Repression of an Interleukin-4-responsive Promoter Requires Cooperative BCL-6 Function*

BCL-6 functions as a potent transcriptional repressor that binds with specificity to DNA elements bearing marked similarity to STAT recognition sequences. Previous studies have demonstrated that BCL-6 and Stat6 can both bind and regulate the Iϵ promoter that controls immunoglobulin heavy chain class switching to IgE. Examination of BCL-6-/- and BCL-6-/-Stat6-/- mice has demonstrated that BCL-6 is a repressor of IgE and that Stat6 is still required for the interleukin-4 (IL-4) induction of class switching to IgE in B cells lacking BCL-6. To define the mechanisms by which BCL-6 represses IL-4 function, we analyzed the role of BCL-6 in repressing the Iϵ promoter. There are three BCL-6-binding sites within this IL-4-responsive promoter. Analysis of Iϵ promoters that have mutated BCL-6-binding sites demonstrates that at least two of these sites are required for maximal BCL-6 repression of this locus. Footprinting analysis demonstrates that BCL-6 binds cooperatively to the two upstream binding sites in the Iϵ promoter. This cooperative binding requires the POZ domain of BCL-6. Furthermore, activated Stat6 molecules can displace BCL-6 from one of these binding sites. These data demonstrate that cooperative interaction between BCL-6 molecules is required for repression of the Iϵ promoter.

Diffuse large cell and follicular lymphomas account for the overwhelming majority of non-Hodgkin's lymphomas seen in adult populations (1). Accordingly, much attention has been focused on determining the factors responsible for the pathogenesis of these malignancies. A surprisingly high number of diffuse large cell and follicular lymphomas have translocations involving a gene termed BCL-6/LAZ3 (2)(3)(4)(5)(6)(7). BCL-6 encodes a 706-amino acid site-specific transcriptional repressor with an N-terminal POZ domain and six C-terminal zinc finger DNAbinding motifs. Much work has been directed toward elucidating the role of BCL-6 in both normal physiology and tumorigenesis. Gene targeting studies have demonstrated that disruption of BCL-6 through homologous recombination results in a striking Th2-type inflammatory disease (8 -10). These reports indicate an important role for BCL-6 in the normal regulation of signaling events downstream of IL-4 and other cytokines responsible for the generation of Th2 responses. Mod-ulation of chemokine expression provides another potential avenue of immune system regulation by BCL-6, which has been implicated in the regulation of MCP-1, MCP-3, MRP-1, MIP-1a, and IP-10 (11,12). Further studies have implicated BCL-6 in the regulation of normal cell growth and development: mice homozygous for a deletion of BCL-6 are runted, and there is some evidence to indicate that overexpression of BCL-6 may either promote or inhibit apoptosis in various cell lines (8, 10 -15). BCL-6 has also been implicated in the control of plasma cell development through the repression of Blimp-1, an important regulator of this process (16). In addition, DNA chip analysis has identified a number of genes involved in lymphocyte differentiation and cell cycle regulation as targets of repression by BCL-6 (11).
The observation that an in vitro defined binding site for BCL-6 resembles the STAT 1 consensus sequence suggests a mechanism for BCL-6 modulation of cytokine signaling whereby BCL-6 is targeted to cytokine-regulated genes by a recognition element shared between the repressor and the STAT proteins responsible for transducing cytokine-activated transcription pathways. In a previous study, we presented evidence suggesting that BCL-6 regulates the expression of a specific subset of Stat6-dependent genes (17). In particular, BCL-6 was shown to modulate transcription of the murine germline ⑀ promoter, but not the CD23b promoter. Lymphochip analysis of genes regulated by BCL-6 provides additional support for the argument that CD23 is not repressed by BCL-6 (11). The mechanism for the selective activity of BCL-6 is not known. However, preliminary studies have revealed two characteristics of the germline ⑀ promoter that may contribute to its ability to act as a target of BCL-6-mediated repression. The first is the relative strength of the BCL-6-binding site at mI⑀ Ϫ111/Ϫ102 compared with its putative binding site in the CD23b promoter, as assessed by unlabeled competition experiments (17). Second, as described in this report, DNase I footprinting assays indicate the presence of two additional BCL-6binding elements in the mI⑀ promoter. These findings suggest two non-mutually exclusive hypotheses regarding target selection by BCL-6. On the one hand, the cumulative affinities of individual BCL-6-binding elements for a promoter may determine the ability of BCL-6 to regulate transcription of a given promoter. Alternatively, the presence of multiple recognition sites may drive BCL-6 binding via cooperative interactions between BCL-6 molecules bound to a responsive promoter.
Cooperative binding requires a domain capable of associating with other proteins in a homo-or heterotypic fashion. The POZ domain is a 120-amino acid motif generally found at the extreme N terminus of broad-complex, traintrack and bric-abrac/POZ proteins. Family members demonstrate extensive homology throughout the length of this hydrophobic domain, with 30 -50% conservation of amino acid sequence identity (18,19). Numerous overexpression studies have catalogued the ability of POZ domain-containing proteins to form homo-or heterodimers; furthermore, the POZ domain is suspected to mediate the formation of larger oligomeric structures in vivo (4, 18 -24). The recently solved crystal structure of the POZ domain of PLZF (a zinc finger transcription factor implicated in retinoic acid-unresponsive acute promyelocytic leukemia) revealed the PLZF dimer to be considerably more stable than the monomeric form, confirming the multimeric association of the POZ domain (19,25). The hydrophobic dimerization interface of the PLZF POZ domain is extensively distributed throughout the domain and involves ϳ25% of the monomer surface area in a tightly intertwined dimer, as is characteristic for an obligate homodimer (19). Thus, both biochemical and structural studies indicate a role for the POZ domain in mediating multimeric protein-protein interactions.
In this study, we investigate the cooperative and competitive associations of BCL-6 with the murine germline ⑀ promoter. Our observations suggest a mechanism for BCL-6-mediated repression that is dependent upon its ability to cooperatively bind to multiple promoter elements in a manner that requires the presence of an intact POZ domain. We further demonstrate the ability of Stat6 and C/EBP␤ to efficiently compete with BCL-6 for binding to a shared site in the mI⑀ promoter. These findings support a model of germline ⑀ promoter transcriptional regulation in which basal repression mediated by BCL-6 is relieved in an activation-dependent manner through the competitive displacement of BCL-6 by positively acting transcription factors.

EXPERIMENTAL PROCEDURES
Plasmid Construction-The eukaryotic expression vector pMT2T-BCL-6, wild-type germ line ⑀ promoter Ϫ167/ϩ55, and the S2mut (S2m4) mutant-driven reporters have been described (17,26). The S3mut (S3m1) mutant was generated using the wild-type mI⑀ promoter Ϫ167/ϩ55 construct as a template in PCR using the following primers: S3m1-5Ј, ataac gcgtC AGGTG TGTCa gCTAG AAAGAG; and mI⑀-3Ј, tttagatctC CCCTG TGCAG GCT. The products of these reactions were cloned into the MluI and BglII sites of the pGL2-basic vector (Promega). All sequences were verified by automated sequencing at the DNA sequencing facility of Columbia University.
The products of these reactions were cloned into the BglII site of the pBlueBacHis2B baculovirus transfer vector (Invitrogen). All sequences were verified by automated sequencing at the DNA sequencing facility of Columbia University. Infection-competent recombinant baculoviral DNA was generated by cotransfection of transfer vectors and Bac-N-Blue DNA (Invitrogen) in Sf9 cells following the manufacturer's protocol. The baculovirus expression vector for Stat6 was obtained from Dr. Michael Berton. The Jak3 baculovirus was the gift of Dr. John Krolewski.
DNase I Footprinting-Expression and purification of the recombinant proteins used in these assays were as described above. For footprinting experiments, wild-type and mutant germ line ⑀ promoter fragments (Ϫ167 to ϩ55) were end-labeled on the noncoding strand. The probes (5000 cpm) were incubated for 20 min at room temperature with the indicated amounts of purified recombinant proteins using the Amersham Biosciences Biotech SureTrack footprinting system as described previously (17).
Cell Lines, Transient Transfection, and Reporter Gene Assays-M12.4.1 cells were maintained in RPMI 1640 medium supplemented with 10% fetal calf serum, 100 units/ml penicillin, 100 units/ml streptomycin, 2 mM L-glutamine, and 0.1 mM ␤-mercaptoethanol. Cells (5 ϫ 10 6 ) were transfected by electroporation as described (28). The transfection mixture contained 5 g of either germ line ⑀ promoter Ϫ167/ϩ55-luciferase reporter construct or mutant germ line ⑀ promoter-luciferase reporter construct (S2mut or S3mut as described above), 0.5 g of pRL Renilla reporter plasmid (Promega), and the indicated amount of BCL-6 expression construct. Vector DNA (pMT2T) was added as necessary to achieve a constant amount of transfected DNA. Following transfection, cells were incubated in the presence or absence of 10 units/ml murine recombinant IL-4 for 24 h. After 24 h, the luciferase activity of cells transfected with the germ line ⑀ reporter was measured. Transfection efficiency was normalized relative to Renilla activity.

Characterization of a Novel BCL-6-binding Element within the mI⑀ Promoter-Initial
DNase I footprinting studies of the murine germline ⑀ promoter using the zinc finger DNA-binding domain of BCL-6 fused to GST (GST-BCL-6ZF) and purified Stat6 demonstrated the ability of these proteins to protect an overlapping region of the promoter at Ϫ111 to Ϫ102 relative to the transcriptional start site (17). Additional footprinting assays of the mI⑀ promoter suggested the presence of two additional BCL-6-binding sites: one at the transcriptional start site (S1) and another (S3) located ϳ50 bp upstream of the BCL-6/ Stat6 site at Ϫ111 to Ϫ102 (S2) (Fig. 1). Previous linker scanning studies demonstrated the importance of these regions in maintaining a low basal level of transcription of the mI⑀ promoter (29). However, although the significance of the initiation site is understood, no transcription factor has yet been shown to bind within the 5Ј-sequence in either the murine or human germline ⑀ promoter. Careful analysis of this region revealed an element at Ϫ157 to Ϫ149 that resembles the core sequence of the artificial BCL-6-binding site identified by cyclic amplification and selection of targets PCR (B6BS). However, the B6BS-like site found in the mI⑀ promoter lacks a critical thy-midine that is thought to confer STAT binding activity to B6BS ( Fig. 2A).
To assess the ability of BCL-6 and Stat6 to recognize this region of the murine germline ⑀ promoter, an oligonucleotide probe corresponding to mI⑀ Ϫ167/Ϫ135 was generated and used in EMSA. The probe was labeled and incubated with whole cell extracts prepared from either the Mutu I (BCL-6positive) or Mutu III (BCL-6-negative) cell line. These cells had been cultured for 1 h prior to harvest in either the absence or presence of IL-4 (to induce the phosphorylation and DNA binding activity of Stat6). These extracts have previously been shown to have Stat6 binding activity (17). EMSA analysis of extracts prepared from the BCL-6-positive Mutu I cell line and incubated with the mI⑀ Ϫ167/Ϫ135 probe revealed a single complex absent in EMSAs performed with extracts from the BCL-6-negative Mutu III cells (Fig. 2B). This constitutive complex was ablated upon incubation with an antiserum specific to BCL-6, but remained intact following incubation with preimmune serum, confirming that this complex contains BCL-6. The absence of an IL-4-induced complex in extracts prepared from either the Mutu I or Mutu III cell line, as well as the lack of a complex sensitive to incubation with antiserum specific to Stat6, indicates the inability of Stat6 to bind to this region of the mI⑀ promoter.
As we had previously demonstrated a correlation between the strength of BCL-6 interaction with its binding site within a promoter and the efficiency of BCL-6 repression of promoter activity (17), we wished to compare the affinity of BCL-6 for the S1, S2, and S3 binding sites found in the murine germline ⑀ promoter. Unlabeled oligonucleotides derived from these three potential BCL-6 sites were assessed for their ability to compete for BCL-6 binding with a labeled probe generated from the 5Ј-BCL-6 recognition element of the mI⑀ promoter (S3). Extracts prepared from cells of the murine B lymphoma line M12.4.1 were incubated with labeled S3 probe and increasing concentrations of the appropriate unlabeled competitor, and the reaction products were analyzed by EMSA (Fig. 2C). The results demonstrate that S2 and S3 competed for binding to the labeled probe with comparable efficiency, indicating that S3, like S2, is a high affinity binding site for BCL-6. However, the binding element at the transcriptional start site (S1) was unable to compete for BCL-6 binding within the range of concentrations used in this assay, and it may therefore be considered a weak binding site for BCL-6.
These experiments define a novel BCL-6 recognition element within the murine germline ⑀ promoter. Interestingly, this sequence is the first reported BCL-6-binding element that does not also serve as a Stat6-binding site. Furthermore, unlabeled competition studies reveal a hierarchy of affinity for BCL-6 among the three potential binding sites for BCL-6 found within the mI⑀ promoter, whereby the two most 5Ј sites (S2 and S3) strongly bind BCL-6, and the 3Ј-element found at the transcriptional initiation site (S1) binds BCL-6 only weakly.
Repression of the mI⑀ Promoter by BCL-6 Requires Multiple BCL-6-binding Sites-The experiments detailed above describe the presence of multiple binding sites for BCL-6 in the murine germline ⑀ promoter. It remained uncertain, however, whether the presence of these additional binding sites is required for the efficient repression of promoter activity by BCL-6. To explore the functional significance of the BCL-6-binding site at Ϫ156 to Ϫ147, we created a series of mutants within the context of the IL-4-responsive region of murine germline ⑀ promoter Ϫ167/ ϩ55 in which the binding of BCL-6 to S2, S3, or both sites was disrupted (S2mut, S3mut, and S2/S3mut, respectively). Stat6 binding to S2 remained intact in the S2 mutant as described previously (17). These promoter mutants were used to drive a luciferase reporter in transient transfection assays. The constructs were cotransfected with either a control plasmid or a BCL-6 expression vector into M12.4.1 murine B lymphoma cells. Fig. 3B demonstrates the results of experiments using a single concentration (2.5 g) of BCL-6, whereas Fig. 3C shows a dose-response curve with increasing concentrations of BCL-6. Transfected cells were cultured either alone or in the presence of IL-4 for 24 h and then harvested and assayed for luciferase activity. Interestingly, the S3 mutants, and perhaps also the S2 mutants, were hyperactivated upon cytokine induction, perhaps due to the inability of endogenous BCL-6 to bind to the mutant promoter (Fig. 3B). As expected, cotransfection with BCL-6 resulted in repression of the IL-4-induced activation of the wild-type promoter. In contrast, the ability of BCL-6 to repress IL-4-induced transcription of the mI⑀ promoter bearing a mutant S2 BCL-6-binding site was compromised (Fig. 3,  B and C). The ability of BCL-6 to repress IL-4-induced transcription of the mI⑀ promoter mutated at S3 was variably decreased, although never to the levels of the S2 mutant (Fig.  3, B and C). Like the S3 mutant, the S2/S3 double mutant showed markedly increased levels of induction compared with the wild-type promoter; like the S2 mutant, the double mutant was not efficiently repressed upon cotransfection of BCL-6. These results demonstrate the requirement of intact BCL-6binding sites at both mI⑀ Ϫ111/Ϫ102 and Ϫ156/Ϫ147 for the proper function of the repressor.
BCL-6 Binds Cooperatively to Multiple Sites in the mI⑀ Promoter in a POZ Domain-dependent Manner-The identification of two functionally important BCL-6-binding elements within the murine germline ⑀ promoter suggests at least two mechanisms whereby BCL-6 might regulate mI⑀ transcription. BCL-6 might independently bind the various sites found in the germ- line ⑀ promoter; in this manner, a graded response may be effected through sequential occupancy of multiple BCL-6 sites of varying affinities. Alternatively, efficient binding of BCL-6 to the germline ⑀ promoter may require the cooperative interaction of several molecules binding to multiple elements in the promoter. Cooperativity is often employed to generate a threshold response, in which small differences in protein concentration can lead to triggered responses. The sensitivity of the trigger may determined by the affinity of the stronger binding site (30 -32).
To distinguish between these two models of BCL-6-mediated gene regulation, we performed in vitro DNA binding studies in which we assessed the ability of recombinant BCL-6 to bind wild-type and mutant mI⑀ promoters. Increasing concentrations of baculovirus-expressed BCL-6 were incubated with either wild-type murine germline ⑀ promoter Ϫ167/ϩ55 or promoters in which BCL-6 binding to either S2 or S3 was specifically disrupted (S2mut or S3mut, respectively). DNase I footprinting analysis of the resultant complexes revealed striking differences in the ability of BCL-6 to protect S2 and S3 from digestion in the wild-type and mutant promoters (Fig. 4). These studies demonstrated a 20 -30-fold increase in the amount of BCL-6 required to protect the binding site at S2 in the S3 mutant promoter compared with the wild-type mI⑀ promoter (Fig. 4, lanes 1-5 and lanes 19 -23). Conversely, 10 -20-fold more BCL-6 is required to protect S3 from DNase I digestion in the S2 mutant compared with the wild-type promoter (Fig. 4,  lanes 1-5 and 10 -14). It therefore appears that efficient binding of BCL-6 to the murine germline ⑀ promoter is dependent upon its ability to cooperatively interact with S2 and S3.
A number of studies have demonstrated the homo-or heterodimerization of POZ protein family members (4, 18 -24). The crystal structure of the PLZF POZ domain has recently been solved, revealing the POZ homodimer to be significantly more stable than the monomer (19, 25, 33). An interesting corollary to this observation is the tendency of the POZ domain to destabilize binding of proteins within this family to their target DNA sequences. However, one recent study determined that, although the POZ domain of the Drosophila GAGA protein inhibits GAGA binding to single site elements, it mediates the oligomerization and subsequent cooperative DNA binding of GAGA to promoters that contain multiple GAGA-binding sites (21). To determine whether the cooperativity demonstrated for BCL-6 binding to S2 and S3 of the mI⑀ promoter is likewise dependent on the presence of the POZ domain, we generated a recombinant baculovirus-expressed BCL-6 mutant lacking the POZ domain (⌬POZ) for use in DNA binding assays. DNase I footprinting studies were performed in which increasing amounts of the ⌬POZ mutant were bound to the wild-type, S2mut, or S3mut promoter prior to digestion with nuclease. These assays demonstrated little difference in the ability of the ⌬POZ mutant to protect the S2 and S3 sites of the wild-type promoter compared with either of the mutants (Fig. 4, lanes  6 -9, 15-18, and 24 -27). Furthermore, 20 -30-fold more ⌬POZ was required to protect S2 and S3 from DNase I digestion compared with the full-length BCL-6 protein (Fig. 4, lanes 2-5  and 6 -9). Interestingly, a prominent footprint at S1 was detected in assays using the ⌬POZ mutant, but not in experiments with full-length BCL-6; disruption of the S2 and S3 sites had no affect on the pattern of DNase I digestion at S1. This behavior is characteristic of the manner in which a POZ protein family member interacts with a promoter containing a single protein-binding site. Therefore, the results of these studies suggest that BCL-6 regulates transcription of the murine germline ⑀ promoter by cooperatively binding to at least two sites within the promoter: S2 (Ϫ111 to Ϫ102), a site that is shared with Stat6, and S3 (Ϫ157 to Ϫ149), a newly identified site that is not recognized by Stat6. Binding of BCL-6 to the low affinity site at S1, if it occurs at all under physiologic conditions, is non-cooperative and requires very high concentrations of the repressor. Finally, the cooperative binding of BCL-6 to S2 and S3 requires an intact POZ domain and is likely mediated through the POZ-dependent dimerization of BCL-6. FIG. 2. mI⑀ S3 is a high affinity bind-ing site of BCL-6. A, shown is a sequence comparison of the in vitro defined BCL-6-binding site (B6BS) and three potential BCL-6-binding sites (S1-S3) within murine germline ⑀ promoter Ϫ167/ ϩ55. B, two Epstein-Barr virus-transformed human B cell lines that differentially express BCL-6, Mutu I (BCL-6positive), and Mutu III (BCL-6-negative) were cultured in the presence or absence of human recombinant IL-4 (10 units/ml) for 1 h prior to harvest. EMSAs were performed using 5 g of whole cell extract and an oligonucleotide probe corresponding to germline ⑀ promoter Ϫ156/Ϫ147 (S3). The DNA-binding complexes were identified upon supershift with antibody (Ab) to BCL-6 (␣BCL-6). Supershift with antibody to Stat6 (␣Stat6) is shown for comparison. C, the affinity of BCL-6 for mI⑀ S3 at Ϫ156 to Ϫ147 of the germline ⑀ promoter relative to the two other mI⑀ promoter BCL-6-binding sites was assessed in unlabeled competition assays. Whole cell extract (5 g) prepared from the BCL-6-positive M12.4.1 murine B cell lymphoma line was incubated with a labeled probe generated from I⑀ S3 (Ϫ156 to Ϫ147) in the absence (lanes 1, 7, and 13) or presence of increasing concentrations of unlabeled competitor oligonucleotides (3, 10, 30, 60, and 100 ng): mI⑀ S3 (lanes 2-6), mI⑀ S2 (lanes 8 -12), and mI⑀ S1 (lanes 14 -18).
Stat6 and C/EBP␤ Effectively Compete for Binding to mI⑀ S2-Stat6 and C/EBP␤ are cytokine-inducible transcription factors that bind to mI⑀ S2 and may therefore regulate germline ⑀ expression in part through competitive or cooperative association with BCL-6. In an effort to delineate the interactions of BCL-6, Stat6, and C/EBP␤ on the murine germline ⑀ promoter, we performed a series of binding studies in which combinations of recombinant full-length BCL-6, recombinant phosphorylated Stat6, and GST-C/EBP␤ were incubated with a labeled probe corresponding to wild-type mI⑀ Ϫ167/ϩ55 (Fig.  5). DNase I footprinting analysis of the resultant complexes demonstrated the ability of both Stat6 (Fig. 5A) and GST-C/ EBP␤ (Fig. 5B) to displace BCL-6 from its binding site at S2. The minimal amount of Stat6 required to fully protect S2 in this assay was sufficient to displace BCL-6 from the site, as demonstrated by the extension of the footprint at S2 in the pattern that is characteristic of Stat6 (Fig. 5A, compare lanes  1-3 with lanes 4 -6); in fact, even at a concentration 10-fold that required to obtain a footprint in the absence of Stat6, BCL-6 was unable to fully compete with Stat6 for binding to S2 (Fig. 5A, lane 12). Interestingly, these experiments revealed an additional region of the mI⑀ promoter protected by Stat6 at Ϫ91 to Ϫ81. This sequence coincides with a site believed to bind NF-B (29, 34, 35) and appears to correspond to a reverse Stat6 consensus element. As with the major Stat6 site at S2, this new footprint was also competed out by very high levels of BCL-6 and may indicate the formation of a Stat6 tetramer at these two sites. Further study will be required to determine whether these two sites, which are separated by two helical turns, represent the association of two Stat6 dimers to form a tetramer, as has been reported for other members of the STAT family (36 -38).
These studies also reveal a potential role for the IL-6-and lipopolysaccharide-inducible transcription factor C/EBP␤ in the regulation of BCL-6 and Stat6 binding activity. Recombinant C/EBP␤ expressed as a fusion protein with GST was an extremely effective competitor of BCL-6 for binding to S2 (Fig.  5B). In contrast, C/EBP␤ appeared to increase the affinity of Stat6 for S2, as was demonstrated by the strengthening of the Stat6 footprint in the presence of GST-C/EBP␤ (Fig. 5C, compare lanes 4 and 5). Stat6 and C/EBP␤ have similarly been reported to cooperatively interact at the corresponding site in the human germline ⑀ promoter (39). These observations correlate with the results of studies that demonstrate the requirement of an intact C/EBP-binding element for efficient IL-4-and Stat6-mediated germline ⑀ promoter transcription (29). It should be noted, however, that the use of a bacterially expressed GST fusion protein confounds interpretation of these results, as C/EBP␤ binding is known to be affected by posttranslational modifications (40,41) and as the 26-kDa GST

FIG. 3. Repression of germline ⑀ promoter transcription is dependent on intact BCL-6-binding sites at both S2 and S3.
A, shown is a schematic diagram of the mutations used in this study. S2mut binds Stat6 (but not BCL-6) at S2. BCL-6 binding at S3 is intact. S3mut does not bind BCL-6 at S3, but has an intact S2 site. S2/S3mut binds Stat6 at S2, but does not bind BCL-6 at either S2 or S3. These mutations were generated within the context of germline ⑀ promoter Ϫ167/ϩ55. WT, wild-type murine I⑀ promoter. B, 5 ϫ 10 6 M12.4.1 cells were cotransfected with a luciferase reporter driven by the indicated variations in the germline ⑀ promoter (wild-type, S2mut, S3mut, and S2/ S3mut) and either control plasmid or 2.5 g of BCL-6 expression vector. Following electroporation, the cells were divided and cultured either alone or in the presence of IL-4 (10 units/ml) for 24 h. The results are given as means Ϯ S.D. from three separate experiments and are normalized to Renilla activity. A schematic of which binding sites have been mutated is shown below each set of transfections, where B and S represent BCL-6-and Stat6-binding sites, respectively; sites that have been mutated are indicated with an X. C, 5 ϫ 10 6 M12.4.1 cells were cotransfected with a luciferase reporter driven by the indicated variations in the germline ⑀ promoter (wild-type, S2mut, S3mut, and S2/S3mut) and either control plasmid or increasing concentrations of BCL-6 expression vector (1, 2.5, and 5 g). Following electroporation, the cells were divided and cultured either alone or in the presence of IL-4 (10 units/ml) for 24 h. The results are given as the mean total luciferase activity of two separate experiments and are normalized to Renilla activity. moiety may itself interfere with BCL-6 binding to S2.
In conclusion, we have determined that both Stat6 and C/EBP␤ are strikingly effective competitors of BCL-6 for binding to S2 in the murine germline ⑀ promoter. This is in contrast to the apparent cooperative interaction demonstrated by Stat6 and C/EBP␤ at the same locus. Finally, we have evidence that suggests the presence of a novel Stat6-binding site ϳ20 bp downstream of the Stat6 element at S2, which may indicate the formation of a Stat6 tetramer in the mI⑀ promoter.

DISCUSSION
In a previous study, we presented evidence indicating that BCL-6 is involved in the transcriptional regulation of only a subset of Stat6-responsive genes; we further established the murine germline ⑀ promoter as a physiologic target of BCL-6mediated repression (17). In this study, we have described experiments designed to identify characteristics of the mI⑀ promoter that allow its regulation by BCL-6. To this end, we have identified a novel BCL-6-binding site, which we have termed S3, ϳ50 bp upstream of the BCL-6/Stat6-binding site at Ϫ111 to Ϫ102 (S2) of the mI⑀ promoter. DNA binding studies using wild-type and mutant mI⑀ promoters demonstrated the highly cooperative nature of BCL-6 binding to a promoter with multiple BCL-6-binding sites. Similar studies utilizing mutant BCL-6 proteins illustrated the dependence of this cooperativity on the presence of an intact POZ domain. Consistent with a model of synergistic BCL-6 binding to S2 and S3 of the murine germline ⑀ promoter, disruption of BCL-6 binding to either one of these sites compromised the ability of BCL-6 to properly regulate the IL-4-induced activation of the promoter from the remaining BCL-6-binding site in transient transfection assays. As the mutants used in these experiments did retain some BCL-6 binding activity, the levels of BCL-6 achieved in tran-sient transfection appeared to overwhelm the system, and the results of these transfection experiments are not as striking as might be predicted in an uncomplicated cooperative system. However, the biochemical and in vivo data together support a model in which BCL-6 modulates gene transcription by cooperatively binding to promoters that have multiple high to medium-affinity binding elements recognized by BCL-6. In addition to elucidating a mechanism of mI⑀ promoter regulation, we believe that these studies have important implications regarding the general regulation of gene expression by BCL-6.
The results of our earlier studies indicated a correlation between the affinity of BCL-6 for an individual binding site within a promoter and the efficiency of BCL-6 repression of promoter activity, leading to the speculation that BCL-6 target selection might be determined through some equation of protein concentration and binding site affinity (17). Similar gradient affinity models have been proposed for the action of other transcriptional regulators, such as was originally suggested for the Drosophila morphogen bicoid, which is responsible for determining stripe pattern in the anterior half of the fly embryo (42). This model can be extended to promoters containing multiple independent binding sites for a given transcription factor. Exquisitely graded responses might be elicited through the modulation of both protein levels and the binding affinities of the various recognition elements for that protein found within a promoter. However, the discovery of an additional, high affinity BCL-6-binding site in the murine germline ⑀ promoter that appears to cooperate with the originally defined site at Ϫ111 to Ϫ102 adds another layer of complexity to any discussion of BCL-6 target specificity.
Cooperative interactions can result in a 10 -1000-fold increase in the efficiency with which a transcription factor recognizes a low affinity binding site, and are therefore less dependent on the intrinsic affinities of the involved binding sites than would be predicted by the gradient affinity model (43,44). In a cooperative system, responsiveness is instead determined by a combination of factors related to promoter topology: the number of binding sites, the intrinsic affinities of each of those sites, the level of cooperativity among the various sites, the spacing between the sites, and the presence of binding sites for other factors with which it might also cooperatively interact. Responsiveness is also dependent upon characteristics of the protein itself, such as flexibility, potential for oligomerization, and the presence of interactive domains capable of recruiting additional factors that enhance its ability to regulate transcription. These features are therefore likely to dictate BCL-6 target specificity. However, although we have established the basic elements required to demonstrate cooperativity of BCL-6 binding and activity at the mI⑀ promoter (i.e. significantly enhanced binding and repressional synergy with the presence of an additional BCL-6-binding site), much remains to be determined regarding the structural characteristics of a promoter responsive to regulation by BCL-6.
The murine germline ⑀ promoter provides an ideal system with which to dissect the factors that determine BCL-6 target selection. The mI⑀ promoter contains three elements protected by recombinant BCL-6 in DNase I footprinting studies, yet only two of these sites (S2 and S3) appear to demonstrate a cooperative association. These two sites exhibit a relatively high intrinsic affinity for BCL-6 and are separated by 47 bp, whereas the non-interacting site (S1) binds BCL-6 only weakly and is located at the transcriptional start site, 110 and 156 bp downstream of S2 and S3, respectively. At this point, it is unclear whether the failure of S1 to participate in the repressor complex is due to its low affinity for BCL-6 (despite the documented ability of other transcription factors to drive occupancy FIG. 4. BCL-6 binding to the mI⑀ promoter is cooperative and depends on an intact POZ domain. Binding of increasing concentrations of purified recombinant BCL-6 or a truncated protein lacking the POZ domain (⌬POZ) (6,18,54, and 162 M) to either the wild-type murine germline ⑀ promoter (WTI⑀; lanes [1][2][3][4][5][6][7][8][9] or mI⑀ promoter variants in which BCL-6 binding was specifically disrupted at S2 (S2mut; lanes 10 -18) and S3 (S3mut; lanes 19 -27) was compared by DNase I footprinting analysis. Lanes 1, 10, and 19 (controls) were incubated with bovine serum albumin alone. The positions of S1, S2, and S3 are indicated. of lower affinity sites through cooperative interaction with a high affinity site (44)), its distance or phasic configuration relative to the other sites, or some other undefined factor. BCL-6 cooperativity requires an intact POZ domain and is likely mediated through the POZ domain-dependent dimerization of the protein. This domain is well recognized for its ability to mediate homo-and heterotypic interactions between specific POZ protein family members (18). It is therefore possible that BCL-6 may, in some circumstances, cooperatively associate with POZ domain-containing proteins other than BCL-6. Furthermore, protein-protein interactions mediated by the POZ domain are not limited to associations between family members. One of the more interesting developments was the discovery that the POZ domains of BCL-6, PLZF, and BACH2 are able to mediate heterophilic interactions with multiple components of the histone deacetylase repressor complex, including the SMRT/N-CoR corepressors (silencing mediator of retinoid and thyroid hormone receptor/nuclear receptor corepressor), BCoR (BCL-6-interacting corepressor, which appears to specifically interact with BCL-6 to the exclusion of other POZ proteins), mSin3A, and histone deacetylase-1 (45)(46)(47)(48)(49)(50)(51)(52). Although apparently not a general mechanism of POZ domain-mediated repression (53), the ability of some family members to recruit the repressor complex provides a reasonable mechanistic explanation for the dependence of transcriptional repression on the presence of an intact POZ domain.
In vitro DNA binding assays performed with combinations of BCL-6 and transcriptional activators believed to regulate the expression of the mI⑀ promoter demonstrated the ability of Stat6 and GST-C/EBP␤ to compete very effectively with BCL-6 for binding to S2. In contrast, Stat6 and GST-C/EBP␤ appear to cooperatively interact at the same locus; previous studies have also demonstrated the cooperative interaction of Stat6 and NF-B on a fragment derived from the murine germline ⑀ promoter (54). These findings support a model of germline ⑀ transcript regulation whereby BCL-6 bound to the mI⑀ promoter maintains the promoter in an inactive state through the activity of the associated SMRT-histone deacetylase repressor complex. Repression is relieved in an activation-dependent manner both through the competitive displacement of BCL-6 by C/EBP␤ and Stat6, which is further stabilized by cooperative interactions with C/EBP␤ and NF-B, and through the recruitment of CBP/p300, which have been reported to interact with various STAT proteins (including Stat6), C/EBP␤, and NF-B (55)(56)(57)(58)(59). The histone acetyltransferase activity of these coactivators can then reverse the repressive chromatin modifications catalyzed by the BCL-6-associated histone deacetylase, resulting in induction of germline ⑀ promoter transcription.
In contrast to all other known or suspected BCL-6-binding sites, the new element we have identified at S3 diverges from the Stat6 consensus motif TTC(N) 4 GAA. We have demonstrated that Stat6 does not, in fact, recognize this element. The description of a physiologically relevant BCL-6-binding site that is not a target of Stat6 complements the results of studies that have demonstrated the persistence of inflammatory disease in BCL-6 Ϫ/Ϫ Stat6 Ϫ/Ϫ mice (17,60), providing further support of an additional role for BCL-6 in the regulation of non-Stat6-dependent pathways. Recent data using Lymphochips have also identified BCL-6-regulated genes that have not been identified as Stat6 target genes. Analysis of the regions of these genes required for BCL-6 repression will determine whether cooperative BCL-6 binding is a general requirement for BCL-6 regulation.