The Plasma Membrane Ca2+-ATPase Isoform 4 Is Localized in Lipid Rafts of Cerebellum Synaptic Plasma Membranes*

Here we describe the association of the synaptosomal plasma membrane Ca2+-ATPase (PMCA) from pig cerebellum with cholesterol/sphingomyelin-rich membrane domains (rafts). The PMCA4 was localized exclusively in rafts prepared by floatation in Nycodenz density gradients of ice-cold Brij 96 extracts. This was corroborated by its colocalization with the raft markers cholesterol, ganglioside GM1, and PrPC. The remaining PMCA isoforms were found in the detergent-soluble fractions, with the majority of the membrane proteins. Activity assays confirmed the bimodal distribution of the PMCA isoforms in the density gradient, with a lower activity for PMCA4 and greater stimulation by calmodulin than for the other isoforms. By providing an ordered membrane microenvironment, lipid rafts may contribute to the interaction of PMCA4 with proteins involved in Ca2+ signaling at discrete functional positions on the synaptic nerve terminals.

The Ca 2ϩ -ATPases from the plasma membrane (PMCA) 3 provide an important pathway for the removal of the intracellular Ca 2ϩ excess by pumping this ion out of the cell. PMCAs are encoded by at least four different genes resulting in four basic isoforms (PMCA1-4). Additional diversity in each isoform is generated by the alternative splicing of their mRNAs (reviewed in Ref. 1). It is clear that these pumps play major roles in Ca 2ϩ homeostasis maintenance and participate in the local regulation of Ca 2ϩ signaling. Moreover specific demands for Ca 2ϩ regulation must involve the preferential targeting of PMCAs to specific locations in the plasma membrane through a yet unknown mechanism.
The organization of biological membranes into specialized subdomains, known as lipid rafts, is an emergent source of research given their involvement in many essential cellular processes such as signal transduction (2)(3)(4). Lipid rafts are characterized by their high content of cholesterol and glycosphingolipids and serve as suitable platforms for the assembly of protein complexes involved in signaling pathways (3). In neurons, lipid rafts have been found in dendrites where they sustain a variety of postsynaptic protein complexes (5). In fact, their disruption results in the depletion of excitatory and inhibitory synapses and in the loss of dendritic spines, among other processes (5). In addition, accu-mulated evidence shows the involvement of lipid rafts in neurodegenerative diseases such as prion diseases, cerebellar ataxia, and Alzheimer disease (6 -12).
In addition to these raft domains, others, referred to as caveolae, are marked by the presence of caveolin-1 and can share or not a solubility response to detergent similar to that of rafts (2,13,14). In this sense immunocytochemical studies of human fibroblast cultures (15) and Western blots of caveolae isolated from the luminal rat lung endothelial cell surface using Triton X-100 (16) have shown the presence of PMCA in caveolae domains.
In this work we have investigated the differential association of the distinct PMCA isoforms from synaptosomal plasma membranes (SPM) from pig cerebellum with lipid rafts. The analysis of Brij 96 cold extracts fractioned by gradient ultracentrifugation shows that of all PMCA, only the isoform 4 (PMCA4) appears associated with the low density and high cholesterol and the sphingomyelin content of raft fractions. This domain segregation is promoted by palmitoylation because treatment with hydroxylamine reduces its floatation properties. The differences observed in the membrane-domain segregation are likely to relate to the PMCA isoform specificity of the role that they play in cell signaling and other molecular processes. The fact that different proteins involved in neurotransmitter release have been associated with lipid rafts in the nerve terminal suggests a direct role of PMCA4 in these processes.

EXPERIMENTAL PROCEDURES
Preparation of Synaptosomal Plasma Membrane Vesicles-SPM from adult pig cerebellum were prepared by osmotic lysis of synaptosomes as described by Salvador and Mata (17) and stored at Ϫ80°C until use. The protein content was measured using the Bradford method (18) with bovine serum albumin as standard.
Electrophoresis and Immunoblots-Equal volumes of each fraction were separated in 6.5% (PMCA analysis) or 13.5% (PrP C and flotillin analysis) polyacrylamide gels according to Laemmli's procedure (21). Proteins were either visualized with Coomassie Brillant Blue or electrotransferred to polyvinylidene difluoride membranes using a Trans-Blot SD semidry electrotransfer system (Bio-Rad). After blocking for * This work was supported by Grants BFI2002-02548 (to A. M. M.) and BIO2003-00285 (to M. G.) from Direcció n General de Investigació n, MEC, Spain. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. 1 Recipient of a Ph.D. studentship Formació n de Profesorado Universitario from Ministerio de Educació n y Ciencia, Spain. 2 To whom correspondence should be addressed. Tel. and Fax: 34-924289419; E-mail: anam@unex.es. 3 The abbreviations used are: PMCA, plasma membrane Ca 2ϩ -ATPase; PrP C , cellular prion protein; SPM, synaptosomal plasma membrane; TBS, Tris-buffered saline; pAb, polyclonal antibody; mAb, monoclonal antibody; MAGUK, membrane-associated guanylate kinase; HRP, horseradish peroxidase. . After washing, the membranes were incubated for 1 h at room temperature with peroxidase-conjugated secondary antibody and stained with 4-methoxy-1-naphthol. Immunoreactive bands were quantified with the TINA software (Raytest, Straunbenhardt, Germany). Ganglioside GM1 Detection-For the lipid raft-associated ganglioside GM1, 0.5 l of each gradient fraction was spotted onto a nitrocellulose membrane sealed on a dot-blot set-up (Bio-Rad). After washing with TBS-0.5% Tween 20, the membrane was incubated with 6 M guanidine chloride for 5 min at room temperature and then washed extensively  with TBS-0.5% Tween 20. After blocking with TBS containing 3% (w/v) defatted milk for 2 h at room temperature, membranes were incubated with the cholera toxin B subunit-HRP (1/20 000, Calbiochem) overnight at 4°C. The detection was performed by chemoluminescence (Bio-Rad Immune-Star HRP substrate kit). The intensities of spots were quantified with the TINA software.
Cholesterol Determination-Cholesterol content was determined colorimetrically using the ferric chloride reaction in a sulfuric acidacetic acid solution (22). Briefly, 10 l of each fraction was mixed with 80 l of chloroform/methanol (2/1, v/v), and the phase containing the lipids was transferred onto a glass tube and dried with nitrogen. Cholesterol was extracted by the addition to each tube of 0.75 ml of acetic acid and 0.5 ml of ferric chloride reagent (prepared by mixing 1 ml of 2.5% (w/v) FeCl 3 in 85% H 3 PO 4 with 11.5 ml of concentrated H 2 SO 4 ). The samples were incubated for 30 min, and the absorbance was read at 550 nm. Cholesterol standard solutions were used for calibration.
Ca 2ϩ -ATPase Activity-The enzymatic activity was measured in 7.5 l of each fraction by using a coupled enzyme assay as described previously (23). The activities were measured in the presence of 100 nM thapsigargin and 5 mM sodium azide to inhibit the contamination of the endoplasmic reticulum Ca 2ϩ -ATPase (SERCA) and other ATPases. Calmodulin (0.32 g/ml) was added when indicated. The membrane vesicles were permeabilized with 0.01% saponin to obtain the activity of all PMCA molecules. The reaction was started by the addition of 1 mM ATP. The PMCA activity of these fractions was obtained after subtraction of the Mg 2ϩ -ATPase activity, measured in the presence of 6 mM EGTA, and was given in mol P i ⅐min Ϫ1 ⅐mg Ϫ1 .
Determination of Hydroxylamine Sensitivity-SPM (10 mg protein) were treated with 1 M hydroxylamine (pH 7.4) or 1 M Tris (pH 7.4) as a control for 2 h at 4°C under constant stirring. After 1 h of ultracentrifugation at 260,000 ϫ g at 4°C, the membrane pellets were resuspended in 25 mM Tris, pH 7.4, 150 mM NaCl, cold-extracted with Brij 96, and fractioned on a Nycodenz gradient as described above.

RESULTS
To investigate the differential membrane domain segregation of PMCA isoforms we used cold Brij 96-extracts of SPM vesicles isolated from pig cerebellum fractioned by ultracentrifugation along a Nycodenz   JANUARY 6, 2006 • VOLUME 281 • NUMBER 1 FIGURE 5. Alignment of the amino acid sequences of the four PMCA isoforms. The sequences of PMCA isoforms (Swiss-Prot accession numbers P20020, Q01814, P16720, and P23634 of hPMCA1, -2, -3, and -4, respectively) were analyzed using the BLAST algorithm. Gaps were introduced into the sequences to maintain the alignment. Amino acids conserved in PMCA1, -2, and -3, but not in PMCA4, are displayed as bold letters in gray boxes. The transmembrane domains (TMD) are underlined, and the cytoplasmic (C) and extracellular (E) loops are indicated by arrows. The C terminus of hPMCA4a is shown in italics and denoted by an asterisk. The PDZ-binding motif and the potentially palmitoylated Cys (OE) are labeled. The insert depicts a schematic view of the PMCA domains. gradient. The choice of Brij 96 instead of Triton X-100 as detergent for the cold extraction was based on the reported improved preservation of raft entities (20). Regarding the density gradient for the floatation fractionation, Nycodenz instead of sucrose was used to reduce the centrifugation time needed to achieve the density equilibrium, thus favoring the preservation of the functional integrity of the membrane proteins.

PMCA4 Association with Lipid Rafts of Cerebellum Membranes
Fractions from SPM Brij 96 extracts were first analyzed for the lipid raft markers, sphingolipid GM1 and cholesterol (Fig. 1). As described (19,20) the GM1 distribution showed the highest levels at the top of the gradient (fractions 1-4), indicating that these low density fractions were raft-enriched fractions (Fig. 1A). Similarly, cholesterol was found mainly in low density fractions 2-4 at the top of the gradient, with very low levels in the remaining fractions (Fig. 1B). As protein markers of rafts, we used both the cellular prion protein (PrP C ) and flotillin (24,25). Fig. 2A shows that the vast majority of PrP C appeared in low density fractions 2-6, separated from most membrane proteins (Fig. 2C). Flotillin showed a broader distribution, with its highest content found in fractions 3 and 4 (Fig. 2B). This floatation distribution of markers was preserved for SPM cold detergent extractions performed at protein concentrations in the 9 -14 mg/ml range.
The presence and distribution of PMCA in the gradient were first analyzed with an anti-PMCA antibody that recognizes all isoforms, giving a broad staining in all fractions (result not shown). Hence we used isoform-specific antibodies to identify the four PMCA isoforms along the gradient (Fig. 3A). PMCA1, -2, and -3 were detected in the non-raft high density fractions, with stronger staining for PMCA1 and -3 than for PMCA2. Interestingly, the anti-PMCA4 antibody recognized two protein bands at the top of the gradient, around low density fractions 2-4. To test whether the observed bands corresponded to both splicing variants a and b, Western blots of SPM were probed with specific anti-PMCA4a (CR4a) and anti-PMCA4b (JA3) antibodies (26,27) (Fig. 3B). It can be seen that each antibody identified only one band corresponding to each splice variant, whereas the anti-PMCA4 antibody stained both variants in the SPM. The quantization of PMCA1-4 bands (Fig.  3C) pointed out more precisely the bimodal distribution of PMCA isoforms along the Brij 96-Nycodenz gradient, with one peak corresponding to PMCA4 located in the low density raft fraction 3 and another peak corresponding to the other PMCA isoforms located in the high density non-raft fraction 9. Furthermore, similar, if not identical, bimodal distribution was obtained from SPM isolated from pig cerebrum (data not shown).
Measurements of Ca 2ϩ -ATPase activity in all fractions (Fig. 4) showed two pools of PMCA activity that overlapped with those obtained by immunoreactivity. The incubation of membranes from each fraction with 0.32 g/ml calmodulin showed stimulation of the activity just around fraction 3.
The preferential ability of PMCA4, among all of the isoforms, to associate with lipid rafts could be the result of an intrinsic propensity; or this ability could be driven indirectly by the existence of modulatory partners that could either promote it, in the case of PMCA4, or preclude it, in the remaining isoforms. The fact that PMCA4b binds to several PDZ (PSD-95/Dig/ZO-1) proteins of the MAGUK family (28,29), some of which are raft-associated (12, 30 -33), could support an indirect raft association. In fact, a ETSV (PMCA4b) to ETSL (PMCAs 1b, 2b, and 3b) sequence change in the class I PDZ-binding motif found at the C terminus could explain the distinct raft segregation among PMCA b splicing variants (28,34). However, this unique PDZ-mediated association with lipid rafts cannot explain the association capacity of PMCA4a and the lack of association of PMCA2b. In this sense, PMCA4a has a shorter C-terminal domain, lacking the PDZ-binding motif (35), and PMCA2b has been described to interact with MAGUK proteins (29) and does not appear as a raft-associated protein.
To explain this controversy we considered the possibility of raft association as an intrinsic property of PMCA4 isoform. If this would be the case, then PMCA4 should contain in its sequence a divergent difference compared with the rest of the PMCAs (Fig. 5). Among the sequence differences found, we identified two regions in which PMCA4 contains differentially a Cys residue (GLQCRIEQ in the C2 of PMCA4a/b, VDCNQ in the C6 of PMCA4b) compared with a Ser in the remaining isoforms. These Cys residues could sustain a covalent modification such as palmitoylation, which has been essential for the lipid raft localization of several transmembrane proteins (36 -38). To test this possibility we investigated the sensitivity of PMCA4 buoyancy to 1 M hydroxylamine, which is known to reverse protein palmitoylations (39). Fig. 6 shows that a pretreatment with 1 M hydroxylamine shifted the maximum of the floatation distribution of PMCA4 from fraction 3 to 9 compared with the control performed with 1 M Tris, whereas that of PMCA2 and the other isoforms remained unchanged at fraction 9.

DISCUSSION
This study shows for the first time that the cerebellar PMCA4 isoform, but not the other PMCA isoforms, is specifically associated with the low density membrane domains from SPM that exhibit high cholesterol and GM1 content and similar floatation properties to the raftassociated proteins PrP C and flotillin. Interestingly, this association is sensitive to hydroxylamine, indicating protein palmitoylation as a driving force (40).
Membrane microdomains, such as the so-called lipid rafts, function as segregation platforms to concentrate different molecules and assemble them into efficient signal transduction machineries (2,41,42). An increasing number of kinases and others proteins appear to use this strategy for best working in signal cascades (43). In neurons, lipid rafts were first thought to be restricted to the axonal compartment (44), although later it was shown that several postsynaptic proteins located in dendrites are also associated to rafts (5). We have recently shown that in pig cerebellum the PMCA is present in the plasma membrane of the soma of Purkinje cells as well as in areas enclosing many synapses, such as in cerebellar glomeruli and dendritic spines of Purkinje cell arborization (23). However, this pattern was not discriminative regarding the specific isoforms because the analysis was done with an antibody that recognized all PMCA isoforms.
The presence of lipid rafts in SPMs was evidenced in the low density fractions of a gradient ultracentrifugation of cold detergent extract, on the basis of the high content of the lipid markers cholesterol and sphingolipid GM1 and the protein markers PrP C and flotillin (19,20,45). The immunodetection of PMCA proteins using specific antibodies showed that only the PMCA4 isoform appears associated in raft, whereas the PMCA isoforms 1, 2, and 3 emerge in the high density non-raft fractions located closer to the bottom of the gradient, where the majority of the proteins were found. The distribution of PMCA isoforms in two different portions of the gradient was corroborated by functional assays. The lower PMCA activity found in fraction 3 with respect to fraction 9 and its stimulation by calmodulin agrees with the presence of PMCA4 in that fraction, because this isoform has the lowest basal activity and the greatest stimulation by calmodulin among isoforms (46,47). The low level or even the lack of calmodulin stimulation found in fractions 3 and 9, respectively, could be expected, provided that PMCA isoforms are already activated by the presence of a high content of phosphatydyl-serine in SPM from the cerebellum as it has been reported in brain (17). It has been reported that the synaptosomal PMCA activity from pig brain is ganglioside-dependent (48), as the activity is reduced by GM1 and to a greater extent by GM2 and GM3. Thus, the high content of gangliosides in the same fraction as PMCA4 (Figs. 1 and 3) may also account for the lower activity found in fraction 3 with respect to fraction 9. Our activity results about the corresponding decrease of PMCA activity with the increase of sphingomyelin and cholesterol content in reconstituted pig brain PMCA are in good agreement with those reported by Pang et al. (49). The distribution pattern of PMCA isoforms was observed in SPM isolated not only from cerebellum but also from cerebrum, giving a more general than regional view on the process.
Screening for molecular determinants of raft association in PMCA4 revealed the presence of isoform-specific Cys residues in regions corresponding to the C2 and C6 cytoplasmic loops, susceptible a priori of being palmitoylated. The key role of this covalent modification was evidenced in the reduction of PMCA4 buoyancy after treatment with 1 M hydroxylamine. However, it must be kept in mind that the accurate attribution of this modification to PMCA4 instead to a second, yet unidentified, partner will require further experimentation.
Although the physiological relevance of PMCA4 in rafts has not been established, it is tempting to speculate on the advantages of this process. Given that, among all isoforms, PMCA4b has the lowest basal activity but the highest stimulation by calmodulin (50), it has been suggested that PMCA4b might be actively involved in the development, shaping, and duration of the Ca 2ϩ signal (47). Such a scenario agrees with a preferential localization at a definite membrane domain as rafts (see Fig. 7) in which a Ca 2ϩ signaling event must be a very tightly regulated process. FIGURE 7. Distribution of PMCA isoforms in the synaptosomal plasma membrane. The scheme represents the association of PMCA4 with cholesterol/sphingomyelin-rich membrane domains through its putative interaction with MAGUK proteins (for PMCA4b) and/or palmitoylation of Cys residues (for PMCA4a/b in C2 and for PMCA4b in C6). The remaining PMCA isoforms are localized in other membrane domains.