As mentioned above, some early studies pointed to a model in which the EGFR kinase domain is normally autoinhibited, but is activated by an inter-molecular allosteric interaction upon ligand-induced dimerization of the receptor [4
]. More compelling evidence for this model came from recent studies using purified kinase domain of EGFR [26
]. These experiments showed that the isolated kinase domain is monomeric in solution and possesses low catalytic activity. The activity dramatically increased when the local concentration of the kinase domain was increased by attaching it to the surface of lipid vesicles. These experiments strongly suggest that an intermolecular interaction between the isolated kinase domain induced by the high local concentration converts the kinase from the CDK/Src-like inactive conformation to the active conformation.
Three structures of the EGFR kinase domain, obtained under different crystallization conditions, showed the active conformation in the same crystal form [15
]. The inactive conformation was seen only in crystals either with ligands that prevent activation (lapatinib or Mig6, see below) or with a specific mutation that prevents the formation of the active crystal form (see below) [22
]. The persistent active conformation in crystals appears to contradict the fact that the kinase domain is stable in the inactive state in solution. This apparent paradox can be explained by the existence of a specific intermolecular interaction between the kinase domain molecules in the crystal lattice which converts the kinase to the active conformation. This provides a link between the structural and biochemical data for understanding the kinase domain activating mechanism. In both cases, high local concentrations promote the specific intermolecular interaction that is required for activating the kinase.
Analyses of structures in the active conformation revealed two prominent dimers in the crystal lattice; one is symmetric and the other asymmetric [26
]. The symmetric dimer is mediated by the interaction between a fragment in the C-terminal tail which is sandwiched between the N-lobe of its own kinase domain and the C-lobe of the dimer partner. Mutational analyses showed that this symmetric dimer is not required for ligand-induced activation of EGFR. However, it may contribute to fine tuning of EGFR activity by exerting an autoinhibitory effect, as suggested by Landau et al
The asymmetric dimer is formed between the bottom of the C-lobe of one kinase monomer (monomer B) and the top of the N-lobe of the other (monomer A) (). It is worth pointing out that an earlier computational study, carried out in the absence of any direct structural information on EGFR, suggested several dimer models, one of which is an asymmetric dimer similar to this crystallographic dimer [42
]. The interaction between monomer B and A resembles that between cyclin A and active cyclin dependent kinase 2, with the C-lobe of monomer B taking the position of cyclin A in engaging the N-lobe of the kinase partner, although the structure of the C-lobe of the EGFR kinase is completely unrelated to that of cyclin [26
]. This asymmetric dimer interaction is incompatible with the CDK/Src-like inactive conformation of the kinase due to large conformational difference in the N-lobe, especially helix αC, of the kinase domain. Taken together, these observations led to a model for the activation of the EGFR kinase domain in which monomer B in the asymmetric dimer acts as a cyclin-like allosteric activator for monomer A.
Mutational analyses confirm the critical role for the asymmetric dimer in the activation of EGFR, both in the context of full length receptor in cells and the isolated kinase domain in the lipid vesicle-based assay [26
]. For example, a Val924 to arginine mutation, which disrupts the C-lobe face of the asymmetric dimer interface but is far away from the kinase active site, abolishes both ligand-induced autophosphorylation of the full length receptor and lipid vesicle-induced activation of the isolated kinase domain [26
]. This Val924Arg mutant kinase domain has been crystallized with an ATP analogue, AMP-PNP, which shows the CDK/Src-like inactive conformation [26
]. The fact that a single point mutation located far away from the active site leads to crystallization of the EGFR kinase in the CDK/Src-like inactive conformation strongly supports that the CDK/Src-like conformation is the preferred inactive state of the kinase domain, and the active conformation seen in the original crystal form is dependent on the asymmetric dimer interface.
The asymmetric dimer interface is dominated by a helix-helix packing interaction between helix αH of monomer B and helix αC of monomer A, which keeps helix C in the active conformation [26
]. The interface buries a large hydrophobic surface area, the core of which is contributed mainly from the hydrophobic patch alongside of helix αC that is largely buried in the CDK/Src-like conformation but exposed in the active conformation (). Therefore, the asymmetric dimer stabilizes the active conformation at least in part by compensating for the free energy penalty associated with the exposure of the hydrophobic patch in the active conformation.
Sequence analyses show that the asymmetric dimer interface is conserved in the two other catalytically active members in the family, ErbB2 and ErbB4, suggesting that ErbB2 and ErbB4 are likely to use the same activation mechanism. This is confirmed by a recent structural study showing that ErbB4 also forms an asymmetric dimer essentially identical to that of EGFR and the dimer is important for ErbB4 activation [28
]. The conserved asymmetric dimer interface also underlies the ability of different members in the EGFR family to form heterodimers to activate one another [44
]. An exception is ErbB3, which shows high sequence homology to other members in the family at the C-lobe face of the dimer interface but not at the N-lobe face. Unlike other members in the family, ErbB3 is a catalytically inactive kinase with several key residues in the active site mutated. The conserved C-lobe face allows ErbB3 to function as a cyclin-like activator for other members in the family through heterodimerization, nicely explaining the functional role of this catalytically dead kinase. The lack of conservation on the N-lobe face of ErbB3 is likely due to loss of selective pressure, since ErbB3 does not need to assume the position of monomer A (the kinase monomer that is activated).