Ligand-activated epidermal growth factor receptor (EGFR) signaling governs endocytic trafficking of unliganded receptor monomers by non-canonical phosphorylation

The canonical description of transmembrane receptor function is initial binding of ligand, followed by initiation of intracellular signaling and then internalization en route to degradation or recycling to the cell surface. It is known that low concentrations of extracellular ligand lead to a higher proportion of receptor that is recycled and that non-canonical mechanisms of receptor activation, including phosphorylation by the kinase p38, can induce internalization and recycling. However, no connections have been made between these pathways; i.e. it has yet to be established what happens to unbound receptors following stimulation with ligand. Here we demonstrate that a minimal level of activation of epidermal growth factor receptor (EGFR) tyrosine kinase by low levels of ligand is sufficient to fully activate downstream mitogen-activated protein kinase (MAPK) pathways, with most of the remaining unbound EGFR molecules being efficiently phosphorylated at intracellular serine/threonine residues by activated mitogen-activated protein kinase. This non-canonical, p38-mediated phosphorylation of the C-tail of EGFR, near Ser-1015, induces the clathrin-mediated endocytosis of the unliganded EGFR monomers, which occurs slightly later than the canonical endocytosis of ligand-bound EGFR dimers via tyrosine autophosphorylation. EGFR endocytosed via the non-canonical pathway is largely recycled back to the plasma membrane as functional receptors, whereas p38-independent populations are mainly sorted for lysosomal degradation. Moreover, ligand concentrations balance these endocytic trafficking pathways. These results demonstrate that ligand-activated EGFR signaling controls unliganded receptors through feedback phosphorylation, identifying a dual-mode regulation of the endocytic trafficking dynamics of EGFR.

The canonical description of transmembrane receptor function is initial binding of ligand, followed by initiation of intracellular signaling and then internalization en route to degradation or recycling to the cell surface. It is known that low concentrations of extracellular ligand lead to a higher proportion of receptor that is recycled and that non-canonical mechanisms of receptor activation, including phosphorylation by the kinase p38, can induce internalization and recycling. However, no connections have been made between these pathways; i.e. it has yet to be established what happens to unbound receptors following stimulation with ligand. Here we demonstrate that a minimal level of activation of epidermal growth factor receptor (EGFR) tyrosine kinase by low levels of ligand is sufficient to fully activate downstream mitogen-activated protein kinase (MAPK) pathways, with most of the remaining unbound EGFR molecules being efficiently phosphorylated at intracellular serine/threonine residues by activated mitogen-activated protein kinase. This non-canonical, p38-mediated phosphorylation of the C-tail of EGFR, near Ser-1015, induces the clathrin-mediated endocytosis of the unliganded EGFR monomers, which occurs slightly later than the canonical endocytosis of ligandbound EGFR dimers via tyrosine autophosphorylation. EGFR endocytosed via the non-canonical pathway is largely recycled back to the plasma membrane as functional receptors, whereas p38-independent populations are mainly sorted for lysosomal degradation. Moreover, ligand concentrations balance these endocytic trafficking pathways. These results demonstrate that ligand-activated EGFR signaling controls unliganded receptors through feedback phosphorylation, identifying a dual-mode regulation of the endocytic trafficking dynamics of EGFR.
Epidermal growth factor receptor (EGFR), 2 one of the most characterized receptor tyrosine kinases, regulates many cellular functions, including survival, proliferation, and differentiation. The aberrant activation of EGFR by overexpression or activating mutations is a major mechanism underlying the pathogenesis of human cancers, including colorectal and lung cancers, and participates in acquired resistance to anticancer agents (1)(2)(3)(4).
Ligand-bound EGFR proteins form an asymmetric homodimer on the plasma membrane, which is followed by activation of its tyrosine kinase. Activated EGFR is then rapidly internalized via clathrin-mediated endocytosis and clathrin-independent endocytosis. Sequential sorting to several vesicular transport systems, including early endosomes, late endosomes, multivesicular bodies (MVBs), and recycling endosomes, directs the fate of internalized EGFR to lysosomal degradation or recycling to the cell surface (5)(6)(7). However, the mechanisms by which structurally identical EGFR proteins are sorted to the different endocytic machineries of clathrin-dependent or -independent endocytosis and recycling or degradation have not yet been elucidated in detail.
Ligand concentrations in the extracellular environment are a key factor affecting EGFR intracellular transportation. Low concentrations mainly induce clathrin-mediated endocytosis, and a large portion of internalized EGFR is recycled back to the plasma membrane. Higher concentrations increase the ratio of lysosomal degradation instead of recycling endocytosed EGFR (7,8). Even in the presence of 50,000 molecules of EGFR on a single HeLa cell surface, the binding of 300 molecules of EGF was found to be sufficient to trigger an EGFR response in 50% of cells (9). Thus, the minimal activation of ligand-bound EGFR is sufficient to evoke intracellular signaling, indicating that most cell-surface EGFR remain in a ligand-unoccupied state. Nevertheless, previous studies did not examine residual unliganded receptors during a ligand stimulation in adequate detail.
Evidence is increasing for the non-canonical activation of receptor tyrosine kinases by the serine/threonine phosphorylation of their intracellular domains in ligand-and tyrosine kinase-independent manners. The phosphorylation of EphA2 Ser-897, for example, plays crucial roles in cell motility and proliferation and has been correlated with poor prognoses for lung cancer and glioblastoma multiforme (10,11). The noncanonical regulation of EGFR has also been investigated in the last decade. We and others have demonstrated that pro-inflammatory cytokines, including tumor necrosis factor ␣ (TNF-␣), and other cellular stresses induce serine/threonine phosphorylation and internalization of EGFR (12)(13)(14)(15)(16)(17). This type of EGFR endocytosis depends entirely on p38 activation and clathrin recruitment. Moreover, endocytosed EGFR is completely recycled to the plasma membrane.
In this study, we attempted to confirm the integrated hypothesis that ligand-activated canonical EGFR signaling provokes p38-mediated non-canonical regulation of residual unliganded EGFR monomers. The results obtained may resolve some of the controversial issues related to the regulation of EGFR endocytic trafficking, including sorting mechanisms for degradation or recycling following stimulation with different ligand concentrations.

p38-dependent and -independent endocytosis of EGFR
To investigate the role of p38 in ligand-induced endocytosis, we examined the effects of a p38 inhibitor (SB203580) and siRNA against p38␣, a major subtype of the p38 family, in HeLa cells (Fig. 1, A-D). As shown previously (12), immunofluorescence analysis confirmed that TNF-␣-induced EGFR internalization was completely dependent on p38 (Fig. 1, A-C, and Fig.  S1A). EGFR endocytosis triggered by low EGF (3 ng/ml) but not by high-EGF (100 ng/ml) was largely inhibited by SB203580 (Fig. 1, A and B) and knockdown of p38␣ ( Fig. 1, C and D, and Fig. S1A). Similar results were obtained in HeLa cells treated with other EGFR ligands, TGF-␣ and heparin-binding EGF-like growth factor (HB-EGF) ( Fig. 1E and Fig. S1B), and in EGFtreated A549 lung cancer cells ( Fig. 1F and Fig. S1C). Although it was difficult to detect membrane-expressing EGFR under permeabilized conditions, EGFR co-localized with the early endosomal marker EEA1 in HeLa cells, indicating that fluorescence dot signals were derived from endocytosed EGFR (Fig.  S2A).
As reported previously (18), endocytosis by low EGF was largely dependent on clathrin, whereas approximately half of EGFR endocytosis by high EGF occurred independent of clathrin (Fig. S2, B and C). We next subjected non-permeabilized cells to flow cytometry (Fig. 1G) and immunofluorescence ( Fig. 1H and Fig. S1D) to analyze cell-surface EGFR. These analyses also revealed the distinct contributions of p38 to EGFR endocytosis by low EGF and high EGF. Collectively, these results demonstrate that EGF induces p38-dependent (non-canonical) as well as -independent (canonical) mechanisms for EGFR endocytosis in a concentration-dependent manner.

Non-canonical phosphorylation of EGFR at low EGF concentration
It has been shown that p38-dependent phosphorylation of EGFR at C-terminal serine/threonine residues is involved in its cytokine-induced endocytosis. To investigate total EGFR phosphorylation, we employed immunoblotting using a Phos tag, which detects phosphorylated proteins as shifted bands. TNF-␣ caused band shifts in all EGFR proteins expressed, and these were abolished by SB203580 but not by PD153035 (an EGFR tyrosine kinase inhibitor) or U0126 (a mitogen-activated protein kinase/extracellular signal-regulated kinase kinase inhibitor), indicating p38-dependent non-canonical EGFR phosphorylation ( Fig. 2A). In analyses with various EGF concentrations, most EGFR molecules shifted unexpectedly at 3 ng/ml (Fig. 2B), at which canonical tyrosine autophosphorylation on Tyr-974, Tyr-1045, and Tyr-1068 was only slightly detected (Fig. 2C). Quantitative analysis demonstrated that low EGF induced the band shift with only slight Tyr(P)-1068 (Fig. 2D). In addition, the shifts induced by low-EGF were also completely dependent on p38, suggesting that low EGF stimulation mainly caused the non-canonical phosphorylation (Fig. 2E). Taken together, these results show that the minimal tyrosine kinase activation of EGFR by low EGF leads to the prominent non-canonical phosphorylation of most cell-surface EGFR molecules, probably in a ligand-unbound form.

p38-dependent endocytosis of ligand-unbound EGFR
Previous studies mainly focused on the internalization of ligand-bound EGFR, described as the canonical endocytosis pathway in this study. Here we investigated whether p38mediated endocytosis is an event in ligand-unbound EGFR. Ligand-receptor co-localization was monitored by immunofluorescence using rhodamine-conjugated EGF. A large amount of internalized EGFR co-localized with the ligand (yellow dots) in the presence of high EGF. In contrast, low EGF induced a small amount of yellow dots but strongly enhanced green dots, which were mainly composed of ligand-unbound EGFR (Fig.  3A). A quantitative analysis showed that, although receptor endocytosis by low EGF was similar to that by high EGF, endocytosis of the ligand markedly increased (Fig. 3B). Moreover, all endocytic events were canceled by the EGFR inhibitor (Fig. 3C); however, the endocytosis of unliganded EGFR was selectively inhibited by the p38 inhibitor (Fig. 3, D and E). These results clearly demonstrate that ligand-dependent activation of the p38-mediated non-canonical pathway selectively induces endocytosis of unliganded EGFR.

p38-dependent endocytosis of inactive EGFR monomers
EGFR markedly changes its conformation to initiate the activation of tyrosine kinase by dimerization (19,20). To investigate whether p38-dependent EGFR endocytosis requires dimer formation, we generated a dimer-deficient EGFR mutant (dd-EGFR) with a C-terminal GFP tag, which lacks the CR1 loop in the extracellular dimerization domain and intracellular docking sites (Ile-682 and Val-924) (19,21) (see also Fig. 4A). We confirmed the non-canonical phosphorylation of dd-EGFR as a similar Phos tag shift pattern in endogenous EGFR (compare Fig. 4B with Fig. 2B). The shift induced by both low EGF and Dual-mode regulation of EGFR endocytic trafficking dynamics Figure 1. p38-mediated endocytosis of EGFR with a ligand stimulation. A, HeLa cells were pretreated with 10 M SB203580 or 5 M U0126 for 30 min and then stimulated with 100 ng/ml TNF-␣ or 3 or 100 ng/ml EGF for another 15 min. The subcellular localization of EGFR was analyzed by confocal fluorescent microscopy. DAPI, 4Ј,6-diamidino-2-phenylindole. Scale bar ϭ10 m. B, the signal intensities of the internalized EGFR dots in A were calculated as a gray value. At least 50 cell profiles were counted, and data represent the mean Ϯ S.D. *, p Ͻ 0.01, n.s., not significant. C and D, HeLa cells were transfected with siRNAs against p38␣ or the negative control and incubated for 48 h. Cells were stimulated with 20 ng/ml TNF-␣ or 3 or 100 ng/ml EGF for 15 min, and the subcellular localization of EGFR (green) was then analyzed by confocal fluorescent microscopy (C). Scale bars ϭ 10 m. The knockdown efficiency of p38 was assessed by immunoblotting (D). E, HeLa cells were pretreated with 10 M SB203580 for 30 min and then stimulated with 3 or 100 ng/ml of TGF-␣ or HB-EGF for another 15 min. The subcellular localization of EGFR was analyzed. Scale bar ϭ10 m. F, A549 cells were pretreated with 10 M SB203580 for 30 min and then stimulated with 3 or 100 ng/ml EGF for another 15 min. The subcellular localization of EGFR was analyzed. G and H, HeLa cells were pretreated with 10 M SB203580 or 5 M U0126 for 30 min and then stimulated with 100 ng/ml TNF-␣ or 3 or 100 ng/ml EGF for another 15 min. Scale bar ϭ10 m. Cell-surface EGFR expression was investigated by flow cytometry (G) and immunofluorescence (H). Cont, control. Scale bar ϭ10 m.

Dual-mode regulation of EGFR endocytic trafficking dynamics
high EGF was dependent on p38 (Fig. 4C). Importantly, Tyr(P)-1068 of GFP-tagged dd-EGFR was not detected even with high EGF stimulation, whereas GFP-tagged WT EGFR and endogenous EGFR were effectively activated, indicating that dd-EGFR did not form dimers with endogenous EGFR (Fig. 4D). In addition, phosphorylation of Ser-1046/1047, typical p38 target sites, of dd-EGFR and endogenous EGFR was detected to a similar extent, indicating that dd-EGFR only receives non-canonical regulation from endogenous EGFR (Fig. 4D).
We then examined the endocytosis of dd-EGFR in CHO-K1 cells expressing negligible levels of endogenous EGFR to assess its potential as a tool for non-canonical regulation. In contrast to the WT EGFR, dd-EGFR did not internalize, even in the presence of high EGF, whereas anisomycin, a p38 activator, efficiently triggered the endocytosis of dd-EGFR and WT EGFR (Fig. 4E). TNF-␣-induced endocytosis of dd-EGFR in a p38-dependent manner was also observed, indicating that it maintains the potential for non-canonical but not canonical endocytosis ( Fig. 4F and Fig. S3A). We investigated whether the ligandinduced activation of endogenous EGFR signaling induces the internalization of dd-EGFR in HeLa cells. As expected, low EGF induced the endocytosis of WT EGFR and dd-EGFR to a similar extent (Fig. 4G). Moreover, endocytosis of dd-EGFR was completely abolished by inhibition of p38, whereas WT EGFR remained partially internalized ( Fig. 4G and Fig. S3B). In addition, the endocytosis of dd-EGFR was clathrin-dependent ( Fig.  4, H and I, and Fig. S4A), and endocytosed dd-EGFR co-localized with EEA1 (Fig. S4B). These results demonstrate that ligand-induced and p38-mediated non-canonical EGFR endocytosis occurs in a dimerization-independent manner.

Identification of Ser/Thr sites controlling EGFR endocytosis
Previous studies reported that the p38-mediated phosphorylation of Ser-1015/Thr-1017/Ser-1018 (region 1, R1) or Ser-1046/Ser-1047 (region 2, R2) is important for stress signal-induced EGFR endocytosis (16,22,23) (Fig. 5A). As described above, dd-EGFR was employed to identify amino acid residues involved in ligand-induced non-canonical endocytosis. The substitution of ATP-binding Lys-721 to alanine (K721A) did not affect ligand-induced non-canonical endocytosis, confirming that tyrosine kinase activity is not required for the endocytosis of dd-EGFR ( Fig. 5B and Fig. S5A). Crucial phosphorylation sites for endocytosis were identified by the alanine substitution of serine/threonine residues in R1 and R2. Ligandinduced shifts in bands disappeared in the R1 mutant (R1m) but not R2m (Fig. 5C). In addition, R1m did not internalize following a low EGF stimulation, although the single S1015A mutation did not impair this, indicating that the multiple phosphorylation of R1 is involved in p38-mediated non-canonical EGFR endocytosis ( Fig. 5D and Fig. S5, B and C). A dileucine motif (Leu-1010 and Leu-1011) near R1, which was identified as an important site for EGFR endocytosis via an unknown mechanism (24,25), was also involved in non-canonical phosphorylation and endocytosis, suggesting a functional interaction between R1 and the neighboring dileucine motif (Fig. S5, D and E).
We investigated whether R1 also regulates canonical endocytosis using WT EGFR in CHO-K1 cells. GFP-tagged WT EGFR was internalized by high EGF and anisomycin (Fig. 5E). The high EGF-induced endocytosis of EGFR-R1m was still intact, whereas anisomycin-induced endocytosis was impaired ( Fig. 5E and Fig. S5F). These results demonstrate that the R1 site only regulates non-canonical endocytosis, which further supports the idea that canonical signaling regulates the noncanonical endocytosis of EGFR via p38-mediated phosphorylation of serine/threonine residues.

Phosphorylation of EGFR at Ser-1015
We generated recombinant monoclonal phospho-specific EGFR (Ser-1015) antibodies to investigate phosphorylation

Dual-mode regulation of EGFR endocytic trafficking dynamics
of R1 of endogenous EGFR. We obtained five clones available for Western blotting (Fig. S6A). Among them, we selected clone 10 for the following analyses because it is also available for immunofluorescence. Surface plasmon resonance analysis demonstrated specific binding of clone 10 to the phosphorylated antigen peptide with high affinity (K d ϭ 1.25 ϫ 10 Ϫ8 M) (Fig. S6B). Moreover, it could not recognize Ser-1015-mutated EGFR in immunoblotting (Fig. S6C).
In a time course analysis, both Ser(P)-1015 in R1 and Ser(P)-1047 in R2 were rapidly induced within 5 min in a p38-dependent manner in TNF-␣-treated HeLa cells (Fig. 6, A and B). Similarly, low EGF induced Ser(P)-1015 in a p38-dependent manner, in which tyrosine autophosphorylation was observed prior to serine phosphorylation (Fig. 6, C and D, and Fig. S6D). In addition, antibodies against Ser(P)-1015 and Ser(P)-1047 recognized the shifted bands in the Phos tag gel (Fig. S6E). Moreover, the mobility of the Tyr(P)-1068 band was shifted down when Ser(P)-1015 was inhibited by SB203580, suggesting that Tyr(P) and Ser(P)-occurred on a single EGFR molecule (Fig. S6F).

Dual-mode regulation of EGFR endocytic trafficking dynamics
was inhibited by SB203580 (Fig. S6G). Moreover, internalized dd-EGFR was also phosphorylated at Ser-1015, indicating that non-canonical phosphorylation occurred on monomeric EGFR (Fig. 6G). In a time course analysis, compared with wildtype EGFR, endocytosis of dd-EGFR was delayed until 15 min (Fig. 6H). In addition, rapid dephosphorylation of Ser-1015 within 30 min in immunoblotting (Fig. 6, A and C) was linked to the disappearance of endocytosed Ser(P)-EGFR at 30 min with low EGF stimulation (Fig. 6I). Together, these results demonstrated that Ser(P)-1015 closely correlated with endocytosis as well as the fate of EGFR after endocytosis.

Endocytic mechanisms determine post-endocytic pathways of EGFR
The mechanisms by which internalized EGFR are sorted to different endosomes for degradation or recycling are not fully understood. Because TNF-␣ induced selective endocytic recycling without degradation (14), we hypothesized that ligandinduced endocytosis via the non-canonical pathway is linked to the preferential recycling of EGFR. On the other hand, the canonically activated ligand-EGFR complex was mainly delivered to degradative lysosomes. Immunofluorescence demonstrated that endocytosed EGFR with a 15-min low-EGF stimu-

. Endocytosis of EGFR monomers via EGF-induced p38 activation.
A, schematic of ligand-induced EGFR endocytosis. Ligand (red circle) binding to WT EGFR caused tyrosine phosphorylation (pY) of the EGFR asymmetric dimer. Activated WT EGFR induced the p38-dependent serine/threonine phosphorylation (pST) of dd-EGFR harboring a deletion (⌬CR1) and point mutations (I682Q and V924R). B and C, HeLa cells were transiently transfected with GFP-tagged dd-EGFR. Cells were stimulated with the indicated concentrations of EGF for 10 min (B). Transfected cells were pretreated with 10 M SB203580 for 30 min and then treated with 3 or 100 ng/ml EGF for another 10 min (C). The Phos tag shift was assessed by immunoblotting (IB) with an anti-GFP antibody. D, HeLa cells were transiently transfected with GFP-tagged WT or dd-EGFR and stimulated with 100 ng/ml EGF for 10 min. The phosphorylation and total expression of endogenous (endo) and exogenous (GFP) EGFR were assessed by immunoblotting. E, CHO-K1 cells were transiently transfected with GFP-tagged WT or dd-EGFR and then stimulated with 100 ng/ml EGF for 15 min or 50 M anisomycin for 30 min. The subcellular localization of EGFR-GFP was analyzed. Cont, control; DAPI, 4Ј,6-diamidino-2-phenylindole. Scale bar ϭ10 m. F and G, HeLa cells were transfected with GFP-tagged WT or dd-EGFR, pretreated with 10 M SB203580 for 30 min, and then stimulated with 20 ng/ml TNF-␣ (F) or 5 ng/ml EGF (G). The subcellular localization of EGFR-GFP was analyzed. H and I, HeLa cells were transfected with siRNAs against CHC or the negative control. After a 48-h incubation, cells were further transfected with GFP-tagged WT or dd-EGFR and stimulated with 5 ng/ml EGF. Scale bar ϭ10 m. The knockdown efficiency of CHC was confirmed by immunoblotting (H). The subcellular localization of EGFR-GFP was analyzed (I). Scale bar ϭ10 m.

Dual-mode regulation of EGFR endocytic trafficking dynamics
lation was strongly recycled at 60 min, whereas high EGF did not induce recycling (Fig. 7A). These results correlated with the ligand-binding capacity to the cell surface, on which binding of rhodamine-conjugated EGF disappeared once after 15 min but largely recovered at 60 min (Fig. 7B). A flow cytometric analysis clearly demonstrated that the recycling/degradation balance was controlled by the concentrations of EGF and TGF-␣ (Fig.  7C). Furthermore, the degradation of EGFR and tyrosine phosphorylation of Cbl ubiquitin ligase showed similar concentration dependences (Fig. S7, B and C). These results strongly suggest that preferential recycling under low-ligand conditions reflects the phenomenon of p38-regulated EGFR via the noncanonical pathway; therefore, we next attempted to clarify the effects of SB203580 on the TGF-␣-induced intracellular trafficking of EGFR. We confirmed that EGF and TGF-␣ induced similar concentration-and time-dependent activation of EGFR and its downstream pathways (Fig. S7D). The kinetics of EGFR internalization were examined in more detail by measuring cell-surface EGFR using flow cytometry. As expected, EGFR was internalized in the first 15 min, and then ϳ50% of EGFR was recycled back to the cell surface within 30 -60 min (Fig. 7D and Fig. S7E). Delayed internalization from 4 -15 min was selectively inhibited by SB203580 whereas early phase internalization within 4 min was not (Fig. 7D). In addition, endocytosed EGFR in the presence of SB203580 was not recycled (Fig. 7D and Fig. S5B). The p38 inhibitor exerted similar effects on EGF-and TGF-␣-treated cells (Fig. 7E).
To confirm whether two independent endocytic mechanisms influence the trafficking routes of EGFR, dd-EGFR was expressed at a similar level to endogenous EGFR in HeLa cells, and their behaviors were monitored. A stimulation with high EGF resulted in the sufficient degradation of exogenous GFPtagged WT EGFR and endogenous EGFR. In contrast, GFPtagged dd-EGFR was not degraded until 180 min; it was recycled back to the plasma membrane, even with high-EGF stimulation (Fig. 7, F-H), indicating that the EGFR status influences endocytic trafficking. Collectively, these results reveal that canonical and non-canonical endocytic mechanisms determine the intracellular trafficking of EGFR; namely, degradation or recycling, respectively. The most important result here is that a major recycling component is ligandunbound EGFR, which is internalized via a p38-dependent mechanism. with GFP-tagged dd-EGFR or dd-EGFR with the S1015/T1017/S1018A mutations in region 1 (R1m) or with the S1046/7A mutations in region 2 (R2m) and then stimulated with 5 ng/ml EGF for 15 min. The Phos tag shift was assessed by immunoblotting (IB) with an anti-GFP antibody. D, HeLa cells were transiently transfected with GFP-tagged dd-EGFR, ddϩR1m, or ddϩR2m and stimulated with 3 ng/ml EGF for 15 min. The subcellular localization of EGFR-GFP was analyzed. Scale bar ϭ10 m. E, CHO-K1 cells were transiently transfected with GFP-tagged WT EGFR or EGFR-R1m and stimulated with 100 ng/ml EGF for 15 min or anisomycin for 30 min. The subcellular localization of EGFR-GFP was analyzed. Scale bars ϭ 10 m.

Dual-mode regulation of EGFR endocytic trafficking dynamics Transient suppression of ligand-induced EGFR activation
We demonstrated previously the TNF-␣-induced transient suppression of ligand-induced EGFR activation by prior noncanonical phosphorylation (14). In this study, we showed that low-ligand conditions established a similar intracellular environment; therefore, we attempted to investigate the effects of a 10-min low-ligand pretreatment on high ligand-induced EGFR activation (Fig. 8A). The pretreatment resulted in a reduction in secondary high ligand-induced Tyr(P)-EGFR from the no pretreatment control (Fig. 8B). In contrast, recycled EGFR at 60 min was sufficiently functional to be activated again by the extracellular ligand (Fig. 8C). These results correlated with The subcellular localization of total EGFR, Ser(P)-1015-EGFR, and Ser(P)-1047-EGFR was analyzed by confocal fluorescent microscopy. DAPI, 4Ј,6-diamidino-2-phenylindole. Scale bar ϭ10 m. G, HeLa cells were transiently transfected with GFP-tagged dd-EGFR and then stimulated with 3 ng/ml EGF or 20 ng/ml TNF-␣ for 15 min. The subcellular localization of EGFR-GFP and Ser(P)-1015-EGFR was analyzed. Scale bars ϭ 10 m. H, HeLa cells were transiently transfected with GFP-tagged WT-EGFR or dd-EGFR and stimulated with 5 ng/ml EGF for 5 or 15 min. The subcellular localization of EGFR-GFP was analyzed by confocal fluorescent microscopy. I, HeLa cells were stimulated with 20 ng/mL TNF-␣ or 3 ng/ml EGF for indicated time. The subcellular localization of total EGFR or Ser(P)-1015-EGFR was analyzed.

Dual-mode regulation of EGFR endocytic trafficking dynamics
ligand-binding capacity to the cell surface, on which binding disappeared once after 15 min but largely recovered at 60 min (Fig. 7B). Collectively, the ligand-induced non-canonical control of EGFR causes transient feedback inhibition in the early phase but subsequently may contribute to the sequential activation of growth factor signaling.

Discussion
A major unsolved issue in the ligand-induced endocytic trafficking mechanisms of EGFR is how it is recruited to the different endocytic machineries and then sorted to different routes for recycling or degradation (5-7, 26, 27). In this study, we demonstrated the dual-mode regulation of the ligand-induced

Dual-mode regulation of EGFR endocytic trafficking dynamics
endocytic trafficking dynamics of EGFR (Fig. 9). Because minimal activation of EGFR was sufficient to fully activate MAPKs, many unliganded cell-surface EGFR are targets for feedback phosphorylation by activated MAPKs. Therefore, non-canonical endocytosis is the main event following stimulation under lowligand conditions. In contrast, the majority of EGFR may initially be occupied by ligands under high-ligand conditions, which induces dimerization-dependent tyrosine kinase activation and internalization of a ligand-EGFR complex before activation of p38. We propose that two different forms of EGFR are internalized following ligand stimulation: the preceding ligand-bound Tyr(P)-EGFR dimers and delayed unliganded Ser/Thr(P)-EGFR monomers, which result in straightforward transport to lysosomal degradation and recycling to the cell surface, respectively. Thus, ligand-induced EGFR trafficking involves the complex parallel events of canonical and non-canonical endocytosis, which are balanced by the number of cell-surface EGFR proteins, initial ligand occupation rate, and MAPK activation levels.
Another factor to consider in the post-endocytic fate of EGFR is ligand specificity. TGF-␣, for example, appears to dissociate from EGFR at endosomal acidic pH, allowing the receptor to escape from lysosomal degradation and instead be efficiently recycled to the cell surface. In contrast, EGF preferentially causes lysosomal degradation over TGF-␣ because EGF-EGFR binding remains stable at acidic pH in endosomes (5,28). An integrated multilayered proteomics approach using a high concentration of ligands (100 ng/ml) recently identified Rab7 tyrosine phosphorylation and the recruitment of the Rabcoupling protein RCP to EGFR as molecular switches dictating TGF-␣-and EGF-dependent EGFR trafficking (29). We demonstrated that EGF and TGF-␣ both provoked the canonical and non-canonical endocytic trafficking of EGFR, indicating that not only the stability of the ligand-EGFR complex in endosomes but also the balance between canonical and non-canonical trafficking are important mechanisms responsible for ligand specificity.
Ligand-induced endocytosis/recycling of ligand-unbound EGFR, the new setting in this study, is similar to TNF-␣induced clathrin-mediated endocytosis and recycling. This study clearly demonstrates, using dd-EGFR, that low EGF as well as TNF-␣ induced tyrosine kinase-independent and clathrin-dependent endocytosis of EGFR monomers through p38-dependent phosphorylation of the C-terminal region 1 around Ser-1015. Low EGF and TNF-␣ both induce transient p38 activation; therefore, EGFR undergoes plasma membrane recycling within 30 -60 min of the dephosphorylation of region 1. Furthermore, ubiquitination is not involved in p38-dependent endocytic trafficking (14,17). Therefore, low EGF and TNF-␣ appear to drive a common endocytosis/recycling system under the control of p38 activation. A recent study reported that cellular stresses, including UV C and the genotoxic agent cisplatin, trigger sustained p38 activation and the non-canonical endocytosis of EGFR, in which EGFR accumulates in a subset of lysobisphosphatidic acid-rich perinuclear MVBs through a mechanism involving the actin polymerization-promoting protein WASH and the endosomal sorting complex containing ALIX and ESCRT (30). Inhibition of p38 results in recycling of intraluminally sorted EGFR to the cell surface. We also demonstrated previously that high osmotic conditions induced the sustained internalization of EGFR during persistent p38 activation (14). We speculate that non-canonically endocytosed EGFR under low-EGF conditions also accumulate in lysobisphosphatidic acid-rich MVBs and are then rapidly sorted to recycling endosomes after the dephosphorylation of region 1 because of the rapid turnover of p38 activation. Thus, p38 activation kinetics may be a critical determinant of the retention time of Ser/Thr(P)-EGFR in MVBs. In any case, endocytosed Ser/Thr(P)-EGFR is not sorted to a distinct subset of degradative MVBs, the main sorting compart-

Dual-mode regulation of EGFR endocytic trafficking dynamics
mentsoftheactivatedligand-EGFRcomplexwithtyrosinephosphorylation and ubiquitination. EEA1 (31), an early endosomal marker, and Eps15 (32), a clathrin adapter protein, are also phosphorylated by p38, suggesting that the clathrin-dependent endocytosis and endosomal sorting of EGFR are completely controlled by p38 signaling pathways.
Convincing evidence has been reported for ubiquitin-dependent targeting of EGFR to lysosomal degradation (6,33,34). A threshold-controlled ubiquitination model was recently proposed in which the E3 ligase Cbl is recruited in complex with Grb2 to EGFR with Tyr(P)-1045 (35). We confirmed that high EGF efficiently induced EGFR degradation, even when p38 was activated. However, p38 is not involved in the early canonical internalization process of the ligand-EGFR complex because canonical endocytosis occurred prior to p38-mediated non-canonical phosphorylation (Fig. 6C). In contrast, the results of the Phos tag analysis suggest that ligand-activated Tyr(P)-EGFR is additionally phosphorylated by p38 after internalization (Fig.  2E). An in vitro study previously demonstrated that the phosphorylation of Ser-1046/1047 reduced the binding affinity of Cbl to Tyr(P)-1045 (36). More importantly, the threshold EGF concentration for EGFR ubiquitination (ϳ3 ng/ml) was similar to the minimally effective concentration for the p38-mediated non-canonical regulation of EGFR in HeLa cells, suggesting a role of p38 in threshold-controlled lysosomal targeting of ubiquitinated EGFR. There is a possibility that Ser-1015/Thr-1017/Ser-1018 in R1 and Ser-1046/Ser-1047 in R2 play distinct roles in triggering non-canonical endocytosis and preventing ubiquitination of EGFR. Further characterization is essential for understanding the potential role of p38-mediated serine/threonine phosphorylation in the ubiquitination-dependent degradation of EGFR.
The intensity and duration of receptor activation are known to affect cellular responses to a ligand (37,38). Low EGF has been shown to stimulate cell proliferation in a similar manner as high EGF, and sustained EGFR signaling is controlled by clathrin-mediated endocytosis in HeLa cells (8). In this study, we clarified the importance of the MAPK-mediated feedback phosphorylation of EGFR following low-EGF stimulation, which may cause transient impairments in association with the extracellular ligand with endocytosed EGFR across the membrane. ERK-mediated Thr-669 phosphorylation in the juxtamembrane domain, a negative feedback site, may also be involved in the inhibition of EGFR tyrosine kinase activity (39). In any case, it is important for recycled EGFR to be sufficiently functional to respond to the ligand, and the continuous existence of active ligands in the extracellular environment may be necessary for sustained EGFR signaling (Fig. 7, A and  B). These results suggest that EGF induces multiphase activation of EGFR on the plasma membrane, which may be beneficial for sustaining receptor signaling.
Mice with kinase-inactive EGFR have some eye and skin defects but better survival than EGFR-deficient mice, indicating kinase-independent roles for EGFR (15). We demonstrated previously that EGFR plays an anti-apoptotic role in the TNF-␣ signaling pathway (12). The most notable finding from recent studies is the role of EGFR in autophagy. Serum starvation induces the endosomal arrest of inactive EGFR via an interaction with the endosomal protein lysosomal-associated protein transmembrane 4 ␤ (LAPTM4B) (40). EGFR then induces the In addition, canonically activated EGFR induces p38 activation, which leads to serine/threonine phosphorylation-and clathrin-dependent endocytosis of monomeric EGFR (non-canonical pathway, shown in green). Thus, ligandinduced EGFR trafficking involves the complex parallel events of canonical and non-canonical endocytosis, which are balanced by the ligand concentration, reflecting the initial ligand occupation rate. Under low-ligand conditions, a large amount of ligand-unbound EGFR is internalized via the non-canonical pathway and then recycled back to the cell surface. Conversely, ligand-occupied EGFR is internalized mainly via the canonical pathway under high-ligand conditions, which is preferentially sorted to the lysosome for degradation. CME, clathrin-mediated endocytosis; CIE, clathrin-independent endocytosis.

Dual-mode regulation of EGFR endocytic trafficking dynamics
dissociation of the Run domain Beclin-1-interacting and cysteine-rich-containing protein (Rubicon) from the Beclin-1 complex, which activates the initiation of autophagy by Beclin-1. EGFR tyrosine kinase inhibitors, including gefitinib and erlotinib, may also trigger autophagy in cancer cells (15). In contrast, EGF, tested at 100 ng/ml, inhibits the initiation of autophagy by decreasing the interaction between EGFR and LAPTM4B (40). In addition, EGFR-mediated tyrosine phosphorylation of Beclin-1 is also involved in the suppression of autophagy (41). EGFR may be sorted to different endosomes/ MVBs in a ligand concentration-dependent manner; therefore, it will be interesting to evaluate the effects of low EGF, which may induce endocytosis of many inactive EGFR, on the initiation of autophagy. Further studies are needed to fully understand the physiological roles of inactive EGFR in the ligand-induced dual-mode activation model.
In summary, here we describe a new concept for the ligandinduced regulation of multiple receptor functions in EGFR systems. The results obtained may be applied to other ligand receptor systems in cellular signaling. Moreover, a comprehensive understanding of the feedback regulation of receptors is the next important challenge in signal transduction research and will contribute to the oncology field by providing information for identifying new therapeutic targets and overcoming resistance to anti-cancer agents.

Dual-mode regulation of EGFR endocytic trafficking dynamics Zn 2؉ Phos tag SDS-PAGE
The procedures for Zn 2ϩ Phos tag SDS-PAGE were described previously (32,42). Cell lysates were prepared with radioimmune precipitation assay buffer (50 mM Tris-HCl (pH7.4), 0.15 M NaCl, 0.25% sodium deoxycholate, 1.0% Nonidet P-40, 1.0 mM EDTA, 20 mM ␤-glycerophosphate, 1 mM sodium orthovanadate, 1 mM phenylmethylsulfonyl fluoride, 1 mM dithiothreitol, 10 g/ml aprotinin, and 10 g/ml leupeptin). Each sample was mixed with a half volume of SDS-PAGE sample buffer (195 mM Tris-HCl (pH 6.8), 3.0% SDS, 15% 2-mercaptoethanol, 30% glycerol, and 0.1% bromphenol blue) and heated at 95°C for 5 min. The acrylamide pendant Phos tag ligand and two equivalents of ZnCl 2 were added to the separating gel before polymerization. The running buffer consisted of 100 mM Tris and 100 mM MOPS containing 0.1% SDS and 5 mM sodium bisulfite. After electrophoresis, the gel was washed twice with a solution containing 25 mM Tris, 192 mM glycine, 10% methanol, and 1.0 mM EDTA for 20 min and then washed once with a solution containing 25 mM Tris, 192 mM glycine, and 10% methanol for 20 min. Gel transfer, blocking, antibody reactions, and detection were performed according to the normal immunoblotting protocol described above.

Flow cytometry
HeLa cells were harvested in PBS. Cells were fixed with 2% paraformaldehyde at room temperature for 15 min. Cells were resuspended in 100 l FACS buffer (PBS containing 0.5% BSA) containing 0.5 g of an anti-EGFR monoclonal antibody (clone LA1, Upstate) and incubated on ice for 1 h. After being washed with FACS buffer, cells were incubated with a fluorescein isothiocyanate-conjugated anti-mouse immunoglobulin G antibody (Dako) on ice for 1 h in the dark and then analyzed using the FACS Calibur system (BD Biosciences).

Immunofluorescence
Cells were seeded on coverglasses. Two days after seeding, cells were incubated with inhibitors and ligands or transfected with plasmid DNAs. Cells were rinsed in cold PBS and fixed in 4% paraformaldehyde for 15 min or methanol for 10 min. After fixation with paraformaldehyde, cells were permeabilized in PBS containing 0.5% Triton X-100 and washed with PBS. Cells were incubated for 1 h with a primary antibody and then washed and incubated with isotype-specific secondary antibodies conjugated to Alexa Fluor (Invitrogen) for 30 min. These antibodies were diluted in PBS containing 0.5% BSA. Microscopy was performed using an LSM 700 confocal microscope (Zeiss, Oberkochen, Germany).
For quantification, signal intensities of internalized EGFR dots were calculated as a gray value. At least 50 cell profiles were counted, and data represent the mean Ϯ S.D. We confirmed the reproducibility of the data in more than two independent experiments, and a representative result is shown.

Generation of rabbit monoclonal antibodies against Ser(P)-1015-EGFR
Phospho-specific monoclonal antibodies were generated using the rabbit immunospot array assay on a chip system, as described previously (43,44). The synthetic peptides EGFR peptide (TPLLSSLSATSNNST), EGFR peptide phosphorylated at Ser-1015 (Ser(P)-EGFR; TPLLSSL(pS)ATSNNST), biotinylated Ser(P)-EGFR peptide, and KLH-conjugated pS-EGFR peptide were obtained from Eurofins (Tokyo, Japan). A rabbit was immunized with the KLH-conjugated Ser(P)-EGFR peptide. IgG was purified, titrated by an enzyme-linked immunosorbent assay, and applied for Western blotting and immunofluorescence. Experiments using rabbits were approved by the Committee on Animal Experiments at the University of Toyama.

Data processing and statistical analysis
We confirmed the reproducibility of the data in more than three independent experiments, and a representative result is shown. Quantitative analysis of immunoblots and immunofluorescence was performed using densitometry with ImageJ software. Values are shown as the mean Ϯ S.D. The significance of differences was assessed by Student's t test. p Ͻ 0.01 was considered to be significant.