In addition to mutations implicated in disease states, a variety of mutants have been generated in order to define interfaces in the Fas/FADD complex (Supplementary Table 2; reviewed in ^{5, 14}). However, despite this information and effort, the nature of the primary Fas/FADD interaction and its implication on the actual mechanism for formation of the Fas/FADD DD network remains elusive¹⁴. In our study we were able to gain insight into this mechanism by elucidating the crystal structure of the Fas/FADD DD complex from which we conclude a model explaining how observed contacts are mediated by a primary Fas/FADD complex, which acts as a sensitive switch governing DISC formation. This extends and completes our view of death domains, which initiated from rigid defined domain complexes such as the constitutive 1:1 complex seen in the Apaf-1/caspase-9 CARD/CARD interaction, to current views which include plasticity and asymmetry of complexes in the death domain superfamily⁵ (see also Supplementary Discussion). Thus the model presented here can furthermore serve as a template for other signaling platforms lacking enzymatic components which are solely mediated by oligomerisation and clustering events. Notorious for its elusiveness and solubility difficulties¹⁴, we succeeded in producing a soluble Fas/FADD complex when E. coli lysates of recombinantly expressed Fas DD and FADD DD were combined prior to purification. After purification we optimized conditions to obtain crystals of the complex and solved the 2.7 Å resolution structure. Crystals indexed in the hexagonal space group P6₁ with two tetrameric assemblies each comprising four Fas DD and four FADD DD in the asymmetric unit (Fig. 1, Supplementary Table 1). The tetrameric arrangement in fact represents a dimer of two Fas/FADD DD complex dimers. In this arrangement Fas provides all contacts (see also Supplementary Discussion). All residues of the Fas DD are well defined in the electron density. FADD is defined from G93 to G191 with the region around residues 107−149 showing more diffuse density due to a high degree of overall motion caused by a lack of crystal contacts (Supplementary Fig. 1c).

Inspection of the interfaces shows that the actual Fas/FADD complex is formed by helix one and six of the FADD DD through employment of a hydrophobic patch surrounded by polar residues (Fig. 2a, d). Unlike the FADD DD, the Fas DD in the complex deviates from the typical DD fold^17-19. Comparison with the solution structure of the isolated Fas DD²⁰ reveals that Fas in the complex has undergone an opening in which helix six has shifted and fused with helix five to form a long helix we term the stem helix (Fig. 2b, c). In addition to Fas opening, a new helix at the C-terminus of Fas dubbed the C-helix is observed (Fig. 2a). The main interaction interface on Fas is formed by hydrophobic residues that become available only upon Fas opening (Fig. 2d). Thus the opening of Fas discovered in the crystal structure is a crucial event in its interaction with FADD. Significantly, other FADD interacting death receptors show similar residues as those observed in Fas at the complex interface (see Supplementary Fig. 4c and Discussion).

Broader investigation of the Fas/FADD interface divulges the disperse nature of the interaction surface, which lacks the presence of defined and focused interaction sites often referred to as hot spots^21-23. This phenomena is used in regulatory complexes characterized by weak binding despite sufficiently large interaction surfaces^{22, 24, 25}. In addition to the lack of hot spots, a loss of compact tertiary structure due to the opening of Fas further enhances flexibility of the system and thus additionally weakens the complex. Besides providing the primary interface for binding of FADD, a second consequence of the opening of Fas is to allow for interaction with another open Fas molecule forming a Fas/Fas bridge. The predominant Fas/Fas association observed in the structure is formed by stem-helices and C-helices of the two Fas molecules building a Fas/Fas dimer (Fig. 2e), which compared to the tetramer, better approximates structural requirements in the context of the natural membrane association (see Supplementary Discussion). Thus opening of Fas provides both the basis for FADD binding and the formation of a regulatory bridge between Fas molecules. Due to the difficult nature of the Fas/FADD complex, biochemical analysis is limited (Supplementary Fig. 2a), but indeed confirms a weak primary Fas/FADD complex which dissociates at concentrations as high as 50 μM (Fig. 2f, Supplementary Fig. 2d). Mutational analysis of the Fas/FADD interface reveals predominately mild phenotypes (Supplementary Fig. 3a-c, and data not shown), in line with previous observations for complexes lacking binding hot spots such as IFNalpha2/IFNAR1²⁶.

In solution, the Fas/FADD DD complex indeed shows a tetrameric behavior (Supplementary Fig. 2b-d), which is further corroborated by previously reported results of a solubility enhanced complex which also pointed to a tetrameric arrangement²⁷. Taken together these results establish that the observed tetramer in the crystal structure reflects the Fas/FADD DD arrangement present in solution and is not a result of crystal formation. In addition the tendency of the tetramer to form higher oligomers can be observed (Supplementary Fig. 2 c, d).

To obtain further insight into the binding of FADD, we overlaid the structure of full length FADD¹⁸ onto the DD of FADD in the primary complex (Fig. 3a). The drastic conformational change in Fas, which is essential to bind FADD, is also accompanied by a change in the FADD death domain. The C-terminal helix of FADD has to shift in order to avoid a steric clash with the newly formed C-helix of Fas, substantiating results which attribute a negative role in cell death induction to this region of Fas (Fig, 3a; Supplementary Fig. 3d)²⁸. This observation is even more crucial in the context of full length FADD, since the FADD C-terminus is located at the interface of the DED and DD. Along this line, an overall conformational rearrangement, namely the relative position of the DED and the DD of FADD, can also be concluded to expose caspase-8 binding residues determined in earlier studies (summarized in ^{5, 18}). This change is readily enabled by the interdomain linker, which is composed exclusively of Ala, Gly and Pro residues. Indeed our studies found that when full length FADD is incubated with Fas a reduced initial complex formation can be observed, underlining a conformational change (Fig. 3b), while longer incubation of the two proteins lead to the formation of protein clusters (Fig. 3c).

Details revealed in the Fas/FADD death domain structure show that Fas opening is key to both FADD binding and formation of a Fas/Fas bridge. To validate the relevance of Fas opening in vivo, we sought to create a mutation that interfered with the closed form of Fas and propagated Fas opening. If Fas opening is a prerequisite for DISC formation, this forced opening of Fas should produce a hyperactive phenotype. By comparing the structures of closed and open Fas we identified Ile313 as a candidate. Ile313 is part of the hydrophobic core and resides in helix 6 of the closed form and is thus at the center of the opening event. At the same time it does not directly participate in the Fas/FADD interaction or the observed Fas/Fas dimer in the complex structure (Fig. 3d, Supplementary Fig. 4a). Indeed mutation of Ile313 to Asp resulted in significantly enhanced killing. Enhanced cell death by Fas I313D compared to wild type could be observed when cells were stimulated with Fas antibody or FasL and even in the absence of a stimulus, further establishing the central role of Fas opening (Fig. 3e).

Based on the analysis of the Fas/FADD structure which displays key regulatory features of DISC formation, and taking into account previous findings, we conclude a model of how receptor signaling in the DISC occurs solely through clustering events. In this model the opening of Fas is central to form the Fas/Fas bridge and to recruit FADD. The Fas/Fas bridge and FADD binding are governed by weak protein:protein interactions only stabilized upon processive clustering, thus representing the actual regulatory switch governing DISC formation (Fig. 4). Until now, the dominant model for DISC formation proposed that trimeric Fas ligand cross-linked units of preformed trimeric Fas in local membrane rafts^8-14, leading to higher order arrangements. This model is substantiated by mutations in the Fas DD (Supplementary Table 2; reviewed in ^{5, 14}) that are not located at the primary Fas/FADD interface, but rather adjacently on the globular units of open Fas observed here. While illustrating the overall process, this initial model did not provide a mechanism for the actual clustering of Fas DDs within the cell and most importantly the relationship between Fas clustering and FADD binding, both of which are explained in our model. In this model only when a sufficient number of Fas molecules are in close proximity, as is expected upon Fas ligand binding, can the open forms of Fas become stabilized (Fig. 4a) since the newly formed stem helices can interact with each other. An interlinked consequence of this interaction is that Fas molecules become connected through this sensitive bridge - setting the stage for a chain reaction. Now the formation of Fas bridges linking the Fas DD trimeric units, which interact via the globular portion of open Fas, directly result in rapid and processive clustering on the inside of the cell membrane (Fig. 4b). Also, FADD is now able to bind the open Fas molecules, key for DISC signaling, and additionally increasing stabilization of the Fas/Fas bridges thereby further fostering DISC formation and clustering (Fig. 4b).

In summary the proximity induced stabilization of open Fas, in combination with the formation of the weak bridge, allows timing and strict regulation of DISC formation solely by clustering as opposed to, for example, a nucleotide exchange mediated oligomerisation as in the apoptosome²⁹.

1