NF-κB2 Is Required for the Control of Autoimmunity by Regulating the Development of Medullary Thymic Epithelial Cells*

Medullary thymic epithelial cells function as antigen-presenting cells in negative selection of self-reactive T cell clones, a process essential for the establishment of central self-tolerance. These cells mirror peripheral tissues through promiscuous expression of a diverse set of tissue-restricted self-antigens. The genes and signaling pathways that regulate the development of medullary thymic epithelial cells are not fully understood. Here we show that mice deficient in NF-κB2, a member of the NF-κB family, display a marked reduction in the number of mature medullary thymic epithelial cells that express CD80 and bind the lectin Ulex europaeus agglutinin-1, leading to a significant decrease in the extent of promiscuous gene expression in the thymus of NF-κB2-/- mice. Moreover, NF-κB2-/- mice manifest autoimmunity characterized by multiorgan infiltration of activated T cells and high levels of autoantibodies to multiple organs. A subpopulation of the mice also develops immune complex glomerulonephritis. These findings identify a physiological function of NF-κB2 in the development of medullary thymic epithelial cells and, thus, the control of self-tolerance induction.


Medullary thymic epithelial cells function as antigen-presenting cells in negative selection of self-reactive T cell clones, a process essential for the establishment of central self-tolerance.
These cells mirror peripheral tissues through promiscuous expression of a diverse set of tissue-restricted self-antigens. The genes and signaling pathways that regulate the development of medullary thymic epithelial cells are not fully understood. Here we show that mice deficient in NF-B2, a member of the NF-B family, display a marked reduction in the number of mature medullary thymic epithelial cells that express CD80 and bind the lectin Ulex europaeus agglutinin-1, leading to a significant decrease in the extent of promiscuous gene expression in the thymus of NF-B2 ؊/؊ mice. Moreover, NF-B2 ؊/؊ mice manifest autoimmunity characterized by multiorgan infiltration of activated T cells and high levels of autoantibodies to multiple organs. A subpopulation of the mice also develops immune complex glomerulonephritis. These findings identify a physiological function of NF-B2 in the development of medullary thymic epithelial cells and, thus, the control of self-tolerance induction.
In the thymus, self-reactive T cells are eliminated through negative selection in which the T cell receptor of a thymocyte engages a high affinity peptide-major histocompatibility complex ligand presented by an antigen-presenting cell, leading to the apoptotic death of the thymocyte (1). Although it has been known for many years that medullary thymic epithelial cells (mTECs) 3 have a crucial role in negative selection by acting as antigen-presenting cells (2)(3)(4), only recently is the underlying mechanism beginning to emerge. mTECs express a broad spec-trum of peripheral tissue-restricted self-antigens, termed promiscuous gene expression (5,6). Evidence for a crucial role of this promiscuous gene expression in self-tolerance induction comes from analysis of mice lacking Aire, a transcription factor that is mutated in the human disease autoimmune polyendocrinopathy-candidiasis-ectodermal dystrophy (APECED) (7,8). mTECs from mice deficient in Aire have diminished expression of tissue-restricted self-antigens, and these mice develop a multiorgan autoimmune syndrome similar to APECED (9). The development of mTECs is accompanied by an increase in the expression levels of CD80 and a carbohydrate epitope that binds the lectin UEA-1 (10 -12). Both UEA-1 Ϫ and UEA-1 ϩ mTECs display promiscuous gene expression (5). However, a more recent study shows a close correlation between the expression levels of CD80 and the extent of promiscuous gene expression (6). The genes and signaling pathways that regulate the development of mTECs are not fully understood.
NF-B2 is a member of the NF-B family of transcription factors that also include p105/p50 (NF-B1), RelA (p65), RelB, and c-Rel. The full-length NF-B2 protein p100 is preferentially associated with RelB in the cytoplasm (13,14), which prevents RelB nuclear translocation and represses RelB-dependent transcription. Phosphorylation of the C terminus of p100 by IKK␣, which itself is activated by NF-B-inducing kinase (NIK), leads to proteolytic processing of p100 into p52 (15,16). The resulting p52-RelB heterodimers then translocate into the nucleus and activate the transcription of their target genes. This alternative NF-B signaling pathway is activated by engagement of receptors for B cell-activating factor, LT␤, and CD40 ligand (14,(17)(18)(19). Previous studies with NF-B2 Ϫ/Ϫ mice demonstrate a crucial role of NF-B2 in B cell development and secondary lymphoid organogenesis. These mice present a marked decrease in the B cell population in peripheral lymphoid organs and the absence of discrete perifollicular marginal and mantle zones and of germinal centers in the spleen (20,21).
Recently, several studies provide convincing evidence for a critical role of the LT␤R signaling pathway in regulation of mTEC development. Mice deficient in LT␤R, IKK␣, or carrying a loss-of-function mutant of NIK (NIK aly/aly ) all display disorganized thymic medulla, reduced numbers of mTECs, and overt autoimmunity (22)(23)(24). As the LT␤R signaling pathway is intimately involved in activation of NF-B2 (25,26), these findings also implicate a role for NF-B2 in the development of mTECs (12). However, defects in LT␤R signaling not only impair processing of NF-B2 p100 into p52 but also result in accumulation of p100, which may lead to repression of RelBdependent transcription. In fact, it was recently suggested that it is the increase in the p100 levels, rather than the absence of p52, that might be responsible for the impaired mTEC development observed in IKK␣-deficient and NIK aly/aly mice (24).
In this report, we describe an autoimmune phenotype for NF-B2 Ϫ/Ϫ mice that lack both p100 and p52 and present evidence for a physiological function of NF-B2 in the development of mTECs. Our findings, in conjunction with studies of other mutant mouse strains, delineate an NF-B2 activationsignaling pathway that links thymic organogenesis to the establishment of self-tolerance.

EXPERIMENTAL PROCEDURES
Mice-NF-B2 Ϫ/Ϫ mice (20) were crossed to B6129SF1/J (Jackson Laboratory), and the heterozygous offspring were interbred to obtain NF-B2 Ϫ/Ϫ , heterozygous, and wild-type littermates. NOD.SCID/NCr mice were purchased from the NCI, National Institutes of Health, at Frederick. All of the animals were maintained under specific pathogen-free conditions at the animal facility of the Medical University of Ohio, and all of the animal procedures were pre-approved by the Institutional Animal Care and Use Committee.
Histology and Immunohistochemistry-Tissues were fixed in 10% neutral buffered formalin, embedded in paraffin blocks, sectioned at 5 m, and stained with hematoxylin and eosin. Histological examination of silver-stained lung sections for possible Pneumocystis infection was conducted by the Research Animal Diagnostic Laboratory at the University of Missouri. For immunohistochemistry, the paraffin was removed and sections were rehydrated according to standard procedures. For retrieval of B220 and CD3 antigen, the sections were subjected to boiling in 10 mM citrate buffer (pH 6.0) or 1 mM EDTA (pH 8.0) for 10 min, respectively. Following quenching of endogenous peroxidase activity with H 2 O 2 and blocking with normal serum, the sections were incubated for 1 h with rat mAb against CD45R/B220 (5 g/ml, RA3-6B2; BD Biosciences) or CD3 (10 g/ml, CD3-12; Serotec). Isotype-matched rat mAb (10 g/ml; BD Biosciences) was used as control. After washing, biotinylated rabbit anti-rat antibody (Vector Laboratories) was applied for 30 min. The sections were then incubated for 30 min with ABC reagent (Vector Laboratories), and the immunostaining was visualized with 3,3Ј-diaminobenzidine (Sigma). The tissue sections were counterstained with hematoxylin and examined under a light microscope.
Real-time PCR-Real-time PCR quantification was conducted with cDNA prepared from total RNA extracted from individual thymi of 4-week-old NF-B2 Ϫ/Ϫ and wild-type mice. The primers and probes for Aire, Spt1, fatty acid-binding protein, glutamic acid decarboxylase 67, and glyceraldehyde phosphodehydrogenase were as previously described (28,29). PCR reactions in triplicate were performed using the TaqMan Universal PCR Master Mix (Applied Biosystems) and run on an Applied Biosystems 7500 real-time PCR system according to the manufacturer's instruction.
Analysis of Thymic Stromal Cells-Thymic stromal cells were prepared as described (30,31). Briefly, thymi of 4-to 6-week-old mice were minced and slowly stirred in the medium (RPMI 1640 plus 2% fetal bovine serum) to release the majority of thymocytes. The tissue fragments were sequentially digested with 0.5 mg/ml collagenase D (Roche Applied Science) and collagenase D/dispase I (0.2 mg/ml each; Roche Applied Science) in the presence of 25 g/ml DNase I (Worthington) in the same medium to release epithelial cells. The digests were pooled, disrupted by 5 mM EDTA, and separated on a discontinuous Percoll gradient (Pharmacia) with densities of ϭ 1.115, 1.06, and 1.0 g/ml (from bottom to top) by centrifugation at 1,350 ϫ g for 30 min at 4°C. The thymic epithelial cell fraction was collected from the interphase between ϭ 1.06 and 1.0 and subjected to three-step staining with Mouse BD Fc Block (2.4G2), with rat anti-mouse Ep-CAM (G8.8) and Biotin-UEA-1 or Biotin-CD80 (16-10A1), and then with PE Cy5-conjugated CD45 (30-F11), PE-conjugated Ly51 (BP-1), FITC-conjugated mouse anti-rat IgG2a (RG7/1.30), and PE Texas Redconjugated streptavidin (all from BD Biosciences). Cells were sorted on Epics Elite (Beckman-Coulter) with appropriate forward and side scatter setting to exclude thymocytes, and the data were analyzed with WinMDI 2.8 software.
Analysis of Regulatory T Cells-Single cell suspensions were prepared from the spleen of 4-to 6-week-old mice according to standard procedures. Red blood cells were depleted as described above. Regulatory T cells were detected by staining for the expression of CD4, CD25, and Foxp3, using a mouse regulatory T cellstaining kit (eBioscience) according to the manufacturer's instructions.

RESULTS
Multiorgan Lymphocytic Infiltration in NF-B2 Ϫ/Ϫ Mice-During a study of apoptosis regulation in thymocytes and T cells by NF-B2 p100 and its processed product p52, we noticed that a significant number of NF-B2 Ϫ/Ϫ mice died prematurely when compared with their heterozygous and wild-type littermates (Fig. 1A). Histological examination of various tissue samples revealed that, of the 20 deceased NF-B2 Ϫ/Ϫ mice, 5 had leukemia or lymphoma, indicated by complete effacement of normal bone marrow and spleen architecture by massive infiltration of medium-sized lymphocytes with pleomorphic nuclei and abundant cytoplasm (data not shown). However, the remaining 15 deceased NF-B2 Ϫ/Ϫ mice showed no obvious sign of leukemia in their bone marrow and spleen samples. Instead, these mice had extensive perivascular infiltration of morphologically normal lymphocytes in multiple organs (data not shown, see also Fig. 1B). The cellular infiltrates might result in organ dysfunction, leading to premature death.
We next examined 20 NF-B2 Ϫ/Ϫ mice at 10 -12 months of age to determine whether the multiorgan lymphocytic infiltration observed in the deceased mice was a consistent feature in NF-B2 Ϫ/Ϫ mice. All of the NF-B2 Ϫ/Ϫ mice showed marked lymphocytic infiltration in the lung, liver, and salivary gland (Fig. 1B). Of note, many of the lung lesions appeared to have lymphocytic invasion of artery walls, consistent with vasculitis (Fig. 1B). Immunohistochemical staining of lung sections revealed that most of the infiltrating cells were CD3 ϩ T cells (Fig. 1C). No significant lymphocytic infiltration was found in major organs of the 15 age-matched wild-type littermates examined (Fig. 1B). This phenotype of multiorgan lymphocytic infiltration in NF-B2 Ϫ/Ϫ mice is very similar to that described for various mouse strains with autoimmunity, such as those lacking Aire, LT␤R, IKK␣, or with NIK mutation (9,(22)(23)(24)28), suggesting that NF-B2 Ϫ/Ϫ mice may also develop autoimmunity.
Given the profound defect in B cell-mediated responses in NF-B2 Ϫ/Ϫ mice (20, 21), we examined the possibility that the Compared with wild-type and heterozygous littermates, NF-B2 Ϫ/Ϫ mice have significantly fewer and smaller lymph nodes, probably due to a defect in lymph node development. 4 Thus, the observed multiorgan lymphocytic infiltration could be a result of homeostatic reorganization. To investigate this possibility, we isolated infiltrating cells from the lungs and performed flow cytometry analysis for expression of T cell activation and memory phenotype markers (Fig. 1D). The majority of the infiltrating cells in the lungs of NF-B2 Ϫ/Ϫ mice were CD4 ϩ T cells. Staining for the activation marker CD69 revealed an average 25-fold increase in the percentage of activated CD4 ϩ cells in the lungs of NF-B2 Ϫ/Ϫ mice. The percentage of the CD4 ϩ cells with the CD44 ϩ memory phenotype was increased by an average of 4-fold. These data indicate that the lymphocytic infiltration in the organs of NF-B2 Ϫ/Ϫ mice was a result of an active ongoing immune response rather than a passive homeostatic process.
Autoimmune Diseases in NF-B2 Ϫ/Ϫ Mice-Given the apparent autoimmune phenotype of NF-B2 Ϫ/Ϫ mice, we further examined these mice for signs of autoimmune disease. Of the deceased NF-B2 Ϫ/Ϫ mice (n ϭ 20), 55% showed patholog-ical evidence of glomerulopathy, including mesangial proliferation, crescent formation, and diffuse interstitial lymphocytic infiltrates ( Fig. 2A). Immunofluorescence staining of cryostat-sectioned renal tissues using anti-IgG revealed deposits of immune complexes with a granular pattern in both the mesangial matrix and capillary loops (Fig. 2B), suggesting that these mice had immune complex glomerulonephritis. We also examined 20 NF-B2 Ϫ/Ϫ mice at 10 -12 months of age and found 30% of them had immune complex glomerulonephritis. None of the 15 age-matched or the two deceased wild-type littermates showed any pathological features of autoimmune renal disease.
It has been shown previously that, despite a marked reduction in the peripheral B cell population and impaired B cell-mediated immune response, NF-B2 Ϫ/Ϫ mice had a significant increase (4.5-fold) in the levels of serum IgM in comparison with control wild-type littermates (20). Immunostaining of NOD. SCID mouse tissues with sera from 1-year-old NF-B2 Ϫ/Ϫ mice (n ϭ 6) revealed the presence of high levels of IgM autoantibodies against multiple organs, including the liver, lung, salivary gland, and pancreas (Fig. 2C). We also detected high levels of antibodies against double-stranded DNA in two of the six serum samples from NF-B2 Ϫ/Ϫ mice but not in any of the six serum samples from age-matched wild-type littermates (data not shown). Together with the results of kidney histopathological analyses, these findings indicate that NF-B2 Ϫ/Ϫ mice developed systemic autoimmune disease, a phenotype consistent with a broad defect in self-tolerance induction.

NF-B2 Deficiency Has No Significant Effect on Activationinduced Thymocyte Apoptosis and the Number of Aire-express-
ing Cells in the Thymus-Self-tolerance is maintained by multiple mechanisms such as expression and presentation of selfantigens in the thymus, induction of apoptosis during negative selection, and production of regulatory T cells (32). Defects in any one of these mechanisms could lead to autoimmunity. As NF-B2 regulates death receptor-mediated apoptosis (33), which has been shown to play a role in negative selection (34), we examined NF-B2 Ϫ/Ϫ mice for the ability of anti-CD3 antibody to induce apoptosis in CD4 ϩ CD8 ϩ thymocytes, a widely used model of negative selection (35). No significant difference was observed in the levels of thymocyte death between NF-B2 Ϫ/Ϫ mice and their age-matched wild-type littermates after injection of anti-CD3 antibody (Fig. 3A). We also performed in vitro assays of anti-CD3-induced apoptosis of thymocytes and obtained similar results (data not shown). Thus, NF-B2 deficiency has no apparent effect on the ability of thy-  RelB has been shown to play a critical role in regulation of Aire expression or the survival of Aire-expressing cells in the thymus (27). As the NF-B2 p52/RelB heterodimer is a major component of the B binding activity found in the thymus (36), we speculated that NF-B2 might have a role similar to that of RelB in the thymus. Immunofluorescence staining of cryostatsectioned thymic tissues using an antibody against Aire revealed no apparent difference in the frequency of Aire-positive cells between age-matched NF-B2 Ϫ/Ϫ and wild-type mice (Fig. 3B), whereas quantitative real-time PCR analysis showed a modest (30%), but consistent, reduction in the Aire mRNA levels in NF-B2 Ϫ/Ϫ mice (Fig. 3C). These findings suggest that NF-B2 does not appear to play a significant role in regulation of Aire expression or the number of Aire-expressing cells in the thymus.
Impaired mTEC Development in NF-B2 Ϫ/Ϫ Mice-To search further for the mechanism underlying the autoimmune phenotype of NF-B2 Ϫ/Ϫ mice, we investigated a possible role of NF-B2 in the development of the thymic medulla and its cellular constituents. Histological examination of the thymi from 4-to 6-week-old NF-B2 Ϫ/Ϫ mice revealed no gross alterations in the thymic architecture (Fig. 4A), as previously reported (20). We also visualized thymic dendritic cells (DCs) with an antibody against the DC marker CD11c and epithelial cells with the antibody G8.8 that recognizes the marker epithelial cell adhesion molecule (Ep-CAM) (5). Immunofluorescence staining showed no differences in the numbers and distribution patterns of these cells between age-matched NF-B2 Ϫ/Ϫ and wild-type mice (Fig. 4B). In contrast, NF-B2 Ϫ/Ϫ mice have a marked reduction in the number of mTECs that bind the lectin UEA-1 (Fig. 4B).
To confirm these findings, we performed flow cytometry analysis of thymic epithelial cell populations (CD45 Ϫ G8.8 ϩ ) in 4-to 6-week-old NF-B2 Ϫ/Ϫ and wild-type mice (Fig. 4C). No significant difference in the numbers of cortical TECs (Ly51 ϩ ) was observed between the knock-out and wild-type mice (Fig.  4D). However, the number of CD80 ϩ cells was reduced by 88% in the thymus of NF-B2 Ϫ/Ϫ mice in comparison with their wild-type littermates (Fig. 4D). CD80 is a marker for mature mTECs (12) but also expressed on activated B cells (37,38). Dual immunofluorescence staining of thymic sections from both wild-type and NF-B2 Ϫ/Ϫ mice revealed that all of the CD80 ϩ cells were negative for the B cell markers B220 and surface IgM (data not shown). Taken together, these data indicate that NF-B2 Ϫ/Ϫ mice have a marked reduction in the number of CD80 ϩ mTECs, suggesting that NF-B2 has a specific and crucial role in the differentiation of mTECs and/or the survival of mature mTECs. NF-B2 Ϫ/Ϫ Mice Show Defects in Promiscuous Gene Expression in the Thymus-It was recently shown that CD80 ϩ mTECs display the highest degree of promiscuous gene expression (6). As NF-B2 Ϫ/Ϫ mice show a marked reduction in the number of CD80 ϩ mTECs, they may have defects in promiscuous expression of tissue-restricted self-antigens in the thymus. We investigated this possibility by examining the expression levels of three representative tissue-restricted self-antigens, Spt1, fatty acid-binding protein, and glutamic acid decarboxylase 67 (9,(22)(23)(24). Real-time PCR analysis revealed that the expression levels of Spt1 and fatty acid-binding protein in the thymus of NF-B2 Ϫ/Ϫ mice were decreased by 65 and 85%, respectively, in comparison with their age-matched, wild-type littermates (Fig.  5A). However, NF-B2 deficiency had no apparent effect on the expression of glutamic acid decarboxylase 67 in the thymus (Fig. 5A). These data suggest that NF-B2 Ϫ/Ϫ mice are defective in promiscuous expression of some, but not all, tissue-restricted self-antigens, which may impair the process of selftolerance induction, leading to the development of autoimmunity.
We next examined the possibility that the reduced number of mTECs and the resulting defect in promiscuous gene expression in NF-B2 Ϫ/Ϫ mice may impair the production of CD4 ϩ CD25 ϩ regulatory T cells (39,40), a population of T cells important for suppression of CD4 ϩ T cell-mediated organ-specific autoimmune diseases (41,42). Flow cytometry analysis revealed that spleens from NF-B2 Ϫ/Ϫ mice actually have a 2.2-fold increase in the percentage of CD4 ϩ CD25 ϩ T cells compared with their wild-type littermates (Fig. 5B). We wanted to point out that NF-B2 Ϫ/Ϫ mice also have more CD4 ϩ T cells in their spleens (Fig. 5B), and therefore the ratio of CD4 ϩ CD25 ϩ to CD4 ϩ T cells is about same between NF-B2 Ϫ/Ϫ (10.5%) and wild-type (10.6%) mice. Given the activated phenotype of peripheral CD4 ϩ T cells in NF-B2 Ϫ/Ϫ mice (Fig. 1D), which may co-express CD25, we further analyzed the CD4 ϩ CD25 ϩ T cells for the expression of Foxp3 (Fig.  5C), a recently identified marker for regulatory T cells (43)(44)(45). The analysis revealed that NF-B2 Ϫ/Ϫ mice had a modest (27%) increase in the number of regulatory T cells in the spleen com-  DECEMBER 15, 2006 • VOLUME 281 • NUMBER 50 pared with their wild-type littermates (Fig. 5D). We also examined the frequency of regulatory T cells in the thymus and found no significant differences between NF-B2 Ϫ/Ϫ and wildtype mice (data not shown). Together, these data indicate that NF-B2 deficiency does not significantly affect the production of regulatory T cells. Thus, the autoimmune phenotype of NF-B2 Ϫ/Ϫ mice probably results from impaired elimination of auto-reactive T cells.

DISCUSSION
In this report we describe an autoimmune phenotype for NF-B2 Ϫ/Ϫ mice, manifesting as multiorgan infiltration of activated T cells, high levels of autoantibodies in the serum, and spontaneous development of immune complex glomerulonephritis in a subpopulation of the mice. We have further shown that NF-B2 Ϫ/Ϫ mice have a specific defect in the generation of UEA-1 ϩ and CD80 ϩ mature mTECs, leading to a marked reduction in promiscuous expression of some peripheral tissue-specific antigens critical for the induction of self-tolerance (5,6,46). The impaired development of mTECs and resulting breakdown in self-tolerance induction probably lead to autoimmunity in NF-B2 Ϫ/Ϫ mice. Thus, NF-B2 not only is essential for B cell development and secondary lymphoid organogenesis, as reported before (20,21), but also has a physiological function in the development of mTECs and, thus, the control of selftolerance induction.
Several recent studies demonstrate a crucial role for the LT␤R signaling pathway in regulation of mTEC development. Mice deficient in LT␤R, IKK␣, or carrying NIK aly/aly mutation all display reduced numbers of mTECs and overt autoimmunity (22)(23)(24), a phenotype shared by NF-B2 Ϫ/Ϫ mice, as demonstrated in this study. Activation of the LT␤R signaling pathway induces the processing of NF-B2 p100 into p52 (17). As a result, defects in this signaling pathway lead to both p52 reduction and p100 accumulation. It was recently suggested that it is the increase in the p100 levels that might be responsible for the impaired mTEC development observed in IKK␣ Ϫ/Ϫ and NIK aly/aly mice (24). However, this model is not supported by our data, which, instead, suggest that the reduced p52 production is most likely the cause of the impaired mTEC development in mice deficient in LT␤R signaling. Thus, our study provides the final evidence for an essential role of NF-B2 activation by LT␤R signaling in the differentiation, proliferation, and/or survival of mTECs. Pooled thymic stromal cells from three NF-B2 Ϫ/Ϫ or wild-type mice 4 to 6 weeks of age were stained with antibodies to CD45, G8.8, CD80, and Ly51. Mature mTECs are CD45 Ϫ G8.8 ϩ Ly51 Ϫ CD80 ϩ , and cortical TECs are CD45 Ϫ G8.8 ϩ Ly51 ϩ . Data in panel D represent means Ϯ S.D. from three independent experiments with a total of nine mice for each genotype. **, Student's t test, p Ͻ 0.01. We noticed that NIK aly/aly and IKK␣ Ϫ/Ϫ mice display more severe structural and cellular defects than do NF-B2-deficient mice. Unlike NF-B2 Ϫ/Ϫ mice, NIK aly/aly and IKK␣ Ϫ/Ϫ mice show disorganization of thymic medulla and a marked reduction in the number of CD4 ϩ CD25 ϩ regulatory T cells (23,24). Although the underlying mechanism remains to be defined, we speculate that it may be related to the control of RelB expression and activation. NF-B2 p100 is an inhibitor of RelB transcriptional activity by forming a complex with RelB in the cytoplasm, which prevents RelB from associating with other NF-B molecules and entering into the nucleus (13,14). Therefore, the absence of p100 in NF-B2 Ϫ/Ϫ mice is expected to result in activation of NF-B complexes containing RelB. On the other hand, the expression of RelB is down-regulated in NIK aly/aly and IKK␣ Ϫ/Ϫ mice (23,24). Thus, RelB may play a more general role in thymic organogenesis and regulatory T cell production. Consistent with the model, mice lacking tumor necrosis factor receptor-associated factor 6 (TRAF6), in which RelB expression is also down-regulated, show a dramatic reduction in the size of the thymic medulla and in the number of CD4 ϩ CD25 ϩ regulatory T cells (47). Also, the phenotype of RelB Ϫ/Ϫ mice more closely resembles that of NIK aly/aly , IKK␣ Ϫ/Ϫ , and TRAF6 Ϫ/Ϫ mice, with complete disruption of the thymic medulla (48,49). A further examination of RelB-deficient mice will reveal whether they also have defects in the production of regulatory T cells.
The distinct phenotypes of knock-out mice lacking individual NF-B members suggest that they have non-redundant physiological functions. Because the NF-B family members exert their biological functions as homo-or heterodimers, the functional specificity must be encoded in the particular NF-B dimers. Thus, identification of particular NF-B dimers that drive distinct biological processes is essential for a molecular understanding of NF-B biology. Our study, coupled to the phenotypic analysis of RelB Ϫ/Ϫ mice (48,49), suggests a cell type-dependent specificity in RelB binding partners. The p52/ RelB heterodimer appears to have an essential role in the control of mTEC differentiation and maturation, as both RelB Ϫ/Ϫ and NF-B2 Ϫ/Ϫ mice display a significant reduction in the number of UEA-1 ϩ mTECs. However, in contrast to NF-B2 Ϫ/Ϫ mice, RelB Ϫ/Ϫ mice also show an absence of thymic DCs (48). Because the predominant B binding activity in thymic extracts is composed of p52/RelB and p50/RelB heterodimers (36), it is most likely that the p50/RelB dimer plays a critical role in the development of thymic DCs. This model is consistent with the reported defect in thymic DC function and development in NF-B1 and NF-B2 double knock-out mice (50).
It was recently demonstrated that mTECs and cortical TECs share a common epithelial precursor (51)(52)(53). However, NF-B2 deficiency has no apparent effect on the development of cortical TECs, despite its essential role in the generation and/or maintenance of mTECs. A molecular understanding of this cell type-dependent activation of NF-B2 will likely provide valuable insights into the process of thymic epithelial cell differentiation. Finally, the specific deficiency of mature mTECs in NF-B2 Ϫ/Ϫ mice provides an experimental system for identifying NF-B2 target genes that regulate the development of mTECs.