Presynaptic Control of Glycine Transporter 2 (GlyT2) by Physical and Functional Association with Plasma Membrane Ca2+-ATPase (PMCA) and Na+-Ca2+ Exchanger (NCX)*

Background: GlyT2 is crucial for glycinergic neurotransmission, but only a few interacting partners for this protein are known. Results: PMCA2/3 and NCX1 interact with GlyT2 and modulate its activity in lipid raft subdomains. Conclusion: Functional interaction of GlyT2 with PMCA2/3 and NCX1 helps Na+ and Ca2+ local homeostasis in glycinergic terminals. Significance: Learning how GlyT2 is regulated might help with developing new therapies for hyperekplexia or neuropathic pain. Fast inhibitory glycinergic transmission occurs in spinal cord, brainstem, and retina to modulate the processing of motor and sensory information. After synaptic vesicle fusion, glycine is recovered back to the presynaptic terminal by the neuronal glycine transporter 2 (GlyT2) to maintain quantal glycine content in synaptic vesicles. The loss of presynaptic GlyT2 drastically impairs the refilling of glycinergic synaptic vesicles and severely disrupts neurotransmission. Indeed, mutations in the gene encoding GlyT2 are the main presynaptic cause of hyperekplexia in humans. Here, we show a novel endogenous regulatory mechanism that can modulate GlyT2 activity based on a compartmentalized interaction between GlyT2, neuronal plasma membrane Ca2+-ATPase (PMCA) isoforms 2 and 3, and Na+/Ca2+-exchanger 1 (NCX1). This GlyT2·PMCA2,3·NCX1 complex is found in lipid raft subdomains where GlyT2 has been previously found to be fully active. We show that endogenous PMCA and NCX activities are necessary for GlyT2 activity and that this modulation depends on lipid raft integrity. Besides, we propose a model in which GlyT2·PMCA2–3·NCX complex would help Na+/K+-ATPase in controlling local Na+ increases derived from GlyT2 activity after neurotransmitter release.

receptive fields in the retina, and spinal cord-sensitive pathways. Glycinergic synaptic inhibition is terminated by two sodium-and chloride-coupled transporters, GlyT1 2 and GlyT2, located in the glial plasma membrane and presynaptic terminals, respectively (1,2). Additionally, GlyT2 transport activity supplies glycine for presynaptic vesicle refilling, a process that is absolutely necessary to preserve quantal glycine content in synaptic vesicles (3)(4)(5). GlyT2 activity dysfunctions reduce presynaptic glycine release and cause significant lacking of inhibitory glycinergic neurotransmission. In humans, this situation causes hyperekplexia or startle disease, a rare disease that is characterized by an exaggerated startle response, usually evoked by tactile or auditory stimuli, leading to hypertonia and apnea episodes that in some cases produce sudden infant death (6 -8). GlyT2 has a crucial role in the pathophysiology of inhibitory glycinergic neurotransmission, and studying modulatory factors that regulate its activity might help the success of future therapies. Indeed, studies of endogenous GlyT2 regulatory mechanisms have revealed that GlyT2 activity is regulated by PKC activation (9,10), lipid raft environment (10,11), P2Y purinergic receptors (12), ubiquitination (13), calnexin function (14), or Na ϩ /K ϩ -ATPase interaction (15). In addition, we have also described that GlyT2 exocytosis is regulated by syntaxin-1A and calcium (Ca 2ϩ ) increases (16). Ca 2ϩ ions are widely known essential regulators of synaptic function because of the following: (a) synaptic neurotransmitter release is driven by Ca 2ϩ influxes through voltage-gated calcium channels (17,18), and (b) Ca 2ϩ ions have an important role as secondary messengers in numerous signal transduction processes (19 -21). Thus, neurons invest large amounts of energy in controlling and maintaining an abrupt gradient between intracellular (ϳ0.1-M) and extracellular (ϳ1.4-mM) Ca 2ϩ concentrations. Plasma membrane calcium ATPases (PMCAs) have the key biochemical function of extruding cytosolic Ca 2ϩ out of the cell (by the spent of ATP) in a calmodulin-stimulated manner (22)(23)(24). These Ca 2ϩ pumps belong to the P-type ATPase superfamily that also encompasses, for example, the Na ϩ /K ϩ -ATPase (NKA) or the Ca 2ϩ -ATPases of the sarco(endo)plasmic reticulum (SERCA) (25) and the secretory pathway (SPCA) (26). Four different genes have been described (ATP2B1-4) that encode four PMCA isoforms (PMCA1-4), and the average amino acid sequence identity amounts to only about 80% (27,28). The four PMCA isoforms show differential patterns of expression, with PMCAs 1 and 4 being ubiquitously expressed and PMCAs 2 and 3 being expressed predominantly in the CNS (29,30). The distribution of PMCA isoforms in isolated DRMs (or lipid rafts) is a point of controversy because in synaptic plasma membranes (SPMs) from pig cerebellum PMCA4 seems to be the only isoform located in raft domains (31), although in primary cortical neurons (32) or hippocampus membranes 3 all isoforms appear to be raft-associated. Recently, Jiang et al. (33) have described by quantitative mass spectrometry that the four isoforms are associated with rafts in synaptic plasma membranes from whole rat brain, representing about 60% of the total PMCA. This DRM-associated PMCA pool has been shown to have higher specific activity (32)(33)(34)(35) according to the initial proposition that PMCAs are more active when included in these membrane subdomains (36). Similarly, we have described that GlyT2 can be found in DRMs where it displays the highest transport activity (11) suggesting that GlyT2 presence/absence in lipid rafts could be a versatile regulatory mechanism for the transporter (9,10).
In this study, we have identified neuronal PMCA isoforms 2 and 3 and the Na ϩ /Ca 2ϩ exchanger (NCX1) as new interacting and regulatory partners of GlyT2. We found that GlyT2, PMCA2/3, and NCX1 are co-enriched in neuronal lipid raft membrane clusters. Pharmacological inhibition of PMCA activity by the specific inhibitor caloxin 2a1, as well as specific inhibition of the reverse mode of NCX by KB-R7943 mesylate (KB-mes), led to a marked reduction in GlyT2 activity suggesting that proper Ca 2ϩ extrusion in presynaptic terminals is somehow necessary for optimal GlyT2 activity. This PMCA/NCX regulation of GlyT2 depends on lipid raft integrity because lipid raft disruption by methyl-␤-cyclodextrin (M␤CD) blocks caloxin 2a1 and KB-mes effects and reduces co-localization of NCX and PMCAs with GlyT2. Here, we suggest that the local functional coupling between GlyT2, PMCA2/3, and NCX1 occurs in lipid raft domains and that this association may help in correcting the local imbalance of Na ϩ produced during high activity periods of 1glycine-3Na ϩ co-transport by GlyT2 after neurotransmitter release. These local increases in cytosolic Na ϩ could not be entirely covered by the Na ϩ /K ϩ -ATPase activity due to its slow Na ϩ -extruding rate (200 s Ϫ1 ) (37) forcing NCX to extrude Na ϩ at its normal higher speed (5000 s Ϫ1 ) (38,39) and producing local Ca 2ϩ increases that will be amended by PMCA activity.

Materials
Male Wistar rats were bred under standard conditions at the Centro de Biología Molecular Severo Ochoa in accordance with the current guidelines for the use of animals in research. All animal procedures were approved by the institutional animal care committee and performed according to European Union guidelines (Council Directive 2010/63/EU). Antibodies against GlyT2 were obtained in-house (rabbit and rat (1,40)), although the other primary antibodies used were as follows: anti-PMCA2, anti-PMCA3, anti-PMCA4, and anti-PMCA (clone 5F10) from Thermo Scientific; anti-Thy-1 (Pharmingen); anti-flotillin1 (BD Biosciences); anti-clathrin heavy chain (BD Transduction Laboratories); PSD95 (Neuromab); and anti-NCX1 (Swant). Fluorophore-coupled secondary antibodies were from Molecular Probes. All chemicals used were from Sigma unless otherwise noted, and Neurobasal medium and B-27 supplement were purchased from Invitrogen. Caloxin 2a1 (VSNSNWPSFPSSGGG-NH 2 ) (41) was custom-synthesized by the proteomics service of the Centro de Biología Molecular Severo Ochoa (Madrid, Spain).

Immunoprecipitation
Synaptosomes or primary neurons from the brainstem or spinal cord (100 g) were lysed for 30 min at room temperature at a concentration of 1.5 mg of protein/ml in TN buffer (25 mM Tris-HCl and 150 mM NaCl (pH 7.4)) containing 0.25% Nonidet P-40 and protease inhibitors (0.4 mM phenylmethylsulfonyl fluoride (PMSF) ϩ Sigma mixture). After 15 min of centrifugation in a microcentrifuge to remove the cell debris, 4 g of protein were separated to quantify total protein, and 5 l of the primary antibody were added and left overnight at 4°C using the following antibodies for immunoprecipitation: rat anti-GlyT2, anti-PMCA (clone 5F10), or anti-NCX1. A negative control was also run in parallel in which an irrelevant antibody was added, denoted as IgG. Subsequently, 50 l of 50% protein G-agarose beads (ABT Beads) were added and incubated for 45 min at 4°C. The beads were collected by mild centrifugation and washed twice for 7 min with lysis buffer at room temperature. Finally, the beads were pelleted, and the immunoprecipitated proteins were eluted in Laemmli buffer at 75°C for 10 min, resolved in SDS-polyacrylamide gels (7.5%), detected in Western blots by ECL, and quantified on a GS-810 imaging densitometer (Bio-Rad). The amount loaded was always 4 g of total protein and 96 g of immunoprecipitated sample or IgG immunoprecipitation controls. This ratio allows having both samples in the same linear range of exposure.

Immunohistochemistry of Rat Brainstem Slices
Immunohistochemistry experiments from rat brainstem slices were performed as described previously (42). The sections shown in this work correspond to the ventral cochlear nucleus of the brainstem.

Ca 2؉ -ATPase Activity Measurement in Brainstem and Spinal Cord Synaptosomes
The Ca 2ϩ -ATPase activity was measured using a coupled assay in different conditions to measure the contribution of PMCA and the intracellular SERCA and SPCA ATPases as described in Sepúlveda et al. (43).

Immunofluorescence Quantification, Generation of Color Maps, and Generation of Pixel Fluorescence Intensity Profiles
At least 45 images for each condition were quantified using ImageJ software (National Institutes of Health). Images were processed with a 2.0-pixel median filter, and the threshold used was automatically determined by the JACoP plugin (44). The Pearson's value was obtained with JACoP by comparing the two thresholded channels and measuring the correlation between them. The value can range from Ϫ1 to 1, with 1 representing the maximal co-localization possible (two identical images), whereas values Ն0.5 can usually be considered as valid co-localization (45). Co-localization color maps were generated in ImageJ using the co-localization color map plugin (46). Pixel fluorescence intensity profiles were also generated using ImageJ software. The linear region of interest was manually drawn to cross several GlyT2-PMCA/NCX-Thy-1 clusters, and the fluorescence intensity profiles were obtained from each individual color channel using the RGB_profiler plugin. Inten-sity profile plots were generated in Origin 8.0 (Figs. 5 and 8, green, PMCAs/NCX1; red, GlyT2; blue, Thy-1), and triple colocalization spots were considered valid only when intensity values of the three proteins were higher than 65% of the maximum at the same region.

Mass Spectrometry Assays
In-Gel Digestion-In-gel digestion was performed as described previously (47,48).
Reverse Phase-Liquid Chromatography RP-LC-MS/MS Analysis (Dynamic Exclusion Mode)-The desalted protein digest was dried, resuspended in 10 l of 0.1% formic acid, and analyzed by RP-LC-MS/MS in an Easy-nLC II system coupled to an ion trap LTQ-Orbitrap-Velos-Pro mass spectrometer (Thermo Scientific). The peptides were concentrated by reverse phase chromatography using a 0.1 ϫ 20-mm C18 RP precolumn (Proxeon) and then separated using a 0.075 ϫ 100-mm C18 RP column (Proxeon) operating at 0.3 l/min. Peptides were eluted using a 90-min gradient from 5 to 40% solvent B (solvent A: 0.1% formic acid in water, and solvent B: 0.1% formic acid, 80% acetonitrile in water). ESI ionization was done using a Nano-bore emitter stainless steel inner diameter of 30 m (Proxeon) interface. The Orbitrap resolution was set at 30,000.
Peptides were detected in survey scans from 400 to 1600 atomic mass units (1 microscan), followed by 15 data-dependent MS/MS scans (top 15), using an isolation width of 2 units (in mass-to-charge ratio units), normalized collision energy of 35%, and dynamic exclusion applied during 30-s periods. Peptide identification from raw data was carried out using the SEQUEST algorithm (Proteome Discoverer 1.3, Thermo Scientific). Database search was performed against uniprot-rattus. fasta. The following constraints were used for the searches: tryptic cleavage after Arg and Lys, up to two missed cleavage sites, and tolerances of 10 ppm for precursor ions and 0.8 Da for MS/MS fragment ions, and the searches were performed allowing optional Met oxidation and Cys carbamidomethylation. Search against decoy database (integrated decoy approach) was performed using a false discovery rate of Ͻ0.01.

Lipid Raft Isolation
Membrane rafts were biochemically isolated from brainstem and spinal cord synaptosomes as described previously (10). Quantification of the relative distribution in lipid rafts of the different proteins was performed as shown in Equation 1, where R is the raft optical density/g of protein, and NR is the non-raft optical density/g of protein.
Densitometric analyses were performed on at least three independent Western blots representing the means Ϯ S.E.

Data Analysis
All statistical analyses were performed using Origin 8.0 (OriginLab Corp.). One-way analysis of variance (ANOVA) was used to compare multiple groups, with subsequent Tukey's post hoc test to determine the significant differences between samples. The Student's t test was used to compare two separate groups. p values are denoted through the text as follows: *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001; and p Ͻ 0.05 or lower values were considered significantly different when compared by oneway ANOVA (Tukey's post hoc test) or Student's t test. When shown, box whisker plots represent the following: median (line); mean (point); 25th to 75th percentile (box); 10th to 90th percentile (whisker); 1st to 99th percentile (ϫ); and minimum to maximum (Ϫ) ranges.

RESULTS
Given the pathophysiological important role of GlyT2 (6) and considering the small number of proteins currently known to interact with this transporter, we carried out a proteomic study to identify new molecular partners involved in the functional regulation of GlyT2. Thus, the native transporter was immunoprecipitated from brainstem and spinal cord synaptosomes using a specific GlyT2 antibody (40), and co-purified proteins were identified by high throughput mass spectrometry. GlyT2 immunoprecipitates were found to selectively enhance the detection of several proteins, including some of the previously described GlyT2 interacting partners such as syntaxin-1 (16) or Na ϩ /K ϩ -ATPase (15). We also found other molecules that are known to be implicated in GlyT2 trafficking, such as clathrin heavy chain (10).
Surprisingly, mass spectrometry also revealed other proteins, the main function of which is related to intracellular calcium homeostasis. In this last group, the highest number of specific peptides detected corresponded to the neuron-specific PMCA isoforms 2 and 3 ( Fig. 1A, highlighted in red), which are responsible for extruding calcium out of cells at the expense of ATP and thus regulating cytosolic calcium concentration at nerve terminals (representative tandem mass spectra of PMCA2-and PMCA3-specific peptides obtained in these assays are shown in Fig. 1, B and C). PMCA2 and PMCA3 have been shown to be richly expressed in synapses (49 -52), where they are proposed to control neurotransmission dynamics (via controlling Ca 2ϩmediated synaptic vesicle fusion) and/or regulate local Ca 2ϩ signaling (53,54).
GlyT2 Interacts with PMCA2 and PMCA3 in Brainstem and Spinal Cord Synaptosomes and in Heterologous Systems-To confirm the interaction between GlyT2 and PMCA2 and 3, we performed reverse immunoprecipitations from brainstem and spinal cord synaptosomal lysates using an antibody that recognizes the four PMCA isoforms (PMCA1-4, clone 5F10) ( Fig.  2A). Western blotting showed that the antibody selectively immunoprecipitated the respective target protein and that GlyT2 specifically co-purified in PMCA immunoprecipitates ( Fig. 2A), indicating that native GlyT2 and PMCA proteins interact specifically under physiological conditions. However, proteomic data only pointed out the neuron-specific PMCA isoforms 2 and 3 as putative interacting partners of GlyT2 (Fig.  1), and consequently, we performed further GlyT2 immunoprecipitations and detected PMCA2 and PMCA3 using isoform-specific antibodies (Fig. 2, B-D). Western blotting showed a specific co-purification of PMCA2 and PMCA3, yet no signal was observed when an irrelevant antibody was added (denoted as IgG). Furthermore, PMCA4 is not found co-purifying in these conditions (Fig. 2D), in agreement with proteomic data.
The interaction between GlyT2 and PMCA2 or PMCA3 was also detected in a heterologous COS7 cell line overexpressing each isoform together with the wild type transporter or with a GlyT2 mutant lacking the C-terminal region, which contains a PDZ (95/Disc large/Zonula occludens-1) ligand motif that could potentially interact with the PDZ binding domain of the pumps, as reported for some PMCA-interacting partners (55)(56)(57). Western blot assays (Fig. 2, E and F) showed that both PMCA2 and PMCA3 clearly maintained their association with the transporter despite the lack of its PDZ-containing C-terminal region, indicating that the PMCA2-and PMCA3-GlyT2 interactions are not PDZ domain-based interactions and suggesting that other regions of GlyT2 are implicated in the physical coupling between both proteins.
To further confirm the interaction between GlyT2 and PMCA2 or PMCA3 in the native system, their co-localization was examined by immunohistochemistry in rat brainstem slices. As shown in Fig. 3, double labeling for GlyT2 and PMCA2 or PMCA3 indicates that GlyT2 clearly co-localizes with PMCA2 (Fig. 3A) and PMCA3 (Fig. 3B), especially in presynaptic structures (Fig. 3, A and B, detail). In addition, we performed immunocytochemistry in synaptosomes from brainstem and spinal cord (Fig. 4). Double labeling for GlyT2 and PMCA2 or PMCA3 (Fig. 4, B and C) indicates that most GlyT2 overlaps with both isoforms, whereas no co-localization is observed with a postsynaptic marker, PSD-95 (Fig. 4A). These differences can be quantified using the Pearson's value of correlation, obtaining a significant co-localization between PMCAs and GlyT2 (Fig. 4D, PMCA2 Pearson's value: 0.635 Ϯ 0.013 S.E., p ϭ 1.98 ϫ 10 Ϫ23 ; PMCA3 Pearson's value: 0.596 Ϯ 0.009 S.E.; p ϭ 7.97 ϫ 10 Ϫ22 ). Together, these immunoprecipitation, immunohistochemical, and immunocytochemical results indicate a specific interaction between GlyT2 and the neuronal PMCA2 and PMCA3 isoforms.
GlyT2 and PMCA Isoforms 2 and 3 Are Co-enriched in Neuronal Lipid Raft Membrane Clusters-DRMs or lipid rafts are small (10 -200 nm), heterogeneous, highly dynamic, and steroland sphingolipid-enriched domains that compartmentalize cellular processes (58 -60). GlyT2 displays optimal transport activity when it is associated with these subdomains at the cell surface, where most of the transporter resides in primary neurons and synaptosomes from the rat brainstem (11,15). Considering that PMCA isoforms are also localized in lipid rafts, Precipitated protein complexes were analyzed by Western blots (WB) probed with anti-GlyT2, anti-PMCA (all isoforms), anti-PMCA2, anti-PMCA3, or anti-PMCA4 antibodies. T, total protein; IP, immunoprecipitated sample; IgG, IgG immunoprecipitation controls. Note that an interaction can be detected between GlyT2 and PMCA isoforms 2 and 3, yet no signal is observed in PMCA4 Western blots or in the IgG controls. E and F, COS7 cells were transfected to express wild type GlyT2 or GlyT2 lacking the C-terminal region (amino acids from 737-799; denoted as del C-Ter) and PMCA isoforms 2 or 3. Lysed cells were incubated with antibodies against GlyT2 or the equivalent IgGs as control and protein complexes were precipitated using protein-agarose beads and analyzed in Western blots probed with anti-GlyT2 (E and F) and anti-PMCA2 (E) or anti-PMCA3 (F) antibodies. Two bands of wild type GlyT2 are detected when expressed in COS cells as reported previously (14). T, total protein; IP, immunoprecipitated sample; IgG, IgG immunoprecipitation controls. Note that the interaction can still be detected between GlyT2 lacking the C-terminal region and the PMCA isoforms, indicating that the PDZ domain of GlyT2 is not necessary for this interaction.
where they are known to be optimally active (32)(33)(34)(35), we investigated whether the coupling shown here between PMCAs and GlyT2 could be facilitated by membrane subdomain compartmentalization of these proteins. As a first approach, we sought to determine the distribution of GlyT2, PMCA2, and PMCA3 in raft subdomains using sucrose gradient centrifugation. As shown in Fig. 5, the three proteins are distributed among raft and non-raft fractions (Fig. 5, A and B). In agreement with previously reported results, GlyT2 is enriched in raft fractions (9,15), showing here a relative distribution in rafts of 77.26 Ϯ 4.46% S.E., whereas PMCA2 and PMCA3 distributions are 51.09 Ϯ 6.25% and 49.46 Ϯ 7.57% S.E., respectively (Fig. 5B). Correct biochemical isolation of raft fractions was confirmed by Western blotting of raft and non-raft markers, flotillin1 and clathrin heavy chain, respectively.
To find out whether GlyT2 and PMCA2 or PMCA3 are present in the same raft clusters, we performed immunocytochemical assays in primary neurons using specific antibodies against GlyT2, PMCA2 or PMCA3, and Thy-1, a marker of neuronal rafts (Fig. 5, C-H) (61). We observed a clear co-localization between GlyT2 and Thy-1, as reported previously (15). Interestingly, these GlyT2-Thy-1 clusters extensively co-localized with PMCA2 or PMCA3, leading to a high degree of triple colocalization as shown by the white color in the overlay image (denoted as Merge in Fig. 5, D and G). To clearly visualize triple co-localization, we generated color maps that highlight pixels of co-localization (Fig. 5, D and G) (46) and pixel fluorescence intensity profiles of each individual color staining that show a substantial overlap among the plots of each color staining (Fig.  5, E and H) (62). In this case, triple co-localization spots are highlighted by inverted arrows. Taken together, these biochemical and immunocytochemical results suggest that GlyT2 and PMCA pumps are associated in lipid raft subdomains, where both proteins have been shown to be optimally active (11,(32)(33)(34)(35).
PMCA Function Regulates Presynaptic Glycine Transport by GlyT2-Given that GlyT2 and PMCAs are co-enriched in raft membrane subdomains, where they have been proposed to be fully functional, we next analyzed the possible regulatory interplay between these proteins. As a first approach, we measured GlyT2 and PMCA activities in brainstem and spinal cord synaptosomes in the presence of ALX1393, a specific GlyT2 inhibitor (63). As expected, ALX1393 reduced about 90% of the transporter activity ( Fig. 6A) but did not affect PMCA activity (Fig. 6B), suggesting that PMCA function is independent of GlyT2 activity. We also tested the reverse hypothesis measuring GlyT2 transport in the presence of the PMCA inhibitor caloxin 2a1, which specifically impairs PMCA function in a dose-dependent manner (Fig. 6, C and D) while not affecting other Ca 2ϩ -ATPases such as SERCA or SPCA (Fig. 6C) (64, 65). Fig. 7A shows that incubation with 400 M caloxin 2a1 for 10 min produced a significant down-regulation of GlyT2 transport, with the remaining activity of 66.38 Ϯ 2.26% S.E. compared with the control (p ϭ 1.36 ϫ 10 Ϫ7 , green box plot). However, the same protocol did not produce any differences when we measured the activity of the highly related glycine transporter GlyT1 (Fig. 7B; p ϭ 0.152) (1). This suggests a specific regulation of GlyT2 by PMCA activity, indicating that somehow Ca 2ϩ extrusion by PMCAs is necessary for optimal GlyT2 glycine recapture. PMCA function is tightly related to the function of the plasma membrane sodium/calcium exchanger (NCX) (66). In virtually all cells, NCX mainly operates in its "forward" mode harnessing the electrochemical gradient of Na ϩ to extrude Ca 2ϩ with a known stoichiometry of 3Na ϩ : 1Ca 2ϩ (67), helping PMCA proteins in the cytosolic calcium clearance (68,69). However, during high frequency action potentials in presynaptic terminals, the activity-mediated increase in cytosolic Na ϩ can switch NCX direction of flux exchange extruding Na ϩ and allowing Ca 2ϩ entry (70 -72). In this mode, NCX produces local increases in Ca 2ϩ concentration that are thought to be corrected by PMCA activity. Based on these data, we hypothesized that PMCA-mediated modulation of GlyT2 transport could be based on NCX locally acting in the reverse mode to correct Na ϩ entry mediated by GlyT2 transport, which is known to introduce 3Na ϩ ions per glycine recaptured (73). NCX reverse mode can be pharmacologically inhibited by the isothiourea derivative KB-mes (74 -77). Brainstem and spinal cord synaptosomes incubated with KB-mes showed a significant down-regulation of GlyT2 transport (42.99 Ϯ 5.71% S.E.; p ϭ 1.53 ϫ 10 Ϫ9 ) in a similar level to that observed in the presence of caloxin 2a1 (66.38 Ϯ 2.26% S.E.; p ϭ 1.36 ϫ 10 Ϫ7 ) (Fig. 7A, purple box plot), indicating that NCX function is also needed for optimal GlyT2 activity and suggesting a PMCA/NCX cooperative effect in regulating GlyT2 transport. Moreover, simultaneous treatment with caloxin 2a1 and KB-mes produced a reduction in GlyT2 activity (49.88 Ϯ 6.17% S.E.) that is not significantly different from KB-mes (p ϭ 0.9627) or caloxin 2a1 (p ϭ 0.17) treatments alone, further supporting the hypothesis that both proteins are acting in the same pathway (Fig. 7A, light blue box plot). In addition, to confirm that Ca 2ϩ extrusion by PMCAs is necessary for optimal GlyT2 glycine recapture, we used BAPTA-AM, a cell-permeant ion chelator that is a highly selective for Ca 2ϩ . Incubation of synaptosomes with BAPTA-AM reduces the availability of cytosolic Ca 2ϩ affecting PMCA activity and interfering on the NCX functional shift toward the reverse mode (71). In these conditions, GlyT2 activity appears to be significantly reduced (71.80 Ϯ 6.10% S.E.; p ϭ 0.00043), suggesting that cytosolic concentration of Ca 2ϩ could affect glycine recapture by affecting NCX/PMCA proteins. Fig.  7 indicate that NCX functions contribute to GlyT2-mediated glycine recapture in the presynaptic terminal. A working hypothesis of this work is that NCX could act by correcting the local imbalance of Na ϩ produced during high activity periods of glycine-Na ϩ co-transport by GlyT2 after neurotransmitter release that could not be covered by Na ϩ /K ϩ -ATPase due to its slow rate (200 Ϫ1 s) (37). This localized process should be facilitated by the proximity of both proteins, and therefore we sought to determine whether GlyT2 and NCX physically interact and/or are compartmentalized in the same membrane clusters. NCX was not detected in the initial proteomic studies (Fig.  1), not even increasing the false discovery rate to 5% (data not shown). However, some proteins are particularly difficult to detect by mass spectrometry or they can be present in amounts below the detection limit of this technique. Therefore, not finding unique peptides of a protein in immunoprecipitation experiments does not exclude the possibility of a real interaction. To test this putative interaction between GlyT2 and NCX, we performed reverse immunoprecipitations from brainstem and spinal cord synaptosomal lysates using antibodies that recognize NCX1 or GlyT2 (Fig. 8, A and B). Western blot assays showed that each antibody selectively immunoprecipitated the respective target protein, that GlyT2 was specifically co-purified in NCX1 immunoprecipitates (Fig. 8A), and that NCX1 was specifically co-purified in GlyT2 immunoprecipitates (Fig. 8B), indicating indeed a physical interaction.

GlyT2 Interacts with NCX1, and Both Proteins Are Present in the Same Neuronal Membrane Raft Clusters-Results from
In addition, to find out whether GlyT2 and NCX1 are present in the same raft clusters, immunocytochemical assays were performed in primary neurons using specific antibodies against GlyT2, NCX1, and Thy-1 as done in Fig. 5 for PMCAs (Fig. 8, C and D). In this case, GlyT2-Thy-1 clusters also extensively co-localized with NCX1, visualized by triple colocalization color maps or pixel fluorescence intensity profiles from each individual protein. Taken together, results from Fig. 8 indicate that NCX1 is physically interacting with GlyT2 and suggest that both proteins are present in the same membrane clusters. Therefore, we suggest that GlyT2 is coupled to NCX1 and PMCAs in lipid raft subdomains where both proteins coordinately act in regulating glycine recapture by correcting local Na ϩ and Ca 2ϩ variations after neurotransmitter release.

PMCA/NCX1 Regulation of GlyT2 Activity Depends on Lipid
Raft Integrity-PMCA pumps are optimally active when they are present in lipid raft subdomains (32)(33)(34)(35). In addition, GlyT2 is more active when it is raft-associated (11). In this work, we show that PMCA, NCX, and GlyT2 are together in the same raft clusters; therefore, we next sought to assess whether the FIGURE 5. GlyT2 is present in neuronal lipid rafts containing PMCA2 or PMCA3. A and B, lipid raft fractions (denoted as Raft) were isolated from brainstem and spinal cord synaptosomal lysates and analyzed by Western blot to determine GlyT2, PMCA2, and PMCA3 distribution in lipid rafts. The purity of the isolation was determined probing anti-flotillin1 (raft marker) and anti-clathrin heavy chain (CHC, non-raft marker). B, histogram shows the quantification of the relative distribution of each protein representing the means Ϯ S.E. from three different experiments as shown in A. C-H, primary cultures of spinal cord and brainstem neurons were fixed and incubated with primary and secondary antibodies against PMCA2 or PMCA3 (green), GlyT2 (red), and the lipid raft marker Thy-1 (blue). Cells were visualized by confocal microscopy, and amplified high definition images of axonal regions were taken to generate triple co-localization maps (Coloc. Map) (D and G). E and H, pixel fluorescence intensity profiles were generated to visualize co-distribution of GlyT2 and PMCA2/3 in Thy-1-positive clusters. The linear region of interest was manually drawn from left to right, and fluorescence intensity profiles were obtained from each individual color channel. Triple co-localization spots are highlighted by inverted arrows. Scale bar in C and F, 50 m. Scale bar in D and G, 5 m.
PMCA/NCX1-mediated regulation of GlyT2 was dependent on lipid raft integrity. A common approach used to disrupt raft structures is based on the use of methyl-␤-cyclodextrin (M␤CD), a cholesterol chelator that extracts cholesterol from the plasma membrane disassembling raft structures (58, 78 -80). Brainstem and spinal cord synaptosomes were incubated with different concentrations of M␤CD (0 -2 and 5 mM) with or without caloxin 2a1 (Fig. 9, A and B) or KB-mes (Fig. 9, C and D). Inhibition of GlyT2 activity by caloxin 2a1 or KB-mes was blocked in a dose-dependent manner (Fig. 9, A and C), showing a complete blockage of the inhibition with 5 mM M␤CD treatments (Fig. 9B, caloxin2a1 ϩ M␤CD remaining transport 97.74 Ϯ 14.39% S.E., p ϭ 0.9998; Fig. 9D, KB-mes ϩ M␤CD remaining transport 91.15 Ϯ 17.45% S.E., p ϭ 0.9986), which points out the importance of raft structures for the GlyT2/NCX/PMCA interplay. To further support these results, we studied the effect of M␤CD on the co-localization of GlyT2/ PMCA/NCX in primary neurons. The treatment used in this work (5 mM, 30 min) has been previously shown to reorganize GlyT2 in primary neurons, showing a specific punctae redistribution that is not observed in other neuronal markers such as MAP2 and that is independent of cellular toxicity (11). We measured Pearson's values on M␤CD-treated and -untreated cells and observed a significant reduction in the co-localization of GlyT2 with PMCA2 (B, p ϭ 0.00167), PMCA3 (D, p ϭ 0.00109), and NCX (F, p ϭ 6.33 ϫ 10 Ϫ5 ), suggesting a reduction in the local compartmentalization of these proteins (Fig. 10,  A-F).
Together, these results show that caloxin2a1 and KB-mesmediated inhibitions of GlyT2 need lipid raft integrity, and they suggest that spatial compartmentalization of these proteins favors the functional coupling between them.

DISCUSSION
In this study, we have used high throughput mass spectrometry to identify proteins that interact with GlyT2 in CNS prep-arations to unravel new mechanisms that might modulate GlyT2 function. In this way, we have identified PMCA isoforms 2 and 3 as new protein partners of GlyT2. Reciprocal co-immunoprecipitation, immunohistochemical studies, and immunocytochemical experiments further confirmed proteomic data and indicated that GlyT2 associates and co-localizes with PMCA2 and PMCA3, whereas the interaction with the ubiquitously expressed PMCA4 isoform could not be found. Both PMCAs 2 and 3 are mainly expressed in the CNS and localize pre-and post-synaptically where they are proposed to function as calcium controllers for Ca 2ϩ -driven vesicle exocytosis and neuronal Ca 2ϩ -dependent regulatory mechanisms (52)(53)(54)81). Besides this important role, PMCA2 also appears to be crucial for the survival of spinal cord cells, in particular motor neurons, under pathological conditions (82). In addition, PMCA2 genetic alterations have been described as a cause of human deafness (83,84), and recently, a genetic alteration of isoform 3 of PMCAs was identified in a family with X-linked congenital cerebellar ataxia (54).
One important finding of this work is that GlyT2 not only interacts with PMCA2/3 isoforms but also with NCX1, an exchanger ubiquitously distributed through the CNS with a crucial role for intracellular calcium homeostasis (66,85). These protein interactions are probably enriched in lipid raft membrane subdomains, as suggested both by triple co-localization between GlyT2, PMCA2/PMCA3/NCX1, and the neuronal raft marker Thy-1 and through the dependence on lipid raft  19, pink). The glycine uptake by GlyT2 was then measured. 100% glycine transport by GlyT2 was 51.30 Ϯ 1.97 pmol of Gly/mg of protein/min. B, synaptosomes from the rat cortex were incubated in presence of vehicle (control, n ϭ 14, orange) or 400 M caloxin 2a1 (calox2a1, n ϭ 14, brown). The glycine uptake by GlyT1 was then measured. 100% of activity was 4.07 Ϯ 0.24 pmol of Gly/mg of protein/min. *, significantly different from control, p Ͻ 0.05; ***, significantly different from control, p Ͻ 0.001; n.s., not significantly different. Comparisons of the means were performed using ANOVA by Tukey's post hoc test.
integrity for GlyT2 inhibition during PMCA or NCX1 blockage. This membrane compartmentalization would favor the interaction between active populations of GlyT2, PMCA2/3, and NCX1, facilitating local presynaptic sodium/calcium homeostasis during glycine recapture while functioning in an ionic microdomain in which Na ϩ and Ca 2ϩ concentrations differ from those in the bulk cytosol (86,87). At synapses, NCX and PMCA are the two only plasma membrane proteins responsible for Ca 2ϩ extrusion and work together to control the recovery of Ca 2ϩ basal levels in the small compartment of axon terminals. They have complementary functional characteristics, whereas the PMCA has high Ca 2ϩ affinity and low transport capacity, NCX has low Ca 2ϩ affinity but high capacity for Ca 2ϩ transport (88). NCX is highly expressed in presynaptic terminals where it is very active (89). Under basal conditions, NCX transports 3Na ϩ into the cell and extrudes 1Ca 2ϩ from the cytoplasm  A and B). Precipitated protein complexes were analyzed in Western blots (WB) probed with anti-GlyT2 or anti-NCX1 antibodies. T, total protein; IP, immunoprecipitated sample; IgG, IgG immunoprecipitation controls. Note that two bands are observed in NCX1 Western blots that correspond to the full-length protein (highlighted by an arrowhead, 100 kDa) and the nonreduced protein around 160 kDa (96). C-E, primary cultures of spinal cord and brainstem neurons were fixed and incubated with primary and secondary antibodies against NCX1 (green), GlyT2 (red), and the lipid raft marker Thy-1 (blue). Cells were visualized by confocal microscopy, and amplified high definition images of axonal regions were taken to generate triple co-localization maps. Top panel indicated ROI longitude that is represented in the abscissa of E. E, pixel fluorescence intensity profiles were generated to visualize co-distribution of GlyT2 and NCX1 in Thy-1-positive clusters. The linear region of interest (ROI) was manually drawn from right to left, and fluorescence intensity profiles were obtained from each individual color channel. Triple co-localization spots are highlighted by inverted arrows. Scale bar, C, 50 m. Scale bar, D, 5 m. functioning in the "forward mode." However, under certain physiological conditions such as neuronal activity, intracellular Na ϩ increases can switch NCX to function in the "reverse mode" to expel Na ϩ excess (Ca 2ϩ influx, Na ϩ efflux) (71,72). Thus, it is conceivable that local increases in intracellular Na ϩ in similar conditions could switch NCX from its forward mode to its reverse mode of action.
The results presented in this paper lead us to propose the existence of a functional complex where GlyT2 would be tightly coupled to PMCA and NCX to promote Na ϩ and Ca 2ϩ homeostasis in specific microdomains of the presynaptic glycinergic membranes (Fig. 11A). In this regard, the decrease in GlyT2 transport activity caused by caloxin 2a1 and KB-mes (selective inhibitors of PMCA and reversed NCX, respectively) suggests that when PMCA and NCX are inhibited, GlyT2 activity could be directly affected by local changes in Na ϩ /Ca 2ϩ gradients due to a decrease in the Na ϩ driving force necessary for glycine transport and/or be inactivated by an unknown Na ϩ -and/or Ca 2ϩ -mediated signaling mechanism that would block glycine recapture. Independently of the molecular mechanism, similar homeostatic regulation has been proposed for other proteins, as the interplay between AMPA receptors and the Na ϩ /Ca 2ϩ exchanger in neocortical interneurons (90), and the interaction between PMCA2 and nicotinic acetylcholine receptors in hippocampal interneurons (91).
It has been previously shown that PMCA isoforms participate in multiprotein complexes at the cell membrane (27). PMCA2 interacts presynaptically with syntaxin-1A, a member of the SNARE complex, and postsynaptically with the scaffolding protein PSD95 (52). Given its fast Ca 2ϩ activation kinetics, PMCA2 is especially suited for rapid clearance of presynaptic Ca 2ϩ in fast-spiking inhibitory nerve terminals (51). Therefore, its association with syntaxin-1A may provide PMCA2 a docking place close to the site of neurotransmitter release where precise and rapid control of presynaptic Ca 2ϩ levels is crucial for neurotransmission (18). It should be noted that some years ago we reported an interaction between GlyT2 and syntaxin-1A that is involved in a calcium-dependent rapid increase of the amount of GlyT2 in the presynaptic membrane during neuronal activity. The GlyT2/syntaxin-1A interaction favors the rapid reuptake and recycling of glycine into the nerve terminal (16). Thus GlyT2, syntaxin-1A, and PMCA2 could be forming a presynaptic protein complex that would regulate GlyT2 via controlling both activity and surface expression of the transporter during neurotransmitter release.
Considering our recent identification of NKA as a partner for GlyT2 that modulates endocytosis and expression of the transporter (15), the above discussion raises the exciting possibility that GlyT2 interactions described so far with syntaxin-1A, NKA, PMCA, and NCX could occur within the same presynaptic surface complex. In agreement with this hypothesis, an interaction between NCX and NKA has also been reported (92)(93)(94)(95). Moreover, NCX-NKA coupling has been proposed to facilitate local intracellular calcium regulation in specific and dynamic calcium signaling microdomains where lipid compo-sition and lipid/protein interactions play an important role (87). As a compendium of these protein-protein couplings, a scheme of the physical and functional interplay between NKA, NCX, PMCA, STX1A, and GlyT2 is shown in Fig. 11B.
In summary, as proposed in Fig. 11A, binding of GlyT2 to PMCA isoforms 2 and 3 and to the NCX1 in lipid raft subdomains is postulated to help in the local homeostasis of Na ϩ and Ca 2ϩ during high rates of GlyT2-mediated recapture of glycine in presynaptic terminals. After neurotransmitter release, active raft-associated GlyT2 reuptakes large amounts of glycine back to the terminal producing Na ϩ imbalances that must be restored. In normal conditions, this recovery is performed by Na ϩ /K ϩ -ATPase, but the pump can be saturated in FIGURE 10. Disruption of lipid raft integrity reduces co-localization of GlyT2 with NCX, PMCA2, and PMCA3 in primary cultures of neurons. A-F, primary cultures of spinal cord and brainstem neurons were treated with 5 mM M␤CD during 30 min, fixed, and incubated with primary and secondary antibodies against GlyT2 and PMCA2 (A), PMCA3 (C), or NCX (E). Cells were visualized by confocal microscopy, and images of primary neurons were taken. Scale bar, 50 m. B, C, and F, quantification of the co-localization using Pearson's value was performed as described under "Experimental Procedures." The box whisker plots show differences on the co-localization between GlyT2 and PMCA2 (B), PMCA3 (D), and NCX (F). ***, significantly different from control, p Ͻ 0.001. Comparisons of the means were performed using ANOVA by Tukey's post hoc test (n ϭ 3; on average, 20 images per condition were analyzed in each experiment).
situations of strong demand due to its low rates of transport (200 Hz) (37). We propose that under these conditions of saturation, Na ϩ is extruded out of the terminal by NCX (a much faster Na ϩ -transporting protein, 2000 -5000 Hz) (88). This protein would work locally in the reverse mode introducing 1Ca 2ϩ ion while extruding 3Na ϩ ions per cycle from the terminal. The local compartmentalization of PMCAs in these membrane raft domains would facilitate the extrusion of the Ca 2ϩ ion introduced by NCX, locally balancing Ca 2ϩ concentration in these situations of high glycine recapture.