Agonist-induced internalization and desensitization of the apelin receptor

Apelin acts via the G protein-coupled apelin receptor (APJ) to mediate effects on cardiovascular and fluid homeostasis. G protein-coupled receptor (GPCR) trafficking has an important role in the regulation of receptor signalling pathways and cellular functions, however in the case of APJ the mechanisms and proteins involved in apelin-induced trafficking are not well understood. We generated a stable HEK-293 cell line expressing N-terminus HA-tagged mouse (m) APJ, and used a semi-automated imaging protocol to quantitate APJ trafficking and ERK1/2 activation following stimulation with [Pyr1]apelin-13. The mechanisms of [Pyr1]apelin-13-induced internalization and desensitization were explored using dominant-negative mutant (DNM) cDNA constructs of G protein-coupled receptor kinase 2 (GRK2), β-arrestin1, EPS15 and dynamin. The di-phosphorylated ERK1/2 (ppERK1/2) response to [Pyr1]apelin-13 desensitized during sustained stimulation, due to upstream APJ-specific adaptive changes. Furthermore, [Pyr1]apelin-13 stimulation caused internalization of mAPJ via clathrin coated vesicles (CCVs) and also caused a rapid reduction in cell surface and whole cell HA-mAPJ. Our data suggest that upon continuous agonist exposure GRK2-mediated phosphorylation targets APJ to CCVs that are internalized from the cell surface in a β-arrestin1-independent, EPS15- and dynamin-dependent manner. Internalization does not appear to contribute to the desensitization of APJ-mediated ppERK1/2 activation in these cells.

As with most G protein-coupled receptors (GPCRs), sustained activation of APJ can cause desensitization and this has been reported to occur for APJ-mediated effects on cytoplasmic Ca 2þ concentration, as well as for effects on activity of adenylyl cyclase, ERK and Akt (Ishida et al., 2004;Masri et al., 2006). APJ also undergoes agonist-induced internalization and down-regulation and so research has focused on the possible role for the canonical pathway for rapid homologous receptor desensitization and trafficking in mediating adaptive responses to APJ activation (Evans et al., 2001;Zhou et al., 2003;Lee et al., 2010). In this pathway, agonist occupied GPCRs are preferred substrates for phosphorylation by G-protein receptor kinases (GRKs) and this phosphorylation mediates binding with b-arrestins that prevent the receptors from activating their cognate G-proteins, thereby causing receptor desensitization. The b-arrestins also target the desensitized receptors for internalization via clathrin-coated vesicles (CCVs). After this the vesicles are uncoated, b-arrestins dissociate, receptors are dephosphorylated and the receptor-containing vesicles may be trafficked back to the plasma membrane (a process that can mediate resensitization to the agonist) or to lysosomes for proteolytic digestion (a process that can cause receptor downregulation). Differing patterns of b-arrestin interaction have allowed the sorting of GPCRs into two classes: Class A receptors, that have a brief interaction with b-arrestins (at the plasma membrane) and preferentially bind b-arrestin2 over -1, and display rapid recycling; and Class B receptors, that form a stable complex with both b-arrestins with equal affinity, and which internalize with the b-arrestins into endosomes. Additional players in this process include epsin and EPS15, which act as adapter proteins for clathrin-mediated endocytosis (CME) (Wolfe and Trejo, 2007), and dynamin, a GTPase that forms a multimeric complex around the neck of nascent endocytic vesicles and mediates their budding off to form endosomes (Damke, 1996).
The adaptive processes outlined above are thought to be relevant for APJ as apelin causes clathrin-mediated APJ internalization (Reaux et al., 2001;El Messari et al., 2004) and also translocation of b-arrestin1 and -2 to the cell surface, indicating translocation to phosphorylated APJ (Lee et al., 2010). Moreover, after agonistinduced internalization, APJ can either be recycled to the cell surface or be degraded in lysosomes (Lee et al., 2010). Interestingly, APJ trafficking displays ligand bias for both Class A and B b-arrestin/ recycling behaviour as when internalization is stimulated by [Pyr 1 ] apelin-13, internalized APJ is rapidly recycled to the plasma membrane with none remaining in the cytoplasm at 60 min, whereas APJ is retained within the cell for up to 120 min after apelin-36-stimulated internalization (Zhou et al., 2003). Similarly, although apelin-13 causes b-arrestin1 translocation to the plasma membrane, the internalized receptors are not associated with b-arrestin1 and are rapidly recycled to the cell surface via early endosomes (Evans et al., 2001;Lee et al., 2010), whereas after apelin-36 stimulation the internalized APJ are co-localized with b-arrestin1 and then undergo rab-7-dependent trafficking to lysosomes (Lee et al., 2010). Finally, truncation of the APJ C-terminus (in order to delete potential GRK phosphorylation sites) prevents homologous desensitization to effects of apelin-13, but not to those of apelin-36, on inhibition of adenylyl cyclase and activation of ERK and Akt (Masri et al., 2006;Lee et al., 2010).
Apelin/APJ has emerged as a major signalling pathway in physiological homeostasis (O'Carroll et al., 2013) and central to ascertaining the precise function of this receptor is an understanding of the system of regulation that dynamically modulates APJ signalling. In peripheral tissues the apelinergic system appears to be down-regulated in hypertensive disease e levels of apelin immunoreactivity in plasma, and in ventricular and aortic tissues, are lower in the spontaneously hypertensive rat, a genetic model of hypertension, than in control Wistar-Kyoto normotensive rats (Zhang et al., 2006a,b;Zhong et al., 2005). Additionally circulating levels of apelin are decreased in patients with essential (Sonmez et al., 2010) and pulmonary (Chandra et al., 2011) hypertension, while there is a negative correlation between plasma apelin levels and blood pressure (Zhu et al., 2013). This suggests a role for decreased peripheral apelin signalling in the pathophysiology of hypertension. Receptor trafficking is a key process for regulating receptor signalling pathways and cellular functions, however in the case of APJ the mechanisms and proteins involved in agonistinduced trafficking are not well understood. To further understand the signalling and regulation of APJ, and thus the efficacy of ligands for potential therapeutic intervention, this study set out to characterize the mechanisms underlying [Pyr 1 ]apelin-13-induced APJ desensitization and internalization, and to determine whether agonist-induced APJ internalization contributes to its functional desensitization. A stable HEK-293 cell line expressing Nterminus HA-tagged mouse APJ (mAPJ) was generated, and a semiautomated imaging protocol was used to quantitate ERK1/2 activation and APJ trafficking in this cell line following agonist activation with [Pyr 1 ]apelin-13. The mechanisms of [Pyr 1 ]apelin-13induced internalization were further explored using dominantnegative mutant (DNM) cDNA constructs of GRK2 (GRK DNM ), b-arrestin1 (bARR DNM ), EPS15 (EPS DNM ) and dynamin (DYN DNM ), known effectors of CME. Anti-HA antibody was from Cambridge Bioscience (Cambridge, UK); rabbit anti-ERL1/2 antibody was from Cell Signalling Technology UK; Hercules II Fusion DNA polymerase was from Agilent Technologies (Stockport, UK), Nanofectamin was purchased from PAA Laboratories (Somerset, UK), and 4 0 ,6-diamidino-2phenyindole (DAPI), adrenaline, EGF and mouse anti-ppERK1/2 antibody were from Sigma-Aldrich (Dorset, UK). [Pyr 1 ]apelin-13 was purchased from Bachem (Bubendorf, Switzerland). Pertussis toxin (PTX), bisindolylmaleimide I (BIM) and 1,4-diamino-2,3dicyano-1,4-bis[2-aminophenylthio] butadiene (U0126) were from Merck Chemicals (Nottingham, UK).

Methods and materials
HEK293 cells, unless otherwise stated, were cultured in 10% FCSsupplemented DMEM containing glutamine (4 mM) and P/S (500 units/ml; 0.5 mg/ml). Cultures were maintained at 37 C in 5% CO 2 . For imaging studies cells were seeded at 17,500 per well into Costar black-walled 96-well plates (Corning, Arlington, UK).

Stable and transient transfection
Untagged and HA-tagged mouse (m)APJ cDNAs were generated by PCR using 150 ng mouse 129SV genomic DNA (PCR conditions: 95 C 2 min; 40 cycles of: 94 C 45 s, 50 C 1 min, 72 C 1 min; and final extension of 72 C 10 min) using Hercules II Fusion DNA polymerase. The integrity of the cDNA constructs was verified by DNA sequencing. Primers for the untagged receptor were directed to 5 0 and 3 0 -regions of mAPJ and corresponded to 8462e10,285 bp of the mouse APJ gene (Genbank Accession number AC117228.2), generating a 1824 bp product. Primers for the tagged receptor were also directed to 5 0 and 3 0 regions of the receptor, but the 5 0 primer contained an additional 27 bp, which coded for the Influenza HA epitope tag and generated a 1851 bp mouse product. The mAPJ gene, in the pcDNA3.1(þ) vector (containing the neomycin resistance gene), was transfected into HEK293 cells by a calcium phosphate procedure (Chen and Okayama, 1988) and selected by G418. Stable cell lines highly expressing APJ were selected by Northern dot blot hybridization. Transient transfection of DNM cDNAs (0.4 mg/well) was performed with Nanofectamin according to the manufacturer's protocol, with DNM cDNA-containing medium removed after 4 h and replaced with fresh DMEM (0.1% FCS). Alongside each transfection, transfection efficiency was estimated using a 5-bromo-4-chloro-3-indolyl-b-D-galactopyranoside (X-gal) staining assay. Control cells were transfected with a mammalian vector inserted with a LacZ gene (pSV-b-Galactosidase control vector; Promega, UK), and subsequent beta-galactosidase (b-gal) activity estimated from the percentage of blue cells. Approximately 40% transfection efficiency was observed with Nanofectamin with HEK293 cells, that did not deviate significantly between experiments.

Receptor imaging studies
The APJ of HA-mAPJ-HEK293 cells have exofacial HA tags enabling cell surface receptor expression to be quantified with anti-HA antibody added to non-permeabilized cells. Cells were incubated in the presence or absence of [Pyr 1 ]apelin-13 in DMEM (0.1% FCS), washed with ice-cold PBS and incubated with mouse anti-HA primary antibody (1:1000 dilution; 1 h). Whole cell APJ levels were also measured by immunohistochemistry but in this case the cells were permeabilized before addition of the primary antibody.
For quantification of HA-mAPJ recovery, HA-mAPJ-HEK293 cells were incubated in the presence or absence of [Pyr 1 ]apelin-13 for 2 h, washed, then incubated in fresh medium as indicated in figure legends, before determination of either cell surface or whole cell APJ levels using anti-HA antibody. This 2 h time point is consistent with that used previously to promote APJ internalization (Masri et al., 2006). APJ internalization was measured by labelling cell surface HA-mAPJ with primary antibody and then washing to remove unbound anti-HA antibody before stimulation with agonist. For one series of experiments clathrin-mediated internalization was blocked with hypertonic sucrose. In this case cells were incubated in physiological salt solution (NaCl (127 nM), NaH 2 PO 4 H 2 O (0.5 mM), CaCl 2 2H 2 O (1.8 mM), MgCl 2 (2 mM), KCl 2 (5 mM), NaHCO 3 (5 mM), HEPES (10 mM), BSA (0.1%) glucose (10 mM), pH 7.4) with or without 0.4 M sucrose for 30 min prior to agonist stimulation, and washing in ice-cold PBS.
To measure recycling of internalized APJ to the cell surface, HA-mAPJ-HEK293 cells were incubated with primary antibody, washed with PBS and incubated in the presence or absence of [Pyr 1 ]apelin-13 for 2 h. After aspiration of the agonist containing medium and two washes with PBS, cells were incubated with fresh medium as indicated in the figure legends.

ERK phosphorylation assay
Cell expression of total (tERK) and di-phosphorylated ERK (ppERK) was visualized in stably transfected HEK293 cells with an immunocytochemistry protocol employing anti-tERK and -ppERK antibodies. Quantification of ERK phosphorylation was performed by incubating mAPJ-HEK293 cells at 37 C with [Pyr 1 ]apelin-13 (100 nM) in DMEM (0.1% FCS) for 5 min. To explore homologous and heterologous desensitization mAPJ-HEK293 cells were preincubated for 2 h with medium in the presence or absence of [Pyr 1 ] apelin-13. This time point has been used in previous studies on desensitization of APJ (Masri et al., 2006). Cells were then washed (Â2) with PBS and exposed either to a second application of [Pyr 1 ] apelin-13 (100 nM, 5 min) or to other ERK inducers, (adrenaline (1 mM), EGF (100 ng/ml), 5 min). Resensitization was monitored by varying the period between primary and secondary agonist incubation. For assays with DNM cDNAs, mAPJ-HEK293 cells were transiently transfected with DNM cDNAs before incubation with [Pyr 1 ]apelin-13.

Semi-automated image acquisition and analysis
Assays were quantified by semi-automated acquisition of digital fluorescence images using a high content imaging platform (IN Cell Analyzer 1000, GE Healthcare UK) and validated algorithms for image segmentation and quantification (IN Cell Analyzer version 1.0 software) as described (Finch et al., 2008). Digital images were taken with a 10Â objective (Plan Apochromat, numerical aperture 0.45), with excitation and emission filters for each channel as follows, blue (360 ± 40 nm; 460 ± 40 nm), green (475 ± 20 nm; 535 ± 50 nm), and red (535 ± 50 nm; 620 ± 60 nm) using a 61002 trichroic mirror. Four fields were acquired per well (each field capturing a 0.602 mm 2 area with a 10Â objective), obtaining on average of 1000 cells per well.
For most experiments (cell surface or whole cell HA-mAPJ measures, and whole cell ppERK measures) image analysis software (In Cell 1000 Multi-target Analysis) was used to define the perimeter of the nucleus (from the DAPI stain) and the perimeter of the cell (from the HA or ppERK stain). Average fluorescence intensity over the entire cell area was calculated for each cell and background values (obtained with no primary antibody) were also determined. The figures show background subtracted and population averaged data in arbitrary fluorescence units (AFU). In most cases these are expressed as percentage of a vehicle control and for some experiments proportional cell surface expression was also calculated (PCSE; (cell surface expression ÷ whole cell expression) Â 100). In receptor internalization assays, agonist exposure caused the internalized receptors to redistribute into punctate regions (presumably endosomes) in the cytoplasm and these "inclusions" were quantified using a Dual Area Analysis Algorithm (In Cell Analyzer version 1.0). The nuclear perimeter was determined from the DAPI stain and this was expanded with a 2 mm collar. The image analysis gave the number of inclusion over the collar and nucleus for each cell and figures show population averaged inclusion counts. The agonist-induced appearance of the antibody in puncta is consistent with the wealth of data showing agonist-induced internalization of these and other GPCRs. We have previously used this methodology to investigate agonist-induced internalization of gonadotropin-releasing hormone receptors (Finch et al., 2009).

Statistical analysis
IN Cell Analyser 1000 experiments were performed in 3 replicate wells with triplicate fields within each well, and experiments were performed at least 3 times. Data are expressed in figures as mean ± SEM. Statistical analysis was with a one-way ANOVA and post hoc Dunnett's test with GraphPad Prism software (version 4.0b) (as detailed in figure legends). p < 0.05 was considered as statistically significant.

Imaging of HA-mAPJ in HEK293 cells and the ppERK response to [Pyr 1 ]apelin-13
To facilitate functional characterization of APJ, a stable HA-mAPJ expressing cell line was generated. In the first experiments receptor expression was confirmed by immunohistochemical detection of the HA tags using automated image acquisition and analysis. As anticipated, essentially all cells expressed HA-mAPJ, that could be detected in permeabilized cells and also when the primary antibody was added to bind the exofacial HA-tag in intact cells (Fig. 1A). HA epitope did not interfere with receptor signalling. Non tagged mAPJ-HEK293 cells and HA-tagged mAPJ-HEK293 cells were stimulated with [Pyr 1 ]apelin-13 (100 nM) for 5 min and compared with control cells treated with 1Â PBS. For (CeF) cells were fixed, stained, and imaged for determination of whole-cell ppERK1/2 intensity using anti ppERK1/2 antibody. The value determined with no primary antibody present was designated as background and was subtracted from raw data to give arbitrary fluorescence units (AFU) and then normalized to a percentage of vehicle control. Data shown are mean ± SEM, of at least three separate experiments, each with triplicate wells and triplicate fields within wells. *p < 0.05, **p < 0.01, and ***p < 0.001 comparing stimulations to basal conditions, analysed by one-way ANOVA and Dunnett's multiple comparison post hoc tests. ns ¼ no statistical significant difference.
We used the pyroglutamyl form of apelin-13, [Pyr 1 ]apelin-13, the most potent and abundant form in the brain  and cardiovascular system (Maguire et al., 2009), to test for expression of functional receptors and found that 5 min stimulation with 100 nM [Pyr 1 ]apelin-13 caused a marked increase in ppERK staining over the cytoplasm and nucleus of mAPJ-HEK293 cells (Fig. 1B). This effect was prevented by pre-treatment with PTX to prevent G i activation; with BIM to prevent PKC activation; or with U0126 to inhibit MEK (Fig. 1CeE). The effects of [Pyr 1 ]apelin-13 on ppERK levels in HEK-293 cells expressing non-tagged mAPJ and HA-tagged mAPJ were also compared and were found to be indistinguishable (Fig. 1F).
mAPJ-HEK293 cells were then treated for varied times with [Pyr 1 ]apelin-13. The ppERK response was rapid (maximal at 5 min) and transient, reducing to near basal values by 10 min (Fig. 2A). We also varied [Pyr 1 ]apelin-13 concentration and this revealed a concentration-dependent effect with an EC 50 value of 3 nM at 5 min (Fig. 2B). No significant variations were seen in

Trafficking of HA-mAPJ
Following the lack of heterologous desensitization described above, that implies that the desensitization of the response to [Pyr 1 ]apelin-13 may be due to upstream APJ-specific (rather than down-stream ERK-specific) adaptive mechanisms, we explored possible changes in the amount and compartmentalization of APJ by stimulating HA-mAPJ-HEK293 cells for varied periods (up to 6 h) with 0 or 100 nM [Pyr 1 ]apelin-13 before determining cell surface and whole cell HA-mAPJ levels with the intact cell and permeabilized cell staining assays used for Fig. 1. As shown (Fig. 4A), [Pyr 1 ]apelin-13 caused a reduction in cell surface HA-mAPJ, which reduced by >50% with a half-time of~30 min. It also reduced whole cell HA-mAPJ (Fig. 4B) but the effect was less marked (reduction tõ 60% of control) and slower (no measurable reduction until 1 h). We also used the cell surface and whole cell HA-mAPJ expression  measures to calculate the proportional cell surface receptor expression (PCSE; (cell surface expression ÷ whole cell expression) Â 100) and found that in control cells~76% of HA-mAPJ were at the cell surface and that this reduced to~40% after 30 min stimulation with [Pyr 1 ]apelin-13 before recovering to near control levels at 6 h (Fig. 4C).

HA-mAPJ internalization and desensitization of APJ-mediated ERK activation
To follow internalization more directly cell surface HA-mAPJ were preloaded with anti-HA antibody in the absence of agonist, and cells were then incubated for varied periods with 0 or 100 nM [Pyr 1 ]apelin-13 before determining the number of anti-HAcontaining inclusions (presumptive endosomes) by automated image analysis. As shown ( Fig. 5A and B), [Pyr 1 ]apelin-13 caused a rapid increase with the inclusion count being maximal after 30 min and remaining significantly elevated for 6 h.
This assay was also used to explore APJ internalization mechanisms using a 2 h [Pyr 1 ]apelin-13 stimulation period. This revealed that pretreatment with hypertonic sucrose to block clathrinmediated endocytosis completely blocked the [Pyr 1 ]apelin-13 effect on inclusion counts (Fig. 6A). The dose-dependent effects of GRK, EPS, DYN and bARR DNMs on HA-mAPJ-HEK293 cells was then assessed. Co-transfection with the individual expression vectors for GRK DNM , EPS DNM and DYN DNM inhibited [Pyr 1 ]apelin-13induced HA-mAPJ internalization in a dose-dependent manner, with an optimal concentration of 0.4 mg/well, however the [Pyr 1 ] apelin-13-stimulated increase in inclusion count was not blocked by bARR DNM (Fig. 6B). The effects of GRK DNM , EPS DNM , DYN DNM and   Fig. 6C.
We also tested for effects of GRK DNM and DYN DNM cDNAs, both of which prevented HA-mAPJ internalization into inclusions (Fig. 6C), on the desensitization of APJ-mediated ERK activation. Acute (5 min) stimulation of mAPJ-HEK293 cells with [Pyr 1 ]apelin-13 caused robust increases in ppERK, that did not alter in cells transfected with GRK DNM or DYN DNM cDNAs (Fig. 6D). Pre-treatment for 2 h with 100 nM [Pyr 1 ]apelin-13 caused the expected reduction of subsequent responses to 5 min stimulation with 100 nM [Pyr 1 ] apelin-13 in control cells (Fig. 6D, see also Fig. 3) and this reduction was also observed in cells transfected with GRK DNM or DYN DNM cDNAs (Fig. 6D).

Recovery of APJ levels after agonist removal
To explore recovery of APJ expression levels following pretreatment with agonist, HA-mAPJ-HEK293 cells were treated for 2 h with 0 or 100 nM [Pyr 1 ]apelin-13, washed and allowed to recover for varied periods (0e6 h) before quantification of cell surface HA-mAPJ and whole cell HA-mAPJ levels. As expected, the [Pyr 1 ]apelin-13 pre-treatment reduced cell surface and whole cell HA-mAPJ levels by 40e50% (Fig. 7A and B; see also Fig. 4). Cell surface HA-mAPJ levels recovered slowly returning to control levels at 4e6 h after the pre-treatment (Fig. 7A), whereas whole cell HA-mAPJ levels remained low and were essentially unaltered during the 0e6 h recovery period (Fig. 7B). These data were used to calculate PCSE and this was reduced (from an initial~76% to~60%) by apelin-13 pre-treatment and recovered to almost 100% at 2e6 h after pre-treatment (Fig. 7C). A similar protocol was used to assess recovery from the effect of [Pyr 1 ]apelin-13 on HA-mAPJ inclusion count. As expected, pre-treatment for 2 h with [Pyr 1 ]apelin-13 increased the number of inclusions by~75% and this effect was rapidly reversed so that there was no measurable increase in inclusions after 30 min of recovery (Fig. 7D).

Resensitization of APJ-mediated ERK activation
We then followed recovery from desensitization (of [Pyr 1 ]apelin-13-stimulated ERK activation) in control cells and in cells transfected with GRK DNM or DYN DNM cDNAs, both of which prevented HA-mAPJ internalization into inclusions (see Fig. 6), or bARR DNM cDNA. mAPJ-HEK293 cells initially exposed to vehicle control (1Â PBS; 2 h) showed significant activation of ERK1/2 after a 5 min exposure to [Pyr 1 ]apelin-13. However 2 h pre-treatment of mAPJ-HEK293 cells with 100 nM [Pyr 1 ]apelin-13 caused the expected reduction in response to a subsequent 5 min stimulation with 100 nM [Pyr 1 ]apelin-13 (Fig. 8A). When cells were allowed to recover for varied periods (0e1 h) before the second stimulus, rapid recovery was observed, with maximal recovery and no measurable desensitization after just 15 min of recovery (Fig. 8A). Recovery was slower in the presence of GRK DNM (Fig. 8B) or DYN DNM cDNAs (Fig. 8C), as for both there was no measurable recovery at 15 min and recovery was near maximal at 1 h. The presence of bARR DNM did not alter the pattern of normal resensitization of [Pyr 1 ]apelin-13-induced ERK1/2 activation.

Discussion
GPCR regulation in response to agonist stimulation is common to nearly all GPCRs and is essential in physiological systems to limit persistent signalling. In this study we have investigated the [Pyr 1 ] apelin-13-induced trafficking and desensitization of mAPJ in mAPJ-HEK293 cells using a semi-automated imaging protocol and clearly  , 2 h), washed, incubated in fresh medium for 0e1 h and then stimulated in the presence or absence of [Pyr 1 ]apelin-13 (100 nM) for 5 min. Cells were fixed, stained and imaged for determination of whole cell ppERK1/2 intensity using anti-ppERK1/2 antibody. The value determined with no primary antibody present was designated as background and was subtracted from raw data to give ppERK1/2 intensity in arbitrary fluorescence units (AFU) and then normalized to a percentage of vehicle control ([Pyr 1 ]apelin-13-induced ERK1/2 signalling in cells initially exposed to vehicle control and designated as "max"). In (BeD) mAPJ HEK293 cell lines were transfected with GRK DNM , DYN DNM or bARR DNM cDNAs respectively before preincubation with or without [Pyr 1 ]apelin-13. Data shown are mean ± SEM, of at least three separate experiments, each with triplicate wells and triplicate fields within wells. ***p < 0.001, comparing stimulations to max conditions, analysed by one-way ANOVA and Dunnett's multiple comparison post hoc tests.
A stable HA-mAPJ expressing cell line was generated and was used to quantify the proportion of APJ at the cell surface and within whole cells using semi-automated acquisition and analysis of digital fluorescence images. While the majority of epitope-tagged mAPJ was localized to the cell surface in these cells, a proportion of tagged APJ was distributed within the cell. This is in contrast to earlier studies that reported enhanced green fluorescent protein (eGFP)-APJ localization, under basal conditions, to be confined to the plasma membrane . APJ acts primarily via G i to inhibit adenylyl cyclase but has also been reported to activate other effectors including PKC, PI3K and ERK (Masri et al., 2002(Masri et al., , 2004. As positioning of differing tags into the native receptor may have implications for receptor trafficking, we verified that the functional integrity of the receptor in our cell line remained intact and that these HA-mAPJ-HEK293 cells, like their non-tagged counterparts, mediate ERK activation. Significant and similar [Pyr 1 ]apelin-13-induced stimulation of ERK1/2 was seen in both HA-tagged and untagged mAPJ transfected HEK293 cells. Many GPCRs show ligand bias (where different agonists bias signalling toward different effectors) and there is evidence that this may occur for APJ (Masri et al., 2006;Brame et al., 2015). It has been shown recently that the cyclic apelin analogue MM07 displays bias towards stimulation of a beneficial G-protein-dependent pathway, stimulating vasodilation and inotropic actions, over a more damaging G-protein-independent b-arrestin-dependent pathway that results in cardiac hypertrophy (Brame et al., 2015). In this regard, it is also of interest that APJ is most closely related to angiotensin 1 receptors (AT 1 ), for which ligand bias has been extensively explored. AT 1A receptors are G q/11 coupled GPCRs that also activate ERK. They undergo a process of rapid homologous receptor desensitization in which arrestins bind to the activated receptors preventing them from activating their cognate G proteins and targeting them for internalization via CCVs. The arrestins can also act as scaffolds for MAPK cascade components and mediate signalling to ERK. Activation of AT 1A receptors can cause an initial phase of G protein-mediated ERK activation followed by a switch to a second phase of arrestin-mediated ERK activation and ligand bias is seen when angiotensin II activates both pathways whereas analogues (such as [Sar(1),Ile(4),Ile(8)]AngII (SII)) engage only the latter (Lefkowitz and Shenoy, 2005;Ahn et al., 2004;Shenoy et al., 2006). We were interested in the possibility that APJ might also mediate such a biphasic response. We established however that when mAPJ-HEK293 cells were treated for varied times with [Pyr 1 ]apelin-13, the ppERK response instead was rapid and transient, with an EC 50 value of~3 nM at 5 min, showing no indication of arrestinmediated activation of ERK1/2.
Having established that the ppERK response to [Pyr 1 ]apelin-13 desensitizes rapidly during sustained stimulation in this model, we explored possible mechanisms. Numerous adaptive mechanisms shape ERK responses and these include inhibitory phosphorylation of Ras by ERK (Dumaz and Marais, 2005) or ERK-driven expression of nuclear-inducible dual specificity phosphatases (DUSP) . However, neither of these down-stream mechanisms seems likely here as APJ-mediated ERK activation is Rasindependent in CHO cells (Masri et al., 2002) and the desensitization is too fast to be mediated by DUSP neosynthesis . We therefore suspected that this rapid desensitization of the response to [Pyr 1 ]apelin-13 was due to upstream APJ-specific (rather than down-stream ERK-specific) adaptive mechanisms and explored this by looking at homologous and heterologous desensitization of this response. Desensitization can be homologous or heterologous in nature; homologous desensitization occurs when there is a loss of response solely to an agonist that is acting at one particular GPCR subtype, whereas heterologous desensitization is agonist-non-specific and involves a broad pattern of unresponsiveness at multiple GPCR subtypes. Homologous desensitization is thought to involve adaptive changes at the level of the GPCR itself, whereas heterologous desensitization may also involve altering the efficiency of downstream signalling components. Adrenaline and EGF are known activators of ERK in HEK293 cells, likely acting via the a 1b adrenergic (Schonbrunn and Steffen, 2012) and EGF receptors (Kramer et al., 2002) respectively, and as expected, both caused robust increases in ppERK in PBS pre-incubated mAPJ-HEK293 cells, that were not different to the increases seen after pre-incubation with [Pyr 1 ]apelin-13. The ppERK response to a subsequent [Pyr 1 ]apelin-13 stimulation was however completely abrogated by pre-incubation with [Pyr 1 ]apelin-13. These data suggest that the desensitization of [Pyr 1 ]apelin-13-induced ERK1/2 phosphorylation was not due to a requirement to reset the intracellular signalling pathway or other post-receptor modifications, but to upstream APJ-specific adaptive changes. These could include receptor internalization, as APJ undergoes agonist-induced internalization via CCVs (Reaux et al., 2001;El Messari et al., 2004), and/ or rapid homologous receptor desensitization, as APJ has been shown to cause translocation of b-arrestin to the cell surface in other systems (Lee et al., 2010). We undertook therefore to monitor APJ compartmentalization, and found that incubation of HA-mAPJ-HEK293 cells with [Pyr 1 ]apelin-13 decreased both cell surface and whole cell expression of APJ in a time-dependent manner. These data are consistent with agonist-induced receptor internalization followed by degradation of a proportion of the internalized receptors, such that down-regulation follows the reduction in cell surface expression.
As the experimental procedure used could reflect agoniststimulation of both anterograde and retrograde APJ trafficking, as has been described for the d-opioid peptide receptor (Zhang et al., 2006a,b), receptor internalization was more directly monitored by loading cell surface HA-mAPJ with anti-HA antibody before washing and incubation with agonist. We found a rapid increase in [Pyr 1 ]apelin-13-induced HA-mAPJ internalization that is consistent with previous confocal microscopy studies showing rapid agonistinduced internalization of eGFP-APJ from the plasma membrane (Lee et al., 2010;El Messari et al., 2004). [Pyr 1 ]apelin-13-induced-HA-mAPJ internalization was inhibited in the presence of sucrose, which prevents formation of clathrin-coated pits (Heuser and Anderson, 1989), and by the expression of DNM cDNAs of GRK2, dynamin and EPS15, known effectors of CME, but not by expression of bARR DNM . We have used these DNMs successfully to assess GRK2-, b-arrestin1-, dynamin-and EPS-dependent internalization of several GPCRs (e.g. oxytocin receptor (Smith et al., 2006); gonadotropin-releasing hormone receptor (Hislop et al., 2001(Hislop et al., , 2005). The GRK2 DNM construct (K220A) reduces agonist-induced GPCR phosphorylation (Mundell et al., 1997); bARR DNM (319e418) competes with wild-type arrestin for clathrin and AP2 binding, and impairs receptor-binding ability (Krupnick et al., 1997); EPS DNM (ED95/295) lacks domains recognising EPS15 itself and is required for coated pit formation; and DYN DNM (K44A) inhibits dynaminmediated scission of CCVs from the plasma membrane (Damke et al., 1994). Together these data suggest that [Pyr 1 ]apelin-13 causes internalization of mAPJ via CCVs, and thereby reduces cell surface mAPJ levels. In accord with many other GPCRs, we suggest that GRK2-mediated phosphorylation targets the receptors to CCVs that are internalized from the cell surface in an EPS15-and dynamin-dependent manner. Interestingly, apelin-13-internalized APJ has been reported to dissociate from b-arrestin1 prior to receptor internalization (Lee et al., 2010), and we have shown that the [Pyr 1 ]apelin-13 effect on inclusion count was not prevented by transfection with the bARR DNM cDNA. Although not tested, it is likely that both bÀarrestin1and bÀarrestin2-mediated endocytosis would be inhibited by transfection of the bARR DNM construct (J.L. Benovic, personal communication). Further work such as using siRNAs specifically against each arrestin will determine whether either is necessary for APJ-mediated internalization. Therefore the means by which these receptors are targeted to CCVs for internalization remains unclear. A similar dynamin-dependent, b-arrestin independent internalization has been reported in the 5hydroxytryptamine 2A (5-HT 2A ) receptor (Bhatnagar et al., 2001).
Desensitization of APJ has been shown not to occur in Cterminally truncated receptors that lack the majority of the serine and threonine residues that are liable to phosphorylation (Masri et al., 2006). Phosphorylation of GPCRs following ligand activation is the first step in receptor desensitization, occurring rapidly upon exposure to the agonist, and is conducted by second messenger kinases (e.g. protein kinase C (Benovic et al., 1985)) and a family of kinases termed GRKs (Benovic et al., 1986). Transfection of GRK2 DNM or DYN DNM cDNA constructs into mAPJ-HEK293 cells pretreated with [Pyr 1 ]apelin-13 for 2 h, both of which constructs had prevented HA-mAPJ internalization, did not affect the abrogation of subsequent responses to stimulation with [Pyr 1 ]apelin-13. This data suggests that internalization is not a major factor in the desensitization of APJ-mediated ERK activation in these cells, and that desensitization of APJ is not dependent upon phosphorylation of APJ by GRK-2, but likely by kinases other than GRK-2.
Focusing on recovery after agonist removal, pre-incubation of HA-mAPJ-HEK293 cells with [Pyr 1 ]apelin-13 and subsequent recovery in agonist-free medium for varied periods revealed that cell surface APJ levels recovered to near control levels at 4e6 h, while APJ levels in the whole cell remained low following up to 6 h of recovery. Similarly, recovery in agonist-free medium rapidly reversed the number of inclusions seen after pre-incubation with [Pyr 1 ]apelin-13 for 2 h. These data are generally consistent with agonist-induced internalization and reduction in cell surface HA-mAPJ occurring relatively rapidly during agonist exposure and recovering more slowly after agonist removal. They also reveal two unexpected features. First, the proportion of HA-mAPJ at the cell surface 2 h after removal of the [Pyr 1 ]apelin-13 pre-treatment is significantly higher than in control cells prior to [Pyr 1 ]apelin-13 pre-treatment (~98% versus~76%), suggesting that all available APJ are specifically compartmentalized at the cell surface upon agonist removal. Second, the rate of loss of inclusions during the recovery period (half-time~15 min) was much higher than the rate of recovery of cell surface HA-mAPJ (half-time~2 h). This may suggest that inclusions are lost as the anti-HA shifts from early endosomes to a sorting compartment. If so, their loss would be expected to precede recovery of cell surface receptor expression. In following recovery from desensitization we transfected GRK2 DNM , DYN DNM and bARR DNM cDNAs into mAPJ-HEK293 cells and found delayed resensitization of [Pyr 1 ]apelin-13-induced ERK1/2 responses after transfection with GRK2 DNM and DYN DNM cDNAs, but not with the bARR DNM cDNA. This indicates that GRK2-and dynamin-dependent receptor internalization may have a role to play in the resensitization of [Pyr 1 ]apelin-13-induced ERK phosphorylation, however bÀarrestin-dependent internalization is not required for this process. Thus regulation of receptor number may determine responsiveness of mAPJ to repeated [Pyr 1 ]apelin-13 stimulation in HEK293 cells.
In summary, these data show that agonist exposure induces internalization and reduction in cell surface HA-mAPJ expression that occurs relatively rapidly during agonist exposure and recovers more slowly after agonist removal. Moreover the ppERK response of mAPJ-HEK293 cells to [Pyr 1 ]apelin-13 desensitizes rapidly during sustained stimulation and this desensitization is due to upstream APJ-specific, rather than down-stream ERK-specific, adaptive changes. We show that GRK2-mediated phosphorylation targets mAPJ to CCVs that are internalized from the cell surface in a bÀarrestin-independent, EPS15-and dynamin-dependent manner.
Our data indicates that receptor internalization is not required for mAPJ desensitization of ppERK responses in the mAPJ-HEK293 cell line. This multifaceted system may be indicative of a complex mechanism in controlling the physiological functions of endogenous apelin and may be important in conditions where there are elevated circulating or tissue levels of APJ. Further study of the signalling and regulation of APJ will help develop ligands for use in potential therapeutic intervention for dysfunctions of physiological homeostasis such as hypertensive disease.