We have developed a strategy to produce a form of the EGF receptor that is activated by ligand binding and suitable for structural, biophysical, and biochemical studies. Steps have been taken to maximize yields and minimize both the time and cost of production as well as heterogeneity arising from glycosylation and proteolysis. In particular, large-scale transient transfection of the EGFR gene into HEK293 GnTi−
cells using PEI as a transfection reagent proved a cost- and time-effective approach that yields ~0.2 mg of purified EGFR per liter of transfected cells and enables efficient removal of N-linked glycosylation using Endoglycosidases H and F. Eliminating the structural and chemical heterogeneity of attached carbohydrates often proves beneficial to growth of diffraction-quality crystals (40
). Heterogeneity in purified EGFR also arose from proteolysis of the C-terminal tail, which proved difficult to avoid completely, and a premature stop codon that eliminated the C-terminal tail was thus introduced immediately following the region encoding the EGFR kinase domain. Coupled with efficient deglycosylation of HEK293 GnTi−
cell-expressed EGFR, truncation of EGFR following the kinase domain allowed purification of a relatively homogenous form EGFR that is activated by ligand binding. As also observed by Springer and colleagues (21
), tEGFR is solubilized well in both Triton X-100 and dodecylmaltoside, is monomeric in the absence of ligand, and is dimeric in the presence of EGF as judged by size-exclusion chromatography.
Additional heterogeneity arose from the formation of unwanted disulfide-linked EGFR oligomers in mildly oxidizing initial purification conditions, presumably mediated by cytoplasmic cysteines that are reduced in normal cell environments. Two approaches, mutagenesis of exposed cytoplasmic cysteines and addition of a small amount of reducing agent, both successfully eliminated unwanted EGFR oligomers. Concerns that a small amount of reducing agent might disrupt the structure and function of the heavily disulfide-linked extracellular region were overcome when the inclusion of 0.5 mM dithiothreitol failed to diminish ligand binding activity or lead to oligomerization through transient reduction of extracellular disulfide bonds. To avoid complications owing to possible effects of cysteine mutations on EGFR enzymatic activity, 0.5 mM dithiothreitol was used during purification to produce native tEGFR for the studies reported here.
Preparation of purified tEGFR allowed quantitative characterization of its kinase activity. Early published reports identified six autophosphorylation sites in the EGFR C-terminal tail but no sites in the kinase region itself (7
). Later studies showed that a tyrosine on the kinase activation loop, Y845, is phosphorylated by the Src kinase and that phosphorylation of Y845 influences EGFR function (34
). Phosphoryation of Y845 is not required for EGFR activity, however (41
). More recently, Arteaga and colleagues showed that Y845 becomes phosphorylated following TGFα stimulation or introduction of oncogenic EGFR mutations and that this Y845 phosphorylation is dependent on EGFR kinase activity (42
). Using purified tEGFR and tandem mass spectrometry we now demonstrate that EGFR itself is capable of phosphorylating Y845. It is well established that phosphorylation of activation loop tyrosines increases the activity of several kinases (43
), and phosphorylation of Y845 seems likely to promote or stabilize EGFR activity and modulate EGFR signaling.
Comparison of the kinase activities of EGF/tEGFR and cetuximab/tEGFR demonstrates that ligand stimulates a ~500-fold increase in catalytic activity (kcat
) and that the effect of ligand binding to tEGFR is predominantly on kcat
rather than substrate Km
s. This observation suggests that substrate binding is likely to be only modestly perturbed in the cetuximab/tEGFR complex. Structural studies on the Abl tyrosine kinase support the idea that interactions with peptide substrate can be maintained despite kinase activation loop and C-lobe conformations that appear inconsistent with catalysis (44
). In this case the orientation of the two substrates and/or placement of key catalytic residues appear distorted from active conformations. A similar mechanism may underlie our observations with EGFR.
Prior quantitative studies of EGFR/ErbB enzymatic activity have primarily utilized soluble intracellular domain fragments or incompletely characterized EGFR (11
). In particular, one recent approach mimicked the high local concentrations achieved in a membrane environment by tethering the EGFR kinase domain to lipid vesicles via a hexa-histidine tag (11
). Stimulation of kinase activity by up to 20-fold were observed following vesicle targeting, consistent with a key role for kinase dimerization in activation.
To our knowledge, our study describes the first detailed kinetic analysis of near full-length EGFR in purified form. By comparing EGF and cetuximab-complexed forms of tEGFR, we have shown a 500-fold activation between autoinhibited and stimulated forms. Elimination of the C-terminal tail may have influenced our results, but this change in activity is some 25-fold greater than the stimulation observed with vesicle activation. This 500-fold differential is also considerably larger than measurements of kinase domains bearing activating mutations (11
). Taken together, these results suggest that kinase-domain only assays incompletely recapitulate full activation/inhibition of the EGFR kinase. The differences observed between isolated kinase regions and a more intact form of the receptor are perhaps not surprising as the ligand-bound extracellular regions and the juxtamembrane regions must contribute to stronger dimerization if not also stereochemically favorable apposition of the kinase domains (47
Of potential clinical interest among the results reported here is the relative inhibitory potencies of the two anti-cancer agents erlotinib and lapatinib, which structural studies have shown bind to active and inactive EGFR kinase conformations, respectively (11
). We observed the expected trend that erlotinib is more effective at inhibiting the ligand-activated EGFR kinase and lapatinib more effective at inhibiting inactive EGFR. This observation suggests the potential for therapeutic synergy between cetuximab and lapatinib. The relative selectivity of erlotinib and lapatinib for their preferred enzyme forms was only 5–10-fold, however, which was less than expected based on the much larger relative difference in kinase activity in the presence and absence of ligand. Either substantial crossover binding of these inhibitors for alternate EGFR kinase conformations occurs, or an unexpectedly high interconversion between active and inactive kinase conformations may occur irrespective of receptor dimerzation. Given the the conditions of our assay and the 7-fold higher Km
for ATP of cetuximab complexed with tEGFR compared with EGF/tEGFR, the Ki
values deduced for lapatinib blocking the inhibited and active enzyme forms are essentially identical. These findings run contrary to the concept of truly conformation-specific EGFR kinase inhibitors and suggest that further investigation of inhibitor-bound conformations of the EGFR kinase may prove illuminating.
A final observation to emerge from our results is that even the ‘inactive’ EGFR conformation observed crystallographically appears capable of catalyzing phosphoryl transfer, albeit at a significantly reduced rate. The differences in substrate Km
s observed for the cetuximab versus EGF bound forms of tEGFR suggest that the residual kinase activity in the antibody-complexed EGFR arises from a different kinase conformation. Had the lower activity of the cetuximab-tEGFR complex resulted solely from shifting the conformational equilibrium of the kinase region away from the active conformation, one would expect to observe lower kcat
s but unaffected Km
s. Although 500-fold lower than the EGF-tEGFR rate, the catalytic power of cetuximab-tEGFR is still more than 10,000-fold greater than the uncatalyzed phosphoryl transfer reaction (49
), indicating substantial transition state-stabilization still occurs in this state.