Membrane-bound Fas ligand only is essential for Fas-induced apoptosis.

Fas ligand (FasL), an apoptosis-inducing member of the TNF cytokine family, and its receptor Fas are critical for the shutdown of chronic immune responses and prevention of autoimmunity. Accordingly, mutations in their genes cause severe lymphadenopathy and autoimmune disease in mice and humans. FasL function is regulated by deposition in the plasma membrane and metalloprotease-mediated shedding. Here we generated gene-targeted mice that selectively lack either secreted FasL (sFasL) or membrane-bound FasL (mFasL) to resolve which of these forms is required for cell killing and to explore their hypothesized non-apoptotic activities. Mice lacking sFasL (FasL(Deltas/Deltas)) appeared normal and their T cells readily killed target cells, whereas T cells lacking mFasL (FasL(Deltam/Deltam)) could not kill cells through Fas activation. FasL(Deltam/Deltam) mice developed lymphadenopathy and hyper-gammaglobulinaemia, similar to FasL(gld/gld) mice, which express a mutant form of FasL that cannot bind Fas, but surprisingly, FasL(Deltam/Deltam) mice (on a C57BL/6 background) succumbed to systemic lupus erythematosus (SLE)-like autoimmune kidney destruction and histiocytic sarcoma, diseases that occur only rarely and much later in FasL(gld/gld) mice. These results demonstrate that mFasL is essential for cytotoxic activity and constitutes the guardian against lymphadenopathy, autoimmunity and cancer, whereas excess sFasL appears to promote autoimmunity and tumorigenesis through non-apoptotic activities.

Although Fas-induced apoptosis is thought to require extensive aggregation of pre-assembled Fas trimers12-15, it has not been resolved whether mFasL, sFasL or both cause cell killing when expressed physiologically 4,5. It is also debated whether either or both of these forms of FasL may have non-apoptotic activities, such as induction of inflammatory responses5,16,17. Indeed, sFasL is capable of activating the NF-κB pathway (Supplementary Fig. 1 and 18,19). We sought to determine the physiological functions of mFasL and sFasL by generating gene-targeted mice that cannot shed FasL but do express membrane-bound FasL (FasL) or conversely mice that lack membrane-bound FasL but are capable of producing sFasL (FasL). Studies using FasL over-expression in transfected cell lines showed that the former can be accomplished, by mutating the sequences in the fasl gene encoding the amino acids required for metalloprotease-mediated cleavage12-14 (Fig. 1a, Supplementary Fig. 2a,b,d). Conversely, the latter can be achieved by replacing the sequences in the fasl gene encoding the trans-membrane and intra-cellular regions of FasL with those encoding the signal peptide of the cytokine G-CSF12-14 (Fig. 1a, Supplementary Fig. 2a,c,e).

Figure 1

Generation of mutant mice that specifically lack either secreted FasL or membrane-bound FasL

a, Schematic diagram of the wt mouse fasl gene, the mutation for creating the FasL allele, in which the metalloproteinase recognition site (middle sequence) was altered to prevent FasL shedding and the mutation for creating the FasL allele, in which sequences encoding the human G-CSF signal sequence were fused in frame with those for the extracellular region of mouse FasL (lower sequence) to preclude insertion of FasL into the plasma membrane but allowing secretion of FasL (sFasL). P, CYTO, TM and EXT represent the promoter, cytoplasmic, transmembrane and extracellular regions, respectively. The positions of PCR primers used for genotyping of the gene-targeted mice are indicated. b, Immunofluorescent staining and confocal microscopy to demonstrate intracellular localisation of FasL (green) in activated T cells from wt, FasL and FasL mice. DAPI (blue) was used to label nuclei. c, ELISA to quantify the levels of FasL in the supernatants of activated T cells from wt, FasL and FasL mice. Each dot represents a single mouse and the bar indicates the mean +/- SD. FasL T cells had significantly (p<0.03) less FasL in their supernatants compared to wt T cells. d,e, Cell surface immunofluorescent staining and FACS analysis to measure expression of membrane-bound FasL on activated T cells from wt, FasL (d) and FasL (e) mice. Solid lines show staining with an anti-FasL antibody and dotted lines show staining with an isotype-matched control antibody. Values represent mean +/- SD of mFasL+ T cells from 3 independent experiments (p<0.05 wt vs FasL mice at 6 and 24 h; p<0.05 wt vs FasL mice at 6 h). For all experiments (b,c,d,e) T cells were stimulated with ConA for 3 days, rested for 2 days in IL-2 (0 h) and then restimulated with PMA plus ionomycin for 6 or 24 h.

To verify that the mutations had the intended consequences, we compared the expression and subcellular localisation of FasL between mitogenically activated T lymphocytes from FasL, FasL and wt mice. Immunofluorescent staining and confocal microscopic analysis of fixed cells showed that intracellular localisation and levels of the FasLΔs and FasLΔm mutant proteins were comparable to those of wt FasL (Fig. 1b). ELISA demonstrated that mitogen-activated T cells from FasL and wt mice contained substantial levels of FasL in their supernatants whereas FasL T cells had significantly less (Fig. 1c). FasL in cellular supernatants can be found in two forms: secreted sFasL derived by metalloprotease-mediated cleavage or mFasL present on vesicles that had been shed by cells12. The latter can efficiently trigger Fas-mediated apoptosis in cultured cells12, although the physiological relevance of this remains unclear. Regardless, FPLC and ultra-centrifugation revealed that in contrast to FasL from supernatants of wt or FasL T cells, a substantial fraction of FasL in supernatants of FasL T cells resided in membranous fractions (Supplementary Fig. 3). Finally, immunofluorescent cell surface staining and FACS analysis identified significantly higher levels of membrane-bound FasL on activated T cells from FasL mice compared to wt T cells (Fig. 1d), consistent with the notion that metalloprotease-mediated cleavage reduces the levels of mFasL11-14. As expected, no FasL was detected on the surface of FasL T cells (Fig. 1e). These results verify that FasL mice produce mFasL that cannot be shed by metalloproteases, whereas FasL mice lack mFasL but produce sFasL.

Since FasL contributes to the killing of virus-infected and other target cells4, we examined which form is critical. We used mitogen-activated T cells from wt or the mutant fasl knock-in mice as killers and the FasL sensitive CH1 mouse B lymphoma cells (Supplementary Fig. 4a,b) as targets. FasL T cells killed CH1 cells with significantly higher efficiency than wt T cells (Fig. 2a). In contrast, FasL T cells possessed only poor cytotoxic activity, comparable to those from FasL-deficient FasL mice (Fig. 2b). FasL neutralisation inhibited the cytotoxicity of wt and FasL T cells but did not further reduce the poor killing by FasL or FasL T cells (Supplementary Fig. 4c-e), demonstrating that only the former triggered a FasL/Fas-dependent apoptotic process. Restimulation of activated T cells in vitro causes activation induced cell death (AICD), which is largely dependent on FasL-mediated (paracrine and/or autocrine) Fas activation4. Stimulation with mitogenic antibodies to CD3 triggered AICD in FasL T cell blasts as efficiently as in wt T cell blasts (Fig. 2c). In contrast, AICD was abnormally reduced in FasL T cells, indeed to a similar extent as seen with their FasL counterparts (Fig. 2d). FasL neutralisation significantly reduced AICD in wt and FasL T cells but did not further diminish the already reduced killing of FasL or FasL T cells (Supplementary Fig. 5). This is consistent with the notion that AICD involves FasL/Fas-dependent as well as -independent mechanisms4,20. Collectively, these results demonstrate that mFasL but not sFasL is essential for Fas-induced killing of target cells and AICD and they indicate that metalloprotease-mediated cleavage of mFasL reduces the cytotoxic activity of activated T cells. This is consistent with certain findings with transfected cell lines over-expressing FasL12-14 and the notion that apoptosis induction requires not only binding of FasL trimers to pre-assembled Fas trimers but more extensive aggregation of Fas trimers12-15. This contrasts with TNFα, which can kill TNF-R1+ target cells in both membrane-bound as well as secreted form4,5.

Figure 2

Membrane-bound but not secreted FasL is essential for target cell killing and AICD

a,b, CH1 target cells (FasL sensitive) were co-cultured with activated T cells from wt (blue squares), FasL (black stars) or FasL (purple triangles) mice (a) or from wt (blue squares), FasL (red circles) or FasL (purple triangles) mice (b) at the indicated effector:target ratio. The percentage of CH1 target cell killing was measured by FACS analysis after 24 h. c,d, Activated T cells from wt (blue squares), FasL (black stars) or FasL (purple triangles) mice (c) or from wt (blue squares) FasL (red circles) or FasL (purple triangles) mice (d) were either cultured in medium plus IL-2 (broken lines) or restimulated for 6, 24 or 48 h with antibodies to CD3 (solid lines). Cell survival was measured by staining with PI plus FITC-coupled annexin V and FACS analysis. Values in graphs in a,b,c,d represent mean +/- SEM from 3 independent experiments. * p<0.05 a,c FasL vs wt, b,d FasL vs wt.

Defects in AICD of mature T cells are thought to be the cause of the lymphadenopathy, hyper-gammaglobulinaemia and autoimmunity in mice and humans deficient in FasL (e.g. FasL) or Fas (e.g. Fas)4. As reported21, by ∼100 days of age FasL mice (on an inbred C57BL/6 background) started to show signs of lymphadenopathy and splenomegaly (Fig. 3a,b), including accumulation of large numbers of ‘unusual’ TCRα/β+CD4-CD8-B220+ T cells (Fig. 3c,d). This was accompanied by hyper-gammaglobulinaemia (Fig. 3e) with high titers of anti-nuclear auto-antibodies (ANA, Fig. 3f,g). FasL mice developed lymphadenopathy, splenomegaly, hyper-gammaglobulinaemia and ANA at a similar rate and extent as FasL mice (Fig. 3). Remarkably, however, the titres of anti-DNA auto-antibodies were significantly higher in FasL mice compared to FasL mice (67% vs 14% symptomatic of SLE-like disease (>12 IU/mL), p<0.05; Fig. 3h). In contrast, FasL mice exhibited none of these abnormalities and had a normal lifespan (Supplementary Fig. 6). These results show that mFasL but not sFasL is essential for the killing of unwanted lymphocytes that is required to prevent lymphadenopathy, hyper-gammaglobulinaemia and accumulation of auto-antibodies.

Figure 3

Membrane-bound FasL but not secreted FasL is essential to prevent lymphadenopathy, splenomegaly and hyper-gammaglobulinemia with anti-nuclear autoantibodies

a, Lymph nodes (axillary, brachial, inguinal and mesenteric) and b, spleens from wt (blue), FasL (purple), FasL (green) or FasL (red) mice of the ages indicated were weighed. c,d, The percentages of the ‘unusual’ CD3+B220+ T cells in the lymph nodes of these mice were determined by FACS analysis (numbers indicate % of cells in each quadrant). e, The levels of the indicated immunoglobulin isotypes in the sera of these mice (age 5 months) were determined by ELISA, mean +/- SD. f, The levels of anti-nuclear autoantibodies (ANA) in the sera of these mice, aged 3 (left panel) or 5 months (right panel) were quantified by indirect immunofluorescence staining (1/100 serum dilution) of slides covered with human HEp-2 epithelial cells and scoring of brightness of fluorescence intensity on a scale of 0: no fluorescence to 3+: maximal fluorescence. g, Examples of ANA quantification in sera of mice of the indicated genotypes. Immunofluorescence intensity score is indicated in brackets. Values in graphs in a,b represent mean +/-SEM and in c,e,f they represent mean +/-SD from a minimum of 10 to 52 FasL mice. h, The levels of anti-DNA auto-antibodies in sera of mice of the indicated genotypes and ages were measured by ELISA (each dot represents the value for a single mouse). Bar at 12.5 IU/ml indicates the level diagnostic for SLE.

FasL mice express a mutated FasLgld protein (in membrane-bound and secreted form) that is unable to bind to its receptor Fas7, whereas the secreted FasLΔm protein (like wt sFasL) can bind to Fas (Supplementary Fig. 1 and12-14). If sFasL plays a role in inflammation, for example through NF-κB activation (Supplementary Fig. 1 and 18,19), one would expect significant differences in morbidity and mortality between FasL and FasL mice. It was therefore remarkable that FasL mice became sick significantly earlier than FasL mice (Fig. 4a, median latency 406 vs 658 days; p<0.0001). At autopsy 62% of terminally ill FasL mice showed signs of SLE-like autoimmune disease, including cellular crescents, protein casts (Fig. 4b,c), deposition of IgM as well as IgG antibodies and complement in renal glomeruli (Supplementary Fig. 7). In contrast, such pathologies were only rarely observed in FasL mice, and then only at a considerably older age and with less severity. Remarkably, by 400 days ∼50% of FasL but only ∼15% of the FasL mice had developed fatal SLE-like autoimmune kidney disease (Fig. 4b).

Figure 4

FasL mice die considerably earlier than FasL mice due to SLE-like fatal glomerulonephritis and histiocytic sarcoma

a, Kaplan-Meyer survival curves for control (blue line, wt and FasL combined), FasL (red line) and FasL mice (purple line) (control vs FasL: p<0.0001; control vs FasL: p<0.0001; FasL vs FasL: p<0.0001). b, Incidence of severe autoimmune kidney disease (glomerulonephritis (GN) score ≥3 in FasL (red line) and FasL (purple line) mice. (FasL vs FasL: p<0.0175). c, H&E stained sections of kidneys from mice of the indicated genotypes and ages were examined for pathological changes, such as hypercellularity, cellular crescents, dilated tubules or sclerotic glomeruli and the % of total glomeruli affected scored on a scale of 0-4 (0=normal, 1=0-25%, 2=25-50%, 3=50-75%, 4=>75% affected; scores indicated in brackets). d, Incidence of histiocytic sarcoma in FasL, FasL and control wt mice (FasL vs FasL: p=0.03), ** 27. e. Pictures of two FasL mice with histiocytic sarcoma (aged 329 and 417 days). Macroscopic lesions are readily seen in the livers and spleens. f, H&E stained sections of livers, spleens and lungs of histiocytic sarcoma-burdened FasL mice (aged 329 and 417 days; * indicates histiocytic lesions). Scale bar represents 155 μm.

By 5 months 46% of FasL mice presented with very high (>3000 ng/mL) serum IgE levels and ∼30% had developed severe dermatitis with lesions appearing on their ears and necks (Supplementary Fig. 8a,b). Although not previously reported, we observed this autoimmune pathology also in some FasL mice, albeit at decidedly lower incidence (∼10%) and later in life compared to FasL mice. In Fas mice lymphadenopathy and accumulation of TCRα/β+CD4-CD8-B220+ T cells are accompanied by abnormally increased serum levels of pro-inflammatory cytokines, including TNFα, IL-6, Ifnγ and FasL itself22. FasL was shown to activate NF-κB transcription factors and expression of pro-inflammatory cytokines and chemokines (Supplementary Fig. 1 5,18,19). We therefore hypothesised that FasL mice may develop autoimmune disease more rapidly and at higher incidence than FasL animals because only the former produce excess sFasL that can bind to its receptor Fas, which may then activate NF-κB and thereby drive production of pro-inflammatory cytokines (Supplementary Figs. 8 and 9). Consistent with this idea, at 3-5 months FasL mice contained substantial numbers of cells with high levels of nuclear (i.e. active) p65/NF-kB in the spleen, liver (Supplementary Fig. 10) and kidneys (not shown) and high serum levels of TNFα (Supplementary Fig. 8; p<0.0001 for TNFα). These abnormalities were significantly less prevalent in FasL mice and were not detected in FasL or wt animals (Supplementary Figs. 8, 9 and 10).

Fas is expressed on hepatocytes and its activation causes apoptosis23,24. FasL can also be found in the liver, produced by infiltrating T lymphocytes or resident myeloid cells, and it has therefore been hypothesised that FasL-Fas induced apoptosis prevents tumorigenesis in this organ25. Interestingly, a significant fraction (27% by 18 months) of both strains of FasL mice developed hepatic tumours with deposits in the spleen and lungs (Fig. 4d,e). Microscopically and by immuno-phenotype (Mac-1+Mac-2+F4/80+B220-Thy-1-CD3-) these tumours were characteristic of histiocytic sarcoma (Fig. 4e,f and Supplementary Fig. 11), being composed of oval cells with eosinophilic cytoplasm and elongated or folded nuclei26. These tumours were transplantable in C57BL/6 mice (Supplementary Fig. 11), confirming their malignant status. Histiocytic sarcoma was rarely seen in FasL or FasL animals and in C57BL/6 (wt) mice such tumours are observed only at very low frequency late in life (∼5% >18 months26,27; Fig. 4d and Supplementary Fig. 6).

Our findings that mFasL but not sFasL is critical for AICD of T cells in vitro and for prevention of lymphadenopathy, hyper-gammaglobulinemia and accumulation of auto-antibodies within the whole animal are consistent with the notion that repeated TCR stimulation kills chronically activated T cells that are specific for auto-antigens or persistent pathogens through FasL-Fas signalling, thereby preventing lymphadenopathy1-4. The observation that FasL mice develop SLE-like glomerulonephritis and histiocytic sarcoma considerably earlier and with higher incidence than FasL mice indicates that the high levels of sFasL produced in the FasL mice, which in contrast to FasLgld can engage its receptor Fas12,13, may promote autoimmunity and tumorigenesis. sFasL may achieve this by triggering non-apoptotic signalling pathways, such as NF-κB-dependent inflammatory processes. Alternatively, differences between FasL and FasL mice may be due to the fact that retrograde signalling through FasL28,29 can only occur in the former but not the latter, although upon challenge with influenza virus in vivo or stimulation with suboptimal doses of anti-CD3 antibodies in vitro, CD8+ T cell responses were indistinguishable between wt, FasL and FasL mice (Supplementary Figures 12 and 13). It is theoretically also possible that FasL mice die earlier than FasL mice because the FasL mutation causes complete loss of function whereas FasL represents a partial loss of function mutation. For two reasons this appears unlikely: (1) lymphadenopathy and hyper-gammaglobulinemia occur in the two mutant strains (FasL and FasL) with comparable kinetics and magnitude (Fig. 3), indicating that the two mutations do not differ markedly in their potency, and (2) histiocytic sarcoma has not been reported in FasL knock-out mice30. We therefore hypothesise that tumorigenesis may be driven by a combination of loss of mFasL-mediated apoptosis of cells undergoing transformation and sFasL-Fas induced non-apoptotic signals, perhaps NF-κB-mediated stimulation of cell proliferation, survival and/or inflammation within an elevated cytokine milieu.

Methods Summary

Generation of FasL mutant mice

The mouse fasl locus and known restriction sites were used to construct the targeting vector and diagnose homologous recombination in embryonic stem (ES) cells and gene-targeted (FasL, FasL) mice. Targeting knock-in vectors were made with the loxP/pGKNeo/loxP cassette cloned into the Pac1 site. Targeting constructs for the mutant FasL mice (Supplementary Fig. 2) were linearised and electroporated into C57BL/6-derived Bruce-4 ES cells.

Analysis of FasL mutant mice

All experiments with mice were performed according to the guidelines of the Animal Ethics committees of our institutions. Mice were killed at 6, 12 or 20 weeks for analysis and further cohorts were monitored daily for morbidity and killed when showing signs of illness. Tissues were fixed for microscopic analysis in 80% Histochoice (Amresco)/20% ethanol or 10% buffered formalin and embedded in paraffin and conventional histopathology was performed on hematoxylin plus eosin stained sections. For detailed methods for immunohistochemical staining, immunofluorescent staining, confocal microscopy, cell preparation, flow cytometric analysis, ELISA, AICD, chromatography, viral infection and target cell killing, T cell proliferation assays and Western blotting refer to the online methods version.

Statistical analysis was performed using the student's T-test, log rank (Mantel-Cox) test for survival curves or one-way analysis of variance using Turkey's comparison test to compare multiple groups where appropriate.