The Cytoplasmic Domain of the Lymphotoxin-β Receptor Mediates Cell Death in HeLa Cells*

Activation of lymphotoxin-β receptor (LT-βR) by conjugation with heterotrimeric lymphotoxin, LT-α1/β2, or by cross-linking with anti-LT-βR antibodies can trigger apoptosis. We have observed that overexpression of either LT-βR or the cytoplasmic domain of LT-βR (LT-βR(CD)) also induces apoptosis, which may be attributed to the tendency of LT-βR(CD) to self-associate. The self-association domain of LT-βR(CD) was mapped to amino acids 324–377, a region of the protein that is also essential for LT-βR-induced apoptosis. Furthermore, we have shown that LT-βR(CD)-induced apoptosis could be inhibited by a TRAF3 dominant negative mutant and by the caspase inhibitors Z-VAD-FMK, DEVD-FMK, and CrmA. The ligand-independent apoptosis induced by LT-βR(CD) will help us to further dissect LT-βR signaling pathway.

Lymphotoxin-␤ receptor (LT-␤R) 1 is a member of the tumor necrosis factor receptor (TNFR) superfamily and is expressed on the surface of most of cell types, including cells of epithelial and myeloid lineages but not on T and B lymphocytes (1,2). LT-␤R can bind specifically to two ligands: the membrane form of lymphotoxin, LT-␣1/␤2, (3,4); and LIGHT, a recently identified member of TNF superfamily (5,6). LT-␤R has been speculated to play an essential role in the development of lymphoid organs. In LT-␣ knock-out mice, the development of lymphoid organs is prevented (7). Studies involving LT-␤ knock-out mice have shown impairment of lymph node development and loss of splenic architecture (8). Similar results were observed when the soluble LT-␤ receptor-immunoglobulin Fc chimera fusion protein was introduced into the embryonic circulation by injecting pregnant mice (9). Direct evidence to demonstrate the role of LT-␤R in lymphoid organ development comes from the fact that LT-␤R deficient mice lack Peyer's patches, colonassociated lymphoid tissues, and all lymph nodes (10). Moreover, injection of the agonist anti-LT-␤R monoclonal antibody into the uteri of pregnant LT-␣ knock-out mice has been shown to result in the genesis of lymph nodes in their progeny (11). In addition to its role in lymphoid organ development, stimulation of LT-␤R on certain cell lines by LT-␣1/␤2 or anti-LT-␤R antibodies can induce cell death (12), chemokine secretion (13), and activation of nuclear factor B (NF-B) (14). Thus, LT-␤R may also have important biological functions in the mature individuals.
The cytoplasmic domain of LT-␤R, like other members of the TNF receptor family, does not contain consensus sequences characteristic of enzymatic activity. Therefore, signaling is thought to be mediated by the proteins interacting with LT-␤R. Two serine/threonine protein kinases, p50 and p80, have been shown to associate with LT-␤R(CD) specifically (15), but the function of p50 and p80 in the LT-␤R signaling pathway is still the subject of intensive study. Moreover, two members of the TNF receptor-associated factor (TRAF) family, TRAF3 and TRAF5, were found to associate with LT-␤R (16,17). Further study has indicated that TRAF3 plays an important role in mediating LT-␤R-induced apoptosis (16,18), whereas TRAF5 has been shown to be involved in the activation of NF-B (17). On the other hand, several members of TNFR superfamily (such as TNFRI, Fas, DR3, DR4, and DR5) contain a common motif, the death domain, in their cytoplasmic region (19 -24). These "death receptors" interact with other death domaincontaining proteins, such as TRADD (25), MORT1/FADD (26), RIP (27), and RAIDD (28), via their death domains. MORT1/ FADD and RAIDD can in turn interact with MACH1/FLICE (caspase-8) and caspase-2, respectively (28 -30), and thus initiate the activation of caspase cascades to execute apoptosis (31). LT-␤R(CD) does not contain a death domain, but signaling through LT-␤R can also induce apoptosis (12). It will be interesting to map the region of LT-␤R responsible for its cytotoxic effect and to determine whether LT-␤R mediates apoptosis via the activation of caspase cascades.
Both TNFRI and TNFRII can induce cell death when bound by TNF. It has also been reported that clustering of TNF receptors due to interaction either with trivalent TNF or with an agonist antibody is a crucial step for subsequent intracellular signaling (32,33). Because the cytoplasmic domain of TNFRI can self-associate through its death domain (34,35), overexpression of TNFRI or of its cytoplasmic domain alone can also induce receptor clustering resulting in the activation of downstream signaling pathways (34,36). In contrast, TNFRII has no death domain and shows no tendency to self-associate, nor does a high level of TNFRII expression result in spontaneous signaling (34,37). In this study, we have shown that LT-␤R(CD) is capable of self-association, despite the absence of a death domain. Moreover, overexpression of LT-␤R or LT-␤R(CD) was sufficient to trigger apoptosis without the need for ligand conjugation. The cytotoxic effect mediated by LT-␤R(CD) relies on the presence of its self-association domain, suggesting that this domain is critical in the LT-␤R signaling pathway.
Plasmid Construction-A cDNA fragment encoding the full-length LT-␤R was amplified by reverse transcription-polymerase chain reaction using a HepG2-derived cDNA template and the primers 5Ј-CGG-GATCCATGCTCCTGCCTTGGGCCAC-3Ј (sense) and 5Ј-CGGGATC-CTCAGTCATGGGTGATAAATTGG-3Ј (antisense). The DNA fragment containing the human LT-␤R cytoplasmic domain was amplified by polymerase chain reaction using the primers 5Ј-GGAATTCCAAGAGC-CACCCTTCTCTCTGC-3Ј (sense) and 5Ј-GGAATTCCTCAGTCAT-GGGTGATAAATTGG-3Ј (antisense). LT-␤R(CD) were subcloned into the pGEX vector as described previously (15). Deletion mutants of LT-␤R(CD) were generated by restriction enzyme digestion. Expression of full-length LT-␤R, LT-␤R(CD), and its deletion mutants in mammalian cells was achieved by subcloning the cDNAs into the pFLAG-CMV2 vector (Eastman Kodak, Co.) in frame with the FLAG tag at 5Ј end. The dominant negative mutant of TRAF3, TRAF3(367-568) (16), was amplified by polymerase chain reaction and subcloned into the pFLAG-CMV2 vector. For inducible expression of LT-␤R(CD), the cDNA encoding the LT-␤R(CD) was subcloned into the tetracycline-controlled expression vector, pTRE (CLONTECH).
In Vitro Binding and Coprecipitation Assays-For surface biotinylation, HeLa cells were incubated in ice-cold phosphate-buffered saline containing 0.5 mg/ml sulfo-NHS-LC-Biotin (Pierce) for 30 min at 4°C. The cells were washed once with phosphate-buffered saline and then incubated in Dulbecco's modified Eagle's medium for 15 min before harvesting. 1 ϫ 10 7 surface biotinylated cells were resuspended in 1 ml of lysis buffer (20 mM Tris, pH 7.7, 0.5% (v/v) Nonidet P-40, 200 mM NaCl, 50 mM NaF, 0.2 mM sodium orthovanadate, 1 mM phenylmethylsulfonyl fluoride, 2 g/ml leupeptin, 2 g/ml aprotonin, and 0.1% (v/v) 2-mercaptoethanol) for 1 h at 4°C, followed by centrifugation at 9000 ϫ g for 10 min. The supernatant was precleared with 15 g of GST bound to glutathione-agarose beads for 2 h at 4°C. The precleared supernatant was mixed with 10 g of GST-LT-␤R(CD) attached to glutathioneagarose beads for 2 h at 4°C. Finally, the beads were washed six times with lysis buffer. The sample was fractionated by SDS-polyacrylamide gel electrophoresis (7.5% (w/v) acrylamide), followed by Western blotting. The blot was probed with peroxidase-conjugated avidin and then processed using ECL detection reagents (Amersham International, Inc.).
Yeast Two-hybrid Interaction Analysis-A cDNA encoding the cytoplasmic domain of LT-␤R was subcloned into the yeast vectors pAS2-1 (GAL4 DNA-binding domain construct) and pACT2 (GAL4 activation domain construct) (CLONTECH), respectively. For protein-protein interaction assays, plasmids were used to transform Saccharomyces cerevisiae strain Y190. Positive clones were selected by prototrophy for histidine and tested by filter assays for ␤-galactosidase activity as described by the vender (CLONTECH).
Apoptosis Assay-To identify the cells transfected with LT-␤R(CD), DNA constructs were cotransfected with a construct designed to express ␤-galactosidase. Cells were then tested for ␤-galactosidase activity as described by Yang et al. (38). Briefly, HeLa cells were plated in 6-well (35 mm) plates at a density of 2 ϫ 10 5 cells/well and left overnight. The next day, the cells were cotransfected with the LT-␤R construct and pCMV-LacZ (at a ratio of 10:1), using LipofectAMINE TM (Life Technologies, Inc.). After 24 h of incubation, cells were fixed in 1% glutaraldehyde in PBS at 4°C for 5 min. Transfected cells became blue after incubation with 1 mg/ml X-gal/5 mM potassium ferricyanide/5 mM ferrocyanide/2 mM MgCl 2 /0.02% Nonidet P-40/0.01% SDS in PBS at 37°C for 4 h. The morphology of transfectants was observed using a phase contrast microscope (Nikon), and the percentage of apoptotic cells was calculated as the number of blue cells with apoptotic morphology divided by the total number of blue cells. To inhibit the cytotoxic effect, DEVD-FMK (Calbiochem, Inc.) or Z-VAD-FMK (KAMIYA Biomedical, Co) was added to culture medium after transfection. At least 1000 blue cells were counted for each sample.
MTT Test-The survival rate of cells was determined by an MTT test. Briefly, cells were seeded in 96-well flat bottom plates at a density of 3 ϫ 10 3 cells/0.1 ml. After the indicated time period, 10 l of 5 mg/ml MTT/well was added, and the cells were incubated at 37°C for 4 h. The cells were then lysed by the addition of 100 l of 10% SDS in 10 mM HCl/well and incubation at 37°C for 24 h. The optical density of each sample was determined by measuring the absorbance at 570 nm versus 650 nm using an enzyme-linked immunosorbent assay reader (TECAN, RainBow).
Immunofluorescence Microscopy-Cells were fixed with 1% paraformaldehyde in PBS at room temperature for 20 min and then permeabilized with acetone at Ϫ20°C for 3 min. Cells were then incubated with anti-FLAG monoclonal antibody (5 g/ml) at room temperature for 1 h, followed by incubation with fluorescein isothiocyanateconjugated goat anti-mouse IgG at room temperature for 1 h after washing with PBS three times. Cells were then examined with a MRC600 scanning confocal microscope (Bio-Rad). All the antibodies were diluted in 1% bovine serum albumin/PBS.

RESULTS
Mapping of the LT-␤R Self-association Domain-In a yeast two-hybrid screen using LT-␤R(CD) as a bait, Chen et al. (39) found that most of the positive clones isolated from a human liver cDNA library corresponded to LT-␤R. This was indicative of a tendency to self-associate. To confirm the ability of LT-␤R to self-associate, we tested whether a GST-LT-␤R(CD) fusion protein is capable of interacting with endogenous LT-␤R from HeLa cells. HeLa cells were surface biotinylated as described under "Experimental Procedures." A GST-pull down assay was carried out, and precipitated proteins were detected with avidin. We found that a 70-kDa protein was coprecipitated with GST-LT-␤R(CD) (Fig. 1, lane 3) but not with GST under the same conditions (Fig. 1, lane 2). Furthermore, the 70-kDa protein associated with GST-LT-␤R(CD) could be precleared from HeLa cell extracts by anti-LT-␤R antibodies (Fig. 1, lane 4). We are confident that the 70-kDa protein is endogenous LT-␤R for following reasons: (i) the 70-kDa protein is a surface protein because it could be surface biotinylated and (ii) the molecular mass (70 kDa) of the LT-␤R(CD)-associated protein (Fig. 1, lane  3) is the same as that of endogenous LT-␤R immunoprecipitated by anti-LT-␤R antibodies (Fig. 1, lane 5). GST-LT-␤R(CD) interacted less strongly with endogenous LT-␤R than did anti-LT-␤R antibodies. This is probably due to the lower affinity of the protein-protein association (Fig. 1, lane 3) than that of the antigen-antibody interaction (Fig. 1, lane 5). Based on the observations above, we concluded that LT-␤R(CD) can associate with endogenous LT-␤R and further confirmed the tendency of LT-␤R to self-associate.
To identify the self-association domain, LT-␤R(CD) and its deletion mutants were subcloned into the vectors pAS2-1 (containing the GAL4 DNA binding domain) and pACT2 (containing the GAL4 activation domain), and the resultant constructs were used to carry out yeast two-hybrid tests. As shown in Table I, LT-␤R(CD), LT-␤R(CD)(aa234-377), and LT-␤R(CD)(aa324-377) clones exhibited ␤-galactosidase activity in filter assays when cotransformed with LT-␤R(CD), indicating that self-association had occurred. In contrast, the deletion mutants LT-␤R(CD)(aa234-324) and LT-␤R(CD)(⌬324-377) did not give rise to ␤-galactosidase activity, suggesting that the absence of amino acids 324-377 resulted in a loss of self-association. These results show that amino acids 324-377 of LT-␤R(CD) are necessary and sufficient for LT-␤R self-association.

Expression of LT-␤R(CD) Triggers Cell Death-Recombinant
LT-␣1/␤2 or anti-LT-␤R antibody has been shown to trigger apoptosis in certain cell lines, including the human cervical carcinoma cell line HeLa (12,39). Here, we asked whether the overexpression of LT-␤R or LT-␤R(CD) has a similar effect. Full-length LT-␤R and LT-␤R(CD) were subcloned into FLAGtagged expression vectors and cotransfected with pCMV-LacZ (which expresses ␤-galactosidase) into HeLa cells. Of the transfected cells that expressed ␤-galactosidase activity, only those expressing either full-length LT-␤R (Fig. 2B) or LT-␤R(CD) (Fig. 2C) became rounded and condensed, which are the typical morphological alterations of adherent cells undergoing apoptosis. In addition, nuclear condensation was observed when the cells were stained with Hoechest 33342 (data not shown). In contrast, cells transfected with the pFLAG vector alone did not undergo apoptosis (Fig. 2A). The proportions of cells expressing LT-␤R or LT-␤R(CD) that underwent apoptosis were about 76 and 74%, respectively, whereas this occurred in only ϳ8% of control cells (Fig. 2D). From this observation, it is clear that overexpression of LT-␤R or LT-␤R(CD) can result in cell death, without the need for ligand binding or receptor cross-linking by an agonist antibody. Similar results were obtained in the human adenocarcinoma cell line HT29 overexpressing LT-␤R or LT-␤R(CD) (data not shown).
To further confirm this observation, we used the Tet-offinducible system to allow the expression of LT-␤R(CD) to be turned on when tetracycline was removed from culture medium. In Tet-off HeLa cells stably transfected with pTRE-LT-␤R(CD), LT-␤R(CD) protein became detectable at day 3 after induction (data not shown). Significant cytotoxicity was observed when tetracycline was removed (Fig. 3A, right panel), whereas cells did not undergo apoptosis in the presence of tetracycline (Fig. 3A, left panel). By staining the surviving cells with MTT, we found that cell death became apparent at day 4 after removal of tetracycline (Fig. 3B). Based on the observations above, we concluded that expression of LT-␤R(CD) alone is sufficient to induce apoptosis.
Signaling through TNFRI and Fas can initiate caspase cascades to execute apoptosis (31). To examine whether LT-␤R(CD) can also activate caspases, the effects of Z-VAD-FMK (a broad spectrum caspase inhibitor), DEVD-FMK (a CPP32 inhibitor), and the CrmA of cowpox virus on LT-␤R(CD)-induced apoptosis were tested. As shown in Fig. 4B, the cytotoxicity induced by LT-␤R(CD) was blocked in cells coexpressing

Mapping of LT-␤R self-association domain by yeast two-hybrid assay
The cDNAs encoding LT-␤R(CD) and its deletion mutants were subcloned into the yeast pAS2-1 (GAL4 DNA-binding domain construct) and pACT2 (GAL4 activation domain construct) vectors as shown. The yeast strain Y190 was transformed with various combinations of these expression constructs. Clones phototrophic for histidine were subjected to filter assays. Blue color indicated the presence of ␤-galactosidase activity.
Protein fused to GAL4 domain CrmA. In addition, LT-␤R(CD)-induced apoptosis was almost completely inhibited by Z-VAD-FMK but only partially by DEVD-FMK. This observation suggests that caspase(s) may participate in the apoptotic pathway activated by LT-␤R(CD). The partial inhibition of DEVD-FMK might result from the lower membrane permeability of HeLa cells for this species, because similar results were obtained when looking at TNFinduced cytotoxicity under the same conditions (data not shown).
Mapping of the Apoptotic Domain of LT-␤R-Because there is no death domain in the cytoplasmic region of LT-␤R, we wished to identify the sequence responsible for LT-␤R(CD)induced apoptosis. As shown in Fig. 5, apoptosis was observed in cells transfected with either LT-␤R(CD) or the deletion mutant LT-␤R(CD)(aa234 -377). In contrast, the deletion mutants LT-␤R(CD)(aa234 -324) and LT-␤R(CD)(⌬324 -377) had no effect on cell viability. These results provided direct evidence that amino acids 324 -377 are essential for LT-␤R(CD)-induced apoptosis. This is also the region required for LT-␤R self-association (Table I).
Localization of the LT-␤R Deletion Mutants-Loss of the extracellular and transmembrane domains would be expected to result in LT-␤R(CD) in the cytoplasm of cells. However, LT-␤R(CD) was shown to associate with endogenous full-length LT-␤R, which is a transmembrane protein (Fig. 1). Therefore, we wished to clarify the localization of the LT-␤R(CD). HeLa cells transiently transfected with LT-␤R(CD) or its deletion mutants were processed for immunostaining (Fig. 6). Interestingly, we found that LT-␤R(CD) and LT-␤R(CD)(aa234 -377) were localized in the proximity of plasma membrane (Fig. 6, A  and B), whereas deletion mutants lacking the self-association domain (LT-␤R(CD)(aa234 -324) and LT-␤R(CD)(⌬324 -377)) were predominantly localized in the cytoplasm (Fig. 6, C and  D). The membrane juxtaposition of LT-␤R(CD) was also observed in Tet-off HeLa cells stably transfected with LT-␤R(CD) (data not shown). Considering that LT-␤R(CD) and LT-␤R(CD)(aa234 -377) contain the self-association domain and have the ability to self-associate ( Fig. 1 and Table I), it is conceivable that they might be directed to the inner side of plasma membrane through their interactions with endogenous LT-␤R. This observation suggests that the self-association domains are able to interact in vivo.

DISCUSSION
The cytoplasmic domains of TNFRI and Fas have been shown to be able to self-associate via their death domains. This self-association can prompt signaling events, giving rise to TNF and Fas effects (34,35). In contrast, the cytoplasmic region of TNFRII has no death domain and shows no tendency to selfassociate, nor does overexpression of this receptor induce cell death (34,37). Like TNFRII, LT-␤R does not contain a death domain. However, studies using the yeast two-hybrid and GSTpull down systems have clearly demonstrated the self-association tendency of LT-␤R(CD) (Ref. 39 and the present report). In addition, we have observed that overexpression of LT-␤R or LT-␤R(CD) is sufficient to trigger apoptosis, without the need to cross-link LT-␤R with ligand or an agonist antibody. It has been shown that a dominant negative mutant of TRAF3, TRAF3(367-568), can inhibit apoptosis induced by LT-␣1/␤2 or anti-LT-␤R antibodies (16). Likewise, our results showed the dominant negative effect of TRAF3(367-568) on LT-␤R(CD)induced apoptosis. This result suggests that apoptotic signaling mediated by LT-␤R(CD) may share the same pathway as that triggered by ligand conjugation. Furthermore, activation of NF-B was also observed in cells overexpressing LT-␤R(CD). 2 This is in accordance with previous reports that stimulation of LT-␤R by recombinant LT-␣1/␤2 or anti-LT-␤R antibodies can transduce signals not only for apoptosis (12) but also for NF-B activation (14). Based on the observations above, overexpression of LT-␤R(CD) seems to be able to activate downstream signaling events equivalent to those induced by LT-␣1/␤2 or anti-LT-␤R antibodies .
It has been shown that apoptotic signaling mediated by TNFRI or Fas occurs via interaction with other death domaincontaining proteins, which can initiate caspase cascades to execute apoptosis (31). Here, we demonstrated that LT-␤R(CD)-induced apoptosis was inhibited by several caspase inhibitors: CrmA, Z-VAD-FMK, and DEVD-FMK, suggesting that the activation of CPP32 or ICE-related caspases might be involved in the apoptotic signaling of LT-␤R. Because LT-␤R does not contain a death domain, it will be intriguing to discover how LT-␤R initiates the caspase pathway. This question might be answered by examining whether TRAF3 or other LT-␤R(CD)-associated proteins could recruit and activate caspase(s).
LT-␣1/␤2 or anti-LT-␤R antibodies cannot trigger apoptosis without the presence of IFN-␥ (12). However, LT-␤R(CD)-induced apoptosis does not require the presence of IFN-␥, nor did IFN-␥ result in a synergistic enhancement of LT-␤R(CD) cytotoxicity (data not shown). The mechanism of IFN-␥ action in apoptosis induced by LT-␤R is still unclear. Recently, IFN-␥ was found to modulate cell death by inducing several apoptosisrelated genes, including the TNFR family members, TNFRI and Fas; a bcl-2 family member, bak; and the caspase family members, ICE, CPP32, and FLICE (40). Nevertheless, IFN-␥ does not up-regulate LT-␤R expression on HT-29 (12) and HeLa (data not shown) cell lines. It is possible that IFN-␥ might modulate LT-␤R-induced cell death by regulating caspases involved in the cytotoxic effect. However, the amounts of LT-␤R(CD) expressed by transfected cells might be sufficient to initiate caspase cascades to execute apoptosis, such that IFN-␥ does not further enhance the cytotoxic effect. This speculation is supported by the fact that LT-␤R(CD)-induced apoptosis can be inhibited by several caspase inhibitors, suggesting that the activation of caspases is important in this apoptotic pathway.
Because there are no consensus sequences for membrane anchorage in the FLAG tag or LT-␤R(CD), the FLAG-LT-␤R(CD) fusion protein would be expected to express as a cytoplasmic protein. However, the FLAG-LT-␤R(CD) was shown to be localized in the proximity of plasma membrane. Interestingly, we also observed that membrane juxtaposition of LT-␤R(CD) and its deletion mutants correlates with their ability to self-associate and to induce apoptosis. Because the self-association domain is required for targeting of LT-␤R(CD) to the plasma membrane (Fig. 6), we speculate that membrane juxtaposition of LT-␤R(CD) might be due to its association with endogenous LT-␤R (Fig. 1), which may therefore induce receptor clustering to activate apoptotic signaling. Nevertheless, our result does not rule out the possibility that LT-␤R(CD) can be linked to the membrane by other proteins, either located on the inner side of cell membrane or associated with endogenous LT-␤R.
Previous FRI has a strong tendency to self-associate and interact with other death domain-containing proteins and plays an obligatory role in signaling cell death. In the present report, selfassociation of LT-␤R(CD) was mapped to amino acids 324 -377, a region that is also required for apoptotic signaling. Interestingly, the self-association domain (amino acids 324 -377) is the same region as that required for the interactions between LT-␤R and its associated kinases, p50 and p80 (15). Moreover, the self-association domain of LT-␤R almost overlaps with a region of LT-␤R, which has been shown to interact with the core antigen of hepatitis C virus (39,41). Nevertheless, the self-association domain of LT-␤R does not show significant sequence similarity with the death domains of TNFRI and Fas; therefore, the self-association domain of LT-␤R represents a novel motif that is essential for receptor self-association, as well as interaction with both cytoplasmic and viral proteins, and that might regulate the receptor signaling for cell death. The identification of proteins capable of interacting with amino acids 324 -377 of LT-␤R will be crucial to understand LT-␤R signaling.