CD95 mutations in mice and humans are associated with defective lymphocyte apoptosis and autoimmunity. The precise molecular mechanism(s) by which these mutations cause reduced apoptosis has not been defined previously. Gene mutations may affect mRNA or protein expression, stability or function. In this study, we examined the molecular mechanisms responsible for loss of CD95-mediated apoptosis, using mutant alleles cloned from eight different patients with heterozygous CD95 mutations. The mutations, affecting either the ICD (P1, P2, P5, P7, P8, P10) or ECD (P4, P6, P11) of CD95, were all associated with markedly reduced CD95-mediated apoptosis of activated T cells. However, different mechanisms appeared to be responsible for the defective apoptosis associated with ICD versus ECD mutations.
ICD mutant alleles were readily detected at the cell surface, indicating that they did not have an appreciable effect on protein expression. However, when they were transfected into Jurkat T cells, which constitutively express WT CD95 and are highly sensitive to CD95-mediated apoptosis (30
), all of the ICD mutants caused a striking reduction in CD95-mediated apoptosis. These data confirm findings from functional studies of similar CD95 mutations in other patients (19
). Ligand-induced clustering of CD95 recruits the adapter molecule, FADD, to the intracytoplasmic death domain of CD95 (3
). This results in the recruitment and activation of FLICE/Mach/caspase-8 (33
) and FLICE2/caspase-10 (35
), initiating sequential caspase activation, intracellular proteolysis, and cell death. We therefore examined the ability of CD95-mutant proteins to recruit FADD, by cotransfection and IP assays using 293T cells, which do not constitutively express CD95. P1, P5, and P10 have truncations of the CD95 death domain, and failure of these mutant proteins to bind FADD was anticipated. Alleles P2 (D244Y), P7 (D244G), and P8 (R234P) had amino acid substitutions of highly conserved residues in subdomains α2 (R234P) and α3 (D244Y, D244G) of the death domain. It is known that the amino acids glycine or proline within a peptide subdomain greatly reduce the probability of α-helix formation (36
). Secondary structure analysis predicted that the death domain substitutions, R234P and D244G, broke the α2 and α3 helices respectively, whereas D244Y did not. Because D244 is solvent-accessible and forms part of the interface involved in CD95 protein–protein interaction (31
), D244 may be a necessary contact residue for FADD binding. When P2, P7, or P8 were coexpressed with FADD, they all failed to recruit FADD. These findings were consistent with the secondary structure predictions and with previous alanine-scanning mutagenesis of the CD95 death domain, which revealed that R234A or D244A abolished both CD95–CD95 and CD95–FADD death-domain interactions in vitro
To determine whether the ICD mutants could prevent binding of FADD to heterotypic aggregates of mutant and wild-type CD95, we titrated mutant CD95 against a constant amount of WT CD95 in a cotransfection study. Co-IP of P10 and WT CD95 demonstrated that the mutant caused a dose-dependent inhibition of FADD binding to the receptor complex. Knockout studies have shown that FADD is absolutely required for CD95-mediated apoptosis (38
), so these observations provide a sufficient explanation for the loss of CD95-mediated apoptosis in patients with ICD mutations.
Of the three ECD mutants — P4, P6, and P11 — two (P6, P11) cause premature truncation within the first 50 amino acids of the protein and were not expressed on the cell surface. Indeed, we found little or no intracellular expression of the P6 peptide in transfectants, suggesting that it was rapidly degraded. Although cell-surface expression of the P4 mutant could not be detected by a panel of anti-CD95 monoclonal antibodies that bind to CRD1, CLB-95/2, a CRD2-specific antibody, revealed cell-surface expression of P4. Whereas the absolute intensity of staining was reduced compared to WT CD95, the relative intensity of intracellular to extracellular staining was similar to that observed with WT CD95 (A.K. Vaishnaw and K.B. Elkon, unpublished observations). These findings suggest that reduced intensity of staining is explained by alteration in the conformation of CRD1 and CRD2 caused by substitution of cysteine at residue 66. Reduction of cysteine residues in the ECD of CD95 has been shown to markedly attenuate anti-CD95 MAB binding (40
), and we have shown previously that the C66R structural alteration leads to loss of binding to CD95L (1
Despite the evidence for lack of expression of P6, we examined whether ECD mutants could interfere with the function of the wild-type allele. Neither of the patients who have ECD mutants, P4 or P6, had a dominant–negative effect on wild-type CD95 function in Jurkat T cells. Cascino et al
) have, however, reported that activated T cells can secrete alternatively spliced CD95 variants that are lacking the transmembrane domain and that are truncated in CRD1, -2, and/or -3. The products of these alternatively spliced forms apparently inhibited anti-CD95 mediated apoptosis. In our studies, supernatants from P6 transfected cells failed to block CD95L-mediated apoptosis, and neither the P4 nor P6 CD95-Fc proteins bound CD95L. Because the P11 mutation causes a truncation immediately after the signal peptide, it is difficult to conceive of any interfering mechanism, even if the peptide is synthesized. In sum, all of the evidence indicates that the ECD mutations are associated with a loss of function rather than a dominant–negative mode of action.
Although lymphadenopathy frequently becomes less severe with age (16
), conceivably leading to underestimation of phenotype in parents of those affected, we compared the penetrance of CSS in families with ICD and ECD mutations. Consistent with our results demonstrating that ICD mutations were dominant–negative, we observed that these alleles were associated with a more highly penetrant CSS syndrome. Analysis of the extended pedigrees for P8, P9, and P10 revealed that P8 and P9 also had additional individuals outside the nuclear families who were affected with a CSS-like syndrome (Vaishnaw, A.K., et al
., manuscript in preparation). Indeed, for P8, the family history for CSS was striking, and the pedigree had been reported previously as containing seven individuals with an autosomal dominant form of “idiopathic hereditary splenomegaly and hypersplenism” (25
). In contrast to ICD mutations, families with an ECD mutation had only one individual affected with CSS, although other family members carried the mutation. In the P4 pedigree, two siblings and the father of the index case were unaffected carriers, and the father, two paternal uncles, and the paternal grandfather of P6 were all unaffected carriers of an ECD mutation.
Although it is clear that ICD mutations are not always associated with CSS, previous reports of ICD and a small number of ECD mutations are consistent with reduced penetrance of ECD mutations (16
). Only one family has been reported with an ECD mutation and multiple affected members, but all of these were compound heterozygotes for an ECD and an ICD mutation (21
). These observations are similar to studies of heterozygous CD95 mutations in mice, in which lprcg
(ICD mutation) shows greater penetrance than the lpr
allele (loss of function mutation) (41
). Together, these findings demonstrate a remarkable complexity in the nature, penetrance, mechanisms of action, and clinical expression of CD95 mutations in the outbred human populations.