Polyubiquitination of the epidermal growth factor receptor occurs at the plasma membrane upon ligand-induced activation.

We have previously shown that, although overexpression of mutant dynamin inhibits clathrin-dependent endocytosis and disrupts high affinity binding of epidermal growth factor (EGF) to the EGF receptor (EGFR), it does not inhibit ligand-induced translocation of the EGFR into clathrin-coated pits. In the present study, we demonstrate that, upon ligand binding and incubation at 37 degrees C, the EGFR was polyubiquitinated regardless of overexpression of mutant dynamin. In cells not overexpressing mutant dynamin, the EGFR was rapidly internalized and deubiquitinated. In cells being endocytosis-deficient, due to overexpression of mutant dynamin, however, the EGFR was upon prolonged chase first found in deeply invaginated coated pits, and then eventually moved out of the coated pits and back onto the smooth plasma membrane. Polyubiquitination occurred equally efficiently in cells with or without intact clathrin-dependent endocytosis, while the kinetics of ubiquitination and deubiquitination was somewhat different. We further found that the EGF-induced ubiquitination of Eps15 occurred both in the absence and presence of endocytosis with the same kinetics as polyubiquitination of the EGFR, but that the EGF-induced monoubiquitination of Eps15 was somewhat reduced upon overexpression of mutant dynamin. Our data show that EGF-induced polyubiquitination of the EGFR occurs at the plasma membrane.

however, the EGFR was upon prolonged chase first found in deeply invaginated coated pits, and then eventually moved out of the coated pits and back onto the smooth plasma membrane. Polyubiquitination occurred equally efficiently in cells with or without intact clathrindependent endocytosis, while the kinetics of ubiquitination and deubiquitination was somewhat different. We further found that the EGF-induced ubiquitination of Eps15 occurred both in the absence and presence of endocytosis with the same kinetics as polyubiquitination of the EGFR, but that the EGF-induced monoubiquitination of Eps15 was somewhat reduced upon overexpression of mutant dynamin. Our data show that EGFinduced polyubiquitination of the EGFR occurs at the plasma membrane.
Binding of epidermal growth factor (EGF) 1 to its receptor at the plasma membrane results in activation and autophosphorylation of the EGF receptor (EGFR), as well as phosphorylation and activation of other molecules (1). Activation of EGFR thereby initiates signal transduction cascades important for cellular growth and differentiation. Furthermore, binding of EGF mediates translocation of the EGFR to clathrin-coated pits, internalization of the EGFR via clathrin-coated vesicles, transport through endosomes, and eventually degradation in lysosomes (2)(3)(4). Activation of the EGFR by ligand binding at 37°C was also demonstrated to induce polyubiquitination of the EGFR (5). No EGF-stimulated EGFR ubiquitination was found in experiments using a mutant kinase-negative EGFR or in experiments where clathrin-dependent endocytosis was prevented by low temperature or K ϩ depletion (5). This suggested a functional coupling between clathrin-dependent endocytosis and ubiquitination of the EGFR.
Ubiquitination has been shown to be essential for endocytosis of the yeast ␣-factor receptor (6) and for the growth hormone receptor (GHR) (7). In cells with a temperature-sensitive defect in ubiquitin conjugation, neither growth hormone (GH)-dependent internalization nor GH-dependent degradation of the GHR was observed at the non-permissive temperature (7). Further studies have demonstrated that when the internalization of the GHR was inhibited, GHR ubiquitination was also inhibited (8). These results also imply a connection between ubiquitin conjugation and endocytosis.
It has so far been unclear whether ubiquitination occurs prior to or following the budding of clathrin-coated vesicles. In a recent study, Levkowitz et al. (9) proposed that polyubiquitination of the EGFR occurs in endosomes. This was, however, not directly demonstrated. Instead, the authors demonstrated increased EGFR polyubiquitination and increased EGFR down-regulation by the overexpression of c-Cbl in Chinese hamster ovary cells. They furthermore demonstrated complex formation between c-Cbl and the EGFR upon incubation with EGF in cells overexpressing both c-Cbl and EGFR. Additionally, they showed that c-Cbl and the EGFR colocalized in endosomes upon addition of EGF. These data strongly support the recently suggested role for c-Cbl in the process of EGFR polyubiquitination (10,11), but do not clarify where in the cell ubiquitination occurs. It has been proposed by others that ubiquitination caused endocytosis of GHR by directing the GHR into clathrin-coated pits (8).
We have previously shown that EGF stimulation of the EGFR induced a rapid relocalization of EGFR from the smooth plasma membrane into clathrin-coated pits in cells being endocytosis-deficient due to overexpression of the GTPase-deficient mutant K44A form of dynamin (12). Clathrin-coated pits do not bud in cells overexpressing this mutant form of dynamin (13), and we took advantage of this system to study the trafficking, tyrosine phosphorylation, and ubiquitination of the EGF-stimulated EGFR in the absence of clathrin-dependent endocytosis. We found that in cells overexpressing K44A dynamin the EGFR transiently localized to coated pits, but that hardly any EGF⅐EGFR complexes were internalized. Nevertheless, polyubiquitination occurred as efficiently in endocytosisdeficient cells as in cells where clathrin-dependent endocytosis took place in a normal fashion. Our data therefore demonstrate for the first time that ubiquitination of the EGFR occurs at the plasma membrane prior to endocytosis.

EXPERIMENTAL PROCEDURES
Materials-Human recombinant EGF was from Bachem Feinchemikalien AG (Budendorf, Switzerland). All reagents were from Sigma unless otherwise noted.
Immunoelectron Microscopy-Cells were incubated with 10 Ϫ8 M EGF for 15 min on ice, washed with ice-cold phosphate-buffered saline (PBS) (140 mM NaCl, 2.7 mM KCl, 10 mM Na 2 HPO 4, 2 mM NaH 2 PO 4 , pH 7.4), and chased in prewarmed EGF-free medium at 37°C. At the end of the chase period, the cells were washed with ice-cold PBS, fixed with 4% paraformaldehyde, 0.1% glutaraldehyde in Soerensen's phosphate buffer, and processed for cryosections and immunogold labeling (15). The sections were labeled using sheep anti-EGFR antibody (Life Technologies, Inc., catalog no. 13287-016), followed by goat anti-sheep IgGcoated 18 nm colloidal gold (Jackson Immunoresearch Laboratory Inc.). To quantify EGFR distribution, the number of gold particles localized to the plasma membrane were counted and sorted into two groups depending on whether the gold localized to clathrin-coated or to smooth plasma membrane areas. Labeling along the tube connecting coated pits with the rest of the plasma membrane was interpreted as labeling of noncoated plasma membrane. From each experiment micrographs from at least 10 randomly chosen cells were examined, and a minimum of 100 gold particles were counted. The percentage of distribution of gold particles to coated pits were calculated as (no. of gold particles on coated plasma membrane/total no. of gold particles at the plasma membrane) ϫ 100.
Labeling for Eps15 was performed using rabbit anti-Eps15 antibodies followed by protein A-coated colloidal gold (purchased from G. Posthuma and J. Slot, Utrecht, Holland).
Brefeldin A Experiments-HeLa cells were preincubated with brefeldin A (17, 18) (5 g/ml) for 1 h at 37°C. Further processing was as for the experiments described above, but in the presence of brefeldin A.
Chase-dependent redistribution of EGFR was quantified after immunolabeling by calculating EGFR labeling density/length of plasma membrane (16) on micrographs from randomly chosen cells.
Western Blotting-EGF (10 Ϫ8 M) was added to cells in 12-well microtiter plates on ice for 15 min before the cells were washed with ice-cold PBS and chased in prewarmed EGF-free medium at 37°C. After incubation the cells were subjected to Western blot analysis as described (19). The cells were lysed in lysis buffer (10 mM Tris-HCl (pH 6.8), 5 mM EDTA, 50 mM NaF, 30 mM sodium pyrophosphate, 2% (w/v) SDS (Bio-Rad), 4% (v/v) ␤-mercaptoethanol, 1 mM Na 3 VO 4 , 1 mM phenylmethylsulfonyl fluoride (PMSF) on ice for 10 min. Then 0.033% (w/v) bromphenol blue and 4% (v/v) glycerol) was added, and the cell lysate was incubated at 95°C for 10 min, centrifuged at 20.000 ϫ g for 15 min, and the supernatant fraction was subsequently subjected to SDS-PAGE (20). The proteins were then electrotransferred to nitrocellulose membranes (21). The reactive proteins were detected using an enhanced chemiluminescence method (Amersham Pharmacia Biotech). To quantify band intensity, and when Western blotting with antibody against ubiquitin, proteins were electrotransferred to polyvinylidene difluoride membranes (Hybond-P, Amersham Pharmacia Biotech) following SDS-PAGE. For quantification, the membrane was incubated with primary antibodies as described above before incubation with alkaline phosphatase-conjugated anti-mouse antibodies or alkaline phosphatase-conjugated anti-rabbit antibodies for 2 h at room temperature. The immunobinding was detected by the enhanced chemifluorescence method (Amersham Pharmacia Biotech), and the chemifluorescence was measured by a phosphorofluoroimager (STORM840, Molecular Dynamics, Sunnyvale, CA). Due to dispersion of the EGFR caused by ubiquitination, both the main band and the smear above the main band were quantified. When membranes were incubated with anti-ubiquitin antibody, 3% (w/v) gelatin in Tris-buffered saline containing 1% (w/v) Tween 20 was used as blocking reagent, and the immunobinding was detected by enhanced chemiluminescence using an ECL Plus kit (Amersham Pharmacia Biotech).
Immunoprecipitation-When immunoprecipitating the EGFR to determine the EGFR phosphorylation, the cells were lysed in immunoprecipitation buffer (PBS with 10 mM EDTA, 1% (v/v) Triton X-100, 10 mM NaF, 200 units/ml aprotinin, 1 mM PMSF, and 1 mM Na 3 VO 4 ). Then protein G-agarose (Amersham Pharmacia Biotech) was incubated with sheep anti-EGFR antibody for 1 h at room temperature. The complexed antibody was washed twice in the buffer used for immunoprecipitation, and immunoprecipitation was performed at 4°C for 60 min. In the experiment demonstrating ubiquitination of the EGFR, the immunoprecipitation was performed essentially as described by Strous et al. (7). The cells were lysed in boiling buffer, containing 1% (w/v) SDS in PBS. After heating the lysates for 5 min at 100°C, the cell lysates were homogenized using QIAshredder columns (Qiagen GmbH, Hilden, Germany). Then protein A-agarose and protein G-agarose (Amersham Pharmacia Biotech) were incubated with rabbit anti-ubiquitin antibody and sheep anti-EGFR antibody, respectively, for 1 h at room temperature. The complexed antibody was washed twice in the immunoprecipitation buffer, consisting of 1% (v/v) Triton X-100, 0.5% (w/v) SDS, 0.25% (w/v) sodium deoxycholate, 0.5% (w/v) bovine serum albumin, 1 mM EDTA, 1 mM PMSF, 2 mM Na 3 VO 4 , 20 mM NaF, 10 g/ml leupeptin, and 200 units/ml aprotinin in PBS, and immunoprecipitation was performed at 4°C for 60 min. The immunoprecipitate was washed and subjected to SDS-PAGE and Western blotting, as described above. Ϫ . After 15 min at 4°C, the cells were washed three times with ice-cold PBS to remove unbound ligand. The cells were then chased in minimal essential medium without HCO 3 Ϫ and with 0.1% (w/v) bovine serum albumin at 37°C for the indicated time periods. The medium was collected, and the 125 I-EGF was precipitated using 5% (w/v) tricloroacetic acid and 1% (w/v) phosphotungstic acid, as described (22). Both the trichloroacetic acid-precipitable and the trichloroacetic acid-soluble radioactivity were measured in a ␥-counter. Cells were washed three times with PBS, treated with 0.2 M sodium acetate buffer (pH 4.5 or 7.4) containing 0.5 M NaCl (SAB pH 4.5/SAB pH 7.4) on ice for 10 min, and washed once with the same buffer. Then 125 I-EGF was precipitated from the cells with trichloroacetic acid and 1% (w/v) phosphotungstic acid. Finally, the precipitate was dissolved with 1 M NaOH, and the radioactivity was measured. Counts/min from cells treated with SAB pH 4.5 represent internalized 125 I-EGF, while counts/min from cells treated with SAB pH 7.5 represent both internalized and surface-localized 125 I-EGF.

Approximately 50% of Initially EGF-bound EGFR Localize to Coated Pits in a Transient
Manner-Overexpression of K44A dynamin has been shown to inhibit clathrin-mediated endocytosis (13). We have recently shown that although the overexpression of K44A dynamin disrupts high affinity binding of EGF to the EGFR, the EGFR is efficiently recruited into coated pits upon ligand binding (12). Due to overexpression of K44A dynamin, coated pits were often found at the end of long tubular plasma membrane invaginations, and sometimes gathered in groups. In our previous study we only examined cells chased for 10 min at 37°C. As coated pits do not bud and form coated vesicles in cells overexpressing K44A dynamin, we wanted to examine the trafficking of the EGFR upon extended chase periods at 37°C. EGF (10 Ϫ8 M) was added to cells on ice for 15 min, the cells were washed free of unbound ligand and chased for increasing time periods at 37°C. At the end of the chase period, the cells were fixed, sectioned, and immunocytochemically labeled using antibod-ies recognizing the EGFR. To estimate the EGF-induced redistribution of the EGFR, we quantified the amount of EGFR localizing to coated pits compared with the total amount of EGFR found at the plasma membrane.
Immunocytochemical labeling showed that in cells not exposed to EGF, as well as in cells fixed immediately after binding of EGF on ice, the EGFR mainly localized outside coated pits (Table I and Fig. 1A). However, upon chasing at 37°C, the EGFR was seen to relocalize rapidly from non-coated, uninvaginated plasma membrane areas into areas where the plasma membrane had a clear cytoplasmic coat (Fig. 1B). At early time points (2-5 min) (Table I and Fig. 1B), several of these coated areas were flat or only slightly invaginated. After a longer chase period (10 -15 min) (Table I and Fig. 1C), however, most of the coat-associated labeling was in coated pits connected to the plasma membrane by long tubular necks. After 10 -15 min of chase, almost 50% of the EGFRs were seen in coated pits (Table I). The labeling was usually found associated with the coated, bulb-shaped blind end of the tubules, and it should be noted that almost no labeling was found along the tubular neck. Upon further chase, however, the EGFR started to appear along the tube while the bulb-shaped end remained coated and the number of EGFR associated with coated areas decreased (Table I and Fig. 1, D and E). Upon 60 min of chase at 37°C, the distribution of plasma membrane-associated EGFR was almost as in cells not exposed to EGF (Table I).
Although the antibodies used recognize the intracellular part of the EGFR, labeling was in some cases found on the extracellular side of the plasma membrane ( Fig. 1). Due to the size of the primary antibody recognizing the EGFR and the secondary antibody coating the colloidal gold, the observed distance from the gold particle to the antigen may be more than the width of a membrane (16). The specificity of the anti-EGFR antibody was confirmed by Western blotting and was furthermore reflected by the EGF-induced change in labeling distribution. The colloidal gold contained a small amount of doublets (see Fig. 1D), but when quantifying, eventual clusters were counted as one gold particle.
Labeling of sections showed that independent of EGF binding and chase at 37°C, some EGFR localized to the Golgi apparatus (data not shown). This labeling most likely represents newly synthesized EGFR. As including newly synthesized EGFR appearing at the plasma membrane during the chase period could complicate the study of ligand-induced trafficking of the EGFR, we treated cells with brefeldin A before addition of EGF as well as during the chase period. Brefeldin A prevents the transport of newly synthesized proteins from the endoplasmic reticulum to the plasma membrane by redistributing the cis and medial stacks of the Golgi apparatus back into the endoplasmic reticulum (17,18). Quantification of labeling on sections from brefeldin A-treated cells showed basically the same EGF-induced EGFR redistribution and transient localization to coated pits as found in cells not exposed to brefeldin A (see Table I). We conclude that, in cells deficient in endocytosis due to overexpression of K44A dynamin, EGFR moves to coated pits upon binding of EGF. The localization to clathrincoated pits is transient, and the EGFR seems to loose its association with coat components and move back onto smooth, uninvaginated parts of the plasma membrane.
Internalization of EGF⅐EGFR Complexes Is Efficiently Inhibited in Cells Overexpressing Mutant Dynamin-To study the ligand-dependent redistribution of EGFR quantitatively, both with and without overexpression of mutant dynamin, we incubated cells with 10 Ϫ8 M 125 I-EGF on ice for 15 min before washing away unbound ligand and chasing the cells at 37°C. At different time points, the fate of the initially bound 125 I-EGF was analyzed as described under "Experimental Procedures." The results show that the majority of the initially bound 125 I-EGF in cells overexpressing K44A dynamin rapidly dissociated from the EGFR. Of the 125 I-EGF that remained cell-associated, only a small amount was endocytosed, and eventually degraded ( Fig. 2A). This is consistent with the findings of Vieira et al. (23). The inhibition of EGFR endocytosis was further confirmed by immunocytochemical labeling for the EGFR. Quantification of labeling on sections from cells exposed to brefeldin A showed that the EGFR labeling density at the plasma membrane remained unchanged for up to 60 min of chase at 37°C (data not shown). In cells grown with tetracycline (not overexpressing mutant dynamin), 125 I-EGF was rapidly endocytosed, with most ligand internalized upon 10 min of chase at 37°C (Fig. 2B).
Autophosphorylated EGFR Is Dephosphorylated upon Entry into Coated Pits and Rephosphorylated upon Endocytosis-Tyrosine-phosphorylated and dimerized EGFR have previously been shown to exist in endosomes (24). In order to correlate the tyrosine phosphorylation of the EGFR with its cellular localization, we incubated cells with 10 Ϫ8 M EGF on ice for 15 min before washing away unbound ligand and chasing the cells at 37°C. Incubation with EGF on ice caused tyrosine phosphorylation of the EGFR both in noninduced cells and in cells overexpressing K44A dynamin (Fig. 3A). However, the EGFR was less efficiently tyrosine-phosphorylated in cells overexpressing K44A dynamin, compared with in cells not overexpressing mutant dynamin (Refs. 12 and 23; Fig. 3B). It should be noted that, upon overexpression of mutant dynamin, more EGFR were found at the plasma membrane (12). We have demonstrated the specific tyrosine phosphorylation of the EGFR (Fig. 3B) by normalizing the intensity of the phosphotyrosine signal with respect to the EGFR signal both in cells with and without overexpression of mutant dynamin. Chase at 37°C caused a rapid dephosphorylation of the EGFR in both induced and noninduced cells. Once dephosphorylated, the EGFR in cells overexpressing K44A dynamin remained dephosphorylated. The kinetics of the dephosphorylation coincided with both entry of EGFR into coated pits and dissociation of ligand. In cells not over-

EGF-induced redistribution of EGFR to coated pits in cells overexpressing mutant dynamin
The distribution of the EGFR to coated versus non-coated areas of the plasma membrane in cells overexpressing mutant (K44A) dynamin was quantified as described under "Experimental Procedures." The distribution of EGFR was determined in cells after binding of EGF on ice and chase at 37°C in EGF-free medium (Pulse-chase) and in cells exposed to brefeldin A prior to, as well as during, a pulse-chase incubation (Brefeldin A). The table shows the percentage of total plasma membrane-associated EGFR labeling, that under each experimental condition localized to coated membrane areas. The control column shows the percentage amount of EGFR localized to coated pits in cells prior to incubation with EGF. The 0, 2, 5, 10, 15, 30 and 60 min columns indicate the percentage amount of EGFR in coated pits upon these times of incubation at 37°C. Labeling along the tube connecting the coated pits with the uninvaginated plasma membrane was interpreted as labeling of non-coated plasma membrane. Results are presented as the mean Ϯ S.E. from a minimum of three independent experiments. expressing mutant dynamin, however, some of the EGFRs were transiently rephosphorylated upon prolonged chase at 37°C (Fig. 3, A and B; 8, 10, and 12 min). When considering the kinetics of trafficking and phosphorylation, the rephosphorylation observed in cells not overexpressing mutant dynamin seemed to occur upon the EGFR reaching endosomes.

Polyubiquitination of the EGFR Occurs at the Plasma Membrane-When
Western blotting with the anti-EGFR antibody, we found that the mobility of the EGFR was shifted as a function of ligand binding and chase at 37°C, and by overexposing the film, a smear was clearly visible (Fig. 4A). As the Western blotting was performed subsequent to SDS-PAGE, this indicated that the decrease in mobility was caused by a covalent modification of the EGFR. It has previously been reported that the EGFR can be polyubiquitinated (5). We therefore investigated whether the decreased mobility of the protein, reactive to both anti-phosphotyrosine antibody (data not shown) and anti-EGFR antibody, was due to polyubiquitination by immunoprecipitating with an antibody to ubiquitin and Western blotting the immunoprecipitate, using an antibody to the EGFR. As demonstrated in Fig. 4B, the slowly migrating EGFR was indeed ubiquitinated. We further immunoprecipitated EGFR and Western-blotted the immunoprecipitate employing an antibody to conjugated ubiquitin. By this procedure, bands with the same apparent molecular weight were visualized (data not shown). As demonstrated in Fig. 4, the EGFR in cells not overexpressing mutant dynamin was clearly polyubiquitinated upon 2, 5, and 10 min of chase at 37°C. However, in cells that were endocytosis-deficient due to overexpression of mutant dynamin, there was a clear polyubiquitination of the EGFR even upon 15 min of chase. In cells not overexpressing mutant dynamin, a large fraction of the EGFR was observed to be endocytosed upon 10 min of chase at 37°C (Ref. 12 and Fig.  2). As demonstrated in Fig. 4, at this time point, the EGFR had started to become deubiquitinated. In conclusion, a transient polyubiquitination of the EGFR was observed upon ligand binding at 4°C, followed by chase at 37°C, and the transient polyubiquitination was observed regardless of EGFR endocytosis, demonstrating that the polyubiquitination of the EGFR happened at the plasma membrane. The polyubiquitination of the EGFR furthermore appeared to precede endocytosis of the EGFR (compare Figs. 2 and 4).
EGF Causes a Transient Ubiquitination of Eps15-Activation of EGFR has previously been shown to induce phosphorylation as well as monoubiquitination of the coated pit localized protein Eps15 (25)(26)(27). Following incubation of cells with EGF on ice and chase at 37°C, we found a rapidly induced shift in Eps15 mobility upon SDS-PAGE and Western blotting with antibody to Eps15 (Fig. 5). EGF induced the same shift in mobility both in cells with and without overexpression of mutant dynamin. However, more Eps15 was shifted in cells that did not overexpress mutant dynamin. The shift in Eps15 mobility has been shown previously to represent monoubiquitination of Eps15 (27). It should be noted that, in cells overexpressing mutant dynamin, Eps15 was slightly ubiquitinated even in the absence of EGF (compare controls with and without tet, Fig. 5). While the shift in Eps15 mobility was reversed upon 10 min of chase at 37°C in cells without overexpression of mutant dynamin, Eps15 remained ubiquitinated for 30 min in cells with deficient clathrin-dependent endocytosis due to overexpression of mutant dynamin (Fig. 5). This demonstrated that ubiquitination of Eps15 is also transient and followed the same kinetics as ubiquitination and coated pit localization of the EGFR. To investigate whether ubiquitination affected the localization of Eps15, we labeled the sections with antibody to Eps15. The labeling for Eps15 was restricted to the rim of the coated area even on coated pits attached to the K44A dynamininduced tubules both in the absence of EGF and upon binding of EGF followed by 30 min of chase at 37°C (Fig. 6).

FIG. 1. In cells overexpressing mutant (K44A) dynamin, EGF induces transient EGFR localization into coated pits.
HeLa cells overexpressing mutant dynamin were fixed either prior to incubation with EGF (A) or after binding of EGF on ice, followed by chase at 37°C in EGF-free medium (B-E). To localize the EGFR, thawed cryosections were labeled with sheep anti-EGFR antibodies followed by goat anti-sheep coated colloidal gold. In cells not exposed to EGF (A), the EGFR localized to the smooth plasma membrane (1) but not to coated pits. Shown within the micrograph is an EGFR-negative, bulb-shaped coated pit. A part of the tubular membrane invagination connecting the coated pit to the plasma membrane is contained within the section and indicated by arrowheads. B, EGFR localization after binding of EGF and 5 min of chase. The EFGR localized to coated pits at different stages of formation. EGFR labeling is seen at the non-coated plasma membrane (1), in a slightly invaginated coated pit (2), as well as in a fully invaginated coated pit connected to the plasma membrane by a tubular membrane invagination (3). The tubular connection is in this micrograph difficult to observe because its major part is not within the plane of the section. After 10 min of chase (C), almost 50% of the EGFR localized to coated pits, and the majority of the labeling was found at coated areas localized to the bulb-shaped end of the tubular membrane invaginations. Prolonged chase, however, caused the EGFR to move out of the coated pits, along the tube, and back onto the non-coated, uninvaginated plasma membrane. D and E show EGFR localization in cells fixed upon 30 min of chase. In D labeling can be seen within the bulb-shaped coated pits (3) as well as along tubular plasma membrane invaginations (4), while in E labeling is shown along the tube connecting two separate coated pits to the plasma membrane. Bar, 100 nm.

DISCUSSION
We have demonstrated that, both in HeLa cells where endocytosis of the EGFR is inhibited due to overexpression of the K44A mutant dynamin and in HeLa cells with intact EGFR endocytosis, the EGFR became polyubiquitinated when the cells were chased at 37°C following binding of EGF at 4°C. At 4°C the EGFR was maximally autophosphorylated, and during chase at 37°C the EGFR rapidly became dephosphorylated. Ubiquitination was initiated by EGF binding, but was delayed with respect to EGFR autophosphorylation. The lag could indicate that ubiquitination follows phosphorylation of an EGFR substrate or, alternatively, that docking of a molecule onto the EGFR with ensuing complex formation is required for ubiquitination to occur. Interestingly, we found that the EGFR substrate Eps15 was ubiquitinated with the same kinetics as was the EGFR. In yeast, genetic interaction between the Eps15 homologue, Pan1p, and the ubiquitin-protein ligase Rsp5p has been demonstrated, and it was suggested that Pan1p could act as a connector bringing Rsp5p close to potential ubiquitination substrates (28). There is therefore the possibility that Eps15 plays a similar role in mammalian cells, and via an ubiquitinprotein ligase induces ubiquitination of membrane proteins like the EGFR. It was recently shown that c-Cbl may act as an ubiquitin-protein ligase (29) and that c-Cbl, synergisti-cally with the ubiquitin-conjugating enzyme UbcH7, promoted ligand-induced ubiquitination of the EGFR (9 -11). c-Cbl has previously been shown to become tyrosine-phosphorylated almost immediately upon EGF stimulation (30), and stimulation with colony-stimulating factor-1 caused a transient phosphorylation, membrane targeting, and ubiquitination of c-Cbl (31). c-Cbl was found to be associated with Crk upon stimulation with EGF (30). Crk has an Src homology 3 domain, which specifically binds to a proline-rich motif in Eps15 (32), further suggesting a role for Eps15 in induction of ubiquitination.
It was recently proposed that polyubiquitination of the EGFR occurs in endosomes (9). However, this hypothesis has not been confirmed, and, up to now, it has been unclear whether ubiquitination occurs before or after internalization of the EGFR. In the recent paper by Levkowitz et al. (9), the authors claimed that they were unable to detect ubiquitinated EGFR on the cell surface by using biotin labeling. However, due to the many technical difficulties in performing such experiments, a negative result does not necessarily mean that ubiquitination of the EGFR does not occur at the plasma membrane. Our present experiments demonstrated that EGF⅐EGFR complexes were not internalized in cells overexpressing mutant dynamin. However, EGF-stimulated EGFR were clearly polyubiquitinated. Our present results therefore show that polyubiquitination of the EGFR occurs at the plasma membrane. We cannot completely rule out the possibility that the EGFR can also be ubiquitinated in endosomes. However, the magnitude of the EGFR ubiquitination was comparable in cells with or without endocytosis. Furthermore, in cells not overexpressing mutant K44A dynamin, the ubiquitination appeared to precede endocytosis, and we observed no additional waves of ubiquitination during the 60-min chase following ligand binding. Interestingly, our unpublished data 2 show that c-Cbl, believed to be instrumental in the ubiquitination of the EGFR (9 -11), was maximally tyrosine-phosphorylated by EGF on ice regardless of overexpression of mutant dynamin.
In yeast, polyubiquitination serves as an internalization signal for the ␣-factor receptor and the a-factor receptor (6,33). In mammalian cells, several tyrosine kinase or kinaselinked receptors are ubiquitinated upon ligand binding (reviewed in Ref. 34). Binding of GH to the GHR induces dimerization, activation and increased ubiquitination of the receptor. Using cells with a temperature-sensitive defect in ubiquitin conjugation (7) or cells expressing truncated GHR (8,35), Strous and colleagues showed that the ubiquitin conjugation system, but not ubiquitination of the GHR itself, is required for ligand-induced endocytosis of the GHR. We here find that the EGF-induced transient localization of EGFR into clathrin coated pits coincides with the transient ubiquitination of both the EGFR and the coated pit localized protein Eps15. This could suggest that ubiquitination of the EGFR is important for localization of the receptor to clathrincoated pits, and it is tempting to suggest that polyubiquitination of the EGFR could function as a recruiting or a retention signal for coated pit localization. However, at no time did we find more than 50% of the plasma membrane associated EGFR localized to coated pits, and we have so far not been able to differentiate between ubiquitinated and non-ubiquitinated receptors at the plasma membrane. Further studies are therefore required in order to conclude as to the ubiquitination status of EGFRs localizing to coated pits.
Although overexpression of mutant (K44A) dynamin pre- vents clathrin-dependent receptor-mediated endocytosis, it does not inhibit the recruitment of receptors into coated pits. Both the constitutive transport of Tfn-R and the ligand-induced transport of EGFR into coated pits appear to occur as in normal cells (12,13). In the present study, we demonstrate that, in HeLa cells that are endocytosis-deficient due to overexpression of K44A dynamin, the EGF-induced localization of EGFR to clathrin-coated pits was transient. Whether binding of EGF causes the EGFR to move into coated pits before or after assembly of clathrin has been debated (36 -38). In cells overexpressing K44A dynamin, invaginated clathrin-coated pits accumulate at the expense of flat clathrin-coated membrane (13). After binding on ice for 15 min, the cells were washed and chased at 37°C in EGFfree medium for the time periods indicated. A, the EGFR was isolated by immunoprecipitation and analyzed by Western blotting using an antibody to phosphotyrosine (␣-pTyr) or to the EGFR (␣-EGFR). B, the intensity of tyrosinephosphorylated EGFR and of the EGFR was quantified as described under "Experimental Procedures." The intensity of the phosphorylated EGFR was normalized with respect to the intensity of the EGFR, and the specific tyrosine phosphorylation of EGFR is therefore demonstrated. The data represent the mean of seven independent experiments, except the values at 8 and 12 min, which represent the means of three and four independent measurements, respectively. was added to the cells on ice. After binding on ice for 15 min, the cells were washed and chased at 37°C for the times indicated. The cells were then lysed as described under "Experimental Procedures." A, the cell lysates were subjected to SDS-PAGE and Western blotting with an antibody to the EGFR, as described under "Experimental Procedures." In order to demonstrate the mobility shift of the EGFR, the film was overexposed. B, the cell lysates were subjected to immunoprecipitation using an antibody to conjugated ubiquitin as described under "Experimental Procedures." The immunoprecipitate was then subjected to SDS-PAGE and Western blotting with an anti-EGFR antibody. This is probably due to a limited cytoplasmic pool of coat components, and a consequence of this could be that new coated pits are prevented from forming. However, we observed coated pits at all stages of invagination, indicating that at least some new assembly of coated pits occurs. After short chase periods, the EGFR localized to flat, or only slightly invaginated, coated areas, but not to fully invaginated coated pits, or along the membrane tubes connecting these pits with the plasma membrane. This suggests that activated EGFR move into growing clathrin-coated pits. It was shown recently that EGF-induced activation of EGFR stimulated Src kinase activation with ensuing phosphorylation and redistribution of clathrin (39). EGFinduced recruitment of clathrin to the plasma membrane is consistent with the suggestion that the EGFR moves into growing coated pits and may explain why only 50% of EGFR were recruited to coated pits in endocytosis-deficient cells. This could be due to lack of clathrin and other molecules important for recruiting the EGFR into coated pits, possibly because of accumulation in preexisting coated pits.
In cells overexpressing K44A dynamin, the dephosphorylation of the EGFR correlated in time with dissociation of EGF. However, we also observed dephosphorylation in cells that did not overexpress mutant dynamin, where the EGF was somewhat more firmly associated with the EGFR via high affinity binding. In cells exhibiting normal endocytosis, the dephosphorylation was, however, transient. The EGFR appeared to be dephosphorylated when localized to coated pits, but some receptors were rephosphorylated upon endocytosis. As rephosphorylation of EGFR in endosomes could be the result of both homo-and heterodimerization, the efficiency of endocytosis of other members of the ErbB family of proteins is expected to affect the specificity of signaling from endosomes. This could explain how EGFR localization to the plasma membrane or to endosomes results in qualitative and quantitative differences of signal transduction, as recently described (19,40). Consist-ent with this notion, EGFR phosphorylation has been shown to be qualitatively different in endosomes, as compared with at the plasma membrane (41). Also consistent with our findings is the reportedly decreased EGF-induced tyrosine phosphorylation of the EGFR in cells deficient in endocytosis due to overexpression of K44A dynamin (23).
In conclusion, we have demonstrated that the EGFR was transiently localized to coated pits in endocytosis-deficient cells, where EGF⅐EGFR complexes were not internalized. Furthermore, we have shown that endocytosed, ligand-bound EGFR became rephosphorylated in endosomes, and that independently of endocytosis, the EGFR was polyubiquitinated at the plasma membrane in a transient manner. Cryosections of cells overexpressing K44A dynamin were labeled with antibodies to Eps15, followed by protein A-coated colloidal gold. The labeling (in C indicated by large arrowhead) showed that Eps15 localized to the rim of the clathrin coat both in uninvaginated coated pits (A), and in coated pits connected to the plasma membrane by K44A dynamin induced tubular invaginations (indicated by small arrowheads) (B-D). When several coated pits were attached to the same tube, each coated pit labeled for Eps15 (D). Micrographs shown in A and D are from cells not exposed to EGF; B and C are from cells chased for 30 min at 37°C. Bar, 100 nm.