Research ArticleCell Biology

AKT Facilitates EGFR Trafficking and Degradation by Phosphorylating and Activating PIKfyve

See allHide authors and affiliations

Science Signaling  11 Jun 2013:
Vol. 6, Issue 279, pp. ra45
DOI: 10.1126/scisignal.2004015


Epidermal growth factor receptor (EGFR) is a receptor tyrosine kinase (RTK) that controls cell proliferation, growth, survival, metabolism, and migration by activating the PI3K (phosphatidylinositol 3-kinase)–AKT and ERK (extracellular signal–regulated kinase)–RSK (ribosomal S6 kinase) pathways. EGFR signaling to these pathways is temporally and spatially regulated. Endocytic trafficking controls the access of EGFR to these downstream effectors and also its degradation, which terminates EGFR signaling. We showed that AKT facilitated the endocytic trafficking of EGFR to promote its degradation. Interfering with AKT signaling reduced both EGFR recycling and the rate of EGFR degradation. In AKT-impaired cells, EGFRs were unable to reach the cell surface or the lysosomal compartment and accumulated in the early endosomes, resulting in prolonged signaling and increased activation of ERK and RSK. Upon EGF stimulation, AKT phosphorylated and activated the kinase PIKfyve [FYVE-containing phosphatidylinositol 3-phosphate 5-kinase], which promoted vesicle trafficking to lysosomes. PIKfyve activation promoted EGFR degradation. Similar regulation occurred with platelet-derived growth factor receptor (PDGFR), suggesting that AKT phosphorylation and activation of PIKfyve is likely to be a common feedback mechanism for terminating RTK signaling and reducing receptor abundance.


Epidermal growth factor receptor (EGFR) is a major regulator of cell proliferation, growth, survival, metabolism, and motility and is overexpressed or inappropriately activated in many cancers (1, 2). EGFR carries out these functions by activating multiple signaling cascades, including the phosphatidylinositol 3-kinase (PI3K)–AKT, mammalian target of rapamycin complex 1–p70 S6 kinase (mTORC1-S6K), and extracellular signal–regulated kinase–p90 ribosomal S6 kinase (ERK-RSK) pathways. PI3K, mTORC1, and ERK variably induce distinct cellular functions depending on the cell type, cell cycle time, and magnitude and duration of pathway activation (3). Upon ligand binding, EGFR molecules transphosphorylate each other on multiple tyrosine residues to create docking sites for the Grb2 and GAB1 adapter proteins (4). GAB1 recruitment results in EGFR activation of type 1 PI3Ks, which generate phosphatidylinositol 3,4,5-trisphosphate [PI(3,4,5)P3]. Pleckstrin homology (PH) domains in AKT, phosphoinositide-dependent kinase 1 (PDK1), and other molecules recognize PI(3,4,5)P3, and the interaction brings AKT and PDK1 to the plasma membrane. PDK1 and mammalian target of rapamycin complex 2 (mTORC2) then activate AKT by phosphorylating AKT at Thr308 and Ser473, respectively. Grb2 recruitment to EGFR results in activation of the RAS–RAF–mitogen-activated or extracellular signal–regulated protein kinase kinase (MEK)–ERK–RSK signaling cascade. In response to EGF stimulation, AKT, ERK, and RSK all contribute to the activation of the mTORC1-S6K pathway. EGFR and these downstream signaling pathways are regulated through a network of feedback and crosstalk mechanisms (3).

Receptor endocytosis is a regulatory mechanism that promotes sustained and spatially regulated signaling by localizing receptors to signaling endosomes and by promoting receptor recycling to the cell surface (57). Alternatively, endocytosis can lead to signal attenuation by culminating in receptor degradation. EGFR endocytosis is initiated by EGF binding to EGFR dimers at the plasma membrane (8). Stabilization of EGFR dimers promotes EGFR activation and transphosphorylation. Active EGFR is ubiquitinated by the E3 ligase Cbl, a posttranslational modification that recruits the endocytic machinery. Both clathrin-dependent (9) and clathrin-independent (10, 11) pathways contribute to EGFR endocytosis. Receptor internalization is followed by localization to early endosome antigen 1 (EEA.1)–positive endosomes, where cargos destined for recycling or degradation are separated (12, 13). EGFR molecules are recycled back to the plasma membrane from the early endosomes and the limiting membrane of multivesicular bodies (MVBs) in a Rab4- and Rab11-dependent manner. Recycled EGFRs engage in additional rounds of endocytosis and signaling (14). Alternatively, protein tyrosine phosphatase 1B (PTP1B) can dephosphorylate EGFR at the limiting membrane of MVBs (15, 16). Dephosphorylated EGFRs enter the MVB lumen through the endosomal sorting complex for transport (ESCRT) complexes (10, 17). These EGFRs are dissociated from signal-transducing molecules, and signaling is terminated. These EGFRs are destined for degradation in the lysosomes.

Proteins involved in EGFR sorting and degradation such as EEA.1 and ESCRT proteins are recruited to the endocytic vesicles through their interaction with phosphoinositides. For example, endomembranes contain phosphatidylinositol 3-phosphate [PI(3)P], which is recognized by FYVE (Fab1, YOTB, Vac1, EEA.1) domains found in these respective proteins. Fab1 is a phosphoinositide kinase that phosphorylates PI(3)P to generate phosphatidylinositol 3,5-bisphosphate [PI(3,5)P2]). The phosphoinositide phosphatase SAC3 (SAC domain–containing inositol phosphatase 3) dephosphorylates PI(3,5)P2 at the 5 position to generate PI(3)P (18). In yeast, deletion of Fab1 disrupts cargo sorting to the yeast vacuoles (19). The human homolog of Fab1 is called PIKfyve [FYVE-containing PI(3)P 5-kinase], which forms a complex with ArPIKfyve (associated regulator of PIKfyve) and SAC3 at the endomembranes. PIKfyve facilitates and SAC3 inhibits the progression of early endosomes toward MVBs, suggesting that PI(3,5)P2 promotes and PI(3)P inhibits vesicle progression (20). However, the regulation of the PIKfyve-ArPIKfyve-SAC3 complex in mammalian cells is not well understood.

Previous work has implicated type I PI3Ks, which are activated by insulin and EGF, in the modulation of vesicular trafficking. Mutagenesis of the binding site of platelet-derived growth factor receptor (PDGFR) for the type I PI3K regulatory subunit p85 blocks degradation of PDGFR (21). The recruitment of the PI3K catalytic subunit p110 to the plasma membrane accelerates clathrin coat dynamics (22). Injection of p110α-blocking antibodies causes transferrin to accumulate within the cells, suggesting that the antibodies inhibit transferrin recycling (23). The PI3K inhibitor wortmannin also suppresses transferrin recycling in cells and reduces the rate of endosomal sorting in cell-free systems (24, 25). Two developments prompted us to investigate whether PI3K signals through AKT to modulate trafficking: AKT inhibitors increase the protein abundance of several receptor tyrosine kinases (RTKs), and AKT knockdown reduces transferrin and EGF uptake (26, 27).

We investigated the role of AKT in EGFR trafficking and discovered a negative feedback loop in which EGF-mediated activation of AKT promotes EGFR progression through the early endosomes and EGFR degradation by activating PIKfyve. Multiple AKT inhibitors and AKT knockdown reduce EGFR progression through the early endosomes, the rate of EGFR degradation, and PIKfyve activity in vitro and in cells. Similarly, knocking down PIKfyve or its activator ArPIKfyve or treatment with an inhibitor of PIKfyve reduces the rate of EGFR degradation. The reduced rate of EGFR degradation produced by AKT inhibition is rescued by knockdown of the phosphatase SAC3. Further, expressing a PIKfyve mutant that cannot be phosphorylated by AKT reduces EGFR degradation. We propose a model in which AKT phosphorylates and activates PIKfyve to facilitate EGFR endosomal progression, thereby increasing EGFR degradation and dampening EGFR signaling.


AKT promotes EGFR degradation

EGF stimulation causes EGFR degradation by delivery to the lysosomes (14). To address the role of AKT in EGFR degradation, we assessed EGFR degradation in human mammary epithelial cells (HMECs), which have 4.5 × 105 EGFR molecules per cell (28). This concentration of EGFRs is within the range of other cell lines commonly used in endocytosis studies. HeLa cells have 1.7 × 105 EGFRs per cell (28). MDA-MB-468 and A431 cells, which harbor amplified EGFR, express three- and sixfold more EGFR than HMECs, respectively (2931). Stable overexpression of EGFR in HMECs led to a 25 to 50% increase in the total EGFR abundance (fig. S1A), so the approximate number of EGFR molecules is 6 × 105. We used the latter cells for our studies because they were more tractable to interrogation and exhibited an intermediate EGFR abundance compared to the above-mentioned commonly used cell lines. Stimulation of these HMECs with a high concentration of EGF (100 ng/ml) induced a decrease in the abundance of EGFR in total cell lysate, which was abrogated by pretreatment with the lysosomal inhibitor chloroquine (fig. S1, B and C). Similar to HeLa cells, a low dose of EGF (1 ng/ml) did not cause EGFR degradation (fig. S1B), suggesting that a specific degree of EGFR activation is needed to either induce degradation or detect the degradation. We hypothesized that perturbations in EGFR trafficking to the lysosomes would alter its degradation rate. We treated cells with vehicle [dimethyl sulfoxide (DMSO)], the dual PI3K-mTOR inhibitor PI103 (32), or the highly specific allosteric AKT inhibitor AKTVIII (33). PI103 interferes with PI3K activity and thus reduces PI(3,4,5)P3 concentrations and phosphorylation of AKT at Ser473 and Thr308, leading to AKT inactivation. AKTVIII stabilizes an AKT conformation in which the PH and kinase domains are locked together (34), which is proposed to interfere with AKT membrane recruitment and the access of PDK1 and mTORC2 to AKT. The inhibitor therefore reduces the phosphorylation of AKT at Ser473 and Thr308 and AKT activation. Membrane recruitment of the PH domain–containing protein general receptor of phosphoinositides 1 (GRP1) is unaffected by AKTVIII, suggesting that the drug specifically interferes with membrane recruitment of AKT (35).

In HMECs pretreated with DMSO, EGF stimulation caused more than 80% of EGFR to be degraded within 120 min (Fig. 1A). Pretreatment with AKTVIII and PI103 blocked phosphorylation of AKT at Ser473 as expected. Furthermore, pretreatment of cells with these inhibitors reduced the rate of EGF-induced EGFR degradation, and a substantial amount of EGFR was detected even after 120 min of EGF stimulation. This phenotype was specific to AKT inhibition because blocking other signaling pathways downstream of EGFR, such as the mTORC1-S6K pathway with rapamycin and the ERK-RSK pathway with U0126, did not affect EGFR degradation (Fig. 1A).

Fig. 1 AKT facilitates EGFR degradation.

(A) Starved HMECs were stimulated with EGF for the indicated times in the presence or absence of the indicated inhibitors. Western blots are representative of at least three independent experiments. (B) Cells were stimulated as above in the presence of CHX. Western blots are representative of at least two independent experiments. EGFR degradation curves are presented in fig. S1E. (C) Cells transfected with siRNAs targeting AKT1 or with nontargeting AllStars control siRNA were deprived of growth factors for 24 hours, treated with CHX for 30 min, and stimulated with EGF. Western blots are representative of at least three independent experiments. EGFR degradation curves are presented in fig. S1F.

Our observation that both PI103 and AKTVIII stabilized EGFR suggests that PI3K acted through AKT to promote EGFR degradation by modulating the trafficking of EGFR-containing vesicles or by modulating EGFR protein synthesis. To test this, we repeated the EGFR degradation assays in the presence of the protein synthesis inhibitor cycloheximide (CHX). Cotreatment of cells with CHX and PI103, AKTVIII, or a structurally distinct adenosine triphosphate (ATP)–competitive AKT catalytic inhibitor, A-443654, reduced EGFR degradation rates (Fig. 1B and fig. S1E). The effect of A-443654 on EGFR degradation was not as potent as that of AKTVIII, potentially because of its decreased specificity (36). Consistent with our previous experiments, rapamycin did not alter EGFR degradation. These data indicate that AKT promotes EGFR degradation in a protein synthesis–independent manner. AKTVIII reduced EGFR degradation in T47D breast cancer cells, which have low EGFR abundance (37), suggesting that AKT stimulates the degradation of EGFR independent of its abundance (fig. S1D). We used small interfering RNA (siRNA) knockdown of AKT1 to further validate that AKT controls EGFR degradation, and found that two different AKT1 siRNAs significantly increased the percent of EGFR left undegraded after 120 min (Fig. 1C and fig. S1F).

AKT inhibition prolongs EGF-EGFR occupancy in the early endosomes

Stimulation of cells with protein kinase C or with G protein (heterotrimeric guanine nucleotide–binding protein)–coupled receptor agonists sequesters EGFR in a perinuclear region (38), which blocks the access of EGF to EGFR and inhibits ligand-induced degradation. To determine whether AKT controls EGFR degradation by facilitating EGF access to EGFR and the resulting EGFR internalization, we quantified the colocalization of EEA.1 with Alexa Fluor 488–labeled EGF. AKTVIII treatment did not significantly alter EEA.1 colocalization with EGF (fig. S2A) or with EGFR after EGF stimulation and did not cause EGFR accumulation in the perinuclear region (fig. S2B). Together with the observation that AKT inhibition does not prevent EGFR phosphorylation (Fig. 1B), these data suggest that AKT does not promote EGFR degradation by controlling EGF-EGFR binding or their subsequent internalization.

To determine whether AKT promotes EGFR degradation by controlling vesicle trafficking, we sought to identify the subcellular location of the EGFR that accumulates upon AKT inhibition. In the presence of AKTVIII, EGFR molecules that were not degraded were still phosphorylated (Fig. 1B). Because PTP1B dephosphorylates EGFR before EGFR is sorted into the lumen of the MVBs (15), we hypothesized the undegraded EGFR accumulates at the early endosomes or the limiting membrane of MVBs. Indeed, AKTVIII treatment increased EEA.1 colocalization with EGF and EGFR 60 min after EGF stimulation (Fig. 2, A and B). In cells in which internalization was synchronized, AKTVIII treatment increased EEA.1 colocalization with EGF and EGFR as well (fig. S2C). To verify that the increased EEA.1-EGF and EEA.1-EGFR colocalization indicated increased EGF and EGFR in EEA.1-positive endosomes, we quantified the integrated intensities of the EGF and EGFR signals in each endosome throughout confocal z stacks of DMSO- and AKTVIII-treated cells. We plotted the cumulative probability distribution of endosomes against the integrated EGF/EGFR intensity values. In AKTVIII cells, endosomal EGF and EGFR intensities were significantly higher for the same probability value, suggesting that AKTVIII treatment increased the amounts of EGF and EGFR in each endosome (Fig. 2, C and D). Thus, we conclude that upon AKT inhibition, undegraded EGF and EGFR molecules accumulate in EEA.1-positive early endosomes.

Fig. 2 AKT regulates the endocytic progression of EGF/EGFR.

(A) Cells plated on coverslips were deprived of growth factors, stimulated with Alexa Fluor 488–labeled EGF (green) for 60 min, and labeled with EEA.1 (red) antibody (n > 10 fields of cells; P < 0.005). (B) Cells were stimulated with unlabeled EGF for 60 min and labeled with EEA.1 (red) and EGFR (green) (n > 15 fields of cells; P < 0.01). EGF/EGFR colocalization with EEA.1 is depicted in yellow and indicated by arrows. Scale bars, 35 μm. Values are averages of percent colocalization ± SD. P < 0.005. Panels to the right are boxplots. The “+” sign indicates outliers, and whiskers indicate maximum and minimum values. The lower end of the boxes, the middle line, and the upper edge of the boxes are 25th, 50th, and 75th quantiles, respectively. The median values indicated by the red lines are statistically significantly different at 95% confidence level. (C) Cells were treated as in (A) and (B). Images were acquired throughout the z plane. Endosomes were created from EEA.1 pixels in the 568-nm channel above a local background with a continuous shape. Endosomes were transferred onto the Alexa Fluor 488–labeled EGF image. The integrated intensity of the Alexa Fluor 488–labeled EGF signal in each endosome was calculated for every other z plane in the z stack. Graph represents the cumulative distribution function plotted against integrated EGF intensities. X axis gives the probability of a randomly selected endosome to have an EGF intensity value less than or equal to the EGF intensity value on the y axis. AKTVIII treatment decreases this probability because it creates endosomes with significantly higher EGF intensity values (n > 2300 endosomes; P < 10−7). (D) Analysis in (C) repeated for EGFR intensity in endosomes. AKTVIII treatment increases the EGFR intensity for any given probability (n > 900 endosomes; P < 10−9).

AKT promotes EGFR recycling

Disruption of receptor recycling with a dominant-negative Rab4 decreases EGF degradation (39). This is thought to be due to the inhibition of continuous rounds of receptor endocytosis in which a subset of the EGFRs would have been degraded with each cycle. We found that Rab11 knockdown reduced EGFR degradation (fig. S3). This suggests that interfering with vesicle recycling induces intracellular retention of vesicles, which prohibits additional rounds of EGF binding to EGFR and therefore additional rounds of the recycling or degradation decision. We hypothesized that AKT facilitates EGFR degradation by promoting receptor recycling and ensuring continuity of the internalization and degradation cycles driven by the EGF in the media. We used an established method to assay receptor recycling in which we measured median cell surface EGFR staining using flow cytometry at different time points: before EGF stimulation (Total), after 15 min of EGF stimulation (Pulse), and after washing the EGF-pulsed cells with acid to remove surface EGF and transferring the cells back to 37°C for 10 to 20 min to allow EGFR recycling back to the plasma membrane (Chase) (40) (fig. S4). We found that AKT inhibition reduced the rate of EGF-induced EGFR recycling (Fig. 3, A and B). Because EGF induces both the degradation and recycling of EGFR, we also assayed EGFR recycling in response to transforming growth factor–α (TGFα), an EGFR ligand that promotes EGFR recycling without substantial degradation (41). TGFα induced more robust EGFR recycling than did EGF, and TGFα-induced EGFR recycling was also reduced by AKTVIII treatment (Fig. 3, A and C).

Fig. 3 AKT regulates EGFR recycling.

Percent recycling was calculated in cells that were pretreated with AKTVIII for 20 min and stimulated with EGF or TGFα for 15 min. (A) Percent recycling values are averages of three independent experiments with SEM. *P = 0.05. (B and C) Flow cytometry histograms are representative of three independent experiments. Solid purple histograms and T represent total EGFR surface labeling before any growth factor stimulation. Green histograms with 0′ represent surface EGFR labeling after 15 min of EGF stimulation. Pink (10′) and blue (20′) histograms represent surface labeling of EGFR after allowing cells to recycle EGFR at 37°C after 0, 10, and 20 min of chase time after a pulse of 15 min with growth factors.

AKT promotes the lysosomal progression of EGFR

AKT1 promotes the localization of CD89-targeted antigen to lysosome-associated membrane protein 1 (LAMP1)–containing vesicles (42, 43). This suggests that AKT is also involved in lysosomal sorting. To determine whether the AKTVIII-induced reduction in EGFR degradation could be due to reduced lysosomal sorting, we quantified the colocalization of LAMP2 with Alexa Fluor 488–labeled EGF. EGF stimulation for 30 min caused a small amount of LAMP2 to colocalize with Alexa Fluor 488–labeled EGF. However, AKT inhibition did not significantly change the colocalization of LAMP2 with EGF (fig. S5A).

The small extent of LAMP2 and EGF colocalization suggested the possibility of EGF degradation in the lysosomal compartment. To reduce the loss of EGF signal, we used chloroquine to inhibit lysosomal degradation (fig. S1C). In the presence of chloroquine, AKTVIII significantly decreased the colocalization of LAMP2 with EGF, suggesting that AKT promotes the sorting of EGF into lysosomes (Fig. 4, A and B). Synchronization of EGF binding to receptors by stimulating the cells at 4°C before internalization also resulted in similar amounts of LAMP2 colocalization with EGF, which was also reduced by AKTVIII pretreatment (fig. S5B). Because EGF causes both EGFR recycling and lysosomal sorting, we also repeated the degradation assay with betacellulin, an EGFR ligand that induces lysosomal sorting and degradation of EGFR without detectable recycling (41). Consistent with a role for AKT in EGFR lysosomal trafficking, AKT inhibition reduced the rate of betacellulin-induced EGFR degradation (Fig. 4C).

Fig. 4 AKT regulates lysosomal progression of EGFR.

(A) Cells were pretreated with chloroquine and stimulated with Alexa Fluor 488–labeled EGF for 30 min. LAMP2 is pseudocolored in red, and EGF in green; yellow represents colocalization. Average percent colocalization values and SDs are given (n > 25 fields of cells; P < 0.005). Scale bars, 35 μm. (B) Boxplot of LAMP2 colocalization with EGF as in Fig. 2. (C) EGFR degradation was calculated in cells that were starved overnight, pretreated with DMSO or AKTVIII for 45 min and with CHX for the last 30 min of inhibitor treatment, and stimulated with betacellulin (BTC) for the indicated time points. n = 2 independent experiments.

AKT promotes EGFR degradation by phosphorylating and activating PIKfyve

We next investigated the mechanism by which AKT controls endocytic vesicle progression. Endosomal identity and the ability of endosomes to progress within the vesicular trafficking system are dictated by the phosphoinositide composition of vesicles (20, 44). For example, an increase in the concentration of the phosphoinositide PI(3,5)P2 promotes the progression of early endosomes into MVBs. Insulin stimulation, which activates AKT, increases the concentration of PI(3,5)P2 by inhibiting the phosphatase activity of SAC3 (45). Thus, we tested whether AKT functioned upstream of or in a pathway parallel with SAC3 to facilitate early endosome progression and EGFR degradation. Whereas SAC3 knockdown did not affect the EGFR degradation rate in untreated HMECs (fig. S6A), it rescued the reduction in EGFR degradation induced by AKT inhibition (Fig. 5A and fig. S7A). When deconvolved, three of four siRNAs in the SAC3 siRNA pool reproduced this phenotype, confirming that acceleration of EGFR degradation was a specific effect of the SAC3 siRNAs (fig. S6B). These data indicate that SAC3 functions downstream of or in a pathway parallel with AKT to regulate EGFR degradation.

Fig. 5 AKT phosphorylates PIKfyve to facilitate EGFR degradation.

(A) Cells were transfected with nontargeting AllStars control siRNA or siRNA pool targeting SAC3. Cells were pretreated with AKTVIII for 45 min and with CHX for 30 min and stimulated with EGF for the indicated time points. Lower panel shows efficient knockdown of SAC3 as assessed by SAC3 mRNA abundance. n = 2 independent experiments. (B and C) PIKfyve (B) or its activator ArPIKfyve (C) was knocked down with siRNA, and cells were stimulated with EGF as in (A). (D) Cells were pretreated with DMSO or YM201636 for 45 min and with CHX for the last 30 min and stimulated with EGF. Representative blots from four or more independent experiments. See fig. S7 for the EGFR degradation decay graphs.

Because insulin stimulation inhibits SAC3, we tested whether AKT directly regulated the phosphatase activity of SAC3 (45). We immunoprecipitated the SAC3-ArPIKfyve-PIKfyve complex from human embryonic kidney (HEK) 293T cells pretreated with or without AKT inhibitor. Pretreatment of cells with AKT inhibitor did not change the phosphatase activity of SAC3 (fig. S6C). The Scansite and PhosphoSite programs do not indicate a consensus AKT phosphorylation motif in SAC3, collectively suggesting that AKT does not directly regulate SAC3 (46, 47). However, the SAC3 binding partner PIKfyve is phosphorylated by AKT in vitro and in cells (48, 49). We perturbed PIKfyve function to test whether PIKfyve regulates EGFR degradation in a similar manner to AKT. Knockdown of PIKfyve or its activator ArPIKfyve or pharmacological suppression of PIKfyve activity with the inhibitor YM201636 (50) reduced the rate of EGFR degradation (Fig. 5, B to D, and fig. S7, B to D). Together, these results suggest that similar to AKT activity, PIKfyve activity also facilitates EGFR degradation.

Incubation of AKT with PIKfyve in vitro increases PIKfyve activity toward PI(3)P to generate PI(3,5)P2, suggesting that AKT activates PIKfyve (48). To confirm that AKT regulates PIKfyve phosphorylation, we immunoblotted FLAG-PIKfyve immunoprecipitates with a phospho-AKT substrate (pAS) antibody, which recognizes the consensus AKT phosphorylation motifs (RXRXXS/T) only when the Ser or Thr residues are phosphorylated, and detected a band comigrating with the FLAG PIKfyve protein (Fig. 6A). AKTVIII treatment or mutation of the AKT consensus phosphorylation site Ser318 to an alanine substantially reduced the pAS signal (Fig. 6A), confirming that AKT phosphorylates Ser318 in PIKfyve.

Fig. 6 AKT phosphorylates and activates PIKfyve.

(A) HEK293T cells that were transfected with FLAG–PIKfyve wild type (WT) or S318A mutant PIKfyve were untreated or pretreated with AKTVIII before lysis. FLAG immunoprecipitates containing PIKfyve were immunoblotted with anti-pAS antibody to determine the phosphorylation status of PIKfyve and with anti-FLAG antibody to determine the total amount of overexpressed PIKfyve. Results of two independent experiments are shown. (B) HMECs were metabolically labeled with myo-[3H]inositol, deprived of growth factors, and treated with the indicated inhibitors for the last 30 min of the growth factor starvation. Cells were then stimulated with EGF for 30 min in the continued presence of inhibitors. Radioactively labeled phosphoinositides were extracted and quantified. Results of two independent experiments are shown. (C) HEK293T cells were transfected as in (A), and the WT PIKfyve complex was immunoprecipitated from cells treated with or without AKTVIII and stimulated with or without EGF (10 ng/ml). One-eleventh of the immunoprecipitate was immunoblotted to ensure equal isolation of the PIKfyve complex. The rest was used in a PIKfyve kinase activity assay with PI(3)P as a substrate. PI(3,5)P2 was extracted, and quantification of PI(3,5)P2 with Quantity One software is presented in the graph. Error bars represent SEM from three independent experiments. (D) HEK293T cells were transfected with the indicated PIKfyve complexes and stimulated with EGF (10 ng/ml). EGFR degradation was quantified, and EGFR degradation curves are presented in fig. S7E. Results are representative of eight independent experiments.

When activated, PIKfyve phosphorylates PI(3)P to generate PI(3,5)P2, which is then dephosphorylated by the myotubularin family of phosphatases to generate phosphatidylinositol 5-phosphate [PI(5)P] (51, 52). PIKfyve depletion in fibroblasts or treatment with PIKfyve inhibitor results in about 85% decrease in the concentration of PI(5)P, suggesting that most cellular PI(5)P is generated by the sequential action of PIKfyve and myotubularins (52). To test whether AKT inhibition and reduced PIKfyve phosphorylation correlated with PIKfyve activity in cells, we measured 3H-labeled phosphoinositide species isolated from HMECs (53). Because of its low abundance in cells, we could not detect PI(3,5)P2 (53). However, the concentration of PI(5)P was reduced upon PIKfyve inhibitor treatment, confirming that it reflects PIKfyve activity (Fig. 6B). AKT inhibitor treatment also reduced the concentration of PI(5)P to varying extents, suggesting that PIKfyve activity is regulated by AKT in cells.

To determine whether AKT directly regulates PIKfyve, we tested the effect of AKT inhibition or mutation of the AKT phosphorylation sites on the in vitro kinase activity of PIKfyve by immunoprecipitating the PIKfyve-ArPIKfyve-SAC3 complex from cells stimulated with or without EGF and pretreated with DMSO or AKTVIII. EGF stimulation increased PIKfyve kinase activity, which was reduced by pretreatment of cells with AKT inhibitor, suggesting that AKT is required for PIKfyve activation upon EGF stimulation (Fig. 6C). Together with the observation that AKT regulates PIKfyve phosphorylation and PIKfyve activity in cells, these data suggest that AKT directly activates the phosphoinositide kinase activity of PIKfyve.

To determine the role of AKT-mediated phosphorylation of PIKfyve in EGFR degradation, we transiently expressed wild-type PIKfyve or the nonphosphorylatable S318A mutant PIKfyve, together with SAC3 and ArPIKfyve, in HEK293T cells. Expression of the S318A mutant caused a small but statistically significant reduction in EGFR degradation and an increase in the phosphorylation of EGFR compared to wild-type PIKfyve (Fig. 6D and fig. S7D). In addition to EGFR degradation, PDGFRβ degradation was also reduced by PIKfyve and AKT inhibitors, suggesting that this pathway is likely to be functional in the degradation of other RTKs (fig. S8).

AKT reduces ERK signaling by facilitating EGFR degradation

Because receptor endocytosis can regulate receptor signaling to downstream pathways, we hypothesized that AKT-regulated degradation of EGFR constitutes a negative feedback loop that reduces EGFR signaling. In this model, EGF stimulation activates EGFR, which activates AKT. AKT directly phosphorylates and activates PIKfyve, which then promotes the progression of early endosomes containing EGFR into the degradation path. This activated endocytic trafficking promotes EGFR degradation and would therefore result in reduced EGFR signaling to the RAS-RAF-MEK-ERK-RSK pathway. We tested the functionality of this negative feedback loop in the MCF10A normal breast epithelial cell line, which is a well-established system to study the crosstalk between the AKT and ERK pathways (54). In these cells, AKT inhibition reduced the degradation of EGFR and led to more sustained ERK signaling as judged by EGFR abundance, phosphorylation of ERK, and phosphorylation of RSK after EGF stimulation (Fig. 7A and fig. S9A). To confirm that AKT feeds back to EGFR rather than a parallel pathway that regulates ERK signaling, we tested whether increased ERK activity by AKT inhibition could be suppressed by EGFR inhibition. We treated MCF10A cells with or without AKTVIII in combination with an EGFR inhibitor. Consistent with the hypothesis that AKT regulates EGFR signaling to modulate ERK activity, EGFR inhibition reduced phosphorylation of ERK even in the presence of AKTVIII (Fig. 7B and fig. S9B).

Fig. 7 AKT inhibits EGFR and ERK signaling by promoting EGFR degradation.

(A) MCF10A cells were deprived of EGF and serum overnight and stimulated with EGF in the presence of CHX for the indicated times. Results are representative of three independent experiments. EGFR degradation curves are presented in fig. S9A. (B) MCF10A cells were deprived of EGF and serum overnight; pretreated with AKTVIII, gefitinib, or their combination; and stimulated with EGF. Results are representative of three independent experiments. Quantification is presented in fig. S9B. (C) Model for how AKT promotes EGFR degradation. AKT phosphorylates and activates PIKfyve for proper EGFR progression through the early endosomes toward the lysosomes. When AKT is inhibited, PIKfyve is less efficient and EGFR progression from the early endosomes is reduced, which leads to reduced rates of EGFR degradation. Reduced EGFR degradation also correlates with increased phosphorylation of EGFR and ERK.


A role for AKT in EGFR trafficking and turnover

Although PI3Ks and their lipid products have been implicated in endocytic trafficking and trafficking of EGFR, the PI(3,4,5)P3-binding proteins that mediate these effects have not been identified and characterized. Here, we showed that AKT regulates EGFR trafficking and determined the molecular mechanism behind this regulation. Our results uncover a feedback loop by which AKT promotes EGFR degradation. Inhibition of AKT reduces recycling and lysosomal sorting of EGFR, which correlates with an increase in the localization of EGFR in the early endosomes and decreased PIKfyve activity in vitro and in cells. Reduced AKT activity enhances ERK and RSK activation in an EGFR-dependent manner. Thus, we propose a model (Fig. 7C) in which AKT negatively feeds back to EGFR to inhibit EGFR signaling to its downstream pathways: (i) AKT phosphorylates and activates PIKfyve, which generates PI(3,5)P2; (ii) PI(3,5)P2 facilitates the progression of EGFR-containing early endosomes to MVBs and late endosomes; and (iii) EGFR degradation increases and EGFR signaling to ERK decreases. Because AKT is upstream of the general endocytic regulator PIKfyve, this model suggests that AKT may regulate the degradation of other RTKs as well. Supporting this hypothesis, we found that AKT and PIKfyve also regulate PDGFRβ degradation.

Role for the PIKfyve-ArPIKfyve-SAC3 complex in RTK degradation

Previous studies on the role of PIKfyve in RTK degradation have been inconclusive. PIKfyve and its product PI(3,5)P2 regulate retrograde trafficking and vesicle budding and the degradation of Notch and the voltage-gated calcium channel Cav1.2 (18). Yet, although two structurally similar PIKfyve inhibitors reduce EGFR degradation and sorting to the lysosomes (50, 55), PIKfyve knockdown does not interfere with EGFR degradation in HeLa cells (56). The different results obtained by different groups could be due to cell type specificity, the nonspecific effects of the inhibitors, or inefficient RNA interference knockdown. Alternatively, manipulating the abundance of one component of the PIKfyve-SAC3-ArPIKfyve complex can change the stoichiometry of proteins within this complex, and therefore could produce a phenotype different from the one observed by directly inhibiting the kinase activity of an intact complex. For example, in contrast to the aberrant vacuolation observed after PIKfyve inhibition (50, 55), expressing a PIKfyve mutant (K2000E) with decreased phosphoinositide kinase activity does not cause a morphological defect unless it is expressed together with SAC3 and ArPIKfyve (57). Similarly, knockdown of SAC3 increases PI(3,5)P2 (20, 45); however, SAC3 knockout decreases the concentration of PI(3,5)P2, possibly because of disruption of complex integrity or function (52).

Given that the relative abundances of these proteins in cells determine the phenotypes observed, we used knockdown and exogenous expression together with pharmacological inhibition to determine whether the PIKfyve-ArPIKfyve-SAC3 complex regulates EGFR degradation. We showed that SAC3 knockdown, which increases the concentration of PI(3,5)P2, rescued the reduction in EGFR degradation produced by AKT inhibitor treatment. Conversely, knockdown of ArPIKfyve or PIKfyve, both of which are required for PI(3,5)P2 generation, reduced EGFR degradation. Inhibiting PIKfyve kinase activity and expressing the AKT phospho-site PIKfyve mutant reduced EGFR degradation. Thus, we conclude that PIKfyve regulates EGFR degradation by generating PI(3,5)P2. The observation that PIKfyve inhibitors reduce c-MET (55) and PDGFRβ degradation suggests that PI(3,5)P2 and enzymes involved in its metabolism are general regulators of RTK degradation. It is likely that PI(3,5)P2 controls RTK degradation either by recruitment of effector proteins that are involved in endosomal sorting or by directly activating the ion channels mucolipin 1 and two-pore channels TPC1 and TPC2, thereby increasing endosome and lysosome fusion events (58, 59).

Receptor endocytosis and signaling

EGFR signaling to the PI3K-AKT and ERK-RSK pathways is regulated by endocytosis at multiple levels. Blocking EGFR internalization by expressing EGFR mutants or dominant-negative dynamin reduces ERK and AKT activation (9, 60). In addition, the route of EGFR internalization determines signal duration. Clathrin-dependent EGFR endocytosis promotes receptor recycling and sustained AKT activation, whereas clathrin-independent EGFR endocytosis promotes transient AKT activation and EGFR degradation (40). Once internalized, EGFR remains associated with the signaling adapter Grb2 in the early endosomes (61). EGFR signaling from the endosomes is sufficient to induce DNA synthesis, activates ERK and AKT in HeLa cells, and regulates apoptosis during zebrafish development (6, 7, 62). EGFR signaling to the ERK pathway is also regulated at the late endosomes by the p14-MEK1-MP1 (MEK partner 1) complex (63, 64).

Our finding that AKT regulates EGFR trafficking and degradation suggests that AKT can regulate EGFR signaling by regulating EGFR trafficking. AKT inhibition increases phosphorylation of EGFR at Tyr1068 and phosphorylation of ERK in MCF10A cells. The AKT1 isoform inhibits ERK signaling in these cells through an unknown mechanism (54). We show that the AKT inhibitor–induced increase in ERK activation requires active EGFR, consistent with AKT facilitating EGFR degradation as a mechanism to reduce ERK signaling. AKT also inhibits ERK by phosphorylating and inhibiting B-Raf (65, 66), and our data add to this mechanism of pathway cross-inhibition.

Trafficking and degradation are important components of oncogenic RTK signaling. Mutant EGFRs that drive lung cancer proliferation degrade at a slower rate than do wild-type EGFRs, associate to a reduced extent with the ubiquitination machinery, and colocalize to a greater extent with markers of the recycling endosomes (67, 68). Constitutively active c-Met mutants require receptor endocytosis to promote migration and in vivo transformation (69). Our data suggest that therapeutic targeting of AKT in settings where RTKs are hyperactive should be performed with caution because of the ability of AKT to facilitate EGFR trafficking and degradation and inhibit ERK signaling. Thus, inhibiting AKT signaling in these cells may actually promote tumorigenesis by reducing the ability of cells to clear surface receptors and inducing continued RTK signaling from the endosomes. Determining the contribution of the AKT-mediated feedback to multiple RTKs in the context of various cancer cell types will enhance our understanding of RTK biology and the intricate signaling networks that govern cellular response to the extracellular environment.

Materials and Methods

Cell culture and transfections

HMECs were immortalized, maintained, and infected with pBabeNeo EGFR retrovirus as previously described (70). For knockdown experiments, cells were plated at a density of 2.5 × 105 cells per 6-cm dish 24 hours before transfection with AllStars nontargeting control siRNA (Qiagen) or siRNAs targeting AKT1, RAB11a, and ArPIKfyve, PIKfyve (Qiagen), or SAC3 (Dharmacon) with Lipofectamine RNAiMAX reagent (Invitrogen). MCF10A cells were cultured as previously described (54). HEK293T cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM) containing 10% fetal bovine serum. Calcium phosphate transfections in HEK293T cells were done as previously described (70).

Antibodies and reagents

EGFR, pEGFR (Tyr1068), pAKT (Ser473), pS6K (Thr389), pPRAS40, PRAS40, and ubiquitin antibodies were from Cell Signaling Technology. The antibodies were from the following companies: glyceraldehyde-3-phosphate dehydrogenase, Ambion; FLAG tag and ppERK1/2, Sigma; S tag, Novagen; EEA.1, Santa Cruz Biotechnology; and Rab11, BD Biosciences. S6K1 and ERK antibodies were made in-house. EGFR antibody used for immunoprecipitation, immunofluorescence, and flow cytometry was from Calbiochem, and ArPIKfyve and PIKfyve were from Abcam. Alexa Fluor 488–, 674–, and 568–labeled secondary antibodies and Alexa Fluor 488–labeled EGF were purchased from Molecular Probes, Invitrogen. Secondary antibodies for Western blotting were from Li-COR. Gefitinib (final concentration, 2 μM), PI103 (final concentration, 5 μM), and U0126 (final concentration, 20 μM) were from Selleck BioChemicals. AKTVIII (final concentration, 10 μM) and YM201636 (final concentration, 200 nM) were from Calbiochem. DMSO, N-ethylmaleimide (NEM), CHX, and chloroquine were from Sigma-Aldrich. EGF (100 ng/ml for stimulation experiments, unless otherwise noted) and insulin were from PeproTech. The AKT catalytic inhibitor A-443654 was a gift from N. Gray.

Cell lysis, immunoprecipitation, and Western blotting

Cell lysis was performed as previously described in the presence of protease and phosphatase inhibitors (70). For ubiquitination experiments, 10 μM NEM was added to the lysis buffer. EGFR was immunoprecipitated with EGFR antibody or mouse immunoglobulin G and with a 1:1 mixture of protein A (Sigma-Aldrich) and protein G–Sepharose beads (GE Healthcare) at 4°C for 90 min. FLAG PIKfyve was immunoprecipitated with agarose-conjugated FLAG antibodies (Sigma, Aldrich). Beads were spun at 5000 rpm for 1 min, and the flow-through was discarded. Beads were washed three times with lysis buffer and boiled with SDS loading buffer at 100°C for 6 min. Western blotting and quantification with the Odyssey Li-COR imaging system were carried out as previously described (71).

Immunofluorescence and image analysis

Cells were plated on HCl-treated coverslips as previously described (70) at a density of 50,000 cells per well in a six-well plate, 48 hours before assay. Cells were deprived of growth factors 24 hours before staining and treated with inhibitors as indicated. Where indicated, EGF entry into the cells was synchronized by stimulating cells with EGF at 4°C for 60 min on ice before moving them to 37°C. All other EGF stimulation experiments were carried out at 37°C. Cells were fixed with 3.7% formaldehyde and stained overnight with primary antibodies to EEA.1 (1:100) and EGFR (1:100). Images were taken using a Nikon Ti inverted microscope with a Yokogawa CSU-10 spinning-disk confocal microscope at ×60 magnification. Images were taken from the focal plane of the nucleus or as z stacks at a 0.3-μm step size. Percent colocalization values were calculated using the measure colocalization function of the MetaMorph software. Pixels with intensity values above a fixed threshold from each channel were used for calculations. Statistical analysis of the percent colocalization values was carried out with bootstrap permutation test (72). Boxplots were generated with MATLAB (R2011b, MathWorks). For measurement of integrated intensities of EGF and EGFR in EEA.1-positive early endosomes, we created early endosomes from EEA.1 pixels that are above the local background and form a continuous shape with a fixed minimum and maximum size. These endosomes were transferred over to the images that contain the EGF/EGFR signals, and only the EGF/EGFR signals within these regions were calculated as integrated intensities after a constant background subtraction. Integrated intensities from every other z plane were pooled for an entire field. Integrated EGF-EGFR intensities in DMSO- and AKTVIII-treated cells were compared by Kolmogorov-Smirnov analysis for nonnormal distributions. Cumulative distribution function plots were generated with MATLAB.

EGFR recycling assay

EGFR recycling was assessed as previously described in (40) with the following changes. Three plates of cells were pulsed with EGF or TGFα (100 ng/ml) for 15 min at 37°C to allow for internalization, and one plate was left unstimulated (to measure total surface EGFR). Internalization was stopped by moving the cells to 4°C. Excess EGF on the cell surface was washed off with acid treatment (150 mM NaCl, 50 mM glycine at pH 3.0) for 3 min at 4°C and washed with phosphate-buffered saline (PBS) three times. One plate was left at 4°C (Pulse). Two plates were moved back to 37°C to allow for recycling for 10 or 20 min (Chase). Cells were moved back to 4°C, fixed with 1% formaldehyde in PBS, washed once with PBS and twice with STE [10 mM tris-Cl (pH 7.5), 10 mM NaCl, 1 mM EDTA], and collected. Cell suspension was blocked with 3% bovine serum albumin (BSA) in PBS at 4°C for 30 min and labeled with EGFR antibody (1:100) at 4°C for 1 hour in 3% BSA and with Alexa Fluor 488 donkey anti-mouse antibody for 30 min at 4°C. Ten thousand cells were evaluated by flow cytometry for surface EGFR staining. Percent recycling at either the 10-min or the 20-min time point was calculated using the formula [(Chase − Pulse)/(Total Surface EGFR − Pulse)] × 100 from median values of each treatment group.

Phosphoinositide phosphatase assays

FLAG-PIKfyve, S-SAC3, and the hemagglutinin (HA)–ArPIKfyve complex were expressed in HEK293T cells. The complex was immunoprecipitated with anti-HA antibodies (30 min at 4°C) and protein A beads (1 hour at 4°C). Immunoprecipitates were washed twice with lysis buffer and twice with phosphatase buffer [50 mM tris-Cl (pH 7.4), 1 mM MgCl2, 1 mM dithiothreitol]. Beads were then resuspended with 35 μl of reaction mixture [31 μl of phosphatase buffer and 4 μl of di-C8 PI(3,5)P2 (Echelon Biosciences Inc.)]. Reactions were carried out at 37°C for 15 min for wild-type SAC3 and for 60 min for D488A mutant. Twenty-five microliters of the reaction was mixed with malachite green assay buffer (Echelon Biosciences). The amount of free phosphate liberated from PI(3,5)P2 via SAC3 activity in each reaction was read at 660 nm along with free phosphate standards and calculated as instructed by the manufacturer.

Phosphoinositide kinase assay

PI(3)P (Avanti Polar Lipids) vesicles were generated by sonication for 10 min in Hepes/EGTA buffer. FLAG-PIKfyve, S-SAC3, and the HA-ArPIKfyve complex expressed in HEK293T cells were immunoprecipitated as above, then washed twice with lysis buffer and twice with TNE buffer (20 mM tris-HCl (pH 7.4), 100 mM NaCl, 0.5 mM EGTA). After the first wash, 1/11 of the immunoprecipitated beads was used for Western blotting to ensure equal loading, and the rest was used in the phosphoinositide kinase assay. After the second wash, the beads were resuspended in TNE buffer, incubated with PI(3)P vesicles for 5 min on ice, and incubated with the ATP reaction mix (6.5 mM Hepes (pH 7.4), 50 μM ATP, 2.5 mM MnCl2, 10 mM MgCl2, 5 mM β-glycerophosphate) and [32P]ATP (PerkinElmer) for 15 min at room temperature. The reaction was stopped by adding 4 M HCl, and phosphoinositides were extracted with methanol/chloroform (1:1). Phosphoinositides were spotted on silica thin-layer chromatography (Millipore) plates and separated with 2 M acetic acid/n-propanol (65:35). Membrane was dried and exposed to a PhosphorImager screen (Bio-Rad), and the volumes of radioactively labeled PI(3,5)P2 spots were quantified with Quantity One software (Bio-Rad).

Measurement of phosphoinositides in cells

HMECs were plated at 1 million cells per 10-cm plate, and phosphoinositides were radioactively labeled in growing cells in inositol-free DMEM/F-12 (U.S. Biologicals) supplemented with myo-[3H]inositol (10 μCi/ml; American Radiolabeled Chemicals) for 48 hours. Cells were then starved for growth factors by incubation in inositol-free DMEM/F-12 without EGF and insulin for 1 hour. Inhibitors were added at the last 30 min of starvation. Cells were then stimulated with EGF (100 ng/ml) for 30 min. Phosphoinositides were then extracted, deacetylated, and separated by high-performance liquid chromatography as described previously (53).

Supplementary Materials

Fig. S1. Stimulation with high EGF concentrations induces lysosomal degradation of EGFR independently of EGFR abundance.

Fig. S2. Internalization of EGF/EGFR does not require AKT activity.

Fig. S3. RAB11a knockdown reduces the rate of EGFR degradation.

Fig. S4. EGFR recycling assay.

Fig. S5. Effects of AKTVIII on the colocalization of EGF with LAMP2.

Fig. S6. SAC3 promotes EGFR degradation but is not regulated by AKT.

Fig. S7. The PIKfyve-ArPIKfyve-SAC3 complex promotes EGFR degradation.

Fig. S8. PIKfyve and AKT promote PDGFRβ degradation in HMECs.

Fig. S9. AKT reduces EGFR signaling to ERK.

References and Notes

Acknowledgments: We thank N. Gray for the AKT catalytic inhibitor A-443654; A. Toker, L. Cantley, J. Zhao, and S. Gygi for critical discussion; P. P. Di Fiore and S. Sigismund for providing the detailed protocol for the EGFR recycling assay; A. Sasaki for his advice and providing a detailed protocol for the in vitro PI3K assay; and M. Vilela for his help with statistical analysis. We thank the Nikon Imaging Facility at Harvard Medical School for their help with light microscopy and the members of the Blenis laboratory for critical discussion. Funding: This work was funded by Susan G. Komen for the Cure (to M.C.M.) and National Cancer Institute grant R37CA46595 (to J.B.). J.B. is an Established Investigator of the LAM Foundation. Author contributions: E.E.E., M.C.M., and J.B. designed the experiments, analyzed the data, and wrote the manuscript. E.E.E. performed the experiments. A.M.M. and L.E.R. designed and performed the high-performance liquid chromatography experiments and analyzed the data. L.E.R. made corrections to the manuscript. Competing interests: The authors declare that they have no competing interests.
View Abstract

Stay Connected to Science Signaling

Navigate This Article