GPR3 Stimulates Aβ Production via Interactions with APP and β-Arrestin2

The orphan G protein-coupled receptor (GPCR) GPR3 enhances the processing of Amyloid Precursor Protein (APP) to the neurotoxic beta-amyloid (Aβ) peptide via incompletely understood mechanisms. Through overexpression and shRNA knockdown experiments in HEK293 cells, we show that β-arrestin2 (βarr2), a GPCR-interacting scaffold protein reported to bind γ-secretase, is an essential factor for GPR3-stimulated Aβ production. For a panel of GPR3 receptor mutants, the degree of stimulation of Aβ production correlates with receptor-β-arrestin binding and receptor trafficking to endocytic vesicles. However, GPR3’s recruitment of βarr2 cannot be the sole explanation, because interaction with βarr2 is common to most GPCRs, whereas GPR3 is relatively unique among GPCRs in enhancing Aβ production. In addition to β-arrestin, APP is present in a complex with GPR3 and stimulation of Aβ production by GPR3 mutants correlates with their level of APP binding. Importantly, among a broader selection of GPCRs, only GPR3 and prostaglandin E receptor 2 subtype EP2 (PTGER2; another GPCR that increases Aβ production) interact with APP, and PTGER2 does so in an agonist-stimulated manner. These data indicate that a subset of GPCRs, including GPR3 and PTGER2, can associate with APP when internalized via βarr2, and thereby promote the cleavage of APP to generate Aβ.


Introduction
Alzheimer's Disease (AD) is a progressive neurodegenerative disorder estimated to affect ,5 million people in the United States and approximately 36 million people worldwide, with numbers predicted to grow further as a result of an aging global population [1][2][3]. Recent advances in molecular pathology and human genetics have reinforced the amyloid hypothesis for the etiology of AD: that the accumulation of Ab peptide (produced by cleavage of APP by BACE1 and the c-secretase complex) is the key initiator of AD pathogenesis [4][5][6][7]. For wild-type APP, cleavage by BACE1 is the rate-limiting step in Ab production [8], but with some mutations found in familial AD -for example, the K670N/ M671L APP695 (Swedish APP) mutant -APP is more readily processed by BACE and the production of Ab is enhanced [9,10]. In addition to efforts to develop clinically useful drugs that inhibit BACE and c-secretase [11][12][13], researchers' attention has also been drawn to indirect modulators of APP processing with a goal of uncovering new potential therapeutic targets.
A cDNA screen for modulators of APP processing uncovered the effects of GPR3 [14], an orphan GPCR most highly expressed in the brain, ovaries and testes [15,16]. GPR3 is a constitutively active G s -coupled receptor that activates adenylyl cyclase, raising intracellular cAMP [17,18]. Thathiah et al showed GPR3 potentiates c-secretase activity and stimulates the production of Ab1-40 and 1-42 in transfected neurons [14]. Further, the authors found a gene dosage-dependent effect of GPR3 on Ab production in vivo, as wild-type, GPR3 heterozygous knockouts and GPR3 homozygous knockouts showed a progressive reduction in soluble Ab levels in hippocampus [14]. Interestingly, although GPR3 exhibits a high constitutive G protein coupling, effects of the receptor on Ab production were independent of G s and cAMP signaling [14].
The finding that GPR3-stimulated APP processing is a G protein-independent process led us to hypothesize that this signaling pathway may involve the b-arrestins. The two b-arrestin isoforms, b-arrestin1 (barr1) and b-arrestin2 (barr2), are ubiquitously expressed adaptor proteins that are recruited to activated GPCRs [19]. Originally the b-arrestins were discovered as key modulators of homologous receptor desensitization, a process that antagonizes the G protein coupling of agonist-occupied GPCRs via phosphorylation by the G protein-coupled receptor kinases (GRKs), leading to b-arrestin recruitment and steric hindrance of G protein activation [20,21]. However, b-arrestins are also essential for endocytosis of receptors via clathrin-coated pits through interactions with clathrin [22] and AP-2 adaptor protein [23]. More recently, it has been shown that b-arrestins coordinate several G protein-independent GPCR signaling cascades [24][25][26]. In these cases, the b-arrestin typically serves as a molecular scaffold, assembling multiple elements of a signaling cascade at activated receptors, thereby regulating the temporal and spatial activity of the pathway.
Here we report that GPR3 can be found in a protein complex with APP, and this interaction is promoted by barr2. Using a set of GPR3 mutants, we show that association of GPR3 with APP correlates with enhanced Ab production, b-arrestin recruitment and localization of the receptor in endocytic vesicles. Testing a wider panel of GPCRs, we found all receptors interact with barrestins, but only GPR3 and PTGER2 showed appreciable interaction with APP and stimulated Ab production. Thus, we propose that a subset of GPCRs is capable of forming a receptor-APP complex in a barr2-dependent manner to facilitate the generation of Ab.

DNA Constructs and Cell Lines
To create the APP-HEK stable cell lines, HEK293 cells were transiently transfected with wild-type (WtAPP-HEK) or Swedish APP (SweAPP-HEK) in the pRK5 Neo vector using Lipofectamine 2000 according to manufacturer's instructions and grown in the presence of 1 mg/mL G418 sulfate for clonal selection. After screening clonal populations for APP expression, selected cell lines were maintained with G418 at 400 mg/mL. Human GPR3, PTGER2 and M1AChR cDNAs were obtained from the Genentech cDNA core facility and subcloned into pcDNA3.1 zeo-(Life Technologies) with an N-terminal leader sequence [27] and FLAG-tag. FLAG-b1AR and FLAG-b2AR constructs were from Robert J. Lefkowitz at Duke University. Rat b-arrestin1 and b-arrestin2 cDNA were also from the Genentech cDNA core facility, subcloned into pcDNA3.1 zeo-and fused in-frame with a C-terminal EGFP tag. Constructs for shRNA knockdown of b-arrestin2 [28] and a firefly luciferase control [29] were created in the pSuper vector, using target sequences as previously described. All GPR3 mutants were created with a QuikChange Multi Site-Directed Mutagenesis Kit (Stratagene) and verified by DNA sequencing.

Cell Culture and Ab ELISA
WtAPP-and SweAPP-HEK cells were maintained in DMEM with 10% fetal calf serum and 1% penicillin/streptomycin and transfected using Lipofectamine 2000 according to manufacturer's instructions. After 72 hours, the culture media was collected and centrifuged at 14,0006 g for 10 minutes at 4uC to remove cell debris. The amount of Ab1-40 in the supernatant was quantified using an MSD ELISA kit, corrected for protein content of the corresponding lysate to adjust for differences in cell density, and normalized to vector controls for each stable cell line. For GPCR agonist experiments, the transfected cells were serum-starved in DMEM +0.4% BSA for at least 1 hour before stimulation.

Co-Immunoprecipitation Assays
SweAPP-HEK cell lines were maintained and transfected as described above. At 48-72 hours post-transfection, the cells were washed twice with ice-cold PBS, lysed in RIPA buffer supplemented with 1% b-octylglucoside, protease and phosphatase inhibitors and tumbled at 4uC. Samples were normalized for total protein concentration and 1 mg per sample was used for immunoprecipitation (IP) with M2 anti-FLAG beads or 22c11 anti-APP followed by protein A/G beads. The beads were washed four times with cold RIPA buffer and incubated with Laemmli sample buffer before separation by SDS-PAGE. For endogenous b-arrestin co-immunoprecipitation with GPCRs, cells were treated with DSP crosslinking reagent before lysis, as previously described [30].

Immunocytochemistry and Image Acquisition
SweAPP-HEK cells or dissociated rat hippocampal cultures (DIV21) were grown on glass coverslips, transfected with the indicated plasmids using Lipofectamine 2000 and fixed in PBS with 4% paraformaldehyde and 4% sucrose 72 hours later. Cells were incubated with primary antibodies in GDB buffer (30 mM phosphate buffer, pH 7.4, containing 0.1% gelatin, 0.3% Triton X-100, and 0.45 M NaCl) at 4uC overnight, washed three times in PBS at room temperature, and labeled with Alexa-conjugated secondary antibodies, followed by another three washes and mounting on glass microscope slides. Images were acquired on a Leica SP5 confocal microscope and binned according to GPR3 subcellular localization with the operator blinded to the transfection conditions. For analysis of receptor clustering, the blinded observer binned transfected cells into clustered or dispersed phenotypes and a minimum of 30 transfected cells were counted per transfection to determine the representative percentages for a given sample, prior to unblinding.

Statistical Analysis
Statistical significance was determined with the aid of Graphpad Prism software using one-way ANOVA and either a Bonferroni (multiple comparisons) or Dunnett (comparing all sets with control) post-hoc test as indicated. All values shown are the mean +/2 SEM.
We also conducted b-arrestin overexpression experiments in a HEK cell line stably expressing wild-type APP (WtAPP-HEK) ( Figure 1E). While the absolute levels of Ab produced were lower than those from SweAPP-HEK cells, GPR3 also enhanced Ab production in WtAPP-HEK cells (1.83+/20.12 fold, compared with WtAPP-HEK transfected with empty vector). However, overexpression of barr2 with GPR3 failed to further enhance Ab levels in these cells (2.00+/20.11). Knockdown of barr2 in WtAPP-HEK cells showed a trend toward reducing Ab production when co-transfected with GPR3 (1.65+/20.05 versus 1.93+/ 20.10 fold for GPR3 plus control shRNA), but this did not reach statistical significance ( Figure 1F). Thus we conclude that GPR3 is capable of boosting wild-type APP processing, but the GPR3enhancing effects of barr2 are only uncovered with the Swedish APP mutant.
Next, we used GPR3 mutants to alter the receptor-b-arrestin interaction and assess the importance of b-arrestin binding on GPR3-stimulated Ab production ( Figure 2). We investigated several mutations in the intracellular domains of GPR3: (i) DRY-AAY, a double point mutant at the base of the third transmembrane domain that impairs the G s -coupling of many GPCRs [31][32][33][34] and reduces GPR3-stimulated cAMP by .85% ( Figure S1); (ii) Q302*, a mutant truncated after the seventh transmembrane segment, eliminating the intracellular carboxyterminal tail required for efficient b-arrestin recruitment to most GPCRs [35]; and (iii) S237A, a point mutant in the third intracellular loop which removes a putative GRK site ( Figure 2A). We examined GPR3-b-arrestin interaction by immunoprecipating the wild-type or mutant receptor (using FLAG antibody) and immunoblotting for co-immunoprecipitated endogenous b-arrestin proteins ( Figure 2B, C). Normalized to wild-type GPR3, both the DRY-AAY (0.48+/20.10) and Q302* (0.25+/20.07) mutants showed reduced recruitment of b-arrestins. Unexpectedly, the S237A mutation resulted in significantly increased b-arrestin association with GPR3 (1.27+/20.08 fold).
We also measured the effect of each GPR3 mutant on Ab production, with and without co-transfected barr2 ( Figure 2D). DRY-AAY in our cell line did not significantly alter Ab production (1.14+/20.05), and overexpression of barr2 with this mutant made no difference (1.10+/20.9). Deletion of the carboxyl tail in the Q302* mutant also prevented the GPR3-mediated Ab increase (1.14+/2.07 fold relative to control for Q302* plus empty vector, and 1.02+/2.7 fold when Q302* was co-transfected with barr2). The S237A mutant stimulated Ab production to a stronger degree than wild-type GPR3, either without (1.68+/2.14) or with barr2- Representative Western blot of endogenous b-arrestins from SweAPP-HEK cells transfected with control (FFLuc) or barr2 shRNA. D) Ab1-40 levels were measured from culture supernatants of SweAPP-HEK293 cells transfected with shRNA plasmids and either empty vector or FLAG-GPR3 as indicated. n = 8, 10, 7, and 10 replicates from left to right. E). Ab ELISA results from WtAPP-HEK cells transfected with GPR3 and barr1 or barr2, as indicated. n = 6, 4, 6, 6, 4 and 6 from left to right. F). Ab levels from WtAPP-HEK cells transfected with luciferase or barr2 shRNA plus FLAG-GPR3 or vector, as indicated. n = 4 for all transfection conditions. Statistical analyses were performed by one-way ANOVA with a Bonferroni post-hoc test comparing all groups with vector-only/luciferase control, and selected comparisons as indicated (*p,0.05, **p,0.01, ***p,0.001). doi:10.1371/journal.pone.0074680.g001 EGFP (1.63+/2.15). Thus, the strength of association with barrestins correlates with the Ab production of the GPR3 variants (S273A.Wt.DRY-AAY = Q302*).
We next examined the subcellular localization of GPR3 in neurons. Mature rat hippocampal neurons were transfected with FLAG-tagged GPR3 after 20 days in culture and immunostained with anti-FLAG antibody 3 days later (DIV20+3). In the majority of cells (68+/23%), wild-type FLAG-GPR3 was observed as intracellular clusters that showed a high degree of co-localization with endogenous APP ( Figure 3A). We found that the GPR3 clusters also partially co-localized with endogenous b-arrestins in many neuronal cell bodies ( Figure S2A) as well as with the early endosomal marker EEA1 ( Figure S2B), the latter being consistent with fractionation experiments in previous studies [36]. In the remainder of the transfected neurons, FLAG-GPR3 showed a more diffuse staining pattern ( Figure 3B). Co-transfection of barr2-EGFP enhanced the ''clustered'' pattern of co-transfected FLAG-GPR3, such that 88+/22% of transfected neurons showed this Figure 2. GPR3 mutants alter Ab production and b-arrestin binding. A). Schematic diagram of GPR3 highlighting in red the mutations used in this study. The diagram was made using the web-based RdBe service [43]. B). Co-immunoprecipiation of endogenous b-arrestins from SweAPP-HEK cells with transfected FLAG-tagged GPR3 (wild-type and mutants). C). Quantification of b-arrestin co-IP with FLAG-GPR3 mutants. n = 5, 5, 5 and 4 independent experiments from left to right. Statistical significance was determined by one-way ANOVA with a Dunnett test comparing all mutants with wild-type GPR3 (*p,0.05, ***p,0.001). D). ELISA measurements of Ab1-40 from SweAPP-HEK293 cells transfected with GPR3 mutants and barr2 as indicated. Data sets for empty vector and wild-type GPR3 are re-plotted from Figure 1B Figure 3C). These data suggest that overexpressed GPR3 is largely present in endosomal compartments and cooverexpression of barr2 promotes GPR3 localization in these compartments, where the receptor can be found in proximity to APP.
Having observed GPR3 co-localization with APP, we asked whether these two molecules could be found together in a protein complex. Because overexpression of barr2 was found to enhance both Ab production and receptor clustering, we also tested whether b-arrestins influenced the formation of such a GPR3-APP complex. We transfected SweAPP-HEK cells with FLAG-tagged GPR3 plus empty vector control, barr1-EGFP, or barr2-EGFP and examined FLAG IPs for co-immunoprecipitation of APP ( Figure 4A-B). With FLAG-GPR3, but not vector control, appreciable amounts of full-length APP were co-immunoprecipitated with FLAG antibodies. Relative to GPR3 alone, cotransfection of barr1 with GPR3 did not significantly increase the amount of APP co-immunoprecipitated with the receptor. On the other hand, co-transfection of barr2 enhanced GPR3-APP association, resulting in 2.78+/20.37 fold more APP co-precipitated than with GPR3 alone. We are unable to tell from these experiments whether the GPR3-APP interaction is direct or indirect. However, it does not appear to be mediated solely through b-arrestin2 binding, as APP was not found to coimmunoprecipitate with b-arrestins pulled down via their EGFP tag ( Figure S3).
Do the GPR3 mutations that affect b-arrestin binding and stimulation of Ab production also show differences in association with APP? To answer this question, FLAG-tagged wild-type GPR3, DRY-AAY, Q302*, and S237A mutants were expressed in SweAPP-HEK cells and immunoprecipitated 3 days post-transfection ( Figure 4C, D). DRY-AAY showed reduced APP co-IP (0.68+/20.09 relative to wild-type GPR3), while the interaction of Q302* with APP was even more robustly attenuated (0.47+/ 20.05). In contrast, S237A showed an elevated interaction with APP (1.50+/20.15 fold increase). Thus among the GPR3 variants, formation of a GPR3-APP complex is positively correlated both with b-arrestin recruitment and Ab production.
We confirmed the biochemical association of GPR3 and APP by performing the reciprocal IP reaction, immunoprecipitating APP from SweAPP-HEK cells transfected with FLAG-GPR3 alone, FLAG-GPR3 plus barr2-EGFP, or barr2-EGFP with FLAG-b2-adrenergic receptor (b2AR; another G s -coupled GPCR). GPR3 was readily found in the APP immunoprecipitates and this interaction was enhanced by co-transfection of barr2-EGFP (1.39+/20.09 fold increase versus to GPR3 alone) ( Figure 4E, F). Notably, b2AR was not detected above background in the APP immunoprecipitates (0.03+/20.01 ( Figure 4E, F). These data confirm that GPR3 can interact with APP, and does so with some specificity, insofar as another GPCR did not associate with APP under the same conditions.
The discovery that APP may selectively interact with some GPCRs, like GPR3, but not others, such as the b2AR, led us to screen a broader panel of GPCRs for Ab production and APP interactions ( Figure 5). We transfected FLAG-tagged GPR3, b1adrenergic receptor (b1AR), b2AR, M1 muscarinic acetylcholine receptor (M1AChR), prostaglandin receptor PTGER2, or empty vector into SweAPP-HEK cells, first looking at the ability of these receptors to stimulate Ab production under basal culture conditions ( Figure 5A). Testing the culture supernatant 3 days post-transfection, only GPR3 (1.49+/20.09 fold relative to control) and PTGER2 (1.23+/20.03 fold) showed significant increases in Ab, while the b1AR (1.08+/20.05), b2AR (1.07+/ 20.06), and M1AChR (0.86+/20.06) were not significantly different from vector control. We find this dichotomy extends to the GPCR-APP complex as well ( Figure 5B-C). APP coimmunoprecipitated with GPR3, but not with b1AR, b2AR and M1AChR, which did not enhance Ab production. The degree of co-IP of APP with PTGER2 was 46+/23% of that seen with GPR3, which corresponds with the more modest enhancement of Ab production by PTGER2. However, when we examined the receptor IPs for b-arrestin recruitment, we found that M1AChR (4.77+/2.75 fold relative to GPR3) and b1AR (2.31+/0.91) showed the highest levels of b-arrestin co-IP under non-stimulated cell culture conditions. Co-IP of endogenous b-arrestins for b2AR (0.75+/20.17) and PTGER2 (1.11+/20.15) were not significantly different from GPR3. Thus there is no correlation between interaction with endogenous b-arrestins and the ability of a GPCR to stimulate Ab production. Rather, our data indicate that Ab production is correlated with a GPCR's interaction with APP.
We also considered the possibility that agonist stimulation may be necessary for some GPCRs to promote an interaction with APP and promote its processing. To this end, we utilized the b1AR and looked at Ab levels, b-arrestin recruitment and APP co-IP in the presence or absence of 10 mM isoproterenol stimulation for 30 minutes (Figure 6). Consistent with the experiments above, GPR3 potentiated Ab production (1.49+/20.27 fold relative to vector control) in the SweAPP cell line, but b1AR had no significant effect (1.08+/20.14); this lack of effect did not change with isoproterenol stimulation of the receptor (1.19+/20.10) ( Figure 6A). Moreover, overexpression of barr2 did not enhance Ab production when co-transfected with b1AR in either the unstimulated (1.08+/20.20) or isoproterenol-stimulated conditions (1.15+/20.20). When we examined the receptor IPs for interaction with APP, only GPR3 (0.65+/20.08 fold relative to GPR3+barr2) and the GPR3+barr2 co-transfected samples showed substantial co-IP with APP ( Figure 6B, C). We observed a b1AR-b-arrestin interaction under basal conditions (1.63+/ 20.07 fold over GPR3) that increased with agonist stimulation of the b1AR (3.30+/20.54) ( Figure 6D). As expected, overexpression of barr2-EGFP robustly increased the amount of total b-arrestin in the b1AR receptor co-IP (7.02+/21.16) and the difference increased even further when b1AR was co-transfected with barr2 and stimulated with isoproterenol (11.94+/21.65) ( Figure 6D, E). Thus for the b1AR, a prototypic GPCR, agonist stimulation and barr2 binding do not correlate with interaction with APP or elevation of Ab levels.
What is the effect of agonist stimulation of a receptor like PTGER2, which does stimulate APP processing under basal conditions ( Figure 5A)? To address this question, we transfected SweAPP cells with barr2-EGFP and co-transfected with empty vector, GPR3, b1AR, b2AR or PTGER2. The cultures were serum starved and then stimulated for 30 minutes with their respective agonists (10 mM isoproterenol for beta-adrenergic receptors, or 10 mM PGE2 for PTGER2). However, under these conditions, only GPR3 (1.87+/20.39 fold over vector control) promoted a significant increase in Ab levels ( Figure 7A). At 30 min of agonist stimulation, we also saw minimal interaction of b1AR, b2AR or PTGER2 with APP and no agonist effects in receptor co-IP experiments ( Figure 7B, C). However, when transfected cells were stimulated for a prolonged period with agonist (16 hours), we found that PTGER2-transfected cells stimulated with PGE2 (2.06+/20.27) increased Ab to levels comparable to GPR3 (2.27+/20.22) ( Figure 7D). Prolonged PGE2 stimulation also potentiated the biochemical association of PTGER2 with APP ( Figure 7E, F). No such effect was observed for either b1AR or  -arrestin. A). SweAPP-HEK cells, transfected as indicated, were immunoprecipitated with FLAG antibody and immunoblotted for APP or FLAG. Arrowhead indicates the band for APP. B). Quantification (by densitometry) of co-IP of APP with FLAG-GPR3. n = 6, 3 and 5 from left to right. Statistical significance was determined by one-way ANOVA with a Bonferroni post-hoc test comparing all data sets (*p,0.05, ***p,0.001). C). Co-immunoprecipitation of APP with FLAG-GPR3 mutants. SweAPP-HEK cells were transfected as indicated and blotted for immunoprecipitated FLAG and co-immunoprecipitated APP as indicated. D) Densitometry for APP co-immunoprecipitated with FLAG-GPR3 constructs. n = 6, 6, 6 and 5 independent experiments left to right. Statistical significance was determined by one-way ANOVA with a Dunnett test comparing all columns with wild-type GPR3. (*p,0.05, **p,0.01, ***p,0.001). E). GPR3, but not b2AR, coimmunoprecipitates with APP. SweAPP-HEK cells were transfected as indicated and lysates immunoprecipitated as indicated. Arrowhead indicates the band for heavy-chain IgG from the 22c11 (anti-APP) IP reaction. F). Quantification of FLAG-GPCR co-IP with APP. n = 5 for all data sets. Statistical significance was determined by one-way ANOVA with a Dunnett test comparing all columns with wild-type GPR3. (***p,0.001). doi:10.1371/journal.pone.0074680.g004 b2AR with isoproterenol. Collectively, these data indicate that agonist stimulation can promote APP interaction and APP processing for a subset of GPCRs like PTGER2. On the other hand, GPR3, which has high constitutive activity, can associate with APP and enhance Ab production in the absence of an exogenous agonist.
Looking across all parameters measured, we examined the correlations of our data sets to better understand the mechanisms of GPCR-stimulated APP processing (Figure 8). For wild-type GPR3 and the GPR3 mutants, Ab production was positively correlated with b-arrestin recruitment to the receptors ( Figure 8A), with intracellular clustering of the receptor in neurons and in SweAPP-HEK cells ( Figure 8B,C), and with co-IP of APP with the receptor ( Figure 8D). However, when we look at the broader panel of GPCRs, APP co-immunoprecipitation with the GPCR correlates positively with stimulation of Ab production, whereas barrestin recruitment is non-correlative ( Figure 7E,F). Further, as expected for GPCR internalization, the GPR3 clustering in  . Agonist stimulation of b1AR promotes b-arrestin recruitment but not interaction with APP or Ab production. A). Ab1-40 ELISA from SweAPP-HEK cells transfected with FLAG-GPR3 or FLAG-b1AR and barr2-EGFP and treated with isoproterenol (ISO, 10 mM, 30 min) as indicated. n = 3 for all conditions. Statistical significance was determined by one-way ANOVA with a Dunnett post-hoc test comparing all treatments with unstimulated vector control. (*p,0.05, ***p,0.001). B). APP co-IP with FLAG-GPR3 or FLAG-b1AR from SweAPP-HEK cells transfected and treated as indicated. C). Quantification (by densitometry) of APP co-IP with FLAG-GPCR IPs. n = 3, 3, 3, 3, 6 and 6 from left to right. Statistical significance was determined by one-way ANOVA with a Bonferroni post-hoc test comparing all conditions with GPR3+barr2 and select comparisons as indicated. (*p,0.05, ***p,0.001). D). b-arrestin co-IP with FLAG-GPR3 or FLAG-b1AR from SweAPP-HEK cells transfected and treated as indicated.
Open arrowheads indicate transfected barr2-EGFP. Closed arrowheads mark endogenous b-arrestin1/2. E). Quantification (by densitometry) of coimmunoprecipitated b-arrestins with FLAG-GPR3 or FLAG-b1AR. n = 3 for all treatments. Statistical significance was determined by one-way ANOVA with a Dunnett post-hoc test comparing all treatments with unstimulated vector control. (*p,0.05, ***p,0.001). doi:10.1371/journal.pone.0074680.g006 endosomes is also correlated with the ability of the receptor to bind b-arrestins ( Figure S4). Thus, we conclude that the ability of GPCRs to stimulate Ab production is related to their capability to interact with APP and that b-arrestin recruitment, while necessary, is not sufficient to drive GPCR-stimulated processing of APP.

Discussion
Here we report that a GPCR-APP complex is formed by GPR3 and agonist-stimulated PTGER2 (GPCRs that potentiate APP processing), but not formed by receptors that do not affect Ab production (such as b1AR or M1AChR). This selective interaction offers a resolution to the question: How can barr2 be integral to GPCR-mediated Ab production, when it interacts nearly universally with GPCRs, and yet only a subset of GPCRs enhances APP cleavage? Our data support the idea that barr2 is crucial for GPCR-mediated enhancement of APP processing, but they suggest that formation of a GPCR-APP complex is also fundamental for receptor-stimulated Ab generation. The degree of the GPCR-APP interaction more closely reflects a receptor's Ab production potential than does b-arrestin binding, and thus formation of a receptor-APP complex defines a new subclass of GPCRs that can promote Ab processing.

Relation to other Studies of GPCR-stimulated Ab Production
Several reports have described stimulation of Ab production by GPCRs including PTGER2 [37], the serotonin receptor 5HTR2C [38], thyrotropin-releasing hormone receptor [39], and the a 2aadrenergic receptor [39]. Thathiah et al discovered GPR3stimulated APP processing in a screen for modulators of Ab production, showing GPR3 stimulates c-secretase activity via signaling pathways independent of the receptor's constitutive G s activity [14]. More recently, the same group found elevated barr2 levels in human AD tissue samples, and moreover, barr2 knockout mice, but not barr1 knockouts, showed lower Ab levels in the hippocampus and cortex when crossed with an AD mouse model [36]. These findings agree with our data that barr2 overexpression potentiates, and barr2 knockdown mitigates, the production of Ab in SweAPP cells ( Figure 1A-B). However, barr2 shRNA knockdown did not significantly reduce Ab production in WtAPP cells ( Figure 1F). One explanation is that barr2 protein has a relatively long half-life of 12-15 h [40], thus the knockdown of endogenous barr2 protein is likely to be incomplete during the time period when GPR3 is co-transfected and expressed. In a genetic experiment, Thathiah et al [36] showed that GPR3 does not increase Ab production in barr2 knockout neurons, which strongly supports the idea that barr2 is necessary for GPR3-stimulated Ab production.
One proposed molecular mechanism for GPR3 stimulation of APP processing is through b-arrestin2-mediated recruitment of active c-secretase complex via an interaction with the Aph-1a subunit [36]. This model is consistent with reports of internalization-dependent, G protein-independent increases in c-secretase activity induced by the b2AR [41] and d-opioid receptor [42]. Though b-arrestins are not specifically investigated in these studies, a role for barr2-mediated internalization and recruitment of c-secretase to the activated GPCRs fits the data. A notable discrepancy is that in our cells, possibly due to differences in the experimental systems, b2AR does not enhance Ab production. We see a synergistic effect of GPR3 and barr2 on Ab production for SweAPP, but not WtAPP ( Figure 1B,E). This finding supports the hypothesis that the effect of barr2 is to enhance c-secretase activity, because for SweAPP (a much better substrate for betasecretase) c-secretase is more rate limiting -in contrast to wildtype APP, for which BACE cleavage is rate limiting.

Molecular Mechanisms of the GPR3 Mutants
Data from our GPR3 mutants implicate the formation of a receptor-APP complex and internalization by barr2 as processes involved in GPR3-stimulated Ab production. However, the mechanisms by which these mutations alter b-arrestin recruitment and APP binding are not completely understood. Only the Q302* cytoplasmic tail truncation mutant, intended to reduce interactions with b-arrestins, behaved as initially hypothesized. The S237A mutant was created to remove a putative GRK phosphorylation site and thereby reduce b-arrestin recruitment. Instead, this mutant showed enhanced b-arrestin binding, along with stronger stimulation of Ab production and enhanced co-IP of APP. The fact that Ab production by S237A could not be further enhanced by barr2 overexpression is consistent with the idea that this mutant binds endogenous b-arrestins so well that this interaction is no longer limiting. S237A may promote a receptor conformation favoring b-arrestin binding, or alternatively, this residue may be a site of inhibitory phosphorylation that is relieved by the S237A mutation.
Multiple lines of evidence indicate GPR3-stimulated Ab production is G protein-independent [14]. Thus, we expected that DRY-AAY, a mutation that reduces G protein coupling, would exhibit Ab production equivalent to wild-type GPR3. Instead, the DRY-AAY mutant showed diminished stimulation of Ab production, which correlated well with reduced b-arrestin recruitment and reduced intracellular clustering of this mutant. Further complicating the issue, Thathiah et al also used a DRY-AAY mutant and observed enhanced b-arrestin binding and Ab production [36]. This discrepancy may be due to differences in the respective SweAPP-HEK cell lines. However, both reports agree that Ab production by mutant GPR3 receptors correlates with barrestin binding.

Implications of GPCR-APP Complex Formation
We find the formation of a protein complex including GPR3 and APP is associated with enhanced Ab production and this is the only parameter tested that correlates with APP processing across multiple GPCRs. While this interaction may define a new subset of Ab-modulating GPCRs, critical questions about the nature of this complex remain to be addressed. First, does APP interact directly or indirectly with GPCRs such as GPR3? Future experiments using mass spectrometry of APP immunoprecipitates may identify additional GPCRs or regulators of the receptor-APP protein complex. Unfortunately, because of a lack of good antibodies and low endogenous expression of GPR3, our study and many others rely on epitope-tagged GPCRs, whose overexpression may lead to artifactual interactions. Second, what are the specific domains and residues required for GPCRs to interact with APP? Experimentally, this could be addressed with chimeric receptors made from GPR3 and a receptor such as b1AR that does not stimulate Ab production. Finally, what is the specific role for b-arrestin in the receptor-APP interaction? Does it participate as a scaffold in the formation of a receptor-barr2-APP ternary complex, or does the b-arrestin-dependent trafficking of GPCRs like GPR3 into endosomes enrich the local concentrations of receptor and APP, promoting their interaction? Experiments in barr2 knockout cells should verify whether barr2 is required for GPR3-APP binding. Finally, we note that PGE2 failed to stimulate APP-PTGER2 association at 30 minutes, a time point when GPCR endocytosis should be occurring. This suggests the APP-receptor complex may be created in the endosomes after internalization, rather than forming at the plasma membrane. More detailed PGE2 time courses with and without inhibition of endocytosis may provide insights into the timing and processes required to form the PTGER2-APP complex. Ultimately, these experiments will provide a deeper understanding of the molecular nature of the GPCR-APP interaction and potentially offer new targets for therapeutic intervention in AD. Figure S1 The DRY-AAY mutation impairs GPR3-stimulated cAMP production. SweAPP-HEK cells were transfected with either empty vector (pcDNA3), GPR3, or DRY-AAY. Two days later, the cells were treated with 10 mM IBMX for 30 minutes to inhibit phosphodiesterases. The cells were lysed and cAMP accumulation was quantified using an HTRF ELISA kit. n = 3, 4 and 3 from left to right. Statistical significance was determined by one-way ANOVA with a Bonferroni post-hoc test, comparing all columns. (*p,0.05, **p,0.01). (TIF) Figure S2 GPR3 clusters partially co-localize with endogenous b-arrestins and the endosomal marker EEA1. Dissociated rat hippocampal neurons were grown on glass coverslips and prepared as described in Materials and Methods. The cells were incubated with anti-FLAG and either A) anti-b-arrestin1/2 or B) anti-EEA1 (middle panel) primary antibodies as indicated. Representative confocal images are shown. Merged images show FLAG in the red channel and either b-arrestin1/2 or EEA1 in green, with yellow pixels indicating colocalization. (TIF) Figure S3 APP co-immunoprecipitates with GPR3, but not with b-arrestin1/2. SweAPP-HEK cells were transfected with empty vector, barr1-EGFP, barr2-EGFP or FLAG-GPR3 as shown and FLAG-or EGFP-immunoprecipitations were blotted for co-immunoprecipitated APP (upper) and levels of immunoprecipitation for the bait proteins (lower). (TIF) Figure S4 GPR3 clustering is a function of b-arrestin recruitment. Correlation graph showing the fraction of SweAPP-HEK cells transfected with the indicated GPR3 mutants in a clustered staining pattern as a function of co-IP with endogenous b-arrestins. (TIF)