BCL-3 and NF-κB p50 Attenuate Lipopolysaccharide-induced Inflammatory Responses in Macrophages*

Lipopolysaccharide (LPS) induces expression of tumor necrosis factor α (TNFα) and other pro-inflammatory cytokines in macrophages. Following its induction, TNFα gene transcription is rapidly attenuated, in part due to the accumulation of NF-κB p50 homodimers that bind to three κB sites in the TNFα promoter. Here we have investigated the inhibitory role of BCL-3, an IκB-like protein that interacts exclusively with p50 and p52 homodimers. BCL-3 was induced by LPS with delayed kinetics and was associated with p50 in the nucleus. Forced expression of BCL-3 suppressed LPS-induced transcription from the TNFα promoter and inhibited two artificial promoters composed of TNFακB sites that preferentially bind p50 dimers. BCL-3-mediated repression was reversed by trichostatin A and was enhanced by overexpression of HDAC-1, indicating that transcriptional attenuation involves recruitment of histone deacetylase. Analysis of macrophages from p50 and BCL-3 knock-out mice revealed that both transcription factors negatively regulate TNFα expression and that BCL-3 inhibits IL-1α and IL-1β. In contrast, induction of the anti-inflammatory cytokine IL-10 was reduced in BCL-3 null macrophages. BCL-3 was not required for the production of p50 homodimers but BCL-3 expression was severely diminished in p50-deficient cells. Together, these findings indicate that p50 and BCL-3 function as anti-inflammatory regulators in macrophages by attenuating transcription of pro-inflammatory cytokines and activating IL-10 expression.

Macrophages are a primary source of the pro-inflammatory cytokine TNF␣, 1 which is expressed in response to viral or bacterial infection and other inflammatory stimuli. TNF␣ evokes numerous responses in immune cells and surrounding tissues, such as expression of ICAM-1 (intercellular adhesion molecule-1) on the surface of epithelial cells to promote adhe-sion and infiltration of neutrophils and monocytes (1,2). TNF␣ also induces the secretion of proteinases that degrade intercellular matrix proteins and allow macrophages to penetrate tissue to the site of injury or infection (3) and stimulates macrophages to produce other pro-inflammatory cytokines that modulate immune functions.
Although TNF␣ plays a beneficial role by rapidly activating immune cells and eliciting other cellular responses that protect the host from infection and injury, prolonged or elevated expression of TNF␣ is associated with a number of pathological conditions, including chronic inflammation (4), septic shock (5), cachexia (3), and autoimmune diseases such as rheumatoid arthritis (6,7). The potentially deleterious effects of TNF␣ require stringent control over its expression. Previously we described a mechanism that attenuates LPS-induced transcription of the TNF␣ gene in macrophages (8). LPS-stimulated macrophages were found to secrete a soluble activity called TNF␣ inhibitory factor (TIF) that suppressed the induction of TNF␣ mRNA by endotoxin. In addition to inhibiting TNF␣ expression, TIF caused the selective nuclear accumulation of NF-B p50 homodimers. NF-B p50, a member of the Rel family of transcription factors, lacks transcriptional activation domains and is capable of inhibiting transcription of target genes (9). p50 dimers accumulate with delayed kinetics in macrophages following LPS stimulation and preferentially bind to three B elements in the murine TNF␣ promoter (B1, B2a, and B3). Binding of p50 dimers appears to play a key role in attenuating TNF␣ gene transcription (8,10,11). This notion is further supported by the fact that a genetic polymorphism that disrupts binding of p50 dimers to the B1 site in the human TNF␣ gene causes increased transcription from this promoter (12).
Although these findings implicate p50 as an inhibitory factor, the mechanism by which p50 represses TNF␣ gene transcription has not been fully elucidated. In particular, it is unclear whether other transcription factors or accessory proteins are required for p50-mediated repression of the TNF␣ gene. In the present study we investigated the role of BCL-3 in attenuating TNF␣ gene transcription. BCL-3 is a predominantly nuclear member of the IB family of NF-B inhibitors whose defining structural feature is the ankyrin repeat motif, which mediates interactions with NF-B dimers (9). BCL-3 associates with p50 and p52 homodimers but not with other NF-B dimers (13,14). In contrast to the cytoplasmic IBs, which are degraded in response to many inducing signals, BCL-3 does not undergo regulated proteolysis. BCL-3 is constitutively expressed in some cells, while in others it is induced by extracellular factors. For instance, BCL-3 levels are increased by mitogens in lymphoid cells (15,16), by IL-4 in T cells (17), by GM-CSF and erythropoietin in erythroid precursor cells (18), * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. § Present address: KaloBios Pharmaceuticals, Inc., Palo Alto, CA 94304.
The role of BCL-3 in modulating NF-B activity has been controversial. In transient transfection assays BCL-3 can either positively (13,18,20) or negatively (19,21) regulate transcription of reporter genes via NF-B sites. It has been proposed that BCL-3 activates transcription by associating with p50 or p52 dimers on promoters, thereby providing a transactivation domain to the NF-B complex (13,20). BCL-3 was also reported to inhibit transcription by enhancing the binding of repressive p50 dimers to DNA (22) or by promoting the assembly of p50 homodimers and facilitating their translocation to the nucleus (23). Mice lacking BCL-3 display multiple immunological defects, including impaired production of antigenspecific antibodies and increased sensitivity to bacterial pathogens (24,25). These phenotypes partially overlap with those observed in p50 knock-out animals (26), and doubly mutant mice display enhanced defects (25). The genetic data thus support a functional relationship between p50 and BCL-3 in vivo and underscore the importance of these proteins in regulating immune responses.
We anticipated that BCL-3 might play a role in TNF␣ gene repression, acting either to increase the levels of nuclear p50 homodimers or to augment their inhibitory activity. We show here that BCL-3 is induced in macrophages by LPS with delayed kinetics and attenuates LPS-induced TNF␣ gene transcription. Analysis of p50 Ϫ/Ϫ and BCL-3 Ϫ/Ϫ macrophages demonstrates that both proteins negatively regulate expression of TNF␣ and several other pro-inflammatory cytokines. BCL-3 also positively regulates expression of the anti-inflammatory cytokine, IL-10. Thus, p50 and BCL-3 function coordinately to limit the pro-inflammatory response of activated macrophages.
Plasmid Constructs-Reporter constructs TNF-luc and (Ϫ514)TNFluc and artificial NF-B promoter-reporter constructs containing four concatomerized copies of individual NF-B sites have been described elsewhere (8). Expression plasmids for p50 and p65 (Rc/CMV-p50 and Rc/CMV-p65) were described previously (8). An expression vector for BCL-3 (obtained from T. W. McKiethan) was derived from the human BCL-3 gene originally cloned into pBluescript. The BCL-3 coding sequence was placed under the control of the CMV promoter in the pcDNA3.0 expression vector (Rc/CMV-BCL-3). The FLAG-tagged pBJ5-HDAC1 vector was provided by C. A. Hassig.
Antibodies-Rabbit antisera against p50 and p65 have been described previously (27,28). A rabbit polyclonal antiserum recognizing BCL-3 (number 1348; kindly provided by N. Rice) was raised against a peptide corresponding to the NH 2 terminus of the human BCL-3 protein (NH 2 -DEGPVDLRTRPKAAC), where the COOH-terminal Cys residue was added to facilitate coupling of the peptide to keyhole limpet hemocyanin. The murine BCL-3 sequence is identical in this region except for the AA dipeptide at the COOH terminus, and the antiserum crossreacts with mouse BCL-3. Fluorescein isothiocyanate-labeled IgG2a and MAC-1 (Pharmingen) and F4-80 (Serotec) antibodies were used for FACS analysis of macrophage cell surface markers. Antibody against human HDAC-1 was obtained from Upstate Biotechnology.
Primary Bone Marrow Macrophages and Macrophage Cell Lines-Primary macrophages and macrophage cell lines were generated using bone marrow from mice lacking p50 (26) (The Jackson Laboratory) or BCL-3 (24) and their wild-type littermates. Bone marrow cells were harvested from femurs and tibias and placed in phosphate-buffered saline (PBS) containing 2% FBS. The cells were disaggregated by passage through an 18-gauge needle and pelleted at 1000 rpm for 5 min, and red blood cells were lysed using NH 4 Cl lysis solution (Sigma). After 5 min, 45 ml of PBS containing 2% FBS was added, and the cells were pelleted. For primary bone marrow macrophages, the cells were cultured in DMEM supplemented with 10% FBS and 100 ng/ml M-CSF for 24 -36 h. The non-adherent cells were collected and seeded at 1 ϫ 10 7 cells per 15-cm 2 dish. The primary cells were cultured in DMEM, 10% FBS containing 100 ng/ml M-CSF for 4 -5 days to obtain adherent, differentiated macrophages.
To generate immortalized macrophage cell lines, 2 ϫ 10 7 bone marrow cells were resuspended in 5 ml CREJ2 cell supernatant (a source of the J2 transforming retrovirus; Ref. 29) with 500 l of 500 g/ml Polybrene (Sigma) and 5 l of 10 6 units/ml GM-CSF (Peprotech). After 24 h at 37 C, the supernatant was removed and the cells were grown in DMEM, 10% FBS containing GM-CSF for 7 days. GM-CSF was then withdrawn and the cells were cultured further in DMEM, 10% FBS to establish growth factor-independent cell lines.
FACS Analysis-FACS analysis of cell surface markers was performed as follows. 10 6 cells in 50 l of phosphate-buffered saline supplemented with 0.1% bovine serum albumin were incubated for 10 min on ice with 1 g of 2.4G2 antibody (Pharmingen) to quench the Fc receptor present on the cell surface. The cells were washed two times with PBS, 0.1% bovine serum albumin and were resuspended in 50 l of PBS, 0.1% bovine serum albumin. Cells were aliquoted into a 96-well V-bottom plate (10 6 cells per well) as appropriate and incubated with 1 g of FITC-labeled IgG2a control, MAC-1, or F4-80 antibodies at 4°C for 15 min. After two additional washes, the cells were fixed with 200 l of 1% paraformaldehyde and analyzed by FACS.
Transfection Assays-RAW264.7 macrophages were transfected with 0.6 -1.0 g of reporter construct, 180 ng of Renilla luciferase vector (pRL-TK; Promega), and, where appropriate, 0.5 g of Rc/CMV (control), Rc/CMV-BCL-3, and/or 1 g of pBJ5-HDAC1-FLAG (HDAC1) using a DEAE-dextran sulfate protocol as described (8). Where indicated, cells were treated with 5 M TSA ϳ16 -18 h prior to LPS treatment. Forty hours after transfection the cells were treated with LPS and cell extracts prepared over a time course. The lysates were analyzed for luciferase activity using the enhanced luciferase assay kit (Analytical Luminescent Laboratory) or the dual-luciferase reporter assay system (Promega). Luciferase expression was normalized to either total protein concentration or Renilla luciferase activity.
Immunoprecipitation-Antibodies were coupled to protein-A-Sepharose beads for immune precipitation assays. 100 l of serum in 2 ml of PBS was incubated with 0.2 ml of protein-A-Sepharose for 1 h at room temperature. The beads were washed twice with 0.2 M Na 2 B 4 O 7 (pH 9.0) and resuspended in 2 ml 0.2 M Na 2 B 4 O 7 , and the antibodies were cross-linked by adding dimethylpimelimidate (0.5 mg/ml) and incubating at room temperature for 1 h. The beads were then washed once with 2 ml of 0.2 M Na 2 B 4 O 7 and incubated for 2 min with 0.5 M (NH 4 ) 2 SO 4 (pH 5.5). The protein-A-Sepharose resin was washed twice with 2 ml of sodium citrate (pH 2.5), twice with 2 ml of PBS, once with 100 mM glycine (pH 3.0), and once with 2 ml of PBS. The antibody-conjugated beads were stored in PBS containing 0.02% sodium azide. Beads (10 l) coupled to anti-p50, anti-BCL-3, or normal rabbit serum were incubated with 100 g nuclear extract at 4°C for 16 h. The beads were pelleted and washed twice with Buffer C. Precipitated proteins were extracted by incubating twice with 100 l gentle elution buffer (Pierce). The eluted proteins were precipitated with trichloroacetic acid and resuspended in 10 l of TE (10 mM Tris, 1 mM EDTA), mixed with sample buffer, and analyzed on 12% SDS-polycrylamide gels. Proteins were transferred to Immobilon and probed with anti-p50 antibody.
RNase Protection Assays-Total RNA was extracted from cells using TRIzol reagent (Invitrogen), and 5-6 g of RNA was used for each RNase protection assay. Probe sets were labeled using the Riboquant in vitro assay kit (Pharmingen), and protection assays were performed using the Riboquant RNase protection assay Kit (Pharmingen) as recommended by the supplier. Samples were analyzed on precast 6% polyacrylamide sequencing gels using the Novex Quickpoint system. Protected fragments corresponding to specific cytokine transcripts were quantitated using the Storm 860 PhosphorImager (Molecular Dynamics) and the data normalized to glyceraldehyde-3-phosphate dehydrogenase expression.

BCL-3 Expression Is Induced by LPS in Macrophages-
We investigated whether BCL-3 expression is activated by LPS in macrophages and thus may be involved in regulating TNF␣ gene transcription. RAW264.7 macrophages were treated with LPS and BCL-3 levels in nuclear and cytoplasmic extracts were analyzed over a time course by Western blotting. Prior to LPS treatment, little or no BCL-3 was detected in nuclear extracts (Fig. 1A, 0 h). Exposure to LPS increased BCL-3 expression after an initial delay, reaching maximal induction by 12 h. Cytoplasmic expression of BCL-3 was also induced by LPS with similar kinetics (Fig. 1A, middle panel). Two major forms of BCL-3 (migrating at ϳ55 and ϳ60 kDa, respectively) were observed in the Western blots. Phosphatase experiments indicate that the slower migrating species is a hyperphosphorylated form (data not shown), in accordance with previous studies showing that BCL-3 is a phosphoprotein (14,22,32). We also examined the kinetics of NF-B p50 and p65 induction (Fig. 1A, lower  panel). p50 and p65 appeared in the nucleus at the earliest LPS time point (1 h); p65 levels declined steadily after 2 h, whereas nuclear p50 continued to accumulate with time. Thus, LPSinduced expression of p50 is prolonged in RAW264.7 cells, consistent with our earlier results from IC-21 macrophages (8).
Previous studies have shown that BCL-3 and p50 interact in vivo (13,14,20,23,33). Since BCL-3 and p50 are present in nuclear extracts from RAW264.7 cells exposed to LPS for several hours, we asked whether these two proteins are physically associated. Nuclear extracts from untreated and LPS-stimulated macrophages were immunoprecipitated with antibodies specific for BCL-3 or p50, and the precipitated proteins were examined for the presence of p50 by Western blotting (Fig. 1B). In unstimulated cell extracts (i.e. lacking BCL-3) p50 was not co-immunoprecipitated by the BCL-3 antibody (lane 3). However, BCL-3⅐p50 complexes were detected in nuclear extracts from LPS stimulated macrophages, as shown by the presence of p50 in BCL-3 immunoprecipitates (lane 6). Thus, at least a portion of the nuclear BCL-3 in LPS-stimulated cells is associated with p50.
Ectopic Expression of BCL-3 Represses the TNF␣ Promoter-The appearance of p50⅐BCL-3 complexes several hours after LPS induction suggested that BCL-3 might play a role in attenuating TNF␣ expression. To investigate whether BCL-3 negatively regulates TNF␣ transcription, we transfected RAW264.7 macrophages with a TNF␣ promoter-luciferase reporter plasmid (TNF-luc) (8) alone or with increasing amounts of BCL-3 expression vector. Luciferase activity was then examined over a time course after LPS stimulation ( Fig. 2A). Overexpression of BCL-3 suppressed LPS-induced transcription from the promoter in a dose-dependent manner, inhibiting luciferase activity by ϳ50% at the highest level of BCL-3 examined.
The TNF␣ promoter contains three NF-B sites (B1, B2a, and B3) that preferentially bind p50 dimers (8). We tested whether a 5Ј deletion mutant lacking these sites ((Ϫ514)TNFluc) (8) retained susceptibility to BCL-3 repression (Fig. 2B). LPS-induced transcription from this truncated promoter in the absence of BCL-3 was diminished compared with the complete promoter (7-fold versus 35-fold, respectively). Notably, BCL-3 failed to repress, and in fact modestly increased, transcription from the Ϫ514 construct (Fig. 2B, right panel). Therefore, the ability of BCL-3 to inhibit TNF␣ promoter activity requires the distal region that contains NF-B sites with high affinity for p50 dimers.
To further examine whether specific NF-B sites mediate BCL-3 repression, we tested the ability of BCL-3 to repress transcription from artificial promoters bearing individual TNF␣ B motifs. Reporter constructs containing four copies of the B2a or B3 sites inserted upstream of the minimal TK promoter (8) were transfected into RAW264.7 cells, without or with the BCL-3 expression vector. Luciferase expression was then analyzed following LPS stimulation. BCL-3 significantly inhibited transcription from both of these constructs (Fig. 2C,  left and middle panels). However, a reporter construct containing the Ig-NF-B element, a site that preferentially binds p50⅐p65 heterodimers (8), was not appreciably affected by BCL-3 (Fig. 2C, right panel). These experiments support the notion that BCL-3 represses TNF␣ gene transcription via p50 dimers bound to the distal B elements in the promoter.
Attenuation of LPS-induced TNF␣ Expression Is Impaired in p50 and BCL-3-deficient Macrophages-BCL-3 and p50 null mice have been generated and display multiple immune disorders (24 -26). To examine the roles of these proteins in regulating TNF␣ expression in macrophages, we derived immortalized macrophage cell lines from mutant or control bone marrow cells by infecting them with the J2 retrovirus, which carries the v-myc and v-raf oncogenes (29). The infected cells were cultured in the presence of GM-CSF to promote macrophage differentiation and survival. After 7 days, GM-CSF was withdrawn, and the cells were further cultured to select for growth

FIG. 1. BCL-3 expression is induced by LPS in macrophages.
A, RAW264.7 macrophages were treated with 1 g/ml LPS and nuclear and cytoplasmic extracts were prepared over a 12-h time course. Twenty g of nuclear extract and 50 g of cytoplasmic extract were assayed for BCL-3 expression by Western blotting. Twenty g of nuclear extract were assayed for NF-B p50 and p65 levels by Western blotting. B, nuclear extracts from LPS-treated RAW264.7 cells were analyzed for the presence of BCL-3⅐p50 complexes by immunoprecipitation (IP). Extracts from control (0 h) or LPS-induced (24 h) cells were immunoprecipitated with normal rabbit serum (NRS), anti-BCL-3 antibody, or anti-p50 antibody coupled to protein-A-Sepharose. Immunoprecipitates were assayed for p50 by Western blotting.
We examined cytokine expression in these the cell lines by stimulating them with LPS, harvesting RNA over a time course and analyzing TNF␣ mRNA levels using a quantitative RNase protection assay (RPA). TNF␣ mRNA was induced to considerably higher levels in p50 Ϫ/Ϫ M⌽ compared with ϩ/ϩ cells (Fig.  3A, left panel). In addition, its expression was prolonged, since TNF␣ levels remained significantly elevated in p50-deficient cells at 24 h post-induction, whereas in wt cells they had returned to base line. Induction of TNF␣ mRNA in BCL-3 Ϫ/Ϫ M⌽ was similar to that of wt cells at 1 and 4 h (Fig. 3A, right  panel). However, TNF␣ expression in BCL-3 Ϫ/Ϫ cells was not efficiently attenuated, remaining elevated at the 8-, 12-, and 24-h time points. These results suggest that both BCL-3 and p50 contribute to the down-regulation of TNF␣ expression after transient induction by LPS.
BCL-3 Positively Regulates IL-10 Expression-Macrophages express pro-inflammatory cytokines as well as anti-inflammatory factors to modulate immune responses, and IL-10 is a major anti-inflammatory cytokine produced by LPS-activated macrophages (34). We therefore investigated whether BCL-3 affects IL-10 mRNA expression. As shown in Fig. 3B, IL-10 mRNA levels were reduced in LPS-treated BCL-3 Ϫ/Ϫ M⌽ compared with ϩ/ϩ cells (3-fold maximal induction versus 7-fold, respectively). By contrast, IL-10 induction in p50 Ϫ/Ϫ M⌽ was similar to that in ϩ/ϩ cells (data not shown). These findings show that BCL-3 is required for efficient induction of the IL-10 gene in LPS-stimulated macrophages, in contrast to its negative effect on TNF␣ expression.
Effects of BCL3 Deficiency on Cytokine Expression in Primary Bone Marrow Macrophages-To verify that our results were not influenced by the use of transformed cell lines and to confirm the effects of BCL-3 on cytokine gene expression, we  also examined cytokine transcription in primary bone marrow macrophages from wt and BCL-3 null mice. Primary cells were first examined for LPS-mediated induction of BCL-3 by West-ern blotting (Fig. 4A). Nuclear levels of BCL-3 were strongly increased by LPS in wt macrophages but not in BCL-3 Ϫ/Ϫ cells. The induction of cytokines by LPS was examined using RPA  (Fig. 4B). Comparison of wt and mutant cells shows that BCL-3 inhibits expression of two pro-inflammatory cytokines (TNF␣ and IL-1␤; also IL-1␣, data not shown). In addition, BCL-3 null cells showed reduced induction of the anti-inflammatory cytokine IL-10, again indicating that BCL-3 positively regulates expression of this gene. The results in primary macrophages therefore corroborate the data from immortalized macrophage cell lines.
Cross-regulation of p50 and BCL-3-A previous study suggested that BCL-3 can promote the formation and nuclear transport of p50 homodimers (23). Thus, BCL-3 could potentially repress TNF␣ gene transcription by increasing the levels of nuclear p50 homodimers in LPS-activated macrophages. To examine whether BCL-3 influences the production of p50 homodimers, we used EMSA to compare p50 dimers in nuclear extracts from immortalized BCL-3 Ϫ/Ϫ and ϩ/ϩ M⌽ that had been treated with LPS. As shown in Fig. 5A, equivalent levels of p50 homodimers accumulated in BCL-3 Ϫ/Ϫ and wt cells. Moreover, p50 dimers were induced with similar kinetics in both cell lines. Therefore, the prolonged induction of p50 homodimers in LPS-stimulated macrophages is not dependent on BCL-3.
Conversely, we investigated whether p50 deficiency affects BCL-3 expression. BCL-3 levels were analyzed in extracts prepared from wt and p50 Ϫ/Ϫ M⌽ cell lines after LPS stimulation (Fig. 5B). Notably, induction of BCL-3 was greatly reduced in both cytoplasmic and nuclear extracts from p50 Ϫ/Ϫ M⌽ compared with wt cells. A minor amount of BCL-3 protein was detected in the mutant cells at 24 h, but the level was much lower than that observed in ϩ/ϩ M⌽, especially in the nuclear fraction. These results suggest that either p50 is required for transcription of the BCL-3 gene or association with p50 stabilizes the BCL-3 protein and may also promote its nuclear localization. Although the mechanism remains to be elucidated, these findings demonstrate that induction of BCL-3 by LPS is strongly dependent on p50 function.
Repression of the TNF␣ Promoter by BCL-3 Requires Histone Deacetylase Activity-Active repression of transcription often involves the recruitment of histone deacetylases (HDACs) to target promoters by repressor proteins (35). To investigate whether HDACs participate in TNF␣ gene repression, we first asked whether BCL-3-mediated inhibition of the promoter is enhanced by ectopic expression of HDAC-1. The TNF-luc reporter was transfected into RAW264.7 cells together with expression vectors for BCL-3 and/or HDAC-1, and luciferase activity was assayed over a time course following LPS treatment (Fig. 6A). Expression of BCL-3 or HDAC-1 individually suppressed TNF␣ promoter activity and together caused a further decrease in transcription. BCL-3 slightly increased transcription from the truncated promoter lacking upstream B sites ((Ϫ514)TNF-luc), while HDAC-1 alone or BCL-3 plus HDAC-1 had little effect. We also examined the effect of treating cells with a specific HDAC inhibitor, TSA (Fig. 6B). TSA dramatically reversed BCL-3-mediated inhibition of the TNF-luc reporter, increasing luciferase activity to a level ϳ2-fold over that of the control (i.e. no BCL-3). In contrast, TSA had almost no stimulatory effect on the (Ϫ514)TNF-luc reporter co-transfected with BCL-3. These results support the idea that p50⅐BCL-3 complexes recruit HDAC-1 (or a related protein) to specific B sites in the TNF␣ promoter and thereby repress transcription.
We analyzed BCL-3 levels in transfected RAW264.7 cells by Western blotting. Transfection of the BCL-3 vector caused only a minor increase in the level of BCL-3 detected by immunoblotting, with or without LPS stimulation (Fig. 6C), probably because the transfection efficiency of these cells is low and the endogenous BCL-3 signal predominates. Interestingly, BCL-3 levels were significantly increased when BCL-3 was co-expressed with HDAC-1, whether or not the cells were treated with LPS (compare lanes 2 and 3 and lanes 5 and 6). One possible explanation is that HDAC-1 stabilizes BCL-3 as a result of physical association between these two proteins.
To determine whether endogenous BCL-3 and HDAC-1 interact in cells, we prepared nuclear extracts from RAW264.7 cells that had been stimulated with LPS or left untreated and performed immunoprecipitation assays using BCL-3 or control antibodies. The immunoprecipitates were analyzed for the presence of HDAC-1 by Western blotting (Fig. 6D). An association between HDAC-1 and BCL-3 was evident in cells stimu-lated with LPS for 6 h but not in cells treated for 2 h or left untreated. The kinetics of complex formation are consistent with our earlier observation that BCL-3 is not appreciably induced in RAW264.7 cells until 4 h following LPS treatment (Fig. 1A). Thus, these results demonstrate that the endogenous BCL-3 and HDAC-1 proteins form a complex in LPS-stimulated cells, involving either direct or indirect interactions. The BCL-3⅐HDAC-1 complex may be responsible for the transcriptional attenuation of TNF␣ and other pro-inflammatory cytokine genes. DISCUSSION Our results demonstrate that BCL-3 is induced in macrophages by LPS and functions to attenuate pro-inflammatory cytokine gene transcription. LPS treatment does not immediately elicit BCL-3 expression; instead, the protein appears with delayed kinetics. BCL-3 induction occurs well after the acute activation of p65⅐p50 dimers and around the time that p50 homodimers begin to predominate over p50⅐p65 heterodimers, and TNF␣ transcript levels start to diminish (8). These kinetics, together with the fact that p50 overexpression inhibits transcription from the TNF␣ promoter (8,10,11), suggested that BCL-3 might negatively regulate the TNF␣ gene through its association with p50 homodimers. In support of this idea, overexpression of BCL-3 suppressed LPS-induced transcription from the TNF␣ promoter in reporter assays. Notably, BCL-3 did not inhibit, and even enhanced, the activity of a truncated promoter lacking the three NF-B sites that preferentially bind p50 homodimers. BCL-3 also repressed two artificial promoter constructs containing multiple copies of these p50 sites but did not inhibit a construct containing the Ig element, which preferentially binds p65⅐p50 heterodimers. Combined with the observation that BCL-3 and p50 physically interact in the nuclei of LPS-stimulated macrophages, these findings support the idea that BCL-3 inhibits TNF␣ transcription via (p50) 2 ⅐BCL-3 complexes bound to specific B sites on the TNF␣ promoter.
Analysis of macrophages from knock-out mice provides further evidence that p50 and BCL-3 inhibit pro-inflammatory cytokine expression. TNF␣ transcripts were induced to considerably higher levels in p50 null cells than in wt cells. This elevated expression was also prolonged, as TNF␣ transcripts remained nearly 20-fold above basal levels at 24 h post-induction, whereas in wt cells expression had returned to background levels. In addition, IL-1␣ mRNA levels were at least 5-fold higher in p50 Ϫ/Ϫ cells compared with wt cells (data not shown), indicating that p50 also represses this gene. It is notable that p50 is completely dispensable for LPS-induced transcription of several pro-inflammatory cytokine genes. This finding is surprising given that p50⅐p65 is the predominant NF-B species induced by LPS and is believed to be the principal activator of cytokine genes. Evidently other NF-B dimers (perhaps p65 homodimers) can substitute for p50⅐p65 complexes to induce cytokine gene expression.
The effect of BCL-3 deficiency on TNF␣ transcription was less pronounced than that seen in p50 Ϫ/Ϫ cells, and there was no discernable phenotype during the induction phase. However, attenuation of TNF␣ expression was impaired, as shown by the elevated levels of TNF␣ mRNA in BCL-3 Ϫ/Ϫ cells during the 8 -24-h interval. Thus, BCL-3 plays a significant role in down-regulating TNF␣ expression. The inhibitory effect of p50 on TNF␣ gene transcription (as determined by comparing transcript levels in wt and p50 Ϫ/Ϫ cells) occurs during both the LPS induction phase, prior to significant BCL-3 expression, and the attenuation phase, when BCL-3 accumulates. We propose that p50 dimers can inhibit TNF␣ gene transcription in the absence of BCL-3 but that BCL-3 augments p50 repression. Our results indicate that HDAC activity is involved in BCL-3-dependent repression of TNF␣, and co-immunoprecipitation assays demonstrate that HDAC-1 is in a nuclear complex with BCL-3 in LPS-stimulated cells. A previous study showed that nuclear p50⅐HDAC-1 complexes are present in unstimulated T cells and, upon stimulation, are replaced by p65 NF-B complexes associated with the CBP co-activator (36). Our findings suggest that LPS stimulation of macrophages causes transient induction of activating NF-B complexes, which are then gradually replaced by p50 dimers associated with BCL-3 and HDAC-1 (or a related protein) to attenuate transcription of specific Bdriven promoters.
It has been reported that nuclear levels of p50 homodimers also become elevated in LPS tolerized macrophages, a situation in which cells exposed to low levels of LPS become resistant to subsequent LPS challenge (10,11,37,38). LPS-tolerant macrophages show greatly reduced inducibility of TNF␣ and other pro-inflammatory cytokines. Notably, macrophages from p50 null mice are refractory to LPS tolerization and are capable of inducing TNF␣ mRNA upon secondary challenge with endotoxin (10). These results are consistent with our finding that TNF␣ transcription is prolonged in p50-deficient cells.
Although the expression of several cytokines was increased in BCL-3 Ϫ/Ϫ macrophages, a notable exception was IL-10. Induction of IL-10 by LPS was diminished at least 2-fold in BCL-3 Ϫ/Ϫ primary macrophages and cell lines, showing that BCL-3 positively regulates this gene. The transcription factors that regulate the IL-10 gene have not been thoroughly characterized. However, an Sp1 site in the human IL-10 promoter is critical for its activation by LPS in monocytes/macrophages (39 -41) and a role for Stat3 in regulating the human gene has also been reported (42). Since IL-10 induction was not impaired in p50-deficient macrophages (data not shown), BCL-3 presumably acts via another DNA-binding protein to activate this promoter.
Nuclear p50 homodimers were induced by LPS with equivalent kinetics and to similar levels in wt and BCL-3 Ϫ/Ϫ macrophages. Thus, BCL-3 does not regulate the synthesis or nuclear translocation of p50 dimers in macrophages, although we cannot exclude such a role for BCL-3 in other cell types, as has been reported (23). Our results support a model in which BCL-3 affects the function but not the expression of p50 dimers. On the other hand, nuclear and cytoplasmic induction of BCL-3 was strongly reduced in p50 Ϫ/Ϫ macrophages. While the underlying mechanism for the impaired expression of BCL-3 is unknown, it is possible that p50 is required for transcriptional regulation of the BCL-3 gene. Consistent with this notion, the BCL-3 promoter was shown to contain two NF-B sites, one of which is required for NF-B-dependent activation of the promoter (43). Alternatively, the BCL-3 protein might be unstable when p50 dimers are unavailable for complex formation.
BCL-3 is probably not activated directly by LPS, and its induction appears to be mediated by an autocrine factor that is secreted in response to LPS. 2 Whether the BCL-3 inducing activity is identical to TIF (defined by its ability to repress TNF␣ expression and induce p50 dimers) (8) or is a distinct factor remains to be clarified. A recent study found that expression of BCL-3 in activated T cells is induced by immunological adjuvants such as LPS and that BCL-3 is involved in suppressing apoptosis (44). Interestingly, BCL-3 induction by LPS in T cells seems to be indirect, and the involvement of other effector cells, particularly antigen-presenting cells, has been proposed (45). Thus, our observation that LPS may induce BCL-3 in macrophages via a soluble autocrine factor could also explain the indirect induction of BCL-3 in T cells.
Although BCL-3 repressed LPS-induced transcription from the TNF␣ promoter in reporter assays, it did not significantly affect promoter activity prior to LPS treatment, and the repressive effects of BCL-3 were apparent only after several hours of LPS stimulation ( Fig. 2A). One explanation for these results is that LPS-induced signals post-translationally regulate BCL-3, in addition to inducing its expression. Since BCL-3 was shown previously to be a phosphoprotein (14,22,32), this regulation could involve a phosphorylation event, elicited directly by LPS signaling or via a secreted factor, that is required for BCL-3 to inhibit transcription. Consistent with this notion, BCL-3 was gradually converted to a slower migrating, hyperphosphorylated form after its induction by LPS or conditioned medium ( Fig. 1 and data not shown). At present, the specific sites of modification are undefined and further studies will be necessary to establish whether phosphorylation plays a role in BCL-3-mediated transcriptional repression.
In summary, we have shown that BCL-3 and NF-B p50 limit the expression of inflammatory mediators in activated macrophages by repressing transcription of pro-inflammatory cytokine genes such as TNF␣, IL-1␣, and IL-1␤. In addition, BCL-3 positively regulates expression of the antiinflammatory IL-10 gene. Thus, BCL-3 and p50 play important roles in modulating the innate immune response to bacterial pathogens.