Our studies provide evidence that low-affinity EGF receptors play crucial roles in cell-fate decisions. EGF titrations demonstrate the existence of two distinct sets of cytoplasmic signaling proteins: one that is activated by low concentrations of EGF when only a small fraction of EGFR molecules are active, and a second, comprising PLCγ1, Stat1, Stat3 and Stat5, that is only activated by higher concentrations of EGF. These results, in conjunction with the observed affinities and ratios of high- and low-affinity receptors, strongly suggest that although many signaling pathways can be activated by high-affinity EGF receptors, PLCγ1 and the Stat proteins depend on low-affinity receptors for their activation. In addition to the concentration-dependent signaling profiles, changes in cell morphology and rates of proliferation coincide with the activation of low-affinity receptors, supporting a role for these receptors in determining phenotypic outcome. Some of the cell lines used in this study maintain autocrine loops that activate EGFR even under serum-starved conditions. While this prohibits the complete elimination of negative feedback loops, the experimental setup using exogenous ligand demonstrates that even in the presence of a basal level of signaling, EGF receptors have distinct functions that are based on their ligand affinity. Accordingly, our data are very similar when cells are grown in the presence of serum, demonstrating that our findings hold true even in a more complex signaling environment.
It has been known for some time that cells exhibit different phenotypic responses to high and low concentrations of EGF. It is also well known that EGFR appears to exist in two different forms: a high-affinity form with an apparent KD of ≈300 pM, and a low affinity form with an apparent KD of ≈2 nM. To date, however, these two observations have not been linked to signaling outputs, as high- and low-affinity EGF receptors have never been shown to elicit specific and distinct intracellular signaling events. Here, we provide evidence that low-affinity EGF receptors activate a distinct set of intracellular signaling proteins (the Stat proteins and PLC-γ) and that the concentration of EGF or TGFα at which this occurs coincides precisely with the point at which different phenotypic outcomes are observed. To date, all mathematical models of EGFR signaling assume that the activated receptor exists in a single form and turns on a specific set of signaling proteins. These models are unable to predict the different phenotypic outcomes that are observed at different ligand concentrations. Our study provides evidence that low-affinity receptors turn on specific and distinct signaling pathways and argues strongly that predictive models of EGFR signaling should take low-affinity receptors into account.
While we would have liked to inhibit each form of the receptor independently, unfortunately there is currently no way to selectively perturb either the high- or low-affinity population of receptors, as they are both encoded by the same transcript. As there was no way to selectively perturb either population, we relied on extremely rigorous and quantitative analyses to observe coincidence between the three phenomena we were studying: ligand binding, activation of signaling proteins, and phenotypic outcome. These studies were performed in a variety of cell lines, both normal and transformed, that exhibited a range of receptor expression levels; they were performed with two different ligands (EGF and TGFα); and they were performed in the presence and absence of serum to control for environmental factors. We went to great lengths to ensure that these observations are not isolated phenomena, but are indeed intrinsic to the receptor, independent of its immediate environment.
It is unclear whether the average concentration of EGF in any tissue ever approaches the concentration required to activate low-affinity receptors. However, in cases of autocrine signaling, the effective concentration of EGFR ligands in the immediate vicinity of cell-surface receptors likely exceeds this threshold. Squamous cell carcinomas of the head and neck often rely on the activation of Stat3 for proliferation and survival 
. Stat3 activity in these cancer cells has been shown to depend on autocrine activation of EGFR by secreted TGFα 
. The fact that Stat3 is only activated by low-affinity receptors in every cell type that we examined suggests that in vivo
concentrations of EGFR ligands can stimulate low-affinity receptors and identifies a possible role for low-affinity receptors in the in vivo
signaling of cancer cells.
Very recently, structural studies of the extracellular ligand-binding domain of Drosophila
EGFR have supported negative cooperativity in ligand binding 
. The authors showed that the first ligand binds with high affinity and induces a conformational change that promotes asymmetry in the dimer. The conformational change restrains the vacant binding site such that its affinity for binding the second ligand is reduced. High- and low-affinity binding sites therefore occur in the same receptor dimer and result from negative cooperativity rather than from distinct populations of receptor. The authors further argue that the second binding event must compromise either ligand-receptor or receptor-receptor contacts, and that therefore a doubly occupied dimer could have different interactions and signaling properties than a singly occupied one. Although this asymmetry has not been observed in the extracellular domain of human EGFR, a similar mode of regulation remains possible. If high- and low-affinity interactions do arise from negative cooperativity, singly-occupied dimers should be most abundant at low concentrations of ligand, and doubly occupied dimers should only form at higher concentrations of ligand. It is therefore possible that the Stat proteins and PLCγ1 can only be activated by doubly occupied dimers that have altered specificity, autophosphorylation or interactions.
EGFR has been extensively studied over the past three decades, and several recent analyses have provided system-level views and models of signaling downstream of the receptor 
. These studies, however, have not accounted for the distinct signaling properties of high- and low-affinity receptors. In addition to the biological implications, our findings should benefit computational efforts to model this signaling network and predict cellular outcomes in response to diverse stimuli.