GPR37 Protein Trafficking to the Plasma Membrane Regulated by Prosaposin and GM1 Gangliosides Promotes Cell Viability*

Background: Intracellular GPR37 accumulation is neurotoxic and associated with parkinsonism, whereas plasma membrane association is protective. Results: The endogenous GPR37 ligand prosaposin promotes GPR37 surface density and association with GM1-enriched lipid rafts. Conclusion: GPR37, prosaposin, and GM1 constitute a pathway that improves cell viability through GPR37 trafficking to the plasma membrane. Significance: Targeting this pathway could reduce toxic intracellular GPR37 accumulation observed in parkinsonism. The subcellular distribution of the G protein-coupled receptor GPR37 affects cell viability and is implicated in the pathogenesis of parkinsonism. Intracellular accumulation and aggregation of GPR37 cause cell death, whereas GPR37 located in the plasma membrane provides cell protection. We define here a pathway through which the recently identified natural ligand, prosaposin, promotes plasma membrane association of GPR37. Immunoabsorption of extracellular prosaposin reduced GPR37tGFP surface density and decreased cell viability in catecholaminergic N2a cells. We found that GPR37tGFP partitioned in GM1 ganglioside-containing lipid rafts in the plasma membrane of live cells. This partitioning required extracellular prosaposin and was disrupted by lipid raft perturbation using methyl-β-cyclodextrin or cholesterol oxidase. Moreover, complex formation between GPR37tGFP and the GM1 marker cholera toxin was observed in the plasma membrane. These data show functional association between GPR37, prosaposin, and GM1 in the plasma membrane. These results thus tie together the three previously defined components of the cellular response to insult. Our findings identify a mechanism through which the receptor's natural ligand and GM1 may protect against toxic intracellular GPR37 aggregates observed in parkinsonism.

tion assays, and PSAP and prosaptides stimulated G i/o and ERK1/2 signaling via GPR37. Moreover, GPR37 was shown to mediate PSAP-induced protection against H 2 O 2 toxicity in primary astrocytes.
It is also known that PSAP localizes to lipid rafts containing GM1 and GM3 gangliosides at plasma membranes (21,22) and that ganglioside-containing lipid rafts are required for PSAPactivated GPCR signaling and neurotrophic effects (22). This study evaluated the hypothesis that PSAP, prosaptides, and gangliosides may regulate the plasma membrane levels of GPR37 suggested to be crucial for neuroprotective actions of GPR37. For this purpose, catecholaminergic N2a cells stably expressing GPR37 genetically fused with the turbo green fluorescent protein (GPR37 tGFP ) were studied using various approaches, most notably confocal microscopy and fluorescence correlation spectroscopy (FCS) in live cells. The FCS technique is unique in its ability to quantitatively characterize molecular diffusion and complex formation in live cells.
Immunoblotting-Samples were subjected to SDS-PAGE on 10% acrylamide gels and then transferred to Immobilon-P membranes (Millipore, Billerica, MA). Membranes were blocked in 5% milk (Tris-buffered saline, 5% fat-free milk powder, 0.12% Tween 20) for 1 h and then incubated with primary antibodies against prosaposin (ab68466, Abcam, Cambridge, UK) or GPR37 (14820-1-AP, Proteintech, Manchester, UK) diluted in 5% milk. Membranes were then washed and incubated with HRPlinked or fluorescently labeled secondary antibodies for film or fluorescence detection, respectively, followed by a second washing. For film detection, membranes were incubated with ECL Plus substrate (PerkinElmer Life Sciences), and the signal was detected on BioMax XAR film (Eastman Kodak). For fluorescence detection, the signal was read on a LI-COR Odyssey CLx (LI-COR Biosciences, Cambridge, UK).
MTT Cell Viability Assays-Cells were seeded in 96-well plates at a density of 1 ϫ 10 4 cells per well and incubated at 37°C, 5% CO 2 for 3 days before addition of prosaposin (PSAP)specific antibody anti-PSAP (sc-32876, Santa Cruz Biotechnology, Dallas, TX) at 40 g ml Ϫ1 or anti-PS769 (generated as described previously (21)) at 8.5 g ml Ϫ1 . Cytotoxicity was assayed 8, 16, and 24 h later or in a separate experiment 48 h later. MTT (0.5 mg ml Ϫ1 ) was added to the medium, and cells were incubated for 1 h at 37°C, 5% CO 2 . The medium was removed, and cells were lysed in isopropyl alcohol containing 0.04 M HCl. Absorbance was measured at 550 nm using an Anthos 2020 microplate reader (Biochrom, Berlin, Germany). Average values for untreated controls within each cell type and time point were set to 100%. Values measured in treatment groups were expressed as a percentage of control within cell type and time point. Data were pooled from three independent experiments.
Immunoprecipitation of PSAP-Cells were seeded in 8-well chambers at a density of 2 ϫ 10 4 cells per well and incubated at 37°C, 5% CO 2 for 3 days before addition of PSAP-specific antibody anti-PSAP (sc-32876, Santa Cruz Biotechnology, 40 g ml Ϫ1 ) or normal IgG (12-370, Millipore, 40 g ml Ϫ1 ) as a negative isotype control. After incubation for 2.5 h at room temperature, medium was removed from the cells and incubated with protein A-coated magnetic beads to pull down the antibodies. Beads and media were separated, and the medium was subjected to Novex SDS-PAGE on 4 -12% BisTris gels (Invitrogen) followed by transfer to Immobilon-P membranes (Millipore). Membranes were subjected to immunoblotting as described above using a PSAP-specific primary antibody (1:200, sc-32876, Santa Cruz Biotechnology) and conformation-specific HRP-labeled anti-rabbit secondary antibody (1:4000, 5127S, Cell Signaling Technology).
Fluorescent Peptides-The PSAP-derivative prosaptide TX14(A) (PS-TX14(A) TAMRA ), and ␤-endorphin (␤-End TAMRA ) peptides fluorescently labeled at the C terminus with carboxytetramethylrhodamine (TAMRA), were purchased from Biomatik (Wilmington, DE). The primary sequences of the peptides were T(D-Ala)LIDNNATEEILY and YGGFMTSEKSQTPLVTLFK-NAIIKNAYKKGQK for PS-TX14(A) and ␤-endorphin. Both peptides were purified by the producer to a purity level of Ͼ98% using HPLC with gradient elution of water/acetonitrile and 0.1% trifluoroacetic acid. The molecular masses of the peptides were 2.1 and 4 kDa for PS-TX14(A) TAMRA and ␤-End TAMRA , respectively, as determined by electrospray ionization-mass spectrometry. Peptides were stored in powder at Ϫ20°C and were used without any further purification. Aqueous solutions of the peptides were prepared just before cell treatment. 5-15-l aliquots of the peptide solution were added to the cell culture medium (350 l) and gently mixed. In experiments with PS-TX14(A) TAMRA and ␤-End TAMRA , the medium was replaced twice with a 45-min interval before addition of the peptide. This was done to reduce the amount of endogenous PSAP in the medium and its potential interference with fluorescently labeled peptides binding to GPR37.
Cholesterol Sequestration-Membrane cholesterol was sequestered by treatment with 4 mM methyl-␤-cyclodextrin (Sigma) for 45 min or 4 units ml Ϫ1 cholesterol oxidase (Sigma) for 1 h. Both treatments were performed at room temperature.
CLSM and FCS/FCCS-CLSM imaging, FCS/FCCS, and fluorescence recovery after photobleaching (FRAP) measurements were performed using the LSM510 ConfoCor3 instrument (Carl Zeiss, Jena, Germany) individually modified to enable fluorescence imaging with silicone avalanche photodiodes (SPCM-AQR-1X; PerkinElmer Life Sciences) (23). Background on FCS/FCCS measurements and data analysis are described below and in Figs. 1 and 2. Image adjustments only included brightness and/or contrast adjustments applied to the entire image and to corresponding controls. Nonlinear image processing and filtering were not applied.
Optical Setting-tGFP fluorescence was excited using the 488-nm line of the Ar/ArKr laser, and the 543-nm He/Ne laser was used for Alexa Fluor 594 and TAMRA. Emitted light was separated from the incident light using the main dichroic beam splitter HFT 488/543 and split using the secondary dichroic beam splitter NFT 545. Further spectral separation of the emitted light was achieved using specialized filters in front of the detectors. For imaging, emitted light was collected in the multitrack acquisition mode using the bandpass filter BP 505-530 for tGFP and the bandpass filter BP 615-680 for Alexa Fluor 594 and TAMRA. Images were acquired without averaging, using a scanning speed of 51.2 s/pixel and a 512 ϫ 512 pixels frame. For FCS/FCCS, emitted light was collected using the bandpass filter BP 505-530 for tGFP and the long pass filter LP 580 or bandpass filter BP 615-680 for Alexa Fluor 594 and TAMRA. All FCS/FCCS measurements were performed on the plasma membrane that is at the apical end of the cell, above the fraction of the plasma membrane that is adhering to the coverslip and the virtually transparent nucleus. The apical plasma membrane was identified by a fluorescence intensity scan across the vertical axis of the cell (Fig. 1A). Fluorescence intensity fluctuations were recorded in arrays of 10 -50 consecutive measurements, each measurement lasting 10 -20 s. Autocorrelation curves (Fig. 1, B and C) were individually inspected to exclude traces where the passage of bright transporting intracellular vesicles interfered with the FCS measurements (Fig.  1C). Thus, the generated average autocorrelation curves ( Fig.  2A) were analyzed using the dedicated ConfoCor3 Zeiss software (Carl Zeiss, Jena, Germany).
Temperature and Atmosphere Control-CLSM imaging, FCS/FCCS, and FRAP were performed under controlled temperature (37°C) and atmospheric conditions (humid air with 5% CO 2 ) and at room temperature (20°C) and atmospheric conditions. Temperature and atmosphere regulations were achieved using a heating insert at the microscope stage (Heat Insert P, PeCon GMBH), temperature regulating unit (TempControl 37-2 digital, PeCon GMBH), and an atmosphere controlling unit (CTI-Controller 3700 digital, PeCon GMBH) with an IR absorption-based CO 2 sensor for continuous monitoring of CO 2 concentration in the heated circulating air stream. Differential effects of temperature/atmosphere on GPR37 trafficking were not observed. All measurements reported here were performed at room temperature.
Background on FCS-FCS is a quantitative method with single molecule sensitivity that uses statistical approaches in the form of temporal autocorrelation and/or photon-counting histogram analysis of fluorescence intensity fluctuations recorded over time to obtain information about molecular numbers and their transporting properties (temporal autocorrelation) or molecular brightness distribution (photon-counting histogram) (24). In this study, fluorescence intensity fluctuations recorded over time (Fig. 1B, inset) were analyzed using temporal autocorrelation to extract information about the average number of molecules in the observation volume element, i.e. the concentration, and the diffusion times. For this purpose, the normalized autocorrelation function G() was first derived as shown in Equation 1, G() gives the correlation between the deviation of fluorescence intensity measured at a certain time point t, ␦I(t) ϭ I(t) Ϫ ͗I(t)͘, which is given as the difference in fluorescence intensity I(t) and the mean fluorescence intensity over the recorded time series (͗I(t)͘), and its intensity measured at a later time, ␦I(t ϩ ) ϭ I(t ϩ ) Ϫ ͗I(t)͘. G() is then plotted as a function of different lag times , yielding the experimental autocorrelation curve (see for example Fig. 1B).
To extract the molecular numbers (N) and diffusion times ( D ), the experimental autocorrelation curves were fitted using autocorrelation functions derived for appropriate model systems (25,26). For measurements performed in the cell cultur- ing medium, FCS curves were fitted using a model for free three-dimensional diffusion with triplet contribution as shown in Equation 2, For measurements performed at the plasma membrane, a model for free two-dimensional diffusion with triplet contribution was used as shown in Equation 3, In Equations 2 and 3, T is the average equilibrium fraction of molecules in the triplet state; T is the triplet state relaxation time; i is the number of components; Di is the diffusion time of the ith component, and y i is its relative amplitude (⌺y i ϭ 1). In Equation 2, w xy and w z are the distances from the center of the laser beam focus in the radial and axial directions, respectively, where the collected fluorescence intensity has dropped by a factor of e 2 compared with its peak value for the Gaussian beam profile. The parameter (w xy /w z ) 2 is determined by instrument calibration using standard aqueous solutions of rhodamine 6G for ex ϭ 488 nm and TAMRA for ex ϭ 543 nm. Data fitting using Equation 2 for a single component (i ϭ 1) yielded diffusion times D, Rh6G ϭ 27 Ϯ 3 s and D, TAMRA ϭ 33 Ϯ 3 s and structural parameters 7 for ex ϭ 488 nm and 8 for ex ϭ 543 nm. For fitting the experimental autocorrelation curves, the simplest model, i.e. the model with the lowest number of components, was always used. For a first hand characterization of twodimensional diffusion, the instrument running software was used (5,9).
Analysis of Autocorrelation Curves for GPR37 tGFP and Interpretation of Autocorrelation Decay Times-Autocorrelation curves for GPR37 tGFP recorded at the apical plasma membrane show several characteristic decay times (Fig. 1B). To identify the most appropriate model to evaluate the experimental autocorrelation curves, we needed to first identify the underlying processes that lead to fluorescence intensity fluctuations. In general, fluorescence intensity fluctuations can be generated by molecular diffusion, chemical interactions, and/or photo-physical processes, such as intersystem crossing (singlet to triplet state transition) (27). To identify the contribution of these processes in our system, we varied the size of the observation volume element and investigated the effect of the observation volume size on the observed decay times (28,29). In this assay, the size of the observation volume element from which the photons FIGURE 2. Fluorescence intensity fluctuations recorded at the apical plasma membrane of N2a cells expressing GPR37 tGFP are generated by molecular movement. A, autocorrelation curves recorded at varying observation volume sizes show that the amplitudes of the GPR37 tGFP autocorrelation curves decrease as the pinhole size in front of the detector is increased, reflecting an increase in the number of molecules in the observation volume element. B, for varying sizes of the observation volume element, a linear relationship was observed between the number of observed molecules (N) and the (pinhole size). C, autocorrelation curves at two different pinhole sizes show that characteristic times increase with increasing the observation volume size. D, linear relationship was observed between the first characteristic time ( D1 ) and the (pinhole size). The positive intercept suggests that partitioning between domains may occur. E, linear relationship was observed between the second characteristic time ( D2 ) and the (pinhole size). The negative intercept suggests that some confinement by the cytoskeletal protein meshwork may also occur. F, linear relationship between the number of molecules characterized by slow (N 2 ) and fast (N 1 ) diffusion was observed for different surface densities of GPR37 tGFP .

Subcellular GPR37 Sorting by Prosaposin and GM1
FEBRUARY 21, 2014 • VOLUME 289 • NUMBER 8 are collected is changed by changing the size of the pinhole in front of the detector. If fluorescence intensity fluctuations are generated by molecular diffusion, the characteristic decay time of the autocorrelation function should increase as the size of the observation volume element increases. In contrast, if fluorescence intensity fluctuations arise as a consequence of photophysical or chemical transformations, such as transition to the triplet state or a chemical reaction where fluorescence is gained/lost, the characteristic decay time of the autocorrelation function would not be affected (27).
We observed that the autocorrelation curves recorded at the apical plasma membrane shift to longer decay times when the size of the observation element is increased (Fig. 2, C-E), indicating that fluorescence intensity fluctuations characterized with decay times Ͼ100 s are generated by molecular movement. Accordingly, a two-component diffusion model with triplet contribution (Equation 2, i ϭ 2) was used for fitting the experimental autocorrelation curves. The two-component diffusion model yielded diffusion times D1 ϭ (3.5 Ϯ 1.5) ϫ 10 Ϫ4 s and D2 ϭ (2.5 Ϯ 1.5) ϫ 10 Ϫ2 s. For all cells investigated (n ϭ 25), the relative contribution of the second component was y 2 ϭ 0.17 Ϯ 0.08 (Fig. 2F). The indicated standard deviation reflects differences between cells and not between different measurements on the same cell.
The pinhole size scan assay also showed that both diffusion times D1 and D2 increase as the size of the observation volume element is increased, but they follow different diffusion laws (30). Thus, a positive intercept was observed for D1 ϭ f (ph 2 ) (Fig. 2D), whereas a small negative intercept was observed for D2 ϭ f (ph 2 ) (Fig. 2E). This indicates that GPR37 tGFP lateral organization and dynamics in the plasma membrane is complex, involving partitioning between microdomains and possibly some confinement by the cytoskeletal protein meshwork (30,31). It is also possible that the faster component originates from an intracellular pool of GPR37 tGFP , as was suggested for serotonin 5-HT 1A (31). However, one would expect in this case that the fraction of molecules characterized by fast or slow diffusion would be independent from one another, showing different relative contributions for different cells. The fact that a linear relationship between the number of molecules characterized by slow (N 2 ) and fast (N 1 ) diffusion was observed for different surface densities of GPR37 tGFP (Fig. 2F) supports the interpretation that a dynamic sorting mechanism is observed at the plasma membrane. In addition, unpublished studies from our laboratory 3 on other GPCRs show that the relative contribution of one fraction with respect to the other is different for different GPCRs. Taken together, these observations strengthen the interpretation that both diffusion times reflect processes at the plasma membrane.
This interpretation is further supported by our observation that a fraction of GPR37 tGFP interacts with GM1 and resides in GM1-enriched domains (see under "Results"). If GPR37 tGFP sorting between the lipid bilayer and GM1-enriched domains is slow compared with the later diffusion time of GPR37 tGFP in the lipid bilayer, two components will be detected by FCS, with the fast one reflecting GPR37 tGFP diffusion in the lipid bilayer and the slow one reflecting the diffusion of the GM1-enriched domains. (A detailed explanation for chemical reactions in three dimensions can be found in Ref. 27.) Recent studies on model membranes verified that peptides/protein can spontaneously partition between liquid disordered and solid ordered raft-like microdomains (32). This study also showed that the diffusion coefficient of the investigated peptides/protein in the fluid phase is in the order of a few m 2 s Ϫ1 , whereas the diffusion coefficient of the microdomains is about 1 ϫ 10 Ϫ1 m 2 s Ϫ1 (32). The difference between the experimentally measured diffusion times for GPR37 tGFP , D1 ϭ (3.5 Ϯ 1.5) ϫ 10 Ϫ4 s, and D2 ϭ (2.5 Ϯ 1.5) ϫ 10 Ϫ2 s, falls in the same range. However, we cannot unequivocally determine the diffusion coefficients based on the lateral diffusion times, because GPR37 tGFP is excluded from some domains in the plasma membrane, and the actual area that is available for GPR37 tGFP diffusion is smaller than the FCS observation area (see "Results"). Analysis of Cross-correlation Curves and Interpretation of Cross-correlation Decay Times-Dual-color FCCS relies on the usage of spectrally distinct fluorescent markers and the co-variance of their signals to establish complex formation (24,33,34). The cross-correlation curves showed two characteristic decay times and were fitted using a two-component diffusion model without triplet contribution. The apparent dissociation constant of the GPR37 tGFP ⅐PS-TX14(A) TAMRA complexes (K D app ) was determined using the uncorrected form of the cross-correlation function to derive the number of free versus bound GPR37 tGFP and PS-TX14(A) TAMRA molecules (35,36) as shown in Equations 4 and 5,

Analysis of Autocorrelation Curves for PS-TX14(A) TAMRA and Interpretation of Autocorrelation Decay Times-Using
where N g is the number of GPR37 tGFP molecules in the observation volume element; N r is the number of PS-TX14(A) TAMRA molecules in the observation volume element, and N gr is the number of their complexes in the observation volume element. In the equivalent Equation 5, A CC , A g , and A r are the amplitudes of the cross-correlation curve, the amplitude of the autocorrelation curve recorded in the "green" channel, reflecting the con-tributions from GPR37 tGFP and GPR37 tGFP ⅐PS-TX14(A) TAMRA complexes, and the amplitude of the autocorrelation curve recorded in the "red" channel, reflecting the contributions from PS-TX14(A) TAMRA and GPR37 tGFP ⅐PS-TX14(A) TAMRA complexes, respectively, at ϭ 1⅐10 Ϫ5 s. The apparent value of the dissociation constant of GPR37 tGFP ⅐PS-TX14(A) TAMRA complexes (K D app ) obtained using the uncorrected cross-correlation function in Equations 4 or 5 should be regarded as an approximate value that may significantly differ from the actual dissociation constant K D . On the technical side, the following factors can affect the amplitudes of auto-and cross-correlation curves, and hence the experimentally determined K D value; the observation volume elements for the spectrally distinct fluorophores have different volumes and can be displaced due to chromatic aberrations and misaligned excitation lasers, cross-talk between the channels, background fluorescence, and resonance energy transfer between the fluorophores (Förster resonance energy transfer (FRET)). These processes have been recognized and meticulously analyzed in several publications, from the first report by Földes-Papp et al. (37) to the recent work by Wohland and co-workers (38) and references therein. The main conclusion of these studies is that technical limitations do not contribute significantly (Ͻ25%) (37), except under conditions when FRET between the fluorophores is very strong (33,35,37,38), which seems not to be the case in our studies. For the purpose of our studies, it is most important to ascertain that cross-talk between the channels does not yield a false-positive result. To verify this, we performed control experiments using the -opioid receptor selective peptide ␤-endorphin fluorescently labeled with TAMRA (␤-End TAMRA ). Cross-correlation curves recorded for GPR37 tGFP /␤-End TAMRA (Fig. 5B, lower left panel, brown curve) and GPR37 tGFP ⅐PS-TX14(A) TAMRA (Fig. 5B, lower left panel, orange curve) clearly show that cross-talk between channels is minimal and cannot account for the cross-correlation observed for GPR37 tGFP ⅐PS-TX14(A) TAMRA .
The formation of complexes with long characteristic times was further confirmed by sequential FCCS (Fig. 5C). Switching between channels was performed at 100 s to enable acquisition of sufficient number of photons. Unfortunately, this resulted in losing the information for lag times shorter than 1 ms. FIGURE 3. WT and GPR37 tGFP -expressing N2a cells both express low levels of endogenous GPR37 and secrete PSAP but respond differently to inhibition of extracellular PSAP. A, Western blot for GPR37 on lysates from WT and GPR37 tGFP -expressing cells showing a weak band at the size corresponding to untagged GPR37 in both cell types and an additional band at the size corresponding to GPR37 tGFP in the GPR37 tGFP -expressing cells. B, Western blot for extracellular PSAP using medium from WT and GPR37 tGFP cells shows that both cell types secrete equal amounts of PSAP (n ϭ 9, p ϭ 0.99). C, estimate of PSAP concentrations in the conditioned medium of differentiated GPR37 tGFP cells based on Western blot analysis. The calibration curve was generated using a standard series of PSAP solutions that were analyzed on the same gel as the conditioned medium. D, immunoprecipitation of PSAP from conditioned cell medium using anti-PSAP (well 1) or normal rabbit IgG (well 2) as a negative isotype control. 40 g ml Ϫ1 anti-PSAP precipitates 83-96% of all PSAP compared with beads alone. Western blot films show three separate immunoprecipitation experiments. E, quantification of GPR37 density at the plasma membrane versus cytoplasm before and after inhibition of extracellular PSAP using anti-PSAP (n ϭ 15 visual fields). Representative images of live cells show distribution of GPR37 with or without anti-PSAP. Data from Western blots and GPR37 density were analyzed by Student's t test. Data from MTT assays were analyzed by two-way analysis of variance followed by t test for pairwise comparisons, ***, pϽ0.001comparedwithWT.F,MTTassayofWTandGPR37-overexpressingcellsuntreatedortreatedwiththePSAP-specificantibodiesanti-PSAP(antibody:F (2,138) ϭ3.6,p Ͻ 0.05; cell type: F (1,138) ϭ 5.0, p Ͻ 0.05; interaction: F (2,138) ϭ 0.9, p ϭ 0.41) or anti-PS769 (antibody: F (1,60) ϭ 42.4, p Ͻ 0.001; cell type: F (1,60) ϭ 9.2, p Ͻ 0.01; interaction: F (1,60) ϭ 9.6, p Ͻ 0.01) (n ϭ 24 in each group). ### , p Ͻ 0.001 compared with WT anti-PS769; *, p Ͻ 0.05 compared with WT anti-PSAP.
In addition to the technical challenges discussed above, one should also note that endogenous nonfluorescent molecules, protein constructs with fluorophores residing in dark states, or protein constructs with photobleached fluorophores may compete in bimolecular interactions, shifting the apparent equilibrium to higher values of K D app (38).
FRAP-FRAP measurements were performed on the same instrument using the same optical settings as for tGFP confocal imaging (described above). Circular regions of interest (ROIs) 2 m in diameter were defined for bleaching. To measure the  ments showed no cross-correlation between GPR37 tGFP and ␤-End TAMRA . Bottom left panel, cross-correlation curves confirming interactions between PS-TX14(A) TAMRA and GPR37 tGFP reflect specific binding (orange), as compared with lack of cross-correlation between ␤-End TAMRA and GPR37 tGFP signal (brown). Right panel, representative images show absence of co-localization between ␤-End TAMRA and GPR37 tGFP . C, autocorrelation and cross-correlation curves for GPR37 tGFP and PS-TX14(A) TAMRA using the sequential illumination mode at a switching rate of 100 s. fluorescence intensity before bleaching, 15 measurements were first performed using low laser intensity (2% transmission), followed by bleaching with the 488 nm line of the Ar/ArKr laser at 100% transmission to achieve 60 -90% bleaching. Fluorescence was then measured at 3-s intervals for 5 min, using the same laser transmission at acquisition as before bleaching. Fluorescence intensities were normalized so that the average intensity before bleaching was 1 and after bleaching was 0 (39).
To investigate the effect of cholesterol depletion on GPR37 dynamics, nearby regions of the same cell were subjected to FRAP before and 45 min after addition of m␤CD or 1 h after addition of cholesterol oxidase. We first ascertained in control experiments that this procedure is valid and that the first and second FRAP measurement yielded the same results when no treatment was applied. Fitting of the FRAP curves was not performed.
Quantification of Fluorescence Intensity and Co-localization-For fluorescence intensity quantification at the plasma membrane versus the cytoplasm, five visual fields per well from three separate experiments were imaged before and 2.5-3 h after addition of the PSAP antibody. This yielded 349 and 346 cells, respectively, for analysis in the untreated and treated group. The membrane was identified in overlaid images of the transmission channel and the tGFP channel. Image analysis was performed using the Carl Zeiss LSM image browser software (Carl Zeiss, Germany). The average fluorescence intensities (I) in defined ROI comprising the plasma membrane and the cytoplasm, respectively, were determined for each cell. The average of all I membrane values in a given image was divided by the average of all I cytoplasm values in the same image, and the ratios of averages for all images were plotted.
In co-localization studies with Alexa 594-tagged cholera toxin ␤-subunit (CTxB), cells were treated with CTxB alone or pretreated with PSAP antibody for 2.5 h followed by addition of CTxB. For quantification of co-localization, 10 visual fields per well from three separate experiments yielded a minimum of 69 cells per group. The plasma membrane was identified as described above, and ROIs comprising the plasma membrane were defined. The co-localization function in the Zeiss LSM image browser was then used to derive the Pearson's correlation (P) and Manders' coefficient of green overlapping red (c2) for each ROI. From these values, the average P and c2 within each group was calculated.
Statistical Analysis-Statistical analysis was performed using GraphPad Prism 5 (GraphPad Software, San Diego). Data sets with two grouping variables were analyzed by two-way analysis of variance, and the post hoc test was performed using Student's t test for pairwise comparisons. Data sets with one variable and two groups were analyzed by unpaired two-tailed Student's t test. Values are reported as mean Ϯ S.E.M., and effects were considered significant at p Ͻ 0.05. Numbers of cells analyzed in each graph are given in the respective figure legends.

RESULTS
N2a Cells Secrete PSAP-Overexpression of GPR37 improves cell survival after toxin exposure (10), which suggests constitutive receptor activity and/or the presence of agonist in the cell culture medium. Here, we studied wild type (WT) N2a cells, expressing low levels of GPR37 and nondetectable levels of GPR37-L1, and GPR37 tGFP -overexpressing cells (Fig. 3A). Because PSAP was recently reported to act as an agonist at GPR37 (20), conditioned medium from WT and GPR37 tGFPexpressing cells was tested for the presence of PSAP. Western blot indicated similar concentrations of PSAP in the conditioned medium from WT and GPR37 tGFP cells (Fig. 3B). The conditioned medium, collected from differentiated cells 3 days after seeding at a density of 5 ϫ 10 4 cells ml Ϫ1 , was estimated to contain 0.1-0.4 M PSAP (Fig. 3C).
Immunoabsorption of Extracellular PSAP Causes Decreased Density of GPR37 at the Plasma Membrane-Differentiation of N2a cells leads to increased density of stably transfected GPR37 tGFP at the plasma membrane (10). Preliminary observations suggested that replacement of the conditioned medium from differentiated cells with new medium temporarily reverses this process. This led us to hypothesize that the density of GPR37 at the plasma membrane depends on secreted PSAP. The extracellular pool of PSAP was therefore inhibited by immunoabsorption with an antibody raised against amino acids 301-385 comprising the neurotrophic saposin C region of PSAP (anti-PSAP). To confirm that most of extracellular PSAP was indeed bound by the amount of antibody added, immunoprecipitation was performed using protein A-coupled magnetic beads. In three separate experiments, 83-96% of extracellular PSAP was precipitated with the antibody compared with magnetic beads alone (Fig. 3D). Quantification of GPR37 tGFP at the plasma membrane compared with the cytoplasm revealed a decrease of membrane density 2.5-3 h after anti-PSAP addition (Fig. 3E).
Immunoabsorption of Extracellular PSAP Decreases Cell Viability in N2a Cells with a Delayed Time Course in GPR37-overexpressing Cells-Protective effects of GPR37 overexpression against dopaminergic toxins in N2a cells correlate with increased receptor levels at the plasma membrane (10). To test for neurotrophic effects of PSAP in differentiated N2a cells and the possible involvement of GPR37 in such effects, we performed MTT cell viability assays after inhibiting extracellular PSAP using anti-PSAP. Cell viability was determined after 8, 16, and 24 h of treatment. Depletion of PSAP decreased cell viability in both WT and GPR37 tGFP cells, but with a delayed time course in the latter cells (Fig. 3F). In WT cells, the maximum effect was reached at 16 h, whereas it was reached at 24 h in GPR37 tGFP cells (Fig. 3F, blue and green). In a follow-up experiment, WT and GPR37 tGFP cells were incubated with another anti-PSAP antibody raised against the neurotrophic 22-mer prosaptide 769 (anti-PS769) (21) for 16 h (Fig. 3F, purple and brown). Based on titration studies, this antibody binds PSAP at a 1:2 antibody/ target ratio (data not shown), and the antibody concentration added in these experiments should thus bind 100 nM PSAP in solution. Also in these experiments, GPR37 tGFP cells showed less reduction in cell viability as compared with WT cells (Fig.  3F). We also confirmed that the reduction in cell viability at 24 h was in fact the maximum effect for both cell types by carrying out an additional MTT assay after 48 h of PSAP depletion. After 48 h, cell viability was reduced by 33 Ϯ 5 and 33 Ϯ 4% in WT and GPR37-overexpressing cells, respectively (n ϭ 24 for each group).

GPR37 Co-localizes, Co-migrates, and Forms Complexes with the PSAP-derived Peptide Prosaptide TX14(A)-
To study interaction in live cells of GPR37 tGFP and the synthetic PSAP derivative prosaptide-TX14(A) fluorescently labeled with carboxytetramethylrhodamine (PS-TX14(A) TAMRA ), we first used CLSM imaging, which revealed co-localization of GPR37 tGFP with PS-TX14(A) TAMRA at the plasma membrane (Fig. 4A). Both GPR37 tGFP and PS-TX14(A) TAMRA fluorescence was discrete, forming punctate patterns at the plasma membrane, indicating compartmentalization in membrane microdomains. Confocal imaging of PS-TX14(A) TAMRA interactions with WT N2a cells showed very limited interaction of PS-TX14(A) TAMRA with the plasma membrane, confirming that PS-TX14(A) TAMRA binding to membranes of GPR37 tGFP -overexpressing cells is specific for GPR37 (Fig. 5A). CLSM also demonstrated co-localization of GPR37 tGFP with PS-TX14(A) TAMRA in intracellular vesicles of live cells (Fig. 4A). Based on this observation, we stained the cells with MitoTracker or LysoTracker to investigate whether GPR37 tGFP co-localized with either mitochondria or lysosomes. Vesicles containing GPR37 tGFP were found to largely co-localize with LysoTracker staining but not with Mitotracker staining (Fig. 6).
GPR37 tGFP interactions with PS-TX14(A) TAMRA at the plasma membrane were further characterized using FCS and dual-color FCCS. FCCS can detect complex formation in living cells in a manner that is independent of distance between the fluorophores by measuring the co-variance of fluorescence intensity fluctuations of the dually labeled complexes (24,33,34).
Single color FCS experiments showed a longer diffusion time of PS-TX14(A) TAMRA at the plasma membrane of GPR37 GFP cells when compared with the extracellular medium, as evident from the shift of the autocorrelation curve toward longer decay times (Fig. 4B, red autocorrelation curve versus blue curve). These longer decay times indicate that PS-TX14(A) TAMRA interacted with components of the plasma membrane. This was also confirmed by the appearance of an additional component in the autocorrelation curve derived from measurements at the plasma membrane, compared with the autocorrelation curve that was measured in the extracellular medium (Fig. 4B). In addition, FCS analysis showed that PS-TX14(A) TAMRA accumulates in the plasma membrane, as evident from the decrease in the amplitude of the autocorrelation curve (Fig. 4B, inset). When WT cells were treated with PS-TX14(A) TAMRA , the autocorrelation curve of PS-TX14(A) TAMRA was not significantly altered, showing that interaction of PS-TX14(A) TAMRA with the plasma membrane of N2a cells required significant density of GPR37 (Fig. 5A).
FCCS analysis at the plasma membrane showed that addition of PS-TX14(A) TAMRA (100 nM, as determined by FCS in the bulk medium (Fig. 4B)) to GPR37 tGFP -expressing cells yielded cross-correlation between fluorescence intensity fluctuations of the two spectrally distinct reporters TAMRA and tGFP, demonstrating complex formation and suggesting ligand-receptor interaction (Fig. 4C). Control measurements with the -opioid receptor-selective peptide ligand ␤-endorphin TAMRA and GPR37 tGFP did not produce any cross-correlation, showing that the cross-correlation did not arise from spectral bleed through between the channels or unspecific binding of peptides to GPR37 tGFP (Fig. 5B). The apparent dissociation constant (K D app ) of the PS-TX14(A) TAMRA ⅐GPR37 tGFP complexes was determined to be K D app ϭ 550 Ϯ 350 nM. It should, however, be noted that the conditioned medium contained high concentrations of endogenously secreted PSAP and that a large fraction of the receptors may thus already have been occupied. The K D values determined in our experiments may therefore be consid- erably higher than K D values in a system lacking extracellular PSAP.
In the Presence of PSAP, GPR37 Accumulates in and Internalizes with GM1-containing Lipid Rafts at the Plasma Membrane-As mentioned above, both GPR37 tGFP and PS-TX14(A) TAMRA fluorescence was discrete, forming punctate patterns at the plasma membrane, indicating compartmentalization in membrane microdomains. To test whether GPR37 compartmentalizes into lipid rafts at the plasma membrane, cells were stained with Alexa 594-labeled CTxB. CTxB binds to GM1 gangliosides and hence labels lipid raft structures with a high density of GM1, as well as endosomes originating from stained plasma membrane microdomains (40). GPR37 tGFP was found to co-localize with CTxB and thus to accumulate in GM1-dense lipid rafts of the plasma membrane (Fig. 7A). This was confirmed by quantification of fluorescent signals using Manders' coefficient of pixel overlap and Pearson's correlation of pixel intensity (Fig. 7C). Indeed, the average Manders' coefficient for all analyzed images showed that a substantial fraction of GPR37 tGFP co-localized with CTxB. The average Pearson's correlation confirmed that fluorescence intensity in the green channel (GPR37 tGFP ) positively correlated with fluorescence intensity in the red channel (CTxB Alexa594 ).
To investigate whether accumulation of GPR37 in GM1dense lipid rafts is dependent on PSAP binding to the receptor, we used the anti-PSAP antibody to inhibit interactions between the extracellular PSAP and GPR37 tGFP . Confocal fluorescence microscopy revealed that accumulation of GPR37 tGFP in strongly CTxB-stained microdomains was reduced 2.5 h after antibody addition (Fig. 7B). This was reflected by an 11% reduction of the average Manders' coefficient and a 43% reduction of the average Pearson's correlation of pixel intensity (Fig. 7C). Importantly, FCCS analysis showed that GPR37 tGFP not only co-localizes but also forms complexes with GM1 gangliosides (Fig. 8A, red cross-correlation curve).
Lipid Raft Disruption Alters Trafficking of GPR37 and Complexes between PS-TX14(A) and GPR37 at the Plasma Membrane-Methyl-␤-cyclodextrin (m␤CD) and cholesterol oxidase are commonly used to sequester membrane cholesterol and to disrupt lipid rafts (41,42). To study whether trafficking of GPR37 is dependent upon the integrity of lipid raft microdomains, FCCS, CLSM, and FRAP analyses were performed before and after m␤CD or cholesterol oxidase treatment and compared (Fig. 8). FCCS experiments at the plasma membrane showed that both treatments largely disrupted GPR37 interactions with GM1, as shown by the loss of cross-correlation between GPR37 tGFP and CTxB Alexa 594 (Fig. 8A, blue curves).
FCCS measurements with GPR37 tGFP and PS-TX14(A) TAMRA at the plasma membrane were also performed at baseline and after m␤CD or cholesterol oxidase treatment (Fig. 8C). Comparison of FCCS curves with or without m␤CD (Fig. 8C, left) or cholesterol oxidase (Fig. 8C, right) shows that the FCCS curves shifted toward shorter diffusion times after disruption of lipid rafts. This is consistent with the interpretation that GPR37⅐PS-TX14(A) complexes are located in and migrate with lipid rafts, which are disrupted after cholesterol sequestration.

DISCUSSION
Emerging evidence points toward a signaling pathway in which GPR37 located at the plasma membrane reduces effects of cellular insult (9,10,20). This signaling pathway appears to be relatively susceptible to perturbation and once compromised may instead exacerbate cellular stress through intracellular accumulation of GPR37 aggregates (3, 6 -8). PSAP and gangliosides also regulate neuronal survival and improve neuronal resistance to various insults via mechanisms that are not yet defined (12,15,43,44). This study provides detailed live cell imaging and fluorescence spectroscopy-based data strongly supporting the notion of functional PSAP-GPR37-ganglioside interactions at the plasma membrane. These interactions reduce intracellular GPR37 accumulation and may both regulate neuroprotective actions and attenuate cytotoxic aggregation of GPR37.
PSAP has long been known to exert neurotrophic and neuroprotective effects via GPCR-dependent signaling at the plasma membrane (45). The PSAP receptor couples to G i/o and activates the MAPK and PI3K signaling pathways as well as FIGURE 8. Lipid rafts disruption with m␤CD alters distribution and trafficking of GPR37, PS-TX14(A) TAMRA , and the complex between GPR37 tGFP and PS-TX14(A) TAMRA at the plasma membrane. A, comparison of FCCS curves for GPR37 tGFP and CTxB from untreated, m␤CD-treated, or cholesterol oxidasetreated cells shows a significant cross-correlation (i.e. complex formation) between GPR37 tGFP and CTxB in control cells with 85% of the complexes diffusing at D1 ϭ 0.5 ms and 15% at D2 ϭ 0.2 s. This cross-correlation was largely disrupted by m␤CD and cholesterol oxidase. B, FRAP measurements for GPR37 tGFP were performed on live cells (n ϭ 23 cells for each treatment). Measurements were made on the same cell before and after treatment. Either m␤CD or cholesterol oxidase increased GPR37 tGFP fluorescence recovery compared with untreated cells, and the average amplitude of the FRAP curves at t ϭ 350 s was significantly higher after m␤CD or cholesterol oxidase treatment than before (**, p Ͻ 0.01; ***, p Ͻ 0.001, Student's t test modulates the activity of voltage-gated calcium channels (45)(46)(47). Very recently, GPR37 and GPR37L1 were identified as receptors for PSAP and PS-TX14(A), an active synthetic peptide derived from the neurotrophic sequence of PSAP. In our present experiments, we extend these observations by demonstrating binding in live cells of PS-TX14(A) to GPR37. Importantly, our study demonstrates that inhibition of extracellular PSAP reduces plasma membrane levels of GPR37. Here, we used an anti-PSAP antibody added to the extracellular medium instead of siRNA to specifically inhibit the function of extracellular PSAP without depleting the cells of all intracellular saposins, which are critical for lysosomal function and cell survival (14). Our observations of PSAP-induced effects on GPR37 trafficking support the idea of functional coupling between extracellular PSAP and GPR37. With inhibition of extracellular PSAP, membrane levels of GPR37 decline. This suggests a pathway for intracellular build up of GPR37, which eventually may reach the critical concentration for aggregation. Given that GPR37 is known to cause degeneration of dopaminergic cells via intracellular aggregation and endoplasmic reticulum stress (3, 5-7), PSAP or its active fragments may via this mechanism protect against the progress of PD, as was indicated in previous experimental studies (15,48).
A mechanism where the GPR37 agonist PSAP promotes GPR37 trafficking to the cell surface may seem at odds with the traditional view on GPCR activation, which in most cases leads to receptor internalization. However, there are now a number of examples of both antagonist-induced GPCR internalization and agonist-induced GPCR insertion into the plasma membrane (49 -55). One example is the ␦-opioid receptor (DOR), which similarly to GPR37 has a low surface density in the absence of agonist (49,56). Agonist binding to DORs at the plasma membrane leads to receptor internalization but also triggers intracellular signaling cascades that promote slow and sustained insertion of DORs from the intracellular pool into the plasma membrane (49). The net effect of these two processes is a positive feedback mechanism leading to a sustained increase of the DOR surface density as long as both extracellular agonist and intracellular receptors are available. It is possible that a similar mechanism applies to GPR37. Such a mechanism would be consistent with the slow kinetics for the anti-PSAP-induced internalization of GPR37 tGFP as well as with the PS-TX14(A)induced GPR37 internalization observed by Meyer et al. (20), and with the co-localization of PS-TX14(A) TAMRA and GPR37 tGFP in intracellular vesicles observed in this study.
Mounting evidence indicates that lipids are critical for GPCR folding and that lipid rafts regulate GPCR trafficking and signaling (57). Moreover, beneficial effects of GM1 gangliosides against PD symptoms and progression have been shown both in animal models and in clinical studies (44,58). We therefore characterized the lipid environment of GPR37 tGFP at the plasma membrane in the presence or absence of PSAP. PSAP localizes to ganglioside-containing membrane microdomains (21,22), and ganglioside-containing lipid rafts are required for PSAP-induced neurotrophic GPCR signaling (22). The mechanism for how lipid rafts mediate PSAP function, however, has been elusive. The present data suggest a model of how lipid rafts mediate PSAP function, as co-localization studies revealed that GPR37 accumulates in GM1-containing lipid rafts in the presence, but not upon inhibition, of extracellular PSAP. Moreover, a significant cross-correlation between GPR37 and CTxB showed that GPR37 not only co-localizes but also forms complexes with GM1 gangliosides in the presence of PSAP. The results from these experiments collectively suggest that PSAP induces trafficking of GPR37 to ganglioside-containing lipid rafts and consequently complex formation of the receptor with GM1 gangliosides. Notably lipid raft disruption using m␤CD or cholesterol oxidase altered trafficking of both GPR37 tGFP and the GPR37 tGFP ⅐prosaptide complex at the plasma membrane. This suggests that not only GPR37, but also GPR37⅐prosaptide complexes, are located in lipid rafts. Lipid rafts thus provide a scaffold in which GPR37, prosaptide, and GM1 may interact to promote cell viability.
We also found that inhibition of extracellular PSAP reduced cell viability in GPR37 tGFP cells with a delayed time course as compared with WT cells. This correlates well with the study of Meyer et al. (20) showing that GPR37 mediates PSAP-induced protection against H 2 O 2 -induced toxicity and our previous study showing that overexpression of GPR37 in N2a cells leads to substantial protection against toxicity when the receptor is located at the plasma membrane (10). Because N2a cells endogenously express low levels of GPR37, the GPR37 tGFP cell line used in our studies can be considered a gain-of-function, rather than a heterologous, system. The accumulated effect of more strongly activated intracellular signaling cascades is likely to cause the delayed response to anti-PSAP in GPR37 tGFP cells. The observed effects of PSAP depletion are therefore consistent with neurotrophic PSAP signaling via GPR37 and suggest that GPR37 at least in part mediates the neurotrophic effects of PSAP in these cells.
In conclusion, our studies show that constitutively secreted PSAP promotes trafficking of GPR37 tGFP to the cell surface, where it forms complexes with GM1 gangliosides and associates with GM1-enriched lipid rafts. Inhibiting extracellular PSAP disrupts GPR37 tGFP interaction with GM1-enriched microdomains, promoting its intracellular accumulation. Thus, our data provide evidence for the functional association between GPR37, PSAP, and GM1 gangliosides in the plasma membrane, which conveys protection against intracellular accumulation of GPR37. These findings define a mechanistic frame wherein PSAP, and its derivative prosaptides, and GM1 may protect against the progress of PD by stabilizing GPR37 in the plasma membrane.