Helios+ Foxp3− CD4SP thymocytes, but both are required for normal accumulation of Helios+ strongly self-reactive cells at the CCR7+ SP2 and SP3 stages in the normal TCR repertoire.
To clarify the requirement for Card11 in accumulation of Helios+ CCR7+ Foxp3− CD4SP cells, we tested the consequences of inhibiting clonal deletion in Card11unm/unm Tg(Vav-BCL2) mice. These had much larger numbers of Helios+ CCR7+ Foxp3− CD4SP cells than Card11unm/unm non-Tg controls, confirming that Helios was still strongly induced in self-reactive cells with defective NF-κB signaling (Fig. 7 D). However, their frequency was half that observed in Card11+/+ Tg(Vav-BCL2) littermates. These data support the conclusion that Card11 signaling to NF-κB is required for Helios+ medullary thymocytes to survive strong TCR signaling.
To extend the genetic evidence of a distinct response induced by strong binding to self-pMHC in CCR7− versus CCR7+ CD4+ thymocytes, mice were analyzed 3 d after BrdU, when the numbers of BrdU+ SP1 and SP2 cells peak (Fig. 2 B) and before many have emigrated (McCaughtry et al., 2007). 3 d after BrdU injection, Card11unm/unm mice had a small but significant increase in the frequency of Helios+ TCRβ+ cells among nascent CCR7− thymocytes, demonstrating that Card11 is required for efficient apoptosis of these cells, although their frequency was still 10 times lower in Card11unm/unm mice than in Bim−/− mice (Fig. 7 E). The characteristic dulling of CD4 and CD8 expression by CCR7− Helios+ TCRβ+ thymocytes was intact in Card11unm/unm mice (Fig. 7 E). Although mutations in Card11 and Bim both increased the frequency of nascent Helios+ thymocytes at the CCR7− stage, these mutations had opposite effects on the frequency of Helios+ cells among nascent CCR7+ CD24+ CD4SP thymocytes, being decreased threefold in Card11unm/unm mice and increased sixfold in Bim−/− mice (Fig. 7 F). In Bim−/− animals, Helios+ cells accounted for 20% of BrdU-labeled SP2 cells (Fig. 7 F).
A potential explanation for the reduced Helios+ cell frequencies at the SP2 and SP3 stages in Card11unm/unm and Rel−/− mice is that Card11 activation of NF-κB delivers a survival signal that opposes the clonal deletion of CCR7+ CD4SP thymocytes that receive a strong TCR signal. For example, signaling to NF-κB induces T cell survival genes such as Bcl2, Bcla1a (A1), and Bcl2l1 (Bcl-X), and these were preferentially induced in strongly self-reactive SP2 cells in DTg mice (Fig. 6 A). To determine whether there was concurrent induction of Bim and NF-κB response proteins in strongly self-reactive SP2 cells, we analyzed Foxp3− CCR7+ CD4SP thymocytes undergoing clonal deletion at the SP2 stage in Card11unm/unm DTg mice. Confirming the microarray data, we found that high levels of CD25, OX40, GITR, and c-Rel were induced selectively in the Bimhi subset of Foxp3− TCR3A9+ CD4SP thymocytes from DTg but not STg mice (Fig. 8 A). The induction of each of these proteins was abolished in Card11unm homozygotes and markedly decreased in heterozygotes (Fig. 8 A), establishing that these are exquisitely dependent on Card11 signaling. Homozygosity for Card11unm diminished the accumulation of Helios+ CCR7+ SP2 cells (Fig. 8, B and C), so that these were 3.6-fold lower in Card11unm/unm DTg mice than in Card11+/+ DTg mice (Fig. 8 C). In contrast, Card11 was not required for accumulation of CCR7+ cells in STg mice where they bound self-pMHC weakly. Collectively, these results indicate that strongly self-reactive CD4 cells are deleted earlier in the SP1–SP2 transition when they are unable to make a concurrent NF-κB survival response.
Thymic selection of CD4+ T cells also encompasses differentiation of Foxp3+ T reg cells, which are crucial for self-tolerance (Rudensky, 2011; Sakaguchi, 2011). Although T reg cells have heightened affinity for self–pMHC II ligands (Jordan et al., 2001; Hsieh et al., 2004), it has been debated whether this reflects active induction of T reg cell differentiation by strong recognition of self-pMHC (Cozzo Picca et al., 2011) or heightened resistance of T reg cell precursors to clonal deletion (van Santen et al., 2004). On a Card11+/+ background, DTg mice had almost fivefold more TCR3A9hi Foxp3+ CD4SP thymocytes than STg mice (Fig. 8 D), arguing that strong binding to self-pMHC actively induced the formation of organ-specific T reg cells in the thymus of DTg mice. The number of TCR3A9hi Foxp3+ CD4SP thymocytes was decreased by 2.6-fold in Card11unm/+ heterozygous DTg mice and by 250-fold in Card11unm/unm DTg mice (Fig. 8 D). These data show that many Foxp3− CCR7+ CD4SP cells that are Bimhi and destined to die nevertheless mount a Card11-dependent program that has the dual function of curbing clonal deletion and enabling T reg cell differentiation.
The findings above address several key questions about thymic negative selection. The results identify the Ikaros family member Helios as a unique molecular marker that qualitatively differentiates the thymocyte response to weak or strong self-pMHC stimulation in vivo. Flow cytometric measurement of Helios provides a way to enumerate and analyze the fraction of nascent thymocytes that are strongly self-reactive, both in normal mouse repertoires and in animals that are genetically predisposed to autoimmune disease. We show that Helios and Bim are coinduced in strongly self-reactive thymocytes at two major waves of Bim-mediated deletion of CD4+ cells: one in CCR7-negative thymocytes that is induced in 55% of newly formed cells and a second in 20% of cells that reach the CCR7+ CD4SP stage. This in turn enabled analysis of the other proteins and mRNAs that are induced in strongly self-reactive thymocytes at each of these stages, addressing the question of how the thymocyte response to strong TCR stimulation varies with maturation stage.
Helios was shown to be induced selectively in strongly self-reactive thymocytes and concurrently with Bim in several settings: in the subset of cells in the normal repertoire that are normally eliminated by Bim, in Vβ5+ and Vβ11+ thymocytes that bind strongly to self-superantigens, and in TCR Tg thymocytes that bind strongly to self-pMHC expressed in AIRE+ medullary thymic epithelium. In contrast, Helios was down-regulated during positive selection of CD4+ thymocytes that bind self-pMHC weakly. This splitting of the response distinguishes Helios from other TCR-induced proteins like Bim, Nur77, CD69, and GITR, which are induced during both positive and negative selection and differentiate the weak and strong response only by their magnitude of induction. T cell development was reported to be normal in the absence of Helios (Cai et al., 2009; Thornton et al., 2010), although it is possible that when the Helios protein is absent this allows other Ikaros family members such as Ikaros itself to fill the niche and substitute, as was found to be the case from comparison of mice with null alleles and non-DNA–binding alleles of Ikaros (Papathanasiou et al., 2003). Acute retroviral-mediated expression of Helios decreased the proliferation of activated splenic CD4+ T cells (Zhang et al., 2007), raising the possibility that Helios has an antiproliferative function. Consistent with this possibility, non-DNA–binding dominant-negative short isoforms of Helios have been associated with T cell leukemia in humans (Sun et al., 2002; Fujii et al., 2003) and induce T cell leukemia in mice (Zhang et al., 2007). Continuous activating receptor stimulation has recently been shown to induce Helios in NK cells and dampen their responsiveness (Narni-Mancinelli et al., 2012).
Helios expression provides a direct way to measure the percentage of thymocytes that bear strongly self-reactive TCRs and are susceptible to deletion. Our analysis of CCR7− thymocytes, including CD69+ TCRβ+ cells selected from BrdU-labeled precursors in the previous 24 or 72 h, suggests that 55% of all TCR-signaled thymocytes are deleted by Bim at the CCR7− stage. At the subsequent CCR7+ stage, 20% of nascent BrdU+ SP2 cells were Helios+ in Bim−/− mice versus only 3% in wild-type controls, indicating that additional CD4SP thymocytes are deleted at this stage. Although a few Helios+ SP2 cells may derive from Helios+ SP1 precursors, it seems unlikely that many Helios+ SP1 cells up-regulate CCR7 to become SP2 cells, given that the superantigen-reactive Vβ5+ and Vβ11+ populations in Bim−/− mice had lower Helios+ frequencies at the SP2 stage than at the SP1 stage (Fig. 3). After accounting for the 55% of TCR-signaled CCR7− cells that arrest their maturation and fail to up-regulate CCR7 in Bim−/− mice, we calculate that the 20% of nascent SP2 cells that were Helios+ in Bim−/− mice (Fig. 7 F) represents 9% of the pool that was initially selected. It is conceivable that these cells bound weakly to self-pMHC in the cortex but then bound strongly to a self-pMHC ligand confined to the medulla, as occurs in the DTg mice. It was recently suggested that the absence of both Bim and Puma (gene name Bbc3) results in more autoreactive thymocytes escaping negative selection than when only Bim is absent (Gray et al., 2012). As the percentage and number of CD69+ thymocytes at the DP stage, a population shown here to include cells TCR-signaled for positive or negative selection, were similar in Bim-deficient and Bim/Puma double-deficient mice, Puma would not seem to delete additional autoreactive cells in the early wave of negative selection. Bim/Puma double-deficient mice had a striking accumulation of CD24lo CD4SP cells (Gray et al., 2012), which indicates that Puma mediates cell death of Bim-deficient CD4+ cells that are more mature than the SP2 stage. Our estimate that a total of ∼64% of TCR-signaled CD4+ thymocytes undergo Bim-dependent negative selection either at the CCR7− (55%) or CCR7+ (9%) stage agrees with earlier estimates based on the frequency of mature CD4SP thymocytes (Tourne et al., 1997; van Meerwijk et al., 1997) or the proportion of CD4+ T cells with self-reactive TCRs (Ignatowicz et al., 1996) in mice in which deletion was defective.
The results here provide molecular insight into the acquisition of self-tolerance by analyzing the thymocyte response to strong TCR stimulation and how it differs from the mature T cell response, independent of the normally rapid Bim-mediated deletion. At the CCR7− stage, strongly self-reactive T cells make a unique response, even compared with CCR7+ thymocytes, which is dominated by developmental arrest at the CCR7− stage, PD-1 induction, CD4 and CD8 down-regulation, and lack of any induction of CD25 or OX40. The induction of PD-1 and dulling of CD4 and CD8 have been observed in HYcd4 TCR Tg thymocytes encountering MHC I–presented antigen on cortical epithelium (McCaughtry et al., 2008), in thymocytes lacking CD28 co-stimulation (Pobezinsky et al., 2012), and in TCR-activated DP thymocytes undergoing apoptosis in vitro (Swat et al., 1991; Vasquez et al., 1992; Page et al., 1993). The fact that Helios+ CCR7− cells remain developmentally arrested at the CCR7− stage in the thymus when Bim-mediated deletion is abolished and that they are not found in the periphery as shown for Vβ5+ and Vβ11+ CD4 cells is consistent with a comparable developmental arrest shown for MHC I–restricted T cells binding strongly to self-pMHC in the thymus cortex (Hu et al., 2009; Kovalovsky et al., 2010; Suen and Baldwin, 2012). This unique response thus appears to be a general, stage-specific response program that may either be intrinsic to very immature CCR7− thymocytes or reflect the nature of the APCs they encounter at that stage. Thymocyte negative selection at this stage has interesting parallels with negative selection of immature B cells that are strongly self-reactive in the bone marrow, which also become developmentally arrested at a stage where they cannot home to lymph nodes when their apoptotic response is blocked (Hartley et al., 1993).
At the CCR7+ stage, however, the thymocyte response to strong pMHC stimulation was found to be surprisingly similar to mature T cell activation, with induction of a broad program of T cell survival and growth genes through Card11 signaling to NF-κB. This hollow activation response opposes Bim-mediated deletion, and by activating c-Rel and CD25, it sets the stage for Foxp3 expression and diversion of a minority of cells to the Foxp3+ T reg cell lineage. The observation that this Card11-mediated hollow activation response is required to oppose Bim-mediated deletion explains why a constitutively active STAT5 transgene failed to rescue thymic T reg cell formation in Card11-deficient mice (Molinero et al., 2009). It is significant that the CCR7+ thymocyte response to strong self-antigen lacks key growth mediators like IL-2 and Myc that are central to mature T cell activation and that there was no evidence that the thymocytes were induced into cell cycle. Because IL-2 and Myc are both direct NF-κB targets in mature T cells, their failure to be activated in CCR7+ SP2 thymocytes may reflect stage-specific programming, for example at the level of chromatin accessibility, or the effect of Helios itself as Ikaros suppresses Il2 transcription in anergic T cells (Bandyopadhyay et al., 2007; Thomas et al., 2007). Collectively, these results reveal that the acquisition of self-tolerance in CD4SP cells in the thymic medulla walks a fine line between T cell activation and Bim-mediated deletion and would appear to provide much scope for inherited variability in susceptibility to autoimmune disease.
MATERIALS AND METHODS
Mice were bred and housed in the Australian Phenomics Facility. Non-Tg mice were either B10.BR (H-2k) or C57BL/6 (B6; H-2b), and all mice were 6–14 wk old when used. 3A9 TCR Tg (Ho et al., 1994), insHEL Tg (originally called ILK-3; Akkaraju et al., 1997), Bim-deficient (Bcl2l11−/−, here referred to as Bim−/−; Bouillet et al., 1999), Vav-BCL2 Tg (Ogilvy et al., 1999), Zap70mrt (Siggs et al., 2007), Card11unm (Jun et al., 2003), and Rel−/− (Köntgen et al., 1995) mice have been described previously. 3A9 TCR, insHEL, and Vav-BCL2 Tg mice were hemizygous for the relevant transgenes and genotyped by PCR. Mice bearing the Bim− allele were genotyped by PCR. A mutagenically separated PCR assay (Rust et al., 1993) was used to genotype mice for the Card11unm allele. Except for Card11unm/unm mice on the B6 background in Fig. 7, all Bim−/− and Card11unm/unm mice were offspring of at least the third generation backcross to B10.BR, and at least half of the Card11+/+ or Bim+/+ animals represented in each figure panel were their littermates, whereas the remainder were wild-type controls from our facility. Mice from a homozygous Rel−/− strain were used. To make bone marrow chimeras, CD45.1+ insHELneg/pos mice were irradiated with x rays (two doses of 4.5 Gy given 4 h apart) and then injected intravenously with 3 × 106 bone marrow cells from a Bim−/− STg donor alone or mixed at a ratio of 1:19 with CD45.1+ non-Tg marrow. They were allowed to reconstitute for 4–5 wk before analysis. The Animal Experimentation Ethics Committee of the Australian National University approved all procedures.
Single cell thymocyte suspensions were incubated for 60 min at 37°C in prewarmed FACS buffer (PBS containing 2% vol/vol heat-inactivated bovine serum and 0.01% m/v sodium azide) or culture supernatant from the 1G12 hybridoma (specific for TCR3A9; American Type Culture Collection) containing anti-CCR7. Cells were then pelleted by centrifugation and incubated for 30 min at 4°C in FACS buffer containing assortments of fluorochrome- or biotin-conjugated monoclonal antibodies against cell surface proteins or mouse IgG1 (to detect 1G12). After washing in FACS buffer, cells were fixed, permeabilized, and stained for intracellular proteins using the eBioscience Foxp3 staining buffers or the FITC BrdU Flow kit (BD) with anti-Helios (clone 22F6; BioLegend; Thornton et al., 2010) included in the same incubation step as anti–BrdU-FITC. Finally, cells were washed in FACS buffer and then incubated for 15 min at 4°C in FACS buffer containing Qdot 605–streptavidin conjugate (Invitrogen) to detect biotinylated antibodies. Except for anti-Bim (clone 3C5; a gift from A. Strasser, Walter and Eliza Hall Institute of Medical Research, Melbourne, Victoria, Australia), which was conjugated to Alexa Fluor 647 using a Protein Labeling kit (Molecular Probes), antibodies were purchased from BD, eBioscience, BioLegend, or Santa Cruz Biotechnology, Inc. (anti–c-Rel–A488). Data were acquired with an LSR II flow cytometer (BD) and analyzed using FlowJo software (Tree Star).
A microarray dataset (Liston et al., 2007) accessible through GEO Series accession no. GSE3997 was mined for T cell activation genes (Shaffer et al., 2001). Thymocyte populations analyzed by microarray were preselection (PreS; CD4+ CD8+ CD69− TCR3A9− from STg and DTg mice), positive selection (S+; CD4+ CD8lo CD69+ TCR3A9+ from STg mice), and negative selection (S−; CD4+ CD8lo CD69+ TCR3A9+ from DTg mice). Microarray analyses were performed before the establishment of the SP1–SP3 categorization system; analysis indicates the percentages of SP1 and SP2 cells were ∼40% and 55%, respectively, in the S+ population, and 75% and 22%, respectively, in the S− population. The genes were categorized using p-values of one-way ANOVA followed by Tukey’s multiple comparison tests. The three categories were (1) genes up-regulated preferentially in negative selection (PreS vs. S− <0.05 and either S− vs. S+ <0.05 or PreS vs. S+ >0.05), (2) genes up-regulated comparably during positive and negative selection (PreS vs. S+ <0.05 and S+ vs. S− >0.05; Irf4 was included in category 2 even though S+ vs. S− was <0.05 on the grounds that S+ exceeded S− by only 1.16-fold), and (3) genes not induced (PreS vs. S+ and PreS vs. S− >0.05) or down-regulated during selection (PreS significantly exceeded both S+ and S−).
Prism version 5.00 for Windows (GraphPad Software) was used to conduct unpaired or paired Student’s t tests as described in the figure legends and for one-way ANOVA followed by Tukey’s multiple comparison tests on microarray data.
We thank Debbie Howard for technical assistance, Andreas Strasser for the gift of the anti-Bim antibody, and Steve Gerondakis for making Rel−/− mice available.
This work was supported by a National Health and Medical Research Ccouncil Program Grant and Australia Fellowship and by the Juvenile Diabetes Research Foundation and involved mutant strains funded by the National Institute of Allergy and Infectious Diseases, National Institutes of Health and the Wellcome Trust.
The authors have no competing financial interests.