INTRODUCTION

The FasL-Fas receptor interaction is a key physiological regulator of programmed cell death (1). Stimulation of the Fas receptor by FasL or agonist antibody results in the rapid induction of apoptosis which is dependent upon a signal transduction cascade involving the caspase family of proteases and is independent of new RNA or protein synthesis (2-5). The Fas receptor cell death signal is antagonized by the cowpox virus protein crmA and the baculovirus p35 protein, as well as aspartic acid containing tri- and tetrapeptides, which specifically inhibit proteases in the caspase family (3, 6, 7). Apoptosis induced by the Fas receptor is characterized by cell shrinkage, plasma membrane blebbing, and DNA fragmentation into nucleosomal sized fragments.

Analysis of lpr, lpr^cg, and gld mice, which have mutations in the Fas or FasL locus, implicate the Fas receptor-ligand system as a critical regulator of the peripheral T cell compartment (1, 8-13). It has been observed that T cell hybridoma lines can be induced to undergo activation-induced cell death when appropriately stimulated by anti-T cell receptor antibodies. This has been used as an in vitro model for the peripheral deletion of lymphocytes and is dependent on interactions between Fas receptor and its ligand (14-16). Defects in this signaling system predispose mice to accelerated autoimmunity characterized by massive lymphadenopathy and splenomegaly, autoantibody production, elevated levels of CD4-CD8 double-negative peripheral T cells, glomerulonephritis, and arthritis (17).

In addition to its role in regulating peripheral lymphoid populations, the Fas receptor-ligand system has also been implicated in maintaining immunoprivilege in the eye and testis, attenuating immunosurveillance against certain types of tumors, such as melanoma and hepatocellular carcinoma, and mediating perforin-independent cytotoxic T lymphocyte target cell killing (18-24). The Fas-FasL system may also be a mediator of Hashimoto's thyroiditis, viral hepatitis, and autoimmune diabetes (25-28).

Much of the current understanding of Fas receptor cell signaling has been derived from the use of agonist antibodies (29-31). Yet there is very little information concerning the requirements for FasL binding to its receptor. The FasL belongs to the TNF¹ family of type II transmembrane proteins, which includes TNF-, lymphotoxin-, CD40 ligand, and CD30 ligand (32, 33). Each family member displays 15-35% sequence identity in the COOH-terminal extracellular region, but does not share sequence similarity in their transmembrane or intracellular domains. The three-dimensional x-ray crystallographic structures of TNF-, lymphotoxin-, and CD40 ligand have been reported (34-37). Soluble forms of each ligand appear to form trimeric complexes with intersubunit interactions stabilized by noncovalent forces.

To study the structure and activity of the human FasL, expression of active protein was necessary. We have transfected FasL cDNAs into heterologous cells and used assays to detect binding of the transiently expressed ligand to the Fas receptor. The biological activity of the FasL was directly monitored after introduction into a Fas-sensitive cell line. Using a series of deletion and single amino acid mutations of the FasL, we have defined an essential receptor binding motif at the COOH terminus of the ligand and a self-association motif located at the NH₂ terminus of the extracellular domain. The results suggest that self-assembly and receptor binding are properties of the ligand which require distinct protein sequences.

MATERIALS AND METHODS

The human FasL cDNA (cloned by KBE) in pBluescript SK (Stratagene) was excised using KpnI and BamHI and ligated into the mammalian expression vector pCDNA3 (Invitrogen). The resulting pCDNA3-hFasL plasmid was purified using the Qtip500 system (Qiagen).

All mutant human FasL mammalian expression constructs were prepared using a PCR-based approach. A common 5-PCR primer (5-catggttctggttgccttggtaggattg-3) was used in conjunction with unique 3-PCR primers for each particular mutant (gld 5-gcggatccttagagcttatataagccgaaaagcgtctg-3; C21, 5-gcggatccttagttgacatataaatgatcagcact-3; C30 5-gcggatccttaaagattgaacactgccccca-3; C52 5-gcggatccttacatcttcccctccatcat-3; C67 5-gcggatccttacctcatgtagaccttgtg; C80 5-gcggatccttaagattgaccccggaagta-3; C98 5-gcggatccttacacaaggccacccttctt-3; and C130 5-gcggatccttagcctgttaaaatgggccacttt-3) and the pCDNA3-hFasL cDNA was used as a template; for the C3 PCR, the 5 (5-ccatttaacaggcaagtccaa-3) and 3 (5-gctctagattataagccgaaaaacgtctg-3) PCR primers were used.

All PCR products, except C3, were cut with StyI and BamHI and ligated into StyI and BamHI digested pBluescript SK-hFasL. The C3 PCR product was digested with BstXI and XbaI and ligated directly into BstXI- and XbaI-digested pCDNA3-hFasL. Each of the mutant human FasL cDNAs in pBluescript SK was isolated by digestion with KpnI and XbaI, introduced into KpnI and XbaI-digested pCDNA3, and verified by DNA sequencing (Cornell University, Ithaca, NY).

The mammalian expression vector pCDNA3 was modified to create a new vector encoding the mouse Ig light chain signal peptide called pCDNA3-IgLCSP (38). PCR was used to amplify the coding region for residues 103-281 of the human FasL. The resulting PCR product was isolated and ligated into the BamHI and EcoRI sites of pCDNA3-IgLCSP to generate pCDNA3-IgLCSP-hFasL. This construct was used to transiently transfect COS-1 cells. Supernatants were confirmed to harbor active soluble FasL by bioassay and Western blot analysis (see below).

Soluble human FasL was prepared using the Celltech glutamine synthetase selection system (Celltech, United Kingdom) (39). The signal peptide and human FasL coding fragment was isolated from pCDNA3-IgLCSP-hFasL and subcloned into pEE14 to create pEE14-hFasL, which was used to transfect CHO-K1. Stable cell lines were made according to the manufacturer's recommendations and screened for secretion of soluble human FasL by bioassay on A20 cells and Western analysis. A single line was expanded and adapted to spinner culture. CHO-K1 cell supernatants were harvested after 48 h and passed through a 0.45-µm filter.

All human FasL N-glycosylation mutants were prepared using a hybrid PCR-based approach to mutate specific asparagine residue(s) to glutamine(s). To generate the singly mutated constructs N184Q, N250Q, and N260Q, the following primers were used: primer 1 (5-ccatttaacaggcaagtccaa-3), primer 2 (5-gctctagattagagcttatataagccg-3), primer 3 (5-tggccttgtgatccaagaaactgggct-3), primer 4 (5-agcccagtttcttggatcacaaggcca-3), primer 5 (5-ggcagtgttccaacttaccagtgc-3), primer 6 (5-gcactggtaagttggaacactgcc-3), primer 7 (5-tcatttatatgtccaagtatctgagctctc-3), and primer 8 (5-gagagctcagatacttggacatataaatga-3). The primers were used in pairs for PCR reactions as follows: A (primers 1 and 4), B (primers 2 and 3), C (primers 1 and 6), D (primers 2 and 5), E (primers 1 and 8), and F (primers 2 and 7). The overlapping PCR products generated using primer pairs A and B or C and D or E and F were mixed, denatured, and annealed. The annealed products were extended with Klenow enzyme in the presence of dNTPs. The extended products were used as templates for a second round of PCRs using primer pair G (primers 1 and 2). The BstXI- and XbaI-digested PCR fragments were ligated into BstXI- and XbaI-digested pCDNA3-hFasL. Plasmids containing the PCR products were isolated and analyzed as described above.

To generate the triple N-glycosylation mutant, in which all three asparagines were mutated to glutamines, the N184Q mutant was used as a template in place of the wild-type human FasL cDNA to generate the N184Q,N250Q double mutant as described above. To generate the N184Q,N250Q,N260Q triple mutant, the N184Q,N250Q mutant was used as a template for PCR as described above. All amplified DNA products were verified by automated DNA sequencing.

All cell lines were cultured at 37 °C in 5% CO₂. COS-1 and 293 cells were cultured in Dulbecco's modified Eagle's medium + 10% fetal calf serum supplemented with penicillin-streptomycin and Fungizone (Life Technologies, Inc.). CHO-K1 cells were cultured in Glasglow minimum essential medium + 10% fetal calf serum (dialyzed) with penicillin-streptomycin and 50 µM methionine sulfoximine (Sigma) and without L-glutamine for glutamine synthetase-based selection. For the production of soluble human FasL, the cells were adapted to CHO-SFMII serum-free medium (Life Technologies, Inc.) in spinner culture, according to the manufacturer's specifications.

All transfections were performed using the calcium phosphate method. COS-1 cells were transfected in 10-cm tissue culture dishes using 20 µg of plasmid DNA; and 293 cells were transfected in 35-mm tissue culture dishes using 1.5 or 2 µg of plasmid DNA as indicated.

Approximately 5-10 × 10⁶ untransfected or transfected COS-1 cells were lysed in 0.5 ml of Nonidet P-40 lysis buffer (1% Nonidet P-40, 20 mM Tris, pH 8.0, 200 mM NaCl, 1 mM EDTA, 2 µg/ml aprotinin, 1 µg/ml leupeptin, and 25 µg/ml phenylmethylsulfonyl fluoride). Cleared lysates were then incubated with the -FasL G-247 (1 µg/ml; Pharmingen) or NOK-1 (1 µg/ml; Pharmingen), -FLAG (2 µg/ml; Eastman Kodak Co.), or -HA (2 µg/ml; Boehringer Mannheim) monoclonal antibody and Sepharose-protein A. The matrix was then washed and immune complexes subjected to Western analysis (see below).

Samples in SDS-PAGE sample buffer were resolved on a 12% SDS-polyacrylamide gel under reducing conditions unless otherwise indicated. Proteins were transferred to nitrocellulose and after blocking incubated with a primary antibody (-FasL G-247, -FLAG, or -HA) at room temperature. The membrane was washed with TBST and then incubated with an -mouse IgG-horseradish peroxidase (Sigma). The membrane was then processed by ECL (Amersham) and exposed to x-ray film.

Approximately 5-10 × 10⁶ untransfected or transfected COS-1 cells were resuspended in 1.0 ml of ice-cold phosphate-buffered saline containing 0.5 mg/ml sulfo-NHS-LC-Biotin (Pierce) and incubated for 30 min at 4 °C. The cells were washed once with phosphate-buffered saline, resuspended in 1.0 ml of Dulbecco's modified Eagle's medium, and incubated on ice for 10 min to quench any residual activated biotin. Cells were lysed using biotinylation lysis buffer (1% Triton X-100, 20 mM Tris, pH 8.0, 150 mM NaCl, 0.2% bovine serum albumin, 5 mM EDTA, 2 µg/ml aprotinin, 1 µg/ml leupeptin, and 25 µg/ml phenylmethylsulfonyl fluoride). Cleared lysates were incubated with streptavidin-agarose (Sigma) to harvest biotinylated protein. The matrix was washed with biotinylation lysis buffer and subjected to Western analysis using the -FasL G-247 antibody.

A human F_c-Fas receptor fusion protein (a gift from C. Smith, Immunex) was used to assess ligand binding to the Fas receptor. Untransfected or transfected COS-1 cells were lysed in 0.5 ml of Nonidet P-40 lysis buffer. Cleared lysates were incubated with the F_c-Fas receptor (1 µg/ml) for 1.5 h at 4 °C followed by Sepharose-protein A for an additional 1 h. The matrix was then washed with Nonidet P-40 lysis buffer and bound ligand detected by Western analysis using the -FasL G-247 antibody.

Mouse -human IgG (F_c-specific) which was adsorbed against mouse serum proteins was purchased from Jackson Immunochemicals and coupled to Affi-Gel-10 (Bio-Rad). This matrix was used to specifically precipitate human F_c Fas receptor from solution with any bound soluble human FasL.

Either 100 µl of CHO-K1 supernatant containing soluble human FasL or CHO-SFMII was diluted to 500 µl with Nonidet P-40 lysis buffer and incubated with or without 1 µg/ml -FasL NOK-1 for 2 h at 4 °C. The F_c-Fas receptor was added to a final concentration of 1 µg/ml and incubated for 1.5 h at 4 °C followed by the addition of 5 µl of the mouse -human IgG (F_c-specific) matrix. The matrix was washed with Nonidet P-40 lysis buffer and F_c-receptor bound ligand was detected by Western analysis using the -FasL G-247 antibody.

Human embryonic kidney 293 cells were transfected in 35-mm tissue culture plates with 2 µg of pCDNA3-hFasL (wild-type or mutant), pCDNA3, or pRC-LacZ as described above. After transfection, the growth medium was replaced with or without 20 µM z-VAD-fmk (Kamiya Biomedical). The transfected cells were analyzed 24 h post-transfection and viability assessed by morphology. pRC-LacZ was used to assess transfection efficiency, which was >50% for all transfections.

Toxicity was also assessed using a quantitative liquid -galactosidase assay after 293 cell cotransfection with 1 µg of wild-type or mutant FasL expression plasmid and 0.5 µg of pRC-LacZ. Since dying 293 cells detach from the tissue culture dish, -galactosidase activity assayed 24 h post-transfection correlated well with the number of viable transfected cells.

To assess whether individual mutant forms of the human FasL were capable of self-assembly, a coprecipitation protocol using specific epitope tags was used. Each mutant was subcloned into pFLAG-CMV-2 (Kodak) or pCDNA3-HA. The pCDNA3-HA mammalian expression plasmid was generated by annealing complimentary oligonucleotides encoding the HA-epitope followed by subcloning into pCDNA3.

The amino-terminal FLAG- and HA-tagged forms of each human FasL mutant were coexpressed in COS-1 cells. After 24-48 h, the cells were lysed in Nonidet P-40 lysis buffer. Cleared lysates were then incubated with either -FLAG or -HA antibody followed by Sepharose-protein A. Immune complexes were prepared for Western analysis as described above.

In addition to analysis of oligomerization status by coimmunoprecipitation, COS-1 expressed FLAG-tagged FasL was subjected to analysis under nonreducing conditions (40). Cell lysates were prepared from transfected cells as described above and incubated with Sepharose--FLAG resin (Kodak). FLAG-tagged protein was isolated using the FLAG peptide (250 µg/ml; Kodak) and subjected to 7% SDS-PAGE under reducing and nonreducing conditions followed by Western analysis using the -FLAG antibody.

RESULTS

The human FasL is synthesized as a 281-amino acid transmembrane protein. A soluble form of this ligand containing most of the extracellular sequence binds the Fas receptor, forms stable trimers under physiological conditions, and retains biological activity (40). The structural requirements for FasL binding to the Fas receptor and self-association have not been defined.

Structure-function analysis was performed using a series of COOH-terminal deletion mutants (C) based upon predicted loops and -stranded segments in the COOH-terminal 140-150 amino acid region of the mouse FasL (41). The mouse and human protein sequences for this ligand are 85% identical in their extracellular domains. The deletion mutants, with the exception of the C3 mutant, remove successive secondary structural motifs proposed for the human FasL extracellular domain. A mutation at position 275 in the human FasL (phenylalanine to leucine), the equivalent of which leads to lymphoproliferative disease in mice, was constructed in the human FasL, as shown in Fig. 1.

[View Larger Version of this Image (20K GIF file)]

Three potential N-glycosylation sites exist in the human FasL at asparagine residues 184, 250, and 260 (see Fig. 1). To assess the contribution of N-linked glycosylation, each site was mutated from asparagine to glutamine, either singly or in combination, to create FasL molecules defective in N-linked glycosylation.

To assess the expression of FasL, each mutant construct was transfected into COS-1 cells using the mammalian expression vector pCDNA3. Lysates from untransfected and transfected cells were subjected to immunoprecipitation with the -human FasL G-247 monoclonal antibody, which binds specifically to a region between residues 103 and 136,² followed by Western blotting with the same antibody (Fig. 2A). The wild-type ligand was found to migrate with a molecular mass of ~40 kDa, similar in size to the human gld mutant protein. Each of the deletion mutants was efficiently expressed and migrated with the expected molecular mass. The observed heterogeneity in size is most likely due to differential glycosylation.

[View Larger Version of this Image (42K GIF file)]

To determine whether the human FasL mutants were expressed on the cell surface, cell surface biotinylation was performed using sulfo-NHS-LC-biotin, a plasma membrane impermeable form of activated biotin. Biotinylated proteins were isolated from cell lysates using streptavidin-agarose and subjected to Western analysis using the -FasL G-247 antibody. The results shown in Fig. 2B confirm that each mutant was expressed appropriately at the cell surface at a level comparable to the wild-type human FasL. Therefore, removal of COOH-terminal sequences from FasL did not appear to interfere with biosynthesis and transport of the protein to the cell surface.

Binding studies of FasL to its receptor have been hampered by the difficulty in generating sufficient quantities of active soluble ligand. COS-1 cells efficiently produce the wild-type or mutant human FasL after transfection. Expression of ligand was confirmed by immunoprecipitation and Western analysis (Fig. 2A). An F_c-Fas receptor fusion protein containing the extracellular domain of the receptor has been used to bind and precipitate FasL from cell lysates and supernatants (32, 40, 42). Therefore, the F_c-Fas receptor fusion protein was used to assay the binding of the human FasL mutants to its receptor. After incubation of lysates with the F_c-Fas receptor, ligand-receptor complexes were precipitated using Sepharose-protein A. Detection of bound ligand was carried out by Western analysis.

The results for the wild-type, gld, and C3 FasLs are shown in Fig. 3. While the wild-type FasL was precipitated by the F_c-Fas receptor in a quantitative manner, only a fraction of the gld FasL was precipitated. Significantly, none of the C3 FasL mutant was precipitated with the F_c-Fas receptor fusion protein. Consistent with this observation, the larger deletion mutants were not precipitated using the F_c-Fas receptor fusion protein (Table I), even though they were fully expressed. These results imply that COOH-terminal amino acid sequences of the human FasL are necessary for binding to the Fas receptor, since removal of as few as 3 amino acid residues from this region completely disrupted receptor binding.

[View Larger Version of this Image (32K GIF file)]

Table I. Summary of properties of human Fas ligand mutants Receptor binding Self-association Biological activity Wild-type Fas Ligand +++ +++ +++ gld mutant + +++   C3   +++   C21   +++   C30   +++   C52   +++   C67   +++   C80   +++   C98   +++   C130

The binding of the gld and C3 FasL proteins to the receptor was drastically reduced. To test further the involvement of the FasL COOH-terminal region in receptor binding, the NOK-1 monoclonal antibody was used. This antibody was identified after immunization of mice with a FasL-expressing human T cell lymphoma line (43). NOK-1 is capable of quantitatively immunoprecipitating the wild-type human FasL from cell lysates and is effective at antagonizing the killing activity of the ligand.

To define the epitope recognized by NOK-1, COS-1 cells were transfected with wild-type or mutant FasL expression constructs. Equal amounts of cell lysate were subjected to immunoprecipitation with either the G-247 or NOK-1 -FasL antibody. Immune complexes were then subjected to Western analysis using the -FasL G-247 antibody (Fig. 4, A and B). Abundant expression of the wild-type, gld, or C3 FasL protein was observed in each transfected cell lysate. However, while NOK-1 was capable of immunoprecipitating the wild-type FasL protein, the antibody was not capable of immunoprecipitating the gld or C3 mutant. Thus, the epitope for NOK-1 appears to reside at the very COOH terminus. Consistent with these results, NOK-1 did not recognize any of the larger human FasL deletion mutants (data not shown).

[View Larger Version of this Image (27K GIF file)]

Since the NOK-1 monoclonal antibody recognized a region implicated in Fas receptor binding, it seemed likely that NOK-1's neutralizing activity resulted from direct competition with the Fas receptor. To assess this possibility, a competition assay was used. For this assay, a soluble, active version of the human FasL, was prepared from CHO-K1 cells. The supernatant containing soluble ligand was preincubated with or without NOK-1. Then, F_c-Fas receptor fusion protein was added to the ligand and precipitated using Affi-Gel 10 mouse -human IgG. This resin was human F_c-specific and did not recognize the NOK-1 antibody. The precipitated material was subjected to Western analysis using the -FasL G-247 antibody. The results demonstrate that NOK-1 directly competed with the Fas receptor for human FasL (Fig. 4B). Hence, the antibody exerts its neutralizing activity by blocking binding of the ligand to its receptor. These results are also consistent with the conclusion that COOH-terminal sequences of FasL are necessary for receptor binding.

Since the primary function of the FasL is to initiate signal transduction events leading to programmed cell death, the biological activities of the mutants were assessed. A human embryonic kidney 293 cell line was ideally suited for this purpose since these cells were found to express the Fas receptor.³ When cells were transfected with a wild-type FasL expression construct, cell death was apparent as early as 6 h post-transfection and was complete by 18-24 h (Fig. 5A). Cell death was further confirmed using a quantitative -galactosidase assay after coexpression of wild-type or mutant FasL with LacZ in 293 cells (Fig. 5B). While the wild-type FasL effectively killed 293 cells after transfection, neither the gld nor the C3 mutant had any toxic activity. Viability was not affected by introduction of the vector pCDNA3 or pRC-LacZ alone. The cell death potential of the other deletion mutants was also assayed (see Fig. 1) and each ligand deletion was found to be biologically inactive (Table I). The requirement for the COOH terminus of the FasL for Fas-mediated cell death is consistent with its necessity in receptor binding.

[View Larger Version of this Image (56K GIF file)]

To confirm that the mechanism of cell death was apoptotic in nature, the pan-caspase inhibitor z-VAD-fmk was incubated with 293 cells immediately after transfection. Proteases of the caspase class were demonstrated to be absolutely required for Fas receptor-mediated programmed cell death (3, 6). The z-VAD-fmk peptide was found to completely inhibit human FasL-mediated cell death at concentrations as low as 20 µM (Fig. 5A).

TNF family members are believed to exist in a trimeric form (44). To address whether the gld and C3 FasLs exist in a multimeric state, the wild-type, gld, or C3 FasL cDNA constructs were subcloned into the pFLAG-CMV-2 and pCDNA3-HA mammalian expression vectors (see "Materials and Methods") and cotransfected into COS-1 cells. Expression of each FLAG- and HA-tagged protein was confirmed by immunoprecipitation and Western analysis using -FLAG and -HA antibodies (Fig. 6A, upper and middle panels).

Self-association of the wild-type, gld, and C3 human FasLs.

[View Larger Version of this Image (22K GIF file)]

Lysates from the co-transfected COS-1 cells were also subjected to immunoprecipitation with the -FLAG antibody and Western analysis using -HA (Fig. 6A, lower panel). The presence of HA-tagged FasL after immunoprecipitation with -FLAG indicated that a stable complex formed between the HA- and FLAG-tagged forms. The results indicate that the gld and C3 human FasL mutants are capable of self-associating to the same extent as the wild-type FasL.

Unlike TNF-, FasL has been demonstrated to be relatively insensitive to detergents (40). SDS-PAGE under nonreducing conditions followed by Western analysis has been used to detect dimeric and trimeric FasL. Ligand from COS-1 cells expressing FLAG-tagged wild-type, gld, or C3 FasL was isolated as described under "Materials and Methods" and analyzed under reducing and nonreducing conditions for dimer and trimer formation (Fig. 6B). Each form of the ligand showed the characteristic pattern of monomer, dimer, and trimer under nonreducing conditions, with the C3 mutant yielding significantly more trimer. Under reducing conditions, only the monomeric form of each was apparent. Consistent with these results, the mouse gld mutant has also been demonstrated to trimerize using the same assay (45). Thus, the observed defect in F_c-Fas receptor fusion protein binding to the gld and C3 FasL mutants is not a consequence of impaired ligand oligomerization, as no significant alterations in quaternary structure can be detected using the above assays.

Since the gld and C3 mutants retain the ability to self-associate upon expression in COS-1 cells, the other human FasL deletion mutants were assayed by coimmunoprecipitation analysis. Lysates were prepared from COS-1 cells cotransfected with all the mutant constructs in FLAG- and HA-tag expression vectors. Expression of FLAG- and HA-tagged ligand was confirmed by immunoprecipitation and Western analysis using equivalent amounts of cell lysates (Fig. 7A, upper and middle panels). In addition, -FLAG antibody precipitated protein was subjected to Western analysis using -HA as the primary antibody to determine if oligomerization occurred (Fig. 7A, lower panel). The C80 and C98 mutants were found to self-associate. Similar results were obtained with the shorter deletion mutants C21, C30, C52, and C67 (Table I). But when the C130 mutant was assayed, self-association was not observed.

[View Larger Version of this Image (20K GIF file)]

Next, the ability of the FasL deletion mutants to associate with the wild-type FasL was assessed. A similar approach was taken as above, except that the FLAG wild-type, -gld, -C80, -C98, and -C130 FasL expression constructs were cotransfected into COS-1 cells with the HA-wild-type FasL construct. Expression of FLAG- and HA-tagged ligands were confirmed as described above (Fig. 7B, upper and middle panels). The ability of the FLAG-tagged wild-type, gld, C80, C98, and C130 FasLs to coprecipitate the HA-tagged wild-type FasL is shown in Fig. 7B (lower panel); as a negative control, the HA wild-type FasL expression construct was transfected alone. The results were very similar to those presented above: deleting the COOH-terminal 98-amino acid residues had little effect on the ability of the FasL to recognize and associate specifically with the wild-type FasL. However, when the next 32 amino acid residues were removed (C98  C130), the association with the wild-type FasL was lost. These results suggest that an important domain for human FasL self-association is located between amino acids 137 and 183.

The human FasL contains three potential N-glycosylation sites at asparagine residues 184, 250, and 260 (see Fig. 1). It is conceivable that the N-glycosylation status of the FasL could influence its biological activity. This possibility is relevant for our study since the COOH-terminal deletions remove these N-glycosylation sites. The three N-glycosylation sites in the extracellular domain of the human FasL were eliminated individually (N184Q, N250Q, and N260Q) and together (N184Q,N250Q,N260Q) by hybrid PCR-based mutagenesis. The N-glycosylation mutants were expressed as demonstrated by immunoprecipitation and Western analysis (Fig. 8A). However, disruption of all three glycosylation sites resulted in markedly reduced expression. Interestingly, all the mutants retained the capacity to bind to the F_c-Fas receptor fusion protein in a quantitative manner (Fig. 8B). Finally, all the N-glycosylation mutants were biologically active, inducing cell death upon transient transfection into 293 cells (Fig. 8C).

[View Larger Version of this Image (49K GIF file)]

The results demonstrate that all three potential N-glycosylation sites are utilized and that the level of expression is influenced by the extent of N-glycosylation. Furthermore, N-glycosylation does not appear to be required for receptor binding or biological activity. These findings suggest that the extracellular protein sequence of the FasL largely dictates biological activity, which depends on receptor binding and oligomerization.

DISCUSSION

This study presents data to indicate that separate regions within the extracellular domain of the FasL sequence are required for self-association and binding to its receptor. Here we have relied upon transfection of the ligand in mammalian cells to obtain biologically active protein. Furthermore, we have used the transmembrane form of the FasL for our study as the ligand is initially synthesized in this state. This is particularly important for FasL, since it has been demonstrated that mice do not produce active soluble ligand, and, thus, rely entirely on the transmembrane protein (46).

Deletion of as few as 3 amino acid residues from the COOH terminus (C3) was found to interfere with receptor recognition, and as a consequence, biological activity. This mutant form of the ligand, however, was still capable of self-associating to the same extent as the wild-type FasL and readily formed SDS-PAGE stable dimers and trimers (Fig. 6, A and B). In fact, removing the three terminal residues yielded more trimeric ligand relative to the wild-type and gld FasLs under our assay conditions. Several other independent experiments support the hypothesis that the COOH-terminal region of the ligand is essential for receptor binding. First, the corresponding gld mutation in the human FasL (F275L) displayed significantly attenuated receptor binding, as assayed by binding to an F_c-Fas receptor fusion protein. The gld phenylalanine to leucine mutation is located 7 amino acids from the COOH terminus of the human FasL. Our results are consistent with a report in which an F_c-receptor fusion protein was able to detect FasL expressed on the surface of cells from normal mice, but not on cells derived from gld/gld mice (47). The phenylalanine residue is also found in the same relative location for several, but not all members of the TNF family, including lymphotoxin-, lymphotoxin-, and the CD27 ligand. The impaired binding of the gld and C3 human FasL mutants was correlated with the inability of these proteins to induce apoptosis.

Second, the NOK-1 blocking antibody also mapped to the COOH terminus of the human FasL. The NOK-1 antibody is capable of antagonizing the apoptosis-inducing activity of the human FasL (43). Our results indicate that NOK-1 directly competes with the Fas receptor for ligand. These findings support the conclusion that the COOH-terminal region of the FasL is a key determinant of receptor binding and biological activity. It is also very likely that other distinctive FasL-Fas receptor contacts are utilized to activate the receptor and initiate apoptosis.

The role of N-linked glycosylation on the expression and activity of the human FasL has been examined, since several of the FasL deletions (C21, C30, and C98) remove N-glycosylation consensus sites (NX(S/T)). The asparagines at residues 184, 250, and 260 within the extracellular domain of the human FasL were mutated to glutamines. The potential glycosylation sites were eliminated individually and together. The results yield the following conclusions about the human FasL: 1) each potential N-glycosylation site is utilized; 2) the level of protein expression is dependent on N-glycosylation; and 3) N-glycosylation is not required for Fas receptor binding or biological activity. Thus, the results using the COOH-terminal deletion mutants can be directly related to alterations in the protein sequence and not the N-glycosylation status of the human FasL.

Another significant finding of this study is the elucidation of a domain within the FasL which is required for oligomerization. This self-assembly domain resides in a region of the FasL extracellular domain well separated from the COOH terminus (Fig. 9). The gld and C3 mutants could self-associate and surprisingly, deletion of as many as 98 amino acid residues from the COOH terminus was still permissive for ligand self-association (Fig. 7A). The results suggest that the COOH-terminal 98 amino acid residues are not necessary for self-association. However, removal of the next 32 amino acids interfered with self-association. These observations therefore define a region within the extracellular domain required for oligomerization.

[View Larger Version of this Image (9K GIF file)]

Additional evidence for ligand oligomerization was provided by coimmunoprecipitation experiments using different FasL deletion mutants expressed with the wild-type ligand (Fig. 7B). Consistent with the self-association results, the C98 mutant formed stable complexes with the wild-type ligand. The C130 mutant, however, did not complex with the wild-type ligand. These findings further suggest that the COOH-terminal 98 amino acids of the FasL appear to be unnecessary for oligomerization.

While the naturally cleaved soluble human FasL exists as a trimer, very little is known about the oligomerization state of the full-length FasL. Several experiments indicate that the transmembrane ligand may form larger aggregates. First, when a recombinant soluble mouse FasL containing the entire extracellular domain was expressed in COS cells, the resulting ligand was found to be biologically active and formed aggregates of ~300 kDa as determined by equilibrium gel filtration (46). This result suggests that this form of the FasL assembles into decameric complexes under physiologic conditions. Second, native gel analysis of the full-length mouse FasL demonstrated the presence of three major forms migrating with apparent molecular masses of 200, 400, and 600 kDa (45).

Unlike many members of this ligand family, it has been very difficult to obtain biologically active recombinant soluble FasL. This suggests that the FasL may have several unique biosynthetic requirements which make this protein distinct from other family members. For example, ligand produced from bacteria or baculovirus-infected cells is not biologically active.² Another observation which suggests that FasL may have distinct structural elements with respect to other members of the TNF family is that FasL is relatively detergent-insensitive. This is in contrast to TNF- which readily dissociates into a monomeric form under such conditions and has been demonstrated to have hydrophobic subunit interfaces (48, 49).

Mutagenesis of TNF- has defined individual residues responsible for receptor binding, trimer formation, and cytotoxicity (50-52). These studies have utilized recombinant ligand engineered with specific point mutations. Residues involved in either receptor binding or oligomerization are widely distributed with respect to the primary structure of TNF-. This is also true of the naturally occurring mutations of CD40 ligand found in X-linked hyper-IgM syndrome, most of which are predicted to disrupt the tertiary and quaternary structures (37). Our study has focused on a series of deletions which remove sequential secondary structural motifs. To our knowledge a similar approach has not been performed on any other family member.

A molecular model has been proposed for the mouse FasL based on the x-ray crystallographic data obtained for TNF- and lymphotoxin-. It is useful to consider our results in light of the proposed structure (41). As TNF- and lymphotoxin- have been demonstrated to adopt a "jelly-roll" structure formed by a series of -strands and intervening loop regions, the FasL was assumed to have similar tertiary and quaternary structures. While the model is helpful in defining the general structural features of FasL, FasL shares only about 30% identity with TNF- or lymphotoxin-. Therefore, details concerning specific residues involved in receptor binding or oligomerization cannot be predicted with any significant degree of certainty. In support of this perspective, the three-dimensional structure of the CD40 ligand has recently been solved (37). While the overall folding pattern was demonstrated to be similar to TNF- and lymphotoxin-, several deviations from the predicted structure were observed. This has allowed many of the X-linked hyper-IgM mutations to be placed in the vicinity of a predicted receptor contact loop which was not appreciated prior to solving the actual crystal structure. Furthermore, while the mouse gld mutation was predicted to demonstrate impaired oligomerization based on its molecular model, this mutant was subsequently shown to oligomerize perfectly well (45).

Our mutagenesis study represents the first attempt to systematically address structure-function relationships for the FasL. The self-association region defined by our deletion study resides between amino acids 137 and 183. This region contains several very well conserved amino acids (Ala-147, Leu-160, Trp-162, and Leu-181) found in all TNF family ligands. A prediction from the sequence alignments of the family members is that the other ligands may also require the same domain for self-association. Furthermore, the observation using the C3 mutant that the COOH terminus is important for receptor binding is substantiated by our results using the -FasL NOK-1 neutralizing antibody and the adjacent gld mutant, which shows impaired receptor binding. Future studies should allow for a more lucid picture of the structural requirements dictating receptor binding and oligomerization for the FasL.