The Pseudo Signal Peptide of the Corticotropin-releasing Factor Receptor Type 2a Decreases Receptor Expression and Prevents Gi-mediated Inhibition of Adenylyl Cyclase Activity*

The corticotropin-releasing factor receptor type 2a (CRF2(a)R) belongs to the family of G protein-coupled receptors. The receptor possesses an N-terminal pseudo signal peptide that is unable to mediate targeting of the nascent chain to the endoplasmic reticulum membrane during early receptor biogenesis. The pseudo signal peptide remains uncleaved and consequently forms an additional hydrophobic receptor domain with unknown function that is unique within the large G protein-coupled receptor protein family. Here, we have analyzed the functional significance of this domain in comparison with the conventional signal peptide of the homologous corticotropin-releasing factor receptor type 1 (CRF1R). We show that the presence of the pseudo signal peptide leads to a very low cell surface receptor expression of the CRF2(a)R in comparison with the CRF1R. Moreover, whereas the presence of the pseudo signal peptide did not affect coupling to the Gs protein, Gi-mediated inhibition of adenylyl cyclase activity was abolished. The properties mediated by the pseudo signal peptide were entirely transferable to the CRF1R in signal peptide exchange experiments. Taken together, our results show that signal peptides do not only influence early protein biogenesis. In the case of the corticotropin-releasing factor receptor subtypes, the use of conventional and pseudo signal peptides have an unexpected influence on signal transduction.

In the case of the CRF 2 R, three splice variants have been described as follows: the CRF 2(a) R, CRF 2(b) R, and CRF 2(c) R. All splice variants bind CRF with low affinity and the urocortins 1-3 with high affinity. They are involved in the regulation of feeding behavior (7) and in recovery from a stress response (8). It is likely that they are also involved in modulating anxietyrelated behavior.
Both the CRF 1 R and the CRF 2(a) R usually couple to the G s / adenylyl cyclase system and consequently increase cytosolic cAMP as a second messenger. However, a promiscuous coupling behavior was described previously in particular for the CRF 1 R involving also G proteins of the G i , G o , and G q families (9 -11). In the case of the CRF 1 R, coupling to G s at low agonist occupancy and to G i at high occupancy leads to a typical bellshaped concentration-response curve in cAMP accumulation assays (12).
The CRF receptors belong to the small group of GPCRs (5-10%) possessing putative N-terminal signal peptides that are cleaved off after mediating the ER targeting/insertion process (13,14). The majority (90 -95%) of the GPCRs do not possess cleavable signal peptides. Here, one of the transmembrane helices of the mature receptors (usually TM1) mediates ER targeting/insertion as an uncleaved signal anchor sequence (13). The reason why some membrane proteins, including GPCRs, require additional signal peptides, whereas others do not, is not completely understood.
An initial function of a signal sequence (cleaved signal peptide or uncleaved signal anchor sequence) is to mediate targeting of the nascent chain to the translocon complex of the ER by binding the signal recognition particle. Moreover, the signal sequence opens the Sec61 protein-conducting channel to integrate the nascent chain into the bilayer. In the case of the CRF 1 R and the glucagon-like peptide-1 receptor, it was shown by deletion mutants that a cleavable signal peptide is indeed necessary for efficient receptor biosynthesis at the ER membrane (15,16). However, additional functions were described for other GPCRs. In the case of the endothelin B receptor, the signal peptide facilitated N tail translocation across the ER membrane (17). The CRF (2a) R, in contrast, possesses an uncleaved pseudo signal peptide that is unable to mediate ER targeting/insertion, remains uncleaved at the extracellular mature N tail of the receptor, and thus forms an additional hydrophobic domain (18). Conventional signal peptide functions are blocked in the pseudo signal peptide by a single amino acid residue (Asn 13 ), and conversion to a conventional cleaved signal peptide is achieved by mutation of this residue (18). To date, the pseudo signal peptide of the CRF 2(a) R is a unique extracellular domain in the GPCR protein family. One function of the pseudo signal peptide may be to increase the portion of correctly folded receptors in the early secretory pathway (18). Here, we have assessed its functions in comparison with the conventional signal peptide of the homologous CRF 1 R. We show that the presence of the pseudo signal peptide leads to a very low receptor expression. Moreover, although the presence of the pseudo signal peptide did not affect coupling to the G s protein, G i coupling of the CRF 2(a) R was impaired.

Materials
The cDNA encoding the rat CRF 1 R and CRF 2(a) R was a gift from U. B. Kaupp (IBI Forschungszentrum Jülich, Germany). [ 3 H]cAMP was purchased from PerkinElmer Life Sciences. The peptidic ligand sauvagine was synthesized in our laboratory (10). Lipofectamine TM 2000 and the vector pSecTag2A were purchased from Invitrogen. The transfection reagent FuGENE TM 6 was from Roche Diagnostics. Polyethyleneimine (PEI) was from Polysciences Europe GmbH (Eppelheim, Germany). The monoclonal mouse anti-CRF 1 R antibody was from LifeSpan Bioscience (Seattle, WA). The phycoerythrin-conjugated goat anti-mouse IgG was from Jackson ImmunoResearch (West Grove, PA). The polyclonal rabbit anti-calnexin antibody was from StressGen (Ann Arbor, MI). The alkaline phosphatase-conjugated anti-rabbit IgG and alkaline phosphatase-conjugated anti-mouse IgG were from Dianova (Hamburg, Germany). The polyclonal rabbit anti-biotin antibody, DyLight 800-conjugated anti-rabbit IgG and DyLight 680-conjugated anti-mouse IgG were from Biomol (Hamburg, Germany). The polyclonal rabbit anti-GFP antiserum 02 was raised against a glutathione S-transferase/GFP fusion protein in our group, and specificity was verified. 4 The monoclonal mouse anti-GFP antibody and the TALON metal affinity resin were from BD Biosciences. DNA-modifying enzymes, PNGaseF and EndoH, were from New England Biolabs (Frankfurt am Main, Germany). Oligonucleotides were from Biotez (Berlin, Germany). Vector plasmid pEGFP-N1 (encoding the enhanced variant of GFP) and the monoclonal anti-GFP antibody were from Clontech. The Roti-Load sample buffer was from Carl Roth (Karlsruhe, Germany). All other reagents were from Sigma. Data of the cAMP RIA were analyzed using the program RadLig Software 6.0 (Cambridge, UK) and GraphPad Prism version 3.02 (GraphPad Software, San Diego).

DNA Manipulations
Standard DNA manipulations were carried out according to the handbooks of Sambrook and Russel (19). The nucleotide sequences of the plasmid constructs were verified using the FS dye terminator kit from PerkinElmer Life Sciences. Sitedirected mutagenesis was carried out with the QuikChange site-directed mutagenesis kit from Stratagene (Heidelberg, Germany).

Plasmid Constructs
The constructs used in this study are schematically shown in Fig. 1 (details of the cloning procedures on request).
Marker Protein Fusions-Constructs CRF 2(a) .NT and CRF 1 . NT represent GFP fusions to the N tails of the CRF 2(a) R (position Ala 121 ) and the CRF 1 R (position Ala 119 ), respectively, in the vector pSecTag2A. In construct SP1-CRF 2(a) .NT, the pseudo signal peptide of the CRF 2(a) R was replaced by the conventional signal peptide of the CRF 1 R. Conversely, in construct SP2-CRF 1 .NT, the signal peptide of the CRF 1 R was replaced by the pseudo signal peptide of the CRF 2(a) R. In construct pN13A-CRF 2(a) , residue Asn 13 of the CRF 2(a) .NT construct was replaced by an alanine residue to convert the pseudo signal peptide into a conventional, cleaved signal peptide (18). In all marker protein fusion constructs, an additional C-terminal His 6 sequence was added to allow their purification.
Full-length Receptor Constructs-Plasmids pCRF 2(a) R and pCRF 1 R encode the full-length CRF 2(a) R and CRF 1 R in the vector plasmid pEGFP-N1. The receptors were C-terminally tagged with a GFP moiety at position Val 411 (CRF 2(a) R) and Thr 413 (CRF 1 R) thereby deleting the stop codons. Plasmids p⌬SP-CRF 2(a) R and p⌬SP-CRF 1 R encode the corresponding signal peptide deletion mutants. Plasmids pSP1-CRF 2(a) R and pSP2-CRF 1 R encode the corresponding signal peptide swap mutants. In construct pN13A-CRF 2(a) R, the pseudo signal peptide of the CRF 2(a) R was converted to a conventional signal peptide (see above).

Cell Culture and Transfection
Cells were cultured at 37°C and 5% CO 2 . Human embryonic kidney cells (HEK 293 cells) and mouse anterior pituitary tumor cells (AtT-20 cells) were grown in Dulbecco's modified Eagle's medium (DMEM) containing 10% (v/v) fetal calf serum (FCS), penicillin (100 units/ml), and streptomycin (100 g/ml). Transfection of the cells with Lipofectamine TM 2000 and PEI was carried out according to the supplier's recommendations.

Confocal Laser-scanning Microscopy, Colocalization of Constructs with Plasma Membrane Marker Trypan Blue
1.5 ϫ 10 5 HEK 293 cells grown for 24 h in a 35-mm diameter dish containing a poly-L-lysine-coated coverslip were transfected with 0.8 g of plasmid DNA and PEI according to the supplier's recommendations. Cells were incubated overnight, washed once with PBS, and transferred immediately into a selfmade chamber (details on request). Cells were covered with 1 ml of PBS (137 mM NaCl, 2.7 mM KCl, 4.3 mM Na 2 HPO 4 , 1.47 mM KH 2 PO 4 , pH 7.4), and trypan blue was added to a final concentration of 0.05%. After 1 min of staining, GFP and trypan blue fluorescence signals were visualized at room temperature on a Zeiss LSM510-META inverted confocal laser-scanning microscope (objective lens, ϫ100/1.3 oil; optical section, Ͻ0.8 m; multitrack mode; GFP, exc , 488 nm, argon laser; BP filter, 500 -530 nm; trypan blue; exc , 543 nm HeNe laser; LP filter, 560 nm). The overlay of both signals was computed using the Zeiss LSM510 acquisition software (release 3.2 SP2). Images 4 C. Rutz and R. Schü lein, unpublished results.
were imported into Photoshop software (Adobe Systems Inc.), and contrast was adjusted to approximate the original image. The overlay of the signals was computed, and the images were processed as described previously above.

Quantification of Plasma Membrane GFP Fluorescence Intensities by Automated Microscopy
For quantification of the plasma membrane fluorescence signals of the receptor, 7 ϫ 10 4 HEK 293 cells were grown for 24 h on 24-well microtiter plates and transfected with PEI according to the supplier's recommendations. Cells were incubated overnight, and the cell culture medium was removed, and the nuclei of the cells were stained for 30 min with 1 M Hoechst-33342 diluted in medium. The staining solution was removed, and the plasma membranes were stained for 1 min with 0.1% (w/v) trypan blue diluted in PBS. After staining, plates were transferred to a Cellomics Array Scan VTI automated microscope and analyzed using the Array Scan VTI software (version 5.6.1.3, Thermo Fisher Scientific). Hoechst-33342 and trypan blue fluorescence signals were used to define nuclei and plasma membrane masks, respectively. Intracellular masks were defined as the difference between plasma membrane and nuclei masks. For each well, colocalization of the GFP fluorescence signals of the receptor with the plasma membrane and intracellular masks was measured for 1 ϫ 10 3 transfected cells using the Array Scan VTI software. In addition to automated microscopy, plasma membrane expression of the receptor constructs was also quantified by conventional cell surface biotinylation assays as described previously (20).

Quantification of Plasma Membrane GFP Fluorescence Intensities by Confocal LSM Microscopy or Flow Cytometry Measurements
Plasma membrane GFP fluorescence signals of individual transiently transfected HEK 293 cells were quantified by confocal LSM microscopy as described previously (15). Briefly, the GFP fluorescence signals were colocalized with the plasma membrane marker trypan blue, and their intensity was quantified using an 8-bit grayscale and the Zeiss LSM510 acquisition software (release 3.2 SP2). Quantification of the plasma membrane GFP fluorescence signals was carried out with at least 20 cells. For the flow cytometry measurements, 1 ϫ 10 6 stably transfected HEK 293 cells were grown on 12-well plates for 48 h. Cells were washed twice with PBS. Cells were transferred in PBS into a flow cytometry tube and incubated with a monoclonal mouse anti-CRF 1 R antibody (1:400, 30 min, 4°C) directed against the N tail of the receptor. Cells were washed three times with PBS and incubated with phycoerythrin-conjugated goat anti-mouse IgG (1:100, 30 min, 4°C). Cell surface fluorescence signals of 10 4 cells were analyzed using a FACSCanto II apparatus (BD Biosciences) and analyzed with FCS Express software (De Novo Software, Los Angeles). Nontransfected HEK 293 cells were used in the measurements to subtract the fluorescence background.

Quantitative Detection of Secreted GFP Marker Protein Fusions
Secreted GFP marker proteins from transiently transfected HEK 293 cells were analyzed by SDS-PAGE immunoblotting and fluorimetric measurements as described previously (15). Briefly, proteins were isolated from the cell culture medium via their His tag using TALON metal affinity resin beads. Proteins were quantified either by measuring their GFP signals fluorimetrically ( exc ϭ 488 nm, em ϭ 507 nm) or by SDS-PAGE immunoblotting using a monoclonal anti-GFP antibody (1:4,000) and alkaline phosphatase-conjugated anti-mouse IgG 1:1,500).

Immunoprecipitation of GFP-tagged Full-length Receptor Constructs; Treatment with PNGaseF and/or EndoH; Detection of Coimmunoprecipitated Calnexin
The previously described immunoprecipitation procedure was used (18). Briefly, receptors were precipitated from transiently transfected HEK 293 cells using the rabbit anti-GFP antiserum 02. Optional treatment of the precipitated receptors with EndoH or PNGaseF was carried out according to the supplier's recommendations. Receptors were detected by SDS-PAGE immunoblotting using a monoclonal anti-GFP antibody (1:4,000) and alkaline phosphatase-conjugated anti-mouse IgG (1:1,500). Coprecipitated calnexin was detected using a polyclonal anti-calnexin antibody (1:1,000) and alkaline phosphatase-conjugated anti-rabbit IgG (1:1,000).

Cell Surface Biotinylation Assay
Cell surface proteins were labeled with sulfo-NHS-biotin as described previously (15). Total receptors were precipitated using the rabbit anti-GFP antiserum 02 as described above. Receptors were deglycosylated with PNGaseF according to the supplier's recommendations and analyzed by SDS-PAGE immunoblotting using a polyclonal rabbit anti-biotin antibody (1:5,000, cell surface receptors) and a monoclonal mouse anti-GFP antibody (1:4,000, total receptors) on the same blot. Following incubation with secondary antibodies (DyLight 800conjugated anti-rabbit IgG and DyLight 680-conjugated antimouse IgG; 1:10,000 each), immunoreactive protein bands were detected using the Odyssey TM infrared imaging system and the application software 2.1 (Li-COR Biosciences, Lincoln, NE).

cAMP Accumulation Assay
Activation of the full-length receptor constructs was monitored by measuring sauvagine-mediated cAMP accumulation as described previously (cAMP RIA) (21).

The Presence of the Pseudo Signal Peptide of the CRF 2(a) R Leads to a Very Low Receptor Expression at the Plasma Membrane-
We have previously shown that the CRF 2(a) R possesses an uncleaved N-terminal pseudo signal peptide, whereas the CRF 1 R contains a conventional signal peptide ( Fig.  1A) (18). Here, we have assessed in a comparative study the influence of these signal peptides on the plasma membrane expression of the different CRF receptor subtypes.
For this study, we used the C-terminally GFP-tagged CRF 2(a) R and CRF 1 R and also constructed signal peptide swap mutants (Fig. 1B); the conventional signal peptide of the CRF 1 R was replaced by the pseudo signal peptide of the CRF 2(a) R (construct SP2-CRF 1 R, Fig. 1B), and conversely, the signal peptide of the CRF 2(a) R was replaced by that of the CRF 1 R (construct SP1-CRF 2(a) R, Fig. 1B). In addition, the N13A mutant of the CRF 2(a) R was used (construct N13A-CRF 2(a) R, Fig. 1B). This point mutation converts the pseudo signal peptide into a conventional and cleaved signal peptide (18).
We first studied whether the properties of the different signals are preserved following the signal swap. To this end, marker protein fusions were constructed (Fig. 1B). The N tails of the above constructs were fused with a His-tagged GFP ( Fig.1B;constructsCRF 2(a) .NT,CRF 1 . NT, SP1-CRF 2(a) .NT, SP2-CRF 1 .NT, and N13A-CRF 2(a) .NT). These constructs do not contain transmembrane domains, and if a conventional signal peptide is present, it directs the soluble GFP marker via the signal recognition particle/ translocon pathway to the ER and subsequently, following cleavage, via the secretory pathway to the cell culture medium (18). The pseudo signal peptide instead fails to target the GFP moiety to the ER; the signal peptide remains uncleaved, and the construct is located in the cytosol and in the nucleus (18) (nonmembrane-bound GFP is able to enter the nucleus, see Ref. 22).
HEK 293 cells were transiently transfected with the constructs, and the GFP fluorescence signals were localized by LSM. In the case of CRF 2(a) .NT, SP2-CRF 1 .NT, and the unfused GFP control protein, the signals were detected diffusely throughout the cell, including the nucleus demonstrating that these fusions were not targeted to the ER membrane ( Fig. 2A). In contrast, in the case of CRF 1 R.NT, SP1-CRF 2(a) R. NT, and the mutant N13A-CRF 2(a) . NT, reticular signals were detected demonstrating that these fusions were able to enter the ER ( Fig. 2A; the validity of this LSM assay has been confirmed previously, see Ref. 18). The reticular signals also colocalized almost completely with the ER stain Rhodamine 6G (data not shown). Consistent with these results, only the constructs CRF 1 R.NT, SP1-CRF 2(a) R.NT, and N13A-CRF 2(a) .NT could be purified via their His tag from cell culture supernatants and be detected by fluorimetric measurements or by immunoblotting (Fig. 2B). Taken together, these results show again that the CRF 2(a) R possesses a pseudo signal peptide failing to mediate ER targeting (18). It could be converted to a conventional signal peptide by the N13A mutation as described previously (18). In contrast, the CRF 1 R possesses a conventional signal peptide that is able to mediate ER association and to direct the GFP moiety to the supernatant as a con-  OCTOBER 22, 2010 • VOLUME 285 • NUMBER 43 sequence of its cleavage. Importantly, signal peptide functions could be swapped; construct SP1-CRF 2(a) .NT behaves like CRF 1 R.NT in these experiments and construct SP2-CRF 1 .NT like CRF 2(a) .NT.

Functional Significance of CRF 2(a) R Pseudo Signal Peptide
To assess for signal peptide cleavage in the case of the fulllength receptors, constructs CRF 2(a) R, CRF 1 R, SP2-CRF 1 R, SP1-CRF 2(a) R, and N13A-CRF 2(a) R (see above) were immunoprecipitated from transiently transfected HEK 293 cells, and the apparent molecular masses of the deglycosylated constructs were compared with their corresponding signal peptide mutants ⌬SP-CRF 1 R (Fig. 1B) and ⌬SP-CRF 2(a) R (Fig. 1B) by immunoblotting. If the signal peptide is cleaved off, constructs should comigrate with the corresponding signal peptide mutants, and if not, the apparent molecular mass should be increased by 2 kDa corresponding to signal peptide size. As expected, constructs CRF 2(a) R and SP2-CRF 1 R indeed possess an uncleaved signal peptide in these experiments, whereas constructs CRF 1 R, SP1-CRF 2(a) R, and N13A-CRF 2(a) R possess a cleavable signal peptide (Fig. 3).
Comparison of the amount of precipitated receptors in Fig. 3 indicates that the CRF 2(a) R is expressed in substantially lower amounts than the CRF 1 R. This lower expression seems to be solely due to the presence of the pseudo signal peptide because construct SP2-CRF 1 R also shows a low expression, whereas construct SP1-CRF 2(a) R is conversely up-regulated. This up-regulation is also seen when the pseudo signal peptide is converted to a conventional signal peptide (construct N13A-CRF 2(a) R).
The deglycosylated, immunoprecipitated receptors shown in Fig. 3 represent a mixture of cell surface and intracellular receptors. To assess the influence of the different signal sequences on the number of mature cell surface receptors, the GFP fluorescence signals of the fulllength receptor constructs were localized in transiently transfected HEK 293 cells by confocal LSM (Fig.  4A, left panel, in green). The cell surface of the same cells was visualized by the use of trypan blue (Fig. 4A,  center panel, in red), and colocalization is indicated in yellow (Fig. 4A,  right panel). A high plasma membrane expression was observed for constructs CRF 1 R and SP1-CRF 2(a) R, a substantially lower plasma membrane expression for constructs CRF 2(a) R and SP2-CRF 1 R. Colocalization of the GFP fluorescence signals with trypan blue was also quantified using automated microscopy (Fig. 4B). Expression of the CRF 2(a) R at the plasma membrane was 25% that of the CRF 1 R in these experiments. In the case of construct SP2-CRF 1 R, the presence at the plasma membrane was reduced to 26% of the CRF 1 R wild type level. Conversely, the amount of SP1-CRF 2(a) R was up-regulated to 115% that of the CRF 1 R.
To verify these results, cell surface biotinylation assays were carried out (Fig. 4, C and D). After labeling of the cell surface receptors with biotin, total receptors were precipitated

Functional Significance of CRF 2(a) R Pseudo Signal Peptide
using an anti-GFP antiserum. Receptors were deglycosylated with PNGaseF and detected by SDS-PAGE immunoblotting using an anti-biotin antibody (cell surface receptors) or by an anti-GFP antibody (total receptors) on the same blot. A similar decrease in cell surface receptor expression was observed for the constructs containing the pseudo signal peptide confirming the above results. Taken together, these results show that the presence of the pseudo signal peptide causes a very low expression of the CRF 2(a) R in the plasma membrane. This property could be transferred to the CRF 1 R by signal peptide exchange.
We next examined the mechanism of the pseudo signal peptide-mediated decrease in receptor expression. One likely possibility was that the presence of the pseudo signal peptide reduces folding efficiency. Misfolded receptors may be recognized by the quality control system (QCS) of the early secretory pathway and retained intracellularly. A decreased folding efficiency and QCS recognition should be accompanied by an increase in the amount of immature, high mannose, and nonglycosylated receptor forms.
To assess for this possibility, the full-length receptor constructs were immunoprecipitated from transiently transfected HEK 293 cells, detected by immunoblotting (Fig. 5A), and the amount of complex high mannose-glycosylated and nonglycosylated receptor forms was determined by densitometric measurements (Fig. 5B). The identity of the three receptor bands was verified by EndoH and PNGaseF treatments (Fig. 5C) (18). In the case of constructs CRF 1 R and SP1-CRF 2(a) R, the majority of the receptors were present in the mature complex glycosylated form (56 and 57% respectively), and the portion of immature receptors was low. In the case of constructs CRF 2(a) R and SP2-CRF 1 R, in contrast, the bulk of the receptor proteins was detectable in their immature (high mannose and nonglycosylated forms) forms (69 and 67%, respectively). Consistently, more intracellular GFP fluorescence signals were detectable for constructs CRF 2(a) R and SP2-CRF 1 R (see also Fig. 4A above), and these intracellular signals colocalized with the ER marker stain Rhodamine 6G (data not shown).
In the case of glycoproteins, a decreased folding efficiency was usually accompanied by a stronger association with the lectin chaperones calnexin and/or calreticulin. To address this question, coprecipitated calnexin was detected in the receptor precipitation samples described above. Immunoreactive calnexin protein bands were detected in the case of constructs CRF 2(a) R and SP2-CRF 1 R, but not in the case of constructs CRF 1 R and SP1-CRF 2(a) R when identical amounts of immunoreactive, deglycosylated receptor protein were loaded (Fig. 6).
Taken together, these results show that the presence of the pseudo signal peptide decreases folding efficiency of the   OCTOBER 22, 2010 • VOLUME 285 • NUMBER 43 JOURNAL OF BIOLOGICAL CHEMISTRY 32883 CRF 2(a) R in comparison with the CRF 1 R. Moreover, this property could be transferred by signal peptide exchange.

The Presence of the Pseudo Signal Peptide of the CRF 2(a) R Prevents G i -mediated Inhibition of Adenylyl Cyclase Activity-
We have previously shown that the pseudo signal peptide of the CRF 2(a) R does not influence the affinity of the receptor for the CRF receptor selective agonist urocortin I and the CRF 2(a) R-specific agonist urocortin II (18). To address the question of whether the presence of the pseudo signal peptide affects receptor activation and second messenger formation, HEK 293 cells transiently transfected with the full-length receptor constructs were treated with the CRF receptor agonist sauvagine, and cAMP formation was measured by radioimmunoassays (cAMP RIA) (Fig. 7A).
Consistent with previous results from other groups (10, 11), a bellshaped concentration-response curve was recorded for the CRF 1 R. This unusual curve results from the fact that the receptor couples to G s at low agonist occupancy but also to G i at high occupancy (10,11). Coupling to G i was confirmed by blunting the concentration-response curve by pertussis toxin pretreatment (data not shown). In the case of the CRF 2(a) R, in contrast, a monophasic concentration-response curve was observed suggesting that the CRF 2(a) R is unable to activate G i . Strikingly, the ability/inability to couple to G i could be transferred by signal peptide exchange; in the case of construct SP1-CRF 2(a) R, a bell-shaped concentration-response curve was observed, whereas construct SP2-CRF 1 R yielded a monophasic curve. These results demonstrate that the presence of the pseudo signal peptide of the CRF 2(a) R was able to prevent G i -mediated inhibition of the adenylyl cyclase activation.
One may argue that the additional G i activation of the CRF 1 R was only detectable in overexpressing transfected cells. To address this question, a cAMP RIA was performed with nontransfected AtT-20 anterior pituitary tumor cells expressing the endogenous CRF 1 R (23). A bell-shaped concentrationresponse curve was also observed under these conditions (Fig. 7B) suggesting that G i activation and impairing this activation in the case of the CRF 2(a) R by the pseudo signal peptide could play a role in the natural CRF system.
The fact that the pseudo signal peptide prevents G i -mediated inhibition of adenylyl cyclase activity (Fig. 7A) may be closely linked to the decreased cell surface expression caused by this domain. If one assumes a limited amount of G s protein in the FIGURE 5. The pseudo signal peptide increases the amount of immature protein present in the early secretory pathway in transiently transfected HEK 293 cells. A, analysis of the glycosylation state of constructs CRF 2(a) R, CRF 1 R, SP2-CRF 1 R, and SP1-CRF 2(a) R. Receptors were precipitated using a polyclonal anti-GFP antiserum and detected by SDS-PAGE immunoblotting using a monoclonal anti-GFP antibody. Nontransfected cells were used as a control (Ϫ). In each lane the same amount of immunoreactive receptor protein was loaded. For each receptor construct, three immunoreactive protein bands are detectable representing the following glycosylation states: mature complex-glycosylated forms (*), immature high mannose forms ( §), and immature nonglycosylated forms (#) (see below). The immunoblot is representative of three independent experiments. B, ratio of the individual immunoreactive protein bands for each construct. Intensity of the protein bands was measured densitometrically, and the relative amount was calculated for each construct. Columns represent mean values (Ϯ S.D.) of protein band intensity of three independent experiments. C, verification of the identity of the three protein bands for the constructs CRF 2(a) R and CRF 1 R (see also Ref. 18). The receptors were precipitated from transiently transfected HEK 293 cells using a polyclonal anti-GFP antiserum. Samples were treated with EndoH or PNGaseF or left untreated (Ϫ). In each lane, the same amount of immunoreactive receptor protein was loaded. Immunoreactive proteins were detected by SDS-PAGE immunoblotting using a monoclonal anti-GFP antibody. Nonglycosylated receptors are EndoH-and PNGaseF-resistant (#); high mannose forms are EndoH-and PNGaseF-sensitive ( §); complex-glycosylated receptors are EndoH-resistant and PNGaseF-sensitive (*). The immunoblot is representative of three independent experiments. cell, a large amount of receptors at the cell surface may deplete the G s pool at high agonist occupancy, and consequently, G i coupling may be allowed (bell-shaped concentration-response curve). A low amount of receptor, in contrast, may not be sufficient to deplete G s , and therefore, G i coupling may not be observed (monophasic concentration-response curve). However, the fact that stimulation of the endogenous CRF 1 receptors of AtT-20 cells also leads to a bell-shaped curve argues against a dominant role of receptor expression, because receptor number is very low in these cells (Fig. 7B). Alternatively, the pseudo signal peptide may favor directly receptor conformations unable to couple to G i .
To study whether the pseudo signal peptide-mediated decrease of receptor expression is responsible for the observed inhibition of G i coupling, constructs CRF 1 R and SP2-CRF 1 R were stably transfected into HEK 293 cells with the aim to select cell clones with matched cell surface receptor expression (expression levels are variable because of different integration sites of the plasmid DNA). Plasma membrane receptors of the individual clones were quantified by measuring their cell surface GFP fluorescence intensities using confocal LSM (Fig. 8A) and by flow cytometry measurements using intact cells and a monoclonal anti-CRF 1 R antibody directed against the N tail (Fig. 8B). In the case of the A6-CRF 1 R clone, cell surface expression is decreased to the level of the B3-SP2-CRF 1 R clone (Fig. 8,  A and B). Nevertheless, the curve for sauvagine-mediated cAMP formation is still bell-shaped (Fig. 8C). Taken together, these results indicate that the pseudo signal peptide-mediated decrease in receptor expression is not responsible for the observed inhibition of G i coupling.

DISCUSSION
We have analyzed the function of the pseudo signal peptide of the CRF 2(a) R in comparison with the conventional signal peptide of the CRF 1 R. Two results were obtained for the significance of this unique GPCR domain as follows. (i) The presence of the pseudo signal peptide causes a very low cell surface receptor expression. (ii) Moreover, it abolishes G i -mediated inhibition of adenylyl cyclase activity.
The observed lower cell surface expression is accompanied by an increase in intracellularly retained receptors that are present in their immature form in the early secretory pathway (high mannose and nonglycosylated forms, Fig. 5). These results indicate that the presence of the pseudo signal peptide decreases folding efficiency, consequently leading to more misfolded receptors that are retained by the QCS of the early secretory pathway. Indeed, we could detect a stronger association of the pseudo signal peptide-bearing constructs with the lectin  chaperone calnexin (Fig. 6). Molinari and Helenius (24) demonstrated that the calnexin/calreticulin system of the QCS interacts directly with membrane proteins when an N-glycosylation site is present within the N-terminal 50 residues. The uncleaved pseudo signal peptide introduces an additional asparagine residue that is indeed N-glycosylated (Asn 13 , see Ref. 18). This may facilitate the observed stronger interaction of unfolded or partially folded receptors with these lectin chaperones. Unfortunately, this hypothesis is difficult to address experimentally because mutation of Asn 13 also converts the pseudo signal peptide into a conventional signal peptide that is cleaved off (Figs. 2 and 3) (18). An alternative way by which the pseudo signal peptide may decrease folding efficiency is by impairing N tail translocation during protein integration into the ER membrane by the translocon machinery.
The experiments in this study compare the functional significance of the CRF 2(a) R pseudo signal peptide with that of the conventional CRF 1 R signal peptide. Interestingly, deletion of the signal peptide of the CRF 2(a) R without replacing it by another sequence leads to a further increase in the amount of immature receptor forms (Ref. 18; mutant ⌬SP-CRF 2(a) R). Thus, when the CRF 2(a) R is considered alone, the pseudo signal peptide also seems to facilitate folding to a certain extent. These data may be put together as follows. The CRF 2(a) R without its pseudo signal peptide or another signal sequence is expressed almost exclusively in its immature form in the early secretory pathway. The presence of the pseudo signal peptide somewhat increases folding efficiency that is, however, still very low in comparison with that mediated by the presence of a conventional signal peptide.
The fact that the pseudo signal peptide prevents G i -mediated inhibition of adenylyl cyclase activity (Fig. 7) seems not to be due to the decreased cell surface expression caused by this domain. Decreasing cell surface expression of CRF 1 R to the level of construct SP2-CRF 1 R still leads to a bell-shaped concentration-response curve (Fig. 8, A-C). Obviously, the presence of the pseudo signal peptide favors receptor conformations that are unable to couple to G i . Such a function should, however, be independent of ligand binding, because affinities of the CRF 2(a) R for both its selective and specific agonists are not influenced by the pseudo signal peptide (18). It was shown previously that heterodimerization may affect selectivity of GPCRs for the different G proteins. An influence on G i coupling, for example, was observed upon coexpression ofand ␦-opioid receptors (25,26) and CCR5/CCR2 chemokine receptors (27). If one speculates that homodimerization may also influence G protein selectivity, the presence of the pseudo signal peptide may prevent receptor dimerization and in turn impair G i coupling. This speculation should be addressed in a future study.
In the case of the CRF 2(b) R subtype, it was shown recently that this receptor is able to couple to G i at high agonist occupancy (12) in contrast to the CRF 2(a) R subtype studied here. It is long known that signal peptide sequences, even of closely related proteins, have a conserved conformation but do not share sequence homologies (28,29). Indeed, the pseudo signal peptide of the CRF 2(a) R and the corresponding sequence of the CRF 2(b) R differ completely in their sequence (MDAALLLSL-LEANCSLALA, CRF 2(a) R versus MGTPGSLPSAQLLLCLFS-LLPVLQVA, CRF 2(b) R). Signal peptide prediction by the Sig-nalP 3.0 program (30) indicates a cleavage probability of 98% for the CRF 2(b) R, and it is thus conceivable that it possesses a conventional signal peptide like the CRF 1 R.
The properties mediated by the pseudo signal peptide of the CRF 2(a) R were entirely transferable to the CRF 1 R in the signal peptide exchange experiments. The sequence may consequently be considered as a novel transport signal negatively influencing receptor processing. The reason why it is advantageous to keep CRF 2(a) R expression low by this unique domain remains elusive. However, the fact that the bell-shaped concentration-response curve and thus G i coupling is also observed in the case of the endogenous CRF 1 R (Fig. 7B) indicates that the use of these different signal sequences may play a role in vivo.
Finally, it is intriguing to speculate that the pseudo signal peptide is part of a novel mechanism regulating cell surface expression of GPCRs. Although the pseudo signal peptide failed FIGURE 8. The presence of the pseudo signal peptide prevents G i -mediated inhibition of adenylyl cyclase activity independent of receptor expression. A, cell surface expression of constructs CRF 1 R and SP2-CRF 1 R in various stably transfected HEK 293 cell clones expressing different amounts of the receptors. The GFP fluorescence signals of the receptor were colocalized with the plasma membrane marker trypan blue using a confocal LSM, and their intensity was quantified using an 8-bit grayscale and the LSM software (15). Columns represent mean values of plasma membrane GFP fluorescence intensity of 20 cells (ϮS.D.). The quantification is representative of three independent experiments. B, flow cytometry quantification of the cell surface receptors of the cell clones A6-CRF 1 R and B3-SP2-CRF 1 R. Plasma membrane receptors of 10 4 cells were quantified using a monoclonal anti-CRF 1 R antibody directed against the extracellular N tail and a phycoerythrinconjugated goat anti-mouse IgG. Columns represent mean values of three independent experiments (ϮS.D.). C, adenylyl cyclase activity assay using the stably transfected HEK 293 cell clones A6-CRF 1 R and B3-SP2-CRF 1 R (see A). Intact cells were stimulated with increasing concentrations of the agonist sauvagine, and a cAMP RIA was performed. Data points represent geometric mean values of a single experiment performed in duplicate. Curves are representative of two independent experiments. The calculated EC 50 values are (95% confidence limits) as follow: CRF 1 R (ascending slope), 0.46 nM (0.38 -0.56); SP2-CRF 1 R, 0.48 nM (0.36 -0.66).
to initiate ER targeting in various cell types and remained uncleaved (18), it is not excluded at the moment that it may gain conventional signal peptide functions under certain conditions. One may speculate, for example, that a protein factor blocks conventional signal peptide function, e.g. by preventing signal recognition particle binding during early receptor biogenesis. Unknown physiological or pathophysiological conditions may prevent binding of this factor, and a strong up-regulation of the CRF 2(a) R at the cell surface accompanied by the ability to activate G i would be the consequence.