As a versatile regulatory protein ubiquitously expressed in almost all mammalian cell types and tissues, PEA-15 controls multiple biological processes in a phosphorylation-dependent manner. The protein itself, however, does not seem to have any dedicated enzymatic or biological functions as PEA-15 knock-out (PEA-15−/−) did not appear to affect the health and fertility in mouse models12. As the functions of PEA-15 are all exerted through engaging protein-protein interactions (PPIs), it is interesting to uncover how such a small, non-catalytic protein can modulate various cellular processes and bind to structurally and functionally diverse proteins in the cell, all by modifying its own phosphorylation states. In this study, we looked into the binding modes of PEA-15 with ERK2, a typical MAP kinase protein responsible for cell-cycle entry and cell proliferation22, and FADD, an adapter protein to facilitate the association of DISC and promote the activation of initiator caspase-8 for extracellular death-ligand induced apoptosis18.
ERK2 and FADD do not share any sequence or structural homology, and they perform very distinct biological functions in the cell. The cellular pathways they each involved do not seem to overlap. Structurally, ERK2 has typical kinase fold, with a small N-lobe composed of a five-stranded antiparallel β-sheet and containing the ATP-binding P-loop, and a large C-lobe with six conserved α-helical segments and the catalytic loop containing four short conserved β strands23. FADD is a death superfamily protein and is composed of an N-terminal DED and a C-terminal DD18, both having a six-helix bundled fold but differing in intermolecular interactions19. They both, however, bind to PEA-15 of proper phosphorylation states. When engaged in binding, PEA-15 utilizes both its DED and the C-terminal tail to interact with partner protein through electrostatic interactions. The main binding surface on PEA-15 DED is helices α5 and α6, which is negatively charged, while the surfaces of ERK2 DEF docking site and FADD helices α1 and α6 of its DED are both positively charged. PEA-15 uses the same binding surface on the DED to engage both ERK2 and FADD in the interactions. The in vivo binding specificity is determined by the phosphorylation states of the two serine residues on PEA-15 C-terminal tail. When unphosphorylated, the C-terminal tail appears to be more positively charged due to the basic lysine residues towards the end of the tail. This positively charged tail will preferably interact with the negative patch of ERK2 DRS. When the C-terminal tail is phosphorylated, particularly at Ser-116, it dramatically changes the charge distribution on the tail, making it significantly more negatively charged. The change of the electrostatic property on the tail promotes the preferential binding to the positive patch on the FADD surface of helix α3 of DED. To our knowledge, this study is the first attempt to explain the phosphorylation dependent binding specificity of PEA-15.
In this MD study, together with our previous NMR studies7,17,24, we pointed to a very flexible DED in PEA-15. PEA-15 DED adopts different conformations depending on its phosphorylation and binding status, and is allosterically controlled by its intrinsically disordered C-terminal tail. When the C-terminal tail is recognized by ERK2 DRS, the DED modulates its conformation, including the allosteric exposure of α5/α6 binding surface and relaxation of α3 to facilitate its interaction with the DEF docking site on ERK2. On the other hand, when the C-terminal tail serine residues are phosphorylated, it allosterically modulates DED conformation to prepare it to engage in DED-DED interactions. When it comes to bind with FADD, little conformational change is needed to form the complex.
This simulated PEA-15pp/FADD model agrees well with our initial NMR assessment on the complex, which demonstrated that the phosphorylation of PEA-15 C-terminal serine residues significantly modulate DED conformation, while binding to FADD does not seem to induce additional conformational change of PEA-15pp17. In an effort to partially assign the NMR spectra of PEA-15 S104D/S116D double mutant (PEA-15DD), mimicking the doubly phosphorylated state and having similar in vivo effects and binding specificity to FADD25, we identify that the residues experiencing the most chemical shift perturbations between the free PEA-15DD and the FADD-bound form are mostly located on helices α5 and α6, including Arg-71, which engages in direct interactions with FADD. Other residues exhibiting significant shifts upon FADD binding include H52, I66, F67, I69, S70, and R72 on the DED, and Q114, D116 (mutated from serine to mimic phosphorylation and involved in FADD binding), and E117 on the C-terminal tail (Supplementary Figure S9).
On almost all DED-containing proteins, a prominent surface feature, termed charge-triad, consisting of three charged residues, D/E-RxDL (first charged residue can be either D or E, and x indicates any amino acids), can be identified to form a hydrogen-bonded and electrostatic network19. For PEA-15, we have illustrated the charge-triad network, D19-R72-D74 in our high-definition NMR structure5. However, the role of the conserved charge-triad has not been determined. Although it was stipulated that D74 is involved directly in ERK2-binding, as D74A mutation caused the loss of binding capacity of PEA-15 to ERK24, later crystal structures16 and NMR dynamics study7 both rejected such hypothesis, and we proposed that the charge-triad, through its hydrogen-bonded network, to facilitate conformational change at DED by reducing transition energy between free- and bound-conformations. The current MD simulations confirmed our hypothesis. In free-form PEA-15, D19 is hydrogen-bonded to R72, while in the ERK2-bound form, R72 is hydrogen-bonded to D74, making the switch between free- and bound-conformation swiftly. For the PEA-15pp/FADD complex, the same trend seems to hold. D19-R72 interaction dominates in the free PEA-15pp, while R72-D74 interaction is more prevalent in the bound form. D74A mutation will disrupt the interaction between R72 and D74, which is essential in complex formation. As we pointed out earlier, other surface polar interactions on PEA-15 also contribute to the conformational transition by lowering the transition energies.
Another significant finding in this study is that FADD undergoes a large conformational change upon interaction with PEA-15pp, involving a complete reorientation of the DD relative to the DED. This reorientation between the two domains has significant implication in the DISC formation in responding extracellular apoptotic signals. However, the understanding of the DISC association and the activation of initiator caspase-8 is limited by the lack of a complete DISC structural model. Two crystal structures have been reported for the complexes formed between the DD of the death receptor Fas, and the DD of the adaptor protein FADD, with PDB IDs 3EZQ26 and 3OQ927. The asymmetric oligomeric DD assembly formed between mouse Fas and human FADD DDs (3OQ9) revealed at least three types of binding interfaces in the association of DD complexes, reflecting the complexity of maintaining an intricate balance between the regulatory threshold and responsive sensitivity in triggering apoptosis27. The tetrameric human Fas-FADD DD complex displayed intriguing new features, in which an elongated new stem helix is formed in place of helices α5 and α6, and an additional C-helix is formed in the C-terminus of Fas DD26. The stem helix and the C-helix may promote the association of Fas and formation of a regulatory bridge between Fas molecules. Another effect of shifting helix 6 and fusing with helix 5 to form the stem helix is to expose a hydrophobic patch between helices α1 and α5 of Fas DD to interact with helices α1 and α6 of FADD DD. This crystal structure implies that relative repositioning of the two domains of FADD seems to be necessary to expose caspase-8 binding surface, and to avoid the steric clash with the newly formed C-helix of Fas. Our MD simulation of PEA-15pp/FADD complex provides initial evidence of the repositioning of FADD DED relative to DD, shifting the binding surface at helices 1 and 6 for DED-DED interaction away from clashing with the Fas stem helix and C-helix. The FADD DD conformation in our simulated complex structure agrees very well with the crystal structure 3EZQ, with a backbone RMSD of 1.610 Å. The rotation of DED relative to DD makes the bound PEA-15 DED, particularly helix α6, to avoid the steric hinderance with the stem helix or the C-helix of Fas DD (Fig. 7A). As a comparison, the NMR structure of full-length FADD (2GF5) had a backbone RMSD of 4.334 Å for the DD with crystal structure 3EZQ, and if we assume PEA-15 binding at the same surface at helices α1 and α6 of FADD DED, significant clashing between PEA-15 helix α6 and the stem helix and C-helix can be detected (Fig. 7B). Since it is reasonable to assume that PEA-15 DED and caspase-8 DED may bind to the same surface of FADD DED, the relative reorientation between the two domains of FADD will be crucial in recruiting caspase-8 to the DISC. Our MD model provides an important insight into the DISC formation and caspase-8 activation.