Identification of a New Murine Tumor Necrosis Factor Receptor Locus That Contains Two Novel Murine Receptors for Tumor Necrosis Factor-related Apoptosis-inducing Ligand (TRAIL)*

Tumor necrosis factor (TNF) ligand and receptor superfamily members play critical roles in diverse developmental and pathological settings. In search for novel TNF superfamily members, we identified a murine chromosomal locus that contains three new TNF receptor-related genes. Sequence alignments suggest that the ligand binding regions of these murine TNF receptor homologues, mTNFRH1, -2 and -3, are most homologous to those of the tumor necrosis factor-related apoptosis-inducing ligand (TRAIL) receptors. By using a number of in vitro ligand-receptor binding assays, we demonstrate that mTNFRH1 and -2, but not mTNFRH3, bind murine TRAIL, suggesting that they are indeed TRAIL receptors. This notion is further supported by our demonstration that both mTNFRH1:Fc and mTNFRH2:Fc fusion proteins inhibited mTRAIL-induced apoptosis of Jurkat cells. Unlike the only other known murine TRAIL receptor mTRAILR2, however, neither mTNFRH2 nor mTNFRH3 has a cytoplasmic region containing the well characterized death domain motif. Coupled with our observation that overexpression of mTNFRH1 and -2 in 293T cells neither induces apoptosis nor triggers NFκB activation, we propose that themTnfrh1 and mTnfrh2 genes encode the first described murine decoy receptors for TRAIL, and we renamed themmDcTrailr1 and -r2, respectively. Interestingly, the overall sequence structures of mDcTRAILR1 and -R2 are quite distinct from those of the known human decoy TRAIL receptors, suggesting that the presence of TRAIL decoy receptors represents a more recent evolutionary event.

In many biological systems, cellular outcomes are often determined by environmental cues delivered through ligand and receptor interactions on the cell surface. One group of ligand/ receptor pairings critical to this decision-making process is the tumor necrosis factor (TNF) 1 ligand and receptor superfamily (1). Upon ligand engagement, TNF receptors trigger intracellular signaling pathways that lead to cell proliferation, differentiation, or apoptosis. The pivotal roles of these TNF ligands and receptors across diverse biological areas are perhaps best illustrated by gene knockout studies demonstrating the essential involvement of the lymphotoxin pathway in lymphoorganogenesis (2,3), the BAFF pathway in B-cell development (4), the RANKL pathway in osteoclastogenesis (5), and the EDA pathway in hair-follicle formation (6). The ability of many members of this family to regulate both innate and adaptive immunity also makes them attractive targets for therapeutic intervention of various immune disorders, as exemplified by the success of anti-TNF therapy in treating rheumatoid arthritis and Crohn's disease (7).
TNF receptor family members are characterized by the presence of cysteine-rich repeats (CRDs) in their extracellular domains (8,9). A CRD typically contains two structural motifs, called modules, that are stabilized by disulfide bridges formed between the cysteine residues. The linear arrangement of modules creates a scaffold that supports the unusual elongated structures seen in all known crystal structures of TNFR family members. In contrast to the absolute conservation of CRDs, the signaling potentials of TNF receptors vary a great deal. Whereas most TNF receptors, such as TNFR1, CD40, and Fas, have cytoplasmic domains containing well characterized signaling motifs such as TRAF-binding sites and/or death domain (10,11), others lack signaling capacity. These non-signaling receptors include soluble receptors OPG and DcR3, the GPIanchored human TRAILR3, and human TRAILR4 that contains a defective signaling cytoplasmic tail. The biological function of these so-called "decoy receptors" is likely to antagonize the pairing between ligands and their signaling receptor counterparts, providing a critical mechanism for ligand desensitization (12)(13)(14)(15).
Whereas many members of the TNF ligand superfamily demonstrate one-to-one pairing with their cognate receptors, others exhibit complex ligand/receptor cross-talks (8). In particular, the TNF ligand TRAIL has five receptors in human, at least based on in vitro binding assays (16,17). Among the hTRAIL receptors, hTRAILR1 and -R2 each contain a death domain in the cytoplasmic region, and as a result hTRAIL can efficiently induce caspase-dependent apoptosis in cell lines expressing these receptors (18,19). As mentioned before, hTRAILR3 and -R4 are both considered decoy receptors, but their in vivo function is not clear. The fifth TRAIL-binding TNF receptor is OPG, which also binds to RANKL (17). Whereas studies of OPG knockout mice have clearly demonstrated OPG as a decoy receptor for RANKL, the in vivo relevance of OPG to TRAIL biology remains to be established (17,20). Interestingly, only one murine TRAIL receptor, mTRAILR2/mKiller, has been identified so far (21). Similar to hTRAILR1 and -R2, mTRAILR2 contains a death domain motif and induces apoptosis when overexpressed or engaged by TRAIL.
Known as the TNF receptor "signature," the uniquely spaced cysteine residues found in these receptors allows identification of potential new family members from unprocessed genomic sequences by bioinformatic means. In this study, we describe the identification through genome mining of three new TNFR family members closely clustered on mouse chromosome 7. All three genes, named mTNFRH1, -2 and -3, encode proteins containing classic TNF receptor-like CRDs. We also demonstrate that, whereas mTNFRH3 remains an orphan receptor, mTNFRH1 and two splice variants of mTNFRH2 can specifically bind murine TRAIL, but not the closely related RANKL nor any other ligand we have tested. Both sequence analysis and transient overexpression studies, however, suggest that mTNFRH1 and -2 are not signaling TRAIL receptors but rather the previously unknown murine TRAIL "decoy" receptors. Given their low overall sequence homology to hTRAILR3 and hTRAILR4, we propose that mTNFRH1 and -2 belong to a new class of TRAIL decoy receptors and thus named them mDcTRAILR1 and -R2, respectively. The identification of these two murine TRAIL decoy receptors will likely facilitate our understanding of the complex biology underlying TRAIL ligand/receptor interactions through the generation of mice deficient in these receptors.
Cell Lines-293T cells were grown in DMEM supplemented with 10% heat-inactivated fetal calf serum (FCS). Jurkat cells were maintained in RPMI ϩ 10% FCS and HEK-293 cells in DMEM:F12 ϩ 2% FCS. All media contained 10 g/ml each penicillin and streptomycin. All the cell lines used for Northern analysis were purchased from ATCC and grown in recommended culture media.
Transient transfections in 293T cells for the production of soluble proteins in serum-free Opti-MEM and establishment of stable transfectants in HEK-293 cells were performed as described previously (22).
Identification and Cloning of mDcTRAILR-1, mDcTRAILR-2 and mTNFRH-3 cDNAs-The cDNA of mTNFRH1/mDcTRAILR1 was obtained from EST clones (GenBank TM accession numbers AI156311, AI747041, and BG077775). A full-length coding cDNA was generated from these clones by a PCR-based method and cloned into the PCR-3 mammalian expression vector (Invitrogen).
The full-length cDNA clone of mTNFRH3 was obtained by screening a mouse spleen phage cDNA library (Stratagene) using a partial cDNA probe amplified from the E14 ES cell line. The screening was performed according to recommended protocol from Stratagene and resulted in one cDNA clone from about 1 ϫ 10 6 independent clones.
The cDNAs of both splicing variants of mTNFRH2/mDcTRAILR2 were obtained by RT-PCR using primer sequences designed on genomic sequences. Briefly, total RNA was isolated from NIH3T3 cells using TRIzol (Invitrogen) followed by first strand synthesis using Superscript II (Invitrogen). PCR was performed using the Touchdown protocol. The cDNA of mTRAILR2/mKiller was obtained similarly using RNA from the J1 ES cell line.
Northern Analysis-Tissue expression patterns were done using premade Northern blots from Ambion. For expression patterns in various murine cell lines, total RNA was isolated using TRIzol reagent (Invitrogen), and 20 g of total RNA was loaded for each lane. Probes for each gene were generated using PCR amplification of the coding regions and labeled with [␣-32 P]dCTP using Ready-To-Go DNA labeling beads (Amersham Biosciences). Blots were hybridized and washed in ExpressHyb solution (Clontech) according to the manufacturer's protocol.
Coimmunoprecipitation and ELISA-Receptors:Fc (ϳ500 ng) mixed with FLAG ligands (200 ng) in 1 ml of PBS were incubated for 2 h at 4°C on a wheel with 2.5 l of protein A-Sepharose (Amersham Biosciences). Beads were harvested, loaded in empty mini-columns, washed 3ϫ with 100 volumes of PBS, eluted with 15 l of 0.1 M citrate NaOH, pH 2.7, neutralized, and analyzed by Western blotting with anti-FLAG M2 antibody. Membranes were subsequently reprobed with goat anti-human IgG antibodies.
The interaction between receptor:Fc and FLAG ligands by ELISA was performed as described previously (23). Briefly, ELISA plates were coated with mouse anti-human IgG and then sequentially incubated at 37°C with receptor:Fc, FLAG ligands, biotinylated M2, and horseradish peroxidase-coupled streptavidin. The binding of receptor:Fc was also probed directly with a horseradish peroxidase-coupled rabbit antihuman IgG polyclonal antibody.
For 293T cell death induced by expression of various TNF receptor family members, both adherent and floating cells, were collected 24 -48 h post-transfection and stained with annexin V and 7-AAD (Pharmingen) according to the manufacturer's recommended protocol. Transfected cells were identified as cells expressing GFP. Dead cells were quantified as the percentage of GFP-positive cells that were also positive for 7-AAD staining.
Caspase and NFB Activity Assays-For caspase activity assay, 293T cells (90-mm dish) were transfected with 7 g of indicated plasmids. For NFB assay, 2 ϫ 10 5 293T cells were plated in each well of 24-well plates overnight and then transfected with various amounts of the indicated TNF receptor expressing vectors in triplicate together with 40 ng of an NFB luciferase reporter construct. Luciferase activity was measured 24 h post-transfection using the LucLite luciferase reporter gene assay kit (PerkinElmer Life Sciences).
mTRAIL Binding Assay-293T cells were transfected with various TNF receptor family members using the Polyfect Tranfection (Qiagen) protocol. Briefly, cells were plated at 5 ϫ 10 5 cells/well in 6-well plates for 24 h and then cotransfected with 500 ng of a GFP expression construct (AN050) and various concentrations of expression constructs containing the individual TNF receptor family members. After 24 -48 h, cells were harvested and sequentially stained with FLAG-tagged mTRAIL, an anti-FLAG M2 monoclonal antibody (Sigma, 1:2000), and phycoerythrin-conjugated anti-murine IgG (Jackson ImmunoResearch, 1:200) each for 30 min at 4°C. All cell samples were analyzed on the BD Biosciences FACSCalibur.
To detect the binding of mTRAIL expressed on the cell surface, 293T cells were transfected with either mock vector or full-length murine TRAIL using LipofectAMINE 2000 (Invitrogen) according to the recommended protocol. The cells were harvested 24 h later with 5 mM EDTA in PBS and incubated for 1 h at room temperature with the following murine receptor Fc fusion proteins diluted in FACS buffer: DcTRAILR2 L :Fc (10 g/ml), BCMA:Fc (10 g/ml), TRAILR2:Fc at (10 g/ml), and DcTRAILR2 S :Fc (100 l of tissue culture supernatant). After washing, the cells were incubated with phycoerythrin-labeled goat anti-human IgG:Fc (Jackson ImmunoResearch) at 1:200 dilution for 30 min at room temperature. After final washes, cells were resuspended in 1% paraformaldehyde/PBS and analyzed using the BD Biosciences FACSCalibur.
Homology Modeling and Visualization-Models of mTRAIL trimer (residues 125-291) complexed to a single subunit of mTNFRH3 (residues 41-148), mDcTRAILR1 (residues 52-160), and mDcTRAILR2 (residues 62-170) were built based on the crystal structure of human TRAIL/DR5 (TRAILR2) complex (Protein Data Bank code 1D4V (24)) using the homology modeling module of the InsightII program (version 2000, Accelrys (25)). The alignment of the receptors used for modeling is shown in Fig. 2A. Multiple models generated for each complex were validated using ProsaII program (25), and the ones having lowest the z scores were selected for further analysis. The models were visualized in the MOLMOL program (26).

Identification of the Murine TNFRHs Locus on Chromosome
7F4 -To identify potential new TNF receptor family members, we used a TNFR signature profile (Prf:TNFR_NGFR_2 at www. expasy.ch/cgi-bin/nicedoc.pl?PDOC00561) to search a data base generated by the Swiss Institute of Bioinformatics that predicts proteins from the public genomic data bases. Our initial screening resulted in one TNFR signature-containing hit from the mouse genomic BAC clone RP23-6I17 (GenBank TM accession number AC068006). By using RT-PCR, we were able to confirm that this TNF receptor-like gene was indeed expressed (data not shown). Subsequent determination of its genomic localization revealed a tight linkage to two potentially novel TNF receptor homologous genes, Tnfrh1 and -2, predicted from the genomic sequencing effort on the distal region of the mouse chromosome 7, the murine syntenic region of the Beckwith-Wiedemann syndrome (BWS) region in human ( (27) Gen-Bank TM accession number of the full genomic locus, AJ276505). Because the TNF receptor-like gene we identified is located to the immediate 3Ј of the predicted Tnfrh1 and -2 genes, we named the gene Tnfrh3. Based on the above information, we hypothesized that there exists a previously unknown TNF receptor cluster on the distal region of mouse chromosome 7. We then proceeded to obtain the full-length coding regions of all three Tnfrh genes (see below), and we identified several BAC clones (Resgen) containing the entire Tnfrhs locus. Upon extensive sequencing efforts, we concluded that, sandwiched between the obph1 and car1 genes, the three mTnfrh genes span roughly 100 kb with no other apparent intervening genes, based on analysis using GenScan (Fig. 1A). Because of our functional data (see below), we propose to rename the first two Tnfrh genes mDcTrailr1 and -r2.
Cloning, Sequence Analysis, and Expression Pattern of mTNFRHs-Upon data base search, we found that a full-length cDNA sequence containing the predicted cDNA sequence for mDcTrailr1 had been isolated previously and reported in a patent filing (international publication number, WO 98/43998). We then employed two different approaches to obtain the complete coding sequences for mDcTrailr2 and mTnfrh3. For mTnfrh3, partial cDNA fragment amplified by RT-PCR from total RNA prepared from the mouse ES cell line E14 was used to probe a Stratagene mouse spleen cDNA phage library, resulting in the isolation of a single mTnfrh3 fulllength cDNA clone (GenBank TM accession number AY165628). For mDcTrailr2, the full coding region was obtained by a combination of computer-assisted exon prediction and PCR amplification from 1st strand DNA synthesized from NIH3T3 total RNA, which revealed the presence of two alternatively spliced mDcTrailr2 variants (GenBank TM accession numbers AY165626 and AY165627) (Fig. 1B).
As expected, each mTNFRH contains a signal peptide sequence at the N terminus followed by TNFR-like cysteinerich repeats, consistent with the typical type I membrane protein topology observed in most TNFR family members (Fig. 1B). The structures of their CRDs are remarkably similar, with three tandem A1/B2 domains in both mDcTRAILR1 and mDcTRAILR2, whereas mTNFRH3 has two A1/B2 followed by one A2/B2 domain (8) (Fig. 1C). The C-terminal portions of the mTNFRHs, however, are surprisingly divergent. Whereas mTNFRH3 contains a typical transmembrane domain (TM), the C terminus of mDcTRAILR1 instead exhibits structural features of a GPI anchor addition signal (Fig. 1B) (28). The two splice variants of mDcTRAILR2 also diverge in their C termini. The cDNA spanning exons 1-7 contains a stop codon within exon 6 and encodes a secreted soluble receptor (mDcTRAILR2 S ), whereas the splice variant lacking exon 6 encodes a longer isoform (mDcTRAILR2 L ) containing a TM region and a short intracellular domain. Finally, mDcTRAILR1 and -R2 are highly homologous with 71% identity at the amino acid level (Fig. 1B), indicating a recent gene duplication event at this locus.
We next examined the expression patterns of mTNFRHs in both mouse tissues and cell lines. Based on Northern blot analysis, the expression of mTnfrh3 is primarily restricted to lymphoid tissues with a single ϳ3-kb message detected in both the thymus and spleen and at low but detectable levels in the lung ( Fig. 2A). The lymphoid-specific expression pattern of mTnfrh3 was largely confirmed in Northern analysis of a number of murine cell lines, with its expression detected almost exclusively in lymphoid cell lines, including both the T and B lineages. The only noticeable exception is Colon 26, a colorectal adenocarcinoma cell line that also expressed mTnfrh3 (Fig. 2B). The analysis of expression patterns of mDcTrailr1 and mDcTrailr2, however, is complicated by the extremely high homology between these two genes. Although we could confirm by RT-PCR that both mDcTrailr1 and -2 are indeed expressed genes (data not shown), we were unable to distinguish their individual expression patterns by Northern analysis due to cross-hybridization of the probe to both mRNAs. Instead, the combined expression of mDcTrailr1 and mDc-Trailr2, as determined by the presence of at least one of the mRNA species, could be detected at low levels in all the mouse tissues upon prolonged exposure (data not shown). In contrast, the levels of expression were considerably higher in murine cell lines, and various levels of combined expression can be detected in most of mouse lines we have tested so far without obvious patterns in the tissue/organ origins of the positive cell lines (Fig. 2C).
Murine TNFRH1/DcTRAILR1 and TNFRH2/DcTRAILR2 Bind Specifically to the TNF Ligand TRAIL-The CRDs of TNF receptors not only provide the overall structural scaffold but also determine their ligand binding specificity (8). In an attempt to "deorphanize" mTNFRHs, we performed sequence alignments of their CRD regions with those of the other known TNF receptors. Our analysis revealed significant homologies between mTNFRHs and TRAIL receptors, particularly in regions of the TRAIL receptors that are involved in binding to TRAIL (Fig. 1C), as suggested by the crystallographic studies of the hTRAIL-hTRAILR2 complex (24,29,30).
To examine experimentally whether these novel TNF recep-tors can indeed bind to the TNF ligand TRAIL, we first used an ELISA-based screening assay that has been optimized for the detection of interactions between TNF family ligands and receptors (22,23). As shown in Fig. 3A, both mDcTRAILR1:Fc and mDcTRAILR2 L :Fc fusion proteins bound soluble mouse TRAIL. Interestingly, mDcTRAILR2 also bound soluble human TRAIL but to a lesser degree, whereas mDcTRAILR1 appeared species-specific and did not interact at all with human TRAIL. The interactions between mDcTRAILRs and soluble TRAIL appeared highly specific because no other TNF ligand, includ- ing the closely related RANKL, bound to mDcTRAILR1:Fc and mDcTRAILR2 L :Fc in this assay ( Fig. 3A and data not shown). Although the absolute binding affinities between these ligand/ receptor pairings have not been measured, our observations are nonetheless suggestive of the possibility that mDcTRAILR1 and -R2 may compete effectively against muTRAILR2 for their common ligand TRAIL. Surprisingly, mTNFRH3:Fc did not bind to TRAIL, RANKL, or any other tested TNF ligand, despite its sequence homology to mDcTRAILR1 and -R2 and other TRAIL receptors (Fig. 3A). The ability of mDcTRAILR1 and -R2 to bind murine TRAIL was further examined by immunoprecipitation followed by Western blot. We found that, similar to hTRAILR2:Fc and mTRAILR2:Fc, both mDcTRAILR1:Fc and mDcTRAILR2 L :Fc efficiently precipitated FLAG-tagged murine TRAIL and again demonstrated the strict specificity of mDcTRAILR1 for murine but not human TRAIL (Fig. 3B). The ability of mDcTRAILR1 and -2 to bind TRAIL was further demonstrated using a biological assay in which both mDcTRAILR1:Fc and mDcTRALR2 L :Fc inhibited significantly mTRAIL-induced cytotoxicity in Jurkat cells (Fig. 3C).
We then confirmed the interactions between mTRAIL and mDcTRAILR1 and -R2 by using full-length mDcTRAILR1 and -R2 expressed on the cell surface (Fig. 3D). We found that mTRAIL readily bound to 293T cells expressing mTRAILR2, mDcTRAILR1, and mDcTRAILR2 L . The failure of soluble mTRAIL to bind 293T cells transfected with the mDcTRAILR2 Sexpressing construct, as shown in Fig. 3D, is not surprising because mDcTRAILR2 S lacks the transmembrane domain and does not express on the cell surface. To circumvent this problem, we tested the ability of mDcTRAILR2 S :Fc fusion protein to bind 293T cells expressing full-length mTRAIL on the surface. As shown in Fig. 3E, both mDcTRAILR2 S :Fc and mDcTRAILR2 L :Fc could bind surface mTRAIL. Based on these results, we conclude that mDcTRAILR1, mDcTRAILR2 L , and mDcTRAILR2 S are indeed previously unknown murine TRAIL receptors.
Murine TNFRH-1/DcTRAILR-1 and TNFRH-2/DcTRAILR-2 Are Putative Murine Decoy TRAIL Receptors-Whereas both mDcTRAILR1 and mDcTRAILR2 L appear to be surface receptors for TRAIL, sequence analysis also suggests that they anchor in the cytoplasmic membrane by different mechanisms. The C-terminal sequence of mDcTRAILR1 is a predicted glycosylphosphatidylinositol (GPI) anchor addition signal, whereas mDcTRAILR2 L seems to contain a conventional transmembrane domain with an extremely short cytoplasmic domain. In both cases, the receptors do not seem to possess the ability, upon ligand engagement, to trigger the various intracellular signal transduction pathways involved in TRAIL-mediated cytotoxicity. In contrast, mDcTRAILR2 S is a secreted soluble receptor. To demonstrate formally that mDcTRAILR1, but not mDcTRAILR2 L , is a GPI-anchored receptor, we employed a well established method for assaying GPI anchorage of recep-tors by measuring the sensitivity of surface receptors to phosphatidylinositol-specific phospholipase C (PI-PLC) (31). As expected, surface expression of mDcTRAILR2 L , as determined by  upper panel. B, mDcTRAILR1, mDcTRAILR2 L , and mDcTRAILR2 S do not induce apoptosis mTRAIL staining, was not altered following PI-PLC treatment, indicating that mDcTRAILR2 L anchors in the plasmic membrane through its conventional TM region (Fig. 4A). On the when other hand, an obvious decrease in mTRAIL staining could be detected in mDcTRAILR1-expressing 293 cells treated with PI-PLC (Fig. 4A), consistent with our hypothesis that mDcTRAILR1 is a GPI-linked surface TRAIL receptor.
Signaling TRAIL receptors hTRAILR1, hTRAILR2, and mTRAILR2 all contain the death domain motif in their cytoplasmic domains that is responsible for TRAIL-induced cytotoxicity, and the decoy receptors do not. Overexpression of these death domain-containing receptors can induce apoptosis in a number of cell lines such as 293 and NIH3T3, likely through self-oligomerization of the receptors. We therefore tested the effect of overexpression of mDcTRAILR1, mDcTRAILR2 L , and -R2 S , on 293T survival. As demonstrated in Fig. 4B, overexpression of death domain-containing receptors TRAMP/DR3, hTRAILR2, or mTRAILR2, but not the human decoy receptors hTRAILR3 and -4, resulted in significant cell death in 293T cell as measured by 7-AAD uptake. Transient transfection of mDcTRAILR1 and either the long or short form of mDcTRAILR2 did not have any noticeable effect on the survival of 293T cells, suggesting that mDcTRAILR1 and -R2 do not possess death-signaling capability. Consistent with this hypothesis, overexpression of mDcTRAILR1 and -R2 did not induce caspase activation that is characteristic of death receptor signaling, as demonstrated in Fig. 4C.
Many TNF receptor family members can also activate the NFB pathway, through direct or indirect association with the TNF receptor-associated factor (TRAF) family members. Although mDcTRAILR1 and -R2 do not seem to possess the TRAF-binding motif according to sequence analysis, we nonetheless examined their ability to trigger indirectly NFB activation (Fig. 4D). As reported previously (21), transient transfection of mTRAILR2 induced considerable NFB activation in a dose-dependent manner as measured by cotransfection of an NFB luciferase reporter construct. Overexpression of mDc-TRAILR1 and both the long and short forms of mDcTRAILR2, however, failed to induce NFB activity above that of the empty vector control level, again in agreement with the notion that both mTRAILR1 and -2 lack signaling capability and are likely murine TRAIL decoy receptors.
Previous studies (32) have suggested that decoy receptors could potentially inhibit ligand-induced cytotoxicity by directly competing for the ligand and/or disrupting the proper assembly of the signaling receptors. We have shown in Fig. 3C that apoptosis of Jurkat cells triggered by mTRAIL can indeed be efficiently blocked by both mDcTRAILR1:Fc and mDcTRALR2 L :Fc, supporting our hypothesis that mDcTRAILR1 and mDcTRAILR2 are likely decoy receptors that could function through ligand competition. To examine the latter possibility, we took advantage of the fact that receptor signaling induced by the overexpression of mTRAILR2 mimics that triggered by engagement of mTRAILR2 with TRAIL ligand. Our data clearly indicate that the presence of mDcTRAILR1 or mDcTRAILR2 L has no apparent effect on 293T cell death induced by overexpression of mTRAILR2, suggesting that neither mDcTRAILR1 nor mDcTRAILR2 L interferes with the proper assembly of mTRAILR2 that is required for transducing the apoptotic signaling (Fig. 4D). DISCUSSION In this study, we described the identification and characterization of a new TNF receptor locus where three TNF receptorlike genes, Tnfrh1, -2 and -3, are located in tandem in the distal region of mouse chromosome 7. The close proximity of these three TNF receptor genes is not surprising as clusters of TNF family ligands and receptors can be found in many parts of the mammalian genome. For example, TNF␣, LT␣, and LT␤ are all tightly clustered within the major histocompatibility region on chromosome 17 (mouse) and 6 (human). Similarly, two recently identified TNF ligands Tweak and April are also closely linked on human chromosome 17p13, whereas TL1A and CD30L are on 9q32-33. On the receptor side, TNF-RI, LT␤-R, and CD27 are located at 12p13 (human), whereas TNFR-RII, HVEM, CD30, and OX40 are clustered at 1p36 (human). Even more strikingly, all four known human receptors for TRAIL are found at 8p21-22. The presence of these clusters has led to the suggestion that the expanding TNF family has evolved from a limited number of ancestor molecules by means of gene duplication (1). Our identification of the Tnfrhs locus thus presents yet another example of TNF family clustering. Perhaps more interestingly, the Tnfrh1/mDcTrailr1 and Tnfrh2/mDcTrailr2 genes are extremely homologous with several stretches of genomic sequences that are nearly identical, indicative of a very recent duplicating event and providing by far the most convincing evidence supporting the gene duplication theory.
We also demonstrated experimentally that mTNFRH1 and -2, but not mTNFRH3, are receptors for TRAIL. Based on the sequence characteristics and a number of in vitro studies, we further propose that mTNFRH1 and -2 are in fact murine decoy receptors for mTRAIL, and we renamed them mDcTRAILR1 and mDcTRAILR2, respectively. Although whether these two murine TRAIL receptors function as decoy receptors in vivo remains to be formally established, several lines of evidence presented in this study, in particular their lack of both death signaling and NFB activating capacity when overexpressed, are consistent with this hypothesis. Interestingly, there are considerable differences in the overall sequences between mDcTRAILR1 and -R2 and the previously known human decoy TRAIL receptors hTRAILR3 and -R4. Whereas the human receptors have the N1-A1-B2-A1-B2 cysteine module structure, both mDcTRAILR1 and mDcTRAILR2 have the A1-B2 domain instead of the N1 module. In addition, mDcTRAILR1 and -R2 do not possess either the membrane-proximal TAPE repeats that are characteristic of the huTRAILR3 or the truncated death domain present in huTRAILR4. These important differences suggest that, despite their apparent functional equivalence, the murine and human TRAIL decoy receptors likely have evolved independently. The divergence between mDcTRAILR1/R2 and hTRAILR3/R4 is also underscored by the potential differences in the mechanisms by which they inhibit TRAIL cytotoxicity. It has been reported previously (33) that, under physiological temperature, the affinities of hTRAILR3 and -R4 for hTRAIL are about 50 -100-fold lower than those between hTRAIL and its cognate receptors hTRAILR1 and -R2. The observation suggests that the hTRAILR3 and -R4 would be very poor competitors for hTRAIL against hTRAILR1 and -R2, leading to the hypothesis that the human decoy TRAIL receptors might inhibit hTRAILinduced cytotoxicity through some other mechanism(s), such as by disrupting the proper assembly of the trimeric cognate receptors (33). In our studies, however, we found that coexpression of mDcTRAILR1 or -R2 has no effect on cell death induced by mTRAILR2, arguing against such a scenario. It is therefore possible that the murine decoy receptors may function by a different mechanism than their human counterparts. Whether they can do so through direct competition for the ligand re-quires further investigation of the relative affinities of all murine TRAIL receptors for the TRAIL ligand.
The inability of mTNFRH3 to bind TRAIL is somewhat surprising given its sequence homology to mDcTRAILR1 and -R2. To understand the molecular basis of ligand selectivity exhibited by these receptors, we performed molecular modeling of the trimeric ligand/receptor interfaces between mTRAIL and various mTNFRHs based on the crystal structure of human TRAIL-TRAILR2 complex (Protein Data Bank code 1D4V) (24). As expected from the significant sequence homology in the CRD regions between hTRAILR2 and mDcTRAILR1 and -R2, the predicted interactions between mTRAIL and mDcTRAILR1 or -R2 mirror those between hTRAIL and hTRAILR2 to a large degree ( Fig. 5A and data not shown).
Whereas the overall modeled structure of mTNFRH3 with mTRAIL is similar to that of mTRAIL with mDcTRAILR1 or -R2, we have identified two distinct features in the mTNFRH3 structure that may be responsible for its inability to bind FIG. 5. A, portion of the modeled interface between mTRAIL (pink) and mDcTRAILR1 (green), mDcTRAILR2 (cyan), or mTNFRH-3 (blue). Intramolecular disulfide bonds are labeled yellow and orange for the extra disulfide bond in mTNFRH3. The hypothetic interface between mTRAIL and mTNFRH3 is likely destabilized due to the clustering of positively charged residues (Arg-86 and Arg-119 from mTNFRH3 and Arg-195 from mTRAIL). B, comparative modeling of simulated ligand/receptor interfaces between the CRD2 regions of mDcTRAILR1 or -R2 and human versus murine TRAIL. Potential interactions (both hydrophilic and hydrophobic) are marked in red. The strict species selectivity of mDcTRAILR1 for murine but not human TRAIL is likely due to an essential electrostatic interaction between Asp-89 of mDcTRAILR1 and Lys-163 of muTRAIL that is not present between mDcTRAILR1 and human TRAIL. mTRAIL. Based on the modeling, cysteines 111 and 116 are positioned favorably to form a disulfide bond, resulting in the A1-B2-A2-B2 CRD module arrangement, rather than A1-B2-A1-B2 seen in mDcTRAILR1 or -R2. We also noticed that mTNFRH3 has two arginines in positions 86 and 119 that are not present in mDcTRAILR1 or -R2. When mapped on the hypothetical model of the mTRAIL-mTNFRH3 complex (Fig.  5A), these residues were found to be located proximally to Arg-195 of the mTRAIL, forming an unfavorable interaction between three positively charged residues. Both mDcTRAILR1 and -R2, however, can form a salt bridge with Arg-195 of mTRAIL via Asp-97 (mDcTRAILR1) or Asp-107 (mDcTRAILR2), stabilizing the interaction. The bulky nature of Arg-119 in mTNFRH3 also prevents its interaction with the side chain of Tyr-247 (mTRAIL) and instead promotes its reorientation toward Arg-86 (mTNFRH3) and Arg-195 (mTRAIL), resulting in the clustering of positive charges at the interface between mTNFRH3 and mTRAIL that makes the formation of a tight complex electrostatically unfavorable.
Our modeling further revealed possible structural basis for the strict species selectivity exhibited by mDcTRAILR1, which binds only mouse, but not human, TRAIL. Because mDcTRAILR2 is highly homologous to mDcTRAILR1 yet binds both human and mouse TRAIL, we closely examined the few residues involved in ligand binding that are different between mDcTRAILR1 and -R2, identifying a small cluster of such residues in the CRD2 regions of mDcTRAILR1 and -R2. As shown in Fig. 5B, it appears that the interface between mDcTRAILR1 and mTRAIL relies upon an important electrostatic interaction between Asp-89 of mDcTRAIL1 and the Lys-163 of mTRAIL. This salt bridge cannot occur with hTRAIL because it has the uncharged Ser-159 at this corresponding position. Although mDcTRAILR2 has Ala-99 at this position, which also precludes an electrostatic interaction with either Lys-163 in mTRAIL or Ser-159 in FIG. 6. Summary of the interactions of TRAIL and RANKL with their cognate receptors in the human and murine systems. Cysteine-rich modules in the extracellular region of the receptors are schematized according to Bodmer et al. (8). hTRAILR1, hTRAILR2, and mTRAILR2 contain an intracellular death domain (large gray boxes), and hTRAILR4 contains a truncated death domain. Consensus TRAF6binding sequences are shown as small gray boxes (in hRANK and mRANK) and consensus binding sequences for other TRAFs are shown as small open boxes. hTRAILR3 and mDcTRAILR-1 are GPI-anchored receptors. The two splice variants of mdcTRAILR-2 (long and short) result in membrane-bound or secreted receptors, respectively. mTNFRH3 interacts neither with TRAIL nor with RANKL but is related in sequence to mdcTRAILR1 and -R2. hTRAIL, the interface between mDcTRAILR2 and either human or mouse TRAIL is stabilized by two additional interactions that do not exist between mDcTRAILR1 and the TRAIL ligand. The aspartic acid residue Asp-98 of the mDcTRAILR2, but not the corresponding His-88 in mDcTRAILR1, can form a salt bridge with Arg-158 of the hTRAIL and the corresponding Arg-162 in mTRAIL. Similarly, a favorable hydrophobic interaction exists between the Tyr-225 of human TRAIL (Tyr-216 of mouse TRAIL) and Ile-101 of mDcTRAILR2 but not the charged corresponding residue Glu-98 in mDcTRAILR1.
The locus linkage of mDcTRAIL1 and -R2 to the murine syntenic region of the human BWS locus is quite intriguing. Beckwith-Wiedemann syndrome is a congenital overgrowth syndrome whose pathogenesis has been linked to the abnormal imprint regulation of several candidate genes located at chromosome 11p15.5, a major imprinting cluster in the human genome (34). Given the role of TRAIL in tumor surveillance, it is tempting to speculate that the dysregulated expression of decoy receptors for TRAIL as a result of loss of imprint could potentially contribute to the failure in tumor suppression that is often associated with BWS. Because the two known human TRAIL decoy receptors are not located in the BWS region, the potential presence in the BWS region of human homologs of mDcTRAILR1 and -R2 that might be involved in the molecular pathology of BWS needs to be investigated.
The TRAIL pathway has been a subject of intense research in the past few years, largely due to the ability of TRAIL to preferentially kill tumor cell lines in vitro and in vivo (35)(36)(37). By using both TRAIL-deficient mice and a TRAIL-neutralizing antibody, more recent studies (38 -40) have established that the TRAIL-mediated cytotoxicity pathway is critical for tumor surveillance by the immune system in vivo. Given the presence of both signaling and decoy receptors for TRAIL, it has been hypothesized that the ability of TRAIL to selectively target certain tumor cells is determined by the relative expression levels of these antagonizing receptors (14,41). Repeated efforts, however, have failed to establish a clear link between TRAIL responsiveness and the expression pattern of various TRAIL receptors in many normal and tumor cells (39,42). As a result, the in vivo function of the decoy TRAIL receptors and their relevance to the tumor-suppressing activity of TRAIL demonstrated in TRAIL-deficient mice remains unclear. In this study, we report the identification of putative decoy receptors for TRAIL in the mouse. Our discovery thus reveals that a comparably complex set of interactions between TRAIL ligand and its cognate as well as decoy receptors exists in both human and mouse (Fig. 6). The identification of these mDcTRAILR1 and -R2 will also make it possible to generate mice deficient in these decoy receptors, thus allowing definitive analysis of their possible in vivo contribution in modulating the sensitivity of different cell types to TRAIL.