Threshold-controlled ubiquitination of the EGFR directs receptor fate

Abstract

How the cell converts graded signals into threshold-activated responses is a question of great biological relevance. Here, we uncover a nonlinear modality of epidermal growth factor receptor (EGFR)-activated signal transduction, by demonstrating that the ubiquitination of the EGFR at the PM is threshold controlled. The ubiquitination threshold is mechanistically determined by the cooperative recruitment of the E3 ligase Cbl, in complex with Grb2, to the EGFR. This, in turn, is dependent on the simultaneous presence of two phosphotyrosines, pY1045 and either one of pY1068 or pY1086, on the same EGFR moiety. The dose-response curve of EGFR ubiquitination correlate precisely with the non-clathrin endocytosis (NCE) mode of EGFR internalization. Finally, EGFR-NCE mechanistically depends on EGFR ubiquitination, as the two events can be simultaneously re-engineered on a phosphorylation/ubiquitination-incompetent EGFR backbone. Since NCE controls the degradation of the EGFR, our findings have implications for how the cell responds to increasing levels of EGFR signalling, by varying the balance of receptor signalling and degradation/attenuation.

Figure 1

Analysis of EGFR ubiquitination. (A) HeLa or NR6 cells were stimulated with the indicated concentrations of EGF for 2 min (in this and all subsequent figures). IP and IB were performed as indicated (Ub, ubiquitin P4D1 antibody). (B) HeLa cells were stimulated with EGF, followed by IB with the indicated antibodies. (C, D) Lysates of HeLa cells stimulated with EGF, as indicated, were subjected to ELISA, forward approach (Supplementary Figure 2A), using the indicated detecting Ab (Ub, FK2 antibody). Results are shown as % of max (see Materials and methods). (E) Lysates of HeLa cells stimulated with EGF, as indicated, were subjected to ELISA, reverse approach (Supplementary Figure 2B), using the indicated detecting Ab (Ub, FK2 antibody). (F) Comparison of the EGFR ubiquitination and phosphorylation curves of HeLa cells obtained by forward and reverse ELISA. In all panels (and in all subsequent figures), error bars indicate s.d. calculated on at least three independent experiments. P-values were calculated using two-way ANOVA analysis. When comparing curves that showed significant differences (in all figures), we show the relative P-values; when comparing curves that did not show significant differences (in all figures), we display R, the Pearson correlation coefficient. Source data for this figure is available on the online supplementary information page.

Figure 2

SILAC-MS for quantitative analysis of ubiquitinated and phosphorylated EGFR. (A) Schematic representation of the SILAC-MS approach. HeLa cells were grown in SILAC-encoded ‘light’ or ‘heavy’ media (Supplementary Experimental Procedures). ‘Light’ (L) cells were stimulated with 100 ng/ml of EGF; ‘heavy’ (H) cells were treated independently with increasing concentrations of EGF, as indicated. Cells were then harvested and mixed (H/L) in 1:1 ratio for each pair. (B) Lysates from the seven H/L mixtures were subjected to anti-EGFR IP and SDS–PAGE. Lanes were cut (shown by red lines) starting from the position of the EGFR (asterisk), to cover potential differentially ubiquitinated forms. (C) Left, LTQ-FTICR mass spectra of Ub (UBC, peptide 11–27, left) and EGFR (peptide 81–98, right) from each H/L mixture (a more detailed representation is in Supplementary Figures 3B and C). (D) Threshold ubiquitination of EGFR, detected by MS. Top, high-accuracy quantification of total EGFR (87) and Ub (13) peptides; see Supplementary Table 1 for raw data. Bottom, comparison of EGFR-Ub data obtained with forward ELISA (Figure 1C) and SILAC-MS (from top panel). R, Pearson correlation coefficient. (E) Top, SILAC ratios of the EGFR ubiquitination site (K692-Ub) shown in Supplementary Figure 3G. Bottom, mean SILAC ratios of EGFR phosphorylation were calculated on the basis of the three pY sites shown in Supplementary Figures 3D–F. Note that, while mean pY increases linearly upon EGF stimulation (R²=0.97, square of correlation coefficient; see also Supplementary Figures 3D–F for the linear behaviour of single pY sites), the abundance of the EGFR-Ub peptide increases with a threshold behaviour, similarly to total Ub (panel D). SILAC ratios are calculated using MaxQuant (see Supplementary Figures 3H–K for more detailed pictures).

Figure 3

The threshold effect for EGFR ubiquitination occurs at the PM. (A) Top, HeLa cells were subjected to dynamin 2-KD and treated for 2 min with EGF at the indicated concentrations. IP and IB were as shown. (B) EGFR internalization kinetics in dynamin 2-KD cells was measured using ¹²⁵I-EGF at low (1 ng/ml) or high (30 ng/ml) EGF concentrations. Results are expressed as the internalization rate constant (Ke, left panel) or as % of Ke in control cells (right panel), and are the mean of triplicate points (s.e.m.<8%). Dynamin 2-KD (Dyn 2-KD) severely impaired EGFR internalization both at low and high EGF concentrations, reducing rates to background levels. Similar background levels have previously been observed by us in clathrin-KD+filipin-treated HeLa cells, in which both CME and NCE are inhibited (Sigismund et al, 2005, 2008). These results confirm that both CME and NCE of the EGFR are dynamin 2 dependent. Comparable results were obtained with two different silencing oligos for dynamin 2 (data not shown). (C) Lysates of HeLa cells, control and dynamin 2-KD, stimulated with EGF for 2 min at the indicated concentrations were subjected to ELISA, forward approach (Supplementary Figure 2A), using anti-Ub (FK2) and anti-pY as detecting antibodies. Results are shown as a percentage of the maximal tyrosine phosphorylation or ubiquitination (% of max, see Materials and methods). Graph error bars indicate s.d. calculated on at least three independent experiments. All P-values were calculated using two-way ANOVA analysis. As shown the Ub curves were not significantly different between control and KD; the same was true for the pY curves. Conversely, the Ub curves were significantly different from the pY curves. Source data for this figure is available on the online supplementary information page.

Figure 4

The EGFR–Cbl interaction is threshold controlled. (A) Q-PCR of c-Cbl, Cbl-b and Cbl-c in HeLa cells. Both Ct values (threshold cycles) and mRNA level of c-Cbl, Cbl-b and Cbl-c (normalized on 18S mRNA and expressed as fraction of c-Cbl mRNA) are reported. (B) HeLa cells were subjected to c-Cbl and Cbl-b-KD, alone or in combination (Contr, HeLa cells transfected with control oligo). IB was as shown (Tub, tubulin; loading control). (C) HeLa cells, transfected with the indicated oligos as in B, were stimulated with EGF as shown. Lysates were subjected to IP and IB as shown. For the Ub blots: l.e., long exposure; s.e., short exposure. Note that two different oligos targeting c-Cbl and Cbl-b were used, with comparable results. In panel B and C, results obtained with UTR1 (for both c-Cbl and Cbl-b) are shown (see Materials and methods for details). (D) Top, HeLa cells were treated with EGF as indicated for 2 min and then IP and IB as shown. Bottom, quantitative assessment. Results are expressed as a percentage of the maximal amount (% of max, see Materials and methods) of EGFR that coimmunoprecipitates (Co-IP) with c-Cbl (from now on Cbl), Grb2 or Shc. Source data for this figure is available on the online supplementary information page.

Figure 5

Models describing the generation of EGFR-Ub threshold. (A) Top, schematic representation of EGFR ubiquitination and EGFR phosphorylation, as a function of ligand concentration. x_T represents the half-maximal EGF dose for EGFR ubiquitination (i.e., the ubiquitination threshold) and it is used to separate in the pictograms underneath (dashed line) the events occurring at low EGF (left) from those occurring at high EGF (right). In the inset, the various symbols used in the models are shown. Various models potentially accounting for the EGFR ubiquitination threshold (in all models the ubiquitination of EGFR by Cbl is indicated by a solid arrow line). 1) Threshold phosphorylation model. The model contemplates that the phosphorylation of individual Cbl-binding sites on EGFR (pY1045 or one between Y1068 and Y1086) increases in a sigmoidal fashion with the doses of EGF. The model is depicted for pY1045, but it could be equally applied to the indirect (Grb2-mediated) binding site(s) (pY1068/pY1086). 2) Threshold Cbl activation model. The model contemplates that the enzymatic function of Cbl is activated in a nonlinear fashion by signalling events (e.g., direct tyrosine phosphorylation of Cbl by the EGFR, indicated by a dashed arrow line) that occur only under high EGF. 3) Competition model. This model invokes the existence of a high affinity, rate-limiting (low amount) competitor X. At low EGF (left), such competitor—that in the model would bind only to activated EGFR—prevents Cbl from interacting with the EGFR or from ubiquitinating the receptor (in this latter case, either directly inhibiting Cbl activity or masking Ub sites on the EGFR, not shown). At high EGF (right), the competitor becomes limiting and Cbl could therefore bind and ubiquitinate the EGFR. 4) Cooperative model. Cbl/Grb2 complex binds stably to EGFR only when pY1045 and at least one of pY1068 and pY1086 are present in the same EGFR molecule. In this case, the EGFR phosphorylation pattern determines the ubiquitination threshold. At low EGF (left), EGFR is poorly phosphorylated and the probability of having the two key sites in the same EGFR molecule is low (possible low-affinity binding of the Cbl:Grb2 complex to single sites is shown by a dotted line). However, this probability increases at high EGF (right) allowing for the cooperative recruitment of Cbl/Grb2. This model implies that phosphorylation sites are phosphorylated independently of one another (as shown experimentally in Figure 7C and Supplementary Figure 7) and therefore the probability of having one site phosphorylated within the same EGFR molecule increases gradually with the EGF concentration, while the probability of having two sites increases sharply. (B) EGF dose–response curve of Cbl phosphorylation. Left, HeLa cells were treated with EGF for 2 min as indicated. Lysates were prepared in RIPA buffer (w/ 1% SDS) and then diluted to 0.2% SDS (see Materials and methods). IP and IB was as shown. Right, quantitation of the blots. Source data for this figure is available on the online supplementary information page.

Figure 6

Grb2 is required to generate the EGFR-Ub threshold in vivo and in vitro. (A) HeLa cells were subjected to Grb2-KD or transfection with control oligos. Lysates were stimulated with EGF for 2 min as indicated and subjected to IP/IB as indicated. (B) Top, HeLa cells, either control or Grb2-KD, were stimulated with EGF for 2 min at the indicated concentrations. Lysates were subjected to IP and IB as shown. Note that the Grb2-KD displays approximately three-fold reduced total ubiquitination with respect to control cells (panel A); thus, we used three-fold more lysate (3 × , 1 mg for Grb2-KD, 300 μg for control) to obtain comparable IB signals. Bottom, quantitative assessment. Results are expressed as a percentage of the maximal amount (% of max, see Materials and methods) of EGFR ubiquitination. (C) Lysates of HeLa cells, control and Grb2-KD, stimulated with EGF for 2 min at the indicated concentrations were subjected to ELISA, reverse approach (see Supplementary Figure 2B), using anti-Ub (FK2) as capturing antibody. Results are shown as a percentage of the maximal ubiquitination (% of max, see Materials and methods). (D) Left, lysates of HeLa cells stimulated with EGF for 2 min at the indicated concentrations were subjected to pull-down assay with 10 μg of GST-Cbl as bait, in absence (left panels) or in presence of 10 × molar excess bacterially purified Grb2 (right panels). IB was as shown. Right, quantitation of the blots. (E) In vitro ubiquitination assay. GST-EGFR cytoplasmic tail (250 ng) was subjected to in vitro autophosphorylation reaction and then bound to beads followed by incubation with ubiquitin (1 μg), purified E1 (100 ng), UbcH5c as E2 (500 ng), Cbl as E3 (500 ng), in absence or presence of purified Grb2. IB was as indicated. Results are representative of at least three experiments. Control reactions without EGFR or without E2 are also shown. All P-values were calculated using two-way ANOVA analysis. Graph error bars indicate s.d. calculated on at least three independent experiments. Source data for this figure is available on the online supplementary information page.

Figure 7

Phosphorylation, ubiquitination and association with Cbl of EGFR add-back mutants. (A) Top, HeLa cells were stimulated with EGF for 2 min at the indicated concentrations. Lysates were prepared in 1% SDS-containing lysis buffer (see Materials and methods) and subjected to IP and/or IB as shown. Eight-tenth of the IP were IB with anti-pY1045; one-tenth each of the IP was IB with anti-pY1068 or anti-EGF. Bottom, quantitation of the blots shown as % of max. Mean and statistical analysis performed on three independent experiments are shown. (B) Top. Scheme of the add-back mutants used in this study. The intracellular domain (kinase domain and C-terminal tail) of the EGFR is shown, with the position of the relevant residues. Critical tyrosine residues involved in Cbl/Grb2 binding are indicated in blue, while the other tyrosine residues in the EGFR tail are depicted in grey. Bottom, NR6 cells stably expressing EGFR-WT or the indicated mutants were analysed by ¹²⁵I-EGF saturation binding and the number of surface receptors was measured. Data are expressed as surface EGFRs/cell. (C) Quantitation of tyrosine phosphorylation of individual phosphosites in the EGFR add-back mutants (in comparison to EGFR-WT), as a function of EGF dose. Results are expressed as absolute values in arbitrary units (a.u., 100=max WT at 100 ng/ml EGF, see Materials and methods). Raw data are in Supplementary Figure 7. (D) Left, cells expressing the indicated mutants were stimulated with EGF (100 ng/ml, 2 min). Lysates were subjected to IP and IB as indicated. Right, quantitation of the blots. Results were normalized for the amount of immunoprecipitated EGFR and are expressed as % of the values obtained in EGFR-WT cells. Mean and statistical analysis performed on three independent experiments are shown. (E) Left, NR6 cells stably expressing EGFR-WT or the indicated mutants cells were stimulated with EGF (100 ng/ml, 2 min). Lysates were subjected to IP and IB as indicated. Right, quantitation of the Ub blot. Results were normalized for the amount of immunoprecipitated EGFR and are expressed as % of the values obtained in EGFR-WT cells. Mean and statistical analysis performed on three independent experiments are shown. Source data for this figure is available on the online supplementary information page.

Figure 8

Dose–response behaviour of Cbl-binding and receptor ubiquitination of EGFR add-back mutants. (A) Left, NR6 cells stably expressing the Y1045+ mutant, the Y1045/68/86+ mutant or EGFR-WT, were stimulated with EGF for 2 min at the indicated concentrations. Lysates were subjected to IP and IB as shown. Note that for the 1045+ mutant, we used two-fold more lysate (2 × , 2 mg for 1045+, 1 mg for EGFR-WT and Y1045/68/86+) and different washing conditions (see Materials and methods). Right, quantitative assessment. (B) Left, NR6 cells stably expressing the Y1045+ mutant, the Y1045/68/86+ mutant or EGFR-WT, were stimulated with EGF for 2 min at the indicated concentrations. Lysates were subjected to IP and IB as shown. Note that the 1045+ mutant displays approximatley five-fold reduced total ubiquitination with respect to EGFR-WT (Figure 8A); thus, we used five-fold more lysate (5 × , 1 mg for 1045+, 200 μg for EGFR-WT and Y1045/68/86+) to obtain comparable IB signals. Right, quantitative assessment. (C) The same samples were subjected to ELISA, forward approach, using anti-Ub (FK2) as detecting antibodies. Results are shown as a percentage of WT ubiquitination (arbitrary units, 100=max WT, left panel) or percentage of maximal ubiquitination in each dose–response curve (% of max, right panel, see Materials and methods). Graph error bars indicate s.d. calculated on at least three independent experiments. All P-values were calculated using two-way ANOVA analysis. R, the Pearson correlation coefficient. Source data for this figure is available on the online supplementary information page.

Figure 9

EGFR ubiquitination and EGFR-NCE are mechanistically linked. (A) ¹²⁵I-EGF internalization kinetics in control HeLa cells or upon KD of the indicated proteins at low (1 ng/ml, top) or high EGF dose (30 ng/ml, bottom). Results are expressed as internalization rate constants (Ke or Ke obs, see Supplementary Experimental Procedures) and are the mean of triplicate experiments. (B) Dose–response curves of EGFR-NCE and EGFR ubiquitination (measured by ELISA, forward approach) in HeLa cells. (B) Dose–response curves of ¹²⁵I-EGF internalization in HeLa cells. Total internalization, CME and NCE are shown, determined as explained in Supplementary data. (C) EGF dose–response curves of EGFR-NCE and EGFR ubiquitination (measured by ELISA, forward approach) in NR6 cells stably expressing EGFR-WT or the indicated add-back mutants. Symbols and error bars are not shown to avoid overcrowding of the figure; actual data are from Figures 8C (for EGFR-Ub) and Figure 9E (for EGFR-NCE). (D) Dose–response curves of ¹²⁵I-EGF internalization in NR6 cells expressing EGFR-WT or the indicated add-back mutants. Total internalization, CME and NCE are shown, determined as explained in Supplementary Experimental Procedures.