Discrimination between RelA and RelB Transcriptional Regulation by a Dominant Negative Mutant of IκBα*

RelA and RelB belong to the nuclear factor-κB (NF-κB-Rel) transcription factor family. Both proteins are structurally and functionally related, but their intracellular and tissue distributions are different. In resting cells, RelB is found mostly in the nucleus, whereas RelA is sequestered in the cytosol by protein inhibitors, among which IκBα is the dominant form in lymphocytes. Upon cellular activation IκBα is proteolyzed, allowing RelA dimers to enter the nucleus and activate target genes. To study the selectivity of gene regulation by RelA and RelB, we generated T cell lines stably expressing a dominant negative mutant of IκBα. We show that selective inhibition of RelA-NF-κB decreased induction ofNFKB1, interleukin-2, and interleukin-2Rα genes but not c-myc. Transcription driven by the IκBα promoter was blocked by the transgenic IκBα; however, wild type IκBα was expressed in the transgenic cell clones but with much slower kinetics than that in control cells. Wild type IκBα expression was concomitant with RelB up-regulation, suggesting that RelB could be involved in transcription of IκBα through binding to an alternative site. These results indicate that RelB and RelA have both distinct and overlapping effects on gene expression.

Nuclear factor B (NF-B) 1 is part of the Rel family of eukaryotic transcription factors which share structural and functional properties. Although ubiquitously expressed in higher eukaryots NF-B has been intensively studied mostly in cells belonging to the immune system where it was first discovered (for review, see Refs. 1 and 2). NF-B-Rel factors were shown to participate in the expression of genes essential for the immune responses and to regulate gene transcription during inflammatory reactions. The prototypical NF-B is a homodimer or heterodimer composed of 50-kDa (p50) and/or 65-kDa (p65 or RelA) polypeptides. In vertebrates other members of the family are c-Rel, RelB, and p52. The tissue and cellular distribution of the three last members is more restrained than that of the prototypical NF-B. For example, the expression of RelB was described as being predominant in dendritic cells from primary and secondary lymphoid organs (3)(4)(5)(6). RelB has also been detected in other cells and tissue but in lower amounts or after specific activation. c-Rel and p52 are also expressed mainly in cells from the hematopoietic lineages. P50 and p52 are generated by proteolytic processing of precursor polypeptides (p105 (NFKB1 gene) and p100 (NFKB2 gene), respectively) (1,7). Each member of the NF-B-Rel family contains a 300-amino acid sequence called the Rel homology domain, which is critical for nuclear translocation, protein-protein interactions, and sequence-specific DNA binding. All members of the NF-B-Rel family form dimers. The dimers can be classified into two pools on the basis of their intracellular localization, which is critical in regulating their activity. One pool of NF-B-Rel dimers is cytosolic in the absence of cellular activators, whereas the second pool is constitutively nuclear. The intracellular location of the dimers depends on the capacity of the NF-B-Rel family members to interact with ankyrin repeat-containing proteins, collectively called IB. The cytoplasmic IBs inhibit NF-B-Rel complexes by preventing both NF-B-Rel nuclear translocation, and their interaction with specific decameric DNA sequences called B (8,9). Thus IBs represent intracellular regulators of NF-B activity. Several members of IB regulatory family have been characterized, including IB␣, IB␤, IB␥, the two NF-B protein precursors p105 and p100, and bcl-3. Except for bcl-3, IB molecules are mostly cytosolic, although nuclear IB␣ has been reported in cultured cells (10 -12) and in vivo. 2 p50 and p52 homodimers as well as RelB-p50 and RelB-p52 heterodimers do not interact efficiently with cytosolic IBs. Consequently they are found in nuclei of cells that produce these complexes (13). Therefore their regulation should be distinct from the cytosolic forms of NF-B. The p50 and p52 homodimers were reported to interact with the nuclear bcl-3. The resulting trimers seem to constitute transcriptional activators, whereas p50 and p52 homodimers are unable to enhance RNA polymerase II-driven transcription (14). In contrast to other members of the NF-B family, RelB contains in its NH 2 -terminal domain a leucine zipper-like structure that is essential for transactivation of target genes (15). However, the regulation of RelB activity is still poorly understood.
Studies of T lymphocytes, isolated from IB␣-deficient mice, demonstrated that the dominant IB regulator of NF-B-Rel is IB␣, the product of the MAD3 gene (16,17). Activation of cells with adequate signals such as T cell receptor triggering, phorbol esters, interleukin 1 (IL-1), tumor necrosis factor (TNF-␣), and others results in IB␣ degradation by 26 S proteasomes (for review, see Ref. 7). This renders dimers, which contain RelA and c-Rel proteins, free to translocate into nuclei where they activate transcription of target genes. The molecular mechanism resulting in IB␣ proteolysis is complex and not completely elucidated. However, at least two post-translational covalent modifications have been reported to be essential for its degradation. The first critical event is phosphorylation of serines 32 and 36 in the NH 2 -terminal region of IB␣, carried out by Ser/Thr kinase(s) including a multienzyme complex of 700 kDa (18). This double phosphorylation of IB␣ does not lead to dissociation from NF-B, but it is prerequisite for the second modification step, which is the ubiquitination of two NH 2 -terminal lysines at positions 21 and 22 (19). Subsequently the phosphorylated and ubiquitinated IB␣ is proteolyzed by the 26 S-proteasome complex (20,21).
Once released from IB␣, NF-B-Rel proteins translocate rapidly to the nucleus where they exert their regulatory functions by interacting with specific decameric B sequences and the general transcription factor TFIIB (22). A plethora of genes have been shown to contain B sequences in their promoters (for review, see Ref. 23). In T cells, gene products involved in cell adhesion (intercellular cell adhesion molecule-1; ICAM-1), cell growth control (IL-2, its receptor IL-2R␣, and c-myc), and proinflammatory mediators (IL-6, TNF-␣) are suspected of being transcriptionally regulated by NF-B. Furthermore, viruses with T cell tropism, such as HIV, are also thought to be transcriptionally regulated by NF-B proteins (24). Specific relationships between distinct NF-B complexes and particular target genes are not yet understood, although preferential binding and preferential transcriptional activation efficiencies have been demonstrated by transfection experiments with discrete NF-B expression vectors and distinct B sequences (25)(26)(27)(28)(29). These observations suggest that distinct NF-B-Rel complexes modulate transcription of different genes selectively.
Transient transfection assays with a mutated form of IB␣, in which serines 32 and 36 were replaced by alanines, demonstrated that the double mutation prevented proteolytic degradation of the transgenic IB␣ by the usual NF-B activators (TNF-␣ and phorbol esters) (30 -32). Thus the double mutation generates a constitutive repressor of the cytosolic NF-B-Rel proteins. Because the 32/36A IB␣ mutant (IB␣ 32/36A ) should not affect the constitutively nuclear pool of RelB proteins, IB␣ 32/36A potentially represents a powerful and selective tool for the study of the respective roles of NF-B and RelB protein complexes in gene expression. We have therefore used cell clones that express both RelA and RelB subunits for stable transfections with the IB␣ 32/36A . We report in the present publication the effects of the expression of the transgenic IB␣ on NF-B-Rel activation and gene expression.

MATERIALS AND METHODS
Cells-The parental HPB-ALL cell line was cultured in RPMI 1640 medium containing glutamine, antibiotics, and 10% fetal calf serum.
Stable Transfections-Expression vector pCMV-IB␣ 32/36A was made by inserting MAD3 cDNA mutant at amino acid positions 32 and 36 into the XbaI/HindIII sites of the pcDNA3 vector from Invitrogen (S32/36A mutant in 32). HPB-ALL cells were transfected with pCMV-IB␣ 32/36A and the empty pcDNA3 vector by electroporation. G418resistant cells were cloned by limiting dilution and genotyped by Southern blotting (33), using a full-length MAD3 cDNA probe. The clones with stable integration of IB␣ 32/36A were grown in RPMI with 10% fetal calf serum in presence of 1 mg/ml G418.
Cell Extracts-In gel shift experiments (electrophoretic mobility shift assay; EMSA) cells were incubated for the indicated periods of time with activators, and nuclear proteins were extracted as described previously (34). In Western blotting, the cytosolic extracts were obtained in the hypotonic buffer described in Ref. 34.
Western Blotting-Equal amounts of protein (30 g) extracted from cytoplasma of control or IB␣ 32/36A -transfected clones were fractionated on 10% polyacrylamide gels by electrophoresis in denaturing conditions, according to Porzio and Pearson (35). Proteins were electrotransferred onto polyvinylidene difluoride membranes (Millipore). The efficacy of the transfer was tested by Ponceau Red staining. The wild type and transgenic IB␣ were determined using a monoclonal antibody (MAD 10B) specific for an NH 2 -terminal domain of IB␣ (36). The antigen-antibody complex was revealed using horseradish peroxidasecoupled anti-mouse antibody and the Amersham enhanced chemiluminescence visualization system (ECL) kit. The autoradiography was carried out for 5 s to 10 min. For RelB Western blotting, the identical procedure was followed except that 70 g of nuclear extracts was used. The RelB-specific antiserum was from Santa Cruz (Tebu, France), and its dilution was 1/500. EMSA-The EMSA was performed using 10 g of nuclear protein extracts/incubation. The B oligonucleotide used was a kind gift from Dr. Leo Lee (NCI, Frederick, MD) and corresponds to the tandem B sequence (PRE) from the HIV LTR. The gel shift experiments were carried out following the procedure described in Ref. 34. To identify the PRE-binding proteins, nuclear extracts from control HPB-ALL cells and one of the stably transfected clones, the A3 clone, were incubated with 2 l of antibodies specific for individual NF-B-Rel proteins before the addition of the radiolabeled PRE oligonucleotide. All antibodies were purchased from Santa Cruz. In addition to the Santa Cruz antibodies, we also used an antiserum specific for the COOH-terminal domain of the RelA molecule (named 1226 in Ref. 37), kindly provided by Dr. Nancy Rice (NCI).
Chloramphenicol Acetyltransferase (CAT) and Luciferase (Luc) Assays-The following vectors were used for B-dependent CAT and Luc assays. The 1168hIL6Lucϩ construct, which contains 534 base pairs of the human IL-6 promoter, was kindly provided by Prof. G. Haegeman (Gent University, Belgium). Dr. A. Israël (Pasteur Institute, France) provided us with the 1.2HN-Luc construct (38) containing the NFKB1 (p105) promoter region and the 0.4SK, 0.2SK, and 0.4SK68⌬B Luc plasmid containing the MAD3 (IB␣) promoter constructs (39). The 0.4SK contains all three B sites from the MAD3 promoter domain, whereas the 0.2SK contains only the proximal B1 site, and the 0.4SK⌬B contains only the B2 and B3 sites. To monitor the tranfection of the three constructs of MAD3 promoter, the p.␤gal-promoter vector (CLONTECH), which contains a functional LacZ gene downstream of the SV40 early promoter, was cotransfected, and the ␤-galactosidase activity was measured by spectrophotometry in the presence of 100 nM o-nitrophenol ␤-D-galactoside. The ICAM-1 promoter-Luc construct (pGL1.3) was described by Ledebur and Parks (40) and was provided by Dr. K. Roebuck (Rush, Chicago). The c-myc promoter (Ϫ2325 to ϩ36) and c-fos promoter (Ϫ711 to ϩ42) CAT constructs are described in Ref. 41. The LTR3 CAT-218 construct containing the 218 base pairs upstream from the transcription initiation start of the HIV 5Ј-LTR (42) was provided by Dr. R. B. Gaynor (UCLA, Los Angeles). Finally, Dr. G. R. Crabtree (HHMI, Stanford, CA) provided us with the IL-2 promoter-Luc construct (pCLN15⌬CX) (Ϫ326 to ϩ45). Transient transfections of the cell clones with CAT and Luc vectors were performed by electroporation at 200 V, 500 microfarads (Bio-Rad electroporation system) with 20 g of plasmid DNA/5⅐10 6 cells. 2 h after transfection, cells were split into two pools. One pool of cells was incubated in RPMI (untreated cells), and the other pool was incubated with 10 ng/ml phorbol 12-myristate 13-acetate (PMA) plus 1 g/ml phytohemagglutinin (PHA) for 24 h (activated cells). The cells were then collected, washed once with phosphate-buffered saline, and lysed by three cycles of freezing/thawing in 150 mM Tris-HCl, pH 8. Cell extracts, normalized for total protein content (43), were assayed for CAT activity using [ 14 C]chloramphenicol (NEN Life Science Products) according to Gorman et al. (44). The chloramphenicol conversion was quantified using a BetaImager 1200 apparatus (Biospace, France). The results were expressed as percent of chloramphenicol conversion/mg of protein (relative CAT units). Transfection experiments were repeated at least three times, using two independent plasmid preparations. Luc assays were performed using the Promega luciferase assay system. The cells were lysed with 25 mM Tris phosphate, 2 mM dithiothreitol, 2 mM 1,2-diaminocyclohexane-N,N,NЈ,NЈ-tetraacetic acid, 10% glycerol, 1% Triton X-100, pH 7.8. The light emission was measured in a luminometer (Bio-Rad). The results were calculated as relative light units (light emission/background/mg of protein).
Northern Blotting-cDNA probes used for Northern blotting were obtained by enzymatic digestion of the following vectors: pKH47-c-myc (PstI/EcoRI digestion generating a 1,200-base pair fragment of the c-myc cDNA), p1IL-2 (PstI/BglII digestion generating a full-length cDNA), pBRchIL6F2 (PstI digestion generating an 855-base pair fragment of the human IL-6 cDNA), p105 (HindIII/ApoI digestion generating full-length cDNA), pCMV-MAD3 (HindIII/XbaI digestion generating full-length IB␣ cDNA), pBr322-actin (PstI digestion generating full-length actin cDNA). The cDNA probes were radiolabeled using [␣-32 P]dCTP (Amersham) and the Rediprime kit (Amersham). Total cytoplasmic RNA was prepared according to a modified method of Chomczynski and Sacchi (45), using the Stratagene RNA kit. Total RNA (10 -20 g) was fractionated by electrophoresis on 0.7% agarose gels containing 2.2 M formaldehyde. Gels were blotted on Hybond N ϩ membranes (Amersham) according to the indications of the manufacturer. Membranes were hybridized with 32 P-labeled probes in Quickhyb solution (Stratagene) according to the protocol supplied by the manufacturer, at 65°C. Membranes were autoradiographed for 1-12 h at Ϫ70°C with intensifying screens. Membranes were stripped by boiling in H 2 O and rehybridized with the ␤-actin probe to normalize loading of RNA samples.
Measurement of CD25 Expression by Flow Cytometry-Control, A3, F10, and D7 cells were activated for 24 h by PMA (10 ng/ml) and PHA (1 g/ml). Unstimulated cells and PMA plus PHA-treated cells were tested for CD25 by flow cytometry using a phycoerythrin-conjugated human CD25-specific monoclonal antibody from Caltag Laboratories and fluorescence-activated cell sorter apparatus from Becton-Dickinson.

Characterization of Stable HPB-ALL Clones
Transfected with the 32/36A Mutant IB␣-The parental HPB-ALL cell line is a T cell tumor producing IL-2 and IL-6 in response to T cell activators, such as phorbol esters, in the presence of a Ca 2ϩ influx activators (PHA, ionomycin, CD3-specific antibodies, etc.). Its phenotype is close to a double positive thymocyte (CD4 ϩ /CD8 ϩ , CD1a ϩ , CD3 ϩ ). We chose this line as a model for studying the inhibition of the inducible NF-B by a dominant negative form of IB␣ (IB␣ 32/36A ). The stability of the integration of the mutant IB␣ was verified by Southern blotting of DNA extracted from several clones isolated by limiting dilution and cultured for 1 month in the presence of the selective antibiotic (not shown). Three clones, A3, D7, and F10, were identified as stably transfected with IB␣ 32/36A . To verify that the IB␣ cDNA was expressed in these clones, we performed Western blot analysis of the cytosolic fractions of the control and the "mutant" clones, using a monoclonal antibody specific for the NH 2 -terminal domain of IB␣ (36). The wild type and the mutant IB␣ are distinguishable on the basis of their electrophoretic migration because the 32/36A mutant migrates slightly slower in SDS gels (32). In the three clones that integrated IB␣ 32/36A , a slower migrating protein was specifically detected by the antibody in addition to the wild type IB␣ (Fig.  1A). Judging by the immunoblot results, the mutant and wild type IB␣ were expressed at comparable levels in clone A3 and F10, whereas in clone D7, the mutant IB␣ was more highly expressed relative to the wild type IB␣. In control cells, only the faster migrating 36-kDa IB␣ was detected. In none of the mutant clones did expression of the transgenic IB␣ prevent constitutive production of the wild type IB␣.
To determine whether the mutant IB␣ could inhibit translocation of NF-B, we performed gel shift experiments with nuclear extracts from resting and PMA plus PHA-treated cell clones (Fig. 1B). In control cells, a 1-h PMA plus PHA activation generated a nuclear translocation of B oligonucleotidebinding proteins, visible as a doublet. In contrast, in the three stable clones, neither constitutive nor inducible B oligonucleotide binding activities were detected, suggesting that the mutant IB␣ prevented NF-B translocation.
IB␣ 32/36A Blocks Nuclear Translocation of RelA-NF-B but Not of RelB-p50 -To investigate the duration of NF-B inhibition by the mutant IB␣, we analyzed NF-B nuclear translocation during a time course of PMA plus PHA treatment. In the control cells, B binding activities were clearly detectable at the 3 h time point and increased in intensity up to 24 h of treatment ( Fig. 2A). In the A3 clone, no significant B binding activity was detected until 7 h of activation. However, by the 7 h time point, a B binding activity, migrating as a doublet, was clearly detected and reached levels similar to the control by 24 h of activation ( Fig. 2A). To identify the proteins in the complexes bound to B oligonucleotide, we tested the abilities of antibodies specific for RelA, c-Rel, p50, and RelB to affect the EMSA patterns of control and A3 cells after 7 and 24 h of PMA plus PHA stimulation. In the absence of specific antibodies, several complexes were detectable in the control cells. The discrimination of these complexes was difficult in this type of gels; but clearly, in the A3 clone, only two bands were detectable after 7 h of cell stimulation, whereas in control cells additional, slower migrating bands, existed (see Fig. 2B and photographically enlarged view in Fig. 2C). Antibodies specific for RelA and RelB demonstrated that the two upper bands from the control cells contained RelA, whereas one of the lower bands contained RelB (Fig. 2B). p50-specific antibodies removed the two lower bands from both A3 and control cells. Thus, the upper band in A3 clone was composed of RelB-p50 dimers, whereas the lower band was the p50 homodimer. Antibodies specific for c-Rel had no effect on the B-binding proteins in control or A3 cell nuclear extracts (Fig. 2), whereas they inhibited efficiently c-Rel-p50 binding in control cells (not shown). Thus, whereas in control cells both RelA and RelB dimers were detected, in the A3 clone only RelB-p50 and p50 dimers were detected. After 24 h of PMA plus PHA activation, both control and A3 nuclear extracts contained only the two faster migrating complexes (p50-p50 and RelB-p50) (Fig. 2B). These results demonstrated that in control HPB-ALL cells, the initial effect of PMA plus PHA activation led to nuclear translocation of cytosolic NF-B proteins (RelA homo-and heterodimers). As expected, in the IB␣ 32/36A -transfected A3 clone, the translocation of these proteins was inhibited. Prolonged stimulation led to RelB activation in both cell clones. Further- more, after 24 h of PMA plus PHA treatment activation of the RelA-containing complexes was also inhibited. It was not surprising to observe comparable levels of RelB in both control and A3 cell clones since RelB activation was reported to be independent of IB␣. Similar results were obtained with the two other IB␣ 32/36A -transfected F10 and D7 cell clones (not shown). Thus, we have generated a cell system in which the prototypical NF-B is inhibited selectively by IB␣ 32/36A , but activation of RelB remains potentially intact. Western blotting analysis of nuclear extracts from A3 and HPB-ALL cells further assessed the presence of RelB. RelB nuclear amounts were increased upon PMA plus PHA stimulation (Fig. 3). In addition, immunochemical analysis with RelB-specific antibodies confirmed the increase of RelB in nuclei of control cells and A3 cells after 24 h of stimulation (not shown). This is suggestive of a transcriptional, or at least, pretranslational, regulation of the RelB in PMA plus PHA-activated HPB-ALL T cells.
IB␣ 32/36A Expression Is Increased in PMA-stimulated Cells-In the absence of stimulation, wild type IB␣ has a rapid turnover that is independent of serines 32 and 36 phosphorylation and of ubiquitination (46). After stimulation with PMA or TNF-␣, IB␣ is modified by phosphorylation and ubiquitination, and the balance between degraded and newly synthesized IB␣ turns transiently in favor of the degradation (47). As a result, IB␣ is detected in lower amounts in cytosol from short term activated cells. However, the resulting activation of NF-B induces newly synthesized IB␣ that is detectable within 1-2 h after activation. This neosynthesized IB␣ is, in turn, probably responsible for the inhibition of the RelAcontaining NF-B at later time points of PMA plus PHA treatment (see Fig. 2B). This cycle of activation-induced proteolysis/ resynthesis of IB␣ is initiated by the phosphorylation of serines 32 and 36. To examine the fate of the IB␣ 32/36A versus the wild type IB␣ in activated cells, we performed kinetic experiments in which the A3 and the control clones were treated with PMA plus PHA for increasing lengths of time. Western blot analysis of IB␣ after up to 2 h (Fig. 4A) and 24 h (Fig. 4B) of activation by PMA plus PHA showed that the wild type IB␣ was degraded almost completely within 30 min in control cells. After 1 h of PMA plus PHA treatment it was resynthesized progressively, reaching initial levels after as soon as 3 h of activation (Fig. 4B). In the A3 clone, the wild type IB␣ was also degraded rapidly in response to cell activation, but neosynthesized wild type IB␣ was detectable only after 7 h of PMA plus PHA treatment, reaching initial levels at the 9 h time point (Fig. 4, A and B). In contrast to the wild type IB␣, the IB␣ 32/36A was not degraded in response to cell activation. In fact, levels of IB␣ 32/36A increased from the 30 min time point to reach a steady maximum at 2 h (Fig. 4A) probably because the CMV promoter is activated independently of NF-B activation. These experiments clearly demonstrated the stability of the mutant IB␣ in activation conditions that lead to wild type IB␣ proteolysis. , and control cells, with a series of reporter gene constructs (CAT or Luc) linked to promoter regulatory regions of seven genes suspected to constitute targets for NF-B. In each of these promoters, at least one B consensus sequence was identified in addition to sites specific for other regulatory transcription factors. The specificity of the IB␣ 32/36A inhibition on NF-B-driven transcription was assessed with a reporter plasmid dependent on serum response element (c-fos CAT) (41). The results are summarized in Tables I and II. All of the promoters used in this study were activated by PMA plus PHA in control cells. The inductions of CAT and Luc constructs by PMA plus PHA ranged from 2.2-fold (c-myc) to 189-fold (p105) in control cells (Table I). The effect of the IB␣ 32/36A transgene on activation of CAT and Luc transcription depended on the promoter used. The transcription of the reporter genes driven by HIV LTR, MAD3, IL-6, IL-2, and p105 promoters was strongly inhibited in the three clones used (Table II). For example, in clone A3, activation of the IL-2 promoter reached only 3% of that obtained in control cells. In contrast, only 50% inhibition was observed with the ICAM-1 promoter. Finally, c-myc promoter driven transcription of CAT was not inhibited at all in the IB␣ 32/36A -transfected clones. Transient transfections performed with the control fos-CAT showed no difference in transcriptional activation between control cells and the A3 clone and a small inhibition in the D7 and F10 clones. Together these results indicated a selective effect of the IB transgene on NF-B-driven transcription. Three sets of promoters could be distinguished on the basis of their responsiveness to NF-B inhibition: promoters that were strongly inhibited by the transgene (IL-6 and IL-2, for example), promoters that were partially inhibited (ICAM-1), and promoters that were not affected by the lack of NF-B translocation (cmyc). Activation of the IB␣ (0.2SK) promoter, which was shown to contain the B1 sequence responsible for the transcriptional induction by RelA-p50 in Jurkat cells (39), was abolished totally in two of the three clones (A3 and D7). To estimate the potential contribution to IB␣-promoter activation by RelB through binding to the two upstream B sites, we performed Luc assays with two additional IB␣-promoter constructs, the 0.4SK-Luc, containing all three B sites, and the 0.4SK⌬B, which contains only the B2 and B3 upstream sites. With the 0.4SK-Luc construct, only a little activation was obtained in A3 clone (2.1-fold) after PMA plus PHA treatment compared with control cells (23.6-fold) (Fig. 5). The 0.4SK⌬B construct was not activated by PMA plus PHA in control and A3 clones (Fig. 5), indicating that B2 and B3 sites are not capable of enhancing IB␣ transcription in the absence of the B1 site. Together, these results suggested that if RelB is implicated in IB␣ gene up-regulation, it does so through the involvement of a site(s) distinct from those contained in the promoter region reported up to now. Thus, to determine the impact of NF-B inhibition on gene expression in the context of the genome, we analyzed the mRNA and/or protein production of several NF-B target genes.
Effect of IB␣ 32/36A on mRNA Production-We compared the induction of five B target genes by a time course analysis of mRNA production by Northern blotting. Hybridization of the ␤-actin probe was used as control (Fig. 6). In the empty vectortransfected cell clone (control), p105 (NFKB1), IL-2, wild type IB␣ mRNA, and IL-6 (data not shown) were induced in a time course-dependent manner. In contrast, no transcripts for IL-2 and IL-6 (data not shown) were detected in the A3 clone. The NFKB1 (p105) transcript was produced constitutively in A3  cells, but no induction by PMA plus PHA was observed. In the control cells, MAD3 (IB␣) mRNA reached a steady state after as little as 30 min of PMA plus PHA treatment. In A3 cells, two transcripts were detected with the MAD3 probe. The higher mobility transcript, corresponding to the mutant IB␣, was strongly augmented in the early time points of activation, peaking at 1 h, whereas the slower migrating wild type messenger was detected at later time points (6 -24 h). Thus, whereas transcription of IL-2 and IL-6 genes was inhibited completely in the A3 clone, induction of the wild type IB␣ messenger RNA was delayed in the A3 clone compared with the control cells. Finally, the c-myc messenger was produced constitutively in these cells, and no induction by PMA plus PHA was detected in either control or A3 cells. Both IL-2 and IL-2R␣ Productions Are Inhibited by the IB␣ 32/36A Mutant-We further investigated the regulation of IL-2, an important T cell proliferation regulator, by measuring its production at the protein level in two of the stable clones (A3 and D7) by ELISA (Fig. 7). In both transgenic clones, IL-2 production was 10% of the control after 12 h of PMA and PHA activation. Thus, the result obtained with the Northern blot analysis of IL-2 induction was confirmed at the protein level.
The ␣ subunit of the IL-2 receptor (CD25) is another potential target for NF-B regulation (48,49). To analyze further the effect of IB␣ 32/36A on the IL-2-regulated growth control we measured the expression of the CD25 by flow cytometry. In the absence of activation, no CD25 was detected on the surface of the clones. After 24 h of PMA plus PHA treatment, 68% of the control cells expressed CD25. In the A3 clone, a strong inhibition was observed because only 7% of cells were labeled with the CD25-specific antibody. In the F10 and D7 clones, the inhibition was less potent; 40% and 28%, respectively, of these cells were found to be CD25 positive (summarized in Table III). Despite the variability among the three clones, together these results indicated that the transcription of IL-2R␣ requires RelA-NF-B activation. DISCUSSION In the present paper, we have reported the effect of a selective inhibition of the RelA-containing NF-B on gene expression in T cell clones. We have generated this cell system by stable transfection of a mutant form of IB␣ which has been  shown previously to block inducible RelA and c-Rel nuclear translocation (32). The particular T cell line that we used also naturally expresses RelB. Three independent cell clones, A3, D7, and F10, were selected by limiting dilution. In agreement with previous reports (30 -32), no proteolysis or ubiquitination of the mutant IB␣ was detected under conditions in which signal-induced degradation of the wild type IB␣ occurred. In all three of the clones, the stability of the transgene led to an efficient inhibition of the RelA-containing NF-B DNA binding, whereas RelB-p50 DNA binding capacity was unchanged, as expected. Therefore the signal-activable NF-B was inhibited selectively, whereas the inducible IB␣-independent RelB-p50 complex remained potentially active.
Expression of RelB is unusual in T cells. RelB is dominantly expressed in dendritic cells from both primary and secondary lymphoid organs (3,5,6) and in B cells at later stages of development. Its involvement in dendritic cells development was demonstrated clearly in RelB knockout mice (4,50), but the role of RelB in gene expression remains obscure. Thus, our cell system is a convenient model for discriminating between gene transcription regulated selectively by the RelA-NF-B and genes that may be regulated by RelB. In this respect, among the genes that we have studied the regulation of IB␣ expression is of particular interest. It has been shown, indeed, that among the three B consensus sites found in the human MAD3 promoter region, it is the most proximal site, B1, that mediates PMA and TNF activation through binding of RelA complexes (39). The B2 site is recognized by RelA complexes, but it is not able to mediate efficient activation of the MAD3 promoter in cells producing only RelA and c-Rel complexes. The B3 site was unable to bind NF-B proteins extracted from myeloid cells (39). In HeLa cells all three B sites were reported to contribute efficiently to TNF-␣ activation of IB␣ promoter (51). In addition, transfection of Jurkat cells with RelB vectors led to increased levels of IB␣, suggesting that RelB is capable of enhancing IB␣ expression (13). We used three-luciferase reporter constructs that contain, respectively, all three (0.4SK), the two upstream (0.4SK⌬B), or only the B1 (0.2SK) sites of the IB␣ promoter. With these constructs the induced transcription of Luc was inhibited potently in all of the IB␣ 32/36A transgenic clones. This result demonstrates that in the MAD3 promoter, the three B sites mediate IB␣ transcription by selectively binding forms of NF-B which are themselves regulated by IB␣. RelB was unable to compensate for the lack of NF-B activation in this assay. However, expression of IB␣ did occur but at later time points of PMA stimulation than in control cells. The expression of the wild type IB␣ mirrored the increase of RelB DNA binding activity and protein. In addition, resynthesis of IB␣ did not occur in the MCF7 cells stably transfected with IB␣ 32/36A , cells that do not express RelB (52). These results strongly suggest that not only does RelB regulate IB␣ transcription through a site different from that which binds the prototypical NF-B, but that NF-B is not required for full IB␣ expression when RelB is produced in sufficient amounts. This effect might be specific for human IB␣ since the porcine IB␣ promoter domain, which contains six B consen-sus sites, was not activated by overexpression of RelB (53). The alternative RelB-specific regulation of IB␣ could have functional consequences in cells that produce high levels of RelB such as dendritic cells. In such cells, NF-B activity could be regulated negatively by IB␣ overexpression due to to RelB. However, the correlation between MAD3 expression and RelB activation remains to be established.
The potential compensatory effect of RelB was not observed with all of the genes that we examined. For example, expression of IL-2 was inhibited dramatically at both RNA and protein levels. IL-2 promoters possess multiple regulatory sequences among which are a single B consensus site and multiple sites capable of interacting with c-jun protein complexes. Expression of a dominant negative mutant of c-jun abolishes IL-2 expression (54) probably because it coordinately blocks the IL-2 transcriptional regulation at multiple sites. In activated T cells, the major forms of NF-B which bind to the B site in the IL-2 promoter were shown to be p50 homodimers (55) and RelA homo-and heterodimers (56). Paradoxically, disruption of the c-rel gene in mice also inhibited induced IL-2 production despite the presence of RelA and p50 (57). A possible explanation of this effect was reported recently by Smith Shapiro et al. (58), who show that c-Rel regulation of the IL-2 promoter might be mediated by AP1 rather than directly through binding to B sites. In contrast to c-Rel, RelA was not able to activate AP1-dependent luciferase expression (58). Here we show that in the absence of RelA dimers, RelB-p50 cannot rescue IL-2 expression. Further, the degree of IL-2 inhibition by IB␣ 32/36A transfection brings additional strong evidence that activation of RelA dimers is a limiting step for IL-2 transcriptional initiation.
Contrary to c-Rel disruption, inhibition of RelA dimers also diminished expression of the IL-2 receptor (CD25). Therefore, the classical NF-B dimers seem to be involved in regulating the whole IL-2 growth control system.
In contrast to MAD3, RelB was not able to enhance the signal induced expression of NFKB1 (p105), indicating that selective activation of RelA dimers is required for the signal-induced expression of p105. However, the p50 protein (the processed, functional product of the NFKB1 gene) and the p105 mRNA were produced in both parental and the IB␣ transgenic cells, independent of cell activation. This suggests that the constitutive expression of NFKB1 is independent of B enhancers.
Interestingly, c-myc expression was not inhibited by the inhibition of NF-B. The c-myc promoter upstream B site was shown to bind to RelA and c-Rel dimers and to be a positive regulator of the c-myc promoter in CAT assays in B lymphoma cells (59). Here we show that c-myc is expressed constitutively, not only in the parental HPB-ALL cells, but also in the IB␣ 32/ 36A transgenic cell clones. The c-myc promoter activity was only feebly enhanced by PMA plus PHA, and it was not decreased by the inhibition of RelA-NF-B. It is therefore possible that RelB is able to activate c-myc expression constitutively. Alternatively, c-myc expression could be independent of the B sites in HPB-ALL cells.
HIV LTR contains two direct repeats of the B site in tandem. These B sites are critical for the initial steps of HIV replication (24,60,61). It has been shown that although RelA-p50 up-regulates the HIV promoter through binding to the B tandem sequence, c-rel behaves as a repressor of the RelA-p50 in the context of HIV LTR and the CD25 promoter (62). In control HPB-ALL cells, PMA plus PHA activated the transcription from the HIV LTR by 42-fold over the basal level. This activation was inhibited by 90% in the IB␣ 32/36A transgenic cell clones. Thus RelB was unable to substitute for RelA dimers. IB␣ 32/36A could be a powerful tool for repressing HIV replication in infected cells. However, since HIV replication is independent of NF-B in the presence of the HIV Tat regulatory factor (61), we are currently investigating by infection experiments whether the effect seen in the CAT assay can be extrapolated to the viral replicative cycle.