New Zeland Black (NZB) mice develop an autoimmune disease involving an abnormal B cell response to peripheral self Ags. This disease is associated with defects in other cell types and thymic stromal organization. We present evidence that NZB cells of various lineages, including thymocytes, fibroblasts, and dendritic precursor cells, show impaired proliferation and enhanced cell death in culture upon stimulation compared with non-autoimmune-prone mice such as C57BL/6. This phenotype explains the reduced efficiency of maturation of bone marrow-derived dendritic cells and the loss of TNF- or IL-1-dependent thymocyte costimulation. Upon TNF-induced activation of NZB thymocytes, nuclear translocation and DNA binding of RelA- and RelB-dependent NF-κB heterodimers are significantly reduced. This phenotype has a transcriptional signature, since the NZB, but not the nonobese diabetic, thymic transcriptome shows striking similarities with that of RelB-deficient thymuses. This partial NF-κB deficiency detected upon activation by proinflammatory cytokines could explain the disorganization of thymic microenvironments in NZB mice. These combined effects might reduce the efficiency of central tolerance and expose apoptotic debris generated during inflammatory processes to self recognition.
Most autoimmune diseases depend upon complex interactions between genetic and epigenetic factors acting in concert to augment or reduce the risk of developing the disease (1). The genetic risk factors can affect various arms of the immune response linked to cell differentiation, migration, or activation. Most frequently, each factor is individually insufficient to trigger the pathology, but their combination can skew immune responses toward self-destruction (2, 3). Identification of phenotypic variations of lymphoid behaviors in autoimmune-prone mice should reflect the participation of candidate susceptibility genes. We have focused our analysis on the lymphoid phenotypes that can be detected in animal models of autoimmunity before the appearance of the disease. Alterations of T lymphocyte activation have been documented in lupic (4) or diabetic (5) patients and in the nonobese diabetic (NOD)4 (6) and New Zeland Black (NZB)/New Zeland White (NZW) (7) mouse models. Such abnormalities might also alter negative selection of thymocytes during ontogeny. Consequently, a less efficient tolerance induction would predispose to the development of autoimmune diseases. Thymocyte selection occurs in specialized thymic microenvironments that are regulated by a cross-talk between thymocytes and stromal cells (8). In mice in which T cell differentiation was abortive, thymic medulla did not develop. Similarly, a deficient activation of medullary epithelial and dendritic cells (DC), as observed in RelB-deficient mice (9, 10), translated into an impairment of negative selection (11). Phenotypically, the thymus of these mice was severely disorganized (12). A similar, although less extensive, disorganization was observed in the thymus of aly, NZB, and NOD mice (13, 14, 15) and correlated with the pathology (16, 17). Since this phenotype represents an attenuated version of that observed in the thymus of a RelB-deficient mouse, we speculated that an alteration of the NF-κB pathway might be associated with autoimmune disorders (18). This hypothesis was recently reinforced by the identification of a mutation of the NIK gene of the aly mouse. This mutation abolishes the activation of the NF-κB pathway in response to lymphotoxin β (19) and explains the attrition of lymph nodes leading to poorly focused lymphocyte responses (14). Several NF-κB agonists (TNF, leukotriene, CD40, etc.) are required for the organization of lymphoid microenvironments and the coordination of peripheral lymphoid responses (20).
We thus focused our approach on the analysis of NF-κB/RelB-dependent activation of the cells involved in the organization of thymic medulla, namely mature thymocytes and DC from NZB mice. In this report we document a deficiency of the NF-κB pathway in NZB mice that could explain the increased rate of activation-induced cell death in the presence of TNF or IL-1. This phenotype correlates with the reduced efficiency in thymocytes and DC maturation in vitro.
Materials and Methods
C57BL/6N Crl BR (B6), NMRI IOPS Han (Swiss), and NZB/Ola Hsb (NZB) mice were purchased from Charles River (l’Arbesle, France), SER/J (Le Genest, St. Isle, France), and Harlan, (Gannat, France), respectively. RelB-deficient mice were derived from D. Lo’s laboratory (Scripps Research Institute, La Jolla, CA) and bred locally on a C56BL/6 background. TCRα-deficient mice were provided by P. Ferrier (Centre d’Immunologie de Marseille-Luminy, Marseille, France). NOD mice were provided by the Laboratory of Diabetology (UPRES-EA2193, Marseille, France).
Thymocytes were harvested from 6- to 10 wk-old mice and cultured overnight to deplete adherent cell populations. Surviving cells were enriched in single-positive (SP) thymocytes (∼21–22% CD4+ and 5–7% CD8+, respectively). Viable thymocytes (6.5 × 105) were cultured on anti-CD3ε mAb-coated wells (145-2C-11, at 10 μg/ml) in the presence of various concentrations of cytokines. All assays are represented using a stimulation index corresponding to the ratio of cytokine and CD3-stimulated vs CD3-stimulated thymocytes. Cell cultures were pulsed on day 2 with [3H]Tdr for 18 h before harvesting on a Packard beta-plate counter (Downers Grove, IL).
Dendritic and embryonic fibroblast cell cultures
Bone marrow-derived DC were grown as previously described (21). Briefly, bone marrow cell suspensions were depleted of lineage-committed cells by complement depletion. Precursor cell-enriched populations were cultured for 5 days in the presence of GM-CSF.
Embryonic fibroblasts (EF) were derived from day 14 embryos. Whereas B6, NMRI, and RelB-deficient EF lines were passaged every 2–3 days, NZB EF lines were passaged every 5–6 days. Cytotoxicity assays were usually performed using an lactate dehydrogenase-based assay (CytoTox 96 Non-Radioactive Cytotoxicity Assay; Promega, Madison, WI) on recently derived cell lines as the background of cell death increases significantly after few passages.
Cytofluorometric analysis and cell sorting
Thymocyte samples were stained with a combination of PE-coupled anti-CD4−, biotinylated anti-CD8− and FITC-coupled anti-CD24 (M1/69) mAb (BD PharMingen, San Diego, CA).
Cultures of bone marrow-derived DC were analyzed with a combination of CD11b-allophycocyanin, CD11c-biotin, MHC class II-PE, and CD40-FITC or CD86-FITC mAb, followed by streptavidin-CyChrome (BD PharMingen). Apoptotic cells were detected using the Topro 3 reagent. All samples were analyzed on a FACSCalibur apparatus (BD Biosciences, Mountain View, CA).
For sorting of thymic stromal cell subsets, thymuses were enzymatically dissociated, and low density cells were enriched on a Percoll gradient as previously described (22, 23). Cell suspensions were stained with the CD11c-PE (BD PharMingen) or 29-FITC anti-EpCAM mAb and sorted on a FACStar+ apparatus. For the preparation of cell-specific mRNA probes, ∼5 × 105 cells were purified from at least 10 dissociated thymuses.
EMSA and Western blot for NF-κB analysis
The protocols for EMSA were previously described and used NF-κB consensus or mutated sequences derived from the human IL-2Rα promoter (24). For Western blot analysis, 10 μg lysates from cytosolic or nuclear extracts were loaded on SDS-PAGE and transferred on Immobilon-P membrane (Millipore, Bedford, MA). Blots were probed with commercial rabbit antisera against p50/p105, IκBα (Upstate Biotechnology, Lake Placid, NY; Euromedex, Mandolsheim, France), p65/RelA (Chemicon, Temecula, CA), p52/p100, c-Rel, and RelB (Santa Cruz Biotechnology, Santa Cruz, CA) and revealed using a peroxidase-coupled donkey anti-rabbit polyclonal Ab (Jackson ImmunoResearch Laboratories, West Grove, PA) and ECL (Amersham, Arlington Heights, IL).
Preparation and screening of high density filters of the thymic cDNA library
The methodology for preparation and screening of high density DNA macroarrays has been previously described (25). The thymic MTB cDNA library (Soares thymus 2NbMT) was obtained from the IMAGE Consortium (http://image.llnl.gov/) and cloned in the DH10B bacterial strain before plating in 386-well plate. This library had been equalized to limit clone redundancy and allow the detection of rarer transcripts. Copies of an Arabidopsis thaliana cytochrome c554 cDNA clone (CG03) and copies of three different clones containing only poly(A) sequences were present on each membrane and serve as control clones (26). Each thymic clone was spotted in duplicate. Membranes were subsequently treated as previously described (27) with one modification, i.e., increase in the duration of proteinase K treatment to 15 h. Hybridization with 33P-labeled probes, normalization, and quantification have been extensively described (28). Screening of the MTB library (∼12,000 clones) was performed in four consecutive rounds using probes derived from whole thymus RNA of different origins. A first screen using RNA from control (B6 and Swiss) vs RelB−/− or NZB thymuses allowed the selection of ∼2,000 underexpressed clones in either or both mice (overexpressed clones were omitted since they were artificially enriched in cortical clones due to the increased representation of the cortex in these thymuses). The two following screens used the same probes and also included RNA from NOD thymuses. The final screen was performed on a selection of 373 selected clones, including some control clones that were spotted on miniarrays and hybridized with various thymic or sorted cell-specific probes as indicated. Clones were selected considering a variation of the intensity of hybridization of 4-fold in primary and 2-fold in the final screenings vs control. Sequencing was performed by Qiagen (Cologne, Germany). Sequence comparisons were performed using BLAST (29).
Inefficient expansion and enhanced cell death in cultured DC and embryonic fibroblasts from NZB mice
The thymus of RelB-deficient (9, 10) and NZB (30) mice contains fewer DC and is disorganized (12, 16, 17). Furthermore, the RelB molecule is required for the differentiation of myeloid-derived DC (31). We thus evaluated the production of bone marrow derived DC from NZB mice. As shown in Fig. 1,A, the number of lineage-negative dendritic precursor-enriched bone marrow cells was 2.5 times lower in NZB mice. Furthermore, when the same number of enriched precursor cells was cultured for 5 days in the presence of GM-CSF, the total number of DC was 2–3 times lower in NZB than in C57BL/6 (B6) mice. As shown in Fig. 1,B, these cultures contained a mixture of immature CD11b−CD11c− and more mature cells expressing variable levels of CD11b and CD11c molecules. In absolute numbers, NZB cultures contained 2–3 times fewer mature cells but 6 times fewer CD11b−CD11c− precursors (2.34 vs 0.4 × 106 cells). This result was explained by the increased proportion of apoptotic cells (25 vs 15%) among NZB vs B6 immature precursors (Fig. 1,C), whereas this proportion was similar among mature populations (∼8–14%). However, maturation is not impaired in NZB cultures. As shown in Fig. 1 D, a fraction of CD11c+ cells expressed maturation markers such as CD86 and MHC class II molecules as well as CD40 (data not shown). The percentage of mature DC was even higher in NZB cultures in part due to the reduction in the number of precursor cells. These results were confirmed by the detection of cell surface expression of MHC class II molecules by confocal analysis of mature DC obtained after LPS stimulation of an enriched population of immature cells (data not shown). Thus, in NZB mice the production of bone marrow-derived DC is reduced in efficiency due to a higher level of cell death among precursors.
RelB is also involved in the control of fibroblast growth (32). Mouse embryonic fibroblasts were derived from control and NZB embryos, and as shown in Table I, the number of passages of NZB MEF was strongly reduced compared with that of control fibroblasts. In control cultures ∼20% cells were spontaneously apoptotic in the presence of cycloheximide, and this value was doubled by incubation in the presence of TNF. In NZB cultures, a similar increase in TNF-induced cell death was observed, whereas in RelB−/− EF lines only TNF-induced cell death was enhanced compared with that in B6 mice. Thus, at least two independent cell types from NZB mice exhibit a higher constitutive and activation-induced cell death.
|Strain .||No. of Passages .||% Cell Death .||.||TNF-Dependent Cell Death .|
|.||.||−TNFα .||+TNFα .||.|
|NMRI||15–22||23 ± 8 (6)||43 ± 6 (4)||20|
|C57BL/6||15–22||25 ± 5 (6)||43 ± 3 (4)||18|
|C57BL/6 RelB−/−||15–22||26 ± 8 (6)||52 ± 13 (4)||26|
|NZB||5–6||40 ± 12b (6)||71 ± 12b (4)||31|
|Strain .||No. of Passages .||% Cell Death .||.||TNF-Dependent Cell Death .|
|.||.||−TNFα .||+TNFα .||.|
|NMRI||15–22||23 ± 8 (6)||43 ± 6 (4)||20|
|C57BL/6||15–22||25 ± 5 (6)||43 ± 3 (4)||18|
|C57BL/6 RelB−/−||15–22||26 ± 8 (6)||52 ± 13 (4)||26|
|NZB||5–6||40 ± 12b (6)||71 ± 12b (4)||31|
Freshly transplanted MEF were cultured in the presence of cycloheximide (30 ng/ml) with or without TNF (100 ng/ml) for 18 h, and viability was evaluated using the lactate dehydrogenase cytotoxicity assay. Results from several independent experiments are represented. The average number of passages in culture before growth arrest are indicated, and the number in parentheses represents the number of experiments. Values from NZB mice were compared with control values from NMRI and B6 mice and were statistically increased as indicated.
p < 0.05, by Student’s t test.
Defective costimulation and increased death of CD3-activated NZB mature thymocytes with IL-1β and TNF-α
Late thymocyte maturation in thymic medulla is accompanied by the acquisition of a competence to proliferate upon TCR-induced stimulation (33). This proliferation can be further enhanced by a costimulatory cytokine such as TNF-α or IL-1β. These cytokines activate the NF-κB pathway upon engagement of the TNF-R2 and IL-1R1 receptors (34, 35), which recruit distinct adapter proteins (36). Although the precise role of these cytokines is not fully understood during thymic maturation, this experimental set-up allowed us to explore a NF-κB-dependent activation step in mature thymocytes. We first checked by flow cytometry and semiquantitative RT-PCR that the expression levels of these receptors were comparable between NZB and control mice, although some quantitative variations could be detected (data not shown). Then thymocyte cultures depleted of adherent cells were plated on CD3 mAb-coated wells. At the beginning of the cultures, the percentages of mature thymocytes were equivalent in all experimental situations; furthermore, we checked that the level of CD3-dependent proliferation was comparable in all mouse strains (data not shown). Under these experimental conditions the addition of IL-1β or TNF-α to the CD3 stimulation of control thymocyte cultures led to a 2- to 3-fold increase in proliferation defining the stimulation index (Fig. 2,A). Similar results were obtained using Swiss, C57BL/6, or BALB/c mice (data not shown). Using NZB thymocytes, the stimulation index remained close to 1 over a large range of cytokine concentrations. The defective proliferation could be due to an increased level of activation-induced thymocyte death. Indeed, we consistently observed a higher proportion of apoptotic NZB vs B6 thymocytes at every time of the cultures (∼10% in B6 vs 20–30% in NZB cultures; data not shown). Thus, we focused our analysis on the mature SP thymocytes, which are susceptible to cytokine-induced costimulation. A cytometric analysis was performed at the end of the culture using the CD4, CD8, and CD24 markers, since the loss of CD24 is coincident with terminal maturation (Fig. 2,B). In both B6 and NZB mice, CD3-mediated activation lead to the expansion of a fraction of double-negative and mature SP thymocytes. The reduction in the percentage of total CD4+ thymocytes after activation is due to the expansion of DN cells, which were not further studied (data not shown). The proportion of CD4+ thymocytes was comparable in B6 and NZB cultures (∼20% cells; data not shown). In contrast, the addition of cytokines (mainly TNF-α) enhanced the proportion of mature CD4+CD24neg cells in B6, but not NZB, cultures (Fig. 2,B). The behavior of CD8+ cells was different in culture. Upon activation of B6 thymocytes, their proportion increased 1.5- to 2-fold in the presence of CD3 and cytokines (data not shown). In NZB cultures the increase in CD8+ thymocytes was not observed, and among them, the proportion of CD24neg thymocytes was significantly reduced after activation in the presence of cytokines such as TNF-α and IL-1β (Fig. 2 B). Thus, the effects of cytokines are complex and vary with the cell types. In NZB thymocytes, TNF-α fails to expand CD4+ and kills CD8+ thymocytes, explaining the lack of costimulation with this cytokine. The effects of CD3 alone or in combination with IL-1 are more modest on the cell phenotypes, but the same tendencies can be detected. In conclusion, these results show a higher susceptibility to TNF-induced cell death among activated mature thymocytes.
Impaired nuclear translocation of NF-κB/Rel complexes
Then, we analyzed the consequences of TNF-α-mediated thymocyte activation on the translocation of active NF-κB dimers from the cytosol to the nucleus. Indeed, TNF-R1 are expressed by most thymocytes and can trigger the NF-κB pathway in a majority of cells (37). Using a consensus human IL-2Rα-derived κB oligonucleotidic probe, we performed EMSA using nuclear extracts from resting or activated thymocytes. As shown in Fig. 3,A, NF-κB complexes were detected in resting thymocytes. The predominant one corresponded to complex a, which disappeared after competition with an unlabeled wild-type and not a mutated κB oligonucleotide. After TNF activation, the proportion of complex a was reduced, and higher m.w. complexes b and c appeared; the latter complexes were not detected in the presence of a competing oligonucleotide. In contrast, complex d was removed by the addition of a SP1/GC-box oligonucleotide attesting the presence of SP1-related complexes. We then identified the relative contributions of the different NF-κB members by supershift and blocking assays using specific Abs. As shown in Fig. 3,B, the amount of complex a was strongly reduced by incubation with an anti-p50 Ab. An anti-RelA Ab lead to the supershift of complex b. Complex c disappeared upon incubation with anti-p52 and anti-RelB Abs, whereas anti-c-Rel Ab had no effect on any of the complexes. These results are compatible with the predominance of p50 homodimers in freshly harvested thymocytes; upon TNF activation, both p50/relA and p52/relB heterodimers become detectable. A similar analysis was performed using thymocytes from NZB mice, and the quantification of the results from a series of three independent experiments using an internal nonspecific band as a control is shown in Fig. 3C. In NZB mice, all NF-κB complexes (a, b, and c) were significantly reduced in intensity. These results suggested that the amount of transcriptionally active NF-κB molecules might be reduced in the nuclei of activated NZB thymocytes. This hypothesis was reinforced by Western blot analysis. As shown in Fig. 3 D, p50, p105, p52, p100, RelB, RelA (p65), and Iκ-Bα molecules were detected in cytosolic extracts. Upon TNF stimulation, one observed a time-dependent nuclear translocation of RelA and RelB proteins, whereas p52 and p50 (not shown) could also be detected in resting conditions. In NZB thymocytes, the proportion of translocated RelA and RelB proteins was reduced in agreement with the results obtained by EMSA analysis.
RelB−/− and NZB thymic transcriptional signatures are comparable
To evaluate the in vivo relevance of the reduced NF-κB-dependent activation observed in vitro, we searched for a transcriptional signature in thymus. The transcriptomes of different mice were compared using a thymic cDNA library gridded on nylon macroarrays. A pilot approach performed on selected clones representative of the cortical and medullary microenvironments (23, 25) showed striking similarities between RelB−/− and NZB, but not B6 or MHC-null, transcriptomes (data not shown). This strategy was repeated with the larger MTB cDNA library (∼12,000 clones). Several rounds of screenings led to the selection of 2,030 underexpressed clones, which were further analyzed with thymic probes from RelB−/−, NZB, and control (B6, Swiss, and NOD) mice or tissue- or cell-derived probes. The results from the final selection (373 clones, most of them without informative sequence) are compiled in Fig. 4. The underexpression ranged from a 1.5- to 10-fold reduction. The clones could be organized in five unequal clusters based on their expression profile in control vs NZB, NOD, and RelB−/− mice. First, most of them were underexpressed in the medulla-less TCRα-deficient mice (38), suggesting that the screening mainly detected transcripts associated with late thymocyte maturation or a medullary stromal expression. The first cluster contained 13 clones under-represented in RelB−/−, NZB, and NOD thymuses, including the IL-4 and thymopoietin transcripts. The second cluster contained six clones under-represented in NOD thymuses, few of them expressed in stromal cells. The third cluster contained 14 clones under-represented in RelB−/− thymuses, including TNF and other ubiquitous transcripts. The two other clusters carried the predominant transcriptional signature. In cluster 4, 79 clones (41 without sequence) were less represented in NZB and RelB−/−, but not NOD, thymuses. Several clones corresponded to the 28S rRNA or transcripts linked to RNA processing (splice factor, helicase, capping enzyme) or DNA repair (Ku autoantigen). Almost all NF-κB, including RelB, transcripts were identified in this cluster, as well as other molecules involved in signal transduction (sox 4, RFLAT-1, pim3). Interestingly, the abundance of transcripts encoded by medullary stromal cells, such as the chemokine TCA-4 (39), or by mature thymocytes such as the tyrosine kinases p59fyn and ZAP70 (40), was also reduced. The remaining unsequenced clones were either preferentially expressed by medullary stromal cells or ubiquitous (data not shown). The last cluster (including 89 unidentified sequences) grouped the clones predominantly underexpressed in NZB thymuses, although most of them were slightly under-represented in a RelB−/− thymus. Among them, several transcripts coded for molecules associated with cell metabolism and/or activation. Thus, the transcriptomes of NZB and RelB−/−, but not NOD, thymuses share many general features linked to abnormal cell activation. These results suggest that the NF-κB-dependent transcriptional activity is diminished in vivo.
The autoimmune-prone NZB mouse strain displays a poorly organized thymic microenvironment. In this paper we show that cultured NZB cells of different types show enhanced apoptosis following activation. The presence of these two phenotypes could be explained by a defective activation of the NF-κB pathway. Here we correlate these functional results with a reduced TNF-induced nuclear translocation of NF-κB/Rel heterodimers in mature thymocytes and the presence of a RelB-deficient-like transcriptional signature in NZB thymuses.
Activation of the NF-κB pathway is required for the initiation of inflammation and the organization of immune responses in vivo (41). In several cell models NF-κB-dependent transcriptional activation leads to the production of antiapoptotic molecules that regulate the balance between survival and death signals (42, 43). In our study of NZB mice one of the most consistent observations is the detection of an inefficient proliferation and increased susceptibility to apoptosis in culture using thymocytes, DC precursors, or embryonic fibroblasts. This phenotype is detected upon activation by growth-inducing agonists and is enhanced upon costimulation with cytokines such as TNF-α or IL-1β, which trigger the NF-κB pathway. Interestingly, NZB mice show a hypoproduction of NO in response to LPS stimulation, another agonist of the NF-κB pathway (44). This impaired expansion of cultured cells in vitro seems to be constitutive to NZB-derived cells, since it is detected in primary embryonic fibroblast or DC cultures and is further enhanced by TNF-α as with mature CD8+ thymocytes. This phenotype could explain the reduced number of dendritic and medullary epithelial cells observed in NZB thymuses (17, 30, 45). The expansion and maturation of myeloid-derived DC are NF-κB/RelB dependent in vivo (31) and in vitro (46, 47). Our results show that the maturation of myeloid DC is not impaired in vitro, but there is a reduction in the amount of transcripts associated with thymic medullary stromal cells in vivo. A typical result is obtained with the chemokine TCA-4, which is under a RelB-dependent transcriptional control (39). A less efficient activation associated with enhanced cell death was also detected with mature thymocytes stimulated with anti-CD3 mAb and IL-1β or TNF-α. This result was corroborated by the transcriptome analysis showing that transcripts coding for tyrosine kinases highly expressed in mature thymocytes were less abundant. Thus, these results suggest that thymocyte maturation might be less efficient in vivo and could explain the impairment of T cell development in NZB fetal thymic organ culture (7), where IL-1β or TNF-α are required for full maturation (48). An increased susceptibility to apoptosis has not been detected under CD3- and CD28-mediated activation (49). This suggests that the CD28 pathway might be sufficient to trigger antiapoptotic mechanisms, whereas cytokines fail to do so. This information might help to position a putative defect in NZB mice, since CD28 can also activate the NF-κB pathway via distinct intracytoplasmic effectors (50). The combination of an inefficient organization of the thymic medulla and a reduced efficiency in thymocyte maturation could lead to a reduction in the efficiency of negative selection of mature thymocytes. Furthermore, the higher susceptibility to cell death in the presence of TNF might contribute to the release of apoptotic self Ags in inflamed tissues. Interestingly, the NZB background predisposes to the development of lupus, a disease in which several autoantigens derive from apoptotic cells. A phenotype of defective lymphocyte activation (51), enhanced lymphocyte activation-induced apoptosis, and abnormal NF-κB activation has been reported in human lupic patients (4, 52, 53, 54), and several reports link the failure to scavenge apoptotic cells in vivo with lupus (55). Furthermore, attempts to enhance NF-κB activation modify disease outcome in lupus-prone (NZB × NZW)F1 mice (56). These results suggest that, as in lupic patients, NZB mice show a defective activation pathway that might enhance the rate of immune cell apoptosis under activation.
These functional results were compatible with an abnormal activation of the NF-κB or NF-κB-linked pathway that was explored in TNF-activated thymocytes. Indeed, the antiapoptotic activity of this cytokine requires NF-κB-dependent transcription (43). Upon TNF stimulation, we observed a reduction in the amount of translocated RelA and RelB proteins that explains the reduced binding of the nuclear p52/relB and relA/p50 heterodimers to DNA. The presence of a quantitatively normal amount of nuclear p52 protein suggests that the mechanisms leading to its production are unaffected. This process involves the activation of the NIK kinase that is mutated in the aly (19), but not in NZB (R. Valéro, unpublished observations), mouse. This mutation explains the disorganization of lymphoid microenvironments in the aly mouse (57). Furthermore, given the homologies between the NZB and RelB-deficient stromal phenotypes, we envisaged the possibility that the RelB gene might be mutated in NZB mice. Indeed, the loss of the RelB gene leads to enhanced apoptosis and reduced terminal maturation of thymocytes (58), as in NZB mice. This possibility was excluded by direct sequencing of the gene and by a Northern blot analysis of thymic RNA, which showed the presence of a normal RelB transcript but at a reduced level compared with that in control mice (M.-L. Baron, unpublished observations). Thus, further explorations are required to provide a molecular explanation to this phenotype.
These in vitro results were correlated with the transcriptome analysis. Indeed, the thymic transcriptomes of NZB and RelB-deficient mice display striking similarities. Results from such a global analysis can reflect changes in the relative proportions of cells as well as direct transcriptional effects. Thymuses from these disabled mice do not show major alterations in the global distribution of thymocyte subsets. They are enriched in apoptotic cells and contain fewer activated medullary stromal cells than controls. The reduction in the amount of the TCA-4 chemokine (directly checked by TaqMan analysis; M.-L. Baron, unpublished observations) and other unknown stromal cell-restricted transcripts can be considered a valid signature of a RelB-dependent transcriptional activity. A majority of transcripts are under-represented in both RelB-null and NZB, but not NOD, thymuses. Among them are found several members of the NF-κB family, a majority of them being positively regulated by NF-κB-dependent transcriptional activation (59) and markers of cell viability (RNA processing mainly), suggesting a possible link between their lower abundance and the proportion of apoptotic cells. Other underexpressed transcripts, such as those coding for the kinases ZAP70 and p59fyn, probably reflect the reduced efficiency of terminal thymocyte maturation and are also undetectable in TCRα-deficient mice. Another predominant category corresponds to transcripts more specifically reduced in NZB mice and are globally related to cell metabolism, reinforcing the point that NZB cells have a reduced metabolic activity. In NOD mice used as an autoimmunity control, the thymic transcriptome was relatively different from that observed in NZB and RelB-deficient strains and does not support the existence of a major deficiency of the NF-κB pathway. Arguments in favor of hyper- (60) or hypoactivation (61) of NF-κB have been reported. Some lymphocyte phenotypes, such as the resistance to activation-induced cell death (62) and a defect in central tolerance in NOD mice (49), might also be related to this issue.
In conclusion, NZB mice exhibit an attenuated NF-κB-deficient phenotype. Although the molecular mechanism requires further investigation, this study represents another example of linkage between NF-κB and autoimmunity (63). This NF-κBlow phenotype could explain both the enhancement of cell death under inflammation and the disorganization of the thymic microenvironment. Whereas the latter might reduce the efficiency of central tolerance induction, the first phenotype might lead to the exposure of abundant apoptotic debris in inflamed sites (64). The combined effects contribute to autoimmunity toward apoptotic cells on the appropriate genetic background.
We thank F. Galland and R. Guinamard for helpful discussions and critical reading of the manuscript, and C. Chabret, G. Victorero, and B. Loriod for their technical help at different stages of the study.
This work was supported by Institut National de la Santé et de la Recherche Médicale, Centre National de la Recherche Scientifique, and Program Hospitalier de Recherche Clinique. M.-L.B. was the recipient of a fellowship from the Ministry of Research and Association pour la Recherche contre le Cancer. R.V. was on a Poste d’Accueil from Institut National de la Santé et de la Recherche Médicale. S.G. was funded by a grant from Association pour la Recherche contre le Cancer. S.B. was the recipient of a grant from the Fondation pour la Recherche Médicale.
Abbreviations used in this paper: NOD, nonobese diabetic; DC, dendritic cell; EF, embryonic fibroblast; SP, single positive.