Hyaluronidase Induction of a WW Domain-containing Oxidoreductase That Enhances Tumor Necrosis Factor Cytotoxicity*

To determine how hyaluronidase increases certain cancer cell sensitivity to tumor necrosis factor (TNF) cytotoxicity, we report here the isolation and character-ization of a hyaluronidase-induced murine WW domain-containing oxidoreductase (WOX1). WOX1 is composed of two N-terminal WW domains, a nuclear localization sequence, and a C-terminal alcohol dehydrogenase (ADH) domain. WOX1 is mainly located in the mitochondria, and the mitochondrial targeting sequence was mapped within the ADH domain. Induction of mitochondrial permeability transition by TNF, staurosporine, and atractyloside resulted in WOX1 release from mitochondria and subsequent nuclear translocation. TNF-mediated WOX1 nuclear translocation occurred shortly after that of nuclear factor- k B nuclear translocation, whereas both were independent events. WOX1 enhanced TNF cytotoxicity in L929 cells via its WW and ADH domains as determined using stable cell transfectants. In parallel with this observation, WOX1 also enhanced TRADD (TNF receptor-associated death domain protein)-mediated cell death in transient expression experiments. Antisense expression of WOX1 raised TNF resistance in L929 cells. Enhancement of TNF cytotoxicity by WOX1 is due, in part, to its significant down-regulation of the apoptosis inhibitors Bcl-2 and Bcl-x L ( > 85%), in vitro translation. C stimulation of L929 cells with hyaluronidase (100–200 units/ml) for 24 h resulted in the increased expression of 46- and 30-kDa WOX1 proteins, as determined by Western blotting using antibodies against a synthetic peptide of WOX1 at the N terminus. At high concentrations ( . 400 units/ml), hyaluronidase suppressed WOX1 expression. The 30-kDa protein is probably a degradation product of 46-kDa WOX1. Red CMXRos. Protein expression in cells was examined by fluorescent microscopy 24 h post-transfection. Also, by successive deletion and expression analyses, the mitochondrial targeting region was mapped to amino acids 209–273 of the ADH domain (GFP-WOX1adh, construct 7). E , however, the truncated GFP-WOX1ww protein (construct 4), which possesses the N-terminal WW domains and the NLS, is present in the nucleus. F , stimulation of an established L929 cell line, which constitutively expresses full-length GFP-WOX1 (construct 2), with TNF (20 ng/ml) resulted in the appearance of GFP-WOX1 in the isolated nuclei at the 40-min time point (detected by anti-GFP antibodies). This correlates with the time course of cytochrome c release from mitochondria to the cytosol.

Most cancer cells are known to secrete the matrix-degrading enzyme hyaluronidase. Elevation of hyaluronidase levels is associated with progression, invasion, and metastasis of breast, ovarian, endometrial, prostate, and other cancers (1)(2)(3)(4)(5). Also, expression of hyaluronidase by tumor cells induces angiogenesis in vivo (6). The growth of murine lung carcinoma and melanoma, for example, is influenced by Hyal-1, a locus determining hyaluronidase levels and polymorphism (7). How hyaluronidase modulates cell growth is not known.
Both in vitro and in vivo studies have shown that exogenous hyaluronidase reverses the resistance to chemotherapeutic drugs in cancer cells and solid tumors by increasing their exposure to the drugs (8 -10). We have determined that hyaluronidase enhances cancer cell susceptibility to tumor necrosis factor (TNF) 1 -mediated cell death (11)(12)(13). For example, pretreatment of murine L929 fibroblasts and human prostate LN-CaP cells with hyaluronidase for at least 12 h significantly increases their sensitivity to TNF-mediated death (100 -700%) (11)(12)(13). The hyaluronidase-enhanced TNF sensitivity in L929 cells is associated, in part, with up-regulation of pro-apoptotic p53 (11)(12)(13).
To further explore the mechanism whereby hyaluronidase enhances TNF cytotoxicity, we report here the isolation of a novel murine WW domain-containing oxidoreductase (Wox1) cDNA. Hyaluronidase increased Wox1 gene and protein expression. Ectopic expression of WOX1 in L929 cells enhanced their sensitivity to TNF cytotoxicity, whereas antisense Wox1 raised TNF resistance. Thus, WOX1 is involved in the hyaluronidaseincreased TNF sensitivity in various cancer cells. We produced polyclonal antibodies against WOX1, examined WOX1 cellular localization, and determined the possibility of WOX1 nuclear translocation in response to apoptotic stimuli and inducers of mitochondrial permeability transition. Overexpression of WOX1 was shown to induce cell death. We then examined whether the cell death was p53-or caspase-dependent. Since hyaluronidase up-regulates p53 expression, the role of p53 and WOX1 in mediating cell death was determined.

EXPERIMENTAL PROCEDURES
Molecular Cloning of Wox1-Isolation of novel cDNAs by differential display, library screening, and functional analysis has been described previously (14,15). L929 cells were treated with bovine testicular hyaluronidase (200 units/ml; Sigma) for 4 h, followed by isolating total cellular RNA and performing first strand cDNA synthesis and differ-* This work was supported by the Guthrie Foundation for Education and Research, the Wendy Will Case Cancer Fund, the American Heart Association, and National Cancer Institute Grants R01CA61879 and R55CA64423 from the National Institutes of Health (to N.-S. C.). 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.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM /EBI Data Bank with accession number(s) AF187014 (murine Wox1) and AI669330 and AF187015 (human WOX3).
‡ To whom correspondence should be addressed: Lab. of Molecular Immunology, Guthrie Research Inst., 1 Guthrie Square, Sayre, PA 18840. Tel.: 570-882-4620; Fax: 570-882-4643; E-mail: nschang@inet.guthrie.org. ential display (14,15). A hyaluronidase-induced cDNA of 300 base pairs was isolated and used to screen a -phage cDNA library from murine NIH/3T3 fibroblasts (CLONTECH, Palo Alto, CA). The isolated fulllength cDNA insert was amplified by polymerase chain reaction (using -phage primers) (see Table I) and subcloned into the TA cloning site of eukaryotic expression vector pCR3.1 (Invitrogen, San Diego, CA). Protein domain analysis was performed using the SMART Simple Modular Architecture Research Tool (16,17). As compared with the existing domains in the universal data bases, each resulting positive domain is determined according to a calculated E-value (16,17).
In Vitro Translation-The full-length murine Wox1 cDNA was constructed in the pCR3.1 vector. An in vitro translation kit (Novagen, Madison, WI) was used to transcribe the full-length Wox1-pCR3.1 construct into mRNA via the T7 promoter and to translate into [ 35 S]methionine-labeled WOX1 protein.
Expression Constructs-Constructs, which were made with the pEGFP-C1 vector (CLONTECH), for expressing N-terminal green fluorescent protein (GFP)-tagged proteins are shown in Table I. These constructs were used to express the full-length coding region (construct 2), the antisense Wox1 mRNA (expressing antisense Wox1 mRNA and GFP protein; construct 3), the N-terminal WW domain region (construct 4), a partial ADH domain (amino acids 180 -392; construct 5), and the potential mitochondrial targeting regions in the ADH domain (amino acids 180 -273 and 209 -273; constructs 6 and 7, respectively). To examine whether the large-size GFP protein (28 kDa) affects the WOX1 function, similar constructs, which were made with a small C-terminal v5 tag (5 kDa) in the pcDNA3.1.TOPO vector (Invitrogen), were used to express full-length WOX1 (construct 8), the first WW domain (construct 9), and the first and second WW domains (construct 10).
p53 Expression Construct-A full-length p53 cDNA clone from normal human tissues was found in the expressed sequence tag data base (GenBank TM /EBI accession numbers AI243172 and AF307851) and obtained from Incyte Genomics (St. Louis, MO). The coding region was cloned into the pcDNA3.1/CTGFP-TOPO vector (Invitrogen) and tagged with a GFP sequence at the C terminus (construct 11) (see Table I).
Site-directed Mutagenesis-Site-directed mutagenesis was performed to alter several indicated sites in the Wox1 cDNA sequence using the QuikChange site-directed mutagenesis Kit (Stratagene, La Jolla, CA). The aspartic acid residues of a putative caspase recognition site (DIND, amino acids 267-270) were mutated to glycine, i.e. D267G and D270G (construct 17) (see Table I). The GKRKRV sequence (amino acids 50 -55) of the nuclear localization sequence (NLS) was also mutated to GQGTGV (construct 18) (see Table I).
Antibody Production-A WOX1 peptide (RLAFTVDDNPTKPT-TRQRY, amino acids 89 -107) was synthesized by Genemed Biotechnologies, Inc. (San Francisco, CA) and conjugated with keyhole limpet hemocyanin for antibody production in rabbits using the Pierce antibody production kit. The selected WOX1 sequence is identical between human and mouse.
Transient Transfection-In most cases, transfection studies were performed by our standard CaPO 4 precipitation method (14,15). Fortyeight h post-transfection, the extent of cell death was measured by crystal violet staining (a measure of both necrosis and apoptosis). Additionally, DNA fragmentation assays (11) were performed to determine the extent of apoptosis. To exclude the possibility of nonspecific killing of cells caused by transfection reagents or vectors alone, the observed results were repeated using liposome-based cell transfection reagents such as LipofectAMINE (Amersham Pharmacia Biotech), GeneFECTOR (Venn Nova, Pompano Beach, FL), and FuGENE 6 (Roche Molecular Biochemicals) and electroporation (BTX ECM830, Genetronics, San Diego, CA).
Confocal Microscopy-Where indicated, confocal microscopy analysis was performed to determine the colocalization of WOX1 and mitochondria. Mitochondria were stained by antibodies against cytochrome c or by the membrane potential-sensitive mitochondrial stain Mitotracker Red CMXRos (Molecular Probes, Inc., Eugene, OR).
Northern and Western Blotting-To perform Northern hybridization, L929 cells were cultured in 100-mm Petri dishes and treated with hyaluronidase (200 units/ml) for 1-24 h. Total cellular RNAs were isolated from these cells, and Northern hybridization was carried out using 40 g of RNA/lane (14,15). Antibodies used in the Western blotting were against IB␣, Bcl-2, Bcl-x L , p53, a phospho-SAPK/JNK peptide, NF-B, and ␣-tubulin (Transduction Laboratories (Lexington, KY) and Santa Cruz Biotechnology (Santa Cruz, CA)). Anti-GFP and anti-cytochrome c oxidase subunit 4 (COX4) antibodies were from CLONTECH. Where indicated, images were analyzed by the NIH Image program. Purification of rat liver mitochondrial proteins for Western blotting was performed as described (19).
Yeast Two-hybrid Interactions-The CytoTrap yeast two-hybrid system was from Stratagene. Unlike the traditional system, which depends upon protein-protein interaction in the nucleus (20), this assay system is based on the binding of an Sos-tagged bait protein to a cell membrane-anchored target protein (tagged with a myristoylation signal) that results in activation of the Ras signaling pathway, thereby permitting mutant yeast cdc25H to grow at 37°C using a selective agarose medium or plate containing galactose. Two constructs of Wox1 as baits (in the pSos vector) and three constructs of p53 as targets (in the pMyr vector) were made (see Table I). Binding interactions using combination of these vectors were performed. Vectors that are included in the system for positive binding interactions are pSos-MafB and pMyr-MafB (21), and those that are included for negative binding interactions are empty pSos and empty pMyr, pMyr and lamin C, or other vectors.

Molecular
Cloning of Murine Wox1-To further explore the mechanism whereby hyaluronidase enhances TNF cytotoxicity, we isolated a murine Wox1 cDNA (2197 bases; GenBank TM /EBI accession number AF187014) by differential display and cDNA library screening. The cDNA possesses an open reading frame, a typical Kozak sequence at the initiation site (ATG), and an upstream in-frame stop codon. The deduced murine WOX1 protein sequence (414 amino acids, 46 kDa) possesses two N-terminal WW domains (first domain, amino acids 18 -47; and second domain, amino acids 59 -87), an NLS (GKRKRV, amino acids 50 -55), and a C-terminal short-chain ADH domain (amino acids 121-330) (Fig. 1). WW domains are known to bind proteins with a particular proline motif, (A/P)PP(A/P)Y (22,23). Whether the WW domains of WOX1 bind to this motif is not known.
The human gene coding for WOX cDNAs (or known as WWOX or FOR) has been mapped to a fragile site on chromosome 16 (24 -26). Murine WOX1 is highly homologous to fulllength human WWOX (46 kDa) (24) and FOR II (46.7 kDa) (25). Three alternatively spliced variants are FOR I (41.2 kDa), FOR III (21.5 kDa), and FOR IV (4.1 kDa), which possess distinct C-terminal ends ( Fig. 1) (25). Two alternatively spliced variants we have identified and sequenced are human prostate WOX3 (identical to FOR III; GenBank TM /EBI accession numbers AI669330 and AF187015) ( Fig. 1) and FOR I-related human WOX5 (a partial clone; GenBank TM /EBI accession number AI219858), whose ADH domain, but not the N terminus, is identical to the sequence of FOR I.
Gene and Protein Expression-Murine L929 fibroblasts constitutively expressed a low level of Wox1 mRNA, as determined by Northern blotting (Fig. 2A). Exposure of L929 cells to hyaluronidase for 2-24 h resulted in increased Wox1 gene expression (ϳ2.3 kilobases), peaking at 8 -24 h (ϳ150% increase) ( Fig.  2A). The induced Wox1 gene expression correlates positively with the induction of TNF sensitivity in L929 cells, which requires pretreatment with hyaluronidase for at least 10 h (11-13).
As predicted, in vitro translation of the full-length murine Wox1 cDNA produced a protein of ϳ46 kDa, as analyzed by reducing SDS-polyacrylamide gel electrophoresis (Fig. 2B). A 30-kDa product was also observed (Fig. 2B). This protein is most likely a degradation product of 46-kDa WOX1 since Wox1 mRNA, which is derived from the cloned full-length cDNA, is unlikely to undergo alternative splicing.
WOX1 is a single chain protein and does not exist as a dimer, as determined by nonreducing SDS-polyacrylamide gel electrophoresis. A putative caspase recognition site is DIND (267-270). However, this site does not appear to be the proteolytic degradation site. Alteration of the DIND sequence to a noncaspase recognition sequence (GING) by site-directed mutagenesis failed to prevent WOX1 degradation (Fig. 2B). The caspase inhibitor peptide acetyl-Asp-Glu-Val-Asp-CHO (aldehyde) at 100 -200 M failed to block WOX1 degradation during in vitro translation. Furthermore, the serine protease inhibitors leupeptin and leuhistin at 100 M could not inhibit WOX1 degradation during in vitro translation.
Stimulation of L929 cells with hyaluronidase induced WOX1 protein expression (Fig. 2C). Our produced antibodies, which interacted with both human and mouse WOX1, recognized both 46-and 30-kDa WOX1 (Fig. 2C). Whether this 30-kDa WOX1 is a degraded protein from 46-kDa WOX1 remains to be determined. Our produced antibodies are specific since the preimmune serum failed to interact with both WOX1 proteins, the synthetic WOX1 peptide blocked the binding of the antibodies to WOX1, and the antiserum also interacted with the in vitro translated WOX1 protein (data not shown).
WOX1 Is Mainly Located in the Mitochondria, and TNF Mediates WOX1 Nuclear Translocation-Immunostaining of COS-7 fibroblasts with anti-WOX1 and anti-cytochrome c an- tibodies showed the presence of WOX1 in the mitochondria and nuclei, as determined by confocal microscopy and colocalization analysis (Fig. 3A). Similar results were observed using neonatal rat heart H9c2 cells (Fig. 3B). These results were further confirmed using human ovarian ME180 and HeLa, breast MCF-7, and neural SK-N-SH cells; isolated rat heart cardiomyocytes; and murine NIH/3T3 and L929 cells. The presence of WOX1 in the mitochondria was further confirmed by Western blotting using purified rat liver mitochondria (Fig. 3C). COX4 was examined as a marker protein for mitochondria (Fig. 3C).
Tagging of full-length murine WOX1 with an N-terminal GFP sequence (construct 2) (Table I) and expression in COS-7 cells revealed the presence of GFP-WOX1 in the mitochondria, as determined 24 h post-transfection (Fig. 3D). Mitochondria were stained by the membrane potential-sensitive stain Mitotracker Red CMXRos. Less than 10% of the transfected cells had nuclear localization of this protein.
By making successive deletion constructs (constructs 5-7) ( Table I) and expressing these constructs in COS-7 cells, the mitochondrial targeting sequence in WOX1 was mapped within the ADH domain (amino acids 209 -273; construct 7) (Fig. 3D). The truncated GFP-WOX1ww protein (amino acids 1-95; construct 4), which contains only the N-terminal WW domains and NLS, was expressed in the nucleus (Fig. 3E). Similarly, the human prostate GFP-WOX3 protein (see Fig. 1), which possesses the WW domains, the NLS, and a partial ADH domain, was expressed in the nucleus (data not shown).
A time course study showed TNF-mediated GFP-WOX1 nuclear translocation (Fig. 3F). This was determined by examining isolated nuclei by Western blotting using an established L929 cell line stably expressing the GFP-WOX1 protein (Fig.  3F). This observation correlates with the time point of TNFmediated cytochrome c release from mitochondria to the cytosol in L929 cells (Fig. 3F). Similarly, time-dependent endogenous WOX1 protein nuclear translocation was also observed in COS-7 and H9c2 cells upon stimulation with TNF for 20 min (data not shown).
TNF-mediated WOX1 nuclear translocation took at least 20 -40 min, which occurred shortly after TNF-mediated p65 NF-B nuclear translocation (at ϳ10 -15 min), as determined by immunostaining and fluorescent microscopy. Alteration of the NLS sequence (GKRKRV) to a less hydrophilic sequence (GQGTGV) by site-directed mutagenesis (construct 18) abolished the TNF-mediated nuclear translocation of this mutant protein. No nuclear translocation was observed when treating the ADH domain (GFP-WOX1adh, construct 5)-expressing cells with TNF. In parallel with the TNF-mediated mitochondrial permeability transition, treatment of COS-7 cells with atractyloside (2 mM) (27) for 40 min to increase the opening of mitochondrial transition pores also resulted in cytochrome c release to the cytosol and WOX1 nuclear translocation (data not shown). WOX1 Enhances TNF Cytotoxicity by Up-regulation of p53, but Down-regulation of Bcl-2 and Bcl-x L -To determine the effect of WOX1 on TNF cytotoxic functions, we established several L929 cell lines that stably expressed the above indicated GFP-WOX1 proteins (using constructs 2, 4, and 5). Expression of these proteins was determined by Western blotting using specific antibodies against GFP (Fig. 4A). As expected, the protein sizes of the expressed full-length GFP-WOX1 (predicted, 72 kDa; observed, 57 kDa) and GFP-WOX1adh (predicted, 50 kDa; observed, 34 kDa) were reduced by ϳ15 kDa, probably due to C-terminal degradation. Exposure of these GFP-WOX1 stable transfectants to TNF for 24 h resulted in enhancement of TNF-mediated cell death as compared with control cells transfected with GFP alone (Fig. 4B). Both the WW and ADH domains enhanced TNF-mediated L929 cell death (Fig. 4B). Expressing untagged full-length WOX1 (construct 1) in L929 cells also increased their TNF sensitivity (data not shown), indicating that GFP does not affect the WOX1 protein function.
In contrast, constitutive expression of antisense Wox1 mRNA (using construct 3) resulted in cellular resistance to TNF killing (resistance increase by 65-90%). These observations indicate that WOX1 participates in the TNF cytotoxicity pathway.
Although NF-B is believed to play an essential role in FIG. 3. WOX1 is mainly present in the mitochondria, and TNF mediates WOX1 nuclear translocation. A, COS-7 cells were stained with both anti-WOX1 and anti-cytochrome c antibodies, followed by staining with secondary antibodies, and subjected to confocal microscopy. Colocalization analysis showed that WOX1 is mainly located in the mitochondria and nuclei. B, similar results were observed by staining neonatal rat H9c2 cardiomyocytes using both antibodies. C, Western blotting showed the presence of WOX1 in the purified rat liver mitochondria. Also, the presence of COX4, a protein on the inner mitochondrial membrane, is regarded as a marker protein for mitochondria. D, when expressed in COS-7 cells, the full-length murine GFP-WOX1 protein (construct 2) (Table I)  blocking cell death by TNF, ionizing radiation, and anticancer drugs (28 -33), the WOX1-increased TNF cytotoxicity is not due to impaired NF-B activation or nuclear translocation. Time course studies showed that the kinetics of TNF-mediated IB␣ degradation were similar in both GFP-and GFP-WOX1-expressing L929 cells (Fig. 4C). Also, TNF induced NF-B (p65) nuclear translocation in both cells, as determined by immunostaining and fluorescent microscopy (data not shown). Additionally, TNF rapidly induced SAPK/JNK activation (peaking at 5-20 min) in both GFP-WOX1-and GFP-expressing cells, as determined using anti-phospho SAPK/JNK peptide antibodies in Western blotting. Thus, the involvement of NF-B and SAPK/JNK in WOX1-increased TNF killing is unlikely.
To exclude the possibility that the WOX1-increased TNF susceptibility in the established L929 transfectants is due to mutation of these cells, transient transfection experiments were performed. Transient expression of TRADD (34), the first adaptor protein recruited by the TNF receptor, in COS-7 fibroblasts resulted in activation of the TNF killing pathway and cell death (Fig. 4D). The TRADD-mediated death was signifi-cantly enhanced (2-3-fold increase) by WOX1 (using a noncytotoxic concentration) in cotransfection studies (Fig. 4D). Similar results were observed with ovarian ME180 and L929 cells (data not shown).
We next examined whether the WOX1-increased TNF killing is associated with down-regulation or up-regulation of apoptosis regulatory proteins. Western blot analysis showed that p53 expression was significantly increased (ϳ200%) in L929 cells stably expressing full-length GFP-WOX1 or the GFP-WOX1adh as compared with cells expressing GFP alone (Fig.  4E). Both GFP-WOX1 and GFP-WOX1adh proteins were present in the mitochondria (Fig. 3). In contrast, cells stably expressing nuclear GFP-WOX1ww failed to increase p53 expression (Fig. 4E). Notably, the ADH domain significantly suppressed the expression of the apoptosis inhibitors Bcl-2 and Bcl-x L (Ͼ85%), whereas the WW domains had no effect (Fig.  4E). IB␣ levels in these cells were not changed (Fig. 4E). The housekeeping protein ␣-tubulin was examined as control for protein loading (Fig. 4E). These data suggest that enhancement of TNF killing by WOX1 is associated in part with its  As summarized in Table II, both the ADH (construct 5) and WW (construct 4) domains induced DNA fragmentation when overexpressed in NIH/3T3 cells. As a positive control, overexpression of p53 induced apoptosis. In contrast, both antisense Wox1 (construct 3) and NLS-mutated WOX1 (NLSqgtg, construct 18) failed to mediate DNA fragmentation. Failure of NLS-mutated WOX1 in inducing apoptosis suggests that nuclear translocation of WOX1 is necessary for inducing cell death. Also, the presence of the WW domains in WOX1 may suppress the apoptosis-inducing activity of the ADH domain.
The WW domain-induced cell death is independent of caspases and serine proteases. Transient expression of the N-terminal WW domains (first and second domains; construct 10) or the first WW domain (construct 9) in NIH/3T3 cells also resulted in cell death 48 h post-transfection (Fig. 5B). These proteins contain the NLS, thus expressing in the nuclei. Exposure of the transfected cells to the caspase inhibitors acetyl-Asp-Glu-Val-Asp CHO (aldehyde) and benzyloxycarbonyl-Val-Ala-Asp-fluoromethyl ketone and the serine protease inhibitors leupeptin and leuhistin (100 M) failed to block cell death (Fig.  5B). These results indicate that the WW domain-mediated cell FIG. 4. WOX1 enhances TNF killing of L929 cells. A, four stable L929 transfectants were established to continuously express full-length GFP-WOX1 (predicted, 72 kDa; observed, 57 kDa) (construct 2), truncated GFP-WOX1ww (predicted, 37 kDa; observed, 33 kDa) (construct 4), truncated GFP-WOX1adh (predicted, 50 kDa; observed, 34 kDa) (construct 5), or GFP protein (26 kDa) alone. A degradation of 15 kDa was observed in the expressed GFP-WOX1 and GFP-WOX1adh proteins. Anti-GFP antibody was used in Western blotting. B, exposure of these WOX1expressing cells to TNF for 24 h resulted in enhancement of cell death as compared with the control GFP-expressing cells (n ϭ 8). C, a representative time course study showed that TNF-mediated IB␣ degradation was similar in both the GFP-WOX1-and GFP-expressing cells, indicating that WOX1 enhancement of TNF cytotoxicity is not due to impaired IB␣ degradation and NF-B activation. D, transfection of COS-7 cells (in 96-well plates) with a TRADD cDNA construct (in a cytomegalovirus-based pRK vector) by CaPO 4 resulted in cell death in 24 h (white bars), and the cell death was significantly enhanced by cotransfection with the full-length murine Wox1-pCR3.1 cDNA (0.2 g/well; construct 1; black bars). At this concentration, WOX1 could not mediate cell death. In controls, the empty vector pCR3.1 (0.2 g/well) failed to increase TRADD-mediated cell death. E, down-regulation of Bcl-2 and Bcl-x L expression and up-regulation of p53 expression were observed in the cells expressing GFP-WOX1 or GFP-WOX1adh, but not in the cells expressing GFP-WOX1ww or GFP alone. Both IB␣ and ␣-tubulin levels were not changed in these cells.
death is independent of caspases and serine proteases.
WOX1 and p53 Are Partners in Apoptosis-Since WOX1 increased p53 expression, we then investigated whether p53 is involved in the WOX1-mediated cell death. Transient expression of the WW domains mediated the death of NIH/3T3 cells in 48 h, and the killing function was increased by cotransfection with p53 (Fig. 6, A, panels a and b). Also, using non-killing concentrations of p53 and WOX1 in transfecting monocytic U937 cells (by electroporation), a synergistic killing effect was observed when combining p53 and WOX1 (data not shown).
Notably, antisense expression of WOX1 in NIH/3T3 cells abolished p53-mediated cell death (Fig. 6A, panel c). Similarly, antisense expression of WOX1 also inhibited p53 apoptosis of THP-1 cells in cotransfection experiments (Fig. 6B). These results suggest that there is a partnership between p53 and WOX1 in apoptosis.
However, using p53-deficient NCI-H1299 cells, transient expression of both the WW and ADH domains mediated cytochrome c release from mitochondria and cell death, indicating that WOX1-mediated cell death is independent of p53 (Fig. 6C). Similar results were observed using full-length WOX1 (data not shown). Together, these data suggest that WOX1-mediated apoptosis is independent of p53, but p53 apoptosis requires the participation of WOX1.
The WW Domains of WOX1 Bind to the Proline-rich Region of p53-Confocal microscopy and colocalization analysis revealed that p53 and WOX1 colocalized in the cytosol and partly in the nuclei in MCF-7 cells (Fig. 7A). Similar results were obtained using other cells such as ME180 and COS-7.
Immunoprecipitation of L929 cytosolic lysates with anti-p53 antibodies resulted in coprecipitation of both p53 and WOX1 (Fig. 7B), indicating the binding interactions between endogenous p53 and WOX1. The presence of p53 in the precipitates was confirmed using anti-p53 antibodies in Western blotting (data not shown). Stimulation of L929 cells with TNF for 2 h resulted in migration of both proteins to the nuclei (determined by immunostaining) (data not shown) and the disappearance of both proteins from the cytosolic lysates in coprecipitation studies (Fig. 7B).
Yeast two-hybrid analysis showed that the proline-rich region of p53 (amino acids 66 -110) physically interacts with the WW domains of WOX1 in vivo (Fig. 7C). In negative controls, no binding interactions were observed using antisense WOX1ww and p53, empty vector (pSos) and empty vector (pMyr), and MafB and lamin C (Fig. 7C). MafB self-binding interactions were tested as positive binding controls (Fig. 7C).

DISCUSSION
In this study, we cloned and functionally characterized the murine WOX1 protein by antibodies, GFP tagging and expression, and other approaches. The gene encoding WOX1 is located on a fragile chromosomal site (24 -26). Homozygous deletion of this gene has been found in various cancers (24 -26). Whether WOX1 plays a role in cancer development remains to be established. We determined that WOX1 is located mainly in the mitochondria. Mitochondrial intermembrane space is a reservoir for a variety of apoptogenic proteins such as cytochrome c; procaspase-2, -3, and -9; and apoptosis-inducing factor (AIF) (35). Whether WOX1 is present in this intermembrane space remains to be established. Apoptotic stimuli such as TNF and staurosporine induce WOX1 nuclear translocation. WOX1 enhances TNF cytotoxic function via its nuclear targeting WW domains and the mitochondrial targeting ADH domain, suggesting that WOX1 functions at both cytosolic and nuclear levels. Functionally, WOX1 mediates apoptosis when overexpressed. WOX1 binds p53 in the cytosol. WOX1-mediated apoptosis is independent of p53, whereas p53-mediated cell death requires the participation of WOX1. This observation suggests that WOX1 is an essential partner of p53 in apoptosis.
In agreement with other studies (24,25), Wox1 mRNA is ubiquitously expressed in most tissues and organs in mouse, as determined by reverse-transcription-polymerase chain reaction (data not shown). Based on the gene structure, four splice variants of WOX proteins are predicted (25). However, we observed additional WOX protein species at high molecular sizes (65 and 100 kDa) in Western blotting using human organs and cell lines (data not shown). Also, three mRNA transcripts probably encoding high molecular mass WOX proteins have been found (25). Indeed, additional splice variants have also been found in the updated expressed sequence tag data base. Accordingly, the functional properties of these proteins remain to be established.
TNF-mediated WOX1 nuclear translocation is independent of the TNF signaling pathway that leads to phosphorylation  and degradation of IB␣ and activation of p65 NF-B and SAPK/JNK (36 -38). For example, the GFP-WOX1-expressing COS-7 cells were pretreated with the proteasome inhibitor benzyloxycarbonyl-Leu-Leu-Leu CHO (aldehyde) (10 M) to block IB␣ degradation or with the IB␣ phosphorylation inhibitors Bay 11-7082 and Bay 11-7085 (30 M) for 1 h, followed by exposure to TNF. These treatments inhibited TNF-mediated p65 NF-B activation or nuclear translocation, but failed to abolish the GFP-WOX1 nuclear translocation, as determined by immunostaining and fluorescent microscopy (data not shown). Nonetheless, TNF-mediated WOX1 release from mitochondria could be dependent upon activation of BID (39). TNF mediates cleavage of BID to truncated BID, which translocates to the mitochondria. Truncated BID oligomerizes BAK to generate membrane pores, thus allowing cytochrome c and probably WOX1 release. Truncated BID-mediated cytochrome c release does not appear to be involved in the opening of mitochondrial transition pores (40). However, atractyloside, an inducer of mitochondrial permeability transition, also induces WOX1 nuclear translocation. This observation suggests that WOX1 release from mitochondria could be dependent upon mitochondrial transition pores as well as BAK oligomerization.
The WW domains of WOX1 are more potent than full-length WOX1 in sensitizing the TNF-resistant COS-7 cells to TNF killing. We found that when COS-7 cells were transiently transfected with the N-terminal WW domains of WOX1 and cultured for 16 -24 h, followed by exposure to TNF, these cells underwent nuclear fragmentation in 3 h and subsequent rupture of the cytosolic components and nuclear condensation in 6 -16 h. Nonetheless, a prolonged treatment (Ͼ12 h) of the full-length WOX1-expressing COS-7 cells with TNF is required to induce cell death.
An intriguing finding in our study is that stable expression of the ADH domain or full-length WOX1 in the mitochondria resulted in down-regulation of Bcl-2 and Bcl-x L , but up-regulation of p53 in L929 cells. In contrast, the WW domains, when expressed in the nuclei, failed to modulate the expression of these proteins. These results suggest that WOX1 indirectly regulates the expression of p53, Bcl-2, and Bcl-x L .
The anti-apoptotic Bcl-2 and Bcl-x L proteins block mitochondrial permeability transition and prevent cytochrome c release from mitochondria (41). Bcl-x L blocks cytochrome c release by binding to the anion channel voltage-dependent anion channel on the outer membrane of mitochondria (42). Binding of Bcl-x L to voltage-dependent anion channel results in closure of the voltage-dependent anion channel. We determined that WOX1 significantly reduces the expression of Bcl-2 and Bcl-x L in mitochondria. This event may result in opening of the mitochondrial permeability transition pores and release of apoptogenic proteins from the intermembrane space. This notion is supported by the observation that transient overexpression of the ADH domain in cells caused cytochrome c release and death.
Another intriguing finding is that when overexpressed, the mitochondrial targeting ADH domain alone (using three regions of the ADH domains; constructs 5-7) was capable of inducing cell death. However, the apoptosis-inducing activity was not found in NLS-mutated full-length WOX1 (construct 18), which indicates that nuclear translocation is needed for the WW domains to mediate cell death. Also, the presence of the mutated WW domains appears to suppress cell death by the ADH domain. Suppression of the ADH domain function is probably related to protein folding when the WW domains are present in the WOX1 protein. Other types of dehydrogenase domain proteins such as AIF (35) and the CC3 protein (43) have been shown to induce cell death when overexpressed.
Endogenous WOX1 colocalizes with p53 in the cytosol. Ectopic expression of both GFP-WOX1 and red fluorescent protein-tagged p53 in COS-7 cells also results in cytosolic colocalization (Ͼ50% in p53/WOX1-expressing cells) (data not shown). Co-immunoprecipitation studies further support binding of p53 with WOX1 in the cytosol. Yeast two-hybrid experiments show the binding of the WW domains of WOX1 to the proline-rich region (amino acids 66 -110) in p53. Based on these observations, it is reasonable to suggest that both p53 and WOX1 migrate together to the nucleus in response to TNF.
Ectopic expression of p53 and WOX1 showed that both proteins mediate apoptosis in a synergistic manner. Although WOX1 can mediate apoptosis independently of p53, blocking of WOX1 expression by antisense mRNA abolishes p53 apoptosis. The inhibition of p53 apoptosis is not due to blocking of p53 protein synthesis by the antisense Wox1 mRNA (data not shown). These observations strongly indicate that WOX1 is an essential partner of p53 in apoptosis. The proline-rich region has been shown to be necessary for p53-mediated apoptosis (44). Our data suggest that binding of WOX1 to this region in p53 appears to be essential for p53 apoptosis-inducing activity.
A wide range of transcription factors including c-Jun, AP-2, NF-E2, CAAT/enhancer-binding protein-␣, and PEBP2/CBF, contain the WW domain-binding motif (22,23). Thus, most of the WW domain-containing proteins act as gene transcription activators or coactivators (45,46). When overexpressed, the WW domains of WOX1 mediate apoptosis. Whether this is related to the transcriptional activation of apoptotic genes by the WW domains of WOX1 remains to be established. None-theless, our data show that the WW domains fail to increase p53 protein expression and suppress Bcl-2 and Bcl-x L expression.
Susin et al. (35) isolated mitochondrial AIF, a homolog of bacterial oxidoreductase. Once released from mitochondria, AIF translocates to the nucleus. Although recombinant AIF induces apoptosis of isolated nuclei (35), whether nuclear AIF induces chromatin condensation and nuclear DNA fragmentation in vivo is unknown. In response to apoptogenic signals such as staurosporine, both WOX1 and AIF migrate to the nucleus and mediate cell death in a caspase-independent mechanism. TNF induces WOX1 nuclear translocation. Whether AIF migrates to the nucleus in response to TNF is unknown. WOX1 enhances TNF cytotoxicity by increasing the expression of p53 (and probably other pro-apoptotic proteins) as well as suppressing the expression of Bcl-2 and Bcl-x L . In contrast, AIF does not appear to be involved in the regulation of protein expression. Whether AIF and WOX1 act synergistically in mediating cell death is not known.
Although the TNF signaling pathway that leads to caspase activation and cell death has been well defined, we 2 and others (47) have shown that TNF-mediated cell death cannot be blocked by inhibitors of caspases. This raises the possibility that TNF induces a caspase-independent killing pathway. Data from us and Susin et al. (35) support that both WOX1 and AIF are the downstream mediators of the caspase-independent TNF killing pathway.
Finally, a high abundance of WOX1 was observed in rat heart (data not shown). This suggests that WOX1 plays a homeostatic role in this organ. The heart is a TNF-producing 2 N.-S. Chang, N. Pratt, J. Heath, and L. Schultz, unpublished data.

FIG. 7. p53 and WOX1 colocalization and binding interactions.
A, confocal microscopy and colocalization analysis revealed that p53 and WOX1 are colocalized in the cytosol and partly in the nucleus in MCF-7 cells (magnification ϫ 400). B, immunoprecipitation (IP) of endogenous WOX1 with anti-WOX1 antibodies from the cytosolic lysates of L929 cells, followed by blotting with anti-WOX1 antibodies, revealed the presence of 46-kDa WOX1 (third lane). Also, precipitation with anti-p53 antibodies, followed by blotting with anti-WOX1 antibodies, also resulted in the appearance of 46-kDa WOX1 (second lane), indicating binding of endogenous p53 to WOX1. In the control without antibodies added, no precipitated protein was observed (first lane). Exposure of L929 cells to TNF (20 ng/ml) for 1 h resulted in the disappearance of WOX1 from the cytosol, due to migration of both p53 and WOX1 to the nuclei (fifth and sixth lanes). C, yeast two-hybrid analysis was performed (see "Experimental Procedures"). Interactions between target and bait proteins allow Sos-mediated activation of the Ras signaling pathway, which permits the temperature-sensitive yeast cdc25H to grow at 37°C. Positive binding interactions between full-length p53 (construct 14) and full-length WOX1 (construct 12) or WOX1ww (both WW domains; construct 13) are demonstrated, as evidenced by the growth of yeast at 37°C. Similar results were obtained with the N-terminal proline-rich region in p53 (amino acids 1-100 and 66 -100; constructs 15 and 16, respectively) and full-length WOX1. p53 failed to bind to the antisense construct of WOX1ww (reverse orientation of WOX1ww cDNA). MafB protein self-interaction is a positive control for the assay system. In negative controls, the yeast failed to grow at 37°C when testing MafB-lamin C or empty vector (pSos)-empty vector (pMyr) interactions. Two representative colonies (out of [15][16][17][18][19][20][21][22][23][24][25][26][27][28][29][30] are shown from each binding experiments. organ, and TNF plays a key role in the pathogenesis of congestive heart failure (48,49). Patients with chronic and severe congestive heart failure have increased levels of TNF in the circulation and cardiac tissues. TNF exerts a negative inotropic effect and triggers the apoptotic process in cardiomyocytes. Whether WOX1 plays a major role in the TNF-mediated apoptosis of cardiomyocytes remains to be established.