Thymic Medullary Epithelial Cell Differentiation, Thymocyte Emigration, and the Control of Autoimmunity Require Lympho–Epithelial Cross Talk via LTβR

The thymus is the primary lymphoid organ that generates a diverse, nonautoreactive T cell repertoire. Maturation of T lymphocytes is accompanied by a highly ordered migration pattern in the thymus (1, 2). During this movement, developing T lymphocytes constantly interact with a dense network of epithelial reticular cells. Thymic epithelial cells are organized in a three-dimensional network rather than as a sheet of cells on a basement membrane (3). The interaction between T cell receptors on immature thymocytes and peptide–MHC complexes expressed by stromal cells ultimately determines the fate of developing T cells. Specifically, epithelial cells in the cortex mediate positive selection whereas bone marrow–derived dendritic cells and medullary epithelial cells (MECs) mediate negative selection of developing thymocytes (4). Recent work has demonstrated an important role for MECs in the prevention of autoimmunity (5–8). MECs express tissue-specific antigens such as thyroglobulin, insulin, renin, proteolipid protein, and myelin oligodendrocyte glycoprotein in a promiscuous fashion presumably to present the entire immunological self in the thymus for repertoire selection. Mutations in the aire gene, so far the only known regulator of promiscuous gene expression in MECs, lead to the autoimmune polyendocrine syndrome type I in man and to the lack of a subset of promiscuous transcripts with organ-specific autoimmunity in mice (8). Thus, MECs may play a crucial role in the control of autoimmunity by the promiscuous expression of tissue-specific antigens that are thought to negatively select autoreactive T cell clones and possibly to promote the generation of regulatory T cells. How MECs develop and acquire their specific differentiation in three-dimensional space to achieve this control of autoimmunity is poorly understood. The aforementioned interaction between lymphocytes and epithelial cells is reciprocal and, thus, also determines the fate of thymic epithelial cells. In the absence of mature thymocytes (e.g., in SCID or RAG-deficient mice), the thymus has a markedly underdeveloped medulla, which is restored to wild-type configuration after transfer of wild-type bone marrow (9, 10). However, the molecular basis for lympho–epithelial cross talk that governs the differentiation of epithelial cells in the thymus is unknown.

The large family of TNF receptors is involved in such diverse biological processes as inflammation, apoptosis, autoimmunity, and lymphoid organogenesis (11). During organogenesis of peripheral lymphoid organs, the lymphotoxin β receptor (LTβR) is absolutely required for the development of Peyer's patches, peripheral lymph nodes, and gut-associated lymphoid tissues (i.e., all peripheral lymphoid organs apart from the spleen; reference 12). In addition, LTβR, which is expressed in stromal cells but not T and B cells, is required for the organization of microenvironments within the spleen as well as the germinal center reaction (13). Two ligands have been described previously for LTβR, namely LTα1β2 and LIGHT (14, 15). LTα1β2 consists of one LTα and two LTβ subunits and is crucial for lymphoid organogenesis. The absence of either subunit, LTα or LTβ, leads to the lack of most peripheral lymph nodes and a disturbed splenic architecture in mutant mice (16–18). Lack of the recently identified homotrimeric LTβR ligand LIGHT does not impair lymphoid organogenesis (19). However, on an LTβ-deficient background (i.e., in LTβ^−/− LIGHT^−/− mice), the absence of LIGHT enhances the phenotype in that most LTβ^−/− LIGHT^−/− mice now lack mesenteric lymph nodes, which are present in all LTβ^−/− LIGHT^+/− mice (19). Concerning the signal transduction of LTβR, a spontaneously occurring point mutation in the nuclear factor–κB-inducing kinase (Nik), a MAP3K-related kinase, has been shown to specifically affect LTβR but not TNF receptor–mediated NF-κB activation (20). This point mutation in the Nik gene is the defect in aly/aly mice that lacks Peyer's patches and lymph nodes and shows a disturbed splenic and thymic microarchitecture (20, 21). Although the exact role of the Nik point mutation in aly/aly mice is unclear, complete removal of Nik leads to a strikingly similar phenotype (22). Despite the obvious similarities between LTβR^−/− and aly/aly mice, the aly/aly phenotype appears to be more complex than can be explained by disturbed LTβR signaling alone. It has been shown for example that aly/aly lymphocytes, i.e., cells which do not express LTβR, are unable to home to the gut-associated lymphoid tissues (23).

In this paper, we characterize the role of the LTβR signaling pathway for lympho–epithelial cross talk in the thymic medulla. Mice deficient for LTβR ligand, LTβR itself, and mice carrying a point mutation in the signaling component Nik all show abnormally differentiated MECs. LTβR signaling is critically required for proper differentiation and normal steady-state numbers of MECs. Functionally, the absence of LTβR signaling leads to the retention of mature T cells in the thymic medulla and is associated with autoantibody production, which may be a result of incomplete central tolerance induction by MECs.

C.B-17 scid/scid (SCID), RAG2^−/−, Balb/c, C57BL/6, C57BL/6-Ly5.1, LTβR^−/−, and aly/aly mice were kept under specific pathogen-free conditions in the mouse facility of the Max Planck Institute for Immunobiology. LTβR^−/− mice had been backcrossed to C57/BL6 at least six times and were genotyped as described previously (12). aly/aly mice had been backcrossed to Balb/c and were genotyped by PCR plus sequencing, using the primers 5′-TCGAAGGCCTCTCAGAGCTCACG-3′ and 5′-CTCTGTTCTGGGAACCCTTGG-3′. LTβ ^−/−, LIGHT^+/−, and LTβ^−/− LIGHT^−/− mice were generated and genotyped as described previously (19). Unless stated otherwise, age-matched 4–8-wk-old mice were used for experiments.

5-μm frozen sections were fixed in a mixture of 75% acetone and 25% methanol for 15 min at 4°C, incubated with PBS/1% H₂O₂ at room temperature for 15 min, washed, blocked with PBS/0.5% BSA, and incubated with the respective antibody solution overnight. The following antibodies and reagents were used: ulex europaeus agglutinin 1 (UEA-1; Vector Laboratories), Troma-1 (anti-cytokeratin 8, a gift by R. Kemler [Max Planck Institute for Immunobiology, Freiburg, Germany]), MTS10 (BD Biosciences), polyclonal rabbit antiserum to I-O (K535, a gift by C. Surh [Scripps Research Institute, La Jolla, CA]), F4/80 (Serotec), MIDC-8 (Serotec), and CD45.1 (BD Biosciences). Staining was developed directly or after incubation with an appropriate biotinylated second step antibody using the StreptABComplex/HRP according to the manufacturer's recommendations (DakoCytomation). For the detection of autoantibodies, frozen sections from RAG2^−/− mice were processed as described earlier in this paragraph, sera were diluted 1:40, and an anti–mouse Ig–FITC conjugate (Southern Biotechnology Associates, Inc.) was used for visualization on a TCS SP2 confocal microscope system (Leica). Sera from MRL.lpr/lpr mice were used as a positive control.

Mice before and after transfer of bone marrow cells were maintained in individually ventilated cages containing sterilized food and water. SCID and RAG2^−/− mice were irradiated sublethally with 350 rad, LTβR^−/− and aly/aly mice were irradiated sublethally with 450 rad, and irradiated mice were injected i.v. 24 h later with 2 × 10⁶ erythrocyte-lysed bone marrow cells of the appropriate genetic background. Thymi of bone marrow chimeras were analyzed 6–8 wk after bone marrow transfer, and engraftment was confirmed either by the detection of mature CD4 and CD8 single-positive thymocytes by FACS^® or by the detection of thymocytes positive for CD45.1(Ly5.1) by FACS^® or immunohistochemistry.

C57BL/6 mice, two in each group, received weekly intraperitoneal injections of 100 μg LTβR-Fc or 100 μg human polyclonal IgG preparation as control (Sandoglobulin; Novartis), both a gift by J. Browning (Biogen, Cambridge, MA). Thymi of injected mice were harvested after 3 and 6 wk, respectively.

To measure the surface area of thymic epithelial cells, UEA-1–stained frozen sections were prepared from littermates with different genotypes on one and the same slide. High resolution pictures were obtained with a CCD camera mounted on an Axioplan2 microscope (original magnification 20; Carl Zeiss MicroImaging, Inc.). From these pictures, medullary areas of 6.25 × 10⁻⁸ m² were extracted, and the UEA-1–stained brown pixels were counted using the public domain National Institutes of Health (NIH) Image program (written by W. Rasband at NIH, Bethesda, MD). To quantify the number of nuclei of medullary thymic epithelial cells in wild-type versus LTβR^−/− mice, high magnification images were obtained using the described setup, and nuclei were counted by eye. Nuclei of medullary thymic epithelial cells were defined as completely surrounded by UEA-1 staining, large, and weakly hematoxylin-positive when compared with nuclei of thymocytes. Quantification of UEA-1–positive epithelial cells by FACS^® was done by digesting thymic stroma after mechanically removing free thymocytes in 5 ml RPMI 1640 containing 2% FCS, 20 mM Hepes, 0.4 mg/ml collagenase/dispase (Roche), and 25 μg/ml DNase I (Roche) at 37°C for 15 min. Cells in the supernatant were collected, and digestion was repeated two more times. A representative sample from each thymus of four LTβR^−/− and four wild-type mice was analyzed, and the total number of CD45-negative CDR1-negative cells expressing high levels of UEA-1 per thymus was calculated. This staining protocol was adopted because cortical, CDR1-positive epithelial cells, and other nonepithelial cells have been shown to express low levels of UEA-1 (24).

FACS^® analyses were performed on freshly prepared, erythrocyte-lysed single cell suspensions of thymocytes in PBS/0.3% BSA/5 mM EDTA. Stainings were analyzed on a FACSCalibur™ machine (BD Biosciences) using the following: FITC-, PE-, allophycocyanin- or biotin-conjugated monoclonal antibodies purchased from BD Biosciences (clone name given in parenthesis): CD4 (GK1.5), CD8 (53–6.7), CD3ε (145–2C11), TCR-β (H57–597), CD62L (MEL-14), CD69 (H1.2F3), CD45 (30-F11), Ly5.1 (A20), B7.1 (16–10A1), and CD45 (30F11). The monoclonal antibody G8.8 (a gift by B. Kyewski [German Cancer Research Center, Heidelberg, Germany]) was conjugated to Alexa Fluor 647 according to the manufacturer's recommendations (Molecular Probes). The unlabeled monoclonal antibodies CDR1 (a gift by B. Kyewski) and M293 (anti–integrin β7; BD Biosciences) were visualized with a FITC-labeled goat anti–rat antiserum. The biotin label was visualized with streptavidin-Red670 (GIBCO BRL). For stainings with LTβR-Fc, a modified protocol was used to avoid background staining; stainings were performed in PBS/2% normal mouse serum. Thymocytes were preincubated with unlabeled monoclonal antibody to CD16/CD32 (2.4G2; provided by M. Lamers [Max Planck Institute for Immunobiology, Freiburg, Germany]) at 1 μg/10⁶ cells for 10 min on ice, washed, and incubated with LTβR-Fc at 1 μg/10⁶ cells for at least 45 min on ice. LTβR-Fc staining was visualized using a PE-conjugated donkey anti–human IgG F(ab′)₂ fragment (Jackson ImmunoResearch Laboratories). Sorting of MECs for RT-PCR was done largely as described previously (5). In brief, thymi of five mice (4–8-wk-old) were digested by three rounds of collagenase/dispase followed by two rounds of trypsin to free epithelial cells. Cells in suspension were labeled and sorted on a high-speed FACS^® (MoFlo; DakoCytomation) to a purity >95%. RT-PCR was performed on RNA equivalent to 30,000 MECs as described previously (25) using primers reported previously (6).

For 5-bromo-2′deoxyuridine (BrdU) labeling, 12-wk-old LTβR^−/− and wild-type control mice were injected with 0.6 mg BrdU in PBS i.p. on day 0 and kept on drinking water containing 1 mg/ml BrdU. Thymocytes were sorted by FACS^® on day 3, and each population was stained separately as described previously (26). For the detection of reentry of peripheral lymphocytes to the thymus, 3 × 10⁷ lymphocytes isolated from lymph nodes and spleen of C57BL/6-Ly5.1 mice were i.v. injected into LTβR^−/− and wild-type control mice. Thymocytes of these mice were analyzed 16 h later for Ly5.1 expression.

Immunohistochemistry reveals the presence of dendritic cells and macrophages in the thymic medulla of LTβR^−/− and aly/aly mice as shown on Fig. S1. The defect causing abnormal differentiation of MECs in aly/aly mice lies in the stroma compartment. Fig. S2 shows low magnification images of UEA-1–stained thymus sections of the indicated aly/aly bone marrow chimeras.

To investigate a possible role for LTβR signaling in thymic development, we stained sections from LTβR^−/− and aly/aly mice with markers for medullary and cortical epithelial cells (Fig. 1 A). Cortical and medullary regions can be found in the thymi of these mutant mice with the same ratio observed in wild-type mice. However, although cortical epithelial cells marked by anticytokeratin 8 staining appeared unaffected, the normal configuration of MECs is lost in LTβR^−/− and aly/aly mice. In wild-type mice, UEA-1–positive epithelial cells form a homogeneous, three-dimensional network of stellate cells that builds the thymic medulla (Fig. 1 A, middle). Although UEA-1–positive cells can still be detected in LTβR^−/− and aly/aly mice, these cells form clumps rather than a network (Fig. 1 A). FACS^® analyses confirmed the epithelial nature of UEA-1–positive cells from LTβR^−/− thymi as they are large G8.8 positive cells that lack CD45 and CDR1 expression (unpublished data). MECs staining positive for UEA-1 have been shown previously to represent one of two subsets that make up the medullary epithelial stroma (27). The other subset is characterized by the expression of the nonpolymorphic MHC class II antigen I-O, and this subset is similarly affected in LTβR^−/− and aly/aly mice when compared with wild-type mice (Fig. 1 B). Apart from epithelial cells, the stroma of the thymic medulla consists of dendritic cells, macrophages, and poorly defined cells of nonepithelial origin that are cytokeratin-negative but recognized by the monoclonal antibody MTS-10, which also stains epithelial cells (28). Dendritic cells and macrophages are readily detectable in LTβR^−/− and aly/aly mice by MIDC-8 and F4/80 staining, respectively (Fig. S1). MTS-10 staining is strongly reduced in aly/aly mice but only mildly affected in LTβR^−/− mice (Fig. 1 B). Although LTβR is expressed in stromal but not lymphoid cells (29–31), Nik has been shown to be ubiquitously expressed (32), and there is evidence that the aly/aly mutation affects cells which do not express LTβR (e.g., lymphocytes; reference 23). Thus, MTS10-positive cells of nonepithelial origin may be unaffected by the absence of LTβR but depend on the presence of undisturbed signaling through Nik. In hematoxylin and eosin stainings of thymus sections, the demarcation between cortex and medulla is blurred in LTβR^−/− and aly/aly mice due to a significant increase in the density of strongly hematoxylin-positive nuclei (i.e., lymphocytes) in the medulla (Fig. 1 B). Collectively, the data reveal an aberrant development of thymic MECs in mice defective for LTβR signaling that affects UEA-1 as well as I-O–positive MECs.

To locate the defect of the aberrant development of the thymic MECs, we generated bone marrow chimeras. The transfer of Ly5.1-marked, wild-type bone marrow did not restore the normal three-dimensional epithelial network in LTβR^−/− mice, even though Ly5.1-marked thymocytes readily repopulated the thymic medulla (Fig. 2 A). Likewise, transfer of wild-type bone marrow into aly/aly mice could not rescue defective MEC development (Fig. 2 A; and Fig. S2). However, the latter experiment was done without the congenic Ly5.1 marker so that we could not formally demonstrate the presence of wild-type cells in the thymus of these bone marrow chimeras. Therefore, we transferred aly/aly bone marrow into SCID mice to test if aly/aly thymocytes induced MEC differentiation in these mice. The transfer of aly/aly bone marrow into SCID mice completely restored the configuration of MECs in three-dimensional space (Fig. 2 A; and Fig. S2). The same is true for the transfer of LTβR^−/− bone marrow into RAG2-deficient mice. We conclude from these experiments that the defect in the development of UEA-1 as well as I-O–positive MECs in LTβR^−/− and aly/aly mice lies in the stromal compartment.To investigate if LTβR signaling influenced cell proliferation, which affected the steady-state number of MECs in addition to their differentiation in three-dimensional space, we determined the number of MECs in LTβR^−/− versus wild-type thymi in three different ways. First, MECs were counted as large cells with lightly hematoxylin-positive nuclei completely surrounded by UEA-1 staining on high-power magnifications. They are significantly reduced in LTβR^−/− mice (Fig. 2 B). Second, we counted cells expressing high levels of UEA-1 by FACS^® analysis, excluding hematopoietic cells (as identified by CD45) and cortical epithelial cells (as identified by CDR1) in preparations of individual thymi (Fig. 2 C). By this method, thymi of LTβR^−/− mice contained three times fewer cells than wild-type control thymi. Third, another staining protocol for MECs was used. The absolute cell number of CD45⁻G8.8⁺CDR1⁻B7.1⁺ MECs that could be isolated from pools of five LTβR^−/− versus five wild-type thymi was determined by FACS^® sorting. MECs defined by these markers were reduced in LTβR^−/− mice sixfold compared with controls (Fig. 2 D). In contrast, total thymocyte numbers did not differ significantly between mutant mice and appropriate controls. Collectively, our data demonstrate that in LTβR^−/− mice MECs identified by the independent markers UEA-1 and I-O show aberrant differentiation. At the same time, UEA-1^hi CD45⁻CDR1⁻ and even more pronounced CD45⁻G8.8⁺ CDR1⁻B7.1⁺ MECs are significantly reduced in absolute numbers per thymus in LTβR^−/− mice compared with controls.

To identify the cells that express LTβR ligands, we performed FACS^® analyses on thymocytes using a soluble LTβR-Fc fusion protein to stain ligand-positive cells. We found that neither wild-type nor LTβR^+/− thymocytes bound LTβR-Fc (Fig. 3 and not depicted). On the other hand, lymph node B lymphocytes from these mice showed low but significant levels of LTβR ligands as described previously (29, 33). To our surprise, mature CD4 or CD8 single-positive thymocytes of LTβR^−/− mice showed easily detectable levels of LTβR ligand expression, whereas CD4 and CD8 double-positive cells and the majority of coreceptor- and TCR-negative thymocytes showed no staining (Fig. 3). Thus, in the absence of LTβR on stromal cells, mature thymocytes appear to up-regulate LTβR ligands to supraphysiological levels. This overexpression of LTβR ligands in LTβR^−/− mice suggests the presence of a regulatory loop that modulates LTβR ligand expression. In contrast, LTβR ligands could not be detected on the surface of aly/aly thymocytes (Fig. 3). Although the situation is complicated by the fact that Nik is ubiquitously expressed, there is the possibility that the direct interaction between ligand(s) and receptor induces the down-regulation of ligand(s) on thymocytes.

The results presented thus far identify the LTβR signaling pathway as a crucial determinant in MEC development. Next, we investigated thymi of mice rendered deficient for LTα1β2, a LTβR ligand that is crucial for lymphoid organogenesis (16–18). Thymi of LTβ^−/− mice consistently show a loss of the homogeneous distribution of UEA-1–positive cellular extensions covering the entire thymic medulla (Fig. 4 A). Clumps of strongly UEA-1–staining cells are found next to medullary areas entirely devoid of UEA-1 staining. Despite this severe morphological impairment, the overall surface area, as a measurement for MEC functionality, provided by UEA-1–positive cells in the thymic medulla of LTβ^−/− mice was not significantly reduced (Fig. 4 B). In contrast, the surface area of UEA-1–positive MECs in LTβR^−/− mice was significantly smaller than in control mice. This difference between receptor and ligand knockouts could be explained if another LTβR ligand, such as LIGHT (15), that is dispensable by itself for lymphoid organogenesis compensated for the absence of LTα1β2 in LTβ^−/− mice. However, this is not the case as LTβ^−/− LIGHT^−/− thymi are indistinguishable from LTβ^−/− LIGHT^+/− thymi (Fig. 4, A and B). These results demonstrate that proper differentiation of UEA-1–positive MECs depends on the presence of LTβ and suggest the existence of (an) additional LTβR ligand(s) apart from LTα1β2 and LIGHT that signal through LTβR in the absence of the two known ligands.

To investigate lympho–epithelial cross talk in a normal adult mouse, we injected wild-type mice once a week with 100 μg LTβR-Fc, a fusion protein of LTβR and human IgG that blocks interaction of ligands with the LTβR in vivo (34). Three injections were sufficient to completely abolish the physiological configuration of UEA-1–positive thymic epithelial cells (Fig. 5 A) and to significantly reduce the detectable surface area of these cells (Fig. 5 B). Six injections did not have an additional effect. UEA-1-positive cells were still detectable and showed the same aberrant phenotype found in LTβR^−/− mice (Fig. 1 A). Control injections with human immunoglobulin did not affect UEA-1–positive MEC organization. From this experiment, we conclude that thymic MECs crucially depend on the continuous presence of signals mediated by LTβR.

We investigated the basis for the increased density of medullary thymocytes found on hematoxylin and eosin stains in LTβR^−/− and aly/aly mice. By fluorescence immunohistochemistry, we found that double-positive and single-positive thymocytes were correctly located to cortical and medullary regions, respectively, which ruled out the possibility that aberrant thymocyte homing within the thymus leads to increased medullary cellularity (unpublished data). Therefore, we analyzed mature thymocytes by FACS^® analysis, which revealed a mild increase in single-positive T lymphocytes in LTβR^−/− and aly/aly mice compared with heterozygous controls (Fig. 6, A and B). This led us to investigate in more detail the population of thymocytes in the process of leaving the thymus. These cells carry the characteristic CD69-negative CD62L-positive integrin β7-positive phenotype of recent thymic emigrants (RTEs; reference 35). In FACS^® analyses, CD4 single-positive thymocytes from 12-wk-old LTβR^−/− mice showed a marked shift toward the CD69-negative CD62L-positive integrin β7-positive RTE phenotype compared with heterozygous controls (Fig. 6 A), which translated into an eightfold increase in the percentage of CD4 single-positive CD69-negative CD62L-positive RTEs in LTβR^−/− mice compared with heterozygous control thymocytes (4.33 ± 0.60% vs. 0.52 ± 0.19% of total thymocytes; P < 0.0001, Student's t test). Consistently, CD8 single-positive LTβR^−/− thymocytes showed a 4.7-fold increase in the percentage of CD8 single-positive CD69-negative integrin β7-positive RTEs (0.91 ± 0.29% vs. 0.19 ± 0.10%; P < 0.005; Fig. 6 B). Similar results were obtained by comparison of aly/aly mutant and aly/+ heterozygous controls; the observed increase for aly/aly mice is 14.9-fold for CD4 single-positive CD69-negative CD62L-positive RTEs and 10.7-fold for CD8 single-positive CD69-negative integrin β7-positive RTEs. The marked increase in RTEs together with only a mild increase in less mature stages of single-positive thymocytes suggested a disturbed export of thymocytes from the thymus. Consistently, the transfer of 3 × 10⁷ Ly5.1-marked lymphocytes into LTβR^−/− mice did not reveal an increased reentry of peripheral lymphocytes into the thymus when compared with wild-type mice (<0.01% Ly5.1⁺ cells among both LTβR^−/− and wild-type thymocytes in transferred mice). To rule out the possibility that enhanced proliferation rates of RTEs rather than their impaired export caused the observed increase in RTEs, we measured proliferation of double-positive, CD4 single-positive, and CD4 single-positive CD62L-positive thymocytes in LTβR^−/− mice and controls by in vivo BrdU labeling studies (Fig. 6 C). Mature thymocytes of mutant mice showed the same if not a lower propensity for BrdU uptake compared with control cells, ruling out abnormally high proliferation of RTEs in LTβR-deficient mice.

Finally, we asked if the disorganization and absolute reduction of MECs in LTβR^−/− mice was associated with autoimmune processes. There is accumulating evidence that MECs play a crucial role in the control of autoimmunity by the promiscuous expression of tissue-specific antigens that are thought to negatively select autoreactive T cell clones and possibly to promote the generation of regulatory T cells (6, 7). To investigate promiscuous gene expression in LTβR^−/− mice, we performed RT-PCR on RNA isolated from MECs of LTβR^−/− and wild-type mice identified as large CD45⁻G8.8⁺CDR1⁻B7.1⁺ cells by FACS^® analysis. This approach revealed that promiscuous gene expression and the expression of AIRE, so far the only known regulator of promiscuous gene expression (8), is unaffected in LTβR^−/− MECs (Fig. 7 A). Because only 1–5% of MECs express a given tissue-specific antigen (6), central tolerance induction to these promiscuously expressed antigens has been suggested to require extensive interaction of mature lymphocytes with a large number of MECs (7). Given the disturbed three-dimensional organization and the absolute reduction of MECs in LTβR^−/− mice, we hypothesized that interaction between lymphocytes and MECs should be grossly disturbed in these mice and, therefore, we sought signs of autoimmunity. Although LTβR^−/− mice develop perivascular lymphocytic infiltrates in lungs, liver, pancreas, salivary gland, adrenal gland, and kidney (12), these infiltrates could be due to homeostatic rather than autoimmune processes considering that these mice lack lymph nodes. More conclusively, we detect autoreactive antibodies in sera of LTβR^−/− but not wild-type mice directed against salivary gland, stomach, and pancreas (Fig. 7 B). Autoreactive antibodies to liver, adrenal gland, and kidney where lymphocytic infiltrations can also be found were not observed. Older mice appear to be frequently affected: all sera obtained from four 6-mo-old LTβR^−/− mice contained autoantibodies recognizing determinants of at least two of these organs whereas control sera remained negative (Table I). Antibodies directed against dsDNA or immunoglobulin in the sera of LTβR^−/− mice were not detected. These findings suggest that promiscuous expression of tissue-specific genes may not be sufficient to prevent autoimmune processes and that proper three-dimensional organization in conjunction with a critical number of MECs may also be required.

Thymic epithelial cells have long been recognized for their ability to generate a diverse, nonautoreactive T cell repertoire by the interaction with developing thymocytes. This reciprocal lympho–epithelial interaction is a unique feature of thymopoiesis. Interaction of T cell receptors and peptide–MHC complexes represents the classical example of an interaction that determines the developmental fate of thymocytes. Here, we provide to our knowledge the first example of how lymphocytes determine the development of thymic MECs that are thought to control autoimmunity. Although transcription factors (e.g., Foxn1 [36] and Hoxa3 [37]), signaling molecules (e.g., RelB; reference 38, 39), and the Wnt proteins (40) have been implicated in thymus development, none of these have been shown to depend on activation by developing lymphocytes in vivo. Indeed, Wnt proteins, which are produced by thymocytes and even more abundantly by thymic epithelial cells, have been so far the only soluble factors influencing epithelial cell differentiation. Our paper identifies the LTβR signaling axis as an important means of communication between developing lymphocytes and epithelial cells in the thymic medulla. The absence of any of its components, namely ligand (i.e., LTα1β2), receptor, or intracellular signaling molecule (i.e., Nik), leads to the disorganization of MECs. Functionally, malformed MECs result in the retention of mature thymocytes and are associated with autoantibody production, suggesting key roles for LTβR signaling in thymic lymphocyte homeostasis and central tolerance induction.

The requirement of instructive signals for thymic epithelial cell differentiation from developing lymphocytes was suggested by the phenotype of mice with defective T cell development (3, 9, 41), but the nature of these signals has remained elusive. Our results now show that defective lympho–epithelial cross talk in LTβR^−/− mice specifically affects medullary but not cortical epithelial cells. Thus, other signals must exist which organize the three-dimensional network in the cortex. This observation is in good agreement with the fact that supra-physiological levels of LTβR ligands were found in LTβR^−/− mice on mature single-positive (situated in the medulla) but not on double-positive thymocytes (situated in the cortex). More work is required to determine the molecular basis for increased ligand expression on LTβR^−/− thymocytes.

The medullary stroma of the thymus consists of a complex mixture of epithelial cells, interdigitating dendritic cells, macrophages, and fibroblastoid cells. Although a series of markers exist that are MEC-specific, one has to consider that any one may only stain a certain MEC subset. UEA-1 and I-O expressed on mutually exclusive MEC subsets in the adult mouse thymus have been shown convincingly to mark the majority of epithelial cells in the thymic medulla when used in combination (27). Although we cannot rule out the possibility that a minor subset of MECs remains unaffected by the lack of LTβR signaling, our data unequivocally demonstrate disturbed differentiation of UEA-1 as well as I-O-positive MECs. Both of these markers are also detectable on thymic epithelial cells during embryogenesis, which should have allowed the identification of MECs in our experiments even if they remained in an embryonic differentiation state. Studies are under way to determine if MECs deficient for LTβR signaling are locked in their embryonic form.

Our paper shows that the marked reduction in UEA-1 staining for MECs in LTβR^−/− mice is caused by both the loss of the characteristic three-dimensional organization and the reduction in absolute numbers of epithelial cells. The maintenance of this three-dimensional structure in the thymic medulla requires the continuous signal transduction through LTβR as demonstrated by the loss of the medullary epithelial network after three LTβR-Fc injections. This rapid regression from a three-dimensional organization to isolated clusters of cells is an unexpected finding. The view of thymic stroma as a static, radio-resistant scaffold for developing thymocytes that, once it has reached its ultimate state of differentiation, remains unchanged may have to be revised. Additional support for the proliferative effect of LTβR signaling comes from analyses of mice that are transgenic for LIGHT under the control of the proximal lck promotor (42). These mice show a significant enlargement of the thymic medulla with only small cortical areas. Although the situation is complicated by the fact that LIGHT binds the herpes virus entry mediator expressed on lymphocytes (15) in addition to LTβR, this may suggest that overexpression of LTβR ligands leads to increased numbers of MECs and an enlarged medulla.

Concerning the LTβR ligands involved in organizing the thymic medulla, our results show that the phenotype of the LTβR^−/− mouse is more severe than that of the LTβ^−/−LIGHT^−/− mouse that lacks both known LTβR ligands. This would indicate that additional, so far unknown ligand(s) partially compensate for the absence of LTβ and LIGHT in the thymic medulla. A recent analysis of the development of Peyer's patches, peripheral, and mesenteric lymph nodes in LTβ^−/−LIGHT^−/− mice also hinted at additional ligand(s) of the LTβR (19). Because not all functional LTβR ligands are known, the transfer of LTβ^−/−LIGHT^−/− bone marrow into SCID mice (to demonstrate the inability of LTβR ligand–deficient thymocytes to induce normal medullary architecture in SCID mice) will yield inconclusive results; LTβ^−/−LIGHT^−/− bone marrow would still be able to induce differentiation of MECs in SCID mice as can be observed in LTβ^−/− LIGHT^−/− mice.

In the absence of LTβR signaling, malformation of MECs leads to increased numbers of the most mature stages of thymopoiesis, suggesting that these cells are retained in the medulla rather than efficiently exported. This view is supported by the fact that abnormal proliferation of mature T cells or reentry of peripheral lymphocytes to the thymus was not observed in LTβR^−/− mice. This may mean that mature T cells ready to leave the thymus require signals from MECs for emigration that are only inefficiently supplied by malformed MECs in LTβR^−/− mice. Indeed, a checkpoint may exist that allows thymocyte emigration only when extensive interaction with MECs has been completed to prevent autoreactive clones from escaping negative selection. The need for extensive lympho–epithelial interaction in the thymic medulla appears to be even more important keeping in mind that only 1–5% of MECs express a given tissue-specific antigen, presumably necessary to prevent organ specific autoimmunity in a mosaic pattern (6); if each T cell had to interact with at least one MEC expressing a particular tissue-specific antigen to guarantee central tolerance, many cell–cell contacts would have to be made. From our data, it is obvious that such lympho–epithelial interaction in the medulla cannot take place normally in LTβR^−/− mice, which may result in autoimmunity, specifically in lymphocytic infiltrations and autoantibody production. However, our data by no means prove that MEC malformation leads to autoimmunity. They simply reveal a correlation between disturbed lympho–epithelial cross talk and autoimmune processes, and further work will be necessary to investigate this issue in more detail. Preliminary results from Western blot analyses using sera of LTβR^−/− mice on whole tissue lysates show reactivity to multiple proteins, suggesting a polyclonal set of autoantibodies. It is interesting to note that we detect autoantibodies in LTβR^−/− mice that lack peripheral lymph nodes, whereas priming of autoreactive T cells to autoantigens is thought to occur primarily in local lymphoid tissues (43).

Intriguingly, promiscuous expression of tissue-specific antigens in MECs appears to be unaffected in LTβR^−/− mice and does not depend on lympho–epithelial interaction. In situ hybridization studies suggested initially that AIRE, so far the only known regulator of promiscuous gene expression, was undetectable in MECs of LTβR^−/− mice (unpublished data); because AIRE detection via in situ hybridization is not very sensitive and may have been complicated by the aberrant phenotype of MECs in LTβR^−/− mice, we investigated this issue by a different, more sensitive procedure. RT-PCR analyses using sorted CD45⁻G8.8⁺CDR1⁻B7.1⁺ MECs demonstrate unequivocally that on a per cell basis, LTβR^−/− MECs express normal levels of AIRE. Thus, autoimmunity occurs in LTβR^−/− mice, whereas promiscuous gene expression is unaffected. This suggests an important role for the three-dimensional network and the absolute number of MECs and all the factors necessary to maintain these parameters, such as LTβR^−/− signaling, to prevent autoimmunity.

In summary, our data identify the LTβR signaling axis as an important pathway that governs the three-dimensional configuration of MECs via lympho–epithelial cross talk. We describe the ligand, the receptor, and a signaling mediator that are crucial components in this signaling axis and provide evidence for the importance of the proper differentiation of MECs for thymocyte export and possibly the prevention of autoimmunity.

We are indebted to S. Gross and C. Sainz-Rueda for excellent technical support. We would like to thank B. Boehm for advice on autoantibodies.

This work was supported in part by a grant of the Deutsche Forschungsgemeinschaft to C.C. Bleul (Bl397/2-3) and to K. Pfeffer (Pf259/2-5/6).