|Home | About | Journals | Submit | Contact Us | Français|
Clonal deletion of autoreactive thymocytes is important for self-tolerance, but the intra-thymic signals that induce clonal deletion have not been clearly identified. We now report that clonal deletion during negative selection requires CD28 costimulation of autoreactive thymocytes at the CD4+CD8lo intermediate stage of differentiation. Autoreactive thymocytes were prevented from undergoing clonal deletion by either absent CD28 costimulation or transgenic over-expression of the anti-apoptotic factors Bcl-2 or Mcl-1, with surviving thymocytes differentiating into anergic T cell receptor αβ+ double negative thymocytes that preferentially migrated to the intestine where they re-expressed CD8α and were sequestered as CD8αα intraepithelial lymphocytes (IELs). This study identifies CD28 costimulation as the intrathymic signal required for clonal deletion and identifies CD8αα IELs as the developmental fate of autoreactive thymocytes that survive negative selection.
Immunocompetent αβ T cells must be reactive to foreign pathogens but tolerant to self ligands. These critical features of T cell immunity are imposed by selection events in the thymus that determine the developmental fate of each individual T cell depending on the specificity of its T cell antigen receptor (TCR). Differentiation in the thymus proceeds in an ordered sequence characterized by CD4 and CD8 coreceptor expression in which the earliest cells are CD4−CD8− (double negative, DN) thymocytes that differentiate into CD4+CD8+ (double positive, DP) thymocytes that then terminally differentiate into CD4+ or CD8+ (single positive, SP) T cells1, 2. Thymocytes at the DP stage of differentiation are the first cells to express endogenous αβTCR complexes and are the cells subjected to TCR-specific thymic selection. DP thymocytes are intrinsically short-lived cells whose continued survival requires TCR signaling by self-ligands in the thymic cortex. TCR signaling rescues DP thymocytes from ‘death-by-neglect’ and induces either positive or negative selection 1, 2. Positive selection is induced by low affinity ligands and results in the differentiation of TCR-signaled DP thymocytes into conventional SP4 or SP8 T cells possessing helper or cytotoxic function respectively, whereas negative selection is induced by high affinity ligands that prevent TCR-signaled DP thymoctyes from continuing their differentiation into conventional SP T cells3, 4. Thus, the thymus imposes central tolerance by generating conventional SP4 and SP8 mature T cells that express TCRs lacking significant autoreactive potential.
The most definitive way of preventing autoreactive TCRs from appearing on mature SP T cells is the clonal deletion of DP thymocytes bearing autoreactive TCRs during negative selection in the thymus5. However, strong TCR signaling of DP thymocytes does not necessarily result in thymocyte death. Indeed, a few DP thymocytes are strongly signaled by agonist ligands to differentiate into specialized SP4 T cell subpopulations possessing regulatory or NK functions 6, 7, with such specialized differentiation referred to as ‘agonist selection’8. In a similar vein, developing DP thymocytes do not undergo clonal deletion when strongly signaled by agonist ligands in the thymic cortex 9. Consequently, strong TCR signaling of DP thymocytes during negative selection appears to be insufficient by itself to induce clonal deletion. However, it is not known what, if any, additional in vivo signals are needed during negative selection to induce thymocytes to undergo clonal deletion.
A potentially useful insight may have been provided by longstanding in vitro studies which demonstrated that CD28 costimulation was required to induce strongly TCR-signaled thymocytes to die10–13. Indeed in vitro CD28 costimulation blocked TCR up-regulation of the anti-apoptotic protein Bcl-214, and transgenic Bcl-2 (Bcl-2 TG) over-expression rescued costimulated thymocytes from TCR-signaled death in vitro 12, 14. Although only observed in vitro, a requirement for CD28 costimulation in TCR-signaled thymocyte death was potentially consistent with observations that clonal deletion was mediated by thymic dendritic cells and medullary thymic epithelial cells which differed from non-deleting cortical thymic epithelial cells in high expression of the CD28 costimulatory ligands CD80 and CD86 9, 15–17. However, these in vitro results were directly contradicted by multiple in vivo studies that examined a role for CD28 costimulation in clonal deletion. These studies showed that autoreactive thymocytes were prevented from differentiating into either SP4 or SP8 SP T cells and that central tolerance was achieved regardless of the presence or absence of in vivo CD28 costimulation18–22.
Nonetheless, we undertook the present study to determine if in vivo CD28 costimulation during negative selection was important for either clonal deletion or central tolerance. Unlike previous in vivo studies, we distinguished between negative selection and clonal deletion by considering the possibility that strongly signaled DP thymocytes bearing autoreactive TCRs might survive negative selection even though they do not differentiate into conventional SP4 or SP8 T cells. In fact we now report that in vivo CD28 costimulation during negative selection is required for autoreactive thymocytes to undergo clonal deletion, but that neither CD28 costimulation nor clonal deletion is critical for central tolerance. We report that tolerance can be induced during negative selection even in the absence of CD28-mediated clonal deletion because strong TCR signaling diverts DP thymocytes from differentiating into conventional SP T cells bearing autoreactive TCRs and directs their differentiation into TCRαβ+DN thymocytes, a process we refer to as ‘developmental diversion.’ Although developmentally diverted thymocytes express autoreactive TCRs, they are coreceptor-negative and functionally anergic which diminishes their autoreactive potential. Moreover, when developmentally diverted TCRαβ+DN thymocytes leave the thymus, they preferentially migrate to the intestine where they are signaled to express CD8αα coreceptor homodimers and are sequestered in the gut as CD8αα intra-epithelial lymphocytes (IELs). Thus, this is the first study to identify CD28 as critical for inducing autoreactive thymocytes to undergo clonal deletion during in vivo negative selection and additionally identifies CD8αα IELs as the fate of autoreactive thymocytes that avoid clonal deletion during negative selection.
We began our assessment of intrathymic costimulation during negative selection by comparing thymocyte populations from wildtype and costimulation-deficient mice that lacked either the costimulatory receptor CD28 or its two costimulatory ligands CD80 and CD86 (also known as B7.1 and B7.2). We refer to mice genetically deficient in CD28 as Cd28−/− mice and to mice genetically deficient in both CD80 and CD86 as B7-deficient (B7-DKO) mice. In our analysis, we specifically looked for a thymocyte subpopulation that was present in costimulation-deficient (Cd28−/− and B7-DKO) mice but was absent in wildtype mice. Costimulation-deficient mice on two different genetic backgrounds (C57BL/6 and BALB/c) had only modestly changed thymocyte numbers and modestly increased frequencies of CD4−CD8− (DN) cells compared to wildtype mice (Fig. 1a top panels). However, costimulation-deficient mice were substantially enriched in a specific subset of DN thymocytes that were TCRαβ+ (Fig. 1a bottom panels). Unlike DN thymocytes from wildtype mice which contained few TCRαβ+ cells that were mostly NKT cells as determined by CD1d tetramer staining, DN thymocytes from costimulation-deficient mice contained high frequencies of TCRαβ+ cells that were not NKT cells (Fig. 1b,c).
If impaired clonal deletion in costimulation-deficient mice were the basis for increased TCRαβ+DN thymocytes, then impairing clonal deletion in costimulation-sufficient mice should also increase TCRαβ+DN thymocytes. To test this prediction, we attempted to impair clonal deletion in wildtype mice with transgenes encoding anti-apoptotic proteins Mcl-1 and Bcl-2 (Fig. 2a). Mcl-1 and Bcl-2 transgenes substantially increased TCRαβ+DN thymocytes in wildtype mice and these thymocytes were CD5hi (Fig. 2a,b), consistent with their having been strongly signaled in vivo. Thus, thymocyte expression of pro-survival transgenes in wildtype mice had the same effect as costimulation-deficiency in increasing TCRαβ+DN thymocytes, suggesting that the TCRαβ+DN thymocyte subset contained cells that would otherwise have been clonally deleted.
To directly test the possibility that the TCRαβ+DN subset was enriched in thymocytes bearing autoreactive TCRs, we examined mice expressing endogenous super-antigens (SAg). BALB/c mice encode proviral proteins Mtv-6, -8, and -9 which specifically engage Vβ3, Vβ5, Vβ11, and Vβ12 TCRs so that BALB/c thymocytes expressing these TCRs undergo clonal deletion23. Indicative of TCRβ-specific clonal deletion in wildtype BALB/c mice, thymocytes bearing the SAg-reactive TCRs Vβ3, Vβ5, and Vβ11 were substantially reduced in all TCRαβ+ thymocyte subsets (SP4, SP8, and DN) compared to pre-selection DP thymocytes, whereas thymocytes bearing unreactive Vβ8 TCRs were compensatorily increased (Fig. 3a, and Supplementary Fig. 1a). In contrast to wildtype BALB/c thymocytes, costimulation-deficient BALB/c thymocytes bearing Mtv-8, 9-reactive TCRs (Vβ5 and Vβ11) were uniquely increased in both frequency and number in TCRαβ+DN thymocytes, though not in SP4 or SP8 thymocytes (Fig. 3a). Thus, thymocytes bearing Mtv-8, 9-reactive Vβ5 and Vβ11 TCRs were not deleted in costimulation-deficient BALB/c mice, but instead appeared as TCRαβ+DN thymocytes.
We additionally analyzed Mtv-6 reactive TCRs but were initially confused by a discrepancy between thymocytes from Cd28−/− and B7-deficient mice (Supplementary Fig. 1). In Cd28−/− BALB/c mice, the frequency and number of thymocytes bearing Mtv-6 reactive Vβ3 TCRs were depleted in SP4 and SP8 thymocytes (Supplementary Fig. 1a), as we observed with other SAg-reactive TCRs (Fig. 3a), but they were not depleted in B7-deficient BALB/c mice (Supplementary Fig. 1a), as was previously reported24. Our attempts to understand the basis for this discrepancy led us to discover that the gene encoding Mtv-6 was absent in B7-deficient BALB/c mice (Supplementary Fig. 1b). This turned out to be due to the fact that the Mtv-6 gene is on the same chromosomal segment as genes encoding CD80 and CD86 which, in B7-deficient mice, was derived from embryonic stem cells of 129 origin that lacked the proviral Mtv-6 gene. Consequently, B7-deficient mice do not contain the Mtv-6 gene in their genome, explaining the presence of TCR-Vβ3 thymocytes in all B7-deficient BALB/c thymocyte subsets, including SP4 and SP8.
Having resolved this discrepancy, we then wondered if Mcl-1 and Bcl-2 transgenes would prevent deletion of SAg-reactive thymocytes in costimulation-sufficient wildtype mice. Both Mcl-1 and Bcl-2 transgenes did in fact prevent clonal deletion of thymocytes in wildtype mice, as the frequency and number of SAg-reactive TCRs were increased among TCRαβ+DN thymocytes, with the Bcl-2 transgene having a greater effect than the Mcl-1 transgene (Fig. 3b).
Based on these results, we conclude that TCR signaling by high affinity intrathymic ligands is sufficient to prevent autoreactive thymocytes from becoming SP thymocytes, but it is not sufficient to induce clonal deletion which additionally requires intrathymic costimulation. In addition we conclude that Mcl-1 and Bcl-2 pro-survival transgenes prevent clonal deletion, and that thymocytes rescued from clonal deletion appear in the thymus as TCRαβ+DN thymocytes.
All TCRαβ+ thymocytes originally derive from TCR−DN precursor thymocytes that were signaled by pre-TCR to differentiate into DP thymocytes, and it is in DP thymocytes that endogenously encoded TCRαβ surface complexes are first expressed. Consequently, because TCRαβ+DN thymocytes bear endogenously encoded TCRαβ surface complexes, it was likely that they were the progeny of DP thymocytes. However, to assess if TCRαβ+DN thymocytes were indeed derived from DP thymocytes, we examined the methylation status of their Cd8b gene locus because it remains methylated until it is permanently de-methylated when thymocytes first express CD8 to become DP25. In thymocytes from Bcl-2 transgenic mice that were enriched in TCRαβ+DN thymocytes, the Cd8b promoter was methylated in TCR−DN precursor thymocytes but de-methylated in DP thymocytes and in their post-selection SP4 and SP8 progeny (Fig. 4a and Supplementary Fig. 2). The Cd8b promoter was also de-methylated in TCRαβ+DN thymocytes (Fig. 4a and Supplementary Fig. 2), revealing that TCRαβ+DN thymocytes were the progeny of coreceptor-positive, i.e. DP, thymocytes.
Next, we assessed if the generation of TCRαβ+DN thymocytes, like that of post-selection SP4 and SP8 thymocytes, required TCR-mediated thymic selection signals. Because TCR signaling in DP thymocytes is strictly dependent on the protein tyrosine kinase ZAP7026, we examined the impact of ZAP70-deficiency on the appearance of TCRαβ+DN thymocytes (Fig. 4b left panels). Costimulation-deficient and ZAP70-deficient (B7-DKO Zap70−/−) mice were devoid of TCRαβ+DN thymocytes (Fig. 4b left panels), documenting that the generation of TCRαβ+DN thymocytes was strictly dependent on signals transduced by ZAP70. We further assessed if generation of TCRαβ+DN thymocytes required intrathymic MHC expression because TCR-signaled thymic selection is MHC-specific. Costimulation-deficient Cd28−/− mice that were additionally MHC-KO (Cd28−/−B2m−/−H2-Ab1−/−) were essentially devoid of TCRαβ+DN thymocytes, revealing that both MHC and ZAP70 expression was required for in vivo generation of TCRαβ+DN thymocytes (Fig. 4b right panels).
We conclude that TCRαβ+DN thymocytes are the progeny of DP thymocytes and that differentiation of DP into TCRαβ+DN thymocytes requires TCR-mediated, MHC- specific thymic selection signals. Since thymic selection signals normally induce DP thymocytes to differentiate into coreceptor-positive SP4 or SP8 thymocytes, we refer to the altered differentiation of strongly signaled DP into coreceptor-negative TCRαβ+DN thymocytes as ‘developmental diversion’.
During positive selection, TCR-signaled DP thymocytes upregulate surface CD4 and downregulate surface CD8 to phenotypically become CD69+CD4hiCD8lo intermediate (INT) thymocytes, and it is in INT thymocytes that CD4-CD8 lineage choice occurs 1. Consequently, we wondered if clonal deletion and developmental diversion also occurred in INT thymocytes. To assess this possibility, we examined SAg-reactive TCR-Vβs in CD69+CD4hiCD8lo INT thymocytes from both wildtype and costimulation-deficient BALB/c mice. SAg-reactive TCRs were present in INT thymocytes in both wildtype and costimulation-deficient BALB/c mice in frequencies that were essentially equal to those in pre-selection DP thymocytes (Fig. 5a). In fact, SAg-reactive TCRs were present in substantially higher frequencies among TCR-signaled INT thymocytes than among SP4 or SP8 post-selected thymocytes (Fig. 5a), indicating that strongly signaled DP thymocytes differentiated into INT thymocytes before undergoing clonal deletion or developmental diversion (Fig. 5a). That is, in wildtype BALB/c mice, SAg-reactive TCRs were present in INT thymocytes but were depleted in all post-selection populations (Fig. 5a left panels), indicating that autoreactive thymocytes did not survive beyond the INT stage of differentiation; and, in costimulation-deficient BALB/c mice, SAg-reactive TCRs were present in both INT and TCRαβ+DN thymocytes but were depleted in SP4 and SP8 post-selection thymocytes (Fig. 5a middle and right panels), indicating that autoreactive INT thymocytes were developmentally diverted into TCRαβ+DN cells. These results indicate that clonal deletion and developmental diversion are both signaled during negative selection at the INT stage of differentiation, but they do not formally exclude the possibility that some DP thymocytes might become DN cells directly.
For INT thymocytes to receive costimulation-dependent deletional signals, INT thymocytes must be in contact with B7 expressing cells. To assess if INT thymocytes were in fact in contact with B7 expressing cells in the thymus, we made use of the fact that CD28-B7 interactions specifically downregulate CD28 surface expression27. In fact, CD28 surface levels were up-regulated during differentiation of DP into INT and TCRαβ+DN thymocytes in B7-deficient mice, whereas CD28 surface levels were down-regulated in B7 wildtype mice (Fig. 5b and Supplementary Fig. 3), revealing that INT and TCRαβ+DN thymocytes in wildtype mice (which survived because of the Bcl-2 transgene) were both in contact with B7 expressing cells in the thymus and were both strongly costimulated in vivo.
We conclude that CD4+CD8lo INT thymocytes are localized in the thymus where they can contact B7 costimulatory ligands. Furthermore, the INT stage of differentiation is a point in thymocyte development during which clonal deletion and developmental diversion can both occur. Because the INT stage of differentiation is also the point in thymocyte development that CD4–CD8 lineage choice occurs, we suggest that it is in INT thymocytes that different TCR signals are translated into different lineage fates, depending on the intensity and duration of TCR plus coreceptor signals as well as the presence or absence of intrathymic costimulation (schematized in Supplementary Fig. 4).
Next we performed a phenotypic analysis of developmentally diverted TCRαβ+DN thymocytes which revealed that, during differentiation of DP into developmentally diverted TCRαβ+DN thymocytes, a variety of molecules were up-regulated (TCRβ, PD-1, CD5, CD69, CD122, Bcl-2 and α4β7) and other molecules were either unchanged or down-regulated (CD25, CD44, CD124, CD127, CD132, Bcl-XL) (Supplementary Fig. 5 a and 5b). The expression pattern of these surface molecules provided us with some potentially useful insights into the generation and fate of TCRαβ+DN thymocytes.
First, TCRβ, CD5 and CD69 upregulation suggested that TCRαβ+DN thymocytes had been TCR signaled, and their high level expression of PD-1 indicated that those in vivo TCR signals had been strong and persistent as would be expected of autoreactive TCRs (Supplementary Fig. 5a). Because surface PD-1 molecules dampen TCR signal transduction 28, we then assessed the in vitro reactivity of TCRαβ+DN thymocytes. Developmentally diverted TCRαβ+DN thymocytes were determined to be anergic cells because (1) their TCR could not induce calcium mobilization (Fig. 5c and Supplementary Fig. 6a), (2) they were unable to produce their own IL-2 (Supplementary Fig. 6b), and (3) they were unable to proliferate in response to anti-TCR-CD28 stimulation without addition of exogenous IL-2 (Fig. 5d). We then considered whether PD-1 expression was required for developmental diversion. However, germline deletion of the Pdcd1 gene which encodes PD-1 protein did not affect generation of developmentally diverted TCRαβ+DN thymocytes in B7-deficient mice (Supplementary Fig. 6c). In addition, little or no surface expression of CD127 (IL-7Rα) indicated that developmentally diverted TCRαβ+DN thymocytes were probably not dependent on IL-7 for their in vivo survival (Supplementary Fig. 5a), unlike mature SP thymocytes and T cells. Indeed, despite virtually absent CD127 expression, developmentally diverted TCRαβ+DN thymocytes contained high levels of Bcl-2 and were not apoptotic in vivo (as revealed by negative Annexin V staining and absent active caspase-3) (Supplementary Fig. 5b), possibly because they expressed CD122 (IL-2Rβ) and CD132 (common γ chain) which conferred the potential to respond in vivo to IL-2 and IL-15 (Supplementary Fig. 5a). Moreover, developmentally diverted TCRαβ+DN thymocytes expressed α4β7 integrin, indicating that they could potentially leave the thymus and migrate to the intestine (Supplementary Fig. 5a).
To determine their in vivo developmental potential, we adoptively transferred TCRαβ+DN thymocytes from B7-deficient donor mice into Rag2−/− host mice (Fig 6a). Five weeks after transfer, most developmentally diverted TCRαβ+DN thymocytes had homed to the small intestine and differentiated into TCRαβ+CD8αα IELs, whereas SP8 T cells remained primarily TCRαβ+CD8αβ T cells regardless of where they had homed (Fig. 6a). Because IL-15 is present in the gut and because TCRαβ+DN thymocytes are potentially responsive to IL-15 by virtue of their expression of CD122 and CD132, we wondered if IL-15 might contribute to the differentiation of TCRαβ+DN thymocytes into TCRαβ+CD8αα cells. To test this possibility, we signaled developmentally diverted TCRαβ+DN thymocytes in vitro with anti-TCRβ and IL-15 (Fig. 6b). Indeed, by culture day 4, most TCR+IL-15 signaled TCRαβ+DN thymocytes had differentiated into TCRαβ+CD8αα cells (Fig. 6b).
These results revealed that developmentally diverted TCRαβ+DN thymocytes possessed the potential to become TCRαβ+CD8αα IELs and resembled observations made with wildtype TCRαβ+DN thymocytes29, but we did not know if developmentally diverted TCRαβ+DN thymocytes actually differentiated into TCRαβ+CD8αα IELs in vivo. To assess this possibility, we took advantage of the fact that TCRαβ+CD8αα IELs were nearly absent from B6 mice that were β2m-deficient (Fig. 6c left panels). We crossed B2m−/− mice with costimulation-deficient or Bcl-2 transgenic mice, both of which contained substantial frequencies of developmentally diverted thymocytes (Fig. 1 and Fig. 2). TCRαβ+CD8αα IELs were present in both B7-DKOβ2m−/− mice and Bcl-2 TGβ2m−/− mice (Fig. 6c right panels), demonstrating that developmentally diverted T cells do in fact differentiate in vivo into TCRαβ+CD8αα IELs. To confirm this finding, we asked if TCRαβ+CD8αα IELs displayed the same TCR-Vβ repertoire as developmentally diverted TCRαβ+DN thymocytes. Indeed, in B7-deficient BALB/c mice, SAg-reactive TCR-Vβ’s were over-represented in essentially identical frequencies in TCRαβ+CD8αα IEL’s and TCRαβ+DN thymocytes (Fig. 7a). We conclude that developmentally diverted TCRαβ+DN thymocytes differentiate in vivo into TCRαβ+CD8αα IELs.
Having found that developmentally diverted TCRαβ+DN thymocytes can differentiate in vivo into TCRαβ+CD8αα IELs, we wondered if all TCRαβ+CD8αα IELs might be derived from cells that had undergone developmental diversion during negative selection in the thymus. If so, mice in which thymic clonal deletion was absent would contain more TCRαβ+CD8αα IELs than wildtype mice. In fact, costimulation-deficient mice contained remarkably more (~10 times more) TCRαβ+CD8αα IELs than wildtype mice (Fig. 7b), suggesting that developmental diversion was the predominant origin of TCRαβ+CD8αα IELs. We then wondered if the TCRαβ+CD8αα IELs present in normal wildtype mice were also the progeny of thymocytes that had survived negative selection and undergone developmental diversion. In fact we found that, even in wildtype mice, SAg-reactive TCRs were specifically over-expressed in TCRαβ+CD8αα IELs but were not over-expressed in other T cell populations (Fig. 7c). This finding suggests that, in normal wildtype mice, significant numbers of autoreactive T cells avoid clonal deletion and undergo developmental diversion, a conclusion which is also consistent with our observation that the frequency of SAg-reactive TCRs in WT mice is always higher among TCRαβ+DN thymocytes than among SP4 or SP8 thymocytes (see Figs. 3a, ,5a5a and Supplementary Fig. 1a).
The over-representation of autoreactive TCR specificities among CD8αα IELs in normal and experimental mice, none of which were autoimmune, made us question the TCR responsiveness of CD8αα IELs. In fact, CD8αα IELs resembled TCRαβ+DN thymocytes in being unresponsive to anti-TCR+APC stimulation in vitro without exogenous cytokines (Fig. 7d).
We conclude that TCRαβ+CD8αα IELs are the progeny of cells that survived negative selection and underwent developmental diversion in the thymus.
Finally, we wished to gain molecular insight into the differentiation of developmentally diverted TCRαβ+DN thymocytes into CD8αα IELs. Because differentiation of TCRαβ+DN thymocytes into CD8αα IELs requires re-activation of Cd8a gene expression which is a known function of the transcription factor Runx330, we considered that Runx3 might be required for differentiation of TCRαβ+DN thymocytes into CD8αα IELs.
To analyze this possibility, we utilized Bcl-2 transgenic Runx3+/YFP heterozygous mice because their cells contained an endogenous Runx3 allele that had been re-engineered to encode YFP instead of Runx3 proteins31. Developmentally diverted TCRαβ+DN thymocytes in these mice were YFPlow whereas TCRαβ+CD8αα IELs were YFPhi (Fig. 8a), indicating that Runx3 gene expression was up-regulated at some point during differentiation of TCRαβ+DN thymocytes into TCRαβ+CD8αα IELs. To determine if that point was related to Cd8a gene re-activation, we stimulated TCRαβ+DN thymocytes in vitro with anti-TCR and IL-15 (Fig. 8b). Upon in vitro stimulation, TCRαβ+DN thymocytes expressed Runx3 and differentiated into CD8αα cells (Fig. 8b upper panels). To determine if their differentiation into CD8αα cells required Runx3, we utilized Bcl-2 transgenic Runx3YFP/YFP homozygous mice which were Runx3-deficient because they contained two Runx3-YFP alleles (Fig. 8b, lower panels). In vitro stimulation of TCRαβ+DN thymocytes that were Runx3-deficient failed to induce differentiation into CD8αα cells even though it did induce YFP expression (Fig. 8b lower panels). Thus, Runx3 proteins were required for developmentally diverted TCRαβ+DN thymocytes to re-activate Cd8a gene expression and to differentiate into CD8αα cells in vitro.
Applying these in vitro observations to the developmental fate of TCRαβ+DN thymocytes in vivo, we wondered if differentiation of developmentally diverted TCRαβ+DN thymocytes into TCRαβ+CD8αα IELs would be impaired in Runx3YFP/YFP homozygous mice because they were Runx3-deficient. Indeed, Runx3-deficient (Runx3YFP/YFP) Bcl-2 transgenic mice contained significantly fewer TCRαβ+CD8αα IELs than Runx3-sufficient (Runx3+/YFP) Bcl-2 transgenic mice, even though they contained equal numbers of developmentally diverted TCRαβ+DN thymocytes (Fig. 8c).
We conclude that TCR and IL-15 stimulate developmentally diverted TCRαβ+DN thymocytes to express Runx3 which promotes Cd8a gene re-activation and differentiation of developmentally diverted TCRαβ+DN thymocytes into TCRαβ+CD8αα IELs. As a result, lineage choice during positive selection and lineage fate during negative selection can be integrated into a unified picture of thymic development in which the developmental fate of TCR-signaled DP thymocytes is determined by TCR and costimulatory signals (schematized in Supplementary Fig. 7).
The present study identifies CD28 costimulatory signals as necessary for thymocytes to undergo clonal deletion during negative selection in vivo, and demonstrates that neither CD28 costimulation nor clonal deletion is required for self-tolerance. Regardless of the presence or absence of CD28-mediated clonal deletion, strong TCR signaling prevented DP thymocytes bearing autoreactive TCRs from differentiating into conventional SP4 or SP8 T cells, so that the absence of autoreactive TCRs on SP T cells was indicative of in vivo negative selection but was not necessarily indicative of in vivo clonal deletion. Indeed, strongly signaled DP thymocytes bearing autoreactive TCRs that survived negative selection did not differentiate into conventional SP T cells, but instead underwent developmental diversion and differentiated into TCRαβ+DN thymocytes. Developmentally diverted TCRαβ+DN thymocytes, upon leaving the thymus, preferentially migrated to the intestine and were signaled by IL-15 to express Runx3 and to further differentiate into Runx3+CD8αα IELs. Clonal deletion and developmental diversion occurred during negative selection in cells at the intermediate CD4hiCD8lo thymocyte stage of differentiation, the identical point in thymocyte development that CD4 and CD8 lineage choice occurs during positive selection 1. Thus, this study distinguishes in vivo negative selection from in vivo clonal deletion, identifies CD28 costimulation as the in vivo signal required for clonal deletion during negative selection, and identifies TCRαβ+CD8αα IELs as the ultimate fate of autoreactive cells that survive negative selection in the thymus.
During thymic selection, weakly signaled DP thymocytes undergo positive selection into SP4 or SP8 T cells, whereas strongly signaled DP thymocytes undergo negative selection which prevents them from differentiating into either SP4 or SP8 T cells. Importantly, as shown in this study, the absence of autoreactive TCR specificities among in vivo SP T cells is indicative of negative selection but it is not necessarily indicative of clonal deletion. Consequently, this study provides a new conceptual model of negative selection in which ‘positive selection’ is the differentiation of TCR-signaled DP thymocytes into SP T cells; ‘negative selection’ is the prevention of TCR-signaled DP thymocytes from differentiating into SP T cells; ‘developmental diversion’ refers to the differentiation of TCR-signaled DP thymocytes into mature T cells that are neither SP4 nor SP8 but are DN; and ‘clonal deletion’ refers to the death of TCR-signaled DP thymocytes prior to maturation.
We used specific TCR-Vβ’s to follow the developmental fate of DP thymocytes strongly signaled by endogenously encoded Mtv proviral antigens. DP thymocytes are the first cells in the thymus to express endogenously encoded TCRαβ complexes and are the cells that are subjected to thymic selection. As previously observed, DP thymocytes bearing SAg-reactive TCRs underwent negative selection as they did not differentiate into SP4 or SP8 T cells23. However, our study now documents that clonal deletion of SAg-reactive thymocytes during negative selection additionally required CD28 costimulation. In the absence of CD28 costimulation, SAg-reactive thymocytes survived negative selection and underwent developmental diversion into TCRαβ+DN thymocytes. Similarly, SAg-reactive thymocytes that were prevented from undergoing clonal deletion by Bcl-2 or Mcl-1 transgenes also survived negative selection and underwent developmental diversion into TCRαβ+DN thymocytes. Thus, contrary to current concepts, CD28 costimulation during negative selection is required for in vivo clonal deletion and is prevented by transgenic expression of either Bcl-2 or Mcl-1.
Developmental diversion of surviving autoreactive thymocytes into TCRαβ+DN thymocytes avoids autoreactivity by removing the contribution of CD4 and CD8 coreceptor proteins to TCR signaling which lessens their in vivo autoreactive potential, as does their expression of negative costimulatory receptors such as PD-1 28. In addition, upon leaving the thymus, developmentally diverted TCRαβ+DN thymocytes migrate to the intestine where they are sequestered away from the rest of the body and signaled by IL-15 to express Runx3, which re-activates Cd8a gene expression and promotes further differentiation into CD8αα IELs that are hyporesponsive to TCR stimulation. Their CD8αα phenotype may further reduce their autoreactive potential, as CD8αα surface complexes have been suggested to sequester Lck signaling kinases away from the TCR32. Indeed, costimulation-deficient mice are free of autoimmunity despite our finding that they contain huge numbers of CD8αα IELs bearing autoreactive TCR specificities and despite lacking functionally suppressive Foxp3+ regulatory T cells33.
Developmental diversion occurs in TCR signaled CD4hiCD8lo intermediate thymocytes that are transcriptionally Cd4+Cd8−, with the result that differentiation into TCRαβ+DN thymocytes during negative selection only requires termination of Cd4 gene expression. In fact Cd4 gene termination occurs in Cd4+Cd8− intermediate thymocytes during MHC class I-specific positive selection into SP8 T cells, but in that case it is mediated by the transcription factor Runx3 30, 34 which additionally induces Cd8 gene reactivation, events referred to as ‘coreceptor reversal’35. In contrast to MHC class I selected SP8 T cells which are dependent on Runx3, developmental diversion into TCRαβ+DN thymocytes did not involve or require Runx3, suggesting that strong TCR signaling of intermediate thymocytes during negative selection terminates Cd4 gene transcription independently of Runx3 and without Cd8 gene re-activation. Interestingly, Runx3-independent termination of Cd4 gene transcription in Cd4+Cd8− intermediate thymocytes by strong TCR signaling would also explain why negatively selected thymocytes become DN instead of SP4 T cells.
By identifying developmental diversion and clonal deletion as alternative outcomes of negative selection, the present study resolves a number of longstanding experimental contradictions. Previous in vitro experiments demonstrated that TCR signaled thymocyte death required CD28 costimulation10–14 and could be prevented by Bcl-2 transgene expression12, 14, whereas previous in vivo experiments came to opposite conclusions18–22. Our present study reveals that these contradictory observations were primarily due to the presumption that autoreactive thymocytes had been clonally deleted in vivo if they failed to differentiate into either SP4 or SP8 Tcells. In fact our current study documents that the absence of autoreactive TCR specificities on SP T cells in vivo indicated that autoreactive thymocytes had undergone negative selection but did not indicate that they had undergone clonal deletion. Our present study also resolves the discrepant observations that genetic deletion of the pro-apoptotic protein Bim interfered with clonal deletion36 while transgenic over-expression of the anti-apoptotic protein Bcl-2 did not20. In fact many features of Bim-deficient mice37–39 resemble those of Bcl-2 transgenic and Mcl-1 transgenic mice in our current study, so that our study explains the high frequency of TCRαβ+DN thymocytes in Bim- and Puma-deficient mice37, 40, 41 as autoreactive thymocytes that survived negative selection and underwent developmental diversion. Our present study also explains why SP4 and SP8 T cells bearing Mtv-6 reactive Vβ3+ TCRs were previously observed in B7-deficient but not Cd28−/− BALB/c mice24. The discrepancy between B7-deficient and Cd28−/− BALB/c mice did not indicate an as yet unknown B7-specific receptor that signaled clonal deletion in Cd28−/− BALB/c mice24, but was instead due to the fact that Mtv-6 was not encoded in the genome of B7-deficient BALB/c mice.
While resolving many longstanding experimental contradictions, our present study does contradict an experimental finding in perinatal mice in which costimulation blockade by in vivo anti-B7 antibody injections rescued SAg-reactive TCRs which were then expressed on SP4 T cells42. In our current study, genetic deletion of B7 ligands in B7-deficient mice also rescued SAg-reactive TCRs but the rescued TCRs were not expressed on SP4 T cells, only on developmentally diverted TCRαβ+DN thymocytes and CD8αα IELs. We think the explanation for this disparity is that anti-B7 antibody injections depleted SAg-bearing B7+ cells from the thymi of injected mice and so eliminated intra-thymic expression of the negative selecting (SAg) ligand.
Does developmental diversion occur in normal wildtype (i.e. costimulation-sufficient) mice? Costimulation-sufficient TCR transgenic mice contain substantial numbers of TCRαβ+DN thymocytes37, 41, 43 that can become CD8αα IELs 29, 32, 43–46. However, transgenic TCRs differ from endogenous TCRs in that they are expressed by early DN thymocytes prior to the DP stage of differentiation, so it is difficult to know if TCRαβ+DN thymocytes in TCR transgenic mice are post-selection cells resulting from developmental diversion or are pre-selection DN thymocytes that were developmentally arrested because their transgenic TCR engaged high affinity ligands in the thymic cortex47, 48. Importantly, the present study found in normal, non-transgenic mice that many or all TCRαβ+CD8αα IELs were the progeny of thymocytes that survived negative selection and underwent developmental diversion in the thymus. Indeed normal wildtype mice contained substantial numbers (~5 million) of TCRαβ+CD8αα IELs in which autoreactive TCR specificities were over-expressed, indicating that substantial numbers of autoreactive T cells avoid clonal deletion and undergo developmental diversion in normal mice.
In conclusion, by distinguishing in vivo negative selection from in vivo clonal deletion, the present study identified CD28 costimulation as critical for clonal deletion and identified developmental diversion as an alternative outcome of negative selection. Moreover, this study identified TCRαβ+CD8αα IELs as the ultimate fate of autoreactive T cells that survive negative selection and undergo developmental diversion in the thymus. And, by discovering that the fate of thymocytes undergoing negative selection was determined at the same point in differentiation as thymocytes undergoing positive selection, this study integrates negative and positive selection into a unified picture of thymic selection.
BALB/c, C57BL/6 (B6), Cd28−/−, B2m−/−, Ab−/−b2m−/−, Zap70−/−, Rag2−/−, and human Bcl-2 transgenic mice20 (Bcl-2 TG) were maintained in our own colony. Cd80−/−Cd86−/− mice were referred to as B7-deficient and were provided by Arlene Sharpe (Harvard University); RUNX3-YFP reporter mice were provided by Dan Littman (NYU) 31; Pdcd1−/− mice was provided by Tasuku Honjo (Kyoto University). The Mcl-1-transgenic construct was made by ligating Mcl-1 cDNA into a human CD2 (hCD2) enhancer-promoter-based vector and injected it into fertilized B6 oocytes to generate Mcl-1 transgenic mice. All mice were cared for in accordance with National Institutes of Health (NIH) guidelines.
Monoclonal antibodies with the following specificities were used in this study: CD4 (GK1.5 and RM4.5), CD5 (53-7.3), CD8α (53-6.7), CD8β (H35-17.2), TCRγδ (GL3), TCRβ (H57-597), Vβ3 (KJ25), Vβ4 (KT4), Vβ5.1-5.2 (MR9-4), Vβ6 (RR4-7), Vβ8 (F23.1), Vβ10 (B21.5), Vβ11 (RR3-15), Vβ12 (MR11-1), NK1.1 (PK136), PD-1 (J43), CD132 (4G3), CD122 (TM-beta1), CD44 (IM7), CD69 (H1.2F3), Bcl-2 PE-set (3F11), IL4Rα (mIL4R-M1), IL7Rα (SB/199), CD25 (PC61), CD28 (37.51), were obtained from BD Pharmingen; Bcl-XL (54H6) and the cleaved form of Caspase-3 (D175) were obtained from Cell Signaling; IL-2 (JES-5H4) was from eBioscencie; CD8α (5H10) was from Invitrogen. Annexin-V FLUOS staining kit was obtained from Roche Applied Science. PBS57 loaded CD1d tetramers or unloaded CD1d control tetramers were provided by the NIH Tetramer Core Facillity. Cells were analyzed either on a FACSVantage SEM or LSRII (BD). Doublets and dead cells were removed from data analysis. For intracellular staining, live cells were first stained for surface proteins, then fixed and permeabilized, and then stained for intracellular proteins. Data were analyzed with software designed by the Division of Computer Research and Technology of National Institutes of Health or with FlowJo software by TreeStar.
CD8+ and LNT cells were isolated by antibody-mediated magnetic bead depletion of CD4+ and Ig+ cells (BioMag Qiagen). For purification of TCRαβ+DN thymocytes, thymocytes were depleted of CD4+ and CD8+ cells by antibody-mediated magnetic bead depletion and then electronically sorted for either CD5+CD4−CD8β −B220−GL3−NK1.1− or CD5+CD4−CD8β −B220−GL3−CD1dPBS57− cells. Isolation of IELs was performed as described49. For purification of TCRαβ+CD8αα IELs, IELs were electronically sorted for CD45+CD8α+CD8β − CD4−GL3− cells.
Cells were loaded with the calcium sensitive dye Indo-1 (1.8 mM, Invitrogene) at 31 °C and then coated at 4 °C with biotinylated anti-TCRβ. Cells were warmed 2 min prior to stimulation and applied to the flow cytometer. Antibody crosslinking was induced with avidin (4μg/ml, Sigma) and data acquisition was recorded for 5 min.
0.8 × 106 purified TCRαβ+DN thymocytes or CD8+CD5+ LNT cells were injected into Rag−/− host mice. Five weeks after transfer, cells from thymi, peripheral lymphoid organs, and the small intestine were isolated and analysed by flow cytometry.
For in vitro differentiation, purified TCRαβ+DN thymocytes were cultured in medium only or stimulated with immobilized anti-TCRβ mAb (5–10 μg/ml) in the presence of IL-15 (100 ng/ml, R&D Systems) for 4–5 days at which time cells were harvested and analyzed by flow cytometry. For in vitro proliferation, sorted TCRαβ+DN and SP8 thymocytes from B7-deficient mice were stimulated for 72h with immobilized anti-TCRβ (5μg/ml) and anti-CD28 mAbs (10μg/ml) in the presence or absence of recombinant IL-2 (200 U/ml). For in vitro proliferation of IELs, sorted TCRαβ+CD8αα and SP4 LN T cells were stimulated for 48h with anti-CD3 (5μg/ml) in the presence of 105 APCs (3000R irradiated syngenic splenocytes) in the presence or absence of recombinant IL-2 (200 U/ml) or IL-15 (100ng/ml). Cultures were pulsed with [3H]-thymidine (1μCi) 8h before collection. For intracellular IL-2 production, T cells that had been stimulated for 4d were treated with PMA (50ng) and ionomycin (500nM) for 4h in the presence of protein transport inhibitor (BD Biosciences) prior to intracellular staining for IL-2.
PCR reactions were performed using mouse tail DNA to detect genomes of integrated Mtv-6 and Mtv-9 proviruses. Primers for: Mtv-6 forward 5′-GCTGGCTATCATCACAAGAGCG-3′, reverse 5′-GGAGTTCAACCATTTCTGCTGC-3′; Mtv-9 forward 5′-ACCGCAGTCAAAGAACAGGTGC-3′ reverse 5′-CAGGAAACCACTTGTCTCACATCC-3′.
Methylation analysis was performed as described50. Briefly, DNA was isolated using the ZR genomic DNA II kit (Zymo Rearch). Bisulfite conversion of total DNA was made using the EZ-DNA methylation-Gold kit (Zymo Research). The following PCR primers were used: CD8β promoter forward 5′-TTGAAAAGTTAAGGTTTTGATGTT-3′ and reverse 5′-AAACACTATTCCCCTCAATACTCTATC-3′. PCR products were purified and cloned using the TOPO-TA cloning kit. Plasmid DNA from MiniPrep (Qiagen) was sequenced.
Statistical significance was determined by Student’s t-test with two tail distribution.
We are grateful to Naomi Taylor, James DiSanto and Richard Hodes for critically reading the manuscript; Arlene Sharpe for B7-deficient mice, Tasuku Honjo for Pdcd1−/− mice, and Dan Littman for Runx3-YFP reporter mice; and Susan Sharrow, Anthony Adams, and Larry Granger for expert flow cytometry. This research was supported by the Intramural Research Program of the NIH, NCI, Center for Cancer Research.