FOXO proteins regulate tumor necrosis factor-related apoptosis inducing ligand expression. Implications for PTEN mutation in prostate cancer.

Mutations in PTEN occur in 60-80% of prostate cancers and lead to a constitutive activation of the phosphatidylinositol 3-kinase pathway and a resultant loss of activity of the FOXO family of forkhead transcription factors FKHRL1 and FKHR. To provide insight into the role of PTEN mutations in prostate cancer, we used microarrays to identify genes regulated by FKHRL1 and FKHR in LAPC4 prostate carcinoma cells. These studies revealed that adenoviral overexpression of FKHRL1 and FKHR in the LAPC4 prostate cancer cell line resulted in apoptosis and induced the expression of many genes that affect cellular proliferation or survival. The expression of one of these FOXO-regulated genes, TRAIL, a pro-apoptotic member of the tumor necrosis factor family, was decreased in human metastatic prostate tumors. The altered expression of TRAIL in these tumors correlated directly with decreased PTEN expression and the resultant loss of FKHRL1 and FKHR activity. Analysis of the effects of FOXO proteins on the TRAIL promoter localized the FKHRL1 responsive element of the TRAIL promoter to nucleotides -138 to -121 and demonstrated that TRAIL is a direct target of FKHRL1. These findings suggest that the decreased activity of FKHRL1 and FKHR in prostate cancers resulting from loss of PTEN leads to a decrease in TRAIL expression that may contribute to increased survival of the tumor cells.

Mutations in PTEN occur in 60 -80% of prostate cancers and lead to a constitutive activation of the phosphatidylinositol 3-kinase pathway and a resultant loss of activity of the FOXO family of forkhead transcription factors FKHRL1 and FKHR. To provide insight into the role of PTEN mutations in prostate cancer, we used microarrays to identify genes regulated by FKHRL1 and FKHR in LAPC4 prostate carcinoma cells. These studies revealed that adenoviral overexpression of FKHRL1 and FKHR in the LAPC4 prostate cancer cell line resulted in apoptosis and induced the expression of many genes that affect cellular proliferation or survival. The expression of one of these FOXO-regulated genes, TRAIL, a pro-apoptotic member of the tumor necrosis factor family, was decreased in human metastatic prostate tumors. The altered expression of TRAIL in these tumors correlated directly with decreased PTEN expression and the resultant loss of FKHRL1 and FKHR activity. Analysis of the effects of FOXO proteins on the TRAIL promoter localized the FKHRL1 responsive element of the TRAIL promoter to nucleotides ؊138 to ؊121 and demonstrated that TRAIL is a direct target of FKHRL1. These findings suggest that the decreased activity of FKHRL1 and FKHR in prostate cancers resulting from loss of PTEN leads to a decrease in TRAIL expression that may contribute to increased survival of the tumor cells.
The forkhead transcription factors are DNA-binding proteins characterized by the presence of a conserved 110-amino acid winged helix DNA binding domain (1). They play important roles in embryogenesis, tumorigenesis, and maintenance of differentiation status. In Caenorhabditis elegans the forkhead transcription factor DAF-16 is under control of the insulin receptor/PI 3-kinase 1 pathway (2). The human orthologs of DAF-16 include FKHRL1, FKHR, and AFX and belong to the FOXO subfamily of forkhead transcription factors. The PI 3-kinase pathway, via activation of its downstream kinase Akt, phosphorylates each of the FOXO proteins at three different Ser/Thr residues (3). These phosphorylated FOXO proteins interact with 14-3-3 proteins and are subsequently sequestered in the cytoplasm where they are inactive. Inhibition of the PI 3-kinase pathway by PTEN overexpression or pharmacologic means leads to dephosphorylation and nuclear translocation of active FKHRL1, FKHR, and AFX, which in turn leads to cell cycle arrest and apoptosis (4). Conversely, loss of PTEN activity results in increased Akt activity leading to inhibition of FOXO protein activity through their phosphorylation and cytoplasmic sequestration. In addition, the expression of dominant negative FKHRL1 results in the inhibition of apoptosis, demonstrating that FOXO transcriptional activity controls cellular proliferation and apoptosis downstream of PTEN (5).
The PTEN gene was initially identified by its frequent loss in glioblastomas (6), but subsequent studies have shown that PTEN is commonly mutated in prostate cancer (7)(8)(9)(10)(11)(12). In prostate tumors, the loss of PTEN occurs late in the tumorigenic process, suggesting that it is more important for tumor progression than tumor initiation (13,14). Several lines of evidence indicate that PTEN is a tumor suppressor gene. First, PTEN overexpression in tumor cell lines results in cell cycle arrest and apoptosis (4,(15)(16)(17)(18)(19)(20). Second, heterozygous deletion of the PTEN gene in mice leads to the spontaneous development of multiple neoplasias (21). Finally, germline mutation of PTEN in humans results in an increased incidence of endometrial, thyroid, and breast tumors (22,23).
Recent studies indicate that cell cycle arrest downstream of PTEN occurs via the transcriptional induction of p27 kip by AFX and FKHRL1 (24). The resulting increased level of p27 kip inhibits cyclin E-cdk2 and causes G 1 /S-phase arrest. However, the mechanism of apoptosis induced by FOXO proteins is less well understood. In T-cells FKHRL1 has been shown to induce Fas ligand, but this phenomenon has not been reproduced in prostate cancer cell lines (25). In HeLa cells, AFX induces the transcriptional repressor BCL-6, which then suppresses the expression of the anti-apoptotic BCL-X L and leads to apoptosis (26). Together, these results indicate that FOXO proteins are important downstream effectors of PTEN tumor suppressive activity; however, their molecular targets and mechanisms of action in prostate cancer are not understood.
To delineate the molecular events that are perturbed by PTEN mutation in prostate carcinoma cells, we focused on identifying genes that are regulated by FOXO proteins in the prostate. Overexpression of FKHRL1 and FKHR in LAPC4 prostate carcinoma cells using adenoviral vectors induced dramatic apoptosis. Furthermore, expression profiling of LAPC4 cells overexpressing FKHRL1 and FKHR revealed that they regulate many genes encoding proteins involved in survival and proliferation, including TRAIL, a pro-apoptotic factor of the TNF family. Meta-analysis of microarray data of human prostate specimens showed that TRAIL expression is decreased in metastatic prostate tumors but not in normal prostate or localized prostate cancer. We further demonstrated that FOXO proteins directly induce TRAIL expression by an element located between 121 and 138 nucleotides upstream of the transcription start site.
Tissue Specimen-To quantify relative amounts of FKHRL1, FKHR, and AFX mRNA in the prostate, four specimens of normal prostate and six specimens containing prostate cancer were obtained from the Alvin J. Siteman Cancer Center Tissue Procurement Core Facility, Washington University. The prostate tumor samples had greater than 70% tumor tissue by histology. Fifty-m sections of the tissues were prepared using a cryostat and total RNA was isolated using TRIZOL reagent (Invitrogen) according to manufacturer's instructions.
Adenovirus Infection-LAPC4 human prostate carcinoma cells (gift from C. Sawyers, UCLA, Los Angeles, CA) were cultured in Iscove's medium supplemented with 7% fetal bovine serum. HEK293 cells (ATCC, Rockville MD) were cultured in Dulbecco's modified Eagle's medium supplemented with 10% serum. Adenovirus recombinants expressing FKHR-TSSA and FKHRL1-TM were prepared as described below (31). FKHR-TSSA and HA-tagged FKHRL1-TM were gifts from T. Unterman, University of Illinois College of Medicine, Chicago, IL, and M. Greenberg, Harvard Medical School, Boston, MA, respectively (25,32). The cDNA fragments were subcloned into the AdtrackGFP vector and inserted into the adenoviral backbone by recombination with AdEasy vector (31). The AdtrackGFP vector itself was recombined into the AdEasy vector to produce a virus that only expresses GFP for use as a negative control in all experiments. Adenoviral recombinants were purified on a cesium chloride gradient, titrated, aliquoted, and stored at Ϫ80°C until further use. For gene delivery, purified adenovirus was added to LAPC4 cells at a multiplicity of infection (m.o.i.) of 50 -100 for 2 h and washed twice with medium. Thereafter, cells were cultured for the indicated periods of time before harvest. For immunoblot analyses, lysates were harvested from cells infected for the indicated times, resolved by polyacrylamide gel, and subjected to Western blotting. Antibodies used were anti-FKHR (New England Biolabs) or anti-HA (monoclonal 12CA5, Roche Molecular Biochemicals) for FKHRL1 or anti-TRAIL (R & D Systems Inc.).
Cell Survival and DNA Laddering-Cell survival assays were performed with the MTS assay (Promega Corp.) as per the manufacturer's instructions using untreated LAPC4 cells and 2% formaldehyde-fixed cells as controls. DNA ladders were visualized as described (33).
Oligonucleotide Microarray Analysis-Ten g of total RNA was isolated from LAPC4 cells infected with either AdGFP (negative control) or AdFKHRL1-TM or AdFKHR-TSSA for 12 or 15 h. The RNA was subjected to GeneChip analysis using U95A Affymetrix chips as per the manufacturer's protocol (Affymetrix Inc.) at the Siteman Cancer Center GeneChip core facility. Analysis of the arrays was performed using the Dchip software, which pools information across multiple arrays to obtain probe-specific indices that reduce errors because of cross-hybridization and image contamination (34). To analyze expression data, chip intensities were normalized and pooled into sets of duplicate chip experiments prior to comparison analysis. The differences in the profiles were generated by two-way analysis using the pooled duplicate chip values of LAPC4 cells infected with control virus (AdGFP) as baseline (data referred as B) against pooled duplicate chips representing LAPC4 cells infected with either AdFKHR-TSSA or AdFKHRL1-TM or either 12 or 15 h (data referred as E). The filtering criteria were as follows. 1) Each gene must be called "present" (i.e. expression above the level of detection on either E or B). 2) The p value by unpaired t test for testing E B should be less than 0.05. 3) The -fold change in expression be-tween E and B should be greater than 1.2. After applying these filters there were 63 genes in AdFKHRL1-TM and 55 genes in AdFKHR-TSSA whose expression was found to be altered 15 h after infection. To reliably detect differential expression during the qRT-PCR confirmation assays, we selected genes that were differentially expressed Ͼ3fold at the lower bound of the 90% confidence interval calculated by the Dchip software. Twenty-six genes in AdFKHRL1-TM-and 21 genes in AdFKHR-TSSA-infected cells met these filtering criteria (Tables I  and II).
Analysis of Published Gene Expression Profiles-Genes that were previously identified as differentially expressed in prostate cancer (35) were downloaded from www.nature.com/cgitaf/DynaPage.taf?fileϭ/ nature/journal/v412/n6849/abs/412822a0_fs.html&dynoptions ϭ d-oi1011638302 and entered into MSAccess data base software (Microsoft Corp., Redmond, WA). These genes were compared with genes that were differentially regulated in LAPC4 cells overexpressing FKHR and FKHRL1 using MSAccess software. One gene, TRAIL, was found to be common between the two groups. The raw expression values of TRAIL in the different prostate specimens were extracted from data provided at the web address above. These TRAIL expression values in normal adjacent prostate (NAP), localized prostate cancer (PCA), and prostate cancer metastasis (MET) were assigned to their respective groups (32 samples total). The relative differences in TRAIL expression between PCA and NAP or MET and NAP were analyzed with the Mann-Whitney rank-sum test using Sigmastat software (Jandel Scientific, Chicago, IL).
Reporter Assays-LAPC4 cells were transfected using Tfx-20 transfection reagent (Promega Corp.) and HEK293 cells were transfected using LipofectAMINE (Invitrogen Corp.) according to the manufacturer's instructions. Transfections were performed in 12-well plates seeded with 3.5 ϫ 10 4 cells/well using 250 ng of TRAIL promoter luciferase reporter, 100 ng of CMVLacZ reporter (control), 50 ng of plasmid expressing wild type or FKHRL1-TM expression vector (pcDNA3.1), and 600 ng of Bluescript plasmid (Stratagene). The Ϫ1523 to ϩ23 TRAIL promoter luciferase reporter deletion plasmids were described previously (37). Luciferase reporter assays were performed as described (38,39). Mutation of the forkhead binding site in the Ϫ165 to ϩ23 TRAIL promoter luciferase construct was generated using the QuikChange kit (Stratagene Corp.) and complementary primers 5Ј-GAGGAGCTTCA-GACAGAGACCCTCCAGACCAACGCCAACG-3Ј and CGTTGGCGTT-GGTCTGGAGGGTCTCTGTCTGAAGCTCCTC. The sequence of the mutant promoter construct was verified prior to use.

FKHRL1 and FKHR Are the Predominantly Expressed Members of the FOXO Subfamily in the Prostate-
The frequent loss of PTEN in human prostate tumors decreases the activity of the FOXO subfamily of forkhead transcription factors, as they become sequestered in the cytoplasm. To determine the members of the FOXO subfamily that are expressed in human prostate, we performed qRT-PCR on normal human prostate and prostate tumors as well as the prostate carcinoma cell lines LAPC4, LnCAP, DU145, and PC3. Quantitative analysis demonstrated that FKHRL1 and FKHR were expressed at much higher levels than AFX in all the analyzed samples ( Fig. 1). FKHR and FKHRL1 were expressed at similar levels in normal prostate and prostate tumors; however, FKHRL1 expression was increased in the LnCAP, LAPC4, and DU145 cell lines. This expression data indicates that FKHRL1 and FKHR are the predominant FOXO transcription factors in the prostate, thus they are good candidates for mediating transcriptional changes resulting from the loss of PTEN activity.
Overexpression of FKHRL1 and FKHR in LAPC4 Cells Results in Apoptosis-The preferential expression of FKHRL1 and FKHR in the prostate suggested that they would be important in mediating changes caused by the loss of PTEN activity in prostate carcinomas. To examine this role, we developed an in vitro model to examine the effects of increased FKHRL1 and FKHR transcriptional activity in prostate cancer cells. First, we generated replication deficient adenoviruses that express constitutively active mutants of these proteins (AdFKHR-TSSA and AdFKHRL1-TM, respectively). The FKHRL1-TM and FKHR-TSSA mutants have each of the three Akt phosphorylation sites mutated to Ala residues, thus leading to nuclear localization and constitutive transcriptional activity (25,32). Additionally, these recombinant adenovirus coexpress GFP so that the efficiency of infection can be readily assessed. Second, the LAPC4 cell line was selected for adenoviral gene delivery because it is androgen responsive and maintains many characteristics of prostate cancer cells (40). These cells have low levels of functional PTEN and therefore endogenous FKHRL1 and FKHR are cytoplasmically sequestered and inactive (Ref. 40, and data not shown). Additionally, LAPC4 cells were readily infected with adenovirus and showed little evidence of adenoviral toxicity even at high multiplicity of infection (see below and data not shown).
Infection of LAPC4 cells with increasing amounts of Ad-FKHRL1-TM or AdFKHR-TSSA resulted in a dose-dependent increase in protein expression ( Fig. 2A). Infection with a m.o.i. of 50 -100 resulted in Ͼ95% infection of LAPC4 cells after 16 h (as monitored by GFP expression). This multiplicity of infection was therefore used in all subsequent experiments (Fig. 2B). Infection of LAPC4 cells with AdFKHR-TSSA or Ad-FKHRL1-TM resulted in detectable protein expression by 8 h and achieved a maximal level by 12 h (Fig. 2C). Thus, infection with AdFKHR-TSSA and AdFKHRL1-TM results in a rapid, controlled and efficient expression of FKHR-TSSA and FKHRL1-TM in LAPC4 cells. FKHRL1-TM and FKHR-TSSA overexpression in these cells resulted in membrane blebbing and detachment from the plate over time (data not shown). This corresponded with a decrease in cellular viability within 24 h (Fig. 3A). Analysis of DNA morphology showed a classic DNA laddering pattern consistent with the process of apoptosis ( Fig. 3B). In contrast, infection with the control AdGFP virus resulted in little or no decrease in LAPC4 viability and showed no DNA laddering. Thus AdFKHRL1-TM and AdFKHR-TSSA elicit a pro-apoptotic response under our experimental conditions as previously observed (4), demonstrating that adenovirally expressed FOXO mutants (FKHRL1-TM and FKHR-TSSA) are biologically active in LAPC4 cells.
FKHRL1 and FKHR Regulated Genes Influence Cell Survival and Proliferation-To better understand the biologic roles of FKHRL1 and FKHR in the prostate, it is important to identify genes that are regulated by these transcription factors. For this purpose, we monitored the changes in gene expression that occur in response to overexpression of these proteins using microarrays. LAPC4 cells infected with AdFKHRL1-TM, Ad-FKHR-TSSA, or AdGFP control virus for 12 or 15 h was subjected to gene expression profiling. Analysis of these data showed that gene expression was not altered after only 12 h of infection (i.e. profiles of the baseline AdGFP versus AdFKHRL1 or Ad FKHR microarrays were similar). In contrast at 15 h, infection of LAPC4 cells with AdFKHRL1-TM or AdFKHR-TSSA induced changes (Ͼ3-fold) in the expression of 26 and 21 genes, respectively, when compared with AdGFP-infected cells (Tables I and II). The identification of genes that are differentially expressed in LAPC4 cells overexpressing either FKHRL1 or FKHR provides clues to their biological actions. First, a significant number of the induced genes encode proteins with pro-apoptotic or tumor suppressive activity. These included TRAIL, Bcl2 adenovirus E1B interacting protein, DAP kinase 1, and SMAD4. Second, expression of a number of genes encoding proteins with roles in cellular signal transduction was increased, suggesting that FKHR-TSSA and FKHRL1-TM overexpressing cells have alterations in cell signaling. These genes included Jak1, growth factor receptor bound protein 14, and p90 S6 kinase. Third, expression of proteins that modulate global transcription, such as Cbp300/p300, general transcription factor ii, human menopausal gonadotropin box-containing protein 1, and SATB1, was increased. Additionally, the list also included some genes whose functions are as yet unknown and that may be important for the biologic actions of FKHRL1 and FKHR.
Further analysis revealed that the expression of five target genes was increased significantly by both FKHRL1-TM and FKHR-TSSA (Fig. 4A). These included: TRAIL, which is a member of the TNF family of ligands known to have potent pro-apoptotic activity against a wide range of tumors; SMAD4, a transcription factor downstream of the transforming growth factor-␤ signaling pathway that is a tumor suppressor mutated in human tumors; cyclinG 2 , a member of the cyclin family of CDK kinase regulators up-regulated in late S phase; ATPase, aminophospholipid transporter, which is involved in the transport of amphipathic molecules like phosphatidylserine; and BCE-1, a protein of unknown function. Overall, the transcriptional profiles of LAPC4 cells overexpressing FKHRL1 and FKHR suggest that these proteins regulate a genetic program important in cellular proliferation and survival.
TRAIL Expression Is Aberrantly Regulated in Metastatic but Not Localized Prostate Tumors-To determine whether genes regulated by FKHRL1 or FKHR in LAPC4 play a role in prostate cancer in vivo, the expression of the target genes discussed above were analyzed in human prostate specimens. For this purpose, we utilized previously published expression profiles of prostate carcinoma specimens (35). This particular microarray study identified a number of genes that were differentially expressed between NAP, localized (PCA) or metastatic (MET) prostate carcinoma, including PTEN, which regulates indirectly the activity of FKHR and FKHRL1. A comparison of the prostate carcinoma expression profile with our profiles of genes regulated by FOXO proteins in LAPC4 cells revealed that TRAIL was aberrantly expressed in prostate tumors. Statistical analysis of differences in TRAIL expression between NAP, PCA, and MET showed that TRAIL expression was decreased in MET versus NAP (p Յ 0.009), but not statistically different in NAP versus PCA (Fig. 4B). Additionally, TRAIL expression was significantly decreased in MET versus PCA (p Յ 0.023), suggesting that a decrease in TRAIL expression occurs late in tumor progression. Interestingly, the expression of PTEN in these same tumors follows a similar pattern, consistent with the fact that loss of PTEN occurs in late stage prostate tumors (13,35). Thus, the loss of PTEN decreases the activity of the FOXO proteins leading to a decrease in the expression of TRAIL.
TRAIL Is a Direct Target of FKHRL1 and FKHR-TRAIL activates signal transduction using the DR4 and DR5 TRAIL receptors and acts to induce apoptosis in transformed mammalian cells in a paracrine manner (41). This suggests that TRAIL could influence the survival of tumor cells and, as regulators of TRAIL expression, suggests that FOXO proteins could also play such a role. To confirm the FKHRL1-and FKHR-mediated regulation of TRAIL observed in our microarray experiments, we performed additional qRT-PCR experiments using RNA isolated from LAPC4 cells infected for 15 h with either Ad-FKHRL1-TM or Ad-FKHR-TSSA. The qRT-PCR results demonstrated that overexpression of these FOXO proteins resulted in a Ͼ10-fold increase in TRAIL mRNA (Fig. 5A). In addition, Western blot analysis using TRAIL antibodies confirmed that the increase in TRAIL transcription mediated by FOXO overexpression results in increased TRAIL protein levels (Fig. 5B).
To determine whether TRAIL is a direct target of FKHRL1, a luciferase reporter containing Ϫ1523 to ϩ23 nt of the TRAIL promoter was co-transfected with FKHRL1 wild-type or mutant (FKHRL1-TM) expression constructs. FKHRL1-TM induced TRAIL promoter activity 10.5-fold, whereas wild-type FKHRL1 increased expression of the TRAIL reporter only 1.9fold over basal activity, supporting the contention that TRAIL is a direct transcriptional target of FKHRL1 (Fig. 5C). Furthermore, the poor induction of the TRAIL promoter reporter by wild-type FKHRL1 suggests that despite its overexpression, FKHRL1 activity in basal cell culture conditions in the LAPC4 cell line is minimal, consistent with the low PTEN expression in these cells.
To identify sequences in the TRAIL promoter essential for FKHRL1-mediated TRAIL induction, a series of 5Ј deletions of the Ϫ1523 TRAIL promoter luciferase reporter were examined. These deletions range from Ϫ1523 to Ϫ35 nt upstream of the TRAIL gene transcription start site and were previously used by us to localize ␤-interferon responsive sequences in the TRAIL promoter (37). Co-transfection of these TRAIL promoter luciferase reporters with wild-type or FKHRL1-TM in HEK293 cells showed that all but the Ϫ35 TRAIL promoter reporter

TRAIL Is a Gene Target of FKHRL1
construct recapitulated the FKHRL1-TM-mediated induction observed when using the Ϫ1523 TRAIL promoter (Fig. 6A). Through this analysis the sequences important for TRAIL induction by FKHRL1 were localized to a region between 165 and 35 nt upstream of the transcriptional start site. To further delimit the region responsible for FKHRL1-mediated induction of TRAIL transcription, we obtained four overlapping probes that spanned the Ϫ165 to Ϫ35-nt region of the TRAIL promoter. We performed EMSA using these oligonucleotide probes along with nuclear extracts prepared from HEK293 cells infected for 24 h with either AdFKHRL1-TM or AdGFP control. The sequence spanning Ϫ141 to Ϫ98 nucleotides showed a unique gel shift band with FKHRL1 nuclear extracts that was not present in AdGFP-infected cells (Fig. 6B). The specificity of this interaction was confirmed by addition of 100-fold excess unlabeled Ϫ141 to Ϫ98 probe, which completely inhibited the formation of this complex. In contrast, gel shift patterns with the other three oligonucleotide probes showed no differences between AdFKHRL1-TM-or AdGFP-infected cells, indicating that FKHRL1 did not recognize any sequence elements in this region of the promoter. To determine whether FKHRL1 directly binds the Ϫ141 to Ϫ98 sequence, a supershift experiment using anti-FKHRL1 antibodies was performed. The addition of anti-FKHRL1 but not an irrelevant antibody caused a shift in the mobility of the FKHRL1-specific band (Fig. 6C). To further support the idea that FKHRL1 binds the Ϫ141 to Ϫ98 sequence, a 100-fold excess of an oligonucleotide probe containing the FKHR binding element from the glucose-6-phosphatase gene was added to the binding reaction and this also completely abrogated the formation of the FKHRL1-DNA complex (Fig. 6C). Thus, a cis-acting element in the TRAIL promoter located between 141 and 98 nt upstream of the transcription start site serves as a recognition site for FKHRL1.
We next examined whether this FKHRL1 element is important for transcriptional activation of the TRAIL promoter by FKHRL1. For this purpose, the location of the crucial FKHRL1 binding sequence was first narrowed by further EMSA and competition experiments (data not shown). For example, whereas the Ϫ141 to Ϫ98 sequence interacted with FKHRL1, probes containing the flanking sequences Ϫ165 to Ϫ130 or Ϫ115 to Ϫ69 of the TRAIL promoter did not interact with FKHRL1. Thus, we were able to delimit the FKHRL1 binding site in the TRAIL promoter to the region between these two flanking sequences (i.e. Ϫ138 to Ϫ121 nt). Inspection of this region showed that it shares commonality with known FOXO protein DNA binding sites in that it contains multiple stretches of 5Ј-TTT-3Ј (25,32,42). To disrupt FKHRL1 binding and consequent transcription of the TRAIL promoter, the three TTT motifs between Ϫ138 and Ϫ121 were mutated to AGA within the Ϫ165 to ϩ23 TRAIL promoter reporter construct (Fig. 7A). Co-transfection of FKHRL1-TM with this mutated TRAIL promoter luciferase reporter construct completely abrogated FKHRL1-TM-induced luciferase expression (Fig. 7B). This data indicates that a cis-acting element located between Ϫ138 and Ϫ121 nt of the TRAIL promoter is important for FKHRL1-mediated transcription of TRAIL.
sponding to Ϫ141 to Ϫ98 of the TRAIL promoter sequence. Also shown is the effect of preincubation of binding reaction with a 100-fold excess unlabeled  6. FKHRL1 binds to a sequence element in the TRAIL promoter. A, 5Ј deletion constructs of the TRAIL promoter luciferase reporter construct (TRAIL) were transfected into HEK293 cells along with either wild type FKHRL1 (hatched bars) or FKHRL1-TM (solid bars) expression vectors. The -fold change in luciferase activity relative to that obtained using the Ϫ35 to ϩ23 construct is shown. All values are triplicate measurements of mean Ϯ S.E. and the data presented are from one of four experiments with similar results. B, nuclear extracts from HEK293 cells infected with AdGFP (G) or AdFKHRL1-TM (F) were incubated with 32 P-end-labeled double-stranded oligonucleotides corresponding to TRAIL promoter sequences Ϫ165 to Ϫ130, Ϫ141 to Ϫ98, Ϫ115 to Ϫ69, or Ϫ79 to Ϫ35. The binding reactions were electrophoresed through an 8% polyacrylamide gel and autoradiographed. Lane C depicts addition of 100-fold excess unlabeled oligonucleotide in each case. An arrow indicates the specific FKHRL1 gel shift band detected only with the Ϫ141 to Ϫ98 probe. C, nuclear extracts from cells infected with AdGFP (lane 1) or AdFKHRL1-TM (lanes 2-6) were incubated with 32 P-end-labeled double-stranded oligonucleotide corre-FIG. 7. FKHRL1-activated transcription of TRAIL promoter requires an element between ؊138 and ؊121 nt. A, schematic of the TRAIL promoter highlighting the FKHRL1 binding region. B, TRAIL promoter luciferase reporter constructs Ϫ35 to ϩ23 (Ϫ35), wildtype (Ϫ165), and mutated Ϫ165 (Ϫ165m) were co-transfected into HEK293 cells alone (control) or with FKHRL1-TM expressing plasmid. The -fold change in luciferase activity relative to that obtained using the Ϫ35 to ϩ23 construct is shown. All values are triplicate measurements with mean Ϯ S.E. The data presented are from one of two experiments with similar results.

DISCUSSION
PTEN is one of the most commonly mutated genes in human tumors (7,12). In prostate cancer, PTEN mutations are rarely detected in early stage tumors, but are commonly found in late stage carcinomas and metastases, suggesting that PTEN loss aids tumor progression rather than tumor initiation (13). The molecular basis for PTEN action in vivo is associated with its ability to negatively regulate the PI 3-kinase pathway by enzymatic hydrolysis of a phosphate group from the 3-position of phosphatidylinositol phosphate, an intracellular second messenger (20). In vitro experiments have demonstrated that PTEN overexpression can induce G 1 to S-phase arrest associated with apoptosis (19,43,44). This cell-cycle arrest occurs by the induction of p27 kip , a potent cell cycle inhibitor (21). Many studies have shown that important targets of PTEN are the FOXO proteins. FOXO proteins are sequestered in the cytoplasm when the PI 3-kinase pathway is active, however, the inhibition of this pathway by PTEN results in dephosphorylation and consequent nuclear translocation of active FOXO proteins leading to decreased cellular proliferation or apoptosis (4). Loss of PTEN activity leads to decreased FOXO activity leading to inappropriate cell survival thus contributing to tumor progression.
Experiments with the C. elegans forkhead protein DAF-16 have demonstrated its importance in growth factor signal transduction and longevity (2). The FOXO proteins are the DAF-16 mammalian homologs and are downstream effectors of the PI 3-kinase pathway, whose activity is decreased by the loss of PTEN in tumor cells. To help understand the biological roles of the FOXO proteins in prostate cancer, we identified a number of FOXO-regulated genes using microarray analysis. These experiments were performed in LAPC4 prostate carcinoma cells (40) that have low PTEN expression and therefore contain mostly phosphorylated forms of FOXO proteins. Because the endogenous FOXO proteins are sequestered in the cytoplasm by phosphorylation, these cells have minimal basal FKHRL1 or FKHR activity and are sensitive to overexpression of constitutively active mutant forms of these transcription factors (FKHRL1-TM and FKHR-TSSA). Using adenoviral delivery of these transcription factors, we identified a number of FOXO protein target genes in prostate carcinoma cells, including many that function in proliferation and cell survival. Furthermore, FKHRL1 was found to induce the expression of FKHR, further linking the expression and function of these two transcription factors. This finding may have implications for FKHRL1 and FKHR actions in vivo as it has been found that they have largely overlapping tissue expression profiles (45). Finally, the identification of FKHRL1 and FKHR target genes suggest that despite some differences in the transcriptional profiles of cells expressing these two proteins, both of them appear to regulate programs that lead to decreased cell survival and proliferation.
We examined the expression pattern of FOXO protein-regulated genes identified in vitro in human prostate specimens. For this analysis, we exploited the vast amounts of gene expression data now being accumulated on a wide variety of normal and diseased human tissues. For example, there are a number of studies detailing the expression patterns of prostate tumors and one of these was selected for detailed analysis (35). These prostate tissue gene expression profiles were compared with the profiles we obtained in LAPC4 cells overexpressing FKHRL1 and FKHR. By this analysis, we found that TRAIL expression was significantly decreased in metastatic prostate tumors. In this same set of tumors, Dhanasekaran et al. (35) reported that PTEN expression was decreased, a finding that is in accord with previous studies showing that PTEN loss is a late manifestation of prostate tumorigenesis (13,40,46). The correlation of decreased PTEN and TRAIL expression in metastatic prostate tumors, but not in localized prostate carcinomas, strongly supports our in vitro data indicating that TRAIL expression is regulated by FOXO proteins. Thus, tumors that have decreased PTEN activity have decreased expression of TRAIL because of decreased FOXO activity. The re-analysis of publicly available gene expression profiles has enabled us to highlight a potentially biologically important phenomenon in vivo. Such a pooling of resources will enhance the power of future investigations and provide additional insight into understanding the pathogenesis of disease.
TRAIL is a pro-apoptotic member of the TNF family of proteins with the ability to induce apoptosis in a wide variety of tumor cells while being nontoxic to nontransformed cells, a property that is being exploited for therapeutic purposes (47). In prostate cancer cells, it is known that the inhibition of PI 3-kinase by PTEN sensitizes them to TRAIL-mediated apoptosis (48). The decreased activity of the PI 3-kinase pathway also leads to nuclear translocation of FKHRL1 and FKHR and resulting TRAIL transcription. Hence, TRAIL induction by FKHRL1 and FKHR, which lie downstream of PTEN and are effectors of the PI 3-kinase pathway, is a potentially important mechanism for inhibiting cell growth.
Despite the knowledge that TRAIL is an important proapoptotic factor, very little is known about its transcriptional regulation. We have previously shown that TRAIL is induced by ␤-interferons (37) and here we show that both FKHRL1 and FKHR can regulate TRAIL transcription. Furthermore, by promoter analysis, the sequence required for TRAIL induction by FKHRL1 was localized to Ϫ138 to Ϫ121 nt of the TRAIL promoter. Mutation of this sequence completely abrogated TRAIL promoter induction by FKHRL1, indicating that FKHRL1 can directly regulate TRAIL gene transcription. Comparison of the FKHRL1 binding site in the TRAIL promoter with known forkhead sites revealed that these sequences constitute a novel forkhead protein binding site. Furthermore, this site is in close proximity to a ␥-interferon activated sequencelike sequence in the TRAIL promoter, which raises the possibility that FKHRL1 may modulate the ability of interferons to regulate TRAIL transcription under certain circumstances, leading to cross-talk between the Jak-STAT and the PI 3-kinase pathways. This is especially relevant in the light of the fact that interferons also stimulate apoptosis of prostate cancer cells (49). Such cross-talk may be important for prostate homeostasis and could potentially be exploited for therapeutic purposes.