RelB forms transcriptionally inactive complexes with RelA/p65.

RelB is an unusual member of the NF-kappaB transcription factor family that acts as both a transcriptional activator as well as a repressor of NF-kappaB-dependent gene expression. Although RelB promotes gene expression when it associates with p50/NF-kappaB1 or p52/NF-kappaB2, the precise molecular mechanisms through which it represses NF-kappaB remain unclear. To examine this inhibitory function in more detail, we employed reporter gene assays and found that RelB represses at the level of RelA. Furthermore, electrophoretic mobility shift analysis revealed that in vitro translated RelB impaired the DNA binding activity of RelA and that overexpressed RelB significantly reduced tumor necrosis factor-alpha-induced RelA activity in murine embryonic fibroblasts. Intriguingly, this inhibitory effect was due to the formation of RelA.RelB heterodimers that were unable to bind to kappaB sites in vitro strongly suggesting that these newly described NF-kappaB dimers cannot bind DNA. Expression pattern analysis revealed that RelA.RelB heterodimers appeared at relatively low levels in both lymphoid and non-lymphoid cells. However, the presence of these complexes increased following stimulation with phorbolesters or lipopolysaccharide or by overexpression of constitutively active IKKbeta. Functional characterization of RelA.RelB heterodimers in NIH3T3 murine embryonic fibroblasts revealed that they are not regulated by IkappaB proteins and are located in both the cytoplasm and the nucleus. Taken together, our findings demonstrate that sequestration of RelA in transcriptionally inactive RelA.RelB complexes provides a molecular mechanism that may explain the repressive role of RelB on NF-kappaB-dependent gene expression.

NF-B describes a family of ubiquitously expressed transcription factors that play crucial roles in regulating innate and adaptive immunity, cell differentiation, and apoptosis (1). NF-B is composed of a dimer of distinct members of the Rel protein family (RelA, c-Rel, RelB, p50, and p52), and in the majority of non-stimulated cells it is maintained inactive in the cytoplasm via association with members of a family of small inhibitory proteins named the IBs. Following stimulation of cells with a wide variety of agents, including TNF␣, 1 LPS, or phorbol esters, IB proteins become phosphorylated on two amino-terminal serine residues through the activity of a specific upstream kinase complex named the IB kinase (IKK). This phosphorylation targets the IB proteins for ubiquitination and degradation by the proteasomal pathway enabling the free NF-B to translocate to the nucleus where it binds to specific DNA elements in the control regions of various genes and promotes their expression (2).
The generation of mice deficient for each Rel protein has revealed that the individual NF-〉 family members have distinct, non-redundant functions. Furthermore, the separate combinations of NF-〉 dimers recognize distinctive variants of the 〉 DNA binding site (3,4). This modulatory design therefore contributes to the differential patterns of target gene expression induced by the NF-〉 family of proteins. Classic NF-〉 is considered to be a heterodimer of RelA and p50 that is efficiently complexed by I〉 proteins and is thereby highly inducible by stimulation with agents such as TNF␣, which transduce signals via the IKK complex. The importance of RelA is highlighted by the fact that it supports the expression of a wide variety of genes that include various leukocyte adhesion molecules, chemokines, effector enzymes, and certain IB and NF-〉 members (5,6). Due to this widespread action RelA has been demonstrated to be of exceptional importance for the anti-apoptotic and pro-inflammatory roles of NF-〉.
In contrast to RelA and the other NF-〉 family members, RelB has been demonstrated to act as both a positive promoter of NF-〉-dependent gene expression and a repressor of NF-〉 activity (7)(8)(9). Moreover, RelB is only weakly controlled by I〉 proteins and is the major contributor toward the constitutive NF-〉 activity observed in lymphoid cells (10). Another specific feature of RelB is its limited ability to form complexes with the other Rel proteins such that it is unable to homodimerize and only heterodimerizes with p50 and p52, but not with RelA or c-Rel (36). Although only a few RelB target genes have been reported, its positive transcriptional role is clearly manifested in studies using RelB-deficient mice. These mice are not only deficient for a thymic medulla and a subset of dendritic cells, but they also display defects in the development of secondary lymphatic organs suggesting a crucial role of RelB for these developmental processes (12).
Recently, it has been demonstrated that RelB⅐p52 heterodimers function in an alternative NF-〉 activation pathway that might play a central role in the phenotype of RelB-deficient mice. This pathway culminates in the regulated processing through the action of the IKK subunit IKK␣ of p100 to form p52, which then associates predominantly with RelB (13,14). However, the phenotype of RelB Ϫ/Ϫ mice also indicates a repressive role for RelB. Thus, in RelB-deficient fibroblasts, the expression of various chemokines and cytokines is augmented and prolonged through a mechanism that is at least in part, dependent upon IKK-induced NF-〉 activity. Furthermore, because the TNF␣ gene is active in RelB Ϫ/Ϫ but not in wild type fibroblasts, it has been suggested that RelB plays an epigenetic role in silencing the TNF␣ gene similar to the reported RelBspecific demethylation of the Ig locus in B cells (15,16). Another example for the repressive function of RelB is its ability to impair the combined phorbolester and ionomycininduced NF-〉 activity in T cells (7). Recently, we reported a signal-induced proteolysis of RelB in activated T cells, implying a necessity for removal of either the negative or the positive RelB function for a proper T cell activation response (17).
In our current study we analyzed in greater detail the repressive effect of RelB on NF-〉 activity. We found that RelB reduces NF-〉 activity by directly influencing the activity of RelA and that the molecular mechanism underlying this negative effect was inhibition of DNA binding by RelA through its complex formation with RelB. Mapping of the interaction domains demonstrated that only the Rel homology domains (RHDs) are necessary for this interaction similar to all other NF-〉 dimers (11). More importantly, we were unable to detect binding of the RelA⅐RelB heterodimer to known 〉 sites raising the possibility that endogenous RelA⅐RelB complexes are unable to bind to DNA. We further observed that RelA⅐RelB heterodimers exist in lymphoid as well as in non-lymphoid cells and are not, at least in the non-lymphoid cells, controlled by I〉␣ or I〉␤. Because the relb gene itself is a target for NF-〉, the increase in RelB expression is paralleled by an increase in levels of RelA⅐RelB heterodimers. Finally, we show using murine embryonic fibroblasts (MEFs) as a model system, that high levels of RelB protein lead to a drastic reduction in active RelA and that this reduction is due to an increase in the formation of RelA⅐RelB heterodimers.
Preparation of Lymphocytes, Cell Culture, and Reporter Gene Assays-Wild type murine embryonic fibroblasts and human Raji and Daudi B cells as well as human 293 HEK cells were cultured in Dulbecco's modified Eagle's medium plus 10% fetal calf serum. Human Jurkat T leukemic cells, murine 70Z/3 pre B cells, and murine EL-4 thymoma cells were cultured in RPMI medium containing 5% fetal calf serum. Murine splenocytes from Balb/c mice were prepared according to standard protocols. Thymocytes from wild type mice or mice with a tissue-specific constitutively active IK⌲␤ transgene were prepared as previously described (20). Briefly, prepared thymocytes were disrupted by using a cell strainer (Falcon), and the resulting cells were pelleted, washed twice with cold phosphate-buffered saline then loaded onto a Ficoll gradient to remove dead cells and erythrocytes. In all cytokine stimulation experiments, recombinant hTNF␣ (R&D systems) was used at a concentration of 10 ng/ml. Transient transfections and luciferase assays were performed using the Fugen6 transfection reagent (Roche Diagnostics) and the dual luciferase reporter assay system (Promega), respectively, according to the manufacturer's protocols. All experiments were performed in parallel and repeated at least three times with similar results.
In Vitro Transcription and Translation-For in vitro translation, 1 g of the appropriate expression vector was mixed with 20 l of rabbit reticulocyte lysate from the TNT-T7 Quick in vitro transcription and translation system (Promega) and 1 l of 35 S-labeled methionine. The reaction was incubated for 30 min at 30°C, and successful expression of the proteins of interest was analyzed by SDS-PAGE, then exposures were made to x-ray film. Immunoblots, Immunoprecipitations, and DNA Binding Studies-For the preparation of whole cellular extracts 5 ϫ 10 6 to 1 ϫ 10 7 of lymphoid and non-lymphoid cells were pelleted, washed once with icecold phosphate-buffered saline, then resuspended in 200 l of TNT buffer (200 mM Tris-HCl, pH 8.0, 20 mM NaCl, 1% Triton X-100, 1 mM dithiothreitol) supplemented with protease inhibitors (Complete protease inhibitor mixture, Roche Applied Science). The extracts were cleared by centrifugation at 14,000 rpm for 10 min at 4°C. Nuclear and cytoplasmic protein extracts were prepared (21), and immunoblot (IB) assays and immunoprecipitations (IPs) were performed as previously described (22). Briefly, for IP experiments samples were equalized to contain identical protein concentrations (between 500 and 1000 g/ sample) in the TNT buffer. Specific antibodies (2 g per sample) or, as a control, non-immune serum were added, and the samples were incubated on ice for 1 h prior to the addition of 30 l a 50% Protein-A-Sepharose slurry. The samples were then incubated an additional 4 -12 h on a rotor, washed extensively with TNT lysis buffer supplemented with protease and phosphatase inhibitors, and then boiled in 25 l of SDS-sample buffer. Immunocomplexes were separated using SDS-PAGE (8%), transferred to a PVDF membrane, and then blotted with the appropriate antibodies. EMSAs were also performed as previously described (21) using either the palindromic B site, which allows for RelB binding, or the 〉 site from the Ig enhancer that favors RelA binding (13,18). For the supershift assays 1 g of a specific antibody was added to each protein-DNA mixture after 10-min incubation on ice. The samples were incubated for an additional 10 min prior to loading onto the gels.

RelB Exerts a Dual Role on NF-B-dependent Gene
Expression-RelB is the only member of the NF-〉/Rel family that both positively and negatively influences NF-〉-dependent gene expression. It is known that RelB promotes NF-〉-dependent gene expression when complexed with p50 or p52, but the molecular mechanisms that underlie its negative function remain unclear. To study this negative regulatory function in detail, we first cotransfected 293 HEK cells with a 〉-dependent luciferase reporter construct in the absence or presence of increasing concentrations of an expression vector encoding murine RelB then stimulated the cells with either TNF␣ or PMA for 4 h. As shown in Fig. 1A, overexpression of RelB dose-dependently reduced NF-〉-dependent gene expression to ϳ50% of both the TNF␣-and PMA-induced (upper and lower panels, respectively) activity. To determine whether this inhibition was due to possible interference with the function of other NF-〉 proteins, we cotransfected 293 HEK cells with expression vectors for RelA, p50, or RelB, either alone or in combination. As expected, RelB did not induce NF-B activity when transfected alone, whereas cotransfection with p50 led to a significant increase in activity over that induced by p50 alone (Fig. 1B, columns 5-7). Intriguingly however, RelB overexpression dose-dependently reduced RelA-mediated 〉-dependent gene expression (Fig. 1B, columns 2-4). To ensure that this effect was not due to reduced levels of expressed RelA, we analyzed these samples by immunoblotting using antibodies specific for either the HA-tag (RelA) or the FLAG-tag (RelB). As shown in the inset (Fig. 1), the levels of overexpressed RelA remained constant in the presence of increasing amounts of RelB. Furthermore, the fact that coexpression of RelB and p50 significantly increased the NF-〉 activity argues against the possibility that the reduced RelA activity is based on the sequestration of its complex partner p50 in non-binding RelB⅐p50 heterodimers.
RelB Forms a Complex with RelA-Inhibition of RelA activity through interaction with other transcription factors, including glucocorticoid receptors, p53, or c-Myc has been previously described (23)(24)(25). Whereas association with the glucocorticoid receptors maintains the DNA binding activity of RelA, it has been demonstrated that p53 hinders the ability of RelA to bind to DNA. To analyze the molecular mechanism through which RelB inhibits RelA activity, we performed an EMSA experiment using a radiolabeled 〉-oligonucleotide together with in vitro translated versions of full-length RelB and the Rel homology domain (RHD) of RelA (RelA⌬313). We used this deletion mutant of RelA because the full-length protein binds poorly to the 〉 binding site due to an intramolecular interaction (18). In addition, by employing this in vitro assay system we avoided additional binding of other, endogenous NF-B complexes to the B site. In our experiments, both proteins were in vitro translated separately then mixed together with the indicated relative amounts prior to incubation with the radiolabeled B probe. As expected, RelB alone was unable to bind to DNA ( Fig.  2A; lane 6), whereas the robust DNA binding activity of the homodimer of RelA⌬313 (RelA⌬313 (2) ) was dose-dependently reduced by the addition of RelB (lanes 2-5). Because addition of in vitro translated RelB protein reduced the DNA binding of RelA⌬313 but did not lead to binding of an additional, slower migrating band ( Fig. 2A), we conclude that the RelA⅐RelB complex is unable to bind to the 〉-oligonucleotide used in this assay.
To determine whether this impaired DNA binding of RelA was due to complex formation with RelB, we transiently transfected 293 HEK cells with expression vectors encoding HAtagged RelA or FLAG-tagged RelB either individually or in combination and performed immunoprecipitation (IP) analysis using whole cell extracts prepared from these cells. As shown in Fig. 2B, HA-tagged RelA was immunoprecipitated with the FLAG-specific antibody only in the presence of cotransfected FLAG-RelB (lane 4) demonstrating that RelA interacts with RelB to form a complex. We noticed a reduced protein concentration of both ectopic-expressed Rel proteins in this particular experiment, which is probably due to a mutual repression of the expression caused high amounts of plasmid DNA in the transfection. Using various RelA and RelB deletion mutants, we mapped the interaction domains responsible for RelA⅐RelB complex formation and found that the RHDs of both proteins are required for the association (Fig. 2C), hence, indicating that this interaction mechanism strongly resembles the classic heterodimerization between other Rel proteins.
Overexpressed RelB Impairs TNF␣-induced RelA Activity in Fibroblasts-Having established that complex formation with RelB blocks the DNA binding capacity of RelA in vitro, we wished to determine whether this mechanism regulates the activities of the endogenous proteins in vivo. We therefore established a stable RelB-overexpressing murine embryonic fibroblast cell line as a model system, and, as shown in Fig. 3A, the levels of RelA, p50, p52, and their precursors p100 and p105, respectively, remained unaltered in these cells compared with control cells. We therefore compared the 〉-specific DNA binding activity in nuclear extracts from either control cells or the RelB-overexpressing cells that were either untreated or stimulated with TNF␣ for 20 and 60 min. EMSA analysis revealed that overexpression of RelB dramatically altered the basal (Fig. 3B, compare lanes 1 and 4) and TNF␣-induced (lanes 2 and 3 versus lanes 5 and 6) 〉-specific binding complexes. Thus, basal binding activity was very low in untreated control cells but increased strongly after stimulation with TNF␣ for 20 min (complex C1Ј) and decreased thereafter ( Fig.   FIG. 1. RelB overexpression impairs NF-〉-dependent gene transcription. A, 293 HEK cells were transiently transfected with a 〉 controlled firefly luciferase reporter construct (200 ng) and 20 ng of a ␤-actin promoter-driven Renilla luciferase reporter plasmid in the absence or presence of increasing amounts of an RelB-encoding expression vector. Twenty-four hours later, the cells were either left untreated or incubated for a further 4 h with either 10 ng/ml hTNF␣ (upper panel) or 50 ng/ml PMA (lower panel). Cells were then lysed, and the firefly and Renilla luciferase activity was determined. The data are presented as the firefly luciferase readings normalized to the Renilla luciferase counts. B, 293 HEK cells were transiently transfected with the 〉-dependent reporter construct along with an expression vector for RelA in the absence or presence of increasing amounts of the wild type RelB expression vector. In addition, expression vectors for RelB or p50 were either transfected individually or in combination together with the 〉 reporter construct. The luciferase activity in whole cell lysates was then determined 24 h post transfection. Selected samples were subjected to immunoblot analysis using either anti-HA or anti-FLAG antibodies to determine the expressed levels of RelA and RelB. All experiments were performed at least three times with similar results.
3B, lanes 1-3). In contrast, in RelB-overexpressing cells we detected a strong constitutive 〉-binding activity of multiple complexes that we designated as complexes C1, C2, and C3 (Fig. 3B, lane 4). Furthermore TNF␣ stimulation led to a slight increase in the binding activity of only complex C1, whereas the complexes C2 and C3 remained unaltered (Fig. 3B, lanes 5 and   6). To determine the composition of these distinct 〉 complexes, we used nuclear extracts from control and the RelBoverexpressing cells stimulated with TNF␣ for 20 min for supershift analysis employing antibodies specific for RelA, RelB, p50, and p52. As shown in Fig. 3C, antibodies specific for both p50 and RelA reduced binding of the complex C1Ј in control cells extracts, whereas antibodies specific for RelB or p52 had no effect (Fig. 3C, lanes 1-5). In extracts from RelB-overexpressing cells we detected a partial reduction in complex C1 binding using antibodies for RelA and more strongly with antibodies specific for RelB, p50, and p52 (Fig. 3C, lanes 6 -10). The RelB antibody also affected the binding of both complex C2 and C3. Furthermore, complex C2 binding was blocked by the p52 antibody. These findings led us to conclude that the strong inducible binding activity (complex C1Ј) in control cells consists of RelA⅐p50 heterodimers, whereas in RelB-overexpressing cells, complex C1 is comprised of a mixture of RelA⅐p50 and RelB⅐p50 heterodimers, and complex C2 represents a RelB⅐p52 heterodimer. Interestingly, complex C3 appears to represent a FIG. 2. Dimerization with RelB impairs the DNA binding ability of RelA. A, in vitro translated RelA⌬313 and RelB were mixed at the indicated relative amounts (top) and incubated for 20 min at 30°C. Additional reticulocyte lysate was added to ensure equal volumes were used for each sample. The preincubated protein extracts were then subjected to an EMSA using a 32 P-labeled 〉-specific oligonucleotide derived from the Ig enhancer. B, 293 HEK cells were transfected with FLAG-tagged RelB or HA-tagged RelA vectors (1 g each) either alone or in combination as indicated. After 48 h whole cell extracts were prepared and subjected to immunoprecipitation (IP) using anti-FLAGconjugated agarose beads. Immunocomplexes were separated by 10% SDS-PAGE and blotted onto PVDF membrane then immunoblotting (IB) was performed using antibodies specific for the FLAG-or the HA-epitopes. C, schematic representation of full-length and deletion mutants of RelB and RelA used for mapping the interaction domains that mediate the association between the proteins. Mapping of the interaction domains was done after transient transfection of 293 HEK cells with either the HA-tagged RelA or the FLAG-tagged RelB deletion mutants alone or in combination. The interaction was determined by anti-FLAG IP followed by immunoblotting with anti-HA and subsequently anti-FLAG antibodies. Interacting (ϩ) and non-interacting (Ϫ) mutants are shown on the right. The positions of the transactivation domains (TA), the Rel homology domains (RHD), the leucine zipper, and the tags (FLAG, HA) are indicated.

FIG. 3. RelB overexpression reduces the inducible activity of RelA in fibroblasts.
A, whole cell extracts (50 g) from control cells stably transfected with empty vector or RelB-overexpressing cells (RelB) were subjected to immunoblot analysis using antibodies specific for RelB, RelA, p50, or p52. Note that the antibodies for p50 and p52 also recognized p105 and p100, respectively. B, for the EMSA studies 10 g of nuclear proteins extracts from untreated control or RelB-overexpressing cells or from cells stimulated with 20 ng/ml recombinant human TNF␣ for 20 or 60 min were used together with a 32 P-labeled palindromic 〉-oligonucleotide. The samples were separated on a 6% native polyacrylamide gel that was then fixed, dried, and exposed to x-ray film. C, the supershift analysis using nuclear extract from control or RelB-overexpressing cells stimulated with TNF␣ for 20 min was performed similar to the EMSA experiment except that 1 g of the indicated antibody was added to the samples. slower migrating RelB-containing complex, although we were unable to precisely identify any other proteins in this complex. Nevertheless, it seems unlikely that this represents a RelB homodimer, because RelB alone cannot bind to 〉 sites. Taken together, the studies described above reveal that RelB overexpression dramatically reduces the inducible binding activity of RelA through the formation of heterodimeric complexes.
Efficient IB Degradation and Nuclear Translocation of RelA in RelB-overexpressing Fibroblasts-Previous studies using fibroblasts from RelB-deficient mice implied that RelB might negatively regulate IKK activity in cells stimulated with LPS (16). Hence, the possibility remained that the reduction in RelA activity we observed occurs via inefficient degradation of IB proteins due to reduced activity of the IKK complex. We therefore performed immunoblot analysis for IB proteins using cytoplasmic extracts from control cells or RelB-overexpressing cells that were either left untreated or stimulated for 20 and 60 min with TNF␣. As shown in Fig. 4A, we detected an effective inducible degradation of I〉␣ and I〉␤ in the control as well as the RelB-overexpressing cells (compare lanes 1-3 with lanes  4 -6). Moreover, although the basal protein levels of I〉␣ were comparable between control cells and RelB cells (Fig. 4A, lanes  1 and 4) we consistently observed that IB␣ was not resynthesized as efficiently in RelB-overexpressing cells (compare lanes  3 and 6).
In addition to IB degradation, nuclear translocation of the Rel proteins is also critical for their transcriptional activity. Therefore, we performed immunoblot analysis for RelA using nuclear extracts from control and RelB-overexpressing cells. In control cells we detected a low basal level of nuclear RelA, which was increased following 20 min of TNF␣ stimulation and declined thereafter (Fig. 4B, lanes 1-3). The time course of this nuclear localization reflected the DNA-binding activity observed in Fig. 3B. Stable overexpression of RelB led to an increase in both the basal as well as the inducible levels of nuclear RelA (Fig. 4B, lanes 4 -6). Not surprisingly, higher levels of nuclear RelB were detected in RelB cells in comparison to the control cells. In addition, we detected a slight increase of the nuclear localization induced by TNF␣ in both the control cells as well as the RelB-overexpressing cells (Fig. 4B, lower  panel). The effective inducible degradation of I〉␣ and I〉␤ (Fig. 4A) suggests that IKK activity induced by TNF␣ is unaffected by RelB overexpression; indeed, we did not detect significant reduction of inducible IKK activity by in vitro kinase assays (data not shown). In addition, the augmented nuclear translocation in RelA in response to TNF␣ seen in RelB cells also suggests that the mobilization of RelA is not affected by RelB overexpression. We therefore reasoned that, because RelB can associate with RelA (Fig. 2, B and C) and RelB overexpression inhibits RelA activity (Figs. 1 and 2C), this might occur through formation of RelA⅐RelB heterodimers. To investigate this possibility we performed immunoprecipitations from whole cell extracts of control and RelB-overexpressing cells using an antibody specific for RelA. As control, similar IP experiments were performed using non-immune serum to ensure the specificity of the interaction (data not shown). As shown in Fig. 4C, immunoblot analysis of the resulting immunoprecipitated material revealed that RelA⅐RelB heterodimers are present in both control cells and at substantially increased levels in RelB overexpressing cells (Fig. 4C). Note that the ectopic overexpressed RelB is FLAG-tagged and therefore migrates slower in the polyacrylamide gel (upper arrow).

Characterization of RelA⅐RelB Heterodimers in Fibroblasts-
The finding that endogenous RelA⅐RelB heterodimers exist in the control cell line albeit at low levels (Fig. 4C, lane 3), prompted us to perform a more detailed characterization of these complexes using these cells. We began by determining the relative amount of RelB that was complexed with RelA compared with the levels associated with either p50 or p52 by performing coimmunoprecipitation analysis using antibodies specific for RelA, p50, and p52 followed by immunoblotting with anti-RelB. As demonstrated in Fig. 5A we detected a RelB-specific signal after immunoprecipitation with each antibody. Nevertheless, although the amount of RelB that coimmunoprecipitated with p50 and p52 antibodies was equally strong, we consistently detected less RelB associated with RelA. To control for the level of immunoprecipitated RelA, we reprobed the membrane using anti-RelA, and as expected, we detected a strong RelA-specific signal (Fig. 5A, lower panel). Neither RelA nor RelB were detected in similar experiments using nonimmune serum (data not shown). Because the p50 and p52 bands overlapped with the Ig heavy chain signal in our immunoblots, we compared the level of each protein in pre-IP sam-  ). B, nuclear protein extracts (30 g) from untreated or TNF␣-stimulated control cells or RelB-overexpressing cells were separated by 8% SDS-PAGE, transferred to PVDF membrane, then immunoblotted with RelA-(upper panel) or RelB (lower panel)-specific antibodies. C, whole cellular extracts (1.5 mg) from untreated control cells or RelB-overexpressing cells (RelB) were subjected to immunoprecipitation using anti-RelA or non-immune serum (data not shown). The precipitates were washed extensively then the precipitated proteins were separated by 8% SDS-PAGE, transferred to PVDF membrane, then immunoblotted using an anti-RelA antibody. The membrane was stripped and reprobed using anti-RelB. As a control, 10% of the input from each cell line was separated on a parallel lane and treated identically to the immunoprecipitates.
ples with the immunodepleted extracts to ensure the success of the IP (lanes 4 and 5).
In light of the fact that activation of RelA is efficiently controlled by IB␣ and I〉␤, we sought to determine whether RelA⅐RelB heterodimers are bound by either of these I〉 proteins. Consistent with the previous reports we detected efficient complex formation between RelA and either I〉␣ or I〉␤ (Fig. 5B). In contrast, RelB did not coimmunoprecipitate with antibodies against either IB protein suggesting that, at least in fibroblasts, RelA⅐RelB heterodimers are not subject to control by I〉␣ or I〉␤. This finding is consistent with the previous observations that the only IB protein regulating RelB in fibroblasts is the NF-B2/p100 protein (26). The subcellular localization of NF-〉 proteins is mainly controlled by their interaction with I〉s such that NF-〉 heterodimers that are not bound to I〉s are able to enter the nucleus. To analyze the subcellular distribution of RelA⅐RelB heterodimers in unstimulated fibroblasts, we prepared nuclear and cytoplasmic extracts from MEFs and performed IP experiments using an anti-RelA antibody or, as control, non-immune serum (data not shown). Surprisingly, we found that the majority of RelA⅐RelB heterodimers were located in the cytoplasm, although a small fraction was also located in the nucleus (Fig. 5C) suggesting that the nuclear translocation of RelA⅐RelB heterodimers is a regulated process, possibly independent of IB proteins. Inter-estingly, we also detected a small amount of RelA⅐p50 heterodimers in the nucleus (Fig. 5C, lanes 1 and 2). As a control for the purity of the nuclear extracts, we used an I〉␣-specific antibody to immunoblot nuclear and cytoplasmic extracts (lanes 3 and 4).
NF-B-dependent Increase in RelA⅐RelB Heterodimer Formation in Lymphoid Cells-RelB Ϫ/Ϫ mice display various defects in the development and regulation of the immune system (27)(28)(29), and RelB plays an important role in lymphoid cells where it is constitutively located in the nucleus (10). Given this importance of RelB in lymphoid cells, we analyzed different B (Daudi and Raji) and T (Jurkat and EL-4) cell lines for the expression of RelA⅐RelB heterodimers and in each of these we were able to coimmunoprecipitate RelA with RelB using a RelB-antibody. Interestingly, we also detected c-Rel association with RelB that occurred predominantly in the B-cell lines (Fig.  6A, middle panel). To investigate RelB expression in primary lymphoid cells, we used whole cell extracts from splenocytes in a similar IP experiment using RelA-and RelB-specific Abs (Fig.  6B). Once again we were able to coimmunoprecipitate RelA and RelB, demonstrating that RelA⅐RelB dimers exist in primary lymphoid cells.
RelB is constitutively located in the nuclei of mature B cells, but its nuclear concentration increases in B cells stimulated with CD40 or LPS or anti-Ig (32,35). Furthermore, the differ- MEFs were subjected to immunoprecipitation using antibodies specific for RelA, p50, or p52 (3 g each). The immunocomplexes were separated by SDS-PAGE then transferred to PVDF membrane prior to immunoblot analysis using anti-RelB-and anti-RelA antibodies. To ensure efficient precipitation of p50 or p52, 50 g of the extracts used for the IP was compared with 50 g of the immunodepleted extracts in an immunoblot analysis using antibodies specific for p50 or p52. B, a similar IP experiment was performed using antibodies specific for either I〉␣ or for I〉␤. In parallel, an IP experiment was performed using nonimmune serum (data not shown). C, nuclear and cytoplasmic proteins (600 g each) were isolated prior to immunoprecipitation using anti-RelA. The immunoprecipitates were separated by SDS-PAGE, transferred to PVDF membrane, then probed using anti-RelB. The membrane was stripped and subsequently reprobed for p50 and RelA. Portions of the cytoplasmic (50 g) and nuclear (20 g) protein extracts were used for immunoblot analysis to control for the subcellular localization of RelA, RelB, and p50. To ensure the purity of the nuclear extract, the same IB was reprobed using anti-I〉␣.

FIG. 6. RelA⅐RelB heterodimers are expressed in lymphoid cells. A, whole cell extracts (500 g) from T cell (Jurkat and EL-4) and B cell lines (Daudi and Raji) were used in an IP experiment with an
RelB-specific antibody. The resulting membrane was probed for RelA and, after stripping, subsequently reprobed for c-Rel and RelB. To ensure the specificity of the reaction, 4 g of the immunogenic RelB peptide was used in parallel samples (lanes 2, 4, 6, and 8). B, splenocytes were isolated from BL-6 mice then extracts were prepared and used for IP experiments with anti-RelA and anti-RelB. The resulting membrane was then subjected to immunoblot analysis with RelA-and RelB-specific antibodies. Parallel experiments were carried out using 3 g of immunogenic RelB peptide to control for the specificity of the reaction. In addition, expression of the indicated proteins was ensured by loading 20 g of the extracts on a separate lane (lanes 3 and 6). entiation of pre B cells after IL-7 withdrawal correlates with a strong increase in RelB expression (33). To determine whether an increase in the level of RelB expression correlates with an increase in RelA⅐RelB heterodimer formation we stimulated cells of the pre B cell line 70Z/3 with PMA or LPS for various times. Stimulation with PMA for 1 h led to a temporary drop of RelB protein levels in the pre-immunoprecipitation samples followed by an increase after 8 and 24 h (Fig. 7A, lower panels). An increase in RelB expression was also detected after stimulation with LPS for 8 and 24 h and, in addition, we detected a slight increase in the expression of RelA by immunoblot analysis (Fig. 7A, lower panel). To determine the amount of RelB complexed with RelA, we performed immunoprecipitation experiments with anti-RelA-specific antibodies (Fig. 7A, upper  part). Both stimuli led to a significant increase in the RelA⅐RelB heterodimer formation after 8 and 24 h. In control IP experiments with non-immune serum neither RelA nor RelB was detected (data not shown). Because RelB is a NF-〉 target gene and its expression is up-regulated by cell stimulation in a NF-〉-dependent fashion, we directly tested whether the increase in RelA⅐RelB heterodimerization is dependent upon NF-〉 activation. Hence, we examined RelB expression in transgenic mice with a tissue-specific overexpression of a constitutively active isoform of IK⌲␤ (IK⌲␤CA) in the thymus that were reported to exhibit high basal NF-〉 activity and low levels of I〉␣ protein (20). As shown in Fig. 7B (lanes 1 and 2), RelB protein expression levels are higher in IK⌲␤CA transgenic cells than wild type thymocytes. Moreover, in an IP experiment using an RelA-specific Ab, we detected a significant increase in the RelA⅐RelB heterodimer formation compared with thymocytes from control mice (Fig. 6D, compare lanes 3  and 4). This finding strongly suggests, therefore, that increased RelA⅐RelB heterodimer formation in lymphoid cells is dependent upon the induced activity of the IKK complex and the subsequent activation of NF-〉. DISCUSSION The limited ability of RelB to form heterodimers with p50 or p52, but not with RelA or c-Rel, together with its inability to homodimerize identified RelB as an unusual member of the NF-〉 family. However, when combined with p50 or p52, RelB promotes the expression of a limited number of NF-〉-dependent genes (6,30). Recently, several reports emphasized the importance of RelB when complexed with NF-〉2 (p52 or its precursor p100) as the basis of an "alternative" NF-〉 pathway, which is solely dependent on the activity of IKK␣ but not on the activity of the canonical IKK pathway involving NEMO and IKK␤ (13). The negative role of RelB, however, is far less characterized, and the aim of this study was to gain more insight into the molecular mechanism underlying this negative function.
Initially cloned and described as human I-Rel, an inhibitory Rel protein, RelB inhibits phorbolester-induced NF-〉 activation when overexpressed in Jurkat T cells. We also detected an inhibition of NF-〉 in reporter gene assays after stimulation with TNF␣ or phorbolester using 293 HEK cells. However, the level of inhibition reflected the dual role of RelB as an inhibitor and positive effector of NF-〉. Thus, NF-〉 activity declined after overexpression of low amounts of RelB but stayed at the same level when we used higher RelB concentrations (Fig. 1A). This might be due to the fact that RelB formed heterodimers with newly synthesized p50 and thereby contributed positively to the reporter gene expression, whereas the inhibitory function is due to binding of RelA. A clearer result was obtained upon transient transfection of individual NF-〉/Rel proteins such that high RelA-induced reporter gene expression was impaired by addition of RelB, whereas cotransfection of p50 and RelB had a positive effect (Fig. 1B). Similar results were observed in a previous study analyzing the differential control of the Bcl-x(L) promoter and the c-myc 〉 site in NIH3T3 cells by overexpression of the cytomegalovirus immediate-early protein 1 (31). It was reported that immediate-early protein 1 specifically activated RelB⅐p50 heterodimers, which promoted expression of the c-myc 〉-site-driven reporter construct but not the Bcl-x(L) promoter controlled reporter. Furthermore, overexpression of RelB inhibited the expression of the Bcl-x(L) gene induced by exogenously expressed RelA.
Another potential molecular mechanism by which RelB might affect the inducible NF-〉 activity was identified in studies using RelB-deficient fibroblasts. In the absence of RelB the basal as well as the LPS-induced IKK activity and consequently I〉␣ degradation were augmented (16). However, we were not able to detect a negative role of RelB on TNF␣-induced FIG. 7. NF-B activity increases the RelA⅐RelB complex formation. A, 70Z/3 pre-B cells were either left untreated or stimulated with PMA (100 ng/ml) or with LPS (10 g/ml) for the times indicated. Whole cellular extracts were prepared and used for immunoblot analysis with antibodies specific for RelA or RelB. Samples of the same extracts (500 g) were used for immunoprecipitation experiments using anti-RelA. The immunoprecipitates were separated by SDS-PAGE, transferred to a PVDF membrane, and then subjected to immunoblot analysis using anti-RelB and, after stripping of the membrane, anti-RelA. Bands specific for the immunoglobulin heavy chain (IgH) and nonspecific bands (n.s.) are indicated. B, thymocytes from 8-week-old control mice (WT) or heterozygote transgenic mice expressing a constitutively active IK⌲␤ (CA) were isolated, and whole cell extracts were prepared. Immunoprecipitations were then performed as described in A. As control similar IPs were performed using non-immune rabbit serum (data not shown). The expression of RelA, RelB, or the FLAG-tagged IK⌲␤ (CA) in samples prior to immunoprecipitation was examined by immunoblot analysis using the indicated antibodies (lanes 3 and 4). I〉-degradation by overexpression of RelB in murine embryonic fibroblasts (Fig. 4). Furthermore, the results from our reporter gene assays with cotransfected NF-〉 factors using 293 HEK cells together with the DNA binding studies using either in vitro translated RelA and RelB proteins or nuclear extracts from RelB-overexpressing cells strongly suggest that RelB inhibits the activity of RelA by affecting its DNA-binding activity. Moreover, our results imply that inhibition of RelA activity occurs through its sequestration in RelA⅐RelB heterodimers. To our knowledge this is the first demonstration that RelB is a complex partner of RelA.
Two possible explanations may exist for the previous failure to detect RelA⅐RelB heterodimers. These are (a) RelA⅐RelB heterodimers do not bind efficiently to conventional 〉 sites and (b) the amount of RelA⅐RelB heterodimers is low compared with RelB⅐p52 and RelB⅐p50 heterodimers. However, activation of NF-〉, by either a constitutively active IK⌲␤ or stimulation with an NF-〉 agonists like a phorbolester or LPS, led to a significant increase in the RelA⅐RelB complex formation in 70Z/3 pre B cells (Fig. 7, A and B). Consistent with these observations are the recent findings that the relb gene itself is regulated by NF-〉 and its expression is induced by most of the known NF-〉-inducing agonists. In addition, a substantial increase in cellular RelB protein levels has been reported in B cells and pre-B cells stimulated by CD40 or IL-7 withdrawal, respectively (32,33). Currently, the biological function of a RelA⅐RelB heterodimer remains unclear, but because we were unable to detect DNA binding of these complexes in vitro, we conclude that they are transcriptionally inactive through a failure to associate with DNA. Recently, we described a signalinduced proteolytic degradation of RelB in T cells stimulated with anti-CD3/CD28 antibodies or phorbolester and calcium ionophore. This degradation process would therefore remove RelA⅐RelB heterodimers and would release RelA (or p50 and p52) as active NF-〉 factor(s), thus supporting the possibility that the biological function of heterodimer formation between RelB and RelA is the inhibition of the RelA activity (17). Nevertheless, the low amounts of RelA⅐RelB heterodimers and the much higher efficiency of the classic I〉 proteins to inhibit RelA argue against this possibility. Alternatively, because variations in the composition of binding sites for the distinct combinations of NF-〉 family members have been described previously (3,4), it is possible that RelA⅐RelB heterodimers bind to a DNA sequence separate from the known 〉 sites. A change in the composition of the NF-〉 factors after time occurs after treatment with various inducers such as the stimulation of resting mature B cells with anti-Ig antibodies or CD40 ligand (35). Thus in unstimulated B cells, active NF-B is composed of RelB and c-Rel heterodimers, whereas, during stimulation with both agonists for 2 h, RelA heterodimers are induced followed by a signal-specific loss of active RelB after 24 h in the case of the CD40 stimulation. Interestingly, only the DNA binding, but not the nuclear localization of RelB, seemed to be affected (35). Interestingly RelA⅐RelB heterodimers do not seem to be controlled by the NF-〉 inhibitors I〉␣ or I〉␤ in fibroblasts (Fig.  5B). This is consistent with the recent finding that p100 and to a lesser extent p105 are the only I〉 proteins complexed with RelB in this cell type (26). Because we found only a portion of RelA⅐RelB heterodimers in the nucleus, comparable to the amount of nuclear RelA⅐p50 heterodimers, the question of how the cytoplasmic retention of RelA⅐RelB heterodimers is achieved still remains. The control of RelB-containing dimers in general has always been a matter of discussion. For example, RelB⅐p50 heterodimers are only weakly if at all controlled by I〉 proteins, but the majority of RelB⅐p50 heterodimers was found in the cytoplasm. It is still unclear if RelB is complexed with p105 instead of p50 in this compartment or if another unidentified mechanism is responsible for this effect. The molecular mechanisms underlying the different RelB functions are versatile and highly cell type-dependent. It was therefore important to analyze both lymphoid as well as non-lymphoid cells. Interestingly, we detected RelA⅐RelB heterodimers in both cell types. However, because RelB has been demonstrated to bind to IB proteins in lymphoid cells, the control of RelA⅐RelB heterodimers might be different in lymphoid versus non-lymphoid cells.
As mentioned earlier, the expression of RelB is controlled by NF-〉 (6), and therefore an increase in the protein levels of RelB in thymocytes overexpressing a constitutively active isoform of IKK␤ is not surprising. In addition, increased cellular RelB levels have been reported during B cell development (33,34). Taken together, our data describe a new feature of the NF-〉/Rel family member RelB, which helps explain its repressive role on NF-〉 activity. Clearly, other molecular mechanism such as blocking of 〉 sites, sequestering of p50 or p52 in heterodimers, as well as epigenetic effects through the inability to recruit transcriptional cofactors might also contribute for the repressive role of RelB. Nevertheless, our findings clearly demonstrate that complex formation between RelB and RelA inhibits RelA activity. Because the regulation of NF-〉-dependent gene expression is so versatile, such a repression of RelA in one context might redirect NF-〉 to induce a different set of genes and therefore potentially represents a mechanism to switch gene expression patterns during an NF-〉-dependent cellular response.