Extracellular domain mutations of the EGF receptor differentially modulate high-affinity and low-affinity responses to EGF receptor ligands

Abstract

The EGF receptor is mutated in a number of cancers. In most cases, the mutations occur in the intracellular tyrosine kinase domain. However, in glioblastomas, many of the mutations are in the extracellular ligand binding domain. To determine what changes in receptor function are induced by such extracellular domain mutations, we analyzed the binding and biological response to the seven different EGF receptor ligands in three common glioblastoma mutants—R84K, A265V, and G574V. Our data indicate that all three mutations significantly increase the binding affinity of all seven ligands. In addition, the mutations increase the potency of all ligands for stimulating receptor autophosphorylation, phospholipase Cγ, Akt, and MAP kinase activity. In all mutants, the rank order of ligand potency seen at the wild-type receptor was retained, suggesting that the receptors still discriminate among the different ligands. However, the low-affinity ligands, EPR and EPG, did show larger than average enhancements of potency for stimulating Akt and MAPK but not receptor autophosphorylation and phospholipase Cγ activation. Relative to the wild-type receptor, these changes lead to an increase in the responsiveness of these mutants to physiological concentrations of ligands and an alteration in the ratio of activation of the different pathways. This may contribute to their oncogenic potential. In the context of recent findings, our data also suggest that so-called “high”-affinity biological responses arise from activation by isolated receptor dimers, whereas “low”-affinity biological responses require clustering of receptors which occurs at higher concentrations of ligand.

The EGF receptor is a classical receptor tyrosine kinase composed of a ligand-binding extracellular domain and an enzymatically active, intracellular tyrosine kinase domain (1). The extracellular ligand-binding domain of the EGF receptor is comprised of four subdomains. In the absence of a ligand, the extracellular domain adopts a closed conformation that is tethered through an interaction between the dimerization arm in subdomain II and the tethering arm in subdomain IV (2). Ligand is bound between subdomains I and III. The binding of ligand releases the tether, allowing the receptor to open and form a back-to-back dimer mediated via the dimerization arm in subdomain II (3, 4). Extensive data suggest that the receptor also forms higher-order oligomers that play a significant role in its function (5, 6, 7, 8).

Upon binding the agonist ligand, dimerization of the extracellular domain of the EGF receptor induces the formation of an active, asymmetric dimer of the intracellular kinase domain (9). In the asymmetric kinase dimer, the “activator” kinase from one receptor monomer allosterically activates the “receiver” kinase from the other receptor monomer. The active receiver kinase then phosphorylates the C-terminal tail of the “activator” kinase on tyrosine residues (10). These phosphotyrosines serve as sites for the binding of SH2 and PTB domain-containing proteins, leading to the activation of downstream signaling pathways (11). Some downstream signaling responses, such as Akt or MAP kinase activation, occur at concentrations of EGF much lower than those required to elicit other responses, such as phospholipase Cγ and Stat activation (12). The molecular basis for the existence of these “high affinity” versus “low affinity” responses to EGF is not understood.

The EGF receptor recognizes and binds seven different ligands that exhibit a range of affinities from low nM to μM (13). Despite the fact that these ligands bind to the same site on the EGF receptor (2, 3, 14), they can induce different biological outcomes in the same cell (15, 16, 17, 18, 19). This has been associated with differences in the kinetics of downstream signaling stimulated by the different ligands (14, 20) as well as differences in the conformation of the receptor dimer induced by the different ligands (21, 22, 23).

The EGF receptor is mutated or over-expressed in a variety of tumors, including those of breast, lung, colon, and brain (24, 25, 26, 27, 28). Most tumorigenic mutations of the EGF receptor occur in the kinase domain and lead to ligand-independent activation of the tyrosine kinase (29, 30, 31). However, in glioblastoma, a large fraction of the EGF receptor mutations occur in the extracellular domain (26, 32, 33). The principal variant is the EGFR-vIII, which contains an in-frame deletion of residues 6 to 273 in the extracellular domain (32, 33, 34). This deletion precludes the binding of EGF to the receptor and leads to constitutive kinase activity (35). Also common in glioblastoma are several point mutations in the extracellular domain, including R84K, A265V, and G574V, that result in enhanced tumorigenicity of the receptor (26). How these mutations affect ligand binding and downstream signaling within the EGF receptor has not been delineated.

In this study, we show that these three glioblastoma mutations result in an enhancement in the affinity of the receptor for all seven EGF receptor ligands. These affinity increases are associated with enhancements in the potency of the ligands for the stimulation of several downstream signaling events. As a result, full activation of a biological response in the mutants is obtained at a dose of ligand that yields only partial activation of the wild-type receptor. Because the ligand-specific potency changes vary between biological responses, the relative ratio of the activation of the downstream signaling pathways is different in the mutants as compared to the wild-type receptor. The mutations also change the behavior of the low-affinity ligands with respect to the induction of “high-affinity” versus “low-affinity” signaling events. In the context of recent reports (5, 14, 21, 36), the data are consistent with a model in which “high-affinity” biological responses are mediated by isolated receptor dimers while “low-affinity” responses are mediated by clustered receptors. Together, these alterations may contribute to the oncogenic potential of these mutations.

Results

EGF binding by the mutant receptors

The three EGF receptor point mutations characterized in this study are R84K, A265V, and G574V. R84K and A265V are in subdomains I and II, respectively, and form part of the interface between these two subdomains. The interaction between R84 and A265 must be disrupted in order to form the symmetric extracellular EGF receptor dimer induced by high-affinity receptor agonists (21). G574V lies in subdomain IV and functions as part of the tether that stabilizes the unoccupied, closed form of the receptor (2).

CHO cells were transfected with plasmids encoding EGF receptors carrying one of these three point mutations. Stable clones were selected and ¹²⁵I-EGF saturation binding experiments were performed. CHO cells expressing the wild-type EGF receptor served as the control in all experiments. As shown in Figure 1, the wild-type receptor bound EGF with an EC₅₀ of 0.75nM. Because of the heterogeneity in the affinity of EGF for the monomeric and dimeric forms of its receptor (37), this EC₅₀ represents an average of the multiple affinities of the receptor for EGF. As shown in Figure 1B, the R84K-EGF receptor bound EGF with ∼20-fold higher affinity (EC₅₀ = 36 pM). The A265V-EGF receptor bound EGF with approximately 3-fold higher affinity (EC₅₀ = 280 pM) while the G574V-EGF receptor exhibited an approximately 5-fold higher affinity for EGF (EC₅₀ = 170 pM). Thus, all three mutations produced receptors that bound EGF with a significantly higher affinity than the wild-type receptor, with R84K yielding the greatest increase in affinity.

Figure 1.

¹²⁵I-EGF binding to wild type and mutant receptors. CHO cells expressing wild-type, R84K-, A265V-, or G574V-EGF receptors were plated in 6-well dishes and assayed for ¹²⁵I-EGF binding as described in Experimental procedures. Panels show direct saturation binding curves. Insets show the Scatchard transformation of the saturation binding data. Points represent the mean ± standard deviation of triplicate determinations. Diagram at top shows the positions of the three mutants in the EGF receptor structure.

Although the mutant EGF receptors were expressed from a tetracycline-inducible plasmid, they did not express at levels much above ∼50,000 receptors per cell. It was therefore not possible to characterize EGF binding over the wide range of receptor expression levels necessary to allow quantitation of the level of linkage and cooperativity exhibited by the receptors (37). We therefore transformed the saturation binding data via Scatchard analysis to probe for the presence of negative cooperativity in the mutants. The results of these analyses are shown in the insets in Figure 1. As is well documented (37, 38, 39, 40, 41, 42), the wild-type EGF receptor exhibits negative cooperativity, in this case, evidenced by an upwardly concave Scatchard plot. The Scatchard plots for the A265V-EGF receptor and the G574V-EGF receptor were also upwardly concave, demonstrating the presence of negative cooperativity in those receptors, as well. In contrast to what was observed for the other mutants, the Scatchard plot for the R84K-EGF receptor was linear, indicating that this mutation ablates the negative cooperativity that is characteristic of the wild type EGF receptor.

We have previously reported that EGF receptor-selective tyrosine kinase inhibitors modulate the affinity of the receptor for EGF (43). In particular, the type 1 inhibitor, erlotinib (44), increases the affinity of the wild-type receptor for EGF. The type 2 inhibitor, lapatinib (45), decreases the affinity of the receptor for EGF. These observations demonstrate the presence of a transmembrane allosteric connection that links the conformation of the intracellular kinase domain to the ligand-binding properties of the extracellular domain. Given the changes in EGF binding observed among the mutant EGF receptors, we next probed for this allosteric connection by assessing the effect of erlotinib and lapatinib on the binding of EGF to the mutant receptors. The results are shown in Figure 2.

Figure 2.

Effects of erlotinib and lapatinib on the binding of¹²⁵I-EGF binding to wild type and mutant receptors. CHO cells expressing wild-type, R84K-, A265V-, or G574V-EGF receptors were plated in 6-well dishes and incubated with 5 μM erlotinib or 10 μM lapatinib for 1 h at 37° C prior to assay for ¹²⁵I-EGF binding. Erlotinib or lapatinib were included in the overnight binding incubation as described in Experimental procedures. Panels A, C, E, and G are from cells expressing wild type, R84K-, A265V-, and G574V-EGF receptors, respectively, and treated with erlotinib. Panels B, D, F, and H are from the same cells treated with lapatinib. Note that the control curve in panel B is the same as shown in Figure 1A. The insets in panels C and D for the R84K-EGF receptor show the saturation binding curve for ¹²⁵I-cetuximab in cells treated without or with erlotinib or without or with lapatinib. Points represent the mean ± standard deviation of triplicate determinations.

As reported previously (43) and as shown in Figure 2, A and B, erlotinib significantly enhanced the affinity of the wild-type receptor for EGF and lapatinib substantially reduced the affinity of the receptor for EGF. A similar pattern was observed in cells expressing the A265V-EGF receptor, though the effect of both erlotinib and lapatinib was smaller than what was observed in the wild-type receptor (Fig. 2, E and F). A slightly different result was seen in cells expressing the G574V-EGF receptor (Fig. 2, G and H). In this case, erlotinib failed to enhance the affinity of the receptor for EGF but lapatinib was still able to reduce the affinity of the receptor for its ligand.

The effects of erlotinib and lapatinib on the binding of EGF to cells expressing the R84K-EGF receptor were unexpected (Fig. 2, C and D). As was seen for the G574V-EGF receptor, erlotinib failed to increase the affinity of the R84K-EGF receptor for EGF. Surprisingly, however, treatment with erlotinib, increased the apparent number of receptors accessible to EGF for binding by approximately 30%. This is an “apparent” increase as parallel experiments measuring the effect of erlotinib treatment on the binding of the anti-EGF receptor monoclonal antibody, cetuximab, failed to detect a change in receptor number or affinity (Fig. 2C, inset).

As was true for the other two mutants, treatment of cells expressing the R84K-EGF receptor with lapatinib led to an ∼2.5-fold decrease in receptor affinity. However, lapatinib reduced the apparent receptor number at saturation by about one-third. Again, this change in receptor number was not detected by ¹²⁵I-cetuximab binding (Fig. 2D, inset).

EGF signaling through the mutant receptors

The ability of EGF to stimulate downstream signaling pathways was next compared in cells expressing wild-type or mutant EGF receptors. This included the measurement of receptor autophosphorylation as well as the activation of phospholipase Cγ, Akt, and MAP kinase via Western blotting for the activated forms of these enzymes. For these experiments, the number of EGF receptors expressed in the cells was titrated using doxycycline to give equivalent levels of wild-type or mutant receptors on the cell surface as assessed by the binding of ¹²⁵I-cetuximab. The results are shown in Figure 3.

Figure 3.

Comparison of signaling in cells expressing wild type or mutant EGF receptors. Cells expressing the wild-type, R84K-, A265V-, or G574V-EGF receptor were grown overnight in six-well dishes and matched for cell surface receptor expression based on the binding of ¹²⁵I-cetuximab. Receptor levels are reported as receptors per cell and shown above the phosphotyrosine blots. Cells were stimulated for 5 min with the indicated doses of EGF and lysates were prepared an analyzed for autophosphorylation, phospholipase Cγ, Akt, and MAP kinase activation as described in Experimental procedures.

Cells were stimulated for 5 min with increasing doses of EGF and the activation of downstream signaling assessed by Western blotting. The left panel in Figure 3 shows a comparison of signaling in cells expressing wild type and R84K-EGF receptors. EGF-stimulated tyrosine phosphorylation of the R84K-EGF receptor was similar to that of the wild type EGF receptor. Likewise, the R84K-EGF receptor mediated activation of phospholipase Cγ, Akt, and MAP kinase was similar in extent to that seen in cells expressing wild type EGF receptors. The level of EGF-stimulated autophosphorylation of the A265V-EGF receptor was also comparable to that seen in cells harboring the wild-type EGF receptor (Fig. 3, center panel). And the stimulation of phospholipase Cγ, Akt and MAP kinase by EGF in cells expressing the A265V-EGF receptor was essentially the same as in cells expressing the wild-type receptor. Thus, in these cells, these mutations did not markedly affect the maximal level of stimulation of each of these responses.

In cells expressing the G574V-EGF receptor (Fig. 3, right panel), EGF-stimulated autophosphorylation and phospholipase Cγ activation were lower than that seen for the wild-type receptor. However, activation of Akt and MAP kinase was the same or only modestly reduced compared to what was seen in cells expressing the wild-type receptor.

To further characterize receptor autophosphorylation, site-specific anti-phosphotyrosine antibodies were used to assess the pattern of phosphorylation of these receptors over times ranging from 0 to 60 min (Fig. 4). Sites examined were Tyr-1045, Tyr-1068, Tyr-1086, Tyr-1148, and Tyr-1173. Phosphorylation at Tyr-992 was too low yield consistent results. There was no gross difference in the utilization of sites between wild type and mutant receptors. There was, however, a difference in the time course of phosphorylation. In the wild-type receptor, phosphorylation at all sites tested peaked at 2 min, declined by 15 min, and was nearly gone by 60 min. By contrast, in all three mutants, phosphorylation peaked at 2 min for all sites, but held constant for 15 min, and only declined to basal levels by 60 min. Thus, autophosphorylation was somewhat prolonged in the mutants as compared to the wild-type receptor.

Figure 4.

Site-specific phosphorylation of wild type and mutant EGF receptors. Cells expressing the wild-type, R84K-, A265V-, or G574V-EGF receptor were grown overnight in six-well dishes. After stimulation with 30 nM EGF for the indicated times, RIPA lysates were prepared and analyzed by Western blotting for phosphorylation on Tyr-1045, Tyr-1068, Tyr-1086, Tyr-1148, Tyr-1173 or total phosphotyrosine (PY20).

Ligand selectivity in mutant EGF receptors

The EGF receptor can bind seven different ligands (13). Five of these ligands, EGF, TGFα, betacellulin (BTC), heparin-binding EGF (HB-EGF), and amphiregulin (AREG), are high-affinity ligands with binding EC₅₀ values in the low- to mid- nM range. Two of the ligands, epiregulin (EPR) and epigen (EPG), are low affinity ligands with binding EC₅₀ values in the low μM range (13, 46). As mutations in the extracellular ligand-binding domain could have an effect on the ability of the receptor to recognize the different ligands, we used competition binding studies to determine whether ligand discrimination by the mutant EGF receptors was the same as that for the wild- type receptor. The results of this analysis are shown in Figure 5.

Figure 5.

Competition binding by EGF receptor ligands in cells expressing wild-type and mutant EGF receptors. Cells expressing the wild type, R84K-, A265V-, or G574V-EGF receptor were grown overnight in six-well dishes and subjected to competition binding with increasing concentrations of each ligand. Points represent the mean ± standard deviation of triplicate determinations.

Based on the IC₅₀’s, the wild-type EGF receptor bound its ligands in order of affinity: HB-EGF>EGF∼BTC>TGF>AREG>EPG∼EPR (Fig. 5A). The R84K-EGF receptor (Fig. 5B), A265V-EGF receptor (Fig. 5C), and the G574V-EGF receptor (Fig. 5D) all exhibited a similar order of affinity for the seven ligands with HB-EGF showing the highest affinity and EPR showing the lowest affinity. However, relative to what was seen for the wild-type receptor, there was an increase in affinity for all of the ligands in all three mutants. The increases were generally in the range of 5- to 10-fold, with the exception of EPG, which showed an increase of 27-fold at the R84K-EGFR and 13-fold at the G574V-EGF receptor, compared to its affinity at the wild-type receptor. Thus, there was some evidence for a difference in ligand binding by the mutants, but the rank order of affinities was not changed. This indicates that the mutations did not substantially alter the ability of the receptors to discriminate among the seven different ligands. The IC₅₀ values for each ligand in each cell line are provided in Table S1.

Figure 6 compares the dose–response curves for the seven different ligands for stimulating receptor autophosphorylation in all four cell lines. Maximum autophosphorylation stimulated by the different ligands was similar within a given cell line so the mutations did not selectively affect the strength of the signal elicited by any particular ligand (Fig. S1).

Figure 6.

Autophosphorylation stimulated by the seven ligands in cells expressing wild-type or mutant EGF receptors. Cells expressing the wild type, R84K-, A265V-, or G574V-EGF receptor were grown overnight in six-well dishes and stimulated for 5 min with increasing doses of each of the seven EGF receptor ligands. Lysates were prepared and analyzed for autophosphorylation as described in Experimental procedures. Points represent the mean and standard error from 3 to 5 separate experiments for each growth factor. The left and right vertical dotted lines mark the EC₅₀ values for EGF and EPR, respectively, to highlight the leftward shift of these parameters in the cells expressing the mutant receptors.

The graphs in Figure 6 show the average normalized dose–response curves from three to five independent experiments for each ligand. The dotted lines delineate the EC₅₀ for EGF (a high-affinity ligand) and EPG (a low-affinity ligand) for stimulating autophosphorylation in cells expressing the wild-type EGF receptor. This helps to highlight the shift in potency of the different ligands in cells expressing wild type versus mutant EGF receptors. As can be seen from the graph for cells expressing the wild-type receptor, there were basically two groups of ligands: four high potency ligands (EGF, TGF, HB-EGF, and BTC) with EC₅₀ values in the 10 nM range and two lower potency ligands (EPR and EPG) with EC₅₀ values in the 1 μM range. AREG was intermediate between these two groups. While showing more compaction among the high affinity ligands, the rank order potency for stimulating autophosphorylation approximated the rank order of binding affinities exhibited in cells expressing the wild-type receptor.

The pattern for stimulating receptor autophosphorylation in cells expressing the three EGF receptor mutants basically recapitulated what was seen for the wild-type receptor. However, all the ligands were on average about 5- to 10-fold more potent for stimulating autophosphorylation than in cells expressing the wild-type receptor. This is consistent with the observed increase in affinity of the ligands in the different mutants. The largest shifts occurred for AREG at the R84K receptor and TGF at the A265V receptor and were 12- and 13-fold, respectively. Nonetheless, the rank order of potency seen in the wild-type receptor was retained in each of the mutants.

The ability of wild-type and mutant receptors to mediate the activation of downstream signaling events was next assessed. Preliminary experiments demonstrated that, in any given cell line, all ligands stimulated similar levels of activation of phospholipase Cγ, Akt, and MAP kinase (Fig. S1). Figure 7 compares the dose response curves for each of the seven ligands for stimulating these three responses in cells expressing the four different receptors. The EC₅₀ values for each ligand for stimulating each signal as well as autophosphorylation are provided in Table S1. The dotted lines in Figure 7 mark the EC₅₀ values for EGF and EPG for stimulating phospholipase Cγ activity in that particular cell line. As discussed below, this promotes visualization of the differences between the behavior of high- and low-affinity ligands in cells expressing wild-type versus mutant EGF receptors.

Figure 7.

Downstream signaling stimulated by the seven ligands in cells expressing wild type or mutant EGF receptors. Cells expressing the wild type, R84K-, A265V-, or G574V-EGF receptor were grown overnight in six well dishes and stimulated for 5 min with increasing doses of each of the seven EGF receptor ligands. Lysates were prepared and analyzed for activation of phospholipase Cγ, Akt, and MAP kinase as described in Experimental procedures. Points represent the mean and standard error from 3 to 5 separate experiments for each growth factor. The vertical dotted lines mark the EC₅₀ values for EGF and EPR for stimulating phospholipase Cγ activity in that specific cell line. This helps to visualize the difference in the behavior of low affinity versus high affinity ligands in cells expressing the wild type EGF receptor as opposed to the mutant EGF receptors.

The graphs in the top row of Figure 7 correspond to the response in cells expressing the wild-type EGF receptor. As can be seen from these graphs, the rank order of potency of the seven ligands for stimulating these three responses was similar to that for ligand binding, with EGF or HB-EGF being the most potent and EPR and EPG being the least potent. However, there were substantial differences in the potencies of the different ligands for stimulating the different responses.

The five higher affinity ligands were 10- to 25-fold more potent for stimulating the activation of Akt and up to 10-fold more potent for stimulating MAP kinase activity than they were for inducing the activation of phospholipase Cγ. By contrast, the two low-affinity ligands, EPR and EPG, showed roughly the same potency for stimulating phospholipase Cγ activation as they did for stimulating Akt and MAPK activation. These observations are consistent with the findings of Krall and Macbeath (12) who previously reported that Akt and MAP kinase were activated at low concentrations of EGF whereas phospholipase Cγ required much higher doses of EGF for full activation. This has given rise to the concept of “high-affinity” and “low-affinity” biological responses to EGF (12).

Cells expressing the glioblastoma mutations showed several significant changes from this pattern. First, most of the ligands exhibited increased potency for stimulating these responses, as might be expected based on the ligand binding data. However, while the order of potency of the ligands was largely retained, the enhancements in potency were highly variable, ranging from 3-fold to over 100-fold when compared to their potency at the wild-type EGF receptor. The enhancements in potency were generally larger for EPR and EPG than for the high-affinity ligands. Alone among the ligands, the potency of HB-EGF for stimulating these downstream responses was not increased in any of the mutants relative to the wild-type receptor. As a result, HB-EGF was no longer the most potent ligand for stimulating these three responses.

Similar to the cells expressing the wild-type receptor, in cells expressing the R84K- or A265V-EGF receptor, the high-affinity ligands were at least an order of magnitude more potent for stimulating Akt than phospholipase Cγ and ∼4- to 20-fold more potent for stimulating MAP kinase than phospholipase Cγ. Unlike in wild-type cells, however, the low-affinity ligands, EPR and EPG, also demonstrated this behavior in these two mutant cell lines. Activation of Akt and MAP kinase by EPR and EPG generally occurred at concentrations of ligand one to two orders of magnitude lower than those required for stimulating phospholipase Cγ. Thus, the low-affinity ligands exhibited the phenomenon of “high-affinity” and “low-affinity” responses that were restricted to the high-affinity ligands in cells expressing the wild-type receptor.

Similar findings were observed in cells expressing the G574V-EGF receptor. However, the difference in potency for stimulating Akt or MAP kinase versus phospholipase Cγ averaged only about 5-fold for the high-affinity ligands. The low affinity ligand, EPG, also showed a 5-fold greater potency for stimulating Akt and MAP kinase than phospholipase Cγ. However, EPR exhibited the roughly the same potency for stimulating all three responses.

Discussion

Effects of mutations on EGF binding and signaling

We report here that three mutations in the extracellular domain of the EGF receptor that are commonly found in glioblastomas lead to distinct alterations in the binding and signaling properties of the EGF receptor. All three mutations led to an increase in the affinity of the receptor for EGF, with the R84K mutation having the largest effect. Alone among the mutations, the R84K mutation also ablated the negative cooperativity that is characteristic of the EGF receptor (37, 47). Hu et al. (21) recently reported the structure of the extracellular domain of the R84K-EGF receptor. The structure revealed contacts between R84, A265, L38, and F263 that are present in the wild-type symmetric extracellular dimer but absent in the R84K receptor extracellular dimer. The authors suggested that the absence of these interactions is the basis for the lack of negative cooperativity in the R84K mutant. These contacts are not fully broken in the A265V mutant (21), which, in our experiments, continued to exhibit negative cooperativity. Our data are therefore consistent with the involvement of these interactions in generating negative cooperativity in the EGF receptor.

Besides altering the affinity of the receptor for EGF, the mutations also altered the ability of erlotinib and lapatinib to modulate receptor binding affinity. We have shown previously (43) that type 1 inhibitors, such as erlotinib, that bind to and stabilize the active conformation of the kinase domain (44), enhance the affinity of the wild-type receptor for EGF. Type 2 inhibitors such as lapatinib, that bind to and stabilize the inactive conformation of the kinase domain (45), reduce the affinity of the receptor for EGF. The increase in affinity seen with erlotinib can be attributed to the tendency of the active, asymmetric kinase dimer to induce the untethered, open conformation of the extracellular domain (48). This form of the receptor would have a higher affinity for EGF as the tether between subdomains II and IV in the extracellular domain would not need to be broken in order for the ligand to bind. The inactive conformation of the kinase domain does not induce dimerization of the receptor (48, 49) and may stabilize the closed, tethered form of the extracellular domain, resulting in decreased affinity for ligands.

Both the R84K and G574V mutations abolished the ability of erlotinib to increase the affinity of the receptor for EGF. While erlotinib retained the ability to enhance EGF affinity in the A265V-EGF receptor, the size of the effect was significantly smaller than that seen in the wild-type receptor (3-fold versus 17-fold, respectively). This suggests that these mutations may cause the unoccupied extracellular domain to more frequently adopt the open, higher affinity conformation that is linked to the active asymmetric, intracellular kinase domain.

Lapatinib retained its ability to induce a decrease in the affinity of EGF in all three mutant receptors. This suggests that this inhibitor retains its capacity to induce the inactive form of the kinase domain and stabilize the tethered structure of the extracellular domain. Interestingly, tumors harboring these extracellular domain mutations are more sensitive to lapatinib than erlotinib (50). The ability of lapatinib to retain its capacity to decrease receptor affinity in these mutants may be related to this clinical phenomenon.

Surprisingly, erlotinib was found to increase and lapatinib was found to decrease the apparent number of R84K-EGF receptors bound by ¹²⁵I-EGF. These changes are only apparent as they were not observed when ¹²⁵I-cetuximab was used to quantify EGF receptor number. The basis for this finding is unclear but could be related to the formation of an oligomeric form of the unliganded EGF receptor. Zanetti-Domingues et al. (51) have provided evidence for the existence of unliganded EGF receptor oligomers that are comprised of linear chains of monomers in a head-to-head configuration that precludes the binding of EGF to the interior subunits. The oligomers were thought to aid in suppressing the basal activity of the receptors and could be disrupted by treatment of the cells with erlotinib. Presumably, the linear chains would also be disrupted by ligand binding, which like erlotinib, induces asymmetric kinase dimer formation. If so, the R84K-EGF receptor may form oligomers that are more resistant to disruption by the binding of EGF than oligomers of the wild-type EGF receptor. By helping to disrupt the unliganded R84K-EGF receptor oligomers, erlotinib would make those sites more accessible for EGF binding. Lapatinib would stabilize the oligomers, reducing the availability of sites to which EGF could bind.

Despite the fact that the three extracellular domain mutations are reportedly transforming (26), they did not lead to elevated levels of receptor autophosphorylation and downstream signaling as compared to that seen in cells expressing equivalent levels of wild-type receptors. In fact, autophosphorylation and phospholipase Cγ activation were lower in the G574V cells. The mutations did prolong the time course of receptor autophosphorylation in cells expressing all three mutants, indicating that there was an underlying change in the balance of kinase to phosphatase activity in those cells.

We speculate that the absence of an increase in maximal responses is due to the utilization by the cells of a variety of mechanisms to suppress aberrant signaling. First, none of the three mutants could be expressed at high levels in the CHO cells, despite the fact that they were on a doxycycline-inducible plasmid. This suggests that the cells were limiting the expression of, or rapidly degrading, the mutant receptors. This may have allowed the cells to regulate receptor signaling using normal mechanisms of desensitization and/or downregulation. Second, as outlined above, the R84K-EGF receptor, and possibly the other mutant receptors, may have formed unliganded oligomers that served to suppress receptor activation. Finally, if the levels of the effector enzymes are limited in CHO cells, they may be maximally activated by wild-type receptors, leaving no capacity for increasing the extent of activation through the mutant receptors. Thus, depending on the mechanisms available to a particular cell for controlling kinase activity and signaling, the effect of these mutations on receptor function may differ in different cell types.

Ligand bias in the mutant receptors

The EGF receptor binds seven different ligands, some of which have a high affinity and some of which have a low affinity. Competition binding studies using ¹²⁵I-EGF demonstrated that all seven EGF receptor ligands exhibited increases in affinity in the three EGF receptor mutants. Affinity shifts varied among ligands but were, on average, about 5- to 10-fold. Despite these changes, the rank order of binding affinities observed for the wild-type EGF receptor was retained in the mutant receptors. Thus, these mutations did not selectively alter the recognition and binding of any particular ligand.

Consistent with the increase in binding affinity of the ligands in the mutant receptors, the potency of all seven EGF receptor ligands for stimulating receptor autophosphorylation was also increased. The average increase was in the range of 5- to 10-fold in the different mutants. This differs from the observations of Hu et al. (21) who reported no change in the potency of EGF for stimulating autophosphorylation of the EGF receptor but an ∼8-fold increase in potency for EPR. Those authors used an anti-pTyr-845 antibody to follow receptor autophosphorylation while we used a pan-phosphotyrosine antibody. It is possible that limiting the measurement of autophosphorylation to what occurred at a single site failed to detect the increase in EGF potency observed using the pan-phosphotyrosine antibody. Hu et al. (21) suggested that the mutations compromised the ability of the receptor to discriminate between high- and low-affinity ligands. Our data support a different interpretation.

For autophosphorylation as well as all three of the downstream signaling pathways, the ability of the mutant receptors to discriminate among the ligands was retained, as demonstrated by the spread of the dose–response curves shown in Figures 6 and 7. The enhancement of potency for stimulating enzyme activation was generally greater for the low-affinity ligands than for the high-affinity ligands but the relative rank order of potency of the seven ligands was largely conserved in all three mutants. However, the low-affinity ligands did exhibit one distinctive change in their signaling pattern.

In cells expressing the wild-type EGF receptor, the five high-affinity ligands were significantly more potent for stimulating Akt activation, and to some extent MAPK kinase activation, than for stimulating autophosphorylation and phospholipase Cγ activity. By contrast, the low-affinity ligands, EPR and EPG, had nearly the same EC₅₀ values for stimulating Akt, MAP kinase, autophosphorylation, and phospholipase Cγ activation. Thus, the high-affinity ligands exhibited the characteristic “high affinity” and “low affinity” type of responses previously reported (12) while the low-affinity ligands did not. This difference between the behavior of high and low-affinity ligands was essentially erased in the three mutant receptors. In the mutant receptors, all ligands, not just the high-affinity ligands, exhibited the phenomenon of “high affinity” and “low-affinity” responses.

We hypothesize that this change is related to mutation-induced increases in the strength of the receptor dimers formed by the low-affinity ligands as well as the ability of low-affinity ligands to induce receptor clustering. Freed et al. (14) reported that EPR and EPG induce weak dimerization of wild-type EGF receptors, compared to that induced by EGF. Similarly, Mundumbi et al. (36) used single particle tracking to demonstrate that while EGF induced significant clustering of its receptor, EPR induced only a minimal reduction of the EGF receptor diffusion coefficient, reflecting a reduced ability to elicit receptor clustering. Thus, low affinity ligands are compromised in their ability to induce both dimerization and oligomerization of the wild-type EGF receptor. Mutations in the extracellular domain of the EGF receptor alter this situation. Hu et al. (21) reported that, like EGF, EPR induced strong dimers in receptors bearing the R84K or A265V mutations. And Mudumbi et al. (36). reported that EPR was effective at inducing clustering of the R84K-EGF receptor.

While oligomerization of the EGF receptor has long been noted (6, 7, 52), the structure of activated EGF receptor oligomers is unclear. Huang et al. (5) have presented data supporting a model in which an oligomerization surface is present on subdomain IV of the extracellular domain of the receptor. This allows formation of receptor multimers in which the kinase domains are daisy-chained together to allow activation of all kinases in the multimer, rather than the single kinase that is activated in a receptor dimer. These authors showed that this multimerization was necessary to elicit strong phosphorylation of tyrosine residues in the proximal portion of the EGF receptor C-terminal tail.

Phospholipase Cγ activation requires phosphorylation of the receptor on Tyr-992 (53, 54). As this residue lies in the proximal part of the tail, the model of Huang et al. (5) predicts that activation of phospholipase Cγ would require clustering of the receptor. Akt activation (via PI 3-kinase binding to Gab1 (55, 56)) and MAP kinase activation (57, 58, 59) can be achieved using more distal tyrosines that can be efficiently phosphorylated in isolated receptor dimers. Thus, the ‘high-affinity’ responses to EGF (and presumably to other high affinity ligands) may be due to the ability of these ligands to induce strong, kinase-active dimers at low concentrations of ligand. The ‘low-affinity’ responses to EGF may require clustering of the receptors which occurs only at higher concentrations of ligand. Because EPR and EPG fail to stimulate strong, kinase-active receptor dimers, they cannot elicit the ‘high-affinity’ responses that arise from isolated receptor dimers. Instead, clustering is required for effective stimulation of kinase activity by these ligands, so all responses show the same ‘low-affinity’ dose–response relationship.

In the extracellular domain mutants, EPR and EPG induce strong receptor dimers (14) so effectively stimulate kinase activity in isolated receptor dimers, giving rise to the ‘high-affinity’ Akt and MAPK responses. Clustering of the receptors, which occurs at higher concentrations of ligand, would still be required to achieve phospholipase Cγ activation, resulting in a ‘low-affinity’ response for activating this enzyme. So, in the mutants, all ligands signal like high-affinity ligands.

Overall, these extracellular domain mutations had two significant effects on cell signaling. First, they enhanced the potency of six of the seven ligands (all but HB-EGF) for stimulating downstream signaling responses, with the low-affinity ligands showing greater enhancements than the high-affinity ligands. Second, the mutations enabled low-affinity ligands to exhibit the phenomenon of ‘high-affinity’ versus ‘low-affinity’ responses. Both of these effects substantially increase the responsiveness of these receptors to physiological concentrations of the ligands when compared to the response mediated through the wild-type receptor. As a result, signaling in the mutants will be quantitatively greater at a given concentration of ligand than in cells expressing the wild-type receptor. In addition, the mutations alter the relative extent of activation of the different downstream signaling pathways. This leads to an altered ratio of signaling through the various pathways, likely resulting in a change in the integrated response of the cell to any given ligand. An example of this is provided in Fig. S2 which compares the relative degree of activation of phospholipase Cγ, Akt, and MAP kinase in cells expressing wild-type or the mutant EGF receptors when stimulated by the same fixed concentration of a ligand. Together, these differences could contribute to the oncogenic potential of these mutants.

Experimental procedures

Materials

The CHO-K1 Tet-On cell line, the pBI-Tet vector, and doxycycline were from Clontech. EGF and amphiregulin were from Gold Biotechnology. Epigen, epiregulin, and betacellulin were from Prospec. HB-EGF was from Sigma. TGFα was from Abcam. Erlotinib was from Selleck Chemicals and lapatinib was from VWR. The PY-20 antiphosphotyrosine antibody was from BD Transduction Labs. The site-specific anti-phosphotyrosine antibodies plus the anti-T669 antibody were from Cell Signaling Technology. The phospholipase Cγ, phospho-phospholipase Cγ, Akt, phospho-Akt, and phospho-MAP kinase antibodies were also from Cell Signaling Technology. The MAP kinase antibody was from Transduction Labs. FetalPlex was from Gemini Bioproducts. Cetuximab and pertuzumab were obtained from the Barnes-Jewish Hospital pharmacy. Na¹²⁵I was from Perkin Elmer. ¹²⁵I-EGF and ¹²⁵I-cetuximab were made using the ICl method of Contreras et al (60).

DNA constructs and selection of stable cell lines

The R84K-, A265V-, and G574V-EGF receptor mutants were constructed using the following primers with wild type EGF receptor as template:

R84K Forward: CTGCAGATCATCAAAGGAAATATGTAC,

R84K Reverse: GTACATATTTCCTTTGATGATCTGCAG,

A265V Forward: AAATACAGCTTTGGTGTCACCTGCGTGAAGAAG,

A265V Reverse: CTTCTTCACGCAGGTGACACCAAAGCTGTATTT.

G574V Forward: GCCCGGCAGTAGTCATGGGAG.

G574V Reverse: CTCCCATGACTACTGCCGGGC.

After PCR and sequencing to confirm accuracy, the constructs were inserted into the pBI Tet vector between the NheI and DraIII cut sites. The resulting plasmids were co-transfected into CHO-K1 Tet-on cells with the pTK-Hyg selection vector. Stable clonal lines were selected by growth in 500 μg/ml hygromycin. Expression of the EGF receptor was confirmed through ¹²⁵I-cetuximab binding and Western blotting.

Cells and tissue culture

CHO cells were maintained in Dulbecco’s Modified Eagle’s Medium (DMEM) containing 10% FetalPlex, 10 units/ml penicillin, 10 μg/ml streptomycin, 100 μg/ml G418 and 100 μg/ml hygromycin. For experiments, cells were plated on lysine-coated 6-well dishes 24 h prior to use. Doxycycline was added at the time of plating at a concentration that gave approximately 20,000 to 40,000 receptors per cell. For experiments in which signaling mediated via the wild type and mutant EGF receptors were compared, the concentration of doxycycline was adjusted, as necessary, to achieve equivalent cell surface expression of wild type and mutant receptors. Cell surface receptor levels were assessed using ¹²⁵I-cetuximab.

Receptor autophosphorylation and downstream signaling

CHO cells stably expressing wild-type or mutant EGF receptors were plated and cultured in the appropriate concentration of doxycycline. After 24 h, cultures were switched to F12 containing 2.5 μg/ml pertuzumab, 25 mM HEPES pH 7.2, and 1 mg/ml bovine serum albumin at 37° for 60 min to serum-starve the cells and block heterodimerization of the EGF receptor with the few ErbB2 molecules present in CHO cells. The indicated concentration of growth factor was then added to each well. After incubation at 37° for the indicated time, cells were washed in ice-cold phosphate-buffered saline (PBS) and RIPA lysates were prepared. Equal amounts of total protein were analyzed by SDS polyacrylamide gel electrophoresis and subsequently blotted for phosphotyrosine, MAP kinase, phospho-MAP kinase, Akt, phospho-Akt, phospholipase Cγ and phospho-phospholipase Cγ. When necessary, Western blots were quantified using Image J and the data fitted to a three-parameter dose-response curve using GraphPad Prism 9.

¹²⁵I-EGF binding

For experiments assessing EGF affinity, CHO cells were washed with ice-cold PBS and chilled on ice for 20 min. Cells were then incubated for 16 h at 4° C in triplicate with ice cold DMEM containing 10 mg/ml bovine serum albumin, 50 mM HEPES, pH 7.2, 5 μg/ml pertuzumab, and 2 pM to 32 nM ¹²⁵I-EGF. For experiments involving erlotinib or lapatinib, the 16 h incubation included 5 μM erlotinib or 10 μM lapatinib. At the end of the incubation, the cells were washed three times in cold PBS and the monolayer was dissolved in 1 ml 1 N NaOH. The NaOH was transferred to tubes and counted for ¹²⁵I in a Beckman Gamma Counter.

Binding data were analyzed using GraphPad Prism 9. Non-specific binding was determined from a fitted curve and subtracted from the total binding. Specific binding data were fit to a one-site or two-site specific binding model, whichever was determined by Prism to be the statistically significantly better fit. Scatchard plots were generated using the Prism built-in analysis for transforming direct binding data into a Scatchard plot.

Studies examining the binding of the six other EGF receptor ligands were carried out via ¹²⁵I-EGF competition binding. Cells were incubated as above in a medium containing 50 pM ¹²⁵I-EGF in the presence of increasing doses of the unlabeled ligands. Binding data was analyzed using Prism 9 and its One-site Fit log IC₅₀ Competitive Binding equation.

¹²⁵I-Cetuximab binding

Cell surface EGF receptor levels were assessed by incubating triplicate wells of a 48-well dish with 1 nM ¹²⁵I-cetuximab. This represents a saturating concentration of cetuximab. Incubations were for 2 h at room temperature. Non-specific binding was defined by the addition of 50 nM unlabeled cetuximab to replicate wells. Cells were washed and counted as for ¹²⁵I-EGF binding as outlined above for ¹²⁵I-EGF binding.

Data availability

All data are contained within the manuscript and/or supporting information.

Supporting information

This article contains supporting information.

Conflict of interest

The authors declare that they have no conflicts of interests with the contents of this article.

Reviewed by members of the JBC Editorial Board. Edited by Wolfgang Peti