Osteoprotegerin Is a Receptor for the Cytotoxic Ligand TRAIL*

TRAIL is a tumor necrosis factor-related ligand that induces apoptosis upon binding to its death domain-containing receptors, DR4 and DR5. Two additional TRAIL receptors, TRID/DcR1 and DcR2, lack functional death domains and function as decoy receptors for TRAIL. We have identified a fifth TRAIL receptor, namely osteoprotegerin (OPG), a secreted tumor necrosis factor receptor homologue that inhibits osteoclastogenesis and increases bone density in vivo. OPG-Fc binds TRAIL with an affinity of 3.0 nm, which is slightly weaker than the interaction of TRID-Fc or DR5-Fc with TRAIL. OPG inhibits TRAIL-induced apoptosis of Jurkat cells. Conversely, TRAIL blocks the anti-osteoclastogenic activity of OPG. These data suggest potential cross-regulatory mechanisms by OPG and TRAIL.

TRAIL is a tumor necrosis factor-related ligand that induces apoptosis upon binding to its death domaincontaining receptors, DR4 and DR5. Two additional TRAIL receptors, TRID/DcR1 and DcR2, lack functional death domains and function as decoy receptors for TRAIL. We have identified a fifth TRAIL receptor, namely osteoprotegerin (OPG), a secreted tumor necrosis factor receptor homologue that inhibits osteoclastogenesis and increases bone density in vivo.

OPG-Fc binds TRAIL with an affinity of 3.0 nM, which is slightly weaker than the interaction of TRID-Fc or DR5-Fc with TRAIL. OPG inhibits TRAIL-induced apoptosis of Jurkat cells. Conversely, TRAIL blocks the anti-osteoclastogenic activity of OPG. These data suggest potential cross-regulatory mechanisms by OPG and TRAIL.
Members of the TNF 1 ligand and receptor superfamilies play an important role in regulating many biological processes, including cytokine production, apoptosis, cell activation, lymphocyte co-stimulation, proliferation, immunoglobulin secretion, and isotype switching (1). Three of the ligands, FasL, TNF␣, and TRAIL/Apo-2L, which are type II transmembrane proteins, exhibit potent cytotoxic activity, inducing apoptosis of susceptible cells within hours (2)(3)(4). These cytotoxic ligands bind to and aggregate type I transmembrane receptors with cytoplasmic "death domains," which ultimately activate a protease cascade leading to apoptosis (5).
Recently four receptors for TRAIL have been identified. Two of these receptors, DR4 (6) and DR5/TRICK2/TRAIL-R2 (7)(8)(9)(10)(11)(12), contain cytoplasmic death domains and are capable of inducing apoptosis upon overexpression in transfected cells. Additionally, two decoy receptors, DcR1/TRID/TRAIL-R3 (7-10) and DcR2 (13), lack functional death domains and inhibit TRAILinduced apoptosis. The existence of decoy receptors for TRAIL helps explain the absence of massive apoptosis in cells and tissues that express both TRAIL and one of its cytotoxic receptors (7). However, in tissues lacking the decoy receptor, another mechanism must exist to prevent TRAIL-induced apoptosis.
Osteoprotegerin (OPG) is a secreted TNF receptor family member that increases bone density and causes splenomegaly when overexpressed in transgenic mice (14). OPG inhibits osteoclastogenesis in vitro, suggesting that increased bone density results from decreased numbers of mature osteoclasts. It has been suggested that OPG may neutralize a TNF-related ligand that promotes osteoclast differentiation or bind a membrane-anchored TNF-related ligand that regulates osteoclastogenesis via reverse signaling (14).
To identify interactions between novel TNF ligand and receptor family members, we screened a panel of receptor-Fc fusion proteins for binding to TRAIL and have identified a novel interaction between OPG and TRAIL. We demonstrate that OPG binds to TRAIL in vitro and can block TRAILinduced apoptosis. Conversely, TRAIL appears to block OPGmediated inhibition of osteoclastogenesis in vitro. These results suggest that OPG and TRAIL may function to inhibit each other.
Soluble TRAIL tagged at the N terminus with the FLAG epitope (TRAIL-Flag, amino acids 95-281) (4) was constructed in the pCDN mammalian expression vector (23) that had been modified to contain an in-frame tissue plasminogen activator signal sequence upstream of the FLAG-TRAIL coding region. Soluble TRAIL-Flag was purified from 30 liters of conditioned medium from a Chinese hamster ovary stable transfectant by sequential chromatography on hydroxyapatite (Bio-Rad, type I Macro Prep), Q-Sepharose, Mono Q, and Superdex 200 (Amersham). TRAIL-Flag eluted at ϳ90 kDa, slightly larger than expected for a trimer, but appeared to be predominantly trimeric by analytical ultracentrifugation analysis. 2 Polyclonal antibodies to TRAIL were raised in rabbits by injection of TRAIL (amino acids 41-281) that contained an epitope tag and hexa-* 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  histidine tag at its N terminus. The tagged TRAIL protein was expressed in Escherichia coli as insoluble inclusion bodies using the pAS expression system (24), solubilized in guanidine HCl, and purified by nickel nitrilotriacetic acid chromatography.
Receptor Precipitations-TRAIL-Flag (250 ng) was added to 2 g of receptor-Fc in binding buffer (25 mM HEPES, pH 7.2, 0.25% bovine serum albumin, 0.01% Tween in RPMI 1640) and incubated for 2 h on ice. Protein A-Sepharose 4B (Amersham, 30 l of a 75% slurry) was added and incubated for 1 h on ice. Complexes were recovered by centrifugation, washed three times with binding buffer, electrophoresed on 15% SDS-polyacrylamide gel electrophoresis, and transferred to nitrocellulose for Western analysis with an anti-TRAIL polyclonal antiserum. For binding reactions with full-length TRAIL, the entire TRAIL coding region was polymerase chain reaction-amplified and cloned into pCDN (23). HEK293 cells (3 ϫ 10 6 ) were transfected with 10 g of pCDN-TRAIL or pCDN alone using lipofectAMINE reagent (Life Technologies, Inc.). 48 h after transfection, regular growth medium was replaced with methionine/cysteine-free Dulbecco's modified Eagle's medium. After a 1-h starvation, 1.5 mCi of translabel [ 35 S]Met,Cys (ICN) was added to transfected cells along with 2% dialyzed fetal bovine serum for 6 h. Cells were washed in Dulbecco's modified Eagle's medium and lysed in 1 ml of STET (150 mM NaCl, 10 mM Tris, pH 7.5, 1 mM EDTA, 1% Triton X-100, 10 g/ml leupeptin, 20 g/ml aprotinin, 2 mM phenylmethylsulfonyl fluoride) for 20 min on ice. Lysates were centrifuged, and the supernatants were precleared with 100 l of protein A-Sepharose 4B (Amersham). 50 l of the precleared lysate was used in receptor precipitation with 2 g of OPG-Fc as described above. Binding volumes were increased to 1 ml, and the binding reaction proceeded overnight at 4°C. Competition was achieved with 2 g of unlabeled TRAIL-Flag. Complexes were captured with 30 l of protein A-Sepharose 4B, washed twice in binding buffer, and electrophoresed through 10% SDS-polyacrylamide gel electrophoresis. Gels were fixed for 30 min, incubated with Enlighten (Amersham) for 30 min, and dried before autoradiography.
ELISAs-ELISAs were performed according to standard techniques (25). Immunlon 4 (Dynatech Laboratories) plates were coated overnight at 4°C with anti-Flag M2 monoclonal antibody (Eastman Kodak Co.) (2 g/ml, 100 l/well) in 0.05 M bicarbonate buffer, pH 8.6. After washing and blocking with 0.5% gelatin in phosphate-buffered saline, TRAIL-Flag (2 g/ml, 100 l/well) was bound for 1 h at room temperature. Serial dilutions of OPG-Fc or DR5-Fc (1 g/ml to 5 ng/ml, 100 l/well) were added after washing and incubated for 1 h at room temperature. The Fc fusions were detected with a biotinylated goat anti-human IgG antibody (Southern Biotechnology Associates) (1/2000 dilution, 100 l/ well), horseradish peroxidase-conjugated streptavidin (Southern Biotechnology Associates) (1/4000 dilution, 100 l/well), and ABTS peroxidase substrate (100 l/well, Kirkegaard and Perry Laboratories). The FIG. 1. OPG binds to TRAIL. A, immunoprecipitation of TRAIL-Flag by receptor-Fc proteins. TRAIL-Flag was incubated with the indicated receptor-Fc proteins, and complexes were precipitated with protein A-Sepharose. The samples were analyzed by SDS-polyacrylamide gel electrophoresis and transferred to nitrocellulose for Western analysis with an anti-TRAIL polyclonal antiserum. B, cleaved OPG, TRID, and DR5 compete with Fc fusions for binding to TRAIL. Binding reactions were essentially as described above, except that a 10-fold molar excess of cleaved receptor was added immediately before the Fc fusion. C, OPG-Fc binds to cell-associated full-length TRAIL (FL-TRAIL). 293 cells were transiently transfected with a full-length TRAIL expression construct and labeled with [ 35 S]methionine, and cell lysates were subject to immunoprecipitation. Unlabeled soluble TRAIL-Flag was added as above where indicated. D, OPG-and DR5-Fc bind to TRAIL-Flag in ELISA. Microtiter plates were coated with an anti-Flag M2 monoclonal antibody and allowed to bind TRAIL-Flag. The plates were incubated with serial dilutions of OPG-Fc (Ⅺ) and DR5-Fc (q) proteins, and binding was detected with a biotinylated goat anti-human IgG antibody followed by streptavidin-horseradish peroxidase and ABTS peroxidase substrate. E, OPG and DR5 compete with OPG-Fc for binding to TRAIL-Flag. The assay was performed essentially as described above, except that serial dilutions of cleaved competitor OPG (Ⅺ) or DR5 (q) were incubated with the anti-Flag-captured TRAIL-Flag before addition of OPG-Fc. Samples without competitor are indicated by ‚. s, soluble; KD, kilodaltons. plates were read at 405 nm on a UV max microplate reader (Molecular Devices). For competition assays, serial dilutions of OPG or DR5 without Fc domains (50 g/ml to 30 ng/ml, 100 l/well) were added to the plate for 1 h at room temperature before the addition of OPG-Fc (10 g/ml, 10 l/well).

Determination of the Affinity of TRAIL for Its Receptors-
The association and dissociation rates of the interaction of TRAIL with captured receptor fusion proteins were determined by surface plasmon resonance using a BIAcore 1000 (BIAcore Inc.). The capture surface was a protein A (Pierce) modified CM5 sensor chip (26). The sensor surface was equilibrated with a buffer of 20 mM sodium phosphate, 150 mM sodium chloride, and 0.005% Tween 20, pH 7.4, and analyses were performed at 30 l/min at 25°C. The receptor-Fc fusion protein was diluted into the above buffer to 1 g/ml, and a 20 l injection was passed over the capture surface, followed by a 150-l injection of the TRAIL-Flag. After the association phase, 500 s of dissociation data was collected. The surface was regenerated after each cycle. Sets of 3-4 analyte concentrations, 500 20 nM, were collected and analyzed by nonlinear regression analysis (27) using the BIAevaluation software 2.1. The dissociation data was fitted on the basis of the AB ϭ A ϩ B model. The association data was fitted to A ϩ B ϭ AB using the type 1 model.
Osteoclastogenesis Assays-A modified osteoclast formation assay was adapted from Ref. 28. In brief, bone marrow mononuclear cells were isolated from femurs of 4 -6-week DBF1 mice in the presence of early passage stromal cells obtained from a human osteoclastoma at a 400:1 ratio. The co-culture was established in RPMI 1640 supplemented with 10% fetal calf serum, 100 nM dexamethasone, 10 nM vitamin D3. Test reagents were added at day 0, and the cultures were left undisturbed for 7 days, after which the media were removed, and the cell monolayer was stained for tartrate-resistant acid phosphatase (TRAP) or by a TRAP-solution assay (29). In the TRAP-solution assay, enzyme activity was measured by the conversion of p-nitrophenylphosphate (20 nM) to p-nitrophenol in the presence of 80 mM sodium tartrate and expressed as OD405 units.

RESULTS AND DISCUSSION
To identify interactions between novel TNF ligand and receptor family members, we screened a panel of receptor-Fc fusion proteins for binding to TRAIL by immunoprecipitation. TRAIL bound to TRID-Fc and DR5-Fc proteins as expected (Fig. 1A). TRAIL also bound to OPG-Fc. TRAIL did not bind to DR3-Fc and HVEM-Fc (Fig. 1A), two additional members of the TNF receptor superfamily, indicating that the OPG-TRAIL interaction is not Fc-mediated. To confirm that binding of OPG-Fc to TRAIL was not to the Fc region, a 10-fold molar excess of OPG cleaved of its Fc domain was incubated with TRAIL and the Fc fusion proteins before immunoprecipitation. Cleaved OPG, TRID, and DR5 effectively competed away the binding of OPG-Fc to TRAIL (Fig. 1B). To determine if OPG could bind to cell-associated TRAIL, HEK293 cells were transiently transfected with a full-length TRAIL expression vector, labeled with [ 35 S]methionine, and the cell lysate was subjected to immunoprecipitation by the OPG-Fc fusion protein. OPG-Fc precipitated a protein of the expected size in TRAIL but not vector-transfected cells (Fig. 1C), indicating that OPG-Fc binds to cell-associated TRAIL. Unlabeled soluble TRAIL-Flag effectively competed with the 35 S-labeled full-length TRAIL for binding to OPG-Fc.
The ability of TRAIL to bind OPG was also examined by ELISA. Both OPG-Fc and DR5-Fc proteins were capable of binding to anti-Flag-immobilized TRAIL-Flag (Fig. 1D). In blocking assays, cleaved DR5 prevented OPG-Fc from binding to TRAIL-Flag (Fig. 1E), indicating that the DR5 and OPG binding sites overlap. Cleaved OPG was more effective than cleaved DR5 in blocking TRAIL-Flag binding (Fig. 1E), possibly reflecting the fact that the cleaved OPG is a dimer, whereas cleaved DR5 is a monomer (Ref. 14 and data not shown). Although overlapping, the binding sites of OPG-Fc and DR5-Fc are not identical, since only OPG-Fc recognized directly immobilized TRAIL-Flag (i.e. without an anti-Flag antibody, data not shown).
The affinity of each receptor for TRAIL was determined by measuring the kinetics of TRAIL binding to each by surface plasmon resonance (Table I). OPG-Fc binds to TRAIL-Flag with an affinity of 3.0 nM, which is slightly weaker than the affinities of TRID-Fc and DR5-Fc for TRAIL-Flag (1.1 nM and 0.76 nM, respectively). The affinities of each receptor are within the range of affinities reported for other physiologically relevant TNF-like ligand and receptor pairs (30 -34).
The observation that OPG binds TRAIL in vitro suggests that OPG may inhibit the cytotoxic activity of TRAIL. To examine this possibility, Jurkat cells were treated with anti-Flagaggregated TRAIL-Flag in the presence or absence of OPG.
Anti-Flag-aggregated TRAIL-Flag exhibited significant cytotoxicity within 3 h ( Fig. 2A). OPG demonstrated a dose-dependent inhibition of TRAIL-induced cytotoxicity and blocked the appearance of apoptotic nuclei in Jurkat cells treated with TRAIL-Flag (Fig. 2, D and E). However, OPG did not prevent killing by an agonist anti-Fas monoclonal antibody, another potent inducer of apoptosis, indicating that OPG does not activate a nonspecific survival pathway ( Fig. 2A). DR5-Fc exhibited a similar but more potent inhibition of TRAIL-induced cytotoxicity (Fig. 2B), likely due to its higher affinity for TRAIL. DR3-Fc, which does not bind TRAIL, did not inhibit TRAIL-induced killing (Fig. 2C). This data suggests that OPG, which is secreted and detectable in the circulation (14), 3 may be a soluble antagonist receptor for TRAIL. OPG may negatively regulate TRAIL activity, similar to the inhibition of TNF␣ by secreted TNF receptors of poxviruses (35). Furthermore, TNF receptors are shed from the cell surface and found in the circulation following certain stimuli. It has been suggested that soluble TNF receptors may inhibit or clear circulating TNF␣ (36). A similar role for OPG might also underlie the splenomegaly in mice that overexpress OPG (14). In mice lacking the anti-apoptotic Bcl-2 gene, there is a decrease in the size of the spleen, suggesting that spleen cell homeostasis is regulated by apoptotic mechanisms (37). Thus, if TRAIL is involved in regulation of spleen cell homeostasis by inducing apoptosis, then constitutive overexpression of OPG might result in decreased apoptosis and splenomegaly.
OPG inhibits osteoclastogenesis in vitro (14). To determine the significance of the OPG-TRAIL interaction in osteoclastogenesis, we examined the effects of TRAIL in an in vitro osteoclastogenesis assay. Human stromal cells were co-cultured with murine bone marrow cells for 7 days in the presence of OPG, TRAIL, or TL4, another TNF-like ligand that does not bind OPG. 4 OPG inhibited the formation of multinucleate TRAP positive osteoclasts as expected, with an IC 50 of approximately 60 pM (Fig. 3). TRAIL alone had no effect on osteoclastogenesis. However, TRAIL completely blocked the inhibitory effect of OPG (Fig. 3), suggesting that soluble TRAIL may function to regulate the activity of OPG. TL4 had no effect on osteoclastogenesis in the presence or absence of OPG.
It has been suggested that OPG may inhibit osteoclastogenesis by binding to a pro-osteoclastogenic TNF-related ligand (14). Our data suggest that this ligand may not be TRAIL. First, soluble TRAIL alone had no effect on osteoclastogenesis, and the IC 50 of OPG is approximately 50-fold lower than the affinity of OPG for soluble TRAIL. Secondly, anti-Flag-aggregated TRAIL, which may mimic membrane TRAIL, induced massive apoptosis in these co-cultures, which is the opposite activity to what one would expect for the OPG ligand (data not shown). However, further experiments will be necessary to fully define the role of TRAIL in osteoclastogenesis and the existence of other OPG ligands.
The identification of a fifth receptor for TRAIL suggests that complex regulatory mechanisms control its activity. A soluble antagonist receptor, OPG, and two decoy receptors, DcR1/ TRID/TRAIL-R3 and DcR2, may prevent the transduction of apoptotic signaling through the death receptors DR4 and DR5/ TRICK2/TRAIL-R2. It has also been suggested that the generation of two DR5/TRICK2/TRAIL-R2 isoforms by alternate splicing may regulate cellular responses to TRAIL (11). Addi- tionally, secreted TRAIL itself may antagonize OPG activity or prevent signaling by membrane-anchored TRAIL through its death receptors, as has been described for FasL and TNF␣ (reviewed in Ref. 3). A complete understanding of these regulatory circuits awaits the determination of soluble TRAIL and OPG levels in vivo and an analysis of knockout mutants of TRAIL and its receptors.