The Apelin Receptor Is Coupled to Gi1 or Gi2 Protein and Is Differentially Desensitized by Apelin Fragments*

The apelin receptor is a G protein-coupled receptor to which two ligand fragments, apelin-(65–77) and apelin-(42–77), can bind. To address the physiological significance of the existence of dual ligands for a single receptor, we first compared the ability of the apelin fragments to regulate intracellular effectors, to promote G protein coupling, and to desensitize the response in Chinese hamster ovary cells expressing the murine apelin receptor. We found that both apelin fragments inhibited adenylyl cyclase and increased the phosphorylation of ERK or Akt. Using stably transfected cells expressing a pertussis toxin-insensitive αi subunit, we demonstrated that each apelin fragment promoted coupling of the apelin receptor to either Gαi1 or Gαi2 but not to Gαi3. Although preincubation with each apelin fragment induced a desensitization at the level of the three effectors, preincubation with apelin-(42–77) also increased basal effector activity. In addition, a C-terminal deletion of the apelin receptor decreased the desensitization induced by apelin-(65–77) but did not alter the desensitization pattern induced by apelin-(42–77). Finally, in umbilical endothelial cells, which we have recently shown to express the apelin receptor, the Gαi1 and Gαi2 subunits are also expressed, ERK and Akt phosphorylation is desensitized after preincubation with apelin-(65–77), and basal levels of Akt phosphorylation are increased after preincubation with apelin-(42–77). In summary, apelin fragments regulate the same effectors, via the preferential coupling of the apelin receptor to Gi1 or Gi2, but they promote a differential desensitization pattern that may be central to their respective physiological roles.

The recently discovered apelin signaling pathway plays a role in the central and peripheral regulation of the cardiovascular system, in water and food intake, and possibly in immune function (for a review, see Ref. 1). The apelin receptor was first identified in human (2) and amphibians (3), and the murine receptor was later cloned from embryonic tissues (4). The deduced protein sequence of these three orthologs revealed that the apelin receptor is a member of the G protein-coupled, seventransmembrane domain receptor family and displays a structural relationship with angiotensin receptors (2) and CXC chemokine receptors (3,5). However, angiotensin II did not bind to it, and the receptor remained orphan until the identification of apelin, its endogenous ligand (6). Cloning of the apelin gene showed that it codes for a preproprotein of 77 residues containing a signal peptide, which, after proteolytic maturation, generates the apelin fragment of 36 amino acids, apelin-(42-77), first isolated from stomach extracts (6). Although the existence of several endogenous apelin fragments was inferred from the presence of conserved basic doublets in the protein sequence, their physiological occurrence has been validated by the characterization of two active peaks from stomach extracts (6) and by gel filtration chromatography of colostrums (7). In addition, a different tissue distribution of a short and a long apelin fragment was clearly demonstrated (8). The form expressed in the lung was clearly eluted at the position corresponding to apelin-(42-77), whereas the other immunoreactive peak of apelin from the mammary gland is likely to correspond to apelin-(65-77).
Expression of the receptor was primarily observed in brain tissues (2), where the mRNA and the protein were later located in neurons (9,10) as well as in oligodendrocytes and astrocytes (11). Outside the brain, the transcripts coding for the murine apelin receptor were also detected in the endothelium of the primary vessels during embryonic development (4) and that of the retinal vessels during postnatal development (12); expression of the receptor protein was deduced in human heart and saphenous vein by the demonstration there of apelin binding sites (13). More recently, the expression of the receptor was also characterized by immunocytochemistry in vascular smooth muscle cells and cardiomyocytes (14).
Localization of receptor expression may provide a clue to its physiological function, and the various sites of tissue expression correlate well with the observed effects of exogenous apelin. First, the high density of receptor transcripts in hypothalamus (15) is consonant with the diuretic effect of apelin that results from decreased release of vasopressin from the hypothalamic nuclei (16). Second, the endothelial expression of the receptor (4) explains the hypotensive activity of apelin (17) as activation of the receptor leads to nitric oxide production by the endothelial cell (18), which, in turn, induces the relaxation of the smooth muscle cell. Finally, the immunochemical detection of the apelin receptor on cardiomyocytes (14) is in accordance with the very strong inotropic activity of apelin peptide (19 -21).
However, the intracellular transduction cascades that provide the molecular link between activation of the apelin receptor by apelin fragments and their biological effects at the cellular level remain to be clarified. An initial study performed on Chinese hamster ovary (CHO) 3 cells stably transfected with the human apelin receptor revealed that apelin peptides elicit a concentration-dependent inhibition of forskolin-stimulated production of cAMP (22). It was later determined that apelin-(42-77) and apelin-(65-77) produce a sharp rise in intracellular calcium concentrations in human NT2.N neurons (10). In CHO cells expressing the murine apelin receptor, we showed that apelin-(65-77) also activates extracellular-regulated kinases (ERK) via a pertussis toxin (PTX)-* The costs of publication of this article were defrayed in part by the payment of page sensitive G protein (23). We recently demonstrated, in the same cells, that apelin-(65-77) also induces the activation of p70S6 kinase via two intracellular cascades, ERK-dependent and phosphatidylinositol 3-kinase/Akt-dependent, respectively, which are linked to the phosphorylation of different subsets of threonine or serine residues (24). Accordingly, the regulation of intracellular effectors has been principally analyzed after stimulation of apelin receptor by the short fragment, apelin-(65-77). The existence of a single receptor and different fragments of the endogenous ligand raised the question of their ability to promote differential G protein coupling, to activate distinct transduction cascades, and to undergo differential regulation.
In the study described here, we have compared the regulation of the previously mentioned intracellular effectors by each of the two apelin fragments, apelin-(42-77) and apelin-(65-77), determined the nature of the G i protein coupled to the apelin receptor after binding of each fragment, and analyzed the desensitization of the response after receptor stimulation by the two peptides. Our results reveal that although both apelin-(42-77) and apelin-(65-77) can activate the same set of intracellular effectors, they display some differences in their G i protein coupling and differ strongly in their desensitization pattern.
Subcloning of the PTX-insensitive G␣ i Subunits-The mutants G␣ i1 C351I, G␣ i2 C352I, or G␣ i3 C352I obtained from Dr. P. J. Pauwels (25) were excised from the pCR3.1 vector by digestion with HindIII and XhoI and then subcloned into the pcDNA5/FRT expression vector (Invitrogen) using the same sites.
Construction of the C-terminally Truncated Mutant of the Murine Apelin Receptor, ⌬327-377-The expression vector pREN (made by Dr. Maret), containing the entire coding region of the murine apelin receptor, was amplified and purified from the dam Ϫ G3819 bacterial strain. After digestion with BclI and ClaI, the protruding ends were filled in using the Klenow fragment of DNA polymerase and were religated by T4 DNA ligase. After bacterial transformation and purification of plasmid DNA using standard procedures, the deletion of the DNA sequence encoding the 51 C-terminal amino acids was confirmed by dideoxy sequencing of both strands using Sequenase.
Cell Culture-CHO cells stably expressing the msr/apj receptor were routinely grown in Eagle's minimum essential medium (Cambrex) supplemented with 5% fetal calf serum, 2 mM L-glutamine, and 300 g/ml Geneticin as described previously (23). CHO cells expressing the PTXinsensitive G␣ i subunits were grown in the presence of 300 g/ml hygromycin B (Invitrogen).
Primary cultures of human umbilical vein endothelial cells (HUVEC) were prepared by digestion with 0.1% type I collagenase (Invitrogen) from human umbilical cords, provided by donors in compliance with French legislation. They were cultured at 37°C in 5% CO 2 humidified air on fibronectin-coated culture dishes in M199 medium (Invitrogen) containing 10% fetal calf serum, 2 mM L-glutamine, 100 units/ml penicillin, 100 g/ml streptomycin, 18 units/ml heparin, 2 ng/ml basic fibroblast growth factor, 1 mg/ml hydrocortisone, and 1 ng/ml epidermal growth factor. HUVEC were used at passage 2.
Generation of Stable CHO Cell Lines Co-expressing the Murine Apelin Receptor and Pertussis Toxin-resistant G␣ i1 , G␣ i2 , or G␣ i3 Subunit-The Flp-In system was used to generate isogenic, stable mammalian cell lines expressing each PTX-insensitive (PTX-r) G␣ i subunit from CHO cells constitutively expressing the murine apelin receptor (23). Stable transfectants were obtained, as recommended by the manufacturer. Briefly, the CHO host cell line was constructed by transfection with 0.5 g of the pFRT/lacZeo vector and selection with growth medium containing 500 g/ml Zeocin TM (Cayla). Zeocin TM -resistant clones were screened to identify those containing a single integrated FRT site and expressing a high level of ␤-galactosidase. A single clone was selected and amplified for further studies. Cells expressing the PTXinsensitive G␣ i protein were generated by cotransfection into the selected Zeocin TM -resistant CHO host cell line of the hygromycin-resistant pcDNA5/FRT plasmids bearing each form of G␣ i and the pOG44 plasmid that constitutively expresses the Flp recombinase. Two days later, the cells were fed with growth medium supplemented with 300 g/ml hygromycin B (Invitrogen). The clones of interest were selected for hygromycin resistance, Zeocin TM sensitivity, and the absence of ␤-galactosidase activity.
Adenylyl Cyclase Assay-CHO cells were plated at a concentration of 3.10 4 cells/well in 6-well cluster plates in the presence of 2 ml of Eagle's minimum essential medium containing 5% fetal calf serum. CHO cells expressing the PTX-insensitive G␣ i subunit were treated with 25 ng/ml pertussis toxin the day before assay. Two days later, the medium was aspirated, and the cells were washed twice with PBS. Then the PBS was replaced by 1 ml of Eagle's minimum essential medium containing 20 M forskolin and 0.5 mM 3-isobutyl-1-methylxanthine in the presence or absence of apelin fragments. After incubation of the cells for 15 min at 37°C, the medium was aspirated and replaced by 1 ml of a mixture of 95% methanol, 5% formic acid. The plate was then left on ice for 15 min. After resuspension of the lysed cells, the lysates were transferred to tubes, which were then capped, frozen, and kept at Ϫ20°C. The amount of cAMP was determined by radioimmunoassay. The dried lysates were resuspended in the diluent provided by the manufacturer (Immunotech, Beckman Coulter) and incubated overnight at 4°C in monoclonal antibody-coated tubes in the presence of 125 I-labeled tracer. After incubation, the contents of the tubes were aspirated, and bound radioactivity was measured in a gamma counter. The cAMP levels were calculated by interpolation from the calibration curve.
Immunoblotting-Cells at subconfluence were deprived of serum for 18 h. In the case of the CHO cells expressing the PTX-insensitive G␣ i subunits, the cells were treated with 25 ng/ml pertussis toxin the day before assay. After the indicated treatment, the cells were stimulated at 37°C and then washed once in cold PBS and lysed for 15 min on ice in a Hepes buffer (50 mM, pH 7.4) containing 150 mM NaCl, 100 mM NaF, 10 mM EDTA, 10 mM Na 4 P 2 O 7 , 2 mM sodium orthovanadate, 1 mM phenylmethylsulfonyl fluoride, 2 g/ml aprotinin, 20 M leupeptin, and 1% Triton. The mixture was gently agitated for 15 min at 4°C, and cell lysates were clarified by centrifugation at 13,000 ϫ g for 15 min. The protein concentration of the supernatants was determined using a Bio-Rad assay kit (Bio-Rad). Soluble proteins (30 -50 g) were fractionated on 12% SDS-polyacrylamide gels and blotted to Protran nitrocellulose membranes (Schleicher & Schuell). After blocking in TBS-T buffer (50 mM Tris-HCl, pH 7.4, 150 mM NaCl, and 0.2% Tween 20) with 5% nonfat milk for 1 h at room temperature, the membranes were incubated with blocking solution containing the indicated antibody overnight at 4°C. The membranes were washed and incubated with a horseradish peroxidase-conjugated secondary antibody, and horseradish peroxidase activity was revealed using a chemiluminescent detection system (Amersham Biosciences).
Desensitization Assay-HUVEC and CHO cells were cultured for 2 days as described for the adenylyl cyclase assay or immunoblotting. Cells expressing the PTX-insensitive G␣ i subunits were treated with 25 ng/ml pertussis toxin the day before assay. After aspiration of the medium and two washes with PBS, the cells were incubated for 2 h at 37°C in serum-free Eagle's minimum essential medium in the presence or absence of 1 M apelin-(65-77) or apelin-(42-77). The cells were then washed twice with PBS and incubated as described previously for the adenylyl cyclase assay or immunoblotting in the presence of the indicated concentrations of apelin.
Acid Washing Procedure-Following preincubation in the presence or absence of 1 M apelin-(42-77), cultures were washed twice with cold PBS and then treated on ice with a solution containing 0.2 M acetic acid and 0.5 M NaCl at pH 3 or 4. The acid washing was performed for 1 min at pH 3 and 2 min at pH 4. Cells were then washed twice with warm PBS and assayed with apelin fragments for adenylyl cyclase inhibition or Akt phosphorylation.
Data Analysis-Data are expressed as means ϩ S.E. Dose responses were plotted and analyzed (-log IC 50 determination) with GraphPad Prism software. Statistical significance between dose-response curves was determined with a two-way analysis of variance tests.
Apelin-(42-77) induced a concentration-dependent inhibition of adenylyl cyclase in adherent CHO cells (Fig. 1A). When compared with the inhibition of adenylyl cyclase by apelin-(65-77), the curve obtained with apelin-(42-77) was displaced to the right, revealing that the IC 50 value of the long fragment (9.0 ϫ 10 Ϫ10 M) is higher than that of the short fragment (3.4 ϫ 10 Ϫ10 M). This slight difference between apelin fragments in their ability to inhibit adenylyl cyclase has already been observed in CHO cells expressing the human apelin receptor (22).
Using antibodies specifically recognizing phosphorylated forms of ERK or phosphorylated Ser-473 of Akt, we observed that, like the short apelin fragment, apelin-(42-77) promotes the phosphorylation of ERK ( Fig. 1B) and Akt (Fig. 1C). The kinetics were identical for both ERK and Akt, with the peak of phosphorylation occurring at 5 min. In addition, the amplitude of the phosphorylation induced by each fragment was very similar. It is thus clear that the two fragments, apelin-(42-77) and apelin-(65-77), regulate the activity of the same effectors.
Generation of Stable Transfectant Cells Co-expressing the Apelin Receptor and a PTX-insensitive G Protein ␣ i Subunit-We next wished to ascertain whether the same G protein was involved in these different biological responses. Since the reactions in question are sensitive to PTX (22)(23)(24), we used a reconstitution model based upon the selection of cells stably co-expressing the murine apelin receptor and a G protein ␣ i subunit mutated to render it PTX-insensitive (25). We chose this approach because it is based upon the targeted insertion, at the same site, of each mutated ␣ i subunit, ensuring that they will all be expressed at the same level. This precludes the possibility of illegitimate reconstitution, which could occur if the stoichiometry between the apelin receptor and the ␣ i subunit was altered. We checked the expression of each ␣ i subunit by immunoblotting and found, as expected, the transfected protein in each clone expressing the PTX-insensitive subunit (Fig. 2). However, a faint but detectable signal was also detected with the antibody raised against the C terminus of the ␣ i3 subunit in the clones expressing  the PTX-insensitive ␣ i1 or ␣ i2 subunit; this result is due to this antibody cross-reacting with the ␣ i1 subunit and, to a lesser extent, with the ␣ i2 subunit.
Apelin Fragments Inhibit Adenylyl Cyclase by Coupling to G i1 or G i2 but Not to G i3 Protein-We first determined the extent of adenylyl cyclase inhibition elicited by different concentrations of apelin-(65-77) and apelin-(42-77) in the three clones that separately expressed the PTX-insensitive subunits ␣ i1 , ␣ i2 , or ␣ i3 . As shown in Fig. 3, we observed a significant inhibition of adenylyl cyclase activity in the clone expressing the PTX-insensitive ␣ i3 subunit only when the apelin fragments were added at high concentrations. On the other hand, both apelin fragments readily inhibited adenylyl cyclase when the PTX-insensitive ␣ i1 subunit or ␣ i2 subunit was expressed.
For apelin-(65-77), the inhibition curves obtained with the clone expressing the PTX-insensitive ␣ i1 subunit (5.9 ϫ 10 Ϫ9 M) and with that expressing the PTX-insensitive ␣ i2 subunit (5.1 ϫ 10 Ϫ9 M) were almost identical (Fig. 3A). The results were very different for apelin-(42-77). The IC 50 value was significantly lower in the clone expressing the PTXinsensitive ␣ i2 subunit (6.7 ϫ 10 Ϫ10 M) than in the clone expressing the PTX-insensitive ␣ i1 subunit (3.4 ϫ 10 Ϫ9 M), and the extent of inhibition was much higher (Fig. 3B). In addition, it should be noted that the IC 50 values for adenylyl cyclase inhibition induced by the two apelin fragments were very similar in the clone expressing the PTX-insensitive ␣ i1 subunit (Fig. 3, A and B).
Taken together, these data demonstrate that, to inhibit adenylyl cyclase, the apelin receptor preferentially couples to G i1 or G i2 , and, less efficiently, to G i3 . Secondly, regardless of the apelin fragment, the inhibition is more robust when the apelin receptor is coupled to the G i2 protein. Finally, apelin-(42-77) promotes a more efficient coupling to the G i2 protein than does apelin-(65-77), as revealed by a higher apparent affinity and a stronger degree of inhibition.
Apelin Fragments Induce ERK Phosphorylation via a Differential Coupling to G i1 and G i2 Protein-We next asked whether apelin-(65-77) and apelin-(42-77) activate ERK via a coupling identical to that described above in the case of adenylyl cyclase inhibition. Interestingly, we noticed a significant difference in the phosphorylation of ERK induced by apelin-(65-77) and by apelin-(42-77). Although ERK was activated by apelin-(42-77), both in the clone expressing the PTX-insensitive ␣ i1 subunit and in that expressing the ␣ i2 subunit (Fig. 4B), activation was more efficiently induced by apelin-(65-77) in the presence of the ␣ i2 subunit (Fig. 4A). In contrast, neither apelin fragment resulted in detectable phosphorylation of ERK in the clone expressing the PTX-insensitive ␣ i3 subunit. All these findings confirm the inability of the apelin receptor to couple to the G i3 protein and suggest once more that apelin fragments may promote a differential coupling of the receptor to a specific G i protein.
Apelin Fragments Induce Akt Phosphorylation by Coupling to G i1 or G i2 Protein-We previously showed (Fig. 1C) that apelin-(42-77) can also promote the phosphorylation of Akt. It was therefore interesting to determine how the apelin receptor was coupled to initiate this transduction cascade.
Once again, we confirmed that the receptor was not coupled to the G i3 protein (Fig. 5). Although we found that both apelin fragments induced Akt phosphorylation via a coupling of the apelin receptor to either the G i1 or the G i2 protein (Fig. 5), we noted significant differences in the reconstitution efficiency. Thus, the phosphorylation of Akt induced by apelin-(65-77) was clearly more efficient when the PTXinsensitive ␣ i1 subunit was expressed (Fig. 5A), and the opposite specificity was observed for apelin-(42-77) (Fig. 5B).  All these results clearly show that both apelin-(65-77) and apelin-(42-77) regulate the activity of the same effectors and apparently do not modify the conformation of the receptor in such a way that it can interact with the G i3 protein. On the other hand, although the transduction of both apelin fragments can be mediated by either G i1 or G i2 , these proteins differ significantly in the efficiency with which they regulate a given effector.
The Regulation of Intracellular Effectors by Apelin Fragments Undergoes Desensitization-Arguing that a functional difference in the effects of the apelin fragments could also reside in the regulation of their biological response, we next decided to analyze the mechanisms of receptor desensitization induced by each apelin fragment. To analyze possible desensitization of apelin signaling, we preincubated CHO cells in the presence of 1 M apelin-(65-77) or apelin-(42-77) and determined their ability to regulate the three effectors studied earlier. In preliminary experiments, we investigated the concentration dependence of the desensitization process and observed that apelin concentrations in the nanomolar range were required to visualize significant desensitization of the apelin-(65-77)-induced inhibition of adenylyl cyclase. 4 After preincubation with apelin-(65-77), we observed a partial decrease of adenylyl cyclase inhibition induced by nanomolar concentrations of both apelin fragments, revealing both autologous and crossdesensitization. On the other hand, preincubation with apelin-(42-77) led to a strong decrease in the basal levels of cyclic AMP, which were not further reduced by the addition of either apelin fragment (Fig. 6A). As the basal cAMP levels were identical to those of control cells activated by the highest concentrations of apelins, these results can be interpreted as reflecting a sustained activation of the apelin receptor.
Similarly, preincubation with apelin-(65-77) strongly decreased the phosphorylation of ERK (Fig. 6B) and Akt (Fig. 6C) induced by apelin-(65-77) and apelin-(42-77). Furthermore, preincubation with apelin-(42-77) reduced the phosphorylation of ERK and Akt induced by both apelin fragments (Fig. 6, B and C). Interestingly, as described for adenylyl cyclase regulation, we observed an increase in the basal phosphorylation of Akt after preincubation with apelin-(42-77) (Fig. 6C). Basal phosphorylation of ERK was also somewhat enhanced, as could be demonstrated by increasing film exposure time. 5 Taken together, these data show that apelin-(65-77) induces an autologous desensitization of its intracellular responses and promotes a cross-desensitization of the responses induced by apelin- (42-77). Although the pattern of effector desensitization was very similar for the other apelin fragment, there was an obvious difference in that only preincubation with apelin-(42-77) increased basal activities of the three effectors, suggesting a persistent activation of the apelin receptor.
G i Coupling and Receptor Desensitization-Since receptor coupling may alter the desensitization process, we decided to analyze the desensitization induced by each apelin fragment in CHO cells expressing either the PTX-insensitive ␣ i1 subunit or the PTX-insensitive ␣ i2 sub-4 Y. Audigier, unpublished data. 5 B. Masri, unpublished data.  unit. With regard to adenylyl cyclase inhibition, the pattern of receptor desensitization by each fragment was qualitatively identical in clones, expressing, respectively, the PTX-insensitive ␣ i1 and ␣ i2 subunit and corresponded to the results obtained with the CHO cells expressing the murine apelin receptor. Preincubation with apelin-(65-77) induced a partial desensitization of the inhibition induced by both apelin fragments, whereas preincubation with apelin-(42-77) led to a constitutive decrease of cAMP levels and the concomitant disappearance of the inhibitory effects of both apelin fragments (Fig. 7A). As concerns ERK phosphorylation, autologous desensitization was similar for each of the PTX-insensitive ␣ i subunits. As described in Fig.  4, under control conditions, we again demonstrated the superiority of apelin-(42-77) in activating ERK phosphorylation in the cells expressing the PTX-insensitive ␣ i1 subunit (Fig. 7B). If this difference is taken into account, preincubation with apelin-(65-77) also induced a similar and partial desensitization of the ERK phosphorylation induced by both apelin fragments. Similarly, preincubation with apelin-(42-77) decreased the ERK phosphorylation induced by both apelin fragments, and a small but significant increase in basal ERK phosphorylation could be seen (Fig. 7B).
At the level of Akt phosphorylation, we observed no obvious difference in the pattern of desensitization between the cells, regardless of the PTX-insensitive ␣ i subunit they expressed (Fig. 7C). Again, each apelin fragment induced not only its own desensitization but also the desensitization promoted by the other fragment. As observed previously, after preincubation with apelin-(42-77), we noticed an increase in the basal phosphorylation level of Akt in the cells expressing the PTX-insensitive ␣ i2 subunit (Fig. 7C).

Regulation of Intracellular Effectors by Apelin Fragments and Its Desensitization in a C-terminally Truncated Apelin
Receptor-Receptor-mediated phosphorylation of serine/threonine residues in the C-terminal tail of G protein-coupled receptors is crucial for their subsequent internalization (26). Consistent with this assertion, the murine apelin receptor contains 14 serine or threonine residues in its C terminus (4), and the protein is rapidly internalized after ligand binding (27)(28)(29). To investigate the role of phosphorylation in desensitization, we constructed a truncated receptor, ⌬327-377, in which most serine/threonine residues were deleted, and analyzed its activation by apelin fragments.
As expected, truncation of the C terminus prevented desensitization of the biological responses induced by apelin-(65-77), notably the inhibition of adenylyl cyclase (Fig. 9A). Indeed, preincubation with apelin-(65-77) did not result in desensitization of the responses induced by low concentrations of apelin-(65-77) or apelin-(42-77) (Fig. 9A). On the other hand, the consequences of preincubation with apelin-(42-77) were not modified by receptor truncation; we again observed the same decrease of basal cAMP levels and the loss of the activity of both fragments that we already observed in the case of the wild-type receptor (Fig. 9A).
Most of these findings were reproduced when the activity of apelin fragments was assessed with respect to ERK (Fig. 9B). The phosphorylation of ERK induced by both apelin fragments was not significantly affected by preincubation with apelin-(65-77), whereas basal ERK phosphorylation was slightly increased after preincubation with the (42-77) fragment (Fig. 9B).
When Akt phosphorylation was analyzed, we observed a significant desensitization of Akt phosphorylation induced by both apelin fragments after preincubation with apelin-(65-77) (Fig. 9C). Interestingly, preincubation with apelin-(42-77) led to results identical to those obtained for the regulation of adenylyl cyclase, i.e. a strong increase in the basal level of Akt phosphorylation and the inability of either apelin fragment to increase it further (Fig. 9C).
Most of the results so far described were obtained via a methodological approach based on the expression of modified proteins in CHO cells. Nevertheless, it should be underlined that the activation of transduction cascades by apelin-(65-77) has also been described in other cell types as well as in isolated organs (19,24,30). As we recently demonstrated that the apelin receptor is expressed by HUVEC (24), we decided to address the question of the physiological relevance of the activation and regulation of these effectors by apelin fragments in these cells.
G␣ il and G␣ i2 Subunits Are Expressed in HUVEC-Given the preferential coupling of the apelin receptor to G i1 or G i2 proteins, we first investigated whether HUVEC did indeed express the G␣ i1 and G␣ i2 subunits. Immunoblotting of HUVEC lysates clearly revealed the presence of immunoreactive proteins of the expected molecular weight (Fig. 10A).
Apelin Fragments Induce ERK Phosphorylation in HUVEC-Although we were unable to detect inhibition of adenylyl cyclase activity in this cell type, both apelin-(65-77) and apelin-(42-77) elicited a timedependent phosphorylation of ERK and Akt, which peaked at 15 min. 5 As observed in CHO cells, ERK phosphorylation was identical for each apelin fragment. Based on these kinetic experiments, we analyzed the desensitization of ERK phosphorylation at 15 min. Preincubation of HUVEC with apelin-(65-77) or apelin-(42-77) resulted in a clear autol-ogous decrease of ERK in each case (Fig. 10B), demonstrating a desensitization of this biological response. In addition, a significant crossdesensitization between apelin fragments was observed. On the other hand, although basal ERK phosphorylation was not increased after preincubation of HUVEC with apelin-(42-77), it is noteworthy that basal ERK phosphorylation is already high in HUVECs preincubated in the absence of apelin fragments (Fig. 10B).
Apelin Fragments Induce Akt Phosphorylation and Promote a Differential Desensitization of the Response in HUVEC-Apelin-(65-77) and apelin-(42-77) also induced a time-dependent phosphorylation of Akt, which displayed the same kinetics as ERK phosphorylation. As observed in CHO cells, Akt phosphorylation induced by apelin-(65-77) at each time point was slightly more intense than that induced by apelin-(42-77). Based on these kinetic experiments, we analyzed the desensitization of Akt phosphorylation at 15 min. As described for ERK, preincubation of HUVEC with apelin-(65-77) abrogated Akt phosphorylation induced by  apelin-(65-77) or apelin-(42-77) (Fig. 10C). In addition, the basal level of Akt phosphorylation was strongly increased after preincubation of HUVEC with apelin-(42-77), and neither apelin-(65-77) nor apelin-(42-77) were then able to increase it. In conclusion, most data obtained in HUVEC reproduce and confirm the results previously described in CHO cells and thus argue in favor of the physiological relevance of the differential desensitization induced by the two fragments.
Acid Washing Does Not Alter the Desensitization Pattern Induced by Apelin-(42-77)-As residual binding could explain the desensitization pattern induced by apelin- (42-77), we carried out a wash at low pH (31) to strip any remaining peptide from the receptor and restore the basal activity of adenylyl cyclase or the basal level of Akt phosphorylation. We observed that although acid wash at pH 4 decreased the forskolin-induced activation of adenylyl cyclase by 30%, apelin fragments were still able to promote adenylyl cyclase inhibition in CHO cells (Fig. 11A). Interestingly, a wash at pH 4 was without effect on the increased basal activity following preincubation with apelin-(42-77). On the other hand, washing at pH 3 strongly decreased the stimulation of adenylyl cyclase by forskolin and fully abolished the inhibition induced by apelin fragments.
We also analyzed the effect of an acid wash at pH 4 on Akt phosphorylation in HUVEC after a preincubation with apelin-(42-77). Fig. 11B shows that the high basal level of Akt phosphorylation was not changed by a wash at low pH.

DISCUSSION
To determine whether each apelin fragment plays a distinct physiological role, we compared the ability of the long (42-77) and short (65-77) forms of apelin to activate the apelin receptor, to promote G i protein coupling, to regulate intracellular effectors, and to induce receptor desensitization.
The Apelin Fragments Regulate the Same Effectors-Apelin-(65-77) has been shown to inhibit adenylyl cyclase (22) and to induce the phosphorylation of ERK (23) or Akt (24). We report here that the longer fragment apelin-(42-77) regulates these intracellular effectors in a sim-  ilar way, both in transfected CHO cells and in HUVEC that endogenously express the apelin receptor. In addition, the kinetics of ERK or Akt-induced phosphorylation were identical for both apelin fragments in HUVEC. As observed previously for the increase of acidification rate (6), we also found that apelin-(65-77) was more efficient than apelin-(42-77), on the basis of their respective IC 50 values in inhibiting adenylyl cyclase.
The Apelin Fragments Promote Coupling to G i1 or G i2 Proteins-As apelin-induced activation of transduction cascades is blocked by PTX (19,23,24), we addressed the question of the G i protein signaling specificity of each apelin fragment by using PTX-insensitive mutants of ␣ i1 , ␣ i2 , or ␣ i3 subunit stably expressed in CHO cells. The first important observation is that the apelin receptor does not couple indifferently to the three G␣ i subunits. Regardless of the apelin fragment and of the regulated effector, we were unable to observe an efficient coupling between the apelin receptor and the ␣ i3 subunit, clearly demonstrating a coupling specificity of the apelin receptor toward G i1 and G i2 proteins. Interestingly, previous studies using PTX-insensitive ␣ i subunits have shown that the long form of the D2 dopamine receptor (32) and the m2 muscarinic acetylcholine receptor (33) can signal through the mutated ␣ i3 subunit to inhibit forskolin-stimulated adenylyl cyclase. Thus, the poor coupling to G i3 protein reflects an intrinsic property of the apelin receptor.
The second observation is that the transduction of both apelin fragments can be reconstituted in cells expressing either the ␣ i1 or the ␣ i2 subunit. However, the extent of this reconstitution depends on the fragment and on the effector. For example, the degree of inhibition of adenylyl cyclase induced by both apelin peptides is higher in the cells expressing the ␣ i2 subunit than in the cells expressing the ␣ i1 subunit. In addition, the apparent affinity of apelin-(42-77) is five times higher for adenylyl cyclase inhibition in the cells expressing the ␣ i2 subunit. This difference may not represent a coupling preference of the receptor but rather a higher efficacy of activation at the level of the G protein and/or effector. Conversely, we noticed a selective recruitment of the G i2 protein for apelin-(65-77) signaling to activate ERK, whereas G i1 and G i2 proteins were equally operative in mediating ERK phosphorylation induced by apelin-(42-77). As the differential coupling induced by the two apelin fragments may vary as a function of the levels of expression of each ␣ i subunit, the differences observed in adenylyl cyclase inhibition or ERK activation by the apelin fragments need to be substantiated in other cell types.
Apelin Receptor Is Subject to Differential Desensitization by Apelin Fragments-Although internalization of the apelin receptor has been widely observed by the use of tagged receptors or fluorescent apelin-(65-77) (27)(28)(29), the only evidence of desensitization was that repeated exposure to apelin fragments resulted in a decrease of the calcium response in NT2.N neurons (10). In this study, we report interesting features of the differential effects induced by a preincubation with each fragment.
As expected, autologous desensitization is observed for apelin-(65-77) regardless of the effector, of the origin of the expressed apelin receptor, and of the cell type. Desensitization of the response was obtained with respect to inhibition of adenylyl cyclase and to phosphorylation of ERK and Akt. In addition, the activation of the human receptor expressed by HUVEC was desensitized in a manner similar to that of the murine receptor in transfected CHO cells. Finally, after a preincubation with apelin-(65-77), the responses induced by apelin-(42-77) were also decreased, a process referred to as cross-desensitization. This phenomenon was observed in CHO cells with all three effectors examined. Similarly, cross-desensitization between apelin fragments was also observed in HUVEC with respect to the phosphorylation of ERK or Akt by each fragment. In view of the ability of apelin-(65-77) to induce receptor internalization, the decreased number of available receptors at the cell surface would account for the diminished cell responsiveness to each apelin fragment.
On the other hand, preincubation with apelin-(42-77) induced different and very interesting consequences. Thus, the basal level of adenylyl cyclase activity was switched to a level corresponding to that observed after the inhibition induced by apelin fragments. In addition, a similar but smaller increase in the basal phosphorylation of both ERK and Akt was also observed. Finally, similar results were obtained in both cell types (CHO and HUVEC) with respect to ERK and Akt effectors when the preincubation was performed in the presence of apelin-(42-77).
Taken together, these data reveal differential properties of apelin fragments in their ability to regulate the activity of intracellular effectors in a chronic manner. Indeed, the qualitative consequence of a preincubation with apelin-(42-77) clearly corresponds to a persistent activation of apelin receptors, although the increase in the basal activity may vary from one effector to the other.
C-terminal Truncation of the Receptor Does Not Modify the Desensitization Pattern Induced by Apelin-(42-77)-Interestingly, whereas apelin-(65-77) was more efficient than apelin-(42-77) in inhibiting adenylyl cyclase in CHO cells expressing the wild-type apelin receptor, the opposite result was observed in CHO cells expressing the truncated apelin receptor. The loss of G protein-coupled receptor desensitization after truncation of the C-terminal region, which contains the phosphorylation sites for G protein-coupled receptor kinase and the subsequent interaction with ␤-arrestins, is firmly established (for a review, see Ref. 34). Accordingly, the disappearance of apelin-(65-77)-induced desensitization with respect to both adenylyl cyclase inhibition and ERK phosphorylation results from the suppression of receptor internalization, as was reported for a truncated receptor expressed in HEK-293 cells (29). On the other hand, it was interesting to note that the desensitization of the effector response promoted by apelin-(42-77) was unchanged in the truncated receptor. Interestingly, similar but not identical results were obtained for the apelin-induced phosphorylation of Akt. This observation suggests that the effects induced by preincubation with apelin-(42-77) do not depend on receptor internalization and may be upstream of the activation of the intracellular cascades.
Apelin-(42-77) Promotes a Persistent Activation of Apelin Receptors -It was therefore tempting to speculate that most, if not all, of the distinguishing properties that we observed with apelin-(42-77) rely on an increase of the proportion of receptors in an active state. In quantitative terms, the proportion of activated receptors remains to be determined. Clearly, a low receptor/G protein ratio does not facilitate the measurement of the extent of this persistent activity (35). Conversely, a high efficacy linked to a low receptor occupancy and a maximal effector response may provide better conditions for visualizing such a phenomenon. This might be the case for the regulation of adenylyl cyclase (36) since the basal activity after preincubation with apelin-(42-77) strictly corresponds to the full inhibition of adenylyl cyclase induced by each apelin fragment.
Interestingly, a low rate of dissociation of apelin-(42-77) has been demonstrated using a membrane preparation (7). In addition, the corollary of a low degree of ligand dissociation is that the basal level of effector activity should be increased, a phenomenon that we observed after a preincubation with apelin-(42-77). However, acid washing of the cells (31) did not modify the persistent activation induced by apelin-(42-77), suggesting that residual binding of apelin-(42-77) is not the explanation for this persistent activation. Nevertheless, it should be pointed out that this treatment strongly affects the biological response, notably adenylyl cyclase inhibition.
Physiological Consequences of the Differential Desensitization by Apelin Fragments-We can therefore hypothesize that the distinct physiological function of these two peptides may rely on the temporal regulation of the response. Activation of apelin receptors by the short fragment is proposed to be transient and rapidly turned off by receptor internalization (27,28). The duration of desensitization would thus depend on the rate of receptor recycling to the plasma membrane (29). In fact, the receptors internalized by apelin-(65-77) are recycled to the surface within 1 h (27,29); consequently, desensitization of the response is only transient.
In contrast, we suggest that binding of apelin-(42-77) locks some receptors in an active state, thereby increasing the persistence of their activation. Even if a few of these receptors can be internalized, the low dissociation rate of apelin-(42-77) from the receptor is also likely to decrease the rate of receptor recycling, as described previously (29), or may even target the ligand-receptor complex to lysosomes. Regardless of the exact mechanism, the activation of apelin receptors would be longer and the reactivation more difficult. Accordingly, we propose a functional distinction between apelin-(65-77) and apelin-(42-77) based on the duration of their action and the time course of the desensitization process.
Pharmacological Implications of Apelin-(42-77)-induced Persistent Activation-The human apelin receptor can act as a co-receptor for the entry of several HIV-1 and simian immunodeficiency virus strains (9,37). Intriguingly, whereas apelin-(65-77) is more efficient than apelin-(42-77) in increasing acidification rate and inhibiting adenylyl cyclase (6,22), this order of efficiency is reversed for the inhibition of HIV entry (38 -40). As discussed previously, the lower antiviral activity of apelin-(65-77) cannot be explained by its lesser ability to promote receptor internalization or its poorer affinity for the receptor. It is thus likely that the apelin-(42-77)induced locked conformation of the apelin receptor and the low dissociation rate of the ligand-receptor complex are key events that underlie the high efficiency with which this apelin fragment inhibits HIV binding and entry.
Conclusions-Although the two apelin fragments, apelin-(65-77) and apelin-(42-77), regulate the same set of intracellular effectors, they promote a distinct desensitization pattern that may be central to their respective physiological functions. These different properties may be useful for designing more specific pharmacological tools to treat the pathological dysfunctions linked to disturbances in apelin signaling.