Research ArticleCell Biology

PIP3 Induces the Recycling of Receptor Tyrosine Kinases

See allHide authors and affiliations

Science Signaling  14 Jan 2014:
Vol. 7, Issue 308, pp. ra5
DOI: 10.1126/scisignal.2004532


Down-regulation of receptor tyrosine kinases such as the epidermal growth factor receptor (EGFR) is achieved by endocytosis of the receptor followed by degradation or recycling. We demonstrated that in the absence of ligand, increased phosphatidylinositol 3,4,5-trisphosphate (PIP3) concentrations induced clathrin- and dynamin-mediated endocytosis of EGFR but not that of transferrin or G protein (heterotrimeric guanine nucleotide–binding protein)–coupled receptors. Endocytosis of the receptor in response to binding of EGF resulted in a decrease in the abundance of the EGFR, but PIP3-induced internalization decreased receptor ubiquitination and phosphorylation and resulted in recycling of the receptor to the plasma membrane. An RNA interference (RNAi) screen directed against lipid-binding domain–containing proteins identified polarity complex proteins, including PARD3 (partitioning defective 3), as essential for PIP3-induced receptor tyrosine kinase recycling. Thus, PIP3 and polarity complex proteins regulate receptor tyrosine kinase trafficking, which may enhance cellular responsiveness to growth factors.


Growth factors such as epidermal growth factor (EGF) stimulate cells through specific transmembrane receptors acting as intracellular tyrosine kinases that dimerize and cross-phosphorylate each other (1). This leads to the recruitment of effector proteins that recognize the phosphotyrosines and to the assembly of signaling complexes at the plasma membrane (1). Downstream effectors such as phosphoinositide 3-kinase (PI3K) and Ras subsequently drive proliferation, differentiation, survival, and many other cellular events. Modulation of the signal is usually achieved by receptor internalization through endocytosis and finally lysosomal destruction or recycling to the plasma membrane (2). Phosphoinositides, phosphorylated derivatives of phosphatidylinositol, are crucial factors in signaling networks downstream of receptor tyrosine kinases (RTKs) (3). The abundant phosphoinositide phosphatidylinositol 4,5-bisphosphate [PI(4,5)P2] is essential for endocytic vesicle formation (4, 5); however, the role of other phosphoinositides is not well understood in the context of RTK internalization. The PI3K phosphorylation product phosphatidylinositol 3,4,5-trisphosphate (PIP3) is present in negligible amounts in resting cells, but transiently increases after RTK activation (6, 7). We therefore studied the effect of PIP3 on RTK endocytosis initially by monitoring EGF receptors (EGFRs).


Increased PIP3 is sufficient to induce EGFR internalization

First, we investigated whether increased phosphoinositide concentrations influenced EGFR internalization. To increase phosphoinositide concentrations, we used synthetic membrane-permeant phosphoinositide derivatives (8) (fig. S1). Treatment of HeLa cells with the ligand EGF induced EGFR internalization from the plasma membrane into endosomes (Fig. 1A). In the absence of EGF, addition of a membrane-permeant derivative of PIP3 (PIP3/AM, AM—acetoxymethyl ester) also induced internalization of an EGFR–yellow fluorescent protein (YFP) fusion (Fig. 1A), similar to that induced by PI(3,4,5,6)P4/AM, a compound that activates PI3K-dependent signaling (8). Other membrane-permeant phosphoinositide derivatives including PI(3)P/AM, PI(4)P/AM, and PI(2,3,5,6)P4/AM did not induce internalization, whereas a 10-fold higher dose of membrane-permeant PI(4,5)P2 was needed to induce internalization (table S1). The same internalization patterns were observed when endogenous EGFR was monitored by antibodies and when another cell line (U2OS) was used (figs. S2 and S3), demonstrating that the effect of PI(3,4,5,6)P4 on EGFR internalization was not due to the fusion of EGFR with YFP, EGFR overexpression, or a particular property of one cell type. EGF and PIP3/AM induced EGFR internalization with different kinetics (Fig. 1B). After applying EGF, EGFR internalization typically started within 5 min, whereas PIP3/AM stimulation resulted in EGFR internalization after 20 to 30 min (Fig. 1B). The time difference in the onset between PIP3/AM- and EGF-induced EGFR endocytosis probably reflected the time needed for the PIP3 derivatives to enter cells and for the bioactivatable groups (such as AM esters and butyrates) to be removed to liberate an active compound. To test this notion, we used a membrane-permeant, photoactivatable (“caged”) version of PIP3 (cgPIP3/AM) (9). cgPIP3/AM enters cells rapidly, but the presence of the “cage” attached to the crucial 3-O-phosphate maintains the lipid messenger inactive in cells for several hours until it is photolysed by a 405-nm laser pulse (9). In cgPIP3/AM-treated cells, EGFR-YFP was not endocytosed after 2 hours of treatment (Fig. 1C). However, when cells were illuminated with a short 405-nm laser pulse, EGFR endocytosis was triggered within 5 min, an onset similar to that triggered by EGF stimulation (Fig. 1, B and C). EGFR-YFP endocytosis in response to cgPIP3/AM treatment occurred only in illuminated cells (Fig. 1C). Because relatively high extracellular concentrations of membrane-permeant phosphoinositides were necessary, we wanted to determine the approximate intracellular concentration of PIP3/AM and also the concentration of the active, deprotected compound. Thus, we performed a quantitative mass spectrometry (MS) analysis of metabolites from lipids extracted from PC-12 cells incubated with PIP3/AM for various periods. Extracellular application of PIP3/AM resulted in an approximate intracellular PIP3/AM concentration of 7 μM after 25 min (a time point when EGFR internalization typically occurred). The active metabolite (having all AM esters and one butyrate removed) was detected at a 10-fold lower concentration than PIP3/AM, which would correspond to an estimated intracellular concentration of 0.5 to 1 μM (fig. S4 and table S2). The most abundant inactive metabolite was the compound that has all AM esters removed and both butyrates still attached, whereas fully deprotected and active dioctanoyl-PIP3 was just detectable.

Fig. 1 Increased PIP3 concentration induces EGFR internalization in the absence of ligand.

(A) HeLa cells expressing EGFR-YFP were stimulated as indicated and imaged by confocal microscopy. Arrows point to EGFR-containing intracellular vesicles formed after stimulation. Images are representative of at least 100 analyzed cells per condition in three independent experiments. Scale bar, 10 μm. (B) HeLa Kyoto cells stably transfected with EGFR-YFP were imaged by time-lapse microscopy after treatment. EGF or PIP3/AM was added to cells at “time 0.” cgPIP3/AM was added 120 min before time 0, at which point it was uncaged with a 405-nm laser pulse. The graph shows the changes in the number of EGFR-containing intracellular vesicles after each treatment (y axis) over time (x axis), which is representative of three different experiments. (C) An example of an uncaging experiment as described in (B). The uncaged area is marked with a red rectangle. Arrows mark EGFR-containing intracellular vesicles. Scale bar, 10 μm. Images are representative of five independent experiments. (D) HeLa Kyoto cells stably transfected with EGFR-YFP were transiently transfected with constitutively active PI3K (p110*-myc, upper panel) or kinase-dead PI3K (p110KD-myc, lower panel) as a negative control. White arrows mark EGFR internalization into vesicular and perinuclear compartments. Scale bar, 15 μm. (E) Quantification of the phenotype shown in (D) by measuring plasma membrane EGFR abundance. Plasma membrane EGFR/total EGFR ratio was calculated for transfected and nontransfected cells. EGFR surface abundance was also analyzed in EGF- or PIP3/AM-treated cells. Examples of plasma membrane staining are shown in fig. S5. Error bars are SD of the mean of at least 450 cells per condition analyzed in three independent experiments.

To strengthen the finding that an increase in PIP3 concentrations triggered EGFR internalization, we used two complementary methods that increase intracellular PIP3 concentrations. First, we overexpressed p110*-myc, a constitutively active form of PI3K (10) that in transfected cells induced EGFR-YFP internalization into vesicular and perinuclear compartments, but not in nontransfected cells or those overexpressing a kinase-dead mutant (p110KD-myc) (Fig. 1, D and E, and fig. S5). The next strategy was based on a previously published method that uses the chemical dimerizer rapamycin and its orthogonally binding domains FKBP and FRB (11). Cells were transfected with a plasma membrane–anchored FRB domain fused to cyan fluorescent protein (CFP) and a PI3K regulatory subunit fragment (iSH2) fused to FKBP and monomeric red fluorescent protein (mRFP) (12). Addition of rapamycin translocated the iSH2 construct to the plasma membrane, increased PIP3 concentrations, and induced EGFR-YFP internalization (fig. S6, A and B). High expression of iSH2 was sufficient to induce EGFR internalization in 40% of cells in the absence of rapamycin (fig. S6B). Control cells expressing the FKBP-mRFP construct without the iSH2 fragment showed no EGFR internalization after rapamycin treatment (fig. S6B).

PIP3-induced EGFR internalization is mediated by clathrin and dynamin

To characterize the nature of EGFR-transporting vesicles after phosphoinositide treatment, we examined their colocalization with early endosomal markers. EEA1 and Rab5 decorated both phosphoinositide-induced and EGF-induced EGFR vesicles (Fig. 2A and fig. S7). To quantify the colocalization between EEA1- and EGFR-YFP–containing vesicles, we calculated Pearson’s correlation coefficients, which increased from 0.11 ± 0.05 in nonstimulated cells to 0.63 ± 0.1 and 0.61 ± 0.15 in cells stimulated with EGF or PIP3/AM, respectively. Thus, both EGF and phosphoinositides promote translocation of EGFR-YFP to the same compartment. EGFR internalization can be a clathrin- and/or caveolin-mediated process, depending on the cell line and ligand concentration (2, 13). Both PIP3- and EGF-induced EGFR endocytosis were reduced by overexpression of a truncated version of the adapter protein AP180 (mycAP180-C), which blocks clathrin-mediated endocytosis (14) (Fig. 2B), as well as RNA interference (RNAi) directed against key mediators of endocytosis such as clathrin heavy chain (CLTC), dynamin2 (DNM2), and AP2 μ-subunit (Fig. 2, C and D). These results indicate that in HeLa Kyoto cells, under our experimental conditions, EGF- and phosphoinositide-induced internalization engage the same clathrin- and dynamin-based core endocytotic machinery.

Fig. 2 PIP3-induced EGFR internalization is a clathrin- and dynamin-mediated process.

(A) HeLa cells expressing EGFR-YFP were stimulated as indicated and stained with an antibody against the early endosome marker EEA1. In the insets, arrows point to colocalization of EEA1 and EGFR. Images are representative of at least 60 cells per condition and three independent experiments. Scale bar, 10 μm. (B) HeLa cells coexpressing EGFR-YFP and a C-terminal fragment of mycAP180 (a dominant-negative inhibitor of clathrin-mediated endocytosis) were stimulated as indicated and stained with anti-myc antibody to indicate mycAP180 (arrowheads, red). Arrows, intracellular vesicles. Images are representative of at least 60 cells per condition and three independent experiments. Scale bar, 10 μm. (C) HeLa cells expressing EGFR-YFP were incubated with the indicated siRNA, stimulated with PIP3/AM or EGF, and stained with an EGFR antibody that recognizes the extracellular epitope in nonpermeabilized cells. * indicates the positions of cells. Scale bar, 10 μm. (D) Quantification of EGFR surface abundance in (C). Graph shows plasma membrane EGFR/total EGFR ratio (y axis) after various siRNA treatments (x axis) and after stimulation with EGF (red bars) or PIP3/AM (blue bars). Data are normalized to the “no siRNA” control. Error bars are SD of the mean from 40 images containing about 3000 cells per condition and obtained in two independent experiments.

PIP3 does not induce internalization of other membrane receptors

We then tested whether PIP3-induced internalization was general or specific to the EGFR. Heterotrimeric guanine nucleotide–binding protein (G protein)–coupled receptors (GPCRs), such as the muscarinic receptor M2 and the purinergic receptor P2Y2, or glycosylphosphatidylinositol (GPI)–anchored proteins did not internalize after an increase in phosphoinositide concentration (Fig. 3A and fig. S8, A and B) or after treatment of cells with EGF (fig. S8C). When cells expressing M2-YFP were treated with the M2 receptor agonist carbachol, receptors were internalized into EEA1-marked early endosomes (fig. S8D), suggesting that the cells had competent GPCR internalization. The transferrin receptor, which is a clathrin-mediated cargo (15), was also unaffected by phosphoinositide or EGF treatment (Fig. 3, B and C). However, the RTK EphrinA4 receptor was internalized after treatment with PIP3 or the ligand EphrinA1 (fig. S9). These results suggest an extra level of regulation for some RTKs that is mediated by PI3K-derived lipids at the plasma membrane.

Fig. 3 GPCRs and transferrin receptors are not affected by increased PIP3 concentrations.

(A) HeLa cells cotransfected with EGFR-CFP and M2-YFP (muscarinic GPCR) were stimulated with PIP3/AM and imaged with time-lapse microscopy. Arrows point to EGFR-containing vesicles (inset); arrowheads point to M2 receptor at the plasma membrane. Scale bar, 15 μm. Images are representative of at least 20 cells per condition and three independent experiments. (B) HeLa Kyoto cells were stimulated as indicated or not stimulated (control cells), and stained with Alexa Fluor 488–labeled antibody against an extracellular EGFR epitope and with Alexa Fluor 568–labeled transferrin. Images show the surface abundance of EGFR (left) and transferrin receptor (TFRC) (right) after the indicated treatments. The apparent reduction of TFRC abundance after transferrin treatment is due to competition between transferrin–Alexa Fluor 568 and preincubated transferrin and does not indicate that transferrin decreases the plasma membrane abundance of TFRC. Scale bar, 20 μm. (C) Quantification of (B). Graph shows the fractions of the EGFR and TFRC at the plasma membrane (x axis) after various treatments (y axis). Fluorescence intensity data for each receptor were normalized to those in control cells. Error bars represent the SD of the mean fluorescence intensity calculated from 20 images per treatment (each image containing at least 100 cells), randomly acquired and obtained in two independent experiments.

PIP3 induces EGFR recycling in the absence of tyrosine phosphorylation and ubiquitination

The differences between EGF and phosphoinositide-induced endocytosis became apparent after endocytosis was observed for prolonged periods by live cell microscopy. Although addition of EGF resulted in EGFR degradation within 2 hours, likely through lysosomal destruction, phosphoinositide-induced endocytosis led to full recycling of EGFR-YFP (Fig. 4, A and B). Receptor posttranslational modifications, such as tyrosine phosphorylation and ubiquitination, play a crucial role in EGFR trafficking (2). Western blot analysis showed that the EGFR was tyrosine-phosphorylated in cells treated with EGF, but not with phosphoinositides (Fig. 4C). A similar pattern was observed for ubiquitination, a key modification that enables lysosomal targeting (Fig. 4C) (16). Furthermore, the EGFR kinase inhibitor AG1478 blocked EGFR internalization induced by EGF, but not by PIP3/AM (Fig. 4D and fig. S10A). The same pattern was observed when trafficking of endogenous EGFR was examined with antibodies (fig. S10B). Thus, tyrosine phosphorylation of EGFR was not required for PIP3/AM-induced EGFR internalization.

Fig. 4 Increased PIP3 concentrations induce EGFR recycling in the absence of tyrosine phosphorylation and ubiquitination of EGFR.

(A) HeLa cells expressing EGFR-YFP were stimulated as indicated and stained with an antibody against an extracellular EGFR epitope. * indicates the positions of cells. (B) Quantification of (A). Graph shows changes in the plasma membrane abundance of EGFR over time after stimulation with EGF or PI(3,4,5,6)P4/AM. Error bars represent the SD of the mean from 40 randomly taken images containing a minimum of 3000 cells obtained in two independent experiments. (C) HeLa cells expressing EGFR and HA-ubiquitin (Ub) were stimulated as indicated. Tyrosine phosphorylation and ubiquitination were assessed after immunoprecipitation (IP) of EGFR and immunoblotting. Immunoblots in lower panels show total EGFR, phosphorylated mitogen-activated protein kinase to demonstrate activation of intracellular signaling pathways and tubulin as a loading control. Immunoblots are representative of two independent experiments. (D) Changes in the plasma membrane abundance of EGFR, detected by an EGFR-specific antibody as in (B), after stimulation with EGF or PIP3/AM in the presence or absence of the EGFR tyrosine kinase inhibitor AG1478. Error bars represent SD of the mean from 40 randomly taken images containing a minimum of 3000 cells obtained in two independent experiments. (E) Graph shows changes in plasma membrane (PM) PIP3 concentrations (measured by PH-Akt-GFP intensity) after EGF stimulation of HeLa cells treated with different siRNAs as determined by TIRF microscopy. Cells were stimulated with EGF at time 0. Error bars represent the SEM from at least 20 cells from three independent experiments. (F) Graph shows the changes in EGFR-YFP intensity (loss of EGFR-YFP suggests degradation of EGFR) over time after EGF stimulation of HeLa cells treated with different siRNAs. Cells were stimulated with EGF and imaged every 10 min. Error bars represent the SD from at least 50 cells per condition from two independent experiments.

Little is known about how the balance between receptor degradation and recycling is regulated. On the basis of our results, we hypothesized that the PIP3-mediated signaling network could be part of this regulation. To test this idea, we examined the fate of the EGFR after EGF treatment under conditions in which PIP3 concentrations remain increased for extended periods and are typical for some advanced cancer states (17). EGF stimulation of HeLa cells typically led to low and transient increases in PIP3 concentrations (Fig. 4E). To achieve higher and longer-lasting PIP3 concentrations, we used an RNAi approach to target phosphatase and tensin homolog (PTEN) and SH2-containing inositol polyphosphate 5-phosphatase 2 (SHIP2), two lipid phosphatases involved in PIP3 degradation (18). Indeed, knockdown of PTEN and SHIP2 (fig. S11, A and B) resulted in higher and prolonged PIP3 concentrations after EGF stimulation (Fig. 4E). Under these conditions, EGFR was not degraded after EGF stimulation (Fig. 4F and fig. S12A) but rather was recycled to the plasma membrane. In cells treated with nontargeting small interfering RNA (siRNA), the receptor was degraded after EGF stimulation as expected (Fig. 4F). The same pattern was observed when endogenous EGFR was analyzed (fig. S12B). Together, these results suggest that a PIP3-dependent feedback loop could promote EGFR recycling by a mechanism that does not require either receptor tyrosine phosphorylation or ubiquitination to specifically internalize and sort RTKs. This effect might also partially explain the persistent growth factor signaling that occurs in late-stage cancer cells by providing a constant reservoir of activated EGFR at the plasma membrane through recycling.

An siRNA screen reveals polarity complex proteins as key mediators of PIP3-induced endocytosis

This type of RTK internalization could partially depend on molecules downstream of PI3K, and we hypothesized that these molecules would need to reside on the plasma membrane to recognize PIP3 or other lipids. We therefore performed an siRNA screen against 680 proteins of the human genome containing a lipid-binding domain (tables S3 and S4) using reverse transfection on siRNA chips and automated microscopy as previously described (19). Knockdown of 141 proteins inhibited PIP3-mediated EGFR endocytosis (table S5). We next performed a general functional analysis of the candidate proteins. We examined overrepresentation of Gene Ontology (GO) terms (20) with the R package “GOstats” (21). One hundred forty-one candidate proteins from the siRNA screen (genes the knockdown of which resulted in inhibition of internalization) were compared to all 680 genes tested in the screen serving as “background.” Two significant results were returned when checking this gene list for overrepresentation of molecular function terms (table S6)—“neurexin family protein binding” and “protein serine/threonine kinase activity” (table S7). To find the targets (substrates) for these serine/threonine kinases, we used, which produced 122 targets for 9 of the 16 kinases. The top candidates were related to “actin binding or regulation” and “lipid binding” (table S8). In a separate analysis using the protein-protein interaction database STRING and the candidate proteins, we constructed a hypothetical interaction network that potentially mediates PIP3 effects (fig. S13). GO database and manual literature inspection revealed that the hypothetical network consisted of proteins involved in membrane trafficking, PI3K or phospholipid signaling, and actin cytoskeleton regulation (fig. S13).

Together, both functional and network analysis revealed regulation of the actin cytoskeleton and lipid binding or signaling as downstream effects of PIP3, leading us to further investigate polarity complex proteins as potential regulators of PIP3-induced internalization. Several members of the polarity complex [PARD3 (partitioning defective 3), PARD6, and PKCζ (protein kinase Cζ)] were identified in the screen as essential for phosphoinositide-induced endocytosis, with PARD3 being the strongest effector in the screen (table S5). They have been previously linked to actin cytoskeleton regulation (22), and an siRNA screen in Caenorhabditis elegans has linked them to endocytosis (23). PARD3 can bind phosphoinositides (24). Polarity complex proteins are important factors in embryonic and cell polarity establishment and maintenance (25), but the molecular mechanism of their action is largely unknown. To validate a potential role of these proteins in PIP3-induced endocytosis, we knocked down PARD3 and PARD6 using different siRNAs, and found that their knockdown inhibited PIP3-mediated EGFR internalization to varying extents (Fig. 5A). PARD3 siRNAs delayed and partly inhibited EGFR internalization in response to EGF (fig. S14) but not that of transferrin (figs. S15 and S16). Because PIP3 signaling is mediated through recruitment of specific effector proteins to the plasma membrane, we wanted to examine how increased PIP3 concentrations affected subcellular PARD3 localization. Green fluorescent protein (GFP)–PARD3 was localized to the cytoplasm and plasma membrane in unstimulated cells. Total internal reflection fluorescence (TIRF) microscopy revealed that GFP-PARD3 was recruited to the plasma membrane after PI(3,4,5,6)P4/AM addition (Fig. 5B and fig. S17), consistent with a potential role for PARD3 in PIP3-induced internalization. Treatment with PI(3)P/AM did not result in PARD3 accumulation at the plasma membrane, which correlates with the inability of this phosphoinositide to induce EGFR internalization (Fig. 5B). We wondered if PARD3 accumulation was sufficient for EGFR internalization. Overexpression of PARD3-mRFP, but not of PLCδ-PH-mRFP, a domain that also binds phosphoinositides and localizes at the plasma membrane, induced EGFR internalization in 60% of transfected cells in the absence of ligand or PIP3 (Fig. 5, C and D). The M2 GPCR was not internalized by overexpression of PARD3-mRFP (fig. S18). We suggest that PIP3-induced plasma membrane accumulation of PARD3 may be involved in EGFR internalization.

Fig. 5 Polarity complex proteins are key mediators of PIP3/AM-induced EGFR internalization.

(A) Percent inhibition of PIP3-induced EGFR internalization (y axis) in HeLa cells stably transfected with EGFR-YFP and incubated with the indicated siRNAs (x axis) and after stimulation with PI(3,4,5,6)P4/AM normalized to the “no siRNA” control. AP2μ, CLTC (clathrin heavy chain), and DNM2 (dynamin2) siRNAs were positive controls. Error bars are SD of 40 randomly taken images from two independent experiments. (B) Changes in GFP-PARD3 plasma membrane (PM) intensity over time after PI(3,4,5,6)P4/AM or PI(3)P/AM treatment in U2OS cells as determined by TIRF microscopy. Cells were stimulated with PI(3,4,5,6)P4/AM or PI(3)P/AM at time 0. Nine of 13 cells exhibited this translocation. Error bars represent the SEM from at least nine cells and three independent experiments, with the values for all cells being normalized individually to t = 0. Representative images are shown in fig. S17. (C) U2OS cells transfected with EGFR-YFP and PLCδ-PH-mRFP (negative control) or PARD3-mRFP. Inset images (on the side of each panel) are a magnified view of the region marked with a square (white dashed line). Arrows point to EGFR localized to intracellular vesicles. Scale bars, 15 μm. (D) Quantification of (C). Error bars represent the SD of at least 70 cells per condition from two independent experiments.


We have shown that products of type I PI3K, such as PIP3, constitute a sufficient signal to induce rapid and specific clathrin-mediated internalization of EGFR in the absence of an endogenous ligand. By specifically stimulating one part of the signaling network downstream of the receptor level, it was possible to uncouple endosomal recycling and lysosomal trafficking. Tyrosine phosphorylation and ubiquitination are not required for specific cargo recognition and its internalization and recycling. The combined results suggest that in plasma membrane regions of high PIP3 concentrations, predominantly recycling endosomes will be formed. This represents a level of regulation that is potentially responsible for the RTK receptor recycling that occurs after stimulation. In support of this hypothesis, Akt, a major effector of PIP3 signaling, promotes EGFR recycling in human mammary epithelial cells, although it ultimately results in increased EGFR degradation through several rounds of EGFR internalization (26), unlike in our model system. It would be interesting to test if the differences in the duration or amplitude of Akt activity could be making a distinction between continuous recycling and degradation of RTKs. Small differences in the amplitude and duration of Akt activity affect the ability of Akt to promote the cellular unfolded protein response (UPR) (27). Additional support for our hypothesis comes from the demonstration that PI3K activity dictates the route of fibroblast growth factor receptor 2b (FGFR2b) trafficking, in which increased PI3K activity diverts receptors from a degradative to a recycling pathway (28).

An siRNA screen targeting proteins with membrane-binding domain revealed that the mechanism of PIP3-induced EGFR endocytosis involves small Rho guanosine triphosphatases and actin cytoskeleton–associated and polarity complex proteins. PIP3 seems to recruit PARD3 to the plasma membrane, which specifically promotes EGFR internalization and recycling. Therefore, a physiological role of the PIP3-mediated internalization through polarity complex proteins might be to control RTK trafficking in areas of persistently increased PIP3 concentrations, by channeling them into recycling rather than in the default degradation pathway. This appears to be especially useful for continuously dividing cancer cells or under circumstances where it physiologically does not make sense to degrade relevant receptors, such as during axon growth and pathfinding of neuronal cells (29) or the leading edge establishment and maintenance in migrating cells (30). For these biological processes, persistent signaling from RTKs is required, and cells cannot afford to resynthesize and transport new receptors from the endoplasmic reticulum to the sites of action. This sort of “polarized recycling” would provide a constant reservoir of activated RTKs at discrete, “PIP3-rich,” sites of the plasma membrane and could be a general strategy to initiate and maintain cellular asymmetry. In further support to our hypothesis, PARD3 is involved in the spatiotemporal control of VEGFR (vascular endothelial growth factor receptor) internalization during angiogenesis (31). In the future, it will be particularly interesting to study the molecular basis of recycling, cargo recognition, and the effect of the polarity complex proteins on the endocytic machinery in more mechanistic detail. Regarding pharmacological relevance, small molecules such as the membrane-permeant phosphoinositide derivatives used here might become important starting points for the development of drugs aimed at specifically triggering the internalization of receptors from the plasma membrane.


Cell lines and reagents

HeLa, HeLa Kyoto, HeLa Kyoto stably transfected with EGFR–enhanced YFP (EYFP), and U2OS cells were grown in Dulbecco’s modified Eagle’s medium (DMEM; Life Technologies) supplemented with 10% fetal calf serum (PAA Laboratories GmbH), 1% l-glutamine (Life Technologies), and 1% penicillin and streptomycin (Life Technologies) at 37°C and 5% CO2. HeLa Kyoto cells stably transfected with EGFR-EYFP were derived from a single colony using geneticin (G418) (Life Technologies) as a selection marker. PC-12 cells were grown on collagen-coated plates in DMEM (Life Technologies) supplemented with 10% heat-inactivated horse serum (Life Technologies), 5% fetal calf serum (PAA Laboratories GmbH), 1% l-glutamine (Life Technologies), and 1% penicillin and streptomycin (Life Technologies). The following chemicals and proteins were purchased: AG1478 (Sigma), apo-transferrin (Sigma), transferrin–Alexa 568 (Life Technologies), EphrinA1 (R&D Systems), F(ab′)2 fragment goat anti-human immunoglobulin G (Jackson ImmunoResearch), rapamycin (LC Laboratories), and carbachol (Calbiochem). The following antibodies were used: EEA1 (BD Biosciences), c-Myc (9E10, sc-40, Santa Cruz Biotechnology), c-Myc (A-14, sc-789, Santa Cruz Biotechnology), extracellular EGFR epitope (R-1, sc-101, Santa Cruz Biotechnology), α-tubulin (DM1A, NeoMarkers), hemagglutinin (HA) (12CA5, Roche), phosphotyrosine (PY99, sc-7020, Santa Cruz Biotechnology), EGFR (R-1, sc-101, Santa Cruz Biotechnology), PTEN (D4.3, Cell Signaling), SHIP2 (C76A7, Cell Signaling), and anti-mouse Alexa 488 or Alexa 568 secondary antibodies (Life Technologies). For immunoprecipitation of EGFR, antibody 1005 (sc-03, Santa Cruz Biotechnology) was used.

Membrane-permeant phosphoinositide treatment

Membrane-permeant phosphoinositides (8) were dissolved in dry dimethyl sulfoxide (DMSO) to a stock solution of 50 mM. The portion to be used was mixed with 10% pluronic F127 in DMSO (Life Technologies) in 1:1 (v/v) ratio to prevent precipitation in aqueous environment. Cell medium was added, and the mixture was rapidly added to cells.

Complementary DNA transfections

EphA4 was a gift from R. Klein (Max Planck Institute of Neurobiology, Martinsried, Germany), GPI-mRFP was a gift from K. Simons (Max Planck Institute for Molecular Cell Biology and Genetics, Dresden, Germany), EGFR-EYFP and EGFR–enhanced CFP (ECFP) were gifts from P. Bastiaens (Max Planck Institute for Molecular Physiology, Dortmund, Germany), HA-ubiquitin was from I. Dikic (IBCII, Frankfurt, Germany), p110*-myc and p110KD-myc were from A. Klippel (Wyeth Pharmaceuticals, New York, NY) (10), mycAP180-C was from M. G. J. Ford (MRC, Cambridge, UK) (14), GFP-PARD3 was from B. D. Grant (Rutgers University, New Brunswick, NJ) (23), and P2Y2-CFP was from C. Hoffmann (University of Wuerzburg, Wuerzburg, Germany). FRB-ECFP was designed as follows: It consisted of the N-terminal plasma membrane–targeting peptide of rat LCK kinase (amino acids 1 to 10), FRB T2098L domain (amino acids 2021 to 2113 of human mammalian target of rapamycin), and ECFP. The amino acid linker between the signaling peptide and FRB was RSANSGAGAGAGAILSR. The linker between FRB and ECFP was TSYPYDVPDYAPVAT. mRFP-FKBP-iSH2 was designed as follows: It consisted of mRFP, human FKBP1A, and amino acids 420 to 615 of the human PI3K (iSH2), regulatory subunit 1 (α) (PI3KR1). The linker between mRFP and FKBP was SGLRSRAAAGAGGAARAA. The amino acid sequence between FKBP and iSH2 was ARGAAAGAGGAGRSGGKL. For complementary DNA (cDNA) transfection, cells were seeded on eight-well Lab-Tek glass-bottom dishes in antibiotic-free medium for 24 hours. Next, they were transfected with cDNAs using FuGENE 6 (Roche) or, in the case of p110*myc and p110KDmyc, with Lipofectamine 2000 (Life Technologies) according to the manufacturer’s instructions. After 24 hours, cells were treated or not with growth factors and membrane-permeant phosphoinositides, then fixed or examined live by microscopy.

RNAi experiments

For each siRNA, the gene name, RefSeq ID, and catalog number were provided. The following siRNAs were from Qiagen: DNM2 (NM_001005360, SI02654687), CLTC (NM_004859, SI00299880), AP2μ (AP2M1, NM_001025205, SI02777355), AP2β1 (NM_001282, SI02780078), GRB2 (NM_002086, SI00300328), PARD3 (NM_019619, SI00677733), PARD3B (NM_057177, SI00294959), PARD6A (NM_001037281, SI02664340), PARD6B (NM_032521, SI00677761), and PARD3G (NM_032510, SI00143388). The following siRNAs were from Ambion: PTEN (NM_000314, s325) and INPPL1 (SHIP2, NM_001567, s7463). For RNAi experiments, the HeLa Kyoto cells stably transfected with EGFR-EYFP were seeded on eight-well Lab-Tek or MatTek glass-bottom dishes in antibiotic-free medium. After 24 hours, they were transfected with siRNAs using Oligofectamine (Life Technologies) according to the manufacturer’s instructions. The final siRNA concentration was 90 nM (for PTEN and SHIP2 siRNA experiments, 60 nM). After 72 hours, they were stimulated with EGF (100 ng/ml) for 15 min and then with PI(3,4,5,6)P4/AM or PIP3/AM (50 μM) for 45 min. Next, they were fixed, and the surface abundance of EGFR was assessed (see “EGFR endocytosis assay”).

EGFR endocytosis assay

To determine the rate of EGFR endocytosis, we seeded the HeLa Kyoto cells (wild type or stably transfected with EGFR-EYFP) on glass-bottom dishes. After 24 hours, they were, if needed, transfected with siRNAs (for 72 hours) or cDNAs (for 24 hours). The cells were next treated with growth factors or membrane-permeant phosphoinositides, fixed with 3.7% paraformaldehyde in phosphate-buffered saline, and stained with mouse monoclonal primary antibody specific for extracellular EGFR epitope (R-1, sc-101, Santa Cruz Biotechnology). This was followed by anti-mouse Cy5-labeled secondary antibody staining and 4′,6-diamidino-2-phenylindole (DAPI) staining to highlight the nucleus. Because the cells were not permeabilized, the primary antibody bound only to receptors present at the plasma membrane. The images of DAPI (nucleus), YFP (total EGFR), and Cy5 (plasma membrane EGFR) were acquired automatically on an Olympus ScanR microscope with a UPlanAPO 20×/0.7 numerical aperture (NA) air objective. Forty images together containing about 3000 cells were typically acquired for each condition. The EGFR endocytosis efficiency was calculated from individual cells by determining the ratio between the plasma membrane EGFR (Cy5 fluorescence intensity) and total EGFR (EYFP fluorescence). Such a ratio was not calculated when endogenous EGFR was examined.


All the experiments on live cells were performed in imaging medium [minimum essential medium (MEM) + 30 mM Hepes] at 37°C. TIRF microscopy was performed on a Leica AF7000 with an HCX PL APO 100×/1.46 NA oil objective. A 488-nm laser line was used for TIRF excitation. One image was acquired in the YFP channel every 30 s for 30 min. Wide-field and time-lapse microscopy were performed on a Leica AF7000 equipped with an HCX PL APO 63×/1.3 NA glycerol objective. For experiments in which automated acquisition was needed, an Olympus ScanR microscope with a UPlanAPO 20×/0.7 NA air objective was used. Confocal microscopy experiments were performed on a Leica SP5 with an HCX PL APO CS 63×/1.4 NA oil objective. Uncaging experiments were done on an Olympus FV1000 microscope with a UPLSAPO 60×/1.35 NA oil objective. Uncaging was performed with a 405-nm laser pulse (two illuminations, 1 s each). After uncaging, one image was acquired in the YFP channel every 15 s for 15 min.

siRNA screen

All the siRNAs used were from Qiagen. The library consisted of siRNAs against genes that encode proteins with a membrane-binding domain. There are 777 proteins in the human genome with a known membrane-binding domain (table S3). Ninety-seven of them had no RefSeq ID and were not considered for analysis. We created an siRNA library against 680 membrane-binding proteins (table S4) in which each gene was targeted with two independent siRNAs, producing 1360 siRNAs for testing. The sequences can be found in table S4.

To transfect a large number of siRNAs, we used an established high-throughput experimental setup (19). In short, siRNA was deposited together with the transfection reagent onto a glass-bottom cell culture dish. After drying, cells were seeded on top of the siRNA spots for transfection (“reverse transfection”). Three hundred eighty-four different siRNAs can be spotted on one dish (called an “siRNA chip”), thereby allowing high-throughput transfection. We produced four such chips. Apart from the siRNAs from the library, each chip contained 20 negative controls (16 nontargeting siRNAs and 4 siRNAs that target INCENP, a protein with no known function in endocytosis) and 13 positive controls (11 siRNAs targeting AP2μ and 2 targeting CLTC) dispersed throughout the chip (fig. S19A).

To determine the effect of siRNAs on phosphoinositide-induced EGFR internalization, we used HeLa Kyoto cells stably transfected with EGFR-EYFP. Cells were seeded on the “siRNA chips.” After 72 hours, they were treated with PI(3,4,5,6)P4/AM (50 μM) for 45 min, fixed but not permeabilized, and stained as described in “EGFR endocytosis assay.” Images were acquired on an automated Olympus ScanR screening microscope with a UPlanAPO 20×/0.7 NA air objective. One image per spot (per siRNA) was acquired in YFP (total EGFR), Cy5 (plasma membrane EGFR), and DAPI (nucleus) channels. The EGFR endocytosis efficiency was calculated from individual cells by determining the ratio between the plasma membrane EGFR (Cy5 fluorescence intensity) and total EGFR (EYFP fluorescence). Each experiment was repeated five times.

EGFR endocytosis efficiency values for each siRNA spot were normalized to the median value of their corresponding “siRNA chip.” The mean and SD were than calculated for each siRNA from five independent experiments. We decided that the candidates for further analysis were those siRNAs whose “EGFR endocytosis efficiency (E)” value was different from the E value of the negative control (nontargeting siRNA) by more than 2 SDs. Therefore, the mean E values from siRNA spots were converted into a “deviation score” by the following formula: (EsiRNAEcontrol)/2*SDcontrol. Values >1 would mean that the siRNA inhibited EGFR endocytosis, and values <−1 would signify that the siRNA accelerated EGFR endocytosis (fig. S19, B and C). Of 1360 siRNAs tested, 198 affected phosphoinositide-induced EGFR endocytosis. One hundred ninety-eight positive siRNAs corresponded to 156 genes (114 genes were positive with one siRNA and 42 with two siRNAs) (table S5). Forty-three of the 52 spotted positive controls were also identified as effectors, and none of the negative controls (out of 80) were identified as effectors, confirming the efficiency and specificity of the screen.

Statistical analysis

All data were analyzed for significance on a rate scale using a Mann-Whitney U test. If multiple comparisons were made, the resulting P values were adjusted for multiple testing using a Bonferroni correction. To analyze the data in Fig. 5B for statistical significance, a standard linear regression model was fit to the data. To formally examine the model, the slope difference divided by its SE (sum of the square root of the individual SEs) was tested using a two-sided z test. P values <0.05 were considered significant.


Materials and Methods

Fig. S1. Structures of the membrane-permeant phosphoinositides and their parent lipids used in the study.

Fig. S2. Membrane-permeant PIP3 derivatives induce internalization of endogenous EGFR in HeLa cells.

Fig. S3. Membrane-permeant PIP3 derivatives induce internalization of EGFR-YFP in U2OS cells.

Fig. S4. Elution profiles and time courses of the accumulation of Bt2DiC8PIP3(AM)7 and its metabolites in cells.

Fig. S5. Overexpression of constitutively active PI3K induces EGFR internalization.

Fig. S6. PI3K activation at the plasma membrane induces EGFR internalization.

Fig. S7. PIP3/AM-induced endosomes are Rab5-positive.

Fig. S8. GPCRs and GPI-anchored proteins are not internalized after EGF or PIP3/AM stimulation.

Fig. S9. Membrane-permeant PIP3 derivatives induce internalization of EphA4.

Fig. S10. The EGFR tyrosine kinase inhibitor AG1478 inhibits EGF-induced, but not PIP3/AM-induced, EGFR internalization.

Fig. S11. Effectiveness of the PTEN and SHIP2 siRNAs.

Fig. S12. PTEN and SHIP2 siRNAs block ligand-induced EGFR degradation.

Fig. S13. A hypothetical interaction network that mediates PIP3-induced EGFR endocytosis.

Fig. S14. PARD3A knockdown inhibits EGF-induced EGFR internalization.

Fig. S15. PARD3A knockdown has no effect on transferrin internalization.

Fig. S16. PARD3A knockdown has no effect on transferrin recycling.

Fig. S17. PI(3,4,5,6)P4/AM treatment induces GFP-PARD3A translocation to the plasma membrane.

Fig. S18. PARD3 overexpression does not trigger GPCR internalization.

Fig. S19. High-throughput siRNA screen.

Table S1. Effects of different membrane-permeant phosphoinositides on EGFR internalization.

Table S2. Calculated and exact masses for DiC8Bt2PIP3/AM and three metabolites identified in cell extracts.

Table S3. List of membrane-binding domains chosen for the siRNA screen.

Table S4. List of the genes tested in the screen.

Table S5. List of the effectors of PIP3/AM-mediated EGFR endocytosis.

Table S6. Molecular function terms that are overrepresented in the candidate protein list compared to the list of tested genes.

Table S7. Candidate proteins associated with the molecular function term “serine/threonine kinase activity.”

Table S8. Molecular function terms that are overrepresented among substrates of serine/threonine kinases from table S6.

References (3236)


Acknowledgments: We are very grateful to the staff of the Advanced Light Microscopy Facility at the European Molecular Biology Laboratory (EMBL) for extensive support in reverse transfection and spotting techniques, microscope handling, and data analysis. We acknowledge B. Klaus from the EMBL Centre for Statistical Data Analysis for help with data analysis. We thank K. Simons, R. Klein, P. Bastiaens, I. Dikic, A. Klippel, M. G. J. Ford, B. D. Grant, and C. Hoffmann for cDNA constructs. We are grateful to D. Whittington of the University of Washington School of Pharmacy Mass Spectrometry Center for helpful discussions and suggestions. Funding: This work was supported by the Helmholtz Association (SBCancer), the European Science Foundation (ESF), Deutsche Forschungsgemeinschaft (DFG) (Schu 943/8-1), the EMBL Interdisciplinary Postdocs (EIPOD) program, and the Volkswagen Foundation (I/78989 and I/81797). Author contributions: V.L., R.P., J.S.-R., and C.S. designed the experiments. V.L. carried out most of the experiments. A.T.-K. and A.N. performed MS analysis. S.Z., D.S., R.M., and M.M. synthesized membrane-permeant phosphoinositide derivatives. M.P. prepared all the constructs needed for experiments using rapamycin-induced protein dimerization and PARD3 constructs. A.M. and V.L. analyzed the siRNA screening data. V.L. and C.S. wrote the manuscript. Competing interests: The authors declare that they have no competing interests. Data and materials availability: A material transfer agreement (MTA) is required by EMBL for the membrane-permeant phosphoinositide derivatives. An MTA is required by EMBL for the fluorescent fusion proteins.
View Abstract

Stay Connected to Science Signaling

Navigate This Article