The forkhead transcription factor AFX activates apoptosis by induction of the BCL-6 transcriptional repressor.

The activation of the AKT/protein kinase B kinases by mutation of the PTEN lipid phosphatase results in enhanced survival of a diversity of tumors. This resistance to apoptosis is partly accomplished by the inhibition of genetic programs induced by a subfamily of forkhead transcription factors including AFX. Here we describe an AFX-regulated pathway that appears to account for at least part of this apoptotic regulatory system. Cells induced to synthesize an active form of AFX die by activating the apoptotic death pathway. An analysis of genes regulated by AFX demonstrated that BCL-6, a transcriptional repressor, is up-regulated approximately 4-7-fold. An examination of the BCL-6 promoter demonstrated that AFX bound to specific target sites that could activate transcription. BCL-X(L), an anti-apoptotic protein, contains potential BCL-6 target sites in its promoter. An analysis of endogenous BCL-X(L) levels in AFX-expressing cells revealed enhanced down-regulation of the transcript ( approximately 1.3-1.7-fold) and protein, and BCL-6 directly binds to and suppresses the BCL-X(L) promoter. Finally, macrophages isolated from BCL-6-/- mice show enhanced survival in vitro. These results suggest that AFX regulates apoptosis in part by suppressing the levels of anti-apoptotic BCL-XL through the transcriptional repressor BCL-6.

The resistance of tumors to diverse apoptotic stimuli is of major clinical importance. A large number of recent studies has highlighted the significance of hyperactivation of the PI3K 1 pathway to the insensitivity of late stage tumors to chemotherapy and radiation treatments (1)(2)(3)(4)(5)(6). The induction of this pathway by growth factor or oncogenic stimuli results in the production of three phosphorylated phosphatidylinositol lipids, which act as binding sites for the pleckstrin homology domains of the downstream kinases, PDK-1 and AKT/PKB (7)(8)(9)(10)(11)(12)(13). The membrane colocalization of these kinases results in the phosphorylation of a specific activation site in AKT/PKB by PDK-1 (7,8,(13)(14)(15). Importantly, this pathway is reversed by the lipid phosphatase activity of PTEN, a tumor suppressor that is homozygously deleted in a variety of late stage tumors (2)(3)(4)(5)(6). Whereas a number of phosphorylation targets has been described for the AKT/PKB kinases, both biochemical and genetic evidence support a subset of forkhead transcription factors including FKHR, FKHR-L1, and AFX as being important substrates for phosphorylation mediated by these kinases (16 -20). Phosphorylation of these forkhead transcription factors by AKT/PKB kinases results in their sequestration to the cytoplasm in which they are unable to activate transcription of their nuclear targets (17)(18)(19)(20). This outcome is an important component of cell survival mediated by the PI3K pathway, because nuclear localization of these forkhead family members results in the induction of transcriptional programs that lead to rapid cell death by apoptosis (17)(18)(19)(20).
Whereas it is clear that the inhibition of the apoptotic programs activated by these forkhead transcription factors is a critical component of the pro-survival PI3K pathway, the proapoptotic genes that are transcribed by these proteins remain largely unknown. The induction of FKHR-mediated transcription results in a profound up-regulation of IGFBP-1 transcript (21) as well as p27 kip protein levels (20,22,23), and although it is likely that the induction of p27 kip results in cell cycle arrest, it is not clear how either of these targets could induce the rapid cell death that is observed. Recently, the expression of the pro-apoptotic BH3-only protein, BIM, was found to be induced in lymphocytes by growth factor withdrawal or FKHR-L1. However, although overexpression of this protein induced apoptosis, it was unclear whether BIM alone was sufficient to mediate FKHR-L1-mediated cell death in other cell types (24).
To isolate genes that are potentially involved with AFXmediated apoptosis, we have produced cell lines that can be induced to synthesize a constitutively active form of AFX, and we have used these cell lines to examine AFX-modulated transcriptional changes. Here we describe an AFX-induced transcriptional program that involves the induction of the transcriptional repressor, BCL-6. An examination of potential targets for BCL-6 repression revealed that the gene encoding the anti-apoptotic BH3 protein, BCL-X L , contains BCL-6/STATbinding sites, and we demonstrate that the induction of the AFX-transcriptional program results in the down-regulation of BCL-X L transcript and protein levels. Finally, we demonstrate the novel and unexpected finding that macrophages isolated from BCL-6Ϫ/Ϫ mice die with slower kinetics as compared with wild-type macrophages. These results thus describe a novel pro-apoptotic transcriptional repression program that is activated by nuclear localization of the AFX-type forkhead protein.
* 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.

EXPERIMENTAL PROCEDURES
DNA Constructs-The full-length wild-type AFX gene was PCRamplified from human brain Marathon-Ready TM cDNA (CLONTECH) and subcloned with a FLAG tag at its amino terminus into the pEG-FP.N3 vector (CLONTECH). All three AKT/PKB phosphorylation sites in AFX were mutated to alanines by a two-step PCR procedure. The final TM-AFX-GFP sequence was verified by DNA sequencing. This pEGFP.N3.TM-AFX DNA construct along with pEGFP.N3.TM/H152R-AFX that has an additional mutation in the DNA-binding domain was used to transiently transfect 293E cells as well as BJAB cells for promoter assays. To engineer the HeLa Tet-On TM AFX stable cell lines, the TM-AFX-GFP insert was subsequently subcloned into the pTRE response plasmid (CLONTECH). BCL-6 promoter constructs were made by amplifying different length promoter fragments from human genomic DNA (CLONTECH) using primers shown in Fig. 3A and inserting them upstream of the SV40 promoter in the luciferase pGL3.promoter vector (Promega). The BCL-X promoter luciferase construct was created using the pGL3 vector (Promega). The pCDNA3.1.BCL-6 plasmid was constructed by subcloning the BCL-6 gene isolated from human brain Marathon-Ready TM cDNA (CLON-TECH) into pCDNA3.1/V5-His TOPO vector (Invitrogen). Plasmids expressing wild-type and mutant AFX DNA-binding domain fused to GST (pGEX.4T.TM-AFXBD and pGEX.4T.TM/H152R-AFXBD) were created by inserting the domain sequence (Gly 86 -Ser 237 ) in pGEX.4T-1. Finally, a carboxyl-terminal portion of BCL-6 (Phe 510 -Cys 697 ), which contains the zinc-finger domain, was fused to GST in pGEX.4T-1 to generate the plasmid pGEX.4T.BCL-6.ZnF.
Cell Cultures and Stable Cell Lines-The HeLa Tet-On TM cell line (CLONTECH) was maintained in high glucose Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum, L-glutamine, penicillin-streptomycin, and Geneticin (200 g/ml, Invitrogen). HeLa Tet-On TM -derived AFX stable cell lines maintained in the same medium with an addition of hygromycin B (200 g/ml, CLONTECH) were engineered by transfecting HeLa Tet-On TM cells with pTRE.TM-AFX and pTK-Hyg and selecting for hygromycin B-resistant clones.
Differential Gene Expression Analysis (GeneCalling TM )-A differential display was performed essentially as described previously (25). mRNA isolated from cell lines treated under various conditions was reverse-transcribed using oligo(dT) as primer. 96 sets of variously sized fluorescent-labeled fragments were produced by GeneCalling TM chemistry for each sample using linkers and primers matched to 96 pairs of restriction enzymes used to digest cDNA. Sample groups were triplicates, each of which had GeneCalling TM chemistry performed in triplicate. Fragments were amplified by PCR and resolved using capillary gel electrophoresis. GeneCalling TM software was used to make binary comparisons of traces derived from sample sets upon which GeneCalling TM chemistry had been performed and to compare binary comparisons with one another. Sequence information and precise electrophoretic mobility were used to query a data base of predicted fragments to identify modulated genes. Differentially expressed genes identified by GeneCalling TM software were confirmed as modulated by GeneCalling TM poisoning (a competitive PCR reaction) or by TaqMan TM -quantitative PCR.
Luciferase Assays-Cells were plated at subconfluent density in 24well plates and transiently transfected using Effectene TM transfection reagent (Qiagen). Unless otherwise specified, 160 ng/well pEGFP. N3.TM-AFX, pEGFP.N3.TM/H152R-AFX, or pEGFP.N3 was used in transient transfection of 293E cells. BCL-6 or BCL-X promoter constructs were transfected at 36 ng/well along with 4 ng/well pRLTK vector (Promega). When varying amounts of TM-AFX, TM/H152R-AFX, or pCDNA3.1.BCL-6 plasmid were transfected, empty vector was added to equalize the final DNA contents. For luciferase assays in BJAB cells, 4 g of TM-AFX or TM/H152R-AFX, 0.9 g of BCL-6 promoter constructs, and 0.025 g of pRLTK were electroporated into the cells, and luciferase activities were measured 22-24 h after transfection. As described by the manufacturer's protocol, luciferase activities were measured by the dual-luciferase reporter assay system (Promega), which allowed for normalization of the inducible firefly luciferase activity to the constitutive Renilla luciferase activity.
Gel Shift Assays-Gel shift assays were performed with purified GST fusion proteins essentially as described in the Promega Corporation technical bulletin N0.110. The binding reactions contained 50,000 cpm (100 fmol) of 32 P end-labeled double-stranded oligonucleotide in 50 mM Tris, pH 8.0, 250 mM NaCl, 2.5 mM dithiothreitol, 2.5 mM EDTA, 5 mM MgCl 2 , 20% glycerol, and 0.25 mg/ml poly(dI⅐dC). 100 ng of purified GST fusion protein were used to initiate the 30-min binding reaction (25 l) at room temperature. Reaction products were run on a 4.5% polyacrylamide gel (29:1 bis) in 0.25ϫ TBE (22.25 mM Tris, pH 7.5, 22.25 mM boric acid, 0.5 mM EDTA). Gels were dried and exposed to x-ray films or PhosphorImager plates. The following oligonucleotides were used in the AFX gel shift assays: S1, 5Ј Analysis of Cell Death-The levels of apoptosis were determined by trypan blue exclusion. Both floating and adherent cells were harvested, washed once with phosphate-buffered saline, and incubated at room temperature in 2% trypan blue in phosphate-buffered saline for 5 min. Samples were kept on ice while trypan blue-positive and total cell numbers were counted.
Macrophage Survival Assays-Wild-type and BCL-6-deficient macrophages were isolated and cultured as described previously (26). 10 7 cells were plated in complete Dulbecco's modified Eagle's medium with 10 ng/ml murine recombinant mouse M-CSF (R&D Systems, Minneapolis) and allowed to adhere overnight. After replacing the medium with fresh medium with or without murine recombinant mouse M-CSF the following day, adherent cells were counted at the indicated time intervals.

RESULTS
To identify genes that are regulated by the AFX-type forkhead transcription factor, we derived HeLa cell lines that could be induced to generate a mutated form of AFX, which cannot be phosphorylated by the AKT/PKB kinase. Because phosphorylation of this transcription factor regulates its subcellular localization (17)(18)(19)(20), a non-phosphorylatable form of the protein would be localized to the nuclear compartment in which it would constitutively activate the transcription of target genes. The AFX cDNA was under the control of the TET-On system, and its expression could thus be regulated by the addition of doxycycline (DOX). In addition, AFX contained the green fluorescent protein (GFP) at its carboxyl terminus, and we could thus follow the expression of the protein as well as its subcellular localization in real time. Fig. 1A illustrates that cells induced to express the AFX-GFP protein showed complete nuclear localization of the transcription factor. In addition, cells expressing AFX-GFP showed nuclear condensation and cell surface blebbing that are characteristic of apoptosis. An analysis of cell viability revealed that cells began to die ϳ18 -24 h after induction of AFX, and ϳ70% cells were dead at 48 h after induction (Fig. 1B). The 20% apoptosis exhibited by this cell line in the presence of DOX for 18 h was the basal level ( Fig.  1D). Further analysis demonstrated that cells expressing AFX At the indicated times after the addition of DOX, both floating and adherent cells were harvested and analyzed by Western blotting. TM-AFX-GFP protein detected by an anti-FLAG antibody appears at 3 h after DOX induction and continues to accumulate to significant levels over 24 h. The p27 KIP1 protein, a cell cycle inhibitor that is known to be regulated by the AFX-type forkhead transcription factors, begins to accumulate 6 -9 h after the addition of DOX. After 12 h of DOX induction, poly(ADP-ribose) polymerase degradation is initiated. D, caspase inhibitor zVAD-fmk prevents cells from AFX-induced apoptosis. AFX stable cells grown for 48 h in the presence or absence of DOX in combination with 0, 25, 50, or 100 M zVAD-fmk were assayed for levels of apoptosis by trypan blue exclusion. Percentage of apoptosis represents the percent of trypan blue-positive cells in the total cell population. Error bars indicate the means Ϯ S.E. E, IGFBP-1 transcription is activated by AFX induction (21). Inducible AFX stable cells were grown in the presence or absence of DOX for the indicated times. Total RNA was isolated from adherent cells and analyzed by TaqMan-quantitative PCR. A significant increase in IGFBP-1 RNA level is already detected at 6 h after DOX induction. The RNA levels continue to increase over 18 h of DOX incubation. No induction of IGFBP-1 transcription is observed in the absence of DOX. RNA levels were normalized to GAPDH mRNA levels. Every sample was analyzed in triplicate in each TaqMan experiment.  (25) confirmed that these bands correspond to the BCL-6 gene (data not shown). A representative band, l0c0-124.4, highlighted here by a red vertical line is up-regulated 4.8-fold in cells induced with DOX for 18 h (EI18ϩD) as compared with that in non-induced cells (EI18ϪD). B, BCL-6 transcript levels increase with DOX induction. AFX stable cells were grown in the presence or absence of DOX. At the indicated times after DOX treatment, total RNA was isolated from adherent cells and subjected to TaqMan-quantitative PCR. Relative BCL-6 RNA levels are depicted after normalizing to GAPDH mRNA levels. Similar results were obtained when RNA levels were normalized to ␤-actin mRNA levels (data not shown). After 5 h of DOX induction, BCL-6 transcript level is increased ϳ3-fold over that of a 0-h sample. At 20 h after DOX induction, BCL-6 reaches a maximum induction of ϳ6.7-fold, similar to that observed by the GeneCalling TM procedure. After 15 h of DOX induction, the level of BCL-6 transcript levels off. Every sample was analyzed in triplicate in each TaqMan experiment. Data shown are the averages of three independent experiments. Error bars represent the means Ϯ S.E. C, BCL-6 protein levels increase in the presence of DOX-induced AFX expression. After growing in the presence or absence of DOX for the indicated times, the parental HeLa Tet-On cells and the AFX stable cells (cell line  were lysed and analyzed by Western blotting. Left panel, anti-FLAG antibody reveals that TM-AFX-GFP is not present in HeLa Tet-On cells even in the presence of DOX. In the 15-14 AFX stable cells, there is a small but detectable amount of TM-AFX-GFP expression (arrow) in the absence of DOX, although the levels do not increase over time. In contrast, in the presence of DOX, a high level of AFX expression is achieved beginning at 3 h of DOX treatment. Anti-BCL-6 antibody detects a small amount of BCL-6 protein in HeLa Tet-On TM cells at 24 h regardless of DOX treatment (right panel). 15-14 AFX stable cells, which have a slightly leaky expression of AFX, also show small levels of BCL-6 protein in the were undergoing death by apoptosis (27). An examination of caspase 3 activation (data not shown) as well as poly(ADPribose) polymerase cleavage (Fig. 1C) demonstrated that these two components of the apoptotic death pathway were both induced in response to AFX synthesis. In addition, cell death could be almost completely inhibited by the pan-caspase inhibitor zVAD-fmk (Fig. 1D), consistent with the activation of caspases being responsible for the observed cellular mortality. Finally, two genes that are known to be induced in response to AFX-type forkhead transcription factors, p27 kip1 (Fig. 1C) (20,22,23) and IGFBP-1 (Fig. 1E) (21), were both up-regulated after AFX induction. Together, these results demonstrated that the induction and nuclear localization of AFX alone were sufficient to activate the apoptotic cell death pathway in this cell system.
The data of Fig. 1 suggest that the activation of a transcription program by AFX is sufficient to induce the apoptotic death pathway through caspase activation. To examine genes that are regulated by AFX, we analyzed the transcripts in cells at various times after AFX induction using the GeneCalling procedure (25). Previously, we and others have utilized (25,28) this method for the analysis of differentially expressed transcripts in cell lines. An analysis of transcripts in DOX-induced versus non-induced cell lines revealed a large number of AFXaffected mRNAs, and a complete analysis of these changes will be published elsewhere. A potentially interesting gene that appeared to be regulated by AFX encoded BCL-6, a transcriptional repressor that was previously shown to be involved with germinal center formation and apoptosis (29 -34). As Fig. 2A illustrates, the GeneCalling procedure revealed that a fragment corresponding to the BCL-6 mRNA was clearly induced ϳ18 h after AFX synthesis was initiated. To examine this induction in a more quantitative manner, TaqMan TM -quantitative PCR analysis was performed at various times after AFX induction. Fig. 2B illustrates that BCL-6 transcript was upregulated in response to AFX synthesis with maximal transcript at ϳ20 h after AFX induction. In addition, this time course revealed that BCL-6 transcription was significantly induced by 5 h after the addition of DOX and that the transcript began to level off after the 15-h time point (Fig. 2B). Whereas these data suggested that the transcript was regulated in response to AFX synthesis, Fig. 2C illustrates that the protein was also induced in response to AFX with a lag time of ϳ3 h after the appearance of AFX. Importantly, BCL-6 was not induced in the absence of DOX or in cell lines not engineered to express mutant AFX (Fig. 2C). Finally, Fig. 2D illustrates that the BLC-6 transcript is up-regulated by the inhibition of the AKT/PKB pathway with the PI3K inhibitor LY294002, consistent with a role for this survival kinase in the regulation of AFX-transcriptional activity. Similar up-regulation of BCL-6 transcript levels was observed in the human embryonic kidney cell line 293E treated with LY294002 for 24 h (data not shown). Together, these data suggested that BCL-6 is a transcriptional target gene for AFX.
An examination of the BCL-6 promoter region (35) revealed that it contains a number of potential consensus AFX-binding sites (5Ј-(G/C)(A/C)N(G/a)T(A/c)AA(T/c)A(T/g)(T/g)(G/c)-3Ј) (Fig. 3A, S1-S8) (36). To examine the functional role of these sites, we produced various truncations of the promoter region and analyzed the activation of a downstream luciferase reporter gene in response to cotransfected AFX. As a control, we also cotransfected a form of AFX with a single point mutation ( 151 HR152 153 ) that abolishes DNA binding and transcriptional activation (19). Fig. 3B illustrates that a promoter fragment containing ϳ1296 nucleotides upstream of the TATA box gave a 6-fold increase in luciferase production in response to transfected wild-type AFX but showed no response to the DNAbinding mutant of AFX. Fragments corresponding to ϳ600 and ϳ185 nucleotides upstream of the TATA box also showed a significant although decreased response to AFX, whereas a control plasmid lacking the BCL-6 promoter region showed no enhanced response. These data suggested that multiple sequence elements in the BCL-6 promoter might act as functional AFX-binding sites. To examine these sites for AFX binding, we performed gel shift assays using wild-type and mutant AFX DNA-binding domains. The S6 site bound avidly to the wild type but not to the DNA-binding mutant of AFX (Fig. 3, C and  D). Other sites also appeared to bind to the wild-type DNAbinding domain although with apparently less affinity than the S6 site. Furthermore, the S6-binding site can compete specifically for AFX binding with itself as well as with a known forkhead-binding site (irs) derived from the IGFBP-1 promoter (19), whereas the S2 site that did not interact with AFX failed to compete (Fig. 3D). The importance of the S6 target site was further verified by the observation that a single nucleotide mutation of this site that was shown to result in greatly decreased AFX binding (data not shown) resulted in significantly lower AFX activation of the BCL-6 promoter (Fig. 3E), approximately equal to that seen for the truncation mutant lacking the S6-binding site (Fig. 3B). Finally, the examination of the BCL-6 promoter region also revealed a consensus BCL-6-binding site (5Ј-(T/a/c/g)(T/c)(C/a/t)(C/t/g)(T/a)(A/C/g)GAA (A/T)(G/a/c)-3Ј) (Fig. 3A) (29,35). Because BCL-6 is a transcriptional repressor, we reasoned that this site might act to regulate negatively AFX-induced BCL-6 expression. Fig. 3F illustrates that increasing levels of AFX expression initially up-regulate and then down-regulate BCL-6 promoter activity. In addition, the mutation of the BCL-6-binding site significantly enhances the response of the promoter to transfected AFX, although the promoter is still ultimately down-regulated. These data suggest that there is a negative feedback effect of AFX on the BCL-6 promoter that is in part probably due to endogenously produced BCL-6 regulating its own promoter negatively. Similar results were observed when luciferase assays were performed in BJAB cells, a lymphoid cell line (Fig.  3G).
Because the PI3K lipid kinase pathway controls the AKT/ PKB protein kinases (1), we searched for survival-regulating genes that might also be regulated by this pathway through the expression of BCL-6. The levels of the critical pro-survival BH3 protein BCL-X L have been found to be regulated by the PI3K pathway in a number of studies, and several reports have suggested that BCL-X L levels can be down-modulated by the overexpression of BCL-6 (19,33,37,38). Therefore, we examined whether BCL-X L levels were modulated in response to AFX expression. An analysis of endogenous transcript levels of BCL-X L demonstrated that although the levels of the transcript appeared to fall because of the depletion of the medium by proliferating cells, the levels in cells expressing AFX fell ϳ5-6 h earlier than cells not expressing the transcription absence of DOX. In the presence of DOX, BCL-6 protein expression is significantly induced 6 h after the addition of DOX, reaches a maximum level at 10 h, and begins to decline slightly between 10 and 16 h. D, the inhibition of the PI3K pathway activates BCL-6 transcription. HeLa cells not engineered to express inducible AFX were cultured in the presence of Me 2 SO (DMSO) or 25 M LY294002, a potent PI3K inhibitor, for 24 h. Total RNA was isolated and analyzed by TaqMan-quantitative PCR. RNA levels were normalized to GAPDH mRNA levels. The addition of LY294002 increased the endogenous BCL-6 transcript level ϳ2-fold (p ϭ 0.00059). Both samples were analyzed in triplicate in each TaqMan experiment. Data shown are averages of three independent experiments. Error bars represent the means Ϯ S.E. factor (Fig. 4A). Importantly, the levels of BCL-X L protein were significantly down-regulated with time after AFX induction as compared with ␤-tubulin and the pro-apoptotic proteins BCL-X S and BAX (Fig. 4, B and C). The decline in BCL-X L protein level was not detectable until ϳ24 h after the initiation of DOX treatment. We reasoned that because only a small FIG. 3. BCL-6 promoter contains AFX-binding sites. A, the promoter region of BCL-6 contains eight putative AFX-binding sites designated as S1-S8 (boxed) (36). It also contains one BCL-6-binding site (29). Small forward and reverse arrows indicate the locations of the PCR primers used to isolate the promoter fragments from human genomic DNA. 1296.pcr-5, 600.pcr-5, and 185.pcr-5 are forward primers; pcr-3 is the reverse primer. The large arrow with ϩ1 designates the transcriptional start site. B, AFX activates BCL-6 promoter activity. The indicated portions of BCL-6 promoter (1296, 600, and 185 nucleotides) were isolated from human genomic DNA by PCR and subsequently subcloned upstream of a luciferase reporter gene in the pGL3 promoter vector (pGL3). These promoter constructs were cotransfected with plasmids expressing TM-AFX or TM/H152R-AFX into 293E cells and assayed for luciferase activities 16 -18 h after transfection. TM/H152R-AFX in addition to the triple alanine mutations also contains a H152R mutation that abolishes the DNA binding ability of AFX and serves as a negative control for luciferase assays (19). The dual-luciferase reporter assay system was used to measure the luciferase activities. Relative luciferase activities are shown after normalizing the inducible firefly luciferase activity to the constitutive Renilla luciferase activity. In each luciferase experiment, every sample was performed in quadruplicates. Data shown are the averages of two independent experiments. Error bars denote the means Ϯ S.E. p ϭ 8.13E-05 for 1296 versus 600; p ϭ 0.030 for 600 versus 185; p ϭ 2.32E-06 for 1296 versus 185. Both 600 and 185 have p Ͻ 0.001 for TM-AFX versus TM/H152R-AFX. C, the DNA-binding domain of AFX binds to specific sites in the BCL-6 promoter. Oligonucleotides corresponding to the S1-S8 sites in the BCL-6 promoter were end-labeled with 32 P and assayed for binding to a GST fusion protein containing the AFX DNA-binding domain. S(B), bound and, hence, shifted oligonucleotides; S(F), free oligonucleotides. D, the specificity of AFX binding to the S6 site in the BCL-6 promoter. The S6 and IRS oligonucleotides were 32 P end-labeled and assayed for binding to wild-type (W) and H152R (M) forms of the DNA-binding domain of AFX. The binding reactions were also competed with 100-fold excess of cold S6 oligonucleotide as well as the non-binding S2 site. IRS (B) and IRS (F) symbolize bound and free irs oligonucleotides, respectively. E, the AFX-binding site S6 in BCL-6 promoter is essential for full AFX-mediated BCL-6 transcription activation. The 1296 pGL3 promoter construct containing wild-type S6 (S6 wt) or an S6-binding site mutated to prevent AFX binding (S6c, data not shown) were cotransfected with TM-AFX or TM/H152R-AFX expression plasmid into 293E cells. 16 -18 h after transfection, luciferase activities were measured by the dual-luciferase reporter assay system. Relative luciferase activities are shown after normalizing the inducible firefly luciferase activity to the constitutive Renilla luciferase activity. The data represent the averages of three independent experiments. In each experiment, every sample was tested in quadruplicates. Error bars indicate the means Ϯ S.E. TM-AFX activation gave p ϭ 0.045 for S6wt versus S6c (p ϭ 0.023 for S6c versus pGL3; p ϭ 0.0035 for S6wt versus pGL3). To generate the pGL3.1296.S6c plasmid, nucleotide A in the S6 site was mutated to G by a two-step PCR procedure (underlined). F, the BCL-6-binding site in BCL-6 promoter negatively regulates its own transcription. The 1296 pGL3 promoter construct containing wild-type (pGL3.1296) or mutant BCL-6-binding site (pGL3.1296.mutBcl6) was cotransfected with the indicated amounts of TM-AFX or TM/H152R-AFX expression plasmid DNA(ng) (DNA (ng)) into 293E cells. Relative luciferase activities were measured 16 -18 h after transfection using the dual-luciferase reporter assay system. In the mutant BCL-6-binding site, GAA was mutated to CCC by a two-step PCR procedure. p Ͻ 0.001 for 1296 versus pGL3.1296.mutBcl6 promoter constructs at all DNA amounts tested with the exception of TM-AFX at 160 ng, p ϭ 0.010. G, pGL3.1296 or pGL3.1296.mutBcl6 promoter reporter construct was electroporated along with TM-AFX or TM/H152R-AFX-expressing plasmid into BJAB cells. 22-24 h after transfection, relative luciferase activities were quantitated by the dual-luciferase reporter assay system. p Ͻ Ͻ 0.00001 for TM-AFX versus TM/H152R-AFX. percentage of cells had undergone apoptosis at this point (ϳ10% above the basal level), it was difficult to detect any slight change in BCL-X L protein levels by Western blotting of the whole population. Accordingly, we began to observe a more significant reduction in BCL-X L protein levels after 28 h and an almost complete loss of the protein at 48 h at a time when the majority of cells (ϳ70%) had undergone apoptosis. Furthermore, the appearance of cleaved caspase 3, indicative of caspase activation, correlated with the down-regulation of BCL-X L protein levels (Fig. 4C). Fig. 4C also illustrates that the fall in BCL-X L protein levels is not merely because of the activation of caspases, since inclusion of the pan-caspase inhibitor zVAD-fmk modified the rate of BCL-X L down-regulation only slightly. Finally, cotransfection of the triple mutant form of AFX together with a BCL-X L promoter linked to a luciferase reporter revealed down-regulation of this downstream target by AFX (Fig. 4D). Together, these results were consistent with a loss of BCL-X L transcript and protein accompanying the induction of cell death in the AFX-expressing cells.
Previous studies demonstrated that BCL-X L expression is controlled by STAT transcription factor activity (39 -41), and the examination of the BCL-X L promoter region revealed the presence of several potential STAT-binding "GAS" sites that are also possible binding sites for BCL-6 (29,39). To test whether BCL-6 can directly down-regulate the BCL-X L promoter, a cotransfection experiment was performed. A fragment encoding the ϳ1000-nucleotide promoter region of BCL-X L was placed 5Ј to the luciferase reporter, and the expression from this promoter was analyzed in the presence of increasing levels of BCL-6 protein. The expression from the BCL-X L promoter was normalized to that observed for another constitutive promoter to ensure that the observed effects were promoter-specific. Fig. 5A illustrates that the BCL-X L promoter was significantly down-regulated in the presence of increasing levels of  (Fig. 5B). When GAA was mutated to CCC in these consensus BCL-6-binding sites, the binding of BCL-6 to the BCL-X promoter fragments was abrogated, indicating that the binding was specific (data not shown). An examination of the ability of BCL-6 to induce cell death revealed that although this protein alone was not as effective as AFX at mediating apoptosis, there was still a significant increase of death in cells expressing this protein, which is in agreement with previous studies in other cell lines (33,34)   . The down-regulation in HeLa Tet-On TM ϩDOX or ϪDOX and cell line 15-14 ϪDOX is probably caused by the depletion of growth factors in the medium by the rapidly proliferating cells. RNA levels were normalized to GAPDH mRNA levels. B, BCL-X L protein levels decrease in the presence of DOX-induced AFX expression. At the indicated times after growing in the presence or absence of DOX, AFX stable cells were lysed and analyzed by Western blotting with various antibodies. In the presence of DOX, BCL-X L protein decreases over time, whereas the levels of pro-apoptotic BAX and BCL-X S do not appear to decrease. Equal loading was verified by probing with an anti-␤-tubulin antibody. C, the down-regulation of BCL-X L protein levels by AFX induction is independent of caspase activities. After growing in the presence of Me 2 SO (ϪzVAD-fmk) or 50 M zVAD-fmk (ϩzVAD-fmk) in addition to DOX for the indicated time intervals, the 15-14 AFX stable cells were lysed and analyzed by Western blotting. DOX induction leads to a reduction in BCL-X L protein levels over time even when caspase activation is inhibited by zVAD-fmk (bottom) albeit with slower kinetics than that in the presence of active caspases (top). D, TM-AFX represses BCL-X promoter activity. A plasmid containing ϳ1000 nucleotides of the human BCL-X promoter upstream of a luciferase reporter gene was cotransfected into 293E cells with a GFP or TM-AFX-expressing plasmid. 19 h after transfection, luciferase activities were measured by the dual-luciferase reporter assay system. TM-AFX expression results in an ϳ2.6-fold reduction in relative luciferase activity compared with GFP (p ϭ 0.019). Data represent the averages of three independent experiments of quadruplicates. Error bars represent the means Ϯ S.E. (Fig. 5C). Whereas there are probably to be other mechanisms such as targeted protein degradation involved with the decrease in BCL-X L levels, these data suggest that one of the important anti-apoptotic genes that is regulated by AFX through BCL-6 is BCL-X L . In addition, these data are consistent with in vivo data demonstrating that there are low or undetectable levels of BCL-6 in pre-B cells but high level expression of BCL-XL in this population (31).
Whereas all of these results were consistent with a role for AFX-induced BCL-6 in apoptosis, the use of an in vivo-derived non-transfected cell system would solidify the case dramatically. Therefore, we examined the in vitro survival characteristics of macrophages isolated from BCL-6Ϫ/Ϫ knockout mice (30). Fig. 6A demonstrates that macrophages from the mutant animals showed enhanced survival in both the presence as well as the absence of the exogenous survival factor macrophage CSF (M-CSF). A tabulation of the viable cell numbers (Fig. 6B) reveals that the rate of BCL-6Ϫ/Ϫ macrophage cell death is significantly retarded in the absence of the exogenous survival factor. This effect is even more dramatic in the presence of the survival factor in which at 24 h all of the BCL-6-deficient macrophages are still alive while almost half of the wild-type macrophages have died. These data thus support the contention that BCL-6 is an important mediator of cell death in vivo, although they also suggest that other modulators of apoptosis are also utilized in these cells under these in vitro cell culture conditions.

DISCUSSION
The resistance of a diversity of tumors to apoptosis induced by chemotherapy and radiation correlates with the up-regulation of the AKT/PKB kinases through increased PI3K activity (42,43). It is probable that a significant fraction of this resistance is because of the inhibition of transcriptional programs induced by forkhead transcription factors such as AFX (17)(18)(19)(20). The results described here provide for a pro-apoptotic transcriptional mechanism that may be inhibited in tumors endowed with enhanced AKT/PKB kinase activity. Whereas previous data have suggested that the AKT/PKB pathway may regulate the BCL-X L :BAD/BAX pathway by direct phosphorylation of BAD (44) for example, the new results reported here link these two pathways through transcriptional modulation (Fig. 7). Our data thus support a transcriptional repression mechanism for the previously observed modulation of BCL-X L FIG. 5. BCL-6 mediates down-regulation of BCL-X L expression. A, BCL-6 represses BCL-X L promoter activity. The BCL-X promoter luciferase reporter construct was cotransfected into 293E cells with various amounts of pCDNA3.1.BCL-6 and assayed for luciferase activities 19 h after transfection. The BCL-X promoter activity diminishes with increasing amounts of BCL-6-expressing DNA. Relative luciferase activities are graphed after normalizing firefly to Renilla luciferase activities. The data represent the averages of three independent experiments of quadruplicates. Error bars represent the means Ϯ S.E. p ϭ 9.12E-05 for 0 versus 10 ng; p ϭ 0.024 for 10 versus 20 ng; p ϭ 0.022 for 20 versus 40 ng; p ϭ 0.019 for 40 versus 80 ng; p ϭ 0.0022 for 80 versus 160 ng; p ϭ 1.85E-05 for 0 versus 160 ng. The figure also illustrates increasing expression of BCL-6 protein (arrow) in each transfection. B, BCL-6 binds directly to multiple sites in the BCL-X L promoter. The carboxyl-terminal portion of BCL-6 containing the DNAbinding zinc finger domain was fused to GST and assayed for binding to three fragments of the BCL-X L promoter, X-S1, X-S2, and X-S3. All of these fragments contain putative BCL-6-binding sites and are capable of binding specifically to the GST.BCL-6.ZnF fusion protein as the corresponding cold oligonucleotides in 100-fold excess successfully compete for BCL-6.ZnF binding. X(B), bound oligonucleotides; X(F), free oligonucleotide. C, BCL-6 overexpression triggers apoptosis in HeLa cells. Plasmids expressing LacZ, BCL-6, or TM-AFX were transfected into HeLa Tet-On TM  levels by both the PI3K pathway as well as the BCL-6 transcription factor (19,33,37,38). Because BCL-X L is a major component of cell viability (27), the down-regulation of its ex-pression by BCL-6 would be expected to have negative effects on cell survival. These negative effects on cellular viability would be particularly important under circumstances of limiting growth factors in which the PI3K pathway would be downregulated (1). These results thus provide a mechanism whereby transcriptional activation by AFX coupled with transcriptional repression by BCL-6 can together induce a cell death pathway, and they are consistent with recent data that demonstrate that an elevated level of BCL-6 expression is a positive prognostic indicator for patients with B-cell lymphoma (45). Whereas it seems clear from our results that the modulation of BCL-6 and BCL-X L levels by AFX is a component of the apoptotic pathway induced by the inhibition of the PI3K pathway, it is probable that this inductive event is substantially more complex. For example, previous data have implicated the induction of the expression of BH3-only proteins such as BIM by the FKHR-L1 forkhead transcription factor (23), although we have been unable to demonstrate an induction of the BIM transcript by the AFX protein in our system (data not shown). In addition, it is improbable that BCL-6 alone is capable of inducing the observed apoptotic response. BCL-6 is expressed in germinal   FIG. 6. BCL-6 plays an anti-survival role in macrophages. Deletion of BCL-6 allows primary mouse macrophages to survive longer in culture. In both the presence and the absence of murine rM-CSF, wild-type macrophages with functional BCL-6 exhibit a more accelerated rate of reduction in the number of live adherent cells than macrophages containing no functional BCL-6. A, phase micrographs of wild-type (BCL-6ϩ/ϩ) and BCL-6-deficient (BCL-6Ϫ/Ϫ) macrophages cultured in the presence or absence of 10 ng/ml M-CSF for the indicated time intervals. B, percentages of cell survival were calculated by dividing the number of live adherent cells at the various time points by the number of live adherent cells at 0 h. p Ͻ Ͻ 0.01 for BCL-6ϩ/ϩ versus BCL-6Ϫ/Ϫ at 24, 48, and 72 h. center B cells, and BCL-6 null animals are deficient in germinal center formation and suffer from myocarditis and pulmonary vasculitis (30). Interestingly, dysregulation of BCL-6 is correlated with the progression of various lymphomas, and BCL-6 can under some circumstances be anti-apoptotic (32,46). Therefore, it is likely that the induction of BCL-6 by AFX is only part of the transcriptional program that is involved with the initiation of the apoptotic response. Thus, there appears to be a diversity of other factors that are involved with the AFXmediated apoptotic cascade. For example, the up-regulation of IGFBP-1, a protein that modulates the activity of the insulinlike growth factor growth and survival factors, is consistent with the hypothesis that inhibition of survival signals upstream of the PI3K pathway is likely to be a component of AFX-induced cell death (19). In addition, the induction of p27 kip1 , an inhibitor of G 1 to S progression, suggests that proteins that modulate the cell cycle are also likely to be involved with forkhead-mediated apoptosis (20,23). Finally, the transcriptional activation of the forkhead homologue DAF-16 in Caenorhabditis elegans during dauer formation induces a quiescent but not mortal state in these worms, suggesting that forkhead may induce different transcriptional programs under different conditions (16,47). A further analysis of genes regulated in the system described here will undoubtedly increase our understanding of the mechanisms that are induced by this important transcriptional pathway.