Whether intrathymic-positive and -negative selection of conventional αβ T cells occur in anatomically distinct sites is a matter of debate. By using a system composed of two distinct immune receptors, the Y-Ae mAb and the 1H3.1 (Vα1/Vβ6) TCR, both directed against the 52–68 fragment of the I-Eα-chain (Eα52–68) bound to I-Ab, we examined the occurrence of negative selection imposed in vivo by a self-peptide-self-MHC class II complex with differential tissue expression. 1H3.1 TCR-transgenic (Tg) mice were bred to mice having an I-Eα transgene with expression directed to all MHC class II-positive cells, restricted to thymic epithelial cells, or restricted to B cells, dendritic cells, and medullary thymic epithelial cells. All 1H3.1 TCR/I-Eα double-Tg mice revealed a severely diminished thymic cellularity. Their lymph node cells were depleted of Vβ6+CD4+ cells and were unresponsive to Eα52–68 in vitro. The absolute number of CD4+CD8+ thymocytes was drastically reduced in all combinations, indicating that negative selection caused by an endogenously expressed self-determinant can effectively occur in the thymic cortex in vivo. Moreover, both cortical epithelial cells and, interestingly, the few cortical dendritic cells were able to support negative selection of CD4+CD8+ thymocytes, albeit with a distinct efficiency. Collectively, these observations support a model where, in addition to the avidity of the thymocyte/stromal cell interaction, in vivo negative selection of autoreactive TCR-Tg T cells is determined by accessibility to self-peptide-self-MHC complexes regardless of the anatomical site.
The full activation of naive conventional αβ T lymphocytes (CD3high CD4+CD8− or CD4−CD8+) requires the integration of two signals (1). The signal 1 corresponds to the recognition of the cognate peptide-MHC complex by the clonotypic Ag receptor of T cells, whereas the signal 2 corresponds to a non-Ag-specific stimulus provided, for instance, by the interaction of CD28 with B7 molecules on the surface of APCs. It is the recognition of invariant microbial molecular motifs by nonclonotypic receptors that induces the expression of costimulatory molecules on APCs (2). When mature/activated APCs coexpress foreign-peptide-self-MHC complexes and a high level of costimulatory molecules, they still present a vast array of peptides derived from self-Ags. However, this usually does not lead to initiation of autoimmune responses because most T cells that would be susceptible to drive such responses are physically eliminated at their immature stage in the thymus by induction of apoptosis (3, 4, 5, 6). Thus, intrathymic negative selection eliminates virtually all T cell specificities able to react strongly to self-determinants expressed on thymic APCs. This process is a major mechanism of central tolerance. In contrast, positive selection allows CD3low CD4+CD8+ immature thymocytes able to react with a moderate affinity to self-peptide-self-MHC complexes to survive and to differentiate into mature CD4+ or CD8+ single-positive T cells (4, 5, 7).
Radiation bone marrow chimeras (BMC)5 have revealed that efficient intrathymic positive selection requires the interaction of the TCR with appropriate MHC molecules expressed on cortical epithelial cells (8, 9, 10, 11, 12). This compartmentalization of positive selection was further demonstrated in an elegant system where the restricted expression of surface MHC class II molecules to the cortical epithelium is sufficient for positive selection of CD4 T cells (13). In this study, the keratin-14 promoter (K14) was used to drive the expression of the I-Aβb gene on epithelial cells in MHC class II deficient (I-Aβb−/−) mice. It is clear that Ag expression by cells having a hemopoietic origin is sufficient to drive negative selection because transfer of bone marrow (BM)-derived cells from male transgenic (Tg) mice expressing a TCR specific for the male Ag into irradiated recipients leads to an effective deletion of thymocytes regardless of the MHC haplotype of the recipient (14). Clonal deletion that affects immature as well as semimature (CD4+CD8−, heat stable Aghigh) thymocytes (15) seems to require engagement of both TCR and costimulatory molecule receptors such as CD28 to be optimal (6, 16, 17). The thymic medulla is rich in BM-derived cells expressing various costimulatory molecules. Therefore, the prevailing view is that the medulla is the site of negative selection. This is supported by studies describing endogenous superantigen (SAG) and circulating Ag-driven negative selection where a massive apoptosis is observed in the medulla (18, 19, 20, 21), and by the detection of “autoreactive” CD4+ T cells in the K14-Aβb-Tg mice, presumably due to the lack of negative selection in the absence of medullary MHC class II-positive cells (13).
The possibility that clonal deletion of self-peptide-self-MHC complex specific thymocytes can occur in the cortex has been less well examined and remains controversial (22). The cortex is separated from the medulla, presents a complex network of ultrastructurally distinct (but all MHC class II+) epithelial cells, and contains fewer BM-derived MHC class II positive cells. Therefore, although they allow the analysis of a nonmanipulated TCR repertoire, the endogenous SAG-based models of negative selection are problematic with respect to the distribution of the deleting ligand. For instance, expression of mouse mammary tumor virus Ag may be distinct between the two compartments because mammary tumor virus Ags are expressed dominantly by BM-derived cells (23). It is also difficult to effectively detect the fine localization of the MHC class II-SAG complex expression. Finally, adding to this complexity, it is not clear whether SAG-induced deletion exactly mimics the conditions of self-peptide-self-MHC complex-induced negative selection. In this respect, it was reported that CD30-deficient mice show an impaired peptide Ag-induced negative selection but normal Mls-2a-induced deletion of reactive T cells (24).
In the case of peptide Ags, it has been observed that the deletion of anti-male Ag-Tg T cells is characterized by the disappearance of immature CD4+CD8+ thymocytes (14) and that Tg T cells specific for a lymphocytic choriomeningitis virus epitope are already deleted at the CD4+CD8+ stage (25, 26). More directly, i.p. injection of antigenic peptide causes a rapid deletion of CD4+CD8+TCRlow-Tg thymocytes accompanied by apoptosis in the cortical area as shown by electron microscopy (27) and in situ detection of apoptosis (28). However, it is now clear that injection of Ag or antigenic peptide to TCR-Tg animals induces mature T cells to produce soluble factors toxic for immature CD4+CD8+ thymocytes such as glucocorticoids (29) or TNF (30). Thus, in some situations, elimination of CD4+CD8+ thymocytes could be stress-related rather than the reflection of a true Ag-mediated clonal deletion. Indeed, the site of deletion of Tg thymocytes caused by peptide injection (cortical area) has been found to be distinct from the site of deletion imposed by the Tg expression of the relevant Ag (corticomedullary junction) (28). Furthermore, the nonspecific deletion of immature cortical thymocytes caused by Ag administration was recently demonstrated (31).
In this report, we examine the occurrence and anatomy of negative selection by using Tg mice expressing a αβ TCR (1H3.1) specific for the Eα52–68-I-Ab complex that also is specifically recognized by the Y-Ae mAb. Breeding of 1H3.1 TCR-Tg mice to various I-Eαd-Tg lines having a differential expression of the Eα52–68-I-Ab complex (well-characterized by using Y-Ae) was used to endogenously express the deleting ligand in a cell type-targeted fashion. The results indicate that the intrathymic confrontation with the Eα52–68-I-Ab complex expressed on various stromal cell types results in deletion of 1H3.1 TCR-Tg thymocytes irrespective of the thymic compartment in which the activating ligand is expressed.
Materials and Methods
Mice used were 3–6 wk old and were housed in the Yale Immunobiology Mouse Unit (New Haven, CT). C57BL/6, B10.A-H2i5 H2-Tl8a(5R)/SgSnJ (5R), and AKR mice were obtained from The Jackson Laboratory (Bar Harbor, ME). The 5R Rag-deficient (Rag-2−/−) mice were obtained from Dr. K. Bottomly (Yale University, New Haven, CT). The 107-Tg (107-1 line), Igk-Eα-Tg, and 36.5-Tg mice were generated in the laboratory of Dr. R. A. Flavell (Yale University). See Table I for characteristics and references. A PCR assay was developed for the identification of the I-Eαd-Tg mice by using tail DNA. The sequences of the oligonucleotide primers used were: sense, 5′-ATTTCTTGAAATGTTAAGTGGAAA-3′, which is specific for the 5′ end of the I-Eα gene fragment absent in C57BL/6 mice; and antisense, 5′-GAAAAATCTTAACACCAGGGC-3′, which is specific for the sequence immediately downstream of the initiation codon. The PCR product is 240 bp long. The 1H3.1 TCR αβ-Tg mice were generated as described previously (32) by using the pTα and pTβ cassette vectors (a gift from Drs. D. Mathis and C. Benoist, Universite Louis Pasteur, Strasbourg, France), which contain the proximal promoters, enhancer, and transcriptional initiation sites of the α and β loci (33) to ensure a normal timing and regulation of expression.
|Strain .||MHC Class II Genes .||.||.||.||I-E Cell Surface Expression .||Ref. .|
|.||Aβ .||Aα .||Eβ .||Eα .||.||.|
|B10.A (5R)||b||b||b||k||MHC class II+ cells|
|Igκ-Eα Tgb||b||b||b||d||B cells, DCs, mEpc||40|
|36.5 Tg||b||b||b||d||cEpc and mEpc only||39, 41|
|107 Tg||b||b||b||d||MHC class II+ cells||39, 41|
|Strain .||MHC Class II Genes .||.||.||.||I-E Cell Surface Expression .||Ref. .|
|.||Aβ .||Aα .||Eβ .||Eα .||.||.|
|B10.A (5R)||b||b||b||k||MHC class II+ cells|
|Igκ-Eα Tgb||b||b||b||d||B cells, DCs, mEpc||40|
|36.5 Tg||b||b||b||d||cEpc and mEpc only||39, 41|
|107 Tg||b||b||b||d||MHC class II+ cells||39, 41|
Nonfunctional allele (38 ).
I-E expression on most B cells as well as on a large fraction of DCs is lower than I-A expression. I-E expression is not detected on Langerhans cells, on cEpc, or on macrophages (40 ). DCs, Dendritic cells; mEpc, medullary thymic epithelial cells; cEpc, cortical thymic epithelial cells.
Immunostaining and flow cytometry
Depending on experiments, thymus, spleen, and lymph nodes (axillary, lateral axillary, superficial inguinal, and mesenteric) were removed and cell suspensions prepared. Splenic RBC were lysed with Tris-buffered ammonium chloride. Fluorescent-labeled mAbs were used for multicolor staining. Briefly, 0.2 × 106 cells were incubated in microtiter U-bottom plates with a saturating concentration of labeled mAb in 20 μl for 30 min on ice. Cells were washed twice and analyzed immediately. For two-step staining, cells first were incubated with purified mAbs in PBS 2% FCS/0.1% NaN3, followed by a F(ab′)2 of goat anti-mouse Ig-FITC conjugate from Sigma (St. Louis, MO). The mAbs used were anti-Vβ6-FITC (clone RR4-7), anti-Cβ-PE (H57-597), anti-CD90.2/Thy-1.2-PE (53-2.1), anti-B220-PE (RA3-6B2), and anti-Vα2,3.2,8,11-FITC (B20.1, RR3–16, B21.14, RR8-1) from BD PharMingen (San Diego, CA), anti-CD8α-PE/FITC (53-6.7) from Life Technologies (Rockville, MD), and anti-CD4-quantum red (H129.19) from Sigma. The Y3JP (mouse IgG2a, anti-I-Ab), 14.4.4 S and Y17 (mouse IgG2a and IgG2b, anti-I-E), 2.4G2 (rat IgG2b, anti CD16/CD32), 25.9.17 (mouse IgG2a, anti-I-Ab), Y-Ae (mouse IgG2b, anti-Ab+Eα), 10.2.16 (mouse IgG2a, anti-I-Ak,r,f,s), GK1.5 (rat IgG2b, anti CD4), 53-6.72 and 2.43 (both rat IgG2b, anti CD8), and 14.8 (rat IgG2b, anti CD45RA/B220) mAbs were affinity-purified in the laboratory by using standard procedures. A flow cytometer equipped with a 15-mW air-cooled argon-ion laser (FACScan) and CellQuest software, both from Becton Dickinson (Mountain View, CA), were used to collect and analyze the data. Nonviable cells were excluded by using forward- and side-scatter electronic gating. FITC-labeled annexin V was purchased from BD PharMingen and used according to provided instructions.
Immunohistofluorescence and immunohistochemistry
Thymi were fixed overnight in 1% paraformaldehyde lysine periodate buffer, infused with sucrose, embedded in Tissue Tek OCT (Miles, Elkhart, IN), and frozen. Sections measuring 5–7 μm were cut by using a Leica CM1800 cryostat (Leica, Heerbrugg, Switzerland), air dried at room temperature, and stained. For immunofluorescence, sections were treated with a saturating concentration of the anti-mouse CD16/CD32 (FcγRIII/II) 2.4G2 mAb in PBS, incubated with anti-Vβ6-FITC, and biotinylated α-l-fuc(glcNAc)2-specific Ulex europaeus 1 (UEA-1) lectin (Sigma) followed by streptavidine-PE (Caltag, San Francisco, CA). Preparations were examined and photographed on an Axiophot 2 apparatus (Carl Zeiss, Thornwood, NY).
For T cell proliferation assays, T cells were isolated from lymph nodes and cultured in U-bottom 96-well plates (Becton Dickinson) 3–4 days at 37°C in Click’s Eagle-Hank’s amino acid medium (Irvine Scientific, Santa Ana, CA) supplemented with 5% heat-inactivated FCS (Intergen, Purchase, NY), 5 × 10−5 M 2-ME (Bio-Rad, Richmond, CA), 2 mM l-glutamine, and 50 μg/ml gentamicin (Life Technologies). In some cases, Tg T cells were sorted for absence of MHC class II and CD8 expression by using magnetic beads and the Y3JP, 53-6.72, and 2.43 mAbs. Depending on the experiment, T cells (10–15 × 103/well) were stimulated by using irradiated splenocytes of different types (3 × 105 or less/well, 2000 rad) or splenocytes plus serial dilutions of synthetic Eα52–68 peptide or anti-CD3ε mAb (YCD3-1) in a total volume of 150 μl. The cells were incubated in duplicate wells, and 1 μCi [3H]thymidine/well was added to the culture for the last 12 h. The plates then were harvested and counts per minute were determined by using liquid scintillation counting. For inhibition experiments, purified mAbs diluted in complete medium were sterile-filtered and added to microcultures. The Eα52–68 peptide (ASFEAQGALANIAVDKA; single-letter amino-acid code) was synthesized, HPLC-purified, and mass spectrometry-analyzed by the W.M. Keck Biotechnology Resource Center (Yale University).
Generation of 1H3.1 TCR-Tg mice having a wild-type or tissue-restricted expression of the Eα52–68-I-Ab complex
The 1H3.1 αβ TCR and the Y-Ae mAb are immune receptors both specific for the 52–68 fragment of the I-Eα-chain presented in the context of I-Ab MHC class II molecules (34, 35, 36, 37). As expected, T cells from 1H3.1 αβ TCR-Tg mice are induced, in a Y-Ae inhibitable manner, to produce cytokine and proliferate in response to splenocytes from B10.A (5R) (I-Ab+/I-Eα+) mice that naturally assemble and present the Y-Ae epitope. Because C57BL/6 mice lack a functional I-Eα gene (38), their APCs do not cause such activation unless the Y-Ae epitope is recreated by exogenously providing synthetic Eα52–68 peptide (32).
We chose this system to study the intrathymic negative selection of MHC class II-restricted T cells because the endogenous expression of the cognate TCR ligand on thymic stromal cells can be finely analyzed by using Y-Ae. The 1H3.1 TCR-Tg mice were bred to several Tg lines that express I-Eαd in different tissues or tissue compartments depending on the length the MHC class II promoter used. We used the 107-Tg mice (39), which express I-Eα, and therefore surface I-E molecules, on all MHC class II-positive cells but with a slightly higher level than APCs from mice naturally expressing an I-Eα-chain such as B10.A (5R). The Igκ-Eα-Tg mice express I-E molecules at an intermediate level on most B cells and on a large fraction of dendritic cells from lymphoid organs but neither on macrophages nor Langerhans cells (40). In these mice, I-E molecules are also expressed on medullary but not cortical thymic epithelial cells (40). Finally, the 36.5-Tg mice express I-E molecules exclusively on thymic epithelial cells (Refs. 39 and 41 ; see Table I for a synopsis). The breedings were set up with heterozygous I-Eαd-Tg mice to simultaneously generate and analyze TCR-Tg littermates that differ only by the presence or absence of the I-Eαd transgene. The flow cytometry-coupled immunofluorescence analysis of total splenocytes from C57BL/6 I-Eαd-Tg mice (Fig. 1) demonstrates that the Tg expression of I-Eαd (detected through surface expression of I-Eαdβb complexes by using the anti-I-E Y17 and 14.4.4 S mAbs) effectively directs expression of the Eα52–68-I-Ab complex: Y-Ae-positive cells are present among 107-Tg and Igκ-Eα-Tg splenocytes but not 36.5-Tg splenocytes. In accordance with the Y17 and 14.4.4 S stainings, the Y-Ae staining intensity is high for 107-Tg cells and low for Igκ-Eα-Tg cells.
Within the thymus, extensive immunohistochemistry analyses have revealed that I-E expression is reticular in the cortex and rather scattered and dense in the medulla of 36.5-Tg mice (39, 41). Accordingly, both cortical and medullary epithelial cells are Y-Ae-positive in the 36.5-Tg thymus (41). Besides the negative staining of splenocytes (Fig. 1), the lack of I-E expression in 36.5-Tg BM-derived cells is evidenced by several observations. First, the medullary Y17 staining matches the staining obtained with the fucose-binding lectin UEA-1 (41), which reacts to most medullary epithelial cells (42). Second, thymic sections from 36.5-Tg→C57BL/6 BMC show no detectable I-E expression (41). Third, the Y17 staining of thymic sections from 36.5-Tg and C57BL/6→36.5-Tg BMC are indistinguishable (41). In the Igκ-Eα thymus, I-E expression is observed throughout the medulla on stromal cells, which are keratin negative (i.e., presumably BM-derived cells; Ref. 40). In addition, C57BL/6→Igκ-Eα BMC have revealed that I-E expression is found on medullary but not cortical thymic epithelial cells (40). Finally, the staining profile of thymic sections from 107-Tg mice resembles closely that of B10.A (5R): a strong Y-Ae staining is observed throughout the medulla, on cortical epithelial cells, and on cortical macrophages (41).
Drastic reduction of the thymic cellularity and deletion of immature CD4+CD8+ thymocytes in all 1H3.1 TCR-Tg/I-Eαd double-Tg mice
To analyze the phenotype of the lymphoid organs, 1H3.1 TCR/I-Eαd double-Tg mice and control littermates were sacrified at 4–6 wk of age, and cell suspensions were prepared from thymus, spleen, and lymph nodes. The flow cytometry-coupled immunofluorescence analysis of the thymus and lymph nodes is presented in Figs. 2 and 3. All double-Tg mice showed a drastic reduction of the thymic size. The reduction of the absolute number of thymocytes was usually ∼90–95%, except for the 1H3.1 TCR/36.5 double-Tg thymi, which displayed a less severe reduction (75–80%). Unlike normal 1H3.1 TCR-Tg (Fig. 2, top), 1H3.1 TCR/I-Eαd double-Tg completely lack the large population of Vβ6highCD4+ thymocytes, that is, the Tg thymocytes that are beyond the stage of positive selection. The absolute number of CD4+CD8− thymocytes typically dropped from 50–55 × 106 in 1H3.1 TCR-Tg mice to 0.1–1.5 × 106 in 1H3.1 TCR/I-Eαd double-Tg mice. A large number of CD4−CD8− cells accumulated in the 1H3.1 TCR/107 double-Tg thymus when compared with 1H3.1 TCR/Igκ-Eα and 1H3.1 TCR/36.5 double-Tg thymi. The thymic profile of B10.A (5R) 1H3.1 TCR-Tg mice (Fig. 2, bottom) resembled that of 1H3.1 TCR/107 double-Tg mice. We also generated B10.A (5R) 1H3.1 TCR-Tg RAG-2−/− mice, which cannot recombine alternative TCR α-chains because of the lack of recombinase activity. In these mice, virtually no CD4+CD8− thymocytes were detected, whereas a substantial population of CD4−CD8− cells accumulated (not shown). Another major feature of all 1H3.1 TCR-Tg/I-Eαd double-Tg thymi was the strong reduction of the absolute number of CD4+CD8+ thymocytes; that is, the CD3low/int small immature thymocytes that essentially populate the cortical compartment (43). The reduction was typically from 55–60 × 106 to 2–3 × 106 for the 1H3.1 TCR/Igκ-Eα thymus and to 0.1–1 × 106 for the B10.A (5R) 1H3.1 TCR-Tg and 1H3.1 TCR/107 double-Tg thymi. However, the deletion of CD4+CD8+ cells was less severe in the 1H3.1 TCR/36.5 double-Tg thymus (from 55–60 × 106 to 14–18 × 106). In the periphery (Fig. 3), the fraction of Vβ6+CD4+ cells is dramatically reduced. This fraction was lacking in lymph node cells from B10.A (5R) 1H3.1 TCR-Tg RAG-deficient mice (not shown). The CD8+ population, which represents ∼5–10% of Vβ6+ lymph node cells in normal 1H3.1 TCR-Tg mice also is reduced in 1H3.1 TCR-Tg/I-Eαd double-Tg mice (1–5%), except in 1H3.1 TCR/36.5 double-Tg mice, where such cells appear to accumulate. Vβ6+CD4−CD8− cells accumulated in the lymph nodes of most 1H3.1 TCR-Tg Y-Ae+ mice, except again in 1H3.1 TCR/36.5 double-Tg mice (Fig. 3, right).
Thus, despite the differential expression of the Eα52–68-I-Ab complex by thymic stromal cells, all 1H3.1 TCR-Tg/I-Eαd double-Tg mice severely deleted immature CD4+CD8+ thymocytes which, for the vast majority, are localized in the cortical compartment. This deletion was most severe (∼99% reduction in absolute number) in mice that express the Y-Ae epitope on all MHC II+ cells (TCR/107 double-Tg and B10.A (5R) TCR-Tg) and less severe (∼65–70% reduction in absolute number) in mice with a thymic epithelium-restricted expression of the Y-Ae epitope (TCR/36.5 double-Tg). It is remarkable that deletion of CD4+CD8+ thymocytes is also drastic (∼90–95% reduction in absolute number) in mice that express I-E molecules at an intermediate level on many dendritic cells but neither on cortical epithelial cells nor on macrophages (TCR/Igκ-Eα double-Tg mice).
Visualization of thymic clonal deletion imposed by the Y-Ae epitope in 1H3.1 TCR/I-Eαd double-Tg mice
To examine the spatial distribution of negative selection of 1H3.1 TCR-Tg thymocytes in Y-Ae-positive mice, we performed immunostaining of frozen thymic sections (Fig. 4). Because the fucose-binding lectin UEA-1 reacts to medullary thymic epithelial cells (42), sections were costained with an anti-Vβ6 mAb and UEA-1 to simultaneously detect 1H3.1 TCR-Tg thymocytes and delineate the cortical and medullary compartments. The anti-Vβ6 staining reveals that bright Vβ6+ cells are asymmetrically distributed in the TCR/Igκ-Eα double-Tg thymus (Fig. 4, A–B): they are more abundant in the cortex than in the medulla. Despite the differential I-E expression, a relatively similar pattern was observed in the case of the TCR/36.5 double-Tg thymus (Fig. 4, C–D). In sharp contrast, analysis of the TCR/107 double-Tg thymic sections (Fig. 4, E–F) shows that Tg thymocytes are already subject to a massive physical elimination in the cortical zone, as demonstrated by the paucity of bright Vβ6+ cells in the thymic area not stained by the UEA-1 lectin. The lack of Vβ6+ cells was less striking in the medulla. This is most likely corresponding to the sizable fraction of Vβ6+CD4−CD8− cells that is observed in the thymus and also the periphery of these mice (see Fig. 2). The staining pattern of B10.A (5R) 1H3.1 TCR-Tg thymi closely resembled that of TCR/107 double-Tg thymi (data not shown).
Together, the data from the flow cytometry and the immunohistofluorescence analyses provide several indications. First, cortical epithelial cells can drive deletion of immature CD4+CD8+ 1H3.1 TCR-Tg thymocytes in vivo as shown by the reduction of this population in TCR-Tg mice with expression of the Y-Ae epitope restricted to the thymic epithelium (1H3.1 TCR/36.5 double-Tg). The phenotype of these mice also suggest that medullary thymic epithelial cells can support an efficient deletion because these mice lack Vβ6+ thymocytes in their medulla. Second, BM-derived MHC class II+ cells that are present in the cortex can efficiently induce negative selection in vivo as shown by the strong reduction of the absolute number of CD4+CD8+ thymocytes in 1H3.1 TCR/Igκ-Eα double-Tg mice that do not express I-E molecules on cortical epithelial cells (40). In this case, the deletion of CD4+CD8+ thymocytes appears driven by rare cortical dendritic cells because Igκ-Eα-Tg mice do not express I-E molecules on macrophages (40). Third, the deletion of cortical thymocytes is most efficient when all cortical stromal cells express the cognate self-peptide-MHC class II complex as shown by the drastic elimination of cortical thymocytes in TCR/107 double-Tg mice (Fig. 4, E–F). This most probably results from the integration of negative selection caused by both cortical epithelial cells and dendritic cells. Theoretically, macrophages may contribute as well because they are I-E+ in 107-Tg mice (39). However, thymic macrophages have been found to be surprisingly inefficient at causing deletion in fetal thymic reaggregation cultures (44). We conclude that the deletion of 1H3.1 TCR-Tg thymocytes imposed by the Eα52–68-I-Ab complex assembled in vivo can be driven by distinct stromal cell types regardless of the anatomical compartment in which they reside.
“Positive” detection of CD4+CD8+ immature thymocyte deletion in 1H3.1 TCR/I-Eαd double-Tg mice
To confirm the “negative” analysis performed by using anti-Vβ6 mAb and UEA-1, we tried to “positively” visualize the spatial distribution of clonal deletion by detecting apoptosis in situ. TUNEL staining did not reveal a high level of apoptosis on 1H3.1 TCR/I-Eαd double-Tg thymic sections (not shown), most likely as a consequence of the very rapid elimination of apoptotic thymocytes by resident macrophages. Because DNA fragmentation occurs late in the apoptotic process, TUNEL staining may not be the most appropriate approach. Therefore, we chose to take advantage of the phospholipid-binding protein annexin V, which permits the detection of early changes in the membrane of cells undergoing apoptosis (45). Freshly isolated thymocytes from young mice were subjected to CD4/CD8/annexin V three-color staining. In accordance with the massive death due to neglect, a sizable fraction of immature CD4+CD8+ thymocytes were annexin V-positive in unmanipulated C57BL/6 mice (Fig. 5, top). A fairly comparable staining intensity was observed for most CD4+CD8+ thymocytes from 1H3.1 TCR/36.5 double-Tg mice, but there were more annexin V-positive cells. This was in contrast with the high annexin V staining that characterizes all of the few CD4+CD8+ thymocytes present in 1H3.1 TCR/107 double-Tg mice (Fig. 5, bottom) and indicates that they are effectively undergoing apoptosis. In 1H3.1 TCR/Igκ-Eα double-Tg mice, the annexin V staining resembled that of 1H3.1 TCR/36.5 double-Tg mice but with many more annexin Vhigh cells. These results validate the in situ analysis and clearly indicate that autoreactive immature CD4+CD8+ thymocytes can be induced to undergo apoptosis in 1H3.1 TCR/I-Eαd double-Tg mice.
Peripheral T cells from 1H3.1 TCR/I-Eαd double-Tg mice do not react to the Eα52–68-I-Ab complex
Because we do observe the presence of some Vβ6+ cells in the periphery of TCR-Tg+/I-E+ mice, we tested their functional status. This fraction, which can vary from animal to animal, consistently contains a substantial subset of CD4−CD8− cells, although this is less pronounced in the case of 1H3.1 TCR/36.5 double-Tg mice. The few CD4+ lymph node cells derived from 1H3.1 TCR/I-Eαd double-Tg mice revealed a markedly increased expression level of endogenously rearranged TCR α-chains as assessed by CD4/CD8/Vα 2, 3.2, 8, 11 staining of lymph node cells. Looking at CD4 and CD8 cells, respectively, we found 16.1% and 11.6% for a (1H3.1 TCR-Tg × B10.A (5R))F1, 24.9% and 20% for a TCR/107 double-Tg and 22.8% and 5.4% for a TCR/36.5 double-Tg as opposed to 7% and ∼2.5% for a 1H3.1 TCR-Tg/I-E− mouse. This suggested that in 1H3.1 TCR/I-Eαd double-Tg mice, Vβ6+CD4+ T cells express alternate TCR α-chains, which, presumably allow them to escape negative selection. In line with this, Vβ6+CD4+ cells were essentially lacking in the periphery of 1H3.1 TCR-Tg B10.A (5R) RAG-deficient mice (not shown). To perform functional analysis, lymph node and spleen cells from 1H3.1 TCR/I-Eαd double-Tg mice were depleted of MHC class II+ cells and CD8+ T cells. Stimulation with Y-Ae-positive (B10.A (5R)) irradiated APCs revealed that no detectable proliferation occurs, whereas identically treated cells from TCR-Tg+/I-Eαd− littermates showed a dose-dependent response (Fig. 6,B). The cultures also were negative for IL-2 production (not shown). In contrast, both types of populations were able to proliferate in response to anti-CD3ε mAb presented by irradiated C57BL/6 APCs (Fig. 6 A). This indicated that peripheral T cells from 1H3.1 TCR/I-Eαd double-Tg mice can be stimulated through their CD3 complex but are not responsive to APCs presenting the Eα52–68-I-Ab complex.
Taking advantage of mice expressing an I-Eαd transgene only on particular cell types (Table I), we recreated expression of the Y-Ae epitope in distinct cell compartments in the 1H3.1 TCR-Tg mice. This allowed us to examine in vivo the negative selection imposed by an endogenously expressed self-peptide-MHC class II complex, the in situ expression of which is well characterized.
Considering the reported poor capacity of peptide-MHC complex presentation to Th clones in vitro by cortical epithelial cells (46), the nondeletional mechanism involved in tolerization to endogenous SAG and MHC Ags expressed on radioresistant thymic stromal elements (47), the induction of anergy instead of deletion of thymocytes by MHC class I Ag or MHC class II-presented peptide expressed only on medullary epithelial cells (48, 49, 50), and the documented requirement for costimulatory molecules for optimal negative selection (6, 16, 17), one might expect not to observe major deletion of Tg thymocytes in the 1H3.1 TCR/36.5 double-Tg mice. Rather, a massive export of functionally inactivated T cells into the periphery may be predicted. However, these animals displayed a marked thymic size reduction, albeit not as drastic as in the case of 1H3.1 TCR/Igk-Eα and 1H3.1 TCR/107 double-Tg animals, and a strong reduction of the absolute number of both CD4+CD8− and CD4+CD8+ thymocytes. Thus, cortical epithelial cells can effectively mediate clonal deletion of self-peptide-self-MHC class II complex specific thymocytes in vivo. The fact that cortical epithelial cells are B7-negative (51) indicates that there is no absolute requirement for professional costimulatory molecules in negative selection of 1H3.1 TCR-Tg thymocytes. The deletion also is occurring in the medullary zone as judged by Vβ6/UEA-1 costaining. This seems to indicate a strong capacity of medullary epithelial cell in causing negative selection and is consistent with reports where medullary epithelial can support deletion of thymocytes mediated by SAG in vivo and by peptide-Ag in vitro (44, 52). Possibly, the higher efficiency of medullary epithelial cells in causing deletion is related to the fact that they can express B7 molecules (52), whereas CDR1+ cortical epithelial cells do not (51). Alternatively, this higher efficiency may be attributable to the high expression level of I-A molecules by medullary epithelial cells, which is comparable to that of BM-derived cells (41). However, a contribution of dendritic cells to deletion cannot be excluded because transfer of the Eα determinant from radio-resistant (presumably epithelial cells) to dendritic cells has been observed (51). Humblet et al. demonstrated this transfer by using isolated thymic dendritic and cortical epithelial cells from a C57BL/6 (H-2b/I-E−) → BALB/c (H-2d/I-E+) BMC to stimulate the 1H3.1 hybridoma. The purified thymic dendritic cells (in theory H-2b/I-E−) stimulated the 1H3.1 T cells, whereas cortical epithelial and spleen cells did not. The in vivo occurrence of the intercellular transfer was established by the absence of 1H3.1 reactivity when thymic stromal cells were isolated after in vitro mixing of C57BL/6 and BALB/c thymic preparations. This phenomenon is reminiscent of the intrathymic intercellular transfer observed for Mls-1a in radiation BMC (47, 53). The molecular mechanism is not known; it could involve an intercellular transfer of the Eα protein. Alternatively, it could involve a displacement of the Eα peptide itself (54), from I-Ab+ epithelial cells to I-Ab+ dendritic cells in our model and from I-Ad+ epithelial cells to I-Ab+ dendritic cells in the model of Hunt et al. because it is known that Eα52–68 can also bind to I-Ad (55).
A recent study in which the cognate peptide (OVAp) was expressed on thymic cortical (and some medullary) epithelial cells revealed a distinct possible outcome for autoreactive (OT-I TCR-Tg) thymocytes, namely editing of the TCR α-chain (56). The reason for such a difference is unclear. Besides the fact that the OT-I TCR is MHC class I (H-2Kb) restricted, whereas the 1H3.1 TCR is MHC class II restricted, one can imagine differences in the expression level of the two TCR ligands. McGargill et al. used the human K14 promoter to drive expression of the endoplasmic reticulum-targeted OVA peptide sequence and it is known that expression of a given cytokeratin is heterogeneous among the thymic epithelium (43), which contains at least three ultrastructurally distinct types of cells in its cortical zone (57). Thus, in the TAPoOT-I/K14-OVAp double-Tg thymus, some epithelial cells may assemble/present the OVAp-H-2Kb complex less efficiently than others and induce TCR editing rather than deletion. However, intrathymic deletion seems to occur as well because the thymic cellularity is reduced by >50% (56). Indeed, the expression level of the OVAp-H-2Kb complex may be low in general in these mice because when driven by the MHC class I promoter, the endoplasmic reticulum-targeted expression of OVA peptides did not detectably up-regulate the MHC class I expression level seen on TAPo APCs (58). In contrast, expression of the Eα52–68-I-Ab complex in 1H3.1 TCR/36.5 double-Tg mice may not be subjected to such a variability because the different types of cortical epithelial cells are all MHC II high (57). In any case, in sharp contrast with TAPoOT-I/K14-OVAp double-Tg mice, 1H3.1 TCR/36.5 double-Tg mice never showed signs of autoimmune disease and were not subject to premature mortality.
Perhaps the most unexpected observation from 1H3.1 TCR/Igκ-Eα double-Tg mice was the severe reduction of the absolute number of immature CD4+CD8+ thymocytes, whereas the source of Eα52–68 peptide is expressed at a subphysiological level on a large number of dendritic cells but neither in cortical epithelial cells nor in macrophages (40). This result identifies the few cortical dendritic cells as effective inducers of negative selection. Thymic dendritic cells (so-called interdigitating cells) are MHC class IIhigh and are particularly concentrated at the corticomedullary junction but also are present throughout the thymus (57). Their role in cortical deletion may have been underappreciated. In support of this idea is the astonishing stimulatory potential of dendritic cells. For instance, it is known that very few BM-derived APCs are required to induce maximal deletion of TCR-Tg thymocytes in reaggregation thymic organ cultures (59). In addition, it has been estimated that a single dendritic cell is able to activate 100-3000 T cells in a MLR (60) and it is well established that the interactions involved in negative selection are less stringent than those involved in activation of mature T cells (61, 62). Under our experimental conditions, dendritic cells appear more efficient than cortical epithelial cells at deleting CD4+CD8+ thymocytes (TCR/Igκ-Eα double-Tg mice vs TCR/36.5 double-Tg mice). This is in contrast with in vitro data indicating that cortical epithelial and dendritic cells are equally efficient at deleting self-peptide-specific CD4+CD8+ thymocytes (44).
The most dramatic reduction of the absolute number of immature CD4+CD8+ thymocytes was observed in 1H3.1 TCR/107 double-Tg mice and 1H3.1 TCR-Tg B10.A (5R) mice where all MHC class II+ cells express the deleting peptide-MHC complex. The deletion is striking when analyzed in situ. Very few Vβ6+ cells are seen in the cortex as expected based on the flow cytometry analysis. Furthermore, those few cells stain bright for Annexin V, revealing their apoptotic status. These observations directly document clonal deletion of autoreactive thymocytes by an endogenous peptide-MHC complex in the cortical compartment. Consistent with this are reports describing deletion of CD4+CD8+-Tg thymocytes caused by endogenously expressed self-peptide-MHC complexes (14, 21, 25, 63). Because the Eα peptide is synthesized and expressed in the thymus itself and is not injected systemically, we exclude the possibility that destruction of cortical thymocytes reflects an indirect deleterious effect due to activation of mature Tg T cells. The drastic cortical deletion seen in 1H3.1 TCR/107 double-Tg mice can obviously involve rare cortical APCs such as dendritic cells, as mentioned above. In contrast, cortical epithelial cells are likely to contribute to this process for two reasons. First, they do drive deletion of CD4+CD8+ cells in 1H3.1 TCR/36.5 double-Tg mice and they indeed express a higher level of I-E molecules in 107-Tg mice (41, 52). Second, purified cortical epithelial (CDR1+) cells from Y-Ae−→Y-Ae+ BMC are effectively Y-Ae+ and specifically activate the 1H3.1 hybridoma (51). Because the latter result was obtained by using cortical epithelial cells from C57BL/6→(C57BL/6 × BALB/c)F1 BMC (i.e., by using I-Ab+/−/I-Eα+/− cortical epithelial cells that do not carry any transgene), their expression level of the Eα52–68-I-Ab complex is indisputably physiologically relevant.
The pioneering studies supporting a possible deletion of thymocytes at the CD4+CD8+ stage have been questioned because of the Tg nature of the systems used. The main concern is that TCR-Tg thymocytes are capable of displaying a relatively high TCR expression level already at the earliest stages of development. Possibly, such TCR level may modify the overall avidity of the interaction with stromal cells. However, it is clear that negative selection of autoreactive cortical thymocytes can occur in mice where the timing and level of TCR expression are strictly normal. For instance, in mice where T cells are unmanipulated, tolerance to I-E molecules can be observed while they are expressed in the thymic cortex but not in the medulla (64). Perhaps the most convincing demonstration was obtained by using mice Tg for the β-chain of the MCC88–103-I-Ek complex-specific 5C.C7 αβ TCR (65). It has to be emphasized that in TCR β-chain-Tg mice thymocytes undergo a normal developmental process and do not display an early elevated expression of TCR. The CD3 complex expression by CD4−CD8− and CD4+CD8+ thymocytes is comparable to those of wild-type thymocytes, and α-chain selection occurs normally at the appropriate stage (65, 66, 67). These mice were analyzed by using MCC88–103-I-Ek tetramers which stain 95% of 5C.C7 αβ TCR-Tg thymocytes (65). Virtually all of the tetramer-positive cells from thymus and lymph node expressed the parental Vα segment and displayed the characteristic CDR3 loop-length restriction observed among other MCC/I-Ek-reactive T cells. When MCC88–103 is endogenously synthesized, thymocytes with the strongest binding to MCC/I-Ek tetramers were deleted at an early stage and more extensively than those that bind MCC/I-Ek tetramer weakly. Thus, the fact that negative selection can occur throughout the thymus and even before positive selection, indicates that deletion can take place in the cortex and is also supportive of an avidity/accessibility model of negative selection. It could be argued that in this case the endogenous expression of the relevant peptide-MHC class II complex is artificially high. However, this appears not to be the case because the thymic epithelial cells derived from this mouse do not activate 5C.C7 T cells (65). In contrast, cortical epithelial cells from C57BL/6→(C57BL/6 × BALB/c)F1 BMC synthesizing a subnatural level (I-Ab+/−, I-E+/−) of the Eα52–58-I-Ab complex can specifically activate T cells carrying the 1H3.1 TCR (51), which may not be of particularly high affinity for its cognate ligand because it is entirely coreceptor dependent (C.V. and C.A.J., unpublished data).
The presence of autoreactive CD4+ T cells in the K14-I-Aβb-Tg mice that express MHC class II molecules on the thymic cortical epithelium but nowhere else was interpreted as an indication that positive and negative selection occur in anatomically distinct sites (13), negative selection being assigned to the medulla. The occurrence and efficiency of negative selection in the medulla and at the corticomedullary junction is undisputed. However, the detection of autoreactive cells after experimental restriction of MHC class II molecule expression to the cortical epithelium does not exclude the possibility that deletion can also happen in the cortical area. For instance, cortical epithelial cells may well induce deletion of thymocytes reactive to determinants they specifically express, as is the case in 1H3.1 TCR/36.5 double-Tg mice. A key experiment would be to test whether CD4+ T cells from K14-I-Aβb-Tg mice are or are not reactive to cortical epithelial cells isolated from a C57BL/6 thymus. In addition, because of the restricted expression of MHC class II molecules to cortical epithelial cells, the K14-I-Aβb-Tg cortex lacks any MHC class II+ professional APCs. Consequently, if the fraction of BM-derived cells, especially dendritic cells, present in the cortex can mediate negative selection (as suggested above by the analysis of 1H3.1 TCR/Igκ-Eα double-Tg mice), such a phenomenon is clearly knocked out in the K14-I-Aβb-Tg thymic cortex. The phenotype of K14-I-Aβb-Tg mice demonstrates that cortical epithelial cells cannot intrinsically drive tolerance to the entire set of self-determinants expressed intrathymically. This is not incompatible with the occurrence of clonal deletion of autoreactive T cells in the thymic cortex.
Finally, our experiments were conducted by using mice expressing a Tg TCR. Therefore, whether intrathymic deletion of autoreactive T cells can occur in the cortical area of the thymus remains to be documented with animal models where the essential parameters follow an unperturbated physiological expression level, i.e., are not based on transgenesis.
In conclusion, these observations indicate that the clonal deletion of autoreactive TCR-Tg thymocytes imposed in vivo by an endogenously assembled self-peptide-MHC class II complex can occur in the medullary as well as in the cortical compartments. Both epithelial and dendritic cells were found to be able to efficiently drive deletion of cortical thymocytes. The data are most consistent with a model where, in addition to the thymocyte/stromal cell interaction avidity, negative selection is largely determined by accessibility to self-determinants regardless of their anatomical distribution. The involvement of multiple stromal cell types in negative selection may help to minimize the chances of autoreactive T cell escape.
We thank Dr. K. Bottomly (Yale University School of Medicine, New Haven, CT) for the B10.A (5R) RAG-2-deficient mice, Dr. R. Flavell (Howard Hughes Medical Institute, Yale University, New Haven, CT) for the 107, 36.5, and Igκ-Eα-Tg lines, and Drs. D. Mathis and C. Benoist (Institut National de la Santé et de la Recherche Médicale-Centre National de la Recherche Scientifique-Universite Louis Pasteur, Strasbourg, France) for providing the pTα and pTβ cassette used to generate 1H3.1 TCR-Tg mice. We also thank Charles Annicelli for help with animal care. C.A.J. is an investigator of the Howard Hughes Medical Institute.
This study supported in part by the Howard Hughes Medical Institute (Grant AI-14579 to C.A.J.).
Abbreviations used in this paper: BMC, bone marrow chimeras; K14, keratin-14 promoter; BM, bone marrow; Tg, transgenic; SAG, superantigen; 5R, B10.A-H2i5 H2-Tl8a(5R)/SgSnJ mice.