Cytoplasmic IκBα Increases NF-κB-independent Transcription through Binding to Histone Deacetylase (HDAC) 1 and HDAC3

IκBα is an inhibitory molecule that sequesters NF-κB dimers in the cytoplasm of unstimulated cells. Upon stimulation, NF-κB moves to the nucleus and induces the expression of a variety of genes including IκBα. This newly synthesized IκBα also translocates to the nucleus, removes activated NF-κB from its target genes, and brings it back to the cytoplasm to terminate the phase of NF-κB activation. We show here that IκBα enhances the transactivation potential of several homeodomain-containing proteins such as HOXB7 and Pit-1 through a NF-κB-independent association with histone deacetylase (HDAC) 1 and HDAC3 but not with HDAC2, -4, -5, and -6. IκBα bound both HDAC proteins through its ankyrin repeats, and this interaction was disrupted by p65. Immunofluorescence experiments demonstrated further that IκBα acts by partially redirecting HDAC3 to the cytoplasm. At the same time, an IκBα mutant, which lacked a functional nuclear localization sequence, interacted very efficiently with HDAC1 and -3 and intensively enhanced the transactivation potential of Pit-1. Our results support the hypothesis that the NF-κB inhibitor IκBα regulates the transcriptional activity of homeodomain-containing proteins positively through cytoplasmic sequestration of HDAC1 and HDAC3, a mechanism that would assign a new and unexpected role to IκBα.

The paradigm of NF-B activation is based on its cytoplasmic sequestration by IB␣ through the masking of its nuclear localization sequence (NLS) 1 in unstimulated cells (1). Recent studies have challenged this hypothesis suggesting that IB␣ constantly shuttles in and out of the nucleus via nuclear import (2) and export sequences (3,4). Because IB␣ nuclear export is more efficient than its nuclear import process, the NF-B⅐IB␣ complex remains mainly in the cytoplasm of unstimulated cells. Upon stimulation by proinflammatory cytokines, viral infection, or bacterial pathogens, IB␣ is phosphorylated by the IB kinase complex and degraded through the proteasome pathway (5). Then NF-B enters into the nucleus and activates the expression of multiple genes including IB␣. This newly synthesized IB␣ protein moves to the nucleus, removes NF-B from its target genes, and brings it back to the cytoplasm to terminate the phase of NF-B activation. IB␣ is part of a family of ankyrin-containing proteins that also includes IB␤ and IB⑀ (6) as well as p100, p105, and BCL-3 (7). In contrast to the IB proteins, BCL-3 is mainly nuclear and harbors some transactivation potential through heterodimer formation with the NF-B proteins p50 or p52 (8) and recruitment of coactivators such as JAB1 and BARD1 (9).
We previously demonstrated that IB␣ enhances the transactivation potential of HOXB7, a homeodomain-containing protein (10). In this study, we elucidated further the underlying mechanism and showed that cytoplasmic IB␣ enhances the transactivation potential of other homeodomain-containing proteins such as Pit-1 and PAX8 through binding to some but not all HDAC proteins, a family of proteins sharing the ability to deacetylate their substrates and to repress transcription (11). Taken together, our results therefore suggest that IB␣ may have new and NF-B-independent roles.

EXPERIMENTAL PROCEDURES
Cell Culture and Biological Reagents-293, HeLa, HCT116, and MDA-MB231 cells were maintained in Dulbecco's modified Eagle's, McCoy, and RPMI 1640 media, respectively, and supplemented with 10% fetal calf serum (Invitrogen) and antibiotics. Polyclonal anti-HA, -HDAC1, -HDAC3, -p50, -p65, and -IB␣ antibodies were purchased from Santa Cruz Biotechnologies (Santa Cruz, CA). The polyclonal anti-HDAC3 antibody used for the indirect immunofluorescence experiments was purchased from Cell Signaling Technology (Beverly, MA). Monoclonal anti-IB␣ antibody was provided by R. Hay (School of Biological and Medical Sciences, University of St. Andrews, Fife, Scotland, UK). Anti-FLAG antibodies and beads were purchased from Sigma.
The HDAC3, FLAG-HDAC1, -2, -3, -4, -5, and -6 constructs were provided by S. Emiliani (Institut Cochin de Génétique Moléculaire, INSERM U529, Paris, France). The HOXB7-expressing vector (10), the pMT 2 T expression vectors for IB␣ (12), IB␣⌬N, NϩC GST, and IB␣⌬C were described previously (13) as were the pMT 2 T expression vectors for p50 and p65 (14). The Pit-1-expressing vector was constructed by subcloning Pit-1 coding sequence into pcDNA3. The IB␣ NLS-MT construct harbors three point mutations within the nuclear import sequence (2) and was generated by subcloning the IB␣ 110〈3 coding sequence, a gift from M. Hannink (University of Missouri, Columbia, MO), into the pMT 2 T expression vector. The PAX8-expressing vector was a gift from M. Zannini (Dipartimento Biologia e Patologia Cellulare e Molecolare, Universita' di Napoli Federico II, Naples, Italy). The pTCBS plasmid contains an eightfold multimerized form of a homeodomain consensus-binding sequence (CBS) cloned upstream of an HSV-TK promoter and of a luciferase reporter gene, whereas the pT109 construct does not contain the CBS and was used as a negative control (15). Both these reporter plasmids were provided by V. Zappavigna (17). The TSH␤ luciferase reporter plasmid was described previously (18).
Immunoprecipitation-For immunoprecipitations of overexpressed proteins, 293 cells (3 ϫ 10 6 ) were transfected with FuGENE (Roche Applied Science) and expression vectors as indicated. Subsequent lysis and immunoprecipitations were performed as described previously (19) as were the immunoprecipitations of endogenous proteins.
For identification of ternary complexes by immunoprecipitation experiments, 293 cells (10 7 ) were transfected with FLAG-IB␣ as described above. 24 h after transfection, ectopically expressed IB␣ was immunoprecipitated with an anti-FLAG antibody for 2 h at 4°C. Anti-HA immunoprecipitations were performed in parallel as negative controls. The immunoprecipitates were washed five times with the lysis buffer and incubated overnight at 4°C with the FLAG peptide (Sigma) according to the protocol provided by the manufacturer. The supernatants were subsequently incubated for 2 h with anti-HA (negative control), anti-p50, or anti-HDAC3 antibodies and overnight with protein A-agarose conjugate. The resulting immunoprecipitates were washed with the lysis buffer and subjected to SDS-PAGE for anti-p65 Western blot analysis.
Indirect Immunofluorescence-293 cells were seeded in 6-well plates and subsequently transfected with the indicated expression plasmids. 24 h later, cells were washed with phosphate-buffered saline, fixed with 4% paraformaldehyde, and permeabilized with ethanol for 6 min at Ϫ20°C. Fixed cells were incubated with monoclonal anti-IB␣ antibodies overnight at 4°C followed by a 60-min incubation with fluorescein isothiocyanate-conjugated anti-mouse IgG (Dako). Coverslips were then mounted with Fluoprep (Biomerieux, Marcy l'Etoile, France), and cells were observed by fluorescent microscopy (Nikon).
Reporter Assays-HCT116, MDA-MB231, or 293 cells (4 ϫ 10 5 ) were transfected in 6-well plates with various amounts of expression vectors as indicated. Total amounts of transfected DNAs were kept constant throughout by the addition of appropriate amounts of pcDNA3 empty vector. Cells were harvested 24 h after transfection, and luciferase activities were measured and normalized as described previously (10).

IB␣ Enhances Transactivation by Multiple Homeodomain-
containing Proteins-We previously established that IB␣ enhances the transactivation potential of HOXB7, a homeodomain-containing protein ( Fig. 1A and Ref. 10). Therefore, we investigated whether this effect also occurred in another cell line and performed identical luciferase assays in 293 cells. Again, IB␣ enhanced HOXB7 transactivation potential in this cell line. To determine whether IB␣ regulates the transcriptional activity of other homeodomain-containing proteins, we selected PAX8, a member of the Paired domain (PAX) family known to transactivate the bcl-2 promoter (20) and used a small fragment of this promoter as reporter plasmid. IB␣ weakly transactivated this promoter in 293 or HCT116 cells but strongly enhanced PAX8-mediated transactivation in both cell lines (Fig. 1B). The homeodomain-containing protein Pit-1 is known to regulate the expression of GH, TSH␤, and PRL. We therefore performed similar luciferase assays by transfecting 293 or HCT116 cells with corresponding reporter plasmids either alone or with Pit-1 and/or IB␣. We again detected a strong synergistic effect on the PRL promoter in both cell lines ( Fig. 2A) and to a lesser extent on the TSH␤ promoter, especially in HCT116 cells (Fig. 2B, right panel). A weaker effect was observed on the GH promoter in 293 cells, whereas no significant synergistic effect was detected in HCT116 cells when the GH promoter was used (Fig. 2C, left and right panel, respectively). Therefore, our results indicate that IB␣ enhances transactivation by several homeodomain-containing proteins on multiple but not all promoters. Moreover this ability of IB␣ to enhance the transactivation potential of the homeodomain-containing proteins does not appear to be cellspecific. Therefore, it is likely that a general mechanism is involved.
Cytoplasmic IB␣ Binds HDAC1 and HDAC3 through the Ankyrin Repeats-Because transcription is also enhanced through cytoplasmic sequestration of HDACs (21), we examined whether the IB␣-mediated gene activation could be mediated through IB␣ physical interactions with HDAC proteins. 293 cells were transfected with expression vectors coding for IB␣ and for various FLAG-tagged HDAC members. Cell extracts were immunoprecipitated by either anti-IB␣ or -HA (negative control) antibodies, and the immunoprecipitates were subjected to an anti-FLAG Western blot (Fig. 3A). Ectopically expressed IB␣ was associated with FLAG-tagged HDAC1 and HDAC3 (Fig. 3A, upper panel, lanes 2 and 6, respectively) but not with HDAC2, -4, and -5 (Fig. 3A, upper panel, lanes 4, 8, and 10, respectively) or with HDAC6 (Fig. 3B, upper panel, lane  3). Similar experiments performed with extracts from untransfected cells showed that this complex occurs physiologically since endogenous HDAC1 or HDAC3 proteins interacted with endogenous IB␣ (Fig. 3C, upper panels, lanes 2). Therefore, our results suggest that IB␣ associates with some, but not all, histone deacetylase complexes in vivo.
To identify the IB␣ domain mediating the interaction with the HDAC proteins, co-immunoprecipitation experiments were performed to test the ability of IB␣ mutants (Fig. 4A) to interact with HDAC3. 293 cells were transfected with FLAG-HDAC3 as well as either IB␣ or IB␣⌬N, and anti-FLAG immunoprecipitation experiments followed by anti-IB␣ Western blots were carried out on the cell extracts. As expected, FLAG-HDAC3 was found to be associated with IB␣ (Fig. 4B, upper panel, lane 2) and also with IB␣⌬N (Fig. 4B, upper  panel, lane 4). Moreover, when IB␣ or IB␣⌬C were transfected with HDAC3, both proteins were detected in the anti-HDAC3 immunoprecipitates (Fig. 4C, upper panel, lanes 2 and 4, respectively) but not in the anti-HA immunoprecipitates (Fig. 4C, upper panel, lanes 1 and 4). Finally, an IB␣ mutant in which the ankyrin repeats were replaced by a GST cassette and named "NϩC GST" failed to interact with FLAG-HDAC3 in 293 cells (Fig. 4D, upper panel, lane 4), whereas the wild type IB␣ protein was again found in the anti-FLAG immunoprecipitates despite a lower level of expression compared with the NϩC GST mutant (Fig. 4D, upper panel, compare lanes 2 and 4, respectively). Taken together, these results strongly suggest that the ankyrin repeats of IB␣ are critical in mediating the interaction with HDAC3. This conclusion is supported by the ability of a GST-ankyrin fusion protein to interact with in vitro translated HDAC3 (data not shown).
p65 Disrupts the Interaction between IB␣ and HDAC3-We next determined the ability of IB␣ to bind HDAC3 in the presence of increasing amounts of p65 by immunoprecipitating IB␣ from transfected 293 cells and by performing anti-HDAC3 Western blots on the immunoprecipitates (Fig. 5A). Increasing amounts of p65 interfered with the ability of FLAG-HDAC3 to bind IB␣ (Fig. 5A, upper panel, lanes 2-4). These results suggest that p65 and HDAC3 compete to bind a common domain of IB␣, most likely its ankyrin repeats. We then performed double immunoprecipitation experiments to investigate whether p65, HDAC3, and IB␣ are part of a common complex. 293 cells were transfected with FLAG-IB␣, and total cell extracts were subsequently subjected to anti-FLAG or anti-HA immunoprecipitations. The immunoprecipitates were then released from the beads by incubation with the FLAG peptide, and the released material was subsequently immunoprecipitated with either anti-HA or with antibodies to endogenous HDAC3 or p50 proteins. Anti-p65 immunoblots were carried out on the resulting immunoprecipitates. As expected, the p50⅐p65⅐IB␣ ternary complex was detected (Fig. 5B, upper  panel, lane 3). On the other hand, no ternary complex including p65, HDAC3, and IB␣ was detectable (Fig. 5B, upper  lane 6). These results suggest that IB␣ binds HDAC3 independently of p65.
HDAC3, but Not HDAC2, Attenuates IB␣-mediated Positive Effect on Pit-1 and HOXB7 Transactivation Potential-We next examined the biological relevance of the interaction of IB␣ with some HDAC proteins and noticed that FLAG-HDAC3 strongly decreased IB␣-mediated enhancement of Pit-1-dependent transcription on the PRL 164-bp reporter plasmid, whereas similar amounts of FLAG-HDAC2, which does not interact with IB␣ (see Fig. 3A), had a much weaker effect (Fig.  6A). An anti-FLAG Western blot was carried out on the cell extracts to make sure that similar amounts of both FLAGtagged HDAC proteins were expressed (Fig. 6A). We next investigated whether NF-B is somehow involved in the IB␣mediated positive effect on the homeodomain-containing protein transactivation potential by using a reporter plasmid that does not harbor any B binding site and is therefore not induced by p50/p65. Because the PRL 164-bp luciferase reporter plasmid turned out to harbor a functional B site (data not shown), we selected the pTCBS reporter plasmid, which contains an eightfold multimerized form of a homeodomain CBS and whose activity is not induced by p50/p65 (10), and tested the effect of FLAG-HDAC3 on such reporter plasmid. A similar dose-dependent decrease of the luciferase activity was detected when the pTCBS reporter plasmid was cotransfected with HOXB7, IB␣, and increasing amounts of FLAG-HDAC3 as compared with the luciferase activity measured in the absence of FLAG-HDAC3 (Fig. 6B). These results suggest that the HDAC proteins differentially modulate IB␣-mediated enhancement of gene transcription, and this is correlated with their ability to physically interact with IB␣. This newly described IB␣ function is NF-B-independent since it also occurs on promoters lacking any B binding site.
IB␣ Sequesters a Pool of HDAC3 Proteins in the Cytoplasm-The possibility that IB␣ causes HDAC3 cytoplasmic sequestration was investigated by studying the subcellular localization of ectopically expressed HDAC3 in HeLa cells either transfected or not with an IB␣ expression vector. Transfected  2 and 3), respectively. The amounts of transfected plasmids are mentioned below the panels. C, endogenous HDAC3 and HDAC1 interact with IB␣ in vivo. 293 cells were lysed, and extracts were subjected to immunoprecipitations with anti-HDAC1, -HDAC3 (left and right panels, respectively, lanes 2), or -HA (lanes 1) antibodies followed by anti-IB␣ Western blots. The expression of HDAC1, HDAC3, and IB␣ is illustrated in the middle and lower panels, respectively.
HDAC3 is found in the nucleus in most cells, although some cytoplasmic staining was observed in some cases, whereas ectopically expressed IB␣ can be found in both cell compartments. In the absence of ectopically expressed IB␣, HDAC3 was nuclear in 47% of the counted cells, and only 13% of the cells showed a predominant cytoplasmic localization. Interestingly, a greater amount of cells (47%) showed a cytoplasmic staining of HDAC3 when both HDAC3 and IB␣ were co-  2) and mutant (lanes 3 and 4) as well as FLAG-HDAC3 (lanes 2 and 4) in the extracts is shown in the middle and lower panels, respectively. C, IB␣⌬C interacts with HDAC3. IB␣ or IB␣⌬C were co-transfected with HDAC3 and subjected to an anti-HA (negative control, lanes 1 and 3) or anti-HDAC3 immunoprecipitation (lanes 2 and 4) followed by an anti-IB␣ Western blot (upper panel). Anti-IB␣ and -HDAC3 Western analysis was also performed on the cell extracts (middle and lower panels, respectively). The amounts of transfected plasmids are mentioned below the panels. D, IB␣ but not NϩC GST interact with HDAC3. 293 cells were transfected with IB␣ (lanes 1 and 2) or NϩC GST (lanes 3 and 4) with (lanes 2 and 4) or without FLAG-HDAC3 (lanes 1 and 3), and anti-FLAG immunoprecipitation followed by anti-IB␣ Western blots was performed (upper panel). The presence of FLAG-HDAC3 as well as wild type IB␣ or the NϩC GST mutant in the cell extracts is illustrated in the bottom panels. IP, immunoprecipitation; WB, Western blot. expressed (data not shown). Therefore, these results demonstrate that ectopic expression of IB␣ increases the cytoplasmic localization of HDAC3. Recent studies have demonstrated that IB␣ constantly shuttles between the cytoplasm and the nucleus. To explore whether this shuttling affects the ability of IB␣ to interact with HDAC3, we examined by co-immunoprecipitation experiments whether an IB␣ mutant (IB␣ NLS-MT) lacking a functional NLS (2) could still interact with HDAC3. As expected, this mutant was mainly localized in the cytoplasm when transfected into 293 cells, whereas the wild type IB␣ protein was also found in the nucleus (Fig. 7A). Interestingly, despite low expression levels as compared with wild type IB␣ (Fig. 7B, lower panel, compare lanes 3 and 4  with lanes 1 and 2), IB␣ NLS-MT strongly interacted with HDAC3 or HDAC1 as judged by immunoprecipitation experiments of cells transfected with IB␣ (lanes 1 and 2) or IB␣ NLS-MT (lanes 3 and 4) and with either FLAG-HDAC1 or FLAG-HDAC3 (Fig. 7B, upper panels on the left and right,  compare lanes 2 and 4). Moreover IB␣ NLS-MT efficiently enhanced Pit-1-mediated transcription through the PRL promoter (Fig. 7C). Taken together, these results therefore suggest that IB␣ enhances Pit-1 transactivating activity through cytoplasmic sequestration of transcriptional repressors such as HDAC1 and -3.

DISCUSSION
The IB␣ protein is ubiquitously expressed as an NF-B cytoplasmic inhibitor. In response to various stimuli, IB␣ is rapidly degraded to allow NF-B nuclear translocation and NF-B target gene transcription. As IB␣ itself is the product of an NF-B-regulated gene, it is rapidly resynthesized and can then migrate to the nucleus to terminate the first phase of NF-B activation. Our present data demonstrate that this IB␣ inhibitor can also associate with HDAC proteins and sequester them in the cytoplasm. Although we do not have any direct evidence that newly synthesized IB␣ binds to HDACs in response to NF-B-activating stimuli and brings them to the cytoplasm, it is possible that such activity participates in NF-B target gene transcription. Indeed, upon stimulation, two waves of NF-B recruitment to its target genes occur (22). The first wave involves immediately accessible genes and appears to be terminated through IB␣mediated removal of NF-B from its binding sequences (23), whereas the second wave involves genes whose promoter accessibility requires stimulus-dependent modification of chromatin structure (22). This second wave appears to rely on IB␤ and IB⑀, which stabilize NF-B responses during longer stimulations (23) and, as suggested by the present data, could possibly be favored by IB␣-mediated cytoplasmic sequestration of HDACs.
We previously demonstrated that IB␣ physically interacts with HOXB7, and this interaction was required to mediate IB␣ positive effect on HOXB7-driven transcription (10). This mechanism presumably relies on a nuclear localization of IB␣, which has been clearly established and would imply that IB␣ is capable of recruiting some coactivators in the vicinity of HOXB7 as does the transactivating protein BCL-3, another ankyrin-containing protein of the same family, with the coactivators BARD1 and JAB1 (9). We indeed detected physical FIG. 5. IB␣ binds HDAC3 independently of p65. A, p65 and HDAC3 compete for binding to IB␣. Cells were transfected with IB␣, FLAG-HDAC3, and increasing amounts of p65 (0.5 or 5 g, lanes 3 and 4, respectively). Extracts were subjected to anti-HA (lane 1) or anti-IB␣ (lanes 2-4) immunoprecipitations, and the presence of FLAG-HDAC3 in the immunoprecipitates was detected by anti-FLAG Western blot (upper panel). The presence of transfected IB␣, FLAG-⌯DAC3, and p65 in the cell extracts is shown in the three lower panels. The amounts of transfected plasmids are mentioned below the panels. B, evidence for a ternary complex associating FLAG-IB␣ with endogenous p50 and p65 but not with endogenous HDAC3 and p65. 293 cells were transfected with 3 g of FLAG-IB␣, (lanes 1-6). Cells were harvested 24 h later, and extracts were subjected to an anti-HA (negative control, lanes 1 and 4) or anti-FLAG immunoprecipitation (lanes 2, 3, 5, and 6). After extensive washes, the FLAG immunoprecipitates were released from the beads by incubating them with a FLAG peptide overnight. Then a second immunoprecipitation was performed with anti-HA, -p50, or -HDAC3 antibodies followed by an anti-p65 Western analysis (top panel). The bottom four panels show Western analyses for FLAG-IB␣, endogenous HDAC3, p65, and p50 proteins. IP, immunoprecipitation; WB, Western blot.
interactions of IB␣ with BARD1 and JAB1, supporting this hypothesis (data not shown). However, our present report demonstrates that a cytoplasmic IB␣ mutant could still strongly induce Pit-1 transactivation potential, therefore suggesting that sequestration of HDACs is the predominant mechanism that accounts for the observed effects. Such a statement, however, does not rule out the ability of IB␣ to harbor some transactivating potential in the nucleus through physical interactions with some transcription factors and recruitment of co-activators.
Evidently HDAC sequestration indicates that IB␣ also exerts NF-B-independent activities. Indeed the positive effect of IB␣ occurs on promoters limited to the CBS, which lacks any B binding site, and we previously showed that the heterodimer p50/p65 does not enhance the IB␣ effect on the pTCBS reporter but rather attenuates it (10). Moreover, we have no evidence for a ternary complex that includes p65, HDAC3, and IB␣ occurring. Rather IB␣ binds HDAC3 independently of p65, supporting the hypothesis that IB␣ harbors NF-B-independent functions. Most NF-B-activating stimuli (proinflammatory cytokines, viruses, DNA damage, etc.) also activate other transcription factors. Our data therefore suggest that IB␣ could favor the transcription of genes controlled by these regulating complexes through the maintenance of histone acetylation and promoter accessibility. In other words, once the first round of NF-B-dependent transcription is over, HDAC/ IB␣ interactions would favor the activity of other induced transcription factors.
However, our data demonstrated that endogenous IB␣ and HDACs physically interact in the absence of any stimulus. Therefore, the ubiquitous IB␣ cytoplasmic pool could sequester some HDACs and thus favor either basal cellular transcription or transcription in various conditions where NF-B is not activated. It is currently unclear whether IB␣ brings HDAC1 and -3 back to the cytoplasm along with NF-B proteins after stimulation or whether a pool of these HDACs is sequestered in the cytoplasm through binding to molecules such as IB␣ prior to their shuttling to the nucleus. Whether or not nuclear IB␣ brings some HDAC proteins along with p65 back in the cytoplasm remains to be investigated.
HDAC molecules regulate histone deacetylation and can thus influence the activity of many transcription factors. However, IB␣ interacts specifically with HDAC1 and -3 and leaves the other HDACs unaffected. Among the candidate transcription factors whose activity could be influenced by HDAC1 and -3 sequestration, we demonstrated that transcriptional activity of HOX proteins is enhanced by IB␣ expression. Several experiments suggest putative functional interactions between NF-B and HOX proteins. IB␣ knock-out mice are smaller than their wild type littermates, but it remains unclear whether this size difference is due to hormonal defect (GH for instance) or to inflammatory reactions. Moreover, NF-B and several HOX proteins are simultaneously expressed at various developmental stages, and local NF-B inhibition following expression of a mutated IB␣ has been demonstrated to interfere with limb development (24).
As the present report demonstrates a novel NF-B-independent function for IB␣, it will be most important to identify the genes whose expression is regulated by IB␣ and by the interaction of IB␣ with HDACs. These genes are probably highly cell-specific according to the expressed transcription factors and to promoter accessibility. The identification of the IB␣regulated transcription factors and genes will open a new field of investigation related to a novel and unexpected IB␣ activity.