Members of the tumour necrosis factor receptor family that contain a death domain have pleiotropic activities. They induce apoptosis via interaction with intracellular FADD/MORT1 and trigger cell growth or differentiation via TRADD and TRAF molecules. The impact of FADD/MORT1‐transduced signals on T lymphocyte development was investigated in transgenic mice expressing a dominant negative mutant protein, FADD‐DN. Unexpectedly, FADD‐DN enhanced negative selection of self‐reactive thymic lymphocytes and inhibited T cell activation by increasing apoptosis. Thus signalling through FADD/MORT1 does not lead exclusively to cell death, but under certain circumstances can promote cell survival and proliferation.
Apoptosis is an evolutionarily conserved cell death mechanism that can be triggered by a broad range of physiological signals or experimentally applied stress conditions (reviewed by Jacobson et al., 1997). Independent signalling cascades converge upon activation of latent intracellular cysteine proteases (caspases), leading to proteolysis of vital cellular constituents and ultimately collapse of the cell. Cytoplasmic adaptors, such as CED‐4 or FADD/MORT1 (see below), appear to be critical for recruitment and activation of caspase zymogens and are likely to be the targets of anti‐apoptosis proteins, like those of the Bcl‐2/CED‐9 family or FLIP (reviewed by Jacobson et al., 1997).
Apoptosis plays a prominent role in the development and functioning of the immune system (reviewed by Strasser, 1995). Developing B and T cells that do not express functional antigen receptors fail to receive a positive signal and die by apoptosis, while lymphocytes bearing autoreactive antigen receptors undergo apoptosis in response to a death signal from neighbouring cells (Murphy et al., 1990; Surh and Sprent, 1994). Cell death is also prominent during immune responses in peripheral lymphoid organs. As in immature cells, apoptosis of activated mature lymphocytes can result either from a failure to obtain a positive signal (e.g. via cytokine receptors) or as a consequence of repeated antigen receptor cross‐linking provoking a death signal (reviewed by Strasser, 1995). The former is often referred to as ‘death by neglect’ and the latter as ‘activation‐induced cell death’. These two physiologically triggered cell death pathways are subject to distinct control. Bcl‐2 and its homologues block ‘death by neglect’ in many circumstances but are ineffective antidotes to antigen receptor‐induced apoptosis of immature lymphocytes or cycling mature B and T cells. In contrast, members of the tumour necrosis factor receptor (TNF‐R) family play a critical role in ‘activation‐induced cell death’ but are dispensable for ‘death by neglect’ (reviewed by Strasser, 1995; Jacobson et al., 1997).
Cross‐linking of the T cell receptor (TCR) triggers apoptosis in T hybridoma cells and activated normal T cells by inducing CD95 ligand (also called Fas ligand or APO‐1 ligand) expression, followed by autocrine or paracrine stimulation of the death‐inducing surface receptor CD95 (also called Fas or APO‐1) (Alderson et al., 1995; Brunner et al., 1995; Dhein et al., 1995; Ju et al., 1995). Mice carrying spontaneous or experimentally introduced mutations in CD95 (lpr or CD95 gene knock‐out) or CD95 ligand (gld) develop progressive lymphadenopathy and autoimmunity, apparently due to their failure to delete chronically activated lymphocytes (reviewed by Nagata, 1997). Deletion of antigen‐stimulated peripheral T cells is slowed in mutant mice lacking p55 TNF‐R1 or p75 TNF‐R2, indicating that these receptors also play a role in activation‐induced cell death (Speiser et al., 1996; Sytwu et al., 1996).
Negative selection of autoreactive thymocytes resembles activation‐induced cell death of mature T lymphoblasts: it is also a consequence of TCR ligation, occurs through apoptosis (Murphy et al., 1990; Surh and Sprent, 1994) and involves caspase activation (Clayton et al., 1997). Members of the TNF‐R family have been implicated in this process as well, but the precise events in this death pathway have yet to be determined. Immature CD4+8+ thymocytes express CD95 and other members of the TNF‐R family and are susceptible to apoptosis induced by agonistic anti‐CD95 antibodies (Ogasawara et al., 1995), recombinant soluble CD95L (Suda et al., 1996) or cell surface‐bound CD95L (Müller et al., 1995). However, deletion of autoreactive thymocytes occurs normally in mutant lpr (Sidman et al., 1992), CD95−/− (Adachi et al., 1995), TNF‐R1−/− (Pfeffer et al., 1993; Rothe et al., 1993) and TNF‐R2−/− mice (Erickson et al., 1994) and is only partially impaired in CD30−/− mice (Amakawa et al., 1996), indicating that negative selection cannot be attributed to ligation of any one of these receptors alone.
The TNF‐R family consists of type I transmembrane proteins characterized by cysteine‐rich repeats in their extracellular ligand binding domains (reviewed by Wallach et al., 1996). A subset of the family, comprising CD95, TNF‐R1, p75 NGF‐R, DR3 (also called WSL‐1, TRAMP, APO‐3 or LARD), CAR1 and the two TRAIL receptors (DR4/TRAIL‐R1 and TRAIL‐R2) also share a conserved motif in their cytoplasmic regions, known as the death domain, because it is essential for transducing the apoptotic signal (Itoh and Nagata, 1993; Tartaglia et al., 1993; Chinnaiyan et al., 1996a; Pan et al., 1997; Walczak et al., 1997; reviewed by Wallach et al., 1996). Upon receptor ligation the death domain acts as the docking site for homotypic interaction with death domain‐containing cytoplasmic proteins such as FADD (also called MORT1) (Boldin et al., 1995; Chinnaiyan et al., 1995) and TRADD (Hsu et al., 1995). FADD/MORT1 binds directly to CD95 (Boldin et al., 1995; Chinnaiyan et al., 1995) and is recruited to TNF‐R1, DR3 and possibly other related receptors via TRADD (Chinnaiyan et al., 1996a; Hsu et al., 1996b). The death effector domain at the N‐terminal end of FADD/MORT1 then interacts with a related motif in the prodomain of caspase‐8 (also called FLICE, MACH or Mch5) (Boldin et al., 1996; Muzio et al., 1996) or caspase‐10b (also called FLICE2) (Vincenz and Dixit, 1997). Activation of these upstream cysteine proteases is thought to trigger a proteolytic cascade which apparently constitutes the ‘point of no return’ in apoptosis. Under certain circumstances, however, CD95 and other members of the TNF‐R family that contain death domains can stimulate alternative signalling pathways which exert positive effects on cell survival and proliferation, rather than triggering apoptosis (see Discussion; reviewed by Wallach et al., 1996).
Thymocyte negative selection may not be perturbed in mutant mice lacking a single member of the TNF‐R family because of functional redundancy among these receptors. This hypothesis can be tested by inactivating all of these receptors or a common signal transducer. A mutant of FADD/MORT1 which lacks the death effector domain and therefore cannot transduce a signal to caspase‐8 acts as a specific inhibitor of CD95‐, TNF‐R1‐, DR3‐ and TRAIL‐R2‐transduced apoptosis in various cultured cell lines (Chinnaiyan et al., 1996a,b; Hsu et al., 1996b; Walczak et al., 1997) and there is, to our knowledge, no indication that it interferes with any other process. We generated transgenic mice expressing a dominant interfering mutant of FADD/MORT1 (hereafter called FADD‐DN) under control of the mouse lck proximal promoter to investigate the impact of FADD/MORT1‐transduced signals in toto on T lymphocyte development and function. Unexpectedly, we found that FADD‐DN expression enhanced thymocyte negative selection and inhibited peripheral T cell activation by increasing apoptosis. These results provide evidence that signalling pathways operating through FADD/MORT1 do not lead exclusively to apoptosis, but under certain circumstances can promote cell survival and proliferation.
Generation of transgenic mice expressing a dominant interfering mutant of FADD/MORT1
We generated two independent lines of C57BL/6J transgenic mice (strains 60 and 64) expressing a dominant interfering mutant of FADD/MORT1, FADD‐DN, which lacks the death effector domain that is needed for activation of caspase‐8 (Figure 1A). The N‐terminal FLAG epitope tag, utilized to detect protein expression, had no effect on FADD‐DN function in CD95L‐sensitive CH1 murine B lymphoma cells (data not shown). Consistent with the T cell specificity of the lck proximal promoter, the 16 kDa FADD‐DN protein was detected by Western blotting (Figure 1B) and flow‐cytometric analysis (Figure 1C) in thymus, spleen and lymph nodes, but not in bone marrow. A higher level of the protein was observed in strain 64 than in strain 60 (Figure 1B). Two‐colour immunofluorescence staining of surface markers and cytoplasmic FADD‐DN in lymph node cells (Figure 1D) and spleen cells (data not shown) confirmed that transgene expression was restricted to T lymphocytes.
FADD‐DN blocks CD95‐transduced apoptosis in T lymphoid cells
To test whether expression of the FADD‐DN transgene could block apoptosis induced by death domain‐containing members of the TNF‐R family in non‐transformed cells, thymocytes from strains 60 and 64 were exposed in vitro to recombinant soluble human CD95 ligand (rhCD95L). This ligand carries a FLAG epitope tag and a FLAG‐specific antibody was used to enhance cross‐linking of receptor–ligand complexes (Apotech Inc). The FADD‐DN transgenic cells were completely resistant to ligand‐induced apoptosis, while thymocytes from control littermates were killed in a dose‐dependent manner (Figure 2E and F). FADD‐DN also blocked thymocyte apoptosis induced by treatment with rhCD95L alone or anti‐CD95 antibodies plus cycloheximide (data not shown). The level of protection afforded by FADD‐DN expression was equivalent to that conferred by transgenic expression of the cowpox virus serpin CrmA, an inhibitor of caspase‐8 (Zhou et al., 1997) or by the CD95 loss‐of‐function mutation lpr (Figure 2E and F). In contrast, FADD‐DN expression did not inhibit spontaneous death of thymocytes in culture nor did it affect their sensitivity to γ‐radiation, phorbol ester, dexamethasone or the calcium ionophore ionomycin (Figure 2A–C and data not shown).
Like thymocytes, mature resting T cells from both FADD‐DN transgenic strains were resistant to rhCD95L‐induced apoptosis (Figure 2G). Cell surface levels of CD95 were normal in FADD‐DN transgenic thymocytes and mature T cells (Figure 2H–K), excluding decreased receptor expression as a basis for the observed resistance. These results show that FADD‐DN can prevent CD95‐transduced apoptosis in non‐transformed cells. The finding that FADD‐DN expression does not inhibit apoptosis triggered by DNA damage, glucocorticoid treatment or other cytotoxic stresses provides evidence that these stimuli activate signalling pathways that are independent of FADD/MORT1.
Deletion of autoreactive T cells is enhanced by FADD‐DN
To examine specifically the impact of FADD‐DN on negative selection of autoreactive thymocytes we investigated the fate of cells responsive to a self‐superantigen. Mice expressing Mls‐2a as well as class II MHC I‐E molecules delete their TCRVβ3‐bearing T cells (Pullen et al., 1988). The I‐Ed and mls‐2a genes were introduced into FADD‐DN mice by crossing strain 64 with BALB/c‐bcl‐2 transgenic mice. Offspring inheriting both the FADD‐DN and the bcl‐2 transgenes enabled us to investigate whether these two cell death antagonists synergize in blocking thymocyte negative selection. Figure 3A shows the percentage of TCRVβ3+ lymph node T cells in control C57BL/6 and BALB/c mice and in (C57BL/6×BALB/c)F1 mice with the genotypes FADD‐DN, bcl‐2, FADD‐DN/bcl‐2 or control wild‐type. Consistent with previous observations (Pullen et al., 1988; Strasser et al., 1991), TCRVβ3+ T cells are deleted in mice expressing I‐Ed plus Mls‐2a and this is not blocked by Bcl‐2. Transgenic expression of FADD‐DN or FADD‐DN plus Bcl‐2 was also unable to prevent deletion of TCRVβ3+ T cells (Figure 3A).
The impact of FADD‐DN on negative selection of T cells recognizing conventional antigens was investigated by crossing strain 64 mice with anti‐HY TCR transgenic mice. This transgenic TCR recognizes a peptide derived from the male antigen HY presented by class I MHC H‐2Db molecules. Consequently, thymocytes produced in male H‐2Db anti‐HY TCR transgenic mice are autoreactive and undergo deletion at the CD4+8+ stage of development (reviewed by von Boehmer, 1990). Surprisingly, FADD‐DN expression enhanced negative selection. Male H‐2Db FADD‐DN/anti‐HY TCR doubly transgenic mice had on average 3‐fold fewer CD4+8+ thymocytes than did male littermates expressing just the TCR transgene (Figure 3B and D). HY‐TCR+ T cells expressing abnormally low levels of the co‐receptor CD8 are unresponsive to male antigen presented by H‐2Db, thereby escape intra‐thymic deletion and are found as CD4−8low cells in peripheral lymphoid organs of male H‐2Db anti‐HY TCR transgenic mice (reviewed by von Boehmer, 1990). Consistent with the notion that FADD‐DN enhances thymocyte negative selection, the number of CD4−8low T cells was reduced 3‐ to 4‐fold in the lymph nodes and spleen of male H‐2Db FADD‐DN/anti‐HY TCR doubly transgenic mice compared with male littermates expressing just the TCR transgene (Figure 3C and E).
Injection of animals with antibodies to the TCR/CD3 complex triggers deletion of immature CD4+8+ thymocytes (Smith et al., 1989) and this is regarded as a model for negative selection. Consistent with the aforementioned results, anti‐CD3 antibody injection led to a 3‐fold greater reduction in CD4+8+ thymocytes in FADD‐DN transgenic mice compared with control littermates (data not shown).
Collectively these data demonstrate that FADD/MORT1‐transduced signals from death domain‐bearing TNF‐R family members do not promote thymocyte negative selection. Instead, unexpectedly, they antagonize it.
FADD‐DN inhibits mitogen‐induced proliferation of mature T cells
Reports that p55 TNF‐RI and CD95 can transduce stimulatory signals for cell growth (Alderson et al., 1993; Mackay et al., 1994) led us to examine the effect of FADD‐DN expression on T cell activation. To this end spleen cells (Figure 4A–C) or purified T cells (Figure 4D–K) from control and FADD‐DN transgenic mice were activated in vitro with mitogenic antibodies to CD3 and CD28, concanavalin A or phorbolmyristyl acetate (PMA) and ionomycin. The proliferative response of normal T cells, as measured by [3H]thymidine incorporation, was maximal after 3–4 days stimulation (Figure 4). T cells from FADD‐DN transgenic mice responded abnormally to all these stimuli. Thymidine uptake in response to concanavalin A was low throughout an experiment, reaching values that were on average 10‐fold lower than the peak response of normal T cells (Figure 4B and C). Maximal thymidine uptake and maximal production of live T cells after CD3 plus CD28 cross‐linking or treatment with PMA and ionomycin were the same as in control cultures but occurred 2–3 days later than with normal T cells (Figure 4A and D–I). For example, after 3–5 days PMA plus ionomycin treatment (Figure 4G–I) or CD3‐ plus CD28‐cross‐linking (Figure 4J and K) control cultures had reached 1–2×106 viable cells/ml, while FADD‐DN cultures had 10‐ to 20‐fold fewer viable cells. Experiments with purified T cells and CD4+8− and CD4−8+ subsets (Figure 4G–I) demonstrated that the defect caused by FADD‐DN expression was intrinsic to the T cells and that both subpopulations were similarly affected. The abnormal growth of FADD‐DN T cells was apparently not due to impaired autocrine stimulation by IL‐2 because the defect remained when saturating amounts of IL‐2 were added to the cultures (Figure 4 and data not shown). Titration of each of the mitogenic reagents demonstrated that normal and FADD‐DN T cells had the same requirements for a maximal response (Figure 4C and data not shown), indicating that the proliferative defect of FADD‐DN T cells was not due to suboptimal stimulation. In sum, these results show that interference with the normal action of FADD/MORT1 causes a defect in mitogen‐induced proliferation of T lymphocytes.
Mitogenic stimulation selects T cells that do not express the FADD‐DN transgene
After 4 days of mitogenic stimulation the transgenic T lymphoblasts expressed no detectable FADD‐DN protein (Figure 5A), suggesting that stimulation had selected T cells that did not express the transgene. To further investigate the relationship between FADD‐DN expression and activation we subjected T cells to alternating phases of mitogenic stimulation and quiescence (Figure 5B). As in the experiments shown in Figure 4A, during 4 days of stimulation with antibodies to CD3 and CD28, proliferation was reduced in cultures of FADD‐DN T cells compared with normal T cells (Figure 5B). Following this initial period of stimulation the mitogens were removed and replaced by low levels of IL‐7 to promote cell survival. Over the following 4 days the T lymphoblasts ceased proliferating and entered the resting state (Figure 5B). Day 8 quiescent transgenic T cells did not express detectable amounts of FADD‐DN (Figure 5A). If the transgenic lckpr promoter was active in resting T cells but turned off during mitogenic activation then these cells should have expressed FADD‐DN protein. This experiment therefore provided evidence that only those transgenic T cells expressing the lowest levels of FADD‐DN could proliferate upon mitogenic stimulation. In accordance with this hypothesis, transgenic T cells responded to secondary stimulation in the same manner as normal T cells (Figure 5B).
FADD‐DN expression increases apoptosis and reduces clonogenic growth of mitogen‐activated T cells
Delayed and diminished growth of mitogen‐activated FADD‐DN T cell populations could be due to a reduced frequency of responding cells or a reduced rate of proliferation. The characteristics of the T cell growth curves were consistent with the first hypothesis (Figure 4G–I). In cultures of purified normal T cells the number of viable cells decreased slightly (∼10–20%) over the first 1–2 days of stimulation and increased rapidly thereafter. In cultures of FADD‐DN T cells a larger fraction of the cells (∼70–80%) failed to survive the initial period of stimulation and the number of viable cells did not increase until day 3 (Figure 4G–I). Consistent with these observations, the incidence of apoptosis was considerably higher after 3 days in FADD‐DN T cell cultures compared with those from normal animals (Figure 6A). This effect was specific to mitogenic stimulation, since control and transgenic T cells cultured without mitogens underwent apoptosis at the same rate (data not shown).
The maximal growth rates of normal and transgenic T lymphocytes were similar (Figure 4G–I), indicating that those FADD‐DN transgenic T cells that survived the initial activation process could proliferate at the normal rate. In accordance with this notion the percentage of viable FADD‐DN T cells in S phase of the cell cycle was abnormally low (∼2‐ to 3‐fold below control T cells) after 1 and 2 days stimulation (data not shown) but was normal at later time points (Figure 6B). All surviving T cells increased their volume (Figure 6C), demonstrating that they responded to TCR/CD3 ligation. Limiting dilution cultures revealed that FADD‐DN diminished the clonal growth of mitogen‐activated T cells (Figure 6D). Interestingly, there was no significant difference in clone size (50–200 cells/colony) between control and FADD‐DN transgenic T cells. Collectively these results show that FADD‐DN reduces mitogenic responses of T lymphocytes by increasing apoptosis early during stimulation.
Neither CrmA expression nor the lpr mutation inhibits T cell proliferation
We next used crmA transgenic mice and lpr mutant mice to investigate potential mechanisms by which FADD‐DN might increase apoptosis and inhibit proliferation of stimulated T cells. A direct comparison of the mitogenic response of splenic T cells from crmA transgenic mice, lpr mutant mice, FADD‐DN transgenic mice and control mice demonstrated that neither CrmA expression nor CD95 deficiency prevents T cells from proliferating normally. [3H]Thymidine incorporation, growth kinetics and rate of apoptosis of T cells from crmA transgenic mice, mutant lpr mice and control littermates were equivalent at all times (Figure 7A and B and data not shown). Thus inhibition of caspase‐8 (FLICE/MACH) activity or absence of signals from CD95 alone are unlikely to account for increased apoptosis and defective proliferation of mitogen‐activated FADD‐DN T cells.
Another phenotypic difference between FADD‐DN transgenic mice and mutant lpr mice was the absence of lymphadenopathy in the former. Six‐month‐old lpr mice had abnormally increased numbers of lymph node cells, many of them bearing the unusual B220+TCR/CD3+ T cell phenotype (Figure 7C). FADD‐DN transgenic mice and crmA transgenic animals of the same age did not have these cells and displayed normal lymphoid cellularity. This indicates that FADD/MORT1 and caspase‐8 do not transmit all CD95‐transduced signals that are responsible for normal T lymphocyte homeostasis and prevention of lymphadenopathy.
The FADD/MORT1 protein is an essential signalling intermediary for CD95‐, TNF‐R1‐ and DR3‐transduced apoptosis in some tumour‐derived cell lines (Chinnaiyan et al., 1996a,b; Hsu et al., 1996b). Two receptors for TRAIL have been identified; one induces apoptosis via FADD/MORT1 (Walczak et al., 1997) and the other by an alternative route (Pan et al., 1997). We show here that thymocytes and mature T cells from FADD‐DN transgenic mice are resistant to CD95L‐induced apoptosis (Figure 2), demonstrating for the first time that FADD/MORT1 is essential for this process in non‐transformed cells. The impact of FADD‐DN on TNF‐ and TRAIL‐induced apoptosis could not be assessed, since thymocytes and mature T cells are normally resistant to these ligands (Wiley et al., 1995; our unpublished observations). As more members of the TNF‐R family and their corresponding ligands are discovered these animals will serve as useful tools for assessing the role of FADD/MORT1 in signal transduction in non‐transformed cells.
In accordance with previous findings (Suda et al., 1996; Smith et al., 1997), we found substantial differences in biological activity between CD95L and agonistic anti‐CD95 antibodies. A 100‐fold higher concentration of Jo2 monoclonal antibody than of CD95L was required to induce apoptosis in thymocytes (Smith et al., 1997; our unpublished observations). Mature T cells from normal mice were killed by CD95L (Figure 2G) but resisted treatment with Jo2 antibody, even in the presence of cycloheximide, which facilitates killing in thymocytes (Smith et al., 1997; our unpublished observations). A possible explanation is offered by the recent finding that CD95L and agonistic antibodies can stimulate different signalling pathways (Thilenius et al., 1997). Differences in binding affinity between soluble and membrane‐anchored TNF for its two receptors, TNF‐R1 and TNF‐R2, are responsible for differences in biological responses (Grell et al., 1995). It is possible that in the case of CD95L, as for other members of the TNF ligand family, usage of alternative receptors and differences in the extent of receptor oligomerization determine which signalling pathways are activated.
Possible cell death mechanisms involved in thymocyte negative selection
Six death‐domain‐bearing TNF‐R family members have been identified to date and it has been speculated that there is some functional redundancy between them in thymocyte negative selection. Our results (Figure 3) show that signals from the TNF‐R family which engage the cell death machinery via FADD/MORT1 are dispensable for this process. Unexpectedly, FADD‐DN enhanced the deletion of autoreactive thymocytes (Figure 3), indicating that FADD/MORT1 can be part of a signal that promotes survival of T lymphocytes activated through their TCR. The 1.5‐ to 2‐fold reduction in mature CD4−8+ T cells that we saw in FADD‐DN mice may therefore be due to a defect in positive selection or a consequence of impaired proliferative expansion in peripheral lymphoid organs.
The failure of FADD‐DN to inhibit thymocyte negative selection does not exclude other FADD/MORT1‐independent pathways that lead from TNF‐R family members to caspase activation and apoptosis. Two FADD/MORT1‐independent pathways from oligomerized CD95 receptors have been uncovered in tumour‐derived cell lines: (i) RIP can interact with the CD95 death domain (Stanger et al., 1995) and bind the adaptor RAIDD, which in turn recruits and activates caspase‐2 (Nedd2/Ich‐1) (Duan and Dixit, 1997); (ii) Daxx interacts with the CD95 death domain and triggers apoptosis by causing Jun kinase activation (Yang et al., 1997). It is possible that other death domain‐bearing members of the TNF‐R family can also trigger multiple independent signalling routes to apoptosis.
Two members of the TNF receptor subfamily that lack a death domain have been implicated as regulators of thymocyte negative selection. Impaired deletion of autoreactive T cells was observed in CD30‐deficient mice (Amakawa et al., 1996) and in animals injected with neutralizing antibodies to CD40 ligand (Foy et al., 1995). However, in both instances negative selection was not completely abrogated and it has not been established whether CD30 and CD40 act as ‘death receptors’ within the doomed thymocytes or function within thymic antigen‐presenting cells. Members of the TNF‐R subfamily that lack a death domain can recruit proteins of the TRAF and IAP families and thereby activate Rel/NF‐κB, which is thought to promote cell growth and survival rather than apoptosis (Rothe et al., 1994). One such receptor, CD27, binds to a death domain‐containing cytoplasmic protein termed Siva and triggers apoptosis (Prasad et al., 1997). It is likely that related signalling intermediaries exist for other members of this TNF‐R subfamily.
Deletion of autoreactive thymocytes is delayed in bcl‐2 transgenic mice (Sentman et al., 1991; Strasser et al., 1991, 1994). This may be due to the fact that Bcl‐2 protects highly sensitive CD4+8+ thymocytes against the toxic effects of calcium flux caused by TCR/CD3 ligation (Strasser et al., 1991, 1994). Alternatively, Bcl‐2 may exert its effect on thymocyte negative selection by blocking another signalling route, such as the one transduced via Daxx and Jun kinase (Yang et al., 1997).
We speculate that multiple signalling pathways contribute to elimination of autoreactive thymocytes. Generation of transgenic mice expressing dominant interfering mutants of Siva, RAIDD or Daxx and crosses between those animals, FADD‐DN mice, bcl‐2 transgenic mice and animals lacking TNF‐R family members may clarify the mechanisms of thymocyte negative selection.
The effect of FADD‐DN on T cell activation
There is ample evidence that death domain‐bearing members of the TNF‐R family can promote cell growth and survival. TNF‐R1 transduces a proliferative signal in primary mouse fibroblasts (Mackay et al., 1994) and is essential for liver regeneration after partial hepatectomy (Yamada et al., 1997). The p75 NGF‐R delivers a survival signal in some types of neurons (Kaplan and Miller, 1997). Antibody‐mediated cross‐linking of CD95 enhances cell growth in some cases of chronic lymphocytic leukemia (Mapara et al., 1993), solid tumour cell lines (Owen‐Schaub et al., 1994), normal human diploid fibroblasts (Aggarwal et al., 1995) and T cells that have been triggered via the TCR/CD3 complex (Alderson et al., 1993). Interestingly, anti‐CD3 antibody‐induced T cell proliferation is enhanced by membrane‐anchored CD95L (M.Alderson, personal communication) but reduced by soluble ligand (Suda et al., 1996), indicating that the degree of CD95 receptor oligomerization may determine whether a growth or a death signal is transmitted.
The growth stimulatory effects of CD95, TNF‐R1 and DR3 are thought to be mediated via activation of members of the Rel/NF‐κB transcription factor family (Kruppa et al., 1992; Mackay et al., 1994; Chinnaiyan et al., 1996a; reviewed by Wallach et al., 1996). Experiments on cultured cell lines indicated that the pathway activating Rel/NF‐κB and the signal leading to cell death are distinct. CrmA and FADD‐DN are specific inhibitors of the pathway to apoptosis, while dominant interfering mutants of TRAF2 and RIP selectively block Rel/NF‐κB activation (Chinnaiyan et al., 1996a,b; Hsu et al., 1996a,b). Consistent with these observations, we found that FADD‐DN and CrmA inhibit CD95‐transduced apoptosis (Figure 2) and so far have no indication that these molecules inhibit Rel/NF‐κB activation in mitogenstimulated T cells.
In light of this knowledge it was surprising that FADD‐DN inhibited mitogen‐induced proliferation of mature T lymphocytes and increased apoptosis (Figures 4,5,6). These findings indicate that normal T cells must have a growth‐promoting pathway which is transduced via FADD/MORT1 and/or a close homologue that can be blocked by FADD‐DN. Unlike the pathway leading to apoptosis, this growth‐promoting signal does not require activation of caspase‐8, since T lymphocytes from crmA transgenic mice respond normally to mitogens (Figure 7). We speculate that in activated T cells FADD/MORT1 can bind one or several growth‐stimulating proteins (denoted FAX for FADD/MORT1‐associated protein X in Figure 7D). An attractive candidate is FLIP (also called I‐FLICE, FLAME‐1, CASH and Casper), which can bind to the death effector domain of FADD/MORT1 and inhibit its pro‐apoptotic activity (Irmler et al., 1997).
It is interesting to contemplate why no growth inhibitory effect of FADD‐DN has been observed in stably transfected cell lines (Chinnaiyan et al., 1996a,b; Hsu et al., 1996b; our unpublished observations). Transformed cells are continuously cycling, while mitogen‐stimulated normal T cells must first complete the G0–G1 transition to enter the cell cycle. Bearing this in mind, plus the differences in the rate of apoptosis and kinetics of proliferation between normal and FADD‐DN transgenic T cells (Figure 4), it appears most likely that FADD/MORT1 transduces a survival‐ and/or growth‐stimulatory signal during the G0–G1 transition. Human peripheral blood T cells are resistant to anti‐CD95 antibody‐induced apoptosis during the first 2 days of mitogenic stimulation and only later become susceptible (Owen‐Schaub et al., 1992; Klas et al., 1993). These differences in CD95 signalling are consistent with FADD/MORT1 changing its partner from the growth‐promoting FAX to the death‐inducing caspase‐8 during the course of the G0–G1 transition.
The source of the ligand which stimulates the FADD‐DN‐inhibitable growth stimulatory signal must be T lymphocytes themselves, since transgene expression inhibits proliferation in cultures of purified cells (Figure 4). The nature of the ligand is less clear. We observed no proliferative defect in T cells from lpr mice (Figure 7) and T cell activation was reported to be normal in TNF‐R1−/− mice (Pfeffer et al., 1993; Rothe et al., 1993), thus excluding these two receptors as the sole trigger of the stimulatory signal. The physiological roles of DR3 and the two TRAIL receptors are unknown, but their expression pattern is indicative of a function in lymphocyte activation and/or death (Chinnaiyan et al., 1996a; Pan et al., 1997; Walczak et al., 1997). We believe that several members of the death domain‐bearing TNF‐R family may transduce a growth‐stimulatory signal via FADD/MORT1 early during T cell activation.
Implications for CD95L/CD95‐induced cell death signalling
Why do FADD‐DN and CrmA completely block CD95‐transduced apoptosis in thymocytes and mature T cells (Figure 2) but fail to elicit lpr‐like lymphadenopathy (Figure 7C; Smith et al., 1996)? Insufficient levels of transgene expression appears to be an unlikely explanation, since the phenotype of the incomplete loss of function mutation in lpr mice is only quantitatively different from that in CD95−/− animals (Adachi et al., 1995). Absence of transgene expression in B lymphocytes also cannot account for this observation, since T cell hyperplasia does develop in B cell‐deficient lpr mice (Shlomchik et al., 1994). At least two explanations could account for the finding that TCR+B220+ T cells do not accumulate in FADD‐DN and crmA transgenic mice. While all in vitro killing by CD95L or anti‐CD95 antibodies appears to be critically dependent on FADD/MORT1 and a CrmA‐sensitive cysteine protease, presumably caspase‐8, alternative cell death pathways from CD95 could operate in activated T lymphocytes in vivo. As mentioned above, recruitment and activation of caspase‐2 via RIP and RAIDD (Duan and Dixit, 1997) or binding of Daxx leading to Jun kinase activation and ultimately apoptosis (Yang et al., 1997) stand out as likely candidates. Mice that are doubly transgenic for crmA and bcl‐2 or FADD‐DN and bcl‐2 do not develop lymphadenopathy (Smith et al., 1996; our unpublished observations), indicating that other functions of CD95L/CD95 besides the CrmA‐ and FADD‐DN‐inhibitable FADD/MORT1→caspase‐8 signal and the Bcl‐2‐inhibitable Daxx→Jun kinase pathway are important for T cell homeostasis (Figure 7D).
Another, more speculative, explanation is that CD95L, like TNF and TRAIL, has two receptors, CD95 and another, which can trigger cell growth. According to this model (Figure 7D) T cells in lpr and CD95−/− mice would not only fail to receive a death stimulus via CD95 but also obtain abnormally increased growth stimulation via the second putative CD95L receptor, since it would not compete with CD95 for binding of the common ligand. Two logical predictions of this idea are that the gld point mutation in CD95L abolishes binding to CD95 but maintains interaction with its second putative receptor and that CD95L−/− animals should have a different phenotype than gld mutant mice.
In conclusion, our results reveal a novel complexity of signalling from members of the TNF‐R family by showing that FADD/MORT1 does not always transduce an apoptotic signal but can, under certain circumstances, transmit survival and growth stimulatory signals.
Materials and methods
A HincII–XbaI fragment encoding truncated human FADD/MORT1(αα80–208) was isolated from pcDNA3 AUI‐FADD (Chinnaiyan et al., 1995) and inserted into the BamHI (blunt ended)–XbaI site of pEF FLAG B (Huang et al., 1997). The BglII–BamHI fragment encoding FLAG–FADD/MORT1(αα80–208) was then subcloned into the BamHI site of the expression vector p1017 (Chaffin et al., 1990). Inbred C57BL/6J mouse zygotes were injected with the assembled sequences as a NotI fragment, which contained the proximal promoter of the mouse lck gene (lckpr), FLAG epitope‐tagged FADD/MORT1(αα80–208) and the human growth hormone gene. Transgene‐positive mice were identified by direct PCR amplification of DNA from whole blood (McCusker et al., 1992) using oligonucleotide primers specific to the human growth hormone gene (5′‐TAG GAA GAA GCC TAT ATC CCA AAG G and 5′‐ACA GTC TCT CAA AGT CAG TGG GG).
FADD‐DN transgenic mice (strains 60 and 64) were propagated by serially mating heterozygous transgenic animals with C57BL/6J mice. Other transgenic strains, Eμ‐bcl‐2‐36, expressing human Bcl‐2 constitutively in B and T lymphoid cells (Strasser et al., 1991), anti‐HY TCR (von Boehmer, 1990), CD2‐crmA (strain 65), expressing the cowpox virus serpin CrmA (Ray et al., 1992) in T lineage cells (Smith et al., 1996), and mutant CD95‐deficient lpr mice (Cohen and Eisenberg, 1993), have been described previously. All transgenes were used on a C57BL/6J genetic background, with the exception of Eμ‐bcl‐2‐36, which was on a BALB/c background. FADD‐DN/bcl‐2 doubly transgenic mice were generated by crossing heterozygous FADD‐DN transgenic mice (strain 64) with heterozygous Eμ‐bcl‐2‐36 transgenic mice. Inheritance of the bcl‐2 transgene was determined by PCR using oligonucleotide primers (5′‐GGA ACT GAT GAA TGG GAG CAG TGG and 5′‐GCA GAC ACT CTA TGC CTG TGT GG) to amplify the SV40 element of the transgene construct. FADD‐DN/anti‐HY TCR doubly transgenic mice were generated by crossing heterozygous FADD‐DN mice with homozygous anti‐HY TCR transgenic mice.
Western blot analysis
Western blots were prepared as described previously (Strasser et al., 1995). Proteins were solubilized from cells in lysis buffer (0.25M Tris–HCl, pH 6.8, 10% SDS, 20% glycerol, 5% 2‐mercaptoethanol, 0.02% bromophenol blue and 0.5 μg/ml Pefabloc) and resolved by electrophoresis through 4–20% gradient Tris–glycine polyacrylamide gels (Novex) in the presence of SDS. Proteins were then transferred to nitrocellulose membranes by electroblotting. Non‐specific binding sites were blocked (>1 h) in phosphate‐buffered saline containing 5% skimmed milk, 1% casein and 0.1% Tween‐20, and the membranes were then incubated with 3 μg/ml anti‐FLAG M2 monoclonal antibody (IBI). Bound antibody was detected with affinity purified rabbit anti‐mouse IgG (Fcγ‐specific) antibodies (Jackson ImmunoResearch) and 125I‐labelled protein A (1–2×106 c.p.m./ml).
Immunofluorescence staining, flow cytometric analysis and cell sorting
Dispersed cells from bone marrow, lymph nodes, spleen and thymus were surface stained with monoclonal antibodies that had been conjugated in our laboratory according to the manufacturers' instructions with fluorescein isothiocyanate (FITC), R‐phycoerytherin (PE) or biotin (Molecular Probes). The monoclonal antibodies used and their specificities were: RA3‐6B2, anti‐B220; GK1.5, anti‐CD4; H188.8.131.52, anti‐CD4; 53.6.72, anti‐CD8; YTS 169, anti‐CD8; 8C5, anti‐Gr‐1; T3.70, anti‐HY TCRα chain; HB58, anti‐Igκ; JC5, anti‐Igλ; MI/70, anti‐Mac‐1; Ter119, anti‐erythroid cell surface marker; T24.31.2, anti‐Thy‐1; KJ25‐606.4, anti‐TCRVβ3; RR4‐7, anti‐TCRVβ6; F23.2, anti‐TCRVβ8.2; and T3/70, anti‐HY TCRα chain (for references see Strasser et al., 1991). Bound biotinylated antibodies were detected with PE–streptavidin or Tricolor–streptavidin (Caltag). Hamster monoclonal antibody Jo2 anti‐mouse CD95 (Fas/APO‐1) (PharMingen) was detected with a FITC‐conjugated mouse anti‐hamster IgG (Fcγ‐specific) monoclonal antibody cocktail (PharMingen). Between 5000 and 10 000 viable cells (not stained by propidium iodide) were analysed in a FACScan flow cytometer (Becton Dickinson). Lymph node T cells were purified in a FACS II or FACStar+ sorter by negative cell sorting. FITC‐negative, propidium iodide (PI)‐negative cells were selected following surface staining with a cocktail of FITC‐conjugated monoclonal antibodies specific for B220, Igκ, Igλ, Mac‐1, Gr‐1 and Ter119. FITC‐conjugated monoclonal antibodies to CD4 or CD8 were included to isolate CD4−8+ and CD4+8− T lymphocytes respectively. For long‐term cultures T lymphocytes were purified from suspensions of lymph node cells, using magnetic beads coated with goat anti‐rat IgG (Paesel and Lorei) to remove unwanted cells stained with rat monoclonal antibodies to B220, Mac‐1, Gr‐1 and Ter119. Staining of the enriched T cell population with a biotinylated antibody to Thy‐1 plus PE–streptavidin revealed that the purity was 90–95%. Expression of FLAG–FADD‐DN or Bcl‐2 protein was determined by cytoplasmic immunofluorescence staining and flow‐cytometric analysis. Cells were fixed for 10 min at room temperature in 1% paraformaldehyde and stained for 30 min on ice with 3 μg/ml mouse anti‐FLAG M2 monoclonal antibody (IBI) or mouse anti‐human Bcl‐2 monoclonal antibody (Bcl‐2‐100) in the presence of 0.3% saponin (Sigma). FITC‐conjugated goat anti‐mouse IgG (Fcγ‐specific) antibodies (Southern Biotechnology) were used at 10 μg/ml as secondary reagent. Cells were analysed in a FACScan, live and dead cells being discriminated on the basis of their forward and side light scattering properties.
Cell survival and proliferation assays
Thymocytes, spleen cells and purified T cells were cultured in the high glucose version of Dulbecco‘s modified Eagle's (DME) medium supplemented with 13 μM folic acid, 250 μM l‐asparagine, 50 μM 2‐mercaptoethanol and 10% fetal calf serum (Biosciences). Cell viability was determined by flow cytometric analysis of PI stained cells in a FACScan. Recombinant soluble human CD95L (rhCD95L) carried a FLAG epitope tag and a FLAG‐specific monoclonal antibody was used to enhance cross‐linking of receptor–ligand complexes (Apotech Inc). T cell proliferation following mitogenic stimulation was measured as [3H]thymidine incorporation after 100 μl cultures (starting concentration: spleen cells, 1×106 cells/ml; purified lymph node T cells, 1×105 cells/ml) were pulsed for 6 h with 0.5 μCi [3H]thymidine (Amersham). Viable cell numbers were determined by trypan blue exclusion or by staining T cell cultures with PI, adding a known number of FACS calibration beads (Flow Cytometry Standards) and analysing the samples on a FACScan. Beads and viable (PI−) lymphocytes were distinguished by their different forward and side light scattering properties and the ratio of the two was used to calculate the concentration of live T cells in the cultures.
Cell cycle and apoptosis analysis
T cells were fixed overnight at 4°C in 70% ethanol and stained for 20 min at 37°C with 69 μM PI in 38 mM sodium citrate, pH 7.4, containing 5 μg/ml RNase A. Between 5000 and 10 000 cells were analysed in a FACScan and their cell cycle distribution was determined using DNA Cell‐Cycle Analysis Software Version C (Becton Dickinson). Apoptotic cells were identified within the PI stained population by virtue of exhibiting an apparent subdiploid DNA content.
Limiting dilution analysis of T cell growth
Using the single cell deposition unit of the FACSstar+ sorter, purified lymph node T cells were distributed at limiting dilution (1, 3, 9, 27 or 81 cells/well) in flat‐bottomed 96‐well plates (Falcon) containing optimal concentrations (20 μg/ml in the coating solution) of immobilized monoclonal antibodies (KT3 anti‐CD3 and 37N51 anti‐CD28) plus 100 U/ml rmIL‐2. Wells were scored by microscopy as positive or negative for growth after 5–6 days incubation at 37°C. The precursor frequency of clonogenic cells was estimated from the slope of a graph plotting ln(fraction of wells negative for growth) versus the starting number of cells/well.
We thank Drs V.Dixit, S.Cory, D.Vaux, J.Tschopp, H.von Boehmer, R.Perlmutter, K.Tomonari, M.Krummel and J.Allison for gifts of transgenic mice, expression vectors, cytokines and monoclonal antibody‐producing hybridoma cells. We are grateful to L.Gibson for generation of the transgene construct, to A.Mifsud and J.De Winter for animal husbandry, to Dr F.Battye, R.Muir and D.Kaminaris for operating the FACS, to M.Stanley and M.Pakusch for expert technical assistance and to J.Tyers for editorial assistance. We gratefully acknowledge Drs S.Cory, M.Grell, D.Vaux and M.Krummel for insightful discussions and Drs S.Cory, J.Adams, K.Shortman, J.Miller, D.Huang and D.Vaux for critical review of the manuscript. K.N. is supported by a Melbourne University PhD scholarship and A.S. is a Scholar of the Leukemia Society of America and a recipient of a Clinical Investigator Award from the Cancer Research Institute. This work was supported by the National Health and Medical Research Council (Canberra), the Anti‐Cancer Council of Victoria and the US National Cancer Institute (CA43540).
- Copyright © 1998 European Molecular Biology Organization