Receptor endocytosis plays an important role in regulating the responsiveness of cells to specific ligands. Phosphatidylinositol 4,5-bisphosphate [PtdIns(4,5)P2] has been shown to be crucial for endocytosis of some cell surface receptors, such as EGF and transferrin receptors, but its role in G-protein-coupled receptor internalization has not been investigated. By using luciferase-labeled type 1 angiotensin II (AT1R), type 2C serotonin (5HT2CR) or β2 adrenergic (β2AR) receptors and fluorescently tagged proteins (β-arrestin-2, plasma-membrane-targeted Venus, Rab5) we were able to follow the sequence of molecular interactions along the endocytic route of the receptors in HEK293 cells using the highly sensitive method of bioluminescence resonance energy transfer and confocal microscopy. To study the role of plasma membrane PtdIns(4,5)P2 in receptor endocytosis, we used our previously developed rapamycin-inducible heterodimerization system, in which the recruitment of a 5-phosphatase domain to the plasma membrane degrades PtdIns(4,5)P2. Here we show that ligand-induced interaction of AT1, 5HT2C and β2A receptors with β-arrestin-2 was unaffected by PtdIns(4,5)P2 depletion. However, trafficking of the receptors to Rab5-positive early endosomes was completely abolished in the absence of PtdIns(4,5)P2. Remarkably, removal of the receptors from the plasma membrane was reduced but not eliminated after PtdIns(4,5)P2 depletion. Under these conditions, stimulated AT1 receptors clustered along the plasma membrane, but did not enter the cells. Our data suggest that in the absence of PtdIns(4,5)P2, these receptors move into clathrin-coated membrane structures, but these are not cleaved efficiently and hence cannot reach the early endosomal compartment.
Cell surface receptors including nutrient receptors, G-protein-coupled receptors (GPCRs) and receptor tyrosine kinases (RTKs) play a crucial role in a variety of physiological and pathological cell functions. The number of surface receptors is one of the major determinants of hormonal responsiveness of cells and tissues, which often changes and contributes to the pathogenesis of common diseases such as diabetes, hypertension, cancer and many others. Moreover, plasma membrane receptors are among the most important targets of drugs used in therapy, but their effectiveness is often limited by downregulation of the receptors. The steady state level of receptors in the plasma membrane is determined by the balance between delivery of receptors to the cell surface and their removal by endocytosis (Tsao et al., 2001). In many GPCRs, endocytosis of phosphorylated and desensitized receptors initiates a process leading to their resensitization and recycling to the plasma membrane. In addition, evidence is accumulating to show that internalization might also lead to the activation of other types of signaling pathways mediated by the internalized receptors (Calebiro et al., 2010) or by associated proteins, such as β-arrestin-2 (Wei et al., 2003; Lefkowitz and Shenoy, 2005).
The most investigated form of endocytosis is clathrin-mediated endocytosis (CME), which is used by nutrient receptors (e.g. transferrin receptor), GPCRs, RTKs and many other cell surface proteins to enter the cells (Doherty and McMahon, 2009). Although the formation of clathrin-coated pits (CCPs) and the recruitment of dynamin are common features of the process (Schmid and McMahon, 2007), the various types of receptor (cargos) use different sets of adaptor proteins (Traub, 2009). For example, β-arrestin-2 is a key member in the endocytic machinery of AT1 receptors (Gáborik et al., 2001), whereas endocytosis of EGF receptors is mainly ubiquitin dependent, although involvement of other, non-arrestin-dependent routes cannot be excluded, as recently reviewed by the Sorkin laboratory (Goh et al., 2010). Genome-wide survey of transferrin and EGF receptor endocytosis revealed clear differences in the molecular details of their regulation (Collinet et al., 2010), which are due to the different adaptors they use. It was also shown that transferrin and β2 adrenergic receptors are endocytosed by biochemically and functionally distinct subpopulations of CCPs (Cao et al., 1998), and the same was found in case of transferrin and EGF receptors (Leonard et al., 2008). These data suggest that the endocytosis of different receptor classes might have their own unique methods of regulation.
Many endocytic proteins, such as dynamin, epsin, AP180 and CALM, contain PtdIns(4,5)P2-binding domains such as PH, PTB, ENTH/ANTH, F-BAR domains (Balla, 2005; Krauss and Haucke, 2007), whereas others, such as β-arrestin-2 and the AP-2 complex do not have a distinct binding pocket, but still contain regions of basic clusters that interact with phosphoinositides. These interactions were demonstrated mainly by in vitro studies, suggesting a role for PtdIns(4,5)P2 in the regulation of endocytosis. To address this question in intact cells, Jost and co-workers overexpressed the PH domain of PLCδ1 that binds PtdIns(4,5)P2, and showed that it decreased transferrin uptake and inhibited coated vesicle formation (Jost et al., 1998). Using treatment of cells with the Ca2+ ionophore ionomycin, which leads to the activation of calcium-dependent PLCs and thus depletes PtdIns(4,5)P2, Zoncu and colleagues found disruption of CCP formation (Zoncu et al., 2007). Overexpression of PtdIns(4,5)P2-modifying enzymes, such as PIP5K isoforms increased AP-2 recruitment to the plasma membrane and transferrin internalization (Padrón et al., 2003), whereas overexpression of the 5-phosphatase (5ptase) domain of synaptojanin-1 or knockdown of PIP5Ks led to severe defects in the internalization of transferrin receptor (Kim et al., 2006). Although these manipulations certainly change the level of plasma membrane PtdIns(4,5)P2, the long-term perturbation of the lipid as well as the consequent generation of important second messengers, such as DAG and InsP3, can also be responsible for the impairment of the endocytic process.
To overcome the limitation of prolonged PtdIns(4,5)P2 modification, new approaches have recently been developed. Boucrot and co-workers depleted plasma membrane PtdIns(4,5)P2 by treating cells with primary alcohol for 5 minutes, which in turn caused the inhibition of PIP5K and a decrease in plasma membrane PtdIns(4,5)P2 level, resulting in an impairment of the formation of CCPs and transferrin uptake (Boucrot et al., 2006). Importantly, addition of exogenous PtdIns(4,5)P2 reversed this effect (Boucrot et al., 2006). The same manipulation led to acute relocalization of FCHo2, a protein that has recently been described as a nucleator of CME (Henne et al., 2010). In another approach, we and others have developed a rapamycin-controlled PtdIns(4,5)P2 depletion system (Suh et al., 2006; Varnai et al., 2006) and confirmed the PtdIns(4,5)P2 dependence of transferrin and EGF receptor internalization (Varnai et al., 2006; Zoncu et al., 2007; Abe et al., 2008). In these studies, PtdIns(4,5)P2 was found to be important in the plasma membrane recruitment of several proteins of the endocytic machinery and hence the regulation of the endocytosis of specific cell surface receptors. However, no data were available on the importance of PtdIns(4,5)P2 in the regulation of the endocytosis of GPCRs.
In the present study, we investigated the importance of plasma membrane PtdIns(4,5)P2 in the endocytosis of three GPCRs, AT1R, 5HT2CR and β2AR. For accurate quantification of this process, we first developed a sensitive BRET-based method that allows monitoring of the movement of activated receptors along the endocytic pathway in living cells. Plasma membrane PtdIns(4,5)P2 depletion was achieved by the rapamycin-induced acute PtdIns(4,5)P2 depletion system. We found that PtdIns(4,5)P2 depletion strongly inhibited the appearance of both receptors in early endosomes. However, interaction of the receptors with β-arrestin-2 did not require PtdIns(4,5)P2. Interestingly, ligand-induced clustering of the receptors was also preserved after PtdIns(4,5)P2 depletion, with only partial impairment of the subsequent removal of the receptors from the plasma membrane.
Monitoring the clathrin-mediated endocytosis of AT1R by molecular interactions measured by BRET
Upon agonist stimulation, AT1R is phosphorylated, binds β-arrestin-2 and is then targeted to clathrin-coated pits (CCPs) at the plasma membrane (Doherty and McMahon, 2009). After the fission of CCPs, the receptor leaves the plasma membrane and is directed to the early endosomal compartment (Fig. 1A). To gain a more detailed, quantitative view of AT1R internalization, we set out to follow the process through the detection of interactions between the receptor and specific endocytic molecules. To do this, we examined the interactions of the receptor with three different molecules: a plasma-membrane-targeting peptide corresponding to the N-terminal lipid-modified sequence of Lyn (hereafter referred to as PM), β-arrestin-2 and Rab5, a small GTPase protein that localizes to early endosomes. Molecular proximity between AT1R and these molecules was tested by bioluminescence resonance energy transfer (BRET). The receptor was labeled with Renilla luciferase (AT1R–luc) whereas the partner molecules were tagged with either yellow fluorescent protein (YFP) or Venus (β-arrestin-2–YFP, Venus–Rab5 and PM–Venus). HEK293 cells transiently expressing these fusion proteins were treated with AngII at several concentrations (1–100 nM), and the BRET ratio changes were calculated and plotted as a function of time.
AngII treatment generated a robust and dose-dependent increase in the BRET signal between AT1R–luc and β-arrestin-2–YFP (Fig. 1B, top), consistent with the well-documented specific interaction between the two molecules, known to initiate receptor endocytosis (Lefkowitz and Shenoy, 2005; Turu et al., 2006). Arrival of the internalized AT1R–luc in the early endosomal compartment was indicated by a dose-dependent rise in the BRET signal between the receptor and Venus–Rab5 (Fig. 1B, middle). Similar but smaller changes of BRET signal were obtained when the FYVE domain of EEA1 (Simonsen et al., 1998) or WDFY2 (Hayakawa et al., 2006), or the full-length WDFY2 protein, each tagged with Venus, was used as early endosomal markers (data not shown). These changes were accompanied by a rapid decrease in the BRET signal between AT1R–luc and PM–Venus, indicating the removal of the receptor from the plasma membrane compartment marked by PM–Venus (Fig. 1B, bottom).
Confocal microscopy was used to confirm that these BRET changes corresponded to the sequential colocalization of AT1R with these respective molecules. For these studies, HEK293 cells stably expressing AT1R were transiently transfected with plasmids containing the DNA of the Cerulean-labeled version of β-arrestin-2, Rab5 or PM. The receptor was visualized by adding Rhodamine-labeled angiotensin II (Rhod–AngII, 300 nM) to the cells. As a result of the internalization of the receptors, 8 minutes after stimulation Rhod–AngII showed intracellular vesicular localization. In good agreement with BRET measurements, the initially cytoplasmic β-arrestin-2–Cerulean began to accumulate in vesicles, partially colocalizing with the endocytosed ligand (Fig. 1C, top panel). A similar colocalization pattern was observed between Rhod–AngII and Cerulean–Rab5 after AngII stimulation (Fig. 1C, middle panel). By contrast, PM–Cerulean, showing predominantly plasma membrane localization, did not follow the ligand into intracellular vesicles, and its colocalization with the internalized ligands after 8 minutes of stimulation was negligible (Fig. 1C, bottom panel).
To confirm that these signals were due to the process of clathrin-mediated endocytosis of the receptor, BRET experiments were also performed in a hyperosmotic medium containing 300 mM sucrose that is known to inhibit clathrin-mediated endocytosis by inducing abnormal clathrin polymerization into empty microcages at the membrane (Heuser and Anderson, 1989). In these and all subsequent experiments we used 100 nM AngII for stimulation. This hyperosmotic pretreatment almost entirely eliminated the endocytosis of the receptor as indicated by the lack of change in the BRET signal between the receptor and PM–Venus or Venus–Rab5 (Fig. 1D, green curves). Similar experiments were carried out with a mutant form of AT1R–luc in which four known phosphorylation sites of the receptor were changed (TSTS-AAAA mutant) eliminating the binding of β-arrestin-2 and preventing the receptor from entering the endocytic pathway (Hunyady et al., 1994; Thomas et al., 1998). This mutant receptor also failed to show changes in the BRET signals after stimulation (Fig. 1D, red curves). These data together strongly suggested that the changes in BRET signals between AT1R–luc and PM–Venus or Venus–Rab5 reflect the clathrin-mediated endocytosis of the receptor.
Drug-inducible rapid depletion of PtdIns(4,5)P2
We have recently developed a method for the acute depletion of PtdIns(4,5)P2 in the plasma membrane (Varnai et al., 2006). This method is based on the inducible heterodimerization of two protein domains, FKBP and FRB by rapamycin (Varnai et al., 2006). Briefly, the FRB domain was labeled with red fluorescent protein and targeted to the plasma membrane through a signal sequence (PM–FRB–mRFP), whereas FKBP was tagged with Cerulean and the catalytic domain of a phosphoinositide 5-phosphatase enzyme, which is capable of dephosphorylating and thus eliminating PtdIns(4,5)P2, was attached to it (Cerulean–FKBP–5ptase). The rapamycin-induced heterodimerization of FRB and FKBP recruits the enzyme to the plasma membrane where it specifically degrades PtdIns(4,5)P2 (Fig. 2A).
To verify the efficiency of this method and to determine the extent and kinetics of lipid elimination under BRET experiment conditions, we measured energy transfer between the Renilla luciferase- and YFP-tagged PH domains of PLCδ1 protein (PLCδ1PH–Sluc and PLCδ1PH–YFP), which bind PtdIns(4,5)P2 specifically, and are therefore suitable to monitor the plasma membrane PtdIns(4,5)P2 level (Várnai and Balla, 1998). When PtdIns(4,5)P2 is present, both these molecules are bound to the plasma membrane and create a relatively high BRET signal. Upon depletion of the lipid, however, they translocate to the cytoplasm, followed by a large drop in the BRET signal (our unpublished observation). To rule out the possibility that the changes were caused by non-specific actions of rapamycin, two different types of control were used: together with PM–FRB–mRFP we either expressed the mRFP–FKBP construct without the 5-ptase domain (FKBP only) and then treated the cells with rapamycin (Fig. 2B, blue curve), or expressed the mRFP–FKBP–5-ptase construct and treated the cells with only the vehicle DMSO instead of rapamycin (Fig. 2B, green curve). Both controls remained unaffected by rapamycin or DMSO treatment, but a massive decrease in the BRET signal was detected in both cases after the addition of 10 μM ionomycin (Fig. 2B), which leads to complete degradation of PtdIns(4,5)P2 as a result of the increase of cytoplasmic Ca2+ and consequent activation of PLC enzymes (Várnai and Balla, 1998). In cells containing the mRFP–FKBP–5-ptase construct, the administration of rapamycin generated a fast decrease in the BRET signal, indicating the depletion of PtdIns(4,5)P2 (Fig. 2B, red curve). The signal change reached its maximum after 3 minutes, and could only be slightly enhanced with ionomycin treatment.
The effect of depletion of plasma membrane PtdIns(4,5)P2 on the internalization of AT1R
To investigate the role of PtdIns(4,5)P2 in the internalization of AT1R, we combined the BRET-based endocytosis detection method with rapamycin-inducible PtdIns(4,5)P2 depletion. We expressed AT1R–luc and its fluorescently labeled partners (β-arrestin-2–YFP, Venus–Rab5 and PM–Venus) together with the two proteins of the depletion system in HEK293 cells and applied rapamycin (200 nM) to degrade PtdIns(4,5)P2 in the plasma membrane. After 5 minutes, the cells were stimulated with 100 nM AngII and BRET signal changes were monitored for 30 minutes. We used both of the above described controls for each experiment. In cells expressing FKBP without the 5-ptase domain that were treated with rapamycin, AngII induced signal changes equivalent to those measured previously without the lipid-depletion machinery in all three interactions tested (Fig. 3A). Similarly, the signal changes were unaltered in cells expressing the FKBP–5-ptase but treated with DMSO alone (Fig. 3B, green curves). However, depletion of PtdIns(4,5)P2 strongly inhibited the signal changes between the receptor and Venus–Rab5 or PM–Venus, while having no effect on the BRET signal change between the receptor and β-arrestin-2–YFP (Fig. 3B, red curves). To quantify the inhibitory effect of rapamycin, the averages of the last five values of each curve were calculated and expressed as percentage of the corresponding controls. Despite the previously reported PtdIns(4,5)P2 binding ability of β-arrestin-2, interaction between AT1R and β-arrestin-2 was not affected by PtdIns(4,5)P2 depletion (100.3±4.8%, n=3), whereas the BRET signal change was abolished in the case of Venus–Rab5 (–1.9±0.7%, n=3), but only reduced by 50% with the receptor and PM–Venus (49.6±9.2%, n=3).
The effect of depletion of plasma membrane PtdIns(4,5)P2 on the internalization of additional plasma membrane receptors
To test whether this inhibitory effect of PtdIns(4,5)P2 depletion is specific for AT1R or has a more general influence on GPCR internalization, similar experiments were carried out with another Gq-coupled receptor, the type 2C serotonin receptor (5HT2CR), as well as a Gs-coupled receptor, the β2 adrenergic receptor (β2AR). These receptors were tagged with a modified form of Renilla luciferase, termed Super luciferase for its higher effectiveness (5HT2CR–Sluc and β2AR–Sluc) (Woo and von Arnim, 2008).
Under control conditions (without rapamycin), upon stimulation with serotonin (10 μM) or isoprenalin (1 μM), the interaction between the receptors and Venus–Rab5 or the receptors and PM–Venus was similar to those observed with AT1R, whereas the interaction between the receptors and β-arrestin-2–YFP showed different kinetics (compare Fig. 3A with Fig. 4A,B, green curves). The effect of PtdIns(4,5)P2 depletion (rapamycin treatment) was also very similar: although it slightly affected the kinetics, it did not alter the magnitude of the signal between either of the receptors and β-arrestin-2–YFP. For both receptors, rapamycin treatment entirely prevented the signal increase with Venus–Rab5, and partially inhibited the signal decrease between the receptors and PM–Venus (Fig. 4A,B, red curves).
Using fluorescently labeled EGF and confocal microscopy, we previously found that plasma membrane PtdIns(4,5)P2 depletion reduces the endocytosis of EGF receptors (Varnai et al., 2006). To test whether this effect can be observed in the BRET system, we also created a luciferase-tagged EGF receptor (EGFR–Sluc). Because the endocytosis of EGFR is not arrestin dependent, in these experiments we used only the PM-targeted Venus and our endosomal markers. Interestingly, although the decrease of the BRET signal between EGFR–Sluc and PM–Venus showed the disappearance of the receptor from the cell surface, its delivery to the endosomal compartment was more prominent with the Venus–FYVE of WDFY2 (Fig. 4C, green curves) compared with Venus–Rab5 (not shown). Notably, we did not see this difference with the examined GPCRs. In accordance with our previous findings, PtdIns(4,5)P2 depletion reduced the endocytosis of EGFR, but the inhibition was partial in both cases (Fig. 4C, red curves).
Expression of the rapamycin-induced PtdIns(4,5)P2 depletion system using a single ‘self-cleaving’ T2A peptide-based plasmid
To gain further insight into the mechanism by which PtdIns(4,5)P2 depletion caused only a partial blockade of the change in the BRET signal with PM–Venus, we turned to confocal microscopy. The original rapamycin-induced plasma membrane PtdIns(4,5)P2 depletion system requires the expression of PM–FRB and FKBP–5-ptase, which is provided by the co-transfection of two plasmids encoding these two proteins. Unfortunately, complete co-transfection of the cells with both plasmids never occurs, and the relative amounts of the two expressed proteins show great cell-to-cell variation, resulting in a high level of variability in the degree of PtdIns(4,5)P2 depletion, which makes single-cell experiments, such as confocal microscopy uncertain. To overcome this problem, we aimed to develop a system which provides stable PtdIns(4,5)P2 depletion in the cells. To do this, we created a single plasmid which contained the coding sequences of the two proteins connected by the sequence of the viral T2A peptide (Fig. 5A) (Szymczak et al., 2004). During translation, a molecular cleavage occurs within this peptide, leading to the expression of two separate proteins in equimolar amounts – in our case PM–FRB–mRFP–T2A and FKBP–5-ptase. As a consequence, fluorescent labeling of both proteins was no longer necessary. To test whether this method is capable of efficient plasma membrane PtdIns(4,5)P2 depletion, HEK293 cells were transfected with this plasmid together with the plasmid of PLCδ1PH–Venus. As expected, PM–FRB–mRFP–T2A showed clear plasma membrane localization (Fig. 5B). In resting cells, Venus-tagged PH domains were also recruited to the plasma membrane, indicating that the FKBP–5-ptase was not in the membrane. Upon rapamycin treatment, however, the PH domain became cytoplasmic, proving the occurrence of PtdIns(4,5)P2 depletion (Fig. 5B, bottom panel). For quantitative comparison of the old and new depletion systems, BRET measurements were performed between the luciferase- and Venus-tagged PH domains, as described earlier. We found the efficiency of the new system to be very similar to the original: 5 minutes after rapamycin addition, BRET ratios were 32.5±2.1% (n=4) and 28.7±4.2% (n=3) of control values for the new and the original depletion system, respectively (Fig. 5C and Fig. 2B, red curves).
Plasma membrane PtdIns(4,5)P2 depletion did not prevent clustering of AT1R
To examine the effect of PtdIns(4,5)P2 depletion on the endocytosis of AT1R by confocal microscopy, HEK293 cells were transiently transfected with a plasmid coding AT1R and the newly developed T2A-based plasmid (encoding PM–FRB–mRFP–T2A and FKBP–5ptase). During the measurements, we selected cells with a high enough expression of PM–FRB–mRFP–T2A for effective PtdIns(4,5)P2 depletion. Again, because proteins of the lipid depletion system are translated from a single mRNA that originates from a single plasmid, detection of PM–FRB–mRFP–T2A guarantees the presence of FKBP–5-ptase, and thus the PtdIns(4,5)P2 depletion upon rapamycin treatment. Cells were then pretreated with rapamycin (200 nM) or DMSO before stimulation with 100 nM Alexa-Fluor-488-labeled AngII (Alexa488–AngII). In cells expressing the proteins of the PtdIns(4,5)P2 depletion system in which rapamycin pretreatment was not performed (control cells), Alexa488–AngII exhibited normal internalization of the AT1Rs within 2–3 minutes (Fig. 6A control and supplementary material Movie 1, left panel), similar to what is shown in Fig. 1C. By contrast, in cells pretreated with rapamycin (200 nM, 4 minutes), Alexa488–AngII outlined the plasma membrane and accumulated in bright puncta along the membrane, but did not enter the cells for at least 5 minutes (Fig. 6A rapamycin and supplementary material Movie 1, right panel). In some experiments, cells were also transfected with TK PLCδ1PH–Venus, which resulted in a very low expression of this PtdIns(4,5)P2 sensor. The loss of the PLCδ1PH domain from the plasma membrane after rapamycin treatment, clearly showed PtdIns(4,5)P2 depletion as a built-in sensor (supplementary material Movie 1, right panel).
In BRET experiments we found that the interaction between AT1R and β-arrestin-2 did not depend on plasma membrane PtdIns(4,5)P2 levels. In addition, AT1R belongs to the family of class A GPCRs, where β-arrestin-2 stays bound to the activated receptors for a longer period, even after endocytosis. Taken together, these ideas raised the possibility that similar to labeled ligands, fluorescently tagged β-arrestin-2 could also be used as a marker for AT1R. To do this, HEK293 cells were transfected with the T2A-based PtdIns(4,5)P2 depletion system, together with AT1R and YFP-tagged β-arrestin-2. Without PtdIns(4,5)P2 depletion, upon stimulation with AngII (100 nM) β-arrestin-2–YFP was first recruited to the plasma membrane, where it bound the activated receptors, and was then internalized with the receptors (Fig. 6B control and supplementary material Movie 2, left panel). By contrast, in cells, where PtdIns(4,5)P2 depletion was achieved by rapamycin treatment (200 nM, 4 minutes), after recruitment to the plasma membrane, receptor-bound β-arrestin-2–YFP accumulated in plasma membrane clusters formed by the activated receptors (Fig. 6B rapamycin and supplementary material Movie 2, right panel).
To determine whether these receptor clusters corresponded to clathrin-containing structures in the plasma membrane, we co-transfected cells with the DNA of AT1R, the DNA of the GFP-tagged light chain polypeptide of clathrin driven by the TK promoter (GFP–CLA), and with the PM–FRB–Cerulean–T2A–FKBP–5-ptase plasmid, to get reliable PtdIns(4,5)P2 depletion upon rapamycin treatment. Cells were pretreated with rapamycin (200 nM) for 5 minutes and then stimulated with 100 nM Rhod–AngII for another 5 minutes. Cells expressing all of these proteins (PM–FRB–Cerulean–T2A as blue signal, GFP–CLA as green signal, and Rhodamine as red signal) were selected and captured. The GFP–CLA level was crucial, because at low expression levels it was hardly visible, whereas at high expression levels, GFP–CLA in clathrin-coated pits was masked by cytoplasmic GFP–CLA. Interestingly, in cells with distinguishable GFP–CLA-stained pits, rapamycin treatment did not have any effect visible by confocal microscope on the localization and distribution of these pits. After rapamycin addition, as an indication of PtdIns(4,5)P2 depletion, Rhod–AngII formed clusters, remained along the plasma membrane, and showed definite colocalization with GFP–CLA (Fig. 6C). This indicates that in spite of PtdIns(4,5)P2 depletion AT1R was able to move and reach clathrin-containing structures in the plasma membrane, suggesting that the very early steps of the endocytic process are, at least partly, PtdIns(4,5)P2 independent. This PtdIns(4,5)P2-independent lateral movement of the receptor upon stimulation is also in accordance with and might account for the partial BRET decrease seen after PtdIns(4,5)P2 depletion between the receptor and the plasma-membrane-targeted PM–Venus protein in Fig. 3B.
In this study, we investigated whether plasma membrane PtdIns(4,5)P2 depletion can disrupt the endocytic process of GPCRs. To follow the receptor along its route from the cell surface into the cell interior in living cells, we set up a highly sensitive and reliable BRET-based assay that allows the detection of the interaction between luciferase-tagged receptors and various cellular compartments labeled with fluorescent proteins. Using this approach, we were able to follow the delivery of AT1 and 5HT2C receptors into Rab5-specific endosomal vesicles and monitor the sequential disappearance of the receptors from the plasma membrane and their interaction with β-arrestin-2. A crucial early step in GPCR internalization is the movement of receptors into clathrin-coated membrane pits, which can be easily detected by confocal microscopy as a colocalization of the stimulated receptors with GFP-tagged CLA or β-adaptin-2 (data not shown). Interestingly, none of these proteins, as markers of the clathrin-coated membrane compartment, produced a measurable BRET signal change upon stimulation of the AT1 or 5HT2C receptors (not shown), most probably because of the non-optimal spatial orientation of the luciferase enzyme and the fluorescent tag in the large protein complex.
It has now been well documented that plasma membrane PtdIns(4,5)P2 plays a fundamental role in the process of clathrin-mediated endocytosis. Because the majority of endocytic molecules are able to bind PtdIns(4,5)P2 in vitro, the PtdIns(4,5)P2 dependence of such a process was not unexpected. But the relative importance of the PtdIns(4,5)P2 of each step of receptor endocytosis in case of various plasma membrane receptors still remains an open question. It is of similar interest which of these molecular events can take place at reduced PtdIns(4,5)P2 levels. BRET experiments in this study clearly showed that the delivery of AT1R to endosomal vesicles was completely abolished by plasma membrane PtdIns(4,5)P2 depletion. However, initial events, such as β-arrestin-2 binding of the phosphorylated receptor, were not affected by this manipulation. This is an important finding because some forms of GPCR kinases as well as β-arrestin-2 were shown to bind PtdIns(4,5)P2 (DebBurman et al., 1995; Pitcher et al., 1996; Gaidarov et al., 1999). Surprisingly, PtdIns(4,5)P2 depletion only partially inhibited the sequestration of the receptor from the plasma membrane marker PM–Venus. This suggests that the receptors are able to leave their resting position upon stimulation and to proceed to a certain point in the endocytic process even when plasma membrane PtdIns(4,5)P2 is greatly reduced. The fact that AT1R receptors were able to form clusters upon stimulation in the absence of PtdIns(4,5)P2, and that CLA was also present in these clusters suggest that PtdIns(4,5)P2 depletion prevents efficient fission of these pits from the plasma membrane, rather than their formation and assembly. These results are in agreement with the study of Abe and co-workers, who found that PtdInsP2 depletion decreased the AP-2-mediated lateral movement of the transferrin receptor but did not affect the assembly of CCPs in HeLa cells (Abe et al., 2008). However, other studies found that acute depletion of PtdIns(4,5)P2 resulted in the loss of endocytic CCPs in unstimulated COS-7 and BSC-1 cells (Boucrot et al., 2006; Zoncu et al., 2007). These apparent discrepancies can be reconciled because the molecular composition of the endocytic complex is dependent on the cargo, which itself can contribute to the stabilization of CCPs (Ehrlich et al., 2004), and hence modify the PtdIns(4,5)P2 sensitivity of the entire endocytic process. It is also worth pointing out that some of the studies where PtdIns(4,5)P2 depletion was shown to disassemble CCPs, examined the process on the plasma membrane of unstimulated cells, whereas in our studies we focused on CCPs induced by stimulated AT1 receptors.
To examine whether these results are characteristic for AT1Rs or have a more general implication, other GPCRs such as type 2C serotonin (5HT2C) receptor (Gq-coupled, similarly to AT1R), and the Gs-coupled β2-adrenergic (β2A) receptor were also tested. To verify and quantitatively investigate the effect of plasma membrane PtdIns(4,5)P2 depletion on the endocytosis of EGF receptor, which we described in an earlier study (Varnai et al., 2006), we also created its luciferase-tagged version for use in our BRET assay. For GPCRs, in general, all results were very similar to those with AT1R. As a consequence of plasma membrane PtdIns(4,5)P2 depletion, we saw an almost complete inhibition of the delivery of both receptors to the Rab5 compartment, whereas the interaction between the receptors and PM–Venus was only partially reduced. Notably, this inhibition was observed mainly at later time points (2–3 minutes after stimulation with the appropriate agonist), which is in accordance with the conclusion that the initial steps of endocytosis seem to be independent of PtdIns(4,5)P2. Regarding the magnitude, we also found the interaction between the receptors and β-arrestin-2 to be resistant to PtdIns(4,5)P2 depletion, although for β2AR the kinetics of this interaction showed some delay, suggesting some form of inhibition, which definitely requires further attention. For the EGF receptor, we confirmed the PtdIns(4,5)P2 dependence of its endocytosis, but unlike the examined GPCRs, the delivery of EGFR to early endosomes (the Venus-tagged FYVE domain of WDFY2 was used in this experiment because the signal was larger) was not completely abolished. This observation suggests that the clathrin-independent route of EGFR internalization (Goh et al., 2010) might not depend on the plasma membrane PtdIns(4,5)P2 level, or that PtdIns(4,5)P2 dependency might occur at different levels for different receptor types.
It is well documented that the PtdIns(4,5)P2 binding of dynamin through its PH domain has an essential role in clathrin-mediated endocytosis (Achiriloaie et al., 1999; Lee et al., 1999; Vallis et al., 1999). In light of this, the most plausible explanation for our finding is that the action of dynamin is impaired in cells depleted of plasma membrane PtdIns(4,5)P2. However, recent studies also showed that a sequence of complex events, including the appearance and disappearance of inositol lipid-modifying enzymes, takes place during the maturation of the CCPs to CCVs (Antonescu et al., 2011; Taylor et al., 2011). For example, the dephosphorylation of PtdIns(4,5)P2 at highly curved membrane regions by an isoform of the 5-phosphatase synaptojanin 1 (Synj1–145) did facilitate membrane fission (Chang-Ileto et al., 2011). It was also shown that this synaptojanin isoform is not present in CCPs during their entire lifetime, but appears on them immediately before vesicle scission (Perera et al., 2006). In another study, it was reported that SHIP2 – a 5-phosphatase capable of PtdIns(4,5)P2 degradation – can inhibit CCP dynamics, confirming the requirement of PtdIns(4,5)P2 during CCP maturation. To perform its inhibitory effect, SHIP2 is recruited to CCPs from the early stages but leaves before membrane fission (Nakatsu et al., 2010). It was also shown that the PtdIns(4,5)P2-producing enzyme PIPKIγ directly interacts with and is activated by the endocytic AP-2 complex (Bairstow et al., 2006; Krauss et al., 2006). Interestingly, this interaction has been proposed to be regulated by clathrin, because the polymerizing clathrin lattice might displace the enzyme from AP-2, thus terminating PtdIns(4,5)P2 production and allowing the decrease of PtdIns(4,5)P2 level as the clathrin coat assembly advances (Thieman et al., 2009). From all of these papers, it appears that a whole cycle of PtdIns(4,5)P2 removal and reappearance dictates the sequential recruitment of protein factors that are needed for the maturation of CCVs, but at the end of this sequence, there is a need for PtdIns(4,5)P2 to initiate the fission process. In a recent paper Bethoney and colleagues, they suggest that dynamin is able to cluster PtdIns(4,5)P2 at the neck of CCPs, which might be crucial for its role in promoting membrane fission (Bethoney et al., 2009), raising the possibility that the assembly and activation of the fission machinery might be the PtdIns(4,5)P2-sensitive step of endocytosis, which is consistent with our data.
The 5-phosphatase we used in our PtdIns(4,5)P2 depletion system is also able to dephosphorylate phosphatidylinositol 3,4,5-trisphosphate [PtdIns(3,4,5)P3] (Kisseleva et al., 2000). Therefore, we cannot rule out the possibility that our findings are caused by the depletion of PtdIns(3,4,5)P3, not PtdIns(4,5)P2. Nevertheless, the inhibition of PI3Ks by wortmannin hardly affects endocytic CCP dynamics (Shpetner et al., 1996), and – more importantly – does not inhibit the internalization of AT1R (Hunyady et al., 2002), suggesting that the inhibition seen is the consequence of PtdIns(4,5)P2 depletion.
Because activation of Gq-coupled receptors also reduces PtdIns(4,5)P2 levels, it was of interest to determine whether the PtdIns(4,5)P2 depletion connected with Gq protein activation impacts the endocytosis of the receptor. Therefore, we also looked at the DRY/AAY mutant form of AT1R, which is unable to activate PLCβ (Gáborik et al., 2003) but still binds β-arrestin-2 (Wei et al., 2003; Szidonya et al., 2007). We reasoned that this mutant receptor might show increased endocytosis upon stimulation. However, in accordance with previous results (Gáborik et al., 2003), we observed a slight decrease in its internalization compared with that of the wild-type receptor, as measured by BRET between the receptor and either PM or Rab5 (data not shown). The effect of PtdIns(4,5)P2 depletion on the internalization of DRY/AAY mutants was also similar to that on the wild-type receptors in every aspect investigated in this study (data not shown). These experiments suggested little, if any, impact of the agonist-induced changes in hormone-sensitive PtdIns(4,5)P2 pools on the internalization process.
An important part of this study was the further development of the rapamycin-induced plasma membrane PtdIns(4,5)P2 depletion system. By connecting the two proteins required for lipid depletion with the viral T2A peptide, which has a so-called ‘self-cleaving’ property (Szymczak et al., 2004), we managed to create a single plasmid suitable for the cellular expression of both proteins in equimolar amounts. The degree of PtdIns(4,5)P2 depletion on the cell population level was comparable with the original system, and because of the constant protein ratio, the T2A-based system definitely has the advantage of the uniform extent of PtdIns(4,5)P2 depletion at a certain expression level, which is considered a significant step forward, especially in experiments such as confocal microscopy performed at the single-cell level.
In summary, this study showed the PtdIns(4,5)P2 dependence of the internalization of selected GPCRs using a combination of methods to follow the trafficking steps of the receptors and acutely manipulate PtdIns(4,5)P2 in the plasma membrane. We conclude that the receptor movement into CCPs, as well as its association with β-arrestin-2 are not dependent on PtdIns(4,5)P2, but the maturation of CCPs and/or their fission does not occur without this phospholipid. More studies will be needed to further resolve the complex PtdIns(4,5)P2 dependence of CCP maturation in the case of GPCRs and to understand the exact mechanism of the fission block that is probably related to the PtdIns(4,5)P2 dependence of dynamin function (Bethoney et al., 2009). Another protein of interest will be CK2, which is able to inactivate many endocytic proteins through phosphorylation (Korolchuk et al., 2005) and whose activity at the CCPs is regulated by PtdIns(4,5)P2 degradation (Korolchuk et al., 2005).
Materials and Methods
Molecular biology reagents were obtained from Fermentas (Vilnius, Lithuania). Cell culture dishes and plates were purchased from Greiner (Kremsmunster, Austria). Coelenterazine h and Lipofectamine 2000 were from Invitrogen (Carlsbad, CA). Rapamycin was obtained from Merck (Darmstadt, Germany). GeneCellin transfection reagent was from BioCellChallenge (Toulon, France). Angiotensin II, Rhodamine and Alexa Fluor 488 conjugates were purchased from Phoenix Pharmaceuticals (Belmont, CA) and Molecular Probes (Invitrogen), respectively. Unless otherwise stated, all other chemicals and reagents were purchased from Sigma (St Louis, MO).
The Renilla luciferase tagged rat AT1a receptor (AT1R–luc) was created by exchanging the sequence of fluorescent protein in AT1R–YFP (Turu et al., 2006) to the sequence of humanized Renilla luciferase (Promega, Madison, WI) using AgeI and NotI restriction enzymes. The DRY-AAY mutant of the AT1R (D125A,R126A) was described earlier (Gáborik et al., 2003). The TSTS-AAAA mutations (T332A, S335A, T336A and S338A) (Hunyady et al., 1994; Thomas et al., 1998) were introduced by conventional site-directed mutagenesis (Stratagene). After verifying the mutations with dideoxy sequencing, the mutated fragment was exchanged between the wild-type and mutated portion with suitable restriction sites to avoid the generation of unwanted mutations outside the sequenced regions.
To create the Renilla-luciferase-tagged human type 2C serotonin receptor construct (5HT2CR–Sluc), first the receptor sequence was amplified from the cDNA clone of the human 5HT2C receptor (Clone ID: HTR02C0000, GenBank Accession Number: NM_000868) purchased from S&T cDNA Resource Center (Rolla, MO), and was subcloned into the pEGFP-N1 vector using XhoI and KpnI restriction enzymes. Then the sequence of the green fluorescent protein was replaced with the sequence of super Renilla luciferase (Woo and von Arnim, 2008) using AgeI and NotI enzymes. The construct used in this study contained a sequence corresponding to an RNA-edited form of 5HT2C receptor where three amino acids had been altered (I156V, N158G, I160V). This version of the receptor – unlike the non-edited form – had been reported to lack constitutive activity and show high cell-surface expression (Price et al., 2001). These amino acid changes were introduced by site-directed mutagenesis in the cDNA sequence.
To create the Renilla-luciferase-tagged human β2-adrenergic receptor construct (β2AR–Sluc), first the receptor sequence was amplified from the cDNA clone of the human β2-adrenergic receptor (Clone ID: AR0B200000, GenBank Accession Number: NM_000024.3) purchased from S&T cDNA Resource Center, and was subcloned into the pEYFP-N1 vector using XhoI and EcoRI restriction enzymes. Then the sequence of the yellow fluorescent protein was replaced with the sequence of super Renilla luciferase (Woo and von Arnim, 2008) using AgeI and NotI enzymes.
Luciferase-tagged EGFR was created by first subcloning the human EGFR from the pCDNA3.1(-) (Várnai et al., 2005) into the pEGFP-N1 plasmid using NheI and SalI enzymes. Then the sequence of the green fluorescent protein was replaced with the sequence of super Renilla luciferase (Woo and von Arnim, 2008) using AgeI and AflII enzymes.
Cerulean- (Rizzo et al., 2004) and Venus- (Nagai et al., 2002) tagged Rab5 were generated by replacing the fluorescent protein of GFP–Rab5 used earlier (Hunyady et al., 2002). To target Venus and Cerulean to the plasma membrane, the targeting sequence of Lyn kinase was used as described previously (Várnai et al., 2007). This N-terminal 13 amino acids of Lyn is myristoylated and palmitoylated, therefore is believed to partition to ‘rafts’ of the plasma membrane (Zacharias et al., 2002). Both Cerulean and Venus used in this study contained the A206K mutation rendering these proteins monomeric (Zacharias et al., 2002). β-arrestin-2–YFP was generated by subcloning β-arrestin-2 into the pEYFP-N1 vector using NheI and KpnI enzymes as described previously (Turu et al., 2006). β-arrestin-2–Cerulean was generated by replacing YFP with Cerulean. The sequences of human full-length WDFY2 protein and its FYVE domain only (amino acid residues 277–353) were amplified from the cDNA clone of human WDFY2 purchased from Invitrogen (Clone ID: 3842589, GenBank Accession Number: NM_052950) and were subcloned into the pEGFP-C1 vector using EcoRI and KpnI restriction enzymes. The sequence of GFP was then replaced with the sequence of Venus using AgeI and BglII enzymes. GFP–Clathrin light polypeptide was a kind gift from Louis Greene (NHLBI, NIH, Bethesda, MD) (Yim et al., 2010), but its cytomegalovirus promoter was replaced with herpes simplex virus thymidine kinase (TK) promoter to lower expression levels. PLCδ1PH–Sluc and PLCδ1PH–Venus were created by changing the sequence of the fluorescent protein in PLCδ1PH–YFP (Várnai and Balla, 2007) to the sequence of Super Renilla luciferase or Venus, respectively, using AgeI and NotI enzymes.
PM–FRB–mRFP and mRFP–FKBP–5-ptase constructs used for rapamycin-induced PtdInsP2 depletion were described earlier (Varnai et al., 2006), with the difference that for plasma membrane targeting of the FRB protein, we used the N-terminal targeting sequence of Lyn (MGCIKSKGKDSAGA) instead of GAP43 (MLCCMRRTKQVEKNDDDQKI). Creation of the viral T2A peptide sequence plasmid, which allows the expression of both protein constructs required for the plasma membrane PtdIns(4,5)P2 depletion was performed in two steps. First, the T2A peptide sequence (Szymczak et al., 2004) was fused to the N-terminal of the FKBP–5-ptase construct, and then this new protein was fused to the C-terminal of PM–FRB–mRFP protein. The whole process resulted in the DNA sequence of PM–FRB–mRFP–linker1–T2A–linker2–FKBP–5-ptase with only one stop codon at the end. (Linker1: VDSGS, linker2: PVAT) The construct that contained PM–FRB–Cerulean instead of PM–FRB–mRFP was created similarly.
Human embryonic kidney (HEK293) cells were obtained from American Type Culture Collection (ATCC, Manassas, VA) and maintained in Dulbecco's modified Eagle's medium (Lonza 12–604) supplemented with 10% fetal bovine serum, 50 U/ml penicillin, and 50 μg/ml streptomycin in a 5% humidified CO2 incubator at 37°C.
Bioluminescent resonance energy transfer (BRET) measurement of cells
For BRET measurements, HEK293 cells were cultured in 10 cm plastic dishes and were trypsinized before transfection. For transient transfection, cells were plated on poly-lysine-pretreated white 96-well plates at 1×105 cells/well density with the indicated DNA constructs (0.24–0.48 μg total DNA/well) and the cell transfection reagent (0.5 μl/well Lipofectamine 2000 or 1.5 μl/well GeneCellin). Measurements were performed 24–26 hours after transfection. In some measurements, HEK293 cells were switched to serum-free medium (DMEM with 0.1% BSA) for several hours (4–6 hours) to render the cells quiescent. Before measurements were made, the medium of cells was changed to modified Krebs–Ringer buffer containing 120 mM NaCl, 4.7 mM KCl, 1.2 mM CaCl2, 0.7 mM MgSO4, 10 mM glucose, and 10 mM Na-HEPES, pH 7.4. For hyperosmotic conditions, this solution was supplemented with 300 mM sucrose. Measurements were performed at 37°C using a Mithras LB 940 multilabel reader (Berthold). The measurements started with the addition of the cell permeable luciferase substrate, coelenterazine h (Invitrogen) at a final concentration of 5 μM, and counts were recorded using 485 and 530 nm emission filters. Detection time was 0.5 seconds for each wavelength. Measurements were done in triplicate. BRET ratios were calculated by dividing the 530 nm and 485 nm intensities, and showed as difference of stimulated and unstimulated, baseline-corrected curves. A step-by-step description of this calculation is shown in supplementary material Fig. S1.
Confocal analysis of single cells
HEK293 cells were cultured on poly-lysine-pretreated (0.001%, 1 hour) No. 1.5 glass coverslips (3×105 cells/35 mm dish) and transfected with the indicated constructs (1–2 μg DNA total/dish) using 2 μl/dish Lipofectamine 2000 for 24 hours. Confocal measurements were performed at 35°C in a modified Krebs–Ringer buffer described above, using a Zeiss LSM 510 scanning confocal microscope and a 63×/1.4 NA objective. In colocalization experiments, data were acquired in multi-track line-scan mode. For other experiments multi-track frame-scan mode was used to minimize crosstalk between the fluorophores. Post-acquisition picture analysis was performed using Photoshop (Adobe) software to expand to the full dynamic range but only linear changes were allowed.
The authors are grateful to Luca Tóth for generating the β2AR–Sluc construct. The technical assistance of Kata Szabolcsi is also appreciated.
P.V. was supported by the Hungarian Scientific Research Fund [grant number OTKA NF-68563]; and the Medical Research Council [grant number ETT 494/2009]. A.B. was supported by the János Bolyai Research Scholarship of the Hungarian Academy of Sciences. T.B. was supported by the Intramural Research Program of the Eunice Kennedy Shriver National Institute of Child Health and Human Development of the National Institutes of Health. Deposited in PMC for release after 12 months.
Supplementary material available online at http://jcs.biologists.org/lookup/suppl/doi:10.1242/jcs.097279/-/DC1
- Accepted December 22, 2011.
- © 2012.