IκB-α Enhances Transactivation by the HOXB7 Homeodomain-containing Protein*

Combinatorial interactions between distinct transcription factors generate specificity in the controlled expression of target genes. In this report, we demonstrated that the HOXB7 homeodomain-containing protein, which plays a key role in development and differentiation, physically interacted in vitro with IκB-α, an inhibitor of NF-κB activity. This interaction was mediated by the IκB-α ankyrin repeats and C-terminal domain as well as by the HOXB7 N-terminal domain. In transient transfection experiments, IκB-α markedly increased HOXB7-dependent transcription from a reporter plasmid containing a homeodomain consensus-binding sequence. This report therefore showed a novel function for IκB-α, namely a positive regulation of transcriptional activation by homeodomain-containing proteins.

Multiple transcription factors establish combinatorial interactions to achieve their in vivo specificity. These protein-protein interactions modulate the activating or repressing abilities of the complexes. The identification of all the partners interacting with a transcription factor is thus essential for the understanding of its biological functions.
Homeodomain-containing proteins are transcription factors that play a crucial role in the development of many species, including humans (1)(2)(3)(4). They share a highly conserved 60amino acid DNA-binding domain, the homeodomain, and control the expression of many target genes, most of which remain unknown (5). These proteins are encoded by 39 HOX genes, which are organized in four clusters (loci A, B, C, and D) located on chromosomes 7, 17, 12, and 2, respectively (6). Interestingly, their pattern of expression along the anteroposterior axis of the developing embryo is closely related to their chromosomal position on the cluster (7), defining a "spatial colinearity." Although homologous recombination experiments have clearly demonstrated their in vivo specificity, all the HOX gene products bind to very similar sequences in vitro (8 -10). Their specificity may thus be achieved not only through DNA-protein interactions but also through protein-protein interactions with other transcription factors whose identities remain largely unknown. Among these partners, the extradenticle/Pbx homeodomain-containing proteins were the first to be identified as co-factors for HOX proteins (11). Interaction with the PBX protein requires the pentapeptide, a conserved domain located upstream of the DNA-binding domain of most HOX gene products and required for the interactions of HOX proteins with other peptides (11), as well as the HOX cooperativity motif, a sequence C-terminal to the Pbx homeodomain (12). Because AbdB-like HOX proteins do not harbor any pentapeptide-like sequence, they cannot interact with Pbx proteins (13), thus suggesting that other partners might be involved. Indeed, a recent report has illustrated the existence of heterodimeric complexes between HOX and Meis1 proteins (13). Moreover, it is likely that other proteins yet to be identified also interact with HOX proteins and contribute to their biological function.
HOXB7 cDNA was initially isolated from an SV40-transformed human fibroblast cDNA library (14). The HOXB7 protein is involved in a variety of developmental processes, including hematopoietic differentiation and lymphoid development (15)(16)(17). Because of its expression in lymphoid and nonlymphoid cells, the HOXB7 protein might be involved in the regulation of a common transcriptional event rather than in lineage-specific gene expression (18). However, despite the demonstration of HOXB7 protein binding to DNA (19), little is known about its transcriptional properties and interacting partners in vivo. We first demonstrated that the HOXB7 protein as well as a naturally occurring mutant harboring a truncated C-terminal tail both transactivate from a HOX-binding consensus sequence in breast cancer cells (20) and physically interact with the coactivator CREB-binding protein. 1 The NF-B proteins form a family of transcription factors that play a central role in the cellular responses to stress, cytokines, and pathogens (22)(23)(24). Indeed, these transcription factors are activated in response to a variety of extracellular signals such as phorbol esters, tumor necrosis factor-␣, interleukin-1, lipopolysaccharide, UV irradiation, viral infection, and growth factors (24,25) and regulate a wide spectrum of immune and inflammatory responses (26). In unstimulated cells, NF-B activity is inhibited by another class of proteins that includes IB-␣ (27,28), IB-␤ (29), IB-⑀ (30), p105, and p100. These inhibitory proteins all share ankyrin repeats, sequester the NF-B complexes in the cytoplasm, and block their binding to B DNA sequences. Initially described as a cytoplasmic protein (28), IB-␣ has since been detected in the nucleus of transfected Vero cells (31) as well as after serum stimulation * This work was supported by grants from the National Fund for Scientific Research, Télévie (Belgium) and the Centre Anti Cancéreux (University of Liège, Belgium). 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.
§  (32). The nuclear localization of IB-␣ is mediated by its second ankyrin repeat, which acts as a nuclear import sequence (33). Once in the nucleus, IB-␣ can remove NF-B dimers from their B DNA sequences, thus inhibiting NF-B activity (34). When fused to the GAL-4 DNA-binding domain, IB-␣ displays transactivation abilities (35,36), a property not possessed by the naturally occurring IB-␣ protein (32). These results raised the possibility that IB-␣ can interact with other transcription factors and modulate their activity.
In this report, we demonstrated that IB-␣ is able to physically interact with the HOXB7 homeodomain-containing protein and to enhance HOXB7 transcriptional activity. We further identified HOXB7 and IB-␣ domains involved in this interaction. Our results thus demonstrate a novel function of IB-␣.

EXPERIMENTAL PROCEDURES
Cell Cultures-The MDA-MB231 cell line was obtained from the American Type Tissue Collection (Rockville, MD). The cells were maintained in RPMI 1640 medium supplemented with 10% fetal bovine serum (Life Technologies, Inc.) and antibiotics.
The mammalian PMT 2 T expression vectors for p50, RelA, and IB-␣ were previously described (37,38). The IB-␣ coding sequence was also subcloned by PCR into the expression vector pcDNA3. The PMT 2 T expression vectors for IB-␣⌬N and IB-␣⌬C lacking the first 53 codons and the last 42 codons of IB-␣, respectively (39), are schematically illustrated in Fig. 1A. The PMT 2 T expression vector for IB-␣ NϩC GST codes for a protein where the ankyrin repeats of IB-␣ have been replaced by the GST peptide (Fig. 1A) as described (39).
Both the pT109 and pTCBS reporter plasmids were provided by Dr. Zappavigna (Laboratory of Gene Expression, Department of Biology and Technology, Instituto Scientifico H. S. Raffaele, Milan, Italy). The pTCBS plasmid contains an 8-fold multimerized form of a homeodomain consensus-binding sequence (CBS) cloned upstream of an HSV-TK promoter and a luciferase reporter gene, whereas the pT109 construct does not contain the CBS sequence and was thus used as a negative control. The B-ICAM-1 reporter plasmid construct has been previously described; it harbors three NF-B-like sites from the ICAM-1 promoter cloned upstream of the herpes simplex virus thymidine kinase minimal promoter and the luciferase gene (40).
In Vitro Translation-In vitro transcription and translation were performed using the Wheat Germ TNT kit provided by Promega (Madison, WI) with 1 g of various DNA templates and [ 35 S]methionine, according to the protocol provided by the manufacturer.
In Vitro Protein-Protein Interactions-GST fusion proteins were produced in the Escherichia coli BL21 bacterial strain. Bacteria were grown in 500 ml of Luria broth to an A 600 nm of 0.6, induced with 0.5 mM isopropyl-1-thio-␤-D-galactopyranoside for 3 h and harvested. Bacterial pellets were washed once with phosphate-buffered saline, resuspended in 10 ml of NENT buffer (250 mM NaCl, 1 mM EDTA, 20 mM Tris, pH 8, Nonidet P-40 1.5%) and sonicated three times for 15 s at 4°C. Insoluble materials were removed by centrifugation. GST fusion proteins were purified after incubation of 1 ml of the supernatant with 10 l of glutathione-Sepharose beads for 1 h at 4°C (Amersham Pharmacia Biotech). The beads were then washed twice with 1 ml of NENTM buffer (NENT ϩ 0.5% milk) and once with 1 ml of TWB buffer (20 mM

FIG. 1. Schematic illustration of the IB-␣ expression vectors (A) and the GST-IB-␣ constructs (B).
The ankyrin repeats are illustrated by dark rectangles. The primers designed for PCR amplification are numbered from 1 to 6 and represented by arrows. Primers 4 -6 are derived from the complementary strand.

FIG. 2. NF-B and IB-␣ increase transactivation by HOXB7
through a HOX consensus DNA-binding sequence. MDA-MB231 cells were transfected with 1 g of HOXB7, p50, RelA, and/or IB-␣ expression vectors together with 1 g of reporter plasmid, as indicated in the figure. The pT109 does not contain any HOX-binding sequence and was used as a negative control. The figure shows the relative luciferase activity over the basal activity observed with 1 g of the pTCBS or pT109 reporter plasmid alone. Each value represents the mean (Ϯ S.D.) of at least three independent experiments after normalization to the protein concentration of the extracts.
Hepes, pH 7.9, 60 mM NaCl, 1 mM dithiothreitol, 6 mM MgCl 2 , 8.2% glycerin, 0.1 mM EDTA). In each case, the expected fusion proteins were visualized on a 12% polyacrylamide gel stained by Coomassie Blue. Protein-protein interactions were performed by incubating an aliquot of the GST-IB␣ fusion protein bound to the glutathione-Sepharose beads with 10 l of in vitro translated protein in 200 l of TWB buffer for 1 h at 4°C. Beads were then washed six times with 1 ml of NENTM buffer, resuspended into migrating buffer, and loaded on an SDS-polyacrylamide gel before autoradiography.
Transient Transfections and Luciferase Assays-Transfections of MDA-MB231 cells were performed as described (41), using 1 g of reporter plasmid (either pTCBS or pT109) and various amounts of vectors expressing HOXB7, RelA, p50, and/or IB-␣. Total amounts of transfected DNA were kept constant throughout by adding appropriate The homeodomain is illustrated by a large shaded rectangle, whereas the pentapeptide is represented by a small shaded box upstream of the homeodomain, and the acidic C-terminal tail is shown as a crosshatched box. The expected molecular mass of the resulting proteins is mentioned on the right. 35 S-Labeled in vitro translated wild-type and deleted HOXB7 proteins were incubated with a GST-IB-␣ fusion protein attached to glutathione-Sepharose beads (lanes 2, 5, 8, 11, and 14), precipitated and run on an SDSpolyacrylamide gel. Beads carrying the GST protein alone were used as negative controls (lanes 3, 6, 9, 12, and 15). In vitro translated proteins (10% of the amounts used in the precipitation experiments) were run on lanes 1, 4, 7, 10, and 13. amounts of either pcDNA3 or pMT 2 T empty vectors. Cells were harvested 48 h after transfection, and luciferase activities were measured with the Luciferase Reporter Gene Assay kit (Boehringer Mannheim), as recommended by the manufacturer. The luciferase activities were normalized to the protein concentration of the extracts.

RESULTS
p50, RelA, and IB-␣ Enhance Transactivation by the HOXB7 Protein-To investigate whether the HOXB7 protein can interact with transcription factors from other families, we transiently transfected MDA-MB231 cells with a HOXB7 expression vector and a variety of constructs coding for different members of the NF-B/IB families. Both the pTCBS and pT109 constructs were used as reporter plasmids: the pTCBS plasmid contains a luciferase reporter gene driven by a multimerized HOX CBS that is recognized by most HOX proteins, whereas the pT109 vector does not harbor any HOX-binding sequence and was used as a negative control (42). A 3.6-fold induction over basal luciferase activity was measured when the HOXB7 expression vector was transfected with pTCBS ( Fig. 2,  column 3), as described previously (20,41). This effect was mediated by the binding of the HOXB7 protein to the CBS sequence, because no significant effect was observed with the pT109 reporter plasmid (Fig. 2, column 4).
When p50 and RelA expression vectors were transfected with the pTCBS or pT109 reporter plasmids, weak inductions of luciferase activity were observed (Fig. 2, columns 5 and 6).
Moreover, a very weak increase in luciferase activity was observed when the plasmid encoding IB-␣ was co-transfected with either the pTCBS or pT109 reporter constructs (Fig. 2,  columns 7 and 8), indicating that, as expected, IB-␣ did not transactivate through these promoters in MDA-MB231 cells. When the HOXB7 expression construct was co-transfected with p50, RelA, and pTCBS, a 4.9-fold induction over basal luciferase activity was observed (Fig. 2, column 9), indicating that NF-B members enhanced HOXB7 transcriptional activity. To determine whether IB-␣ could inhibit the transactivation observed with HOXB7 and p50-RelA, we co-transfected an IB-␣ expression vector with the plasmids generating the HOXB7, p50, and RelA proteins as well as with the pTCBS construct. Surprisingly, a further increase in luciferase activity (7.2-fold induction over basal luciferase activity) was measured (Fig. 2,  columns 11). Moreover, the luciferase activity was even more elevated (13.7-fold induction over basal luciferase activity) when we co-transfected only the HOXB7 and IB-␣ expression vectors with the pTCBS reporter plasmid (Fig. 2, column 13).
To further characterize the transcriptional properties of the HOXB7 protein, additional transient expression experiments were performed using the B-ICAM-1 reporter plasmid harboring three B-like binding sites upstream of a CAT gene. As expected, transfection of the p50 and RelA expression vectors induced CAT activity (Fig. 3, column 2), and this effect was inhibited by simultaneous expression of IB-␣ (Fig. 3, column   FIG. 5. In vitro protein-protein interaction between HOXB7 and IB-␣ does not require the HOXB7 C-terminal domain. The HOXB7 expression vector generating proteins deleted in their C-terminal domain are schematically represented (B7-⌬C12, B7-⌬C34, B7-⌬C80, and B7-⌬C97). The B7* protein is a naturally occurring mutant deleted of two C-terminal amino acids. The homeodomain is illustrated by a large shaded rectangle, whereas the pentapeptide is represented by a small shaded box upstream of the homeodomain, and the acidic Cterminal tail is shown as a cross-hatched box. The expected molecular mass of the resulting proteins are mentioned on the right. 35 S-Labeled in vitro translated wild-type and deleted HOXB7 proteins were incubated with a GST-IB-␣ fusion protein attached to glutathione-Sepharose beads (lanes 2, 5, 8, 11, 14, and 17), precipitated, and run on an SDS-polyacrylamide gel. Beads carrying the GST protein alone were used as negative controls (lanes 3, 6, 9, 12, 15, and 18). In vitro translated proteins (10% of the amounts used in the precipitation experiments) were run on lanes 1, 4, 7, 10, 13, and 16. 3). Transfection of increasing amounts of HOXB7 expression vector did not lead to any significant induction of CAT activity (Fig. 3, columns 4 -6). Moreover, when we co-transfected HOXB7 with both p50 and RelA expression vectors, the CAT activity was close to that measured in the absence of HOXB7 (Fig. 3, columns 2 and 8) and was attenuated by the inhibitor IB-␣ (Fig. 3, columns 10 -12). No significant induction of CAT activity was measured when HOXB7 and IB-␣ expressing vectors were co-transfected (Fig. 3, columns 14 -16). These results suggest that the HOXB7 protein does not significantly modulate the transcriptional abilities of NF-B members.
IB-␣ Physically Interacts in Vitro with the N-terminal Domain of HOXB7-To determine whether IB-␣ physically interacted with the HOXB7 protein, purified GST-IB-␣ fusion protein bound to glutathione-Sepharose beads was incubated with in vitro translated HOXB7. After precipitation of the beads, a positive signal was detected (Fig. 4, lane 2), whereas HOXB7 did not interact with the GST protein (lane 3), thus demonstrating the existence of an in vitro interaction between HOXB7 and IB-␣. To map the HOXB7 domain involved in this process, we designed several constructs generating HOXB7 gene products progressively deleted in their N-terminal domain and designated as B7-⌬N18, ⌬N54, ⌬N86, and ⌬N129. All these HOXB7 proteins shared an intact homeodomain, whereas only the ⌬N129 peptide was deleted of the pentapeptide. These products were then in vitro translated and incubated with the GST-IB-␣ fusion protein as described above. None of these proteins were able to significantly interact with IB-␣ (Fig. 4, lanes 5, 8, 11, and 14).
To determine whether other domains were involved in the interaction, additional constructs including a naturally occurring HOXB7 mutant that lacks two amino acids at the Cterminal tail (B7*) (20) and HOXB7 products deleted in their C-terminal domain (B7-⌬C12, B7-⌬C34, B7-⌬C80, and B7-⌬C97) were in vitro translated (Fig. 5, lanes 1, 4, 7, 10, 13, and 16) and incubated with the GST-IB-␣ fusion protein bound to glutathione-Sepharose beads. All these proteins were still able to interact with IB-␣ despite the absence of a complete homeodomain sequence for B7-⌬C80 and of the pentapeptide region for B7-⌬C97 (Fig. 5, lanes 5, 8, 11, 14, and 17). These results indicate that a HOXB7/IB-␣ physical interaction can occur independently of the homeodomain sequence and depends exclusively on an intact HOXB7 N-terminal sequence.
The N-terminal Domain of the HOXB7 Protein Is Required for the Interaction with IB-␣ in Vivo-We previously demonstrated that both the N-terminal domain and the acidic Cterminal tail of the HOXB7 protein mediated its transcriptional properties. 1 Because the N-terminal domain of HOXB7 was required for the interaction with IB-␣ in vitro, we transfected MDA-MB231 cells with the B7⌬N129 expression vector and the pTCBS or pT109 reporter plasmid. The B7⌬N129 product, alone or co-expressed with IB-␣, did not induce any luciferase activity (Fig. 6). Moreover, the B7-⌬C12 protein, which lacks the acidic C-terminal domain but still interacts with IB-␣ in vitro (Fig. 5), did not behave as a transcriptional activator (Fig. 6). Interestingly, when both the B7-⌬C12 and IB-␣ expression vectors were transfected simultaneously with the pTCBS reporter plasmid, an induction of the luciferase activity similar to that measured with both HOXB7 wild-type and IB-␣ proteins was observed (Fig. 6). These results suggest that the inhibitor IB-␣ potentiates HOXB7 transactivating activities through a physical interaction with the HOXB7 Nterminal domain.
The Ankyrin Repeats and the C-terminal Domain of IB-␣ Are Required for in Vitro and in Vivo Interaction with HOXB7-To map the IB-␣ domain(s) involved in the physical interaction with HOXB7, we inserted the ankyrin domain of IB-␣ in the pGex-2TK vector (Fig. 7A, lane 2) and produced the corresponding fusion protein. Incubation of this protein with in vitro translated HOXB7 demonstrated a physical interaction between the two proteins (Fig. 7A, lane 2), whereas HOXB7 could not interact with the GST protein purified from E. coli (Fig. 7A, lane 3). Because the signal was weaker than that observed with the full-length GST-IB-␣ fusion protein (Fig. 7A, lane 1), it is likely that other functional domains were also involved in this process. An IB-␣ peptide deleted of the C-terminal domain was fused to GST (Fig. 7B) and purified. As illustrated in Fig. 7B (lane 4), this fusion protein bound to glutathione-Sepharose beads interacted with HOXB7, although this interaction was weaker than that obtained with the full-length IB-␣ product (Fig. 7B, lane 1). Moreover, a GST-IB-␣-C-terminal domain fusion protein was still able to interact with HOXB7 (Fig. 7B, lane 5). To confirm that both the ankyrin and C-terminal domains of IB-␣ were responsible for the interaction, a fusion protein that does not contain the IB-␣-N-terminal domain was produced and incubated with HOXB7. As illustrated in Fig. 7C (lane 6), a signal similar to that obtained with the wild-type IB-␣ protein was observed, suggesting that the N-terminal domain of IB-␣ is not involved in the interaction with HOXB7. Indeed, incubation of the GST-IB-␣-N-terminal domain fusion protein with HOXB7 did not generate any significant signal (Fig. 7C, lane 7). These results indicate that both the IB-␣ ankyrin and C-terminal domains are required for physical in vitro interaction with HOXB7.
To confirm these results in vivo, we performed transient expression experiments using both pT109 and pTCBS constructs as reporter plasmids and a variety of vectors generating distinct IB-␣ peptides (Fig. 1A). Full-length IB-␣, IB-␣ ⌬C, IB-␣ ⌬N, and N IB GSTC IB proteins did not significantly induce luciferase activity when co-transfected with any of the reporter plasmids (Fig. 8, rows 5-12). However, cotransfection of the plasmids generating the IB-␣ ⌬N and HOXB7 peptides led to a 13-fold induction of luciferase activity (Fig. 8, row 17) that was dependent on the binding to the CBS sequence (row 18) and comparable with that measured with the full-length IB-␣ expression vector (row 19). Cotransfection of the expression plasmids for IB-␣ ⌬C and HOXB7 generated only a 4.3fold induction over the basal activity (Fig. 8, row 15). To confirm that the ankyrin domain was also required, we cotransfected the pTCBS or the pT109 reporter plasmids with the HOXB7 plasmid and the N IB GSTC IB construct generating an IB-␣-related protein in which the ankyrin domain had been replaced by the GST sequence. As illustrated in Fig. 8  (row 13), we observed only a 6.4-fold induction of the luciferase activity. Our results clearly demonstrate that the ankyrin repeats and the C-terminal domain of IB-␣ are required for in vitro and in vivo interaction with the HOXB7 homeodomaincontaining protein.

DISCUSSION
This report has demonstrated a physical interaction between the HOXB7 homeodomain-containing protein and IB-␣, resulting in an enhanced transactivation by this HOX gene product. Moreover, we identified the HOXB7 and IB-␣ functional domains mediating this interaction. These results provide new insights into the transcription properties of the homeodomaincontaining proteins and reveal a novel function of the inhibitor IB-␣.
All the homeodomain-containing proteins encoded by the 39 HOX genes share a highly conserved 60-amino acid DNA-binding domain, the homeodomain, and bind to very similar sequences in vitro (8 -10). Their in vivo specificity may thus involve protein-protein interactions with other transcription factors. In this context, the homeodomain proteins derived from the extradenticle/Pbx genes act as co-factors for HOX gene products that contain a pentapeptide sequence (11), whereas the AbdB-like HOX proteins, which do not harbor a pentapeptide, interact with Meis1 (13). We have provided here evidence that the NF-B proteins, including the p50-p65 heterodimer, can enhance the transcription potential of the HOXB7 protein in transient expression experiments. This effect is presumably mediated by physical interactions between the p50-p65 complex and HOXB7. Preliminary in vitro experiments have indeed confirmed this hypothesis (data not shown). It is tempting to speculate the existence of a p50-p65-HOXB7 complex that could bind the CBS consensus sequence and transactivate through the HOXB7 and p65 activation domains. This complex, however, did not display a similar effect on a B-binding site under our experimental conditions.
We have shown that the inhibitor IB-␣ can enhance the HOXB7 transactivating effect. This is the first demonstration that IB-␣ interacts with proteins from other families of transcription factors. Previous reports had demonstrated that IB-␣ can translocate to the nucleus, using its second ankyrin repeat as a nuclear import sequence (33) and subsequently remove the NF-B complex from its binding site (34). Moreover, IB-␤ can repress the 9-cis-retinoic acid-induced transcriptional activity of retinoid X receptor in lipopolysaccharidetreated cells (43). Thus, in both cases, IB proteins localized in the nucleus negatively regulate the transcriptional activity of their interacting partners. Surprisingly, our study demonstrates that IB-␣ can also positively regulate the transcriptional properties of a homeodomain-containing protein. A similar phenomenon was described previously for Bcl3, another member of the IB protein family. Indeed, Bcl3 can transactivate through B sites when physically associated with p52 and p50 (37,44), and it has been demonstrated that the N-and C-terminal domains of Bcl3 are transcriptional activation domains (37). Moreover, Bcl3, but not IB-␤, can also act as a coactivator of the retinoid X receptor (45). These results and the present report strongly suggest that distinct IB proteins can modulate, positively as well as negatively, the transcriptional properties of their interacting partners, including transcription factors that do not belong to the NF-B family.
Interestingly, we demonstrated that both the IB-␣ ankyrin and C-terminal domains mediate interaction with HOXB7. The same domains are also required for the regulation of c-Rel by IB-␣ in the nucleus (46). Taken together, these results suggest a critical role for the ankyrin repeats and C-terminal domains in the function of IB-␣ in the nucleus.
Several models, which are not mutually exclusive, can explain how IB-␣ enhances HOXB7 transcriptional activity (Fig.  9). The first model does not imply a direct interaction between HOXB7 and IB-␣ but rather an indirect mechanism mediated by NF-B. Indeed, we can postulate that NF-B activates the expression of HOX genes encoding repressors. Therefore, NF-B inhibition by IB-␣ would lead to a decreased expression of these HOX genes and to increased luciferase activity in transient expression experiments (Fig. 9A). This first model is the only one that does not require IB-␣ nuclear localization. The second model is based on a report demonstrating that ankyrin repeats stabilize the DNA binding of other transcription factors (47). The enhanced HOXB7 transcriptional activity would then be mediated by a stronger HOXB7 DNA binding affinity for its target sequence in the presence of IB-␣ (Fig.  9B). This hypothesis is supported by the observation that the physical interaction between HOXB7 and IB-␣ requires the ankyrin repeats. In the third model, a HOXB7-IB-␣ complex is bound to the CBS sequence through the HOXB7 homeodomain and transactivates through IB-␣. This hypothesis is supported by previous studies demonstrating that IB-␣ can transactivate when fused to a GAL4 DNA-binding domain (35,36). Moreover, Bcl3, another member of the IB-␣ family, also harbors transactivating domains (37). The fourth model implies a HOXB7-RelA-IB-␣ complex activating the luciferase gene through both the HOXB7 and RelA transactivation domains. The last two hypotheses are supported by the induction of luciferase activity observed with IB-␣ and B7⌬C12 expression vectors, whereas the B7⌬C12 protein is not able to transactivate by itself but can physically interact with IB-␣. However, the last model cannot account for the fact that the transcriptional activity was higher after transfection of the HOXB7 and IB-␣ expression vectors than in the presence of the same vectors plus RelA (Fig. 2). Further experiments are required to determine which of these models is correct. Unfortunately, the available HOXB7 antibodies did not allow us to study more precisely the HOXB7 multimeric complexes in transfected or unmodified cells.
The functional link between NF-B-IB-␣ and homeodomain proteins was unexpected because of the distinct physiological processes they control. However, a first link between these two families during the outgrowth of the vertebrate limb has recently been described (48,49). Indeed, NF-B gene expression has been detected during limb morphogenesis and the alteration of NF-B activity causes an arrest of the outgrowth (48,49). Moreover, IB-␣ is the human homologue of cactus, a protein that plays a crucial role in the dorsoventral patterning of the Drosophila embryo (21). Because HOX genes are clearly required to establish the anteroposterior axis of the developing embryo, it is tempting to speculate that the interaction between IB-␣ and HOX proteins might determine the anteroposterior and dorsoventral polarities of the embryo.