P-Glycoprotein is localized in intermediate-density membrane microdomains distinct from classical lipid rafts and caveolar domains Galina Radeva, Jocelyne Perabo and Frances J. Sharom Department of Molecular and Cellular Biology, University of Guelph, Guelph, Ontario, Canada In recent years, intense interest has been focussed on the properties and biological functions of specialized membrane domains known as lipid rafts [1,2]. Rafts consist of cholesterol–sphingolipid-rich regions within the plasma membrane, stabilized by interactions between cholesterol and the long saturated acyl chains of sphingolipids. They are thought to exist in the liquid-ordered phase, which has properties intermedi- ate between those of the liquid-crystalline and gel phases [3,4]. Acylated and lipid-modified proteins are Key words ABC transporter; caveolin-1; detergent- resistant membranes; lipid rafts; P-glycoprotein Correspondence F. J. Sharom, Department of Molecular and Cellular Biology, University of Guelph, Guelph, Ontario, Canada, N1G 2W1 Fax: +519 837 1802 Tel: +519 824 4120; ext. 52247 E-mail: fsharom@uoguelph.ca (Received 11 May 2005, revised 27 July 2005, accepted 4 August 2005) doi:10.1111/j.1742-4658.2005.04905.x P-glycoprotein (Pgp), a member of the ATP-binding cassette (ABC) super- family responsible for the ATP-driven extrusion of diverse hydrophobic molecules from cells, is a cause of multidrug resistance in human tumours. Pgp can also operate as a phospholipid and glycosphingolipid flippase, and has been functionally linked to cholesterol, suggesting that it might be associated with sphingolipid–cholesterol microdomains in cell membranes. We have used nonionic detergent extraction and density gradient centrifu- gation of extracts from the multidrug-resistant Chinese hamster ovary cell line, CH R B30, to address this question. Our data indicate that Pgp is localized in intermediate-density membrane microdomains different from classical lipid rafts enriched in Src-family kinases. We demonstrate that Brij-96 can selectively isolate the Pgp domains, separating them from the caveolar and classical lipid rafts. Pgp was found entirely in the Brij-96- insoluble domains, and only partially in the Triton X-100-insoluble membrane microdomains. We studied the sensitivity of these domains to cholesterol removal, as well as their relationship to GM 1 ganglioside- and caveolin-1-enriched caveolar domains. We found that the buoyant density of the Brij-96-based Pgp-containing microdomains was sensitive to choles- terol removal by methyl-b-cyclodextrin. The Brij-96 domains retained their structural integrity after cholesterol depletion while, in contrast, the Triton X-100-based caveolin-1 ⁄ GM 1 microdomains did not. Using confocal fluor- escence microscopy, we determined that caveolin-1 and GM 1 colocalized, while Pgp and caveolin-1, or Pgp and GM 1 , did not. Our results suggest that Pgp does not interact directly with caveolin-1, and is localized in inter- mediate-density domains, distinct from classical lipid rafts and caveolae, which can be isolated using Brij-96. Abbreviations ABC, ATP-binding cassette; BSS, buffered saline solution; CTB, cholera toxin B subunit; CTB–HRP, cholera toxin B–horseradish peroxidase conjugate; DRM, detergent-resistant membranes; ECL, enhanced chemiluminescence; MbCD, methyl-b-cyclodextrin; MDR, multidrug resistance ⁄ resistant; MEM, minimal essential medium; MRP, multidrug-resistance-associated protein; NaCl ⁄ P i , phosphate-buffered saline; Pgp, P-glycoprotein. 4924 FEBS Journal 272 (2005) 4924–4937 ª 2005 FEBS often sequestered into lipid rafts, probably as a result of their acyl chain properties; GPI-anchored proteins are found in the outer leaflet, and Src-family tyrosine kinases are found in the inner leaflet. Substantial evi- dence supports the existence of lipid raft microdomains in model membrane systems in vitro, and in intact cells in vivo [5–7], although there is still controversy regard- ing their size and dynamic properties [8]. Pgp (P-glycoprotein, MDR1, ABCB1) is an energy- dependent drug efflux pump that is a member of the ATP-binding cassette (ABC) family of proteins [9]. Pgp decreases the intracellular concentration of a wide variety of drugs and hydrophobic molecules by actively transporting them across the plasma membrane, pow- ered by ATP hydrolysis at two cytoplasmic nucleotide- binding domains. Pgp has been proposed to act as a drug flippase or a hydrophobic ‘vacuum cleaner’ [10]. Under normal physiological conditions Pgp is involved in cellular detoxification leading to cell survival; however, in cancer cells its overexpression confers a multidrug-resistant (MDR) phenotype thus causing chemotherapy failure [11]. An accompanying change in many cells expressing MDR transporters, including Pgp, is elevated levels of certain sphingolipids [12–14]. The ATPase activity of Pgp is modulated by lipids [15–17], and its interaction with drug substrates depends on the lipid surroundings [18]. Pgp has also been associated with active cholesterol redistribution across the plasma membrane [19], and cholesterol affected its drug binding [18,20], transport and ATPase activity [21–24]. Pgp is functional when reconstituted into a sphingomyelin-cholesterol mixture that mimics lipid rafts [25]; however, it can carry out both ATP hydrolysis and drug transport in bilayers of only phos- phatidylcholine [17,26], so cholesterol is not required for its function. As sphingolipids and cholesterol are both components of lipid rafts, the fine interplay between lipid environment and Pgp function may be linked to the membrane microdomain organization of the protein. Raft domains have been isolated from intact cells based on their insolubility in cold nonionic detergents, especially Triton X-100, and their low buoyant density in sucrose gradients. The resulting detergent-resistant membranes (DRM) are believed to arise from the coalescence of smaller raft structures on the cell surface. Caveolin-1, a 21 kDa transmembrane choles- terol-binding protein, is the primary constituent of invaginated plasma membrane structures called caveo- lae. Caveolar and noncaveolar DRM microdomains represent distinct plasma membrane regions [27,28]. Caveolin-1, GM 1 ganglioside and cholesterol are believed to be hallmarks of caveolae which are distinct from the classical lipid rafts that are enriched in GPI- anchored proteins, cholesterol and GM 1 , but do not contain caveolins [29]. Up-regulation of caveolin-1 and caveolae has been observed in MDR cells expressing Pgp, suggesting a functional link between them [30,31]. Interestingly, Pgp was reported to appear in the low density membrane fractions in Triton X-100 extracts [32], as well as in detergent-free cell extracts [21]. Demeule and coworkers found that Pgp was contained in the caveolae in MDR cells and blood–brain barrier endothelial cells [33,34]. In contrast, Hinrichs et al. determined that Pgp was localized in the noncaveolar rafts [35], while flow cytometry and confocal microsco- py showed that a substantial fraction of Pgp was asso- ciated with lipid rafts and the cytoskeleton in human colon carcinoma cells [36]. We recently reported that the nonionic detergents Brij-96 and Triton X-100 isolated different lipid raft microdomains from rat basophilic leukemia (RBL- 2H3) cells [37]. We therefore employed these detergents to investigate the microdomain localization of Pgp in the MDR cell line CH R B30. In the present work, we showed that this ABC transporter is localized in inter- mediate-density membrane microdomains that are dis- tinct from caveolar domains and Src kinase-containing classical lipid rafts. We also showed that these domains are differentially extracted by Brij-96, but not by Triton X-100. In addition, we found that Brij-96 segregates caveolar domains from Src kinase-based classical lipid rafts, leading to distinct sets of fractions containing each class of raft. Triton X-100 extraction apparently leads to the copartitioning of different types of membrane microdomains ino a common pool. Tri- ton X-100 rafts are disrupted by cholesterol removal, whereas the Brij-96 rafts change their buoyant density, but maintain their structural integrity. Results Pgp is localized in intermediate-density membrane microdomains DRM are commonly isolated by cold nonionic deter- gent extraction followed by sucrose density centrifuga- tion. We previously showed that Brij-96 and Triton X-100 isolate lipid rafts with different physical and biochemical properties from RBL-2H3 cells [37]. In this work, we used a similar approach to investigate the membrane domain localization of Pgp. Brij-96 or Triton X-100 extracts of CH R B30 cells expressing Pgp were subjected to sucrose density gradient flotation, and the distribution of Pgp among the fractions was determined by western blotting. G. Radeva et al. P-glycoprotein in intermediate-density microdomains FEBS Journal 272 (2005) 4924–4937 ª 2005 FEBS 4925 Triton X-100 extraction of CH R B30 cells yielded a bimodal distribution for Pgp (Fig. 1A, right panel, top row). This type of distribution is typical for the protein constituents of classical lipid rafts, such as GPI- anchored proteins and the Src tyrosine kinases. A significant amount of Pgp was observed in the low- density lipid raft fractions 4–6, while the majority of the transporter remained in the high-density fractions 9–11. However, in the density gradient fractions from Brij-96 extracts, Pgp displayed a continuous distribu- tion in fractions 2–10, peaking in fractions 5 and 6 (Fig. 1A, left panel, top row). This pattern of Pgp par- titioning along the sucrose density gradient is quite unlike the picture observed for the known constituents of lipid rafts, such as GPI-anchored proteins or Src kinases. In our previous work, we showed that under the same conditions, the lipid raft protein Thy-1 of RBL-2H3 cells is concentrated entirely in the lowest density lipid raft fractions 2–4 [37]. We have therefore termed the fractions in which Pgp is incorporated, as intermediate-density fractions. The data indicated that all of the Pgp in CH R B30 cells is localized in Brij-96- based domains that are completely resistant to solubili- zation with this detergent. Furthermore, Pgp is only partially located in the Triton X-100-resistant DRM, and about half of it can be solubilized by extraction with this detergent. Pgp is an N-glycosylated protein [38], and because glycosylation may affect the membrane domain local- ization of proteins, we examined whether the glycosy- lation status of Pgp had any bearing on its distribution in the density gradient after extraction using Brij-96. To address this, we used the CH R PHA R cell line (a lectin-resistant variant of the parental line used to derive CH R B30), which is deficient in N-linked glyco- sylation. The results presented in Fig. 1A show that the profile for Pgp localization in the sucrose density gradient is very similar when either cell line is used with each of the detergents (compare first and second rows). We conclude that glycosylation does not play a role in the partitioning of Pgp into intermediate- density microdomains. A B Fig. 1. Sucrose density gradient partitioning of P-glycoprotein (Pgp) and markers of clas- sical lipid rafts. (A) CH R B30 cells or CH R PHA R cells (second row only) were lysed in either 0.5% (w ⁄ v) Brij-96 or 1% (w ⁄ v) Triton X-100 at 4 °C, and postnuclear lysates were fractionated by ultracentrifuga- tion on a discontinuous sucrose gradient. A total of 13 fractions was collected from the top of the gradient tube and an aliquot from each fraction was run on SDS ⁄ PAGE. Separ- ated proteins were transferred to a nitrocel- lulose membrane and the presence of Pgp, Yes, caveolin-1 (Cav-1), and CD71 was observed by western immunoblot analysis and enhanced chemiluminescence (ECL) detection, as described in the Experimental procedures. (B) CH R B30, RBL-2H3 and Jur- kat cells were lysed in Triton X-100 at 4 °C. Lysates were precleared by centrifugation at 10 000 g for 5 min. An aliquot from each extract was run on SDS ⁄ PAGE, and the separated proteins were transferred to a nitrocellulose membrane and analyzed for Src-family kinases (Lck, Lyn, and Yes) by western immunoblot analysis and ECL. P-glycoprotein in intermediate-density microdomains G. Radeva et al. 4926 FEBS Journal 272 (2005) 4924–4937 ª 2005 FEBS Comparison of the density gradient partitioning of Pgp and markers of classical lipid rafts CH R B30 cells have little or no expression of the most common GPI-anchored proteins, such as Thy-1, alka- line phosphatase, decay accelerating factor, etc. We therefore employed Src family tyrosine kinases for the identification of classical lipid rafts and comparison with the intermediate-density Pgp membrane domains. First, we determined which members of the Src tyro- sine kinase family are expressed in CH R B30 cells, using extracts from RBL-2H3 and Jurkat cells as positive controls (Fig. 1B). Of the three proteins we tested for (Lck, Lyn and Yes), Lyn was expressed only in RBL- 2H3 cells, and Lck only in Jurkat cells, whereas Yes was seen in both of these cell lines. In CH R B30 cells we found that only Yes kinase was detectable. Next, we investigated how the distribution of Yes kin- ase along the density gradient compared with that of Pgp. When Triton X-100 was used, Yes kinase resided in almost the same fractions as Pgp. A portion of Yes kinase partitioned into the low-density sucrose fractions 3–6, while a significant amount (about half) remained in the high-density fractions 10–13 (Fig. 1A, third row, right panel). When Brij-96 was used, Pgp and Yes kinase did not copartition, as determined by density centrifuga- tion (Fig. 1A, third row, left panel). Yes kinase was localized exclusively in the lowest-density fractions 1–4. This localization is similar to that obtained for another Src-family tyrosine kinase, Lyn, whose sucrose density gradient partitioning was examined in RBL-2H3 cells following extraction with Brij-96 [37]. The data presen- ted in Fig. 1A indicate that Pgp is localized in mem- brane microdomains that are distinct from classical lipid rafts containing Src-family tyrosine kinases. The mem- brane microdomains containing Pgp displayed an inter- mediate density in the sucrose gradient when Brij-96 was used. They can be separated from classical lipid rafts if extracted with Brij-96, but not with Triton X-100. The total protein content of each fraction was meas- ured by the bicinchoninic acid assay, as shown in Fig. 2 (lower panel). Both detergents solubilized the majority of cellular proteins, leaving them in the high- density soluble fractions 11–13. Brij-96 extraction resulted in small, but detectable, amounts of protein in the low density fractions, whereas Triton X-100 extrac- tion resulted in virtually no protein in these fractions. Relationship of the intermediate-density Pgp-containing microdomains to caveolae It is well-documented that at least two types of deter- gent-insoluble membrane microdomains exist. The first class encompasses the classical lipid rafts (or noncaveolar lipid rafts), which contain GPI-anchored proteins and Src-family kinases, while the other class represents the caveolar raft microdomains, with cave- olin as a hallmark protein. We examined the possible relationship between the intermediate-density mem- brane structures in which Pgp is found, and caveolae structures, by assessing copartitioning of Pgp and caveolin-1 in the sucrose density gradient fractions. Caveolin-1 was concentrated in fractions 4–8 in both of the detergent extracts, although in the case of Tri- ton X-100 there was tailing out to fractions 11–12 (Fig. 1A, fourth row). Importantly, the localization of caveolin-1 displayed a significant overlap with that of Pgp in both the low-density fractions from Triton X-100 extracts and in the intermediate-density frac- tions from Brij-96 extracts (Fig. 1A, compare the first row with the third row). These results suggest two interesting possibilities. First, Brij-96 appears to differentially isolate the caveolar (fractions 4–8, caveolin-1 marker) from Fig. 2. Protein and GM 1 profile of Triton X-100 and Brij-96 sucrose density gradients. Post-nuclear lysates of detergent extracts of CH R B30 cells were run on sucrose gradients, and the gradient frac- tions were assayed for the distribution of total protein and GM 1 ganglioside, as described in the Experimental procedures. The pro- tein content is shown for a 20 lL aliquot of each gradient fraction from 5–10 · 10 8 cells lysed in 1 mL of buffer, and the activity of cholera toxin B–horseradish peroxidase conjugate (CTB–HRP) in a 50 lL aliquot of each gradient fraction from 2–3 · 10 8 cells is indi- cated. Data are displayed as the mean ± range; where error bars are not visible, they are contained within the symbols. G. Radeva et al. P-glycoprotein in intermediate-density microdomains FEBS Journal 272 (2005) 4924–4937 ª 2005 FEBS 4927 noncaveolar (fractions 1–4, Yes marker) lipid rafts. Such a distribution was not observed in the sucrose density fractions from the Triton X-100 extracts. The observed differences were not an artefact of the deter- gent used because the integral membrane protein, CD71 (transferrin receptor), was fully solubilized by both Brij-96 and Triton X-100 (Fig. 1A, bottom row). Second, the Pgp distribution profile partially over- lapped with that of caveolin-1, but not with that of Yes kinase, when Brij-96 was used. This observation suggests that the intermediate-density fractions con- taining Pgp might represent caveolar membrane struc- tures. Such an idea is in agreement with previous reports suggesting a close interaction between Pgp and caveolin-1 [33,34]. However, this observation does not necessarily signify molecular colocalization of Pgp and caveolin-1. For example, all three proteins – Pgp, Yes, and caveolin-1 – segregate into membrane domains of similar density when Triton X-100 is used to prepare the cell extracts, but exhibit different distribution patterns in the case of Brij-96 detergent extraction (Fig. 1A). The potential colocalization of Pgp and caveolin-1 was therefore further examined directly by confocal fluorescence microscopy and immunoprecipi- tation experiments, as described below. Identification of lipid raft microdomains by detection of lipid raft-associated GM 1 ganglioside GM 1 ganglioside is a known marker of both classical lipid rafts and caveolae. This glycosphingolipid has been shown to cofractionate not only with markers of various detergent-insoluble microdomains (such as caveolae, GPI-anchored protein-enriched rafts, and glycosphingolipid-enriched domains), but also to colo- calize with caveolin-1 [39]. We employed a cholera toxin B–horseradish peroxidase conjugate (CTB–HRP) enzyme assay to identify the fractions into which lipid raft-associated GM 1 partitions (Fig. 2, upper panel). For Triton X-100, these were fractions 5, 6, and 7. In the density gradient of Brij-96 extracts, GM 1 was detected in fractions 2–5. This pattern is very similar to that observed for GM 1 in RBL-2H3 cells [37]; how- ever, the peak seen for GM 1 localization in the CH R B30 gradient fractions is somewhat broader. The gradient partitioning of GM 1 (Fig. 2, upper panel) par- tially overlaps with the Yes kinase classical lipid rafts fractions on the one hand, and with the caveolin-1- enriched raft fractions on the other (Fig. 1A). This broader profile can be explained by the fact that GM 1 is a constituent of both classical lipid rafts and caveo- lae. The high level of GM 1 apparently present in the high density fractions of the gradient in Fig. 2 (upper panels), which is not observed in the dot-blots (Fig. 6, panel B) is probably spurious. This was also reported by Blank et al. [40], and could arise from the presence of soluble HRP-like activity in the cells. Examination of Pgp and caveolin-1 localization by confocal fluorescence microscopy and immunoprecipitation We wanted to determine whether the intermediate- density fractions containing Pgp in the Brij-96 extract represent caveolar membrane microdomains. Demeule et al. reported that Pgp and caveolin-1 coimmunopre- cipitated in extracts from Pgp-expressing CH R C5 cells and brain capillary membranes [33]. We tested the coimmunoprecipitation of Pgp and caveolin-1 using the pooled lipid raft fractions from the Brij-96 and Triton-100 extracts. However, we were unable to detect coimmunoprecipitation between the two pro- teins under these conditions. We decided therefore to examine their potential association using total cell extracts because these would contain the entire pool of Pgp and caveolin-1. Cell extracts were prepared using various lysis buffer conditions. Combinations of different detergents were used to establish whether the choice of detergent plays a role in the observation of coimmunoprecipitation of the two proteins. Hinrichs and coworkers had already reported a weak associ- ation of multidrug resistance-associated protein 1 (MRP1) with caveolin-1 when Lubrol was used, but saw no such association in the presence of Triton X- 100 [35]. In our experiments, all buffers contained sufficient detergent to disrupt the vesicles previously observed to exist in the DRM fractions [37]. Other- wise, a false impression of coimmunoprecipitation would be obtained if the two proteins were simply located in the same vesicular structure. Under these conditions, only a very faint band of Pgp was seen in the caveolin-1 immunoprecipitates (Fig. 3A, top), but no signal for caveolin-1 was detected in the Pgp immune complexes (Fig. 3A, bottom), suggesting that there is no significant coimmunoprecipitation between the two proteins. A signal for caveolin-1 in Pgp immu- noprecipitates was seen only after prolonged overnight exposure (Fig. 3B, bottom), while an enhanced Pgp band was seen in the caveolin-1 immunoprecipitates when the film was overexposed (Fig. 3B, top). We sug- gest that either a very small fraction of the two pro- teins is associated with each other, or that they are located close together in the membrane, but not directly interacting with one another. This result agrees with the results of confocal immunofluorescence analysis (see below) and is in accordance with the P-glycoprotein in intermediate-density microdomains G. Radeva et al. 4928 FEBS Journal 272 (2005) 4924–4937 ª 2005 FEBS work of Hinrichs et al. [35], who also reported no coimmunoprecipitation between the two proteins in 2780AD human ovarian carcinoma cells. We further examined the possible cellular colocaliza- tion of the two proteins by confocal fluorescence microscopy. We first examined the colocalization of GM 1 and caveolin-1 in CH R B30 cells. The individual staining patterns for GM 1 and caveolin-1 were very similar; bright punctuate spots were observed, mainly on the plasma membrane (Fig. 4A,B). When the sig- nals from the two dyes, Alexa 488 and Alexa 594 , were superimposed, there were several areas of clear over- lap, as indicated by the yellow colour (Fig. 4C, over- lay). This observation indicates a close colocalization of GM 1 and caveolin-1 in some regions of the cell, probably in the caveolar raft domains. Next we investi- gated the localization of Pgp and GM 1 (Fig. 4D,E). We found that the cellular localization patterns of Pgp and GM 1 were distinct and did not overlap, indicating that the two molecules do not directly interact (Fig. 4F). When the localization of caveolin-1 and Pgp was compared, the signals for these proteins were, once again, very distinct (Fig. 4G,H). Both caveolin-1 and Pgp maintained the pattern described above. Pgp displayed staining at the plasma membrane but also A B Fig. 3. Immunoprecipitation of P-glycoprotein (Pgp) and caveolin-1. (A) Lanes 1 and 6, Brij-96 extracts; lanes 2 and 7, Triton X-100 extracts; lanes 3 and 8, Nonidet P-40 ⁄ Triton X-100 ⁄ octylglucoside extracts; lanes 4 and 9, Brij-96 ⁄ radioimmunoprecipitation assay (RIPA) extracts; lanes 5 and 10, Triton X-100 ⁄ RIPA extracts. One microgram of each anti-Pgp or anti-(caveolin-1) immunoglobulin was added to 500 lL of cell extracts. Immune complexes were collec- ted on Protein-A–Sepharose beads and washed twice in the appro- priate buffer. The immunoprecipitated proteins were extracted in Laemmli’s sample buffer. One half of each immunoprecipitation sample was run on 7.5% nonreducing SDS ⁄ PAGE for Pgp analysis (A and B top). The other half of each sample was run on 12% non- reducing SDS ⁄ PAGE for caveolin-1 analysis (A and B, bottom). Sep- arated proteins were transferred to a nitrocellulose membrane and analysed by western immunoblot (IB) analysis and enhanced chemi- luminescence. The film exposure time in (A) was 5 min; (B) is an overnight exposure of (A). GM 1 Cav-1 Cav-1 overlay GM 1 Pgp overlay Pgp overlay A BC D EF G H I Fig. 4. Confocal fluorescence microscopy analysis of P-glycoprotein (Pgp) and caveolin-1 localization. CH R B30 cells grown in monolayer culture were first labelled with cholera toxin B–horseradish peroxi- dase conjugate (CTB–HRP), as described in the Experimental procedures. Cells were then fixed in 4% paraformaldehyde in phos- phate-buffered saline (NaCl ⁄ P i ), pH 7.4, permeabilized in 0.1% (v ⁄ v) Triton X-100 and blocked in 5% (w ⁄ v) skim milk. Pgp and caveolin-1 (Cav-1) proteins were detected with mouse and rabbit immunoglob- ulin, respectively, and localization was revealed with anti-species immunoglobulin conjugated to either Alexa 488 (green) or Alexa 594 (red) fluorophores. CTB–HRP was conjugated to the Alexa 488 fluoro- phore. The overlay image was produced by superimposing the image from the green and red channels, using LCS Lite software. G. Radeva et al. P-glycoprotein in intermediate-density microdomains FEBS Journal 272 (2005) 4924–4937 ª 2005 FEBS 4929 showed intracellular and perinuclear staining. When an overlay image from the two proteins was produced, no association between Pgp and caveolin-1 was observed, as indicated by the absence of any yellow colour (Fig. 4I). Cholesterol distribution in sucrose density fractions following treatment of CH R B30 cells with methyl-b-cyclodextrin Cholesterol is a component of classical lipid rafts that is proposed to be required for their structural integrity. The removal of cholesterol by various agents often leads to the disruption of microdomain structures, which is manifested by detergent extraction of mole- cules residing there or by altered activity of signalling components. We have demonstrated previously that Pgp is localized in intermediate-density membrane microdomains, which are distinct from classical lipid rafts and caveolar rafts. Our next step was to examine whether cholesterol plays a role in the formation and stabilization of these domains. One commonly used agent for the depletion of cholesterol is methyl-b-cyclo- dextrin (MbCD). We tested various concentrations of MbCD (5–50 mm) with CH R B30 cells, and found that treatment with 20 mm MbCD for up to 1 h produced the maximal cholesterol depletion while still preserving cell viability (G. Radeva & S. J. Sharom, unpublished data). We measured the cholesterol content of the density gradient fractions of extracts from untreated CH R B30 cells and from cells treated with MbCD, and found that cholesterol was effectively removed from the lipid raft fractions obtained using both detergents (Fig. 5). In untreated cells, cholesterol displayed a bimodal dis- tribution profile when the extracts were prepared with Triton X-100. One peak of cholesterol was seen around fractions 4–6 and another in fractions 9–13 (Fig. 5, bottom). This pattern corresponds to the lipid raft marker protein distribution for this detergent (Fig. 1A). This cholesterol distribution pattern is also similar to that reported in our previous study in the RBL-2H3 cell line [37]. When CH R B30 cells were trea- ted with MbCD and then extracted with Triton X-100, the cholesterol content was dramatically reduced in lipid raft fractions 4–6, and to a lesser extent in fractions 9–13. When lipid rafts were isolated using Brij-96, cholesterol partitioned into a single peak exclusively in fractions 1–5, which falls into the region where protein markers of classical lipid rafts segregate (Fig. 1A). Upon treatment with MbCD, cholesterol was significantly depleted from these fractions (Fig. 5, top). Effect of cholesterol depletion on Pgp, caveolin-1 and GM 1 distribution in the sucrose density gradient After we determined that cholesterol was effectively depleted from lipid raft fractions upon treatment with MbCD, we examined whether cholesterol removal had an effect on the distribution of Pgp, caveolin-1 and GM 1 in the sucrose density gradient. We found that Pgp located in the low-density raft fractions 4–6 in untreated cells was shifted slightly towards the high- density fractions when cells were extracted with Triton X-100 (Fig. 6A, right panel). In addition, more Pgp appeared in the high-density soluble fractions relative to those of low density upon cholesterol depletion. Fig. 5. Cholesterol distribution in sucrose density fractions follow- ing treatment of CH R B30 cells with methyl-b-cyclodextrin (MbCD). CH R B30 cells treated with 20 mM MbCD (grey bars) or untreated control cells (black bars) were lysed in either 0.5% (w ⁄ v) Brij-96 or 1% (w ⁄ v) Triton X-100. Post-nuclear cell extracts were then run on sucrose gradients, and the separated gradient fractions were assayed for the distribution of cholesterol as described in the Experimental procedures. The cholesterol content is shown for the entire gradient fraction from 1–2 · 10 8 cells lysed in 300 lL of buf- fer, as the mean ± range. The results shown in Figs 5 and 6 were obtained using the same set of gradient fractions. P-glycoprotein in intermediate-density microdomains G. Radeva et al. 4930 FEBS Journal 272 (2005) 4924–4937 ª 2005 FEBS Interestingly, in Brij-96 extracts, there was a significant shift in the Pgp distribution after cholesterol depletion. Pgp was located primarily in fractions 6–11 following cholesterol depletion, as compared to fractions 3–10 for the untreated cells (Fig. 6A, left panel). This indi- cates a substantial change in the buoyant density of the intermediate-density fractions upon cholesterol depletion. GM 1 distribution also shifted towards the higher-density fractions of the gradient after choles- terol was removed (Fig. 6B), for rafts isolated using Brij-96. However, no Pgp was found in the high- density fractions, suggesting that these microdomains retain their structural integrity on cholesterol deple- tion. However, when rafts were extracted with Triton X-100 following cholesterol removal, GM 1 was not only shifted to slightly higher density in the raft frac- tions but could also be seen in the high-density soluble fractions 10–11 (Fig. 6B). Caveolin-1 also showed dif- ferent behaviour in the two detergent extracts. The protein shifted towards the higher-density fractions upon treatment with MbCD in Brij-96 cell extracts (Fig. 6C, left panel). However, if cholesterol-depleted cells were extracted with Triton X-100, a large fraction of the caveolin-1 partitioned into the high-density sol- uble fractions 10–12 (Fig. 6C, right panel), while the remaining protein remained localized in fractions 5–7. This behaviour is similar to that seen for GM 1 under the same conditions. These results suggest that the domains in which GM 1 and caveolin-1 are located prior to cholesterol depletion, corresponding to the low density fractions, require cholesterol for their sta- bilization and are disrupted when it is removed. Discussion The lipid raft hypothesis proposes the existence of discrete microdomains in cellular plasma membranes, which arise from the specific interactions of sphingo- lipids, glycosphingolipids and cholesterol. Pgp has recently been proposed to mediate active cholesterol redistribution in the plasma membrane [19]. It has also been reported that MDR cells display differential expression and accumulation of glycosphingolipids [12–14]. These observations were suggestive of a speci- fic membrane domain organization for Pgp, prompting us to examine this issue using techniques commonly A B C Fig. 6. Effect of cholesterol removal on the distribution of P-glycoprotein (Pgp), GM 1 and caveolin-1 in the sucrose density gradi- ent. CH R B30 cells treated with 20 mM methyl-b-cyclodextrin (MbCD) or untreated control cells were lysed in either 0.5% (w ⁄ v) Brij-96 or 1% (w ⁄ v) Triton X-100. Post-nuclear lysates were fractionated on a 5–30% discontinuous sucrose gradient, and 13 fractions were collected. An aliquot from each fraction was run on SDS ⁄ PAGE, and the separated proteins were transferred to a nitrocellulose membrane and analysed for Pgp (A) and caveolin-1 (C) by western immunoblot (IB) analysis and enhanced chemiluminescence (ECL) detection. GM 1 (B) detection was carried out by dot-blot analysis. The results in Figs 5 and 6 were obtained using the same set of gradient fractions. Note that these experiments were carried out under somewhat different condi- tions from Fig. 1; as a result, the distribution of caveolin-1 in the sucrose gradient is slightly narrower. G. Radeva et al. P-glycoprotein in intermediate-density microdomains FEBS Journal 272 (2005) 4924–4937 ª 2005 FEBS 4931 employed to study lipid rafts, namely cold nonionic detergent extraction and sucrose density centrifugation. We have recently demonstrated that in RBL-2H3 cells, Brij-96 and Triton X-100 isolate DRM with different physical and biochemical properties [37]. Here we present evidence that Pgp is localized in intermediate- density membrane microdomains that are completely insoluble in Brij-96, but partially soluble in Triton X-100. Other ABC transporter proteins appear to reside in Lubrol WX-resistant domains, but not in Tri- ton X-100-resistant domains [35,41,42]. Two yeast ABC transporters have been reported to be involved in trafficking cholesterol specifically from lipid raft micro- domains in the plasma membrane to the endoplasmic reticulum, thus facilitating exogenous sterol uptake into the cell [43]. We used Yes kinase as a marker for classical lipid rafts and found that Pgp does not segregate with this protein upon extraction with Brij-96. Thus, the inter- mediate-density domains containing Pgp generated using Brij-96 are distinct from classical lipid rafts. There have been other reports of the existence of non- classical rafts. For example, hepatitis C core protein was associated with DRM that did not colocalize with GM 1 or caveolin-1, and Drobnik et al. found that the GPI-anchored molecules CD14 and CD55 did not colocalize with ABCA1 after isolation of Lubrol rafts [41]. In polarized HepG2 cells, Lubrol WX-insoluble and Triton X-100-insoluble domains with differing properties were functionally linked to distinct traffick- ing pathways in the apical targeting of proteins [42]. Lubrol WX-based rafts were also described where var- ious ABC proteins were entirely recovered in the Lubrol-insoluble fractions and only partially (or not at all) in the Triton X-100-insoluble fractions [35,41,42]. These observations are consistent with our findings that Pgp extracted from CH R B30 cells is partially solu- bilized by Triton X-100 but is completely resistant to Brij-96 solubilization. We previously showed that the degree of enrichment of microdomain constituents in various regions of the density gradient depends on the ratio of cell number to detergent [37]. The observed differences in microdomain localization of ABC pro- teins might therefore reflect variations in the amount of cellular starting material relative to detergent. Indeed, we found it necessary to double the cell : detergent ratio when using CH R B30 cells, com- pared to RBL-2H3 cells, in order to detect the protein constituents of lipid rafts in the sucrose gradient frac- tions. Proteins that partition into lipid rafts are generally those with lipid modifications, such as GPI-anchored proteins, or acylated proteins that are members of the Src tyrosine kinase family, while many integral mem- brane proteins appear to be excluded. Recent data, including the present work, points out that multispan- ning proteins of the ABC transporter superfamily may display lipid raft domain localization [35,41,42]. Cyc- lic-nucleotide-gated channels also appear to be targeted to lipid rafts [44]. It is conceivable that some proteins with transport functions may be organized into mem- brane microdomains, probably together with regula- tory molecules, thus providing an additional level of control over the entry and exit of their substrates. One of the apparent differences between the Brij-96 and Triton X-100-insoluble microdomains in CH R B30 cells is their buoyant density, which is determined by lipid composition and protein content. Cholesterol is often required for maintaining lipid rafts but may also modulate Pgp catalytic and transport activity [19,21,36]. We found that the Brij-96-insoluble mem- branes contain most of the cellular cholesterol, while the Triton X-100-insoluble domains comprise only a fraction of total cholesterol, the remainder of which is located in the high-density soluble fractions. However, in RBL-2H3 cells, most of the cholesterol in Triton X-100 extracts was detected in the low-density frac- tions [37], indicating that cell-specific differences exist in raft microdomain detergent solubility. Drobnik et al. also detected a lower percentage of cellular cho- lesterol in the low-density fractions of Triton X-100 lysates, as compared to high-density fractions, in human skin fibroblasts but not in monocytes [41]. Their data corroborate our current findings and sug- gest that the ratio of cholesterol in the low-density vs. high-density fractions in Triton X-100 extracts is a cell type-specific phenomenon. Upon depletion of cellular cholesterol by Mb CD treatment, the Pgp-containing intermediate-density domains isolated using Brij-96 showed a shift to higher buoyant densities. However, the domains retained their structural integrity as no Pgp was solubilized into the high-density fractions. Cholesterol may not be neces- sary for the maintenance of some types of membrane microdomains, for example those containing K-ras [45] and galectin-4 [46]. In contrast, the Pgp-containing Tri- ton X-100 microdomains remaining after cholesterol depletion showed only a small shift in buoyant density. However, a significant fraction of these domains appeared to have been disrupted, so that more Pgp appeared in the soluble high-density fractions, indica- ting a strong cholesterol requirement for maintenance of their integrity. This finding also suggests that the reason only a fraction of the cellular Pgp is observed in the Triton X-100-insoluble fractions may be that cholesterol is removed from these domains upon P-glycoprotein in intermediate-density microdomains G. Radeva et al. 4932 FEBS Journal 272 (2005) 4924–4937 ª 2005 FEBS detergent treatment. Indeed, a larger proportion of cel- lular cholesterol is seen in the high-density fractions of Triton X-100 extracts, in contrast to Brij-96 extracts. Interestingly, the effect of cholesterol removal on GM 1 and caveolin-1 distribution was far more profound in the Triton X-100 rafts than in the Brij-96 rafts. The GM 1 - and caveolin-1-containing domains not only showed a shift to higher density, but significant amounts of GM 1 and caveolin-1 were also detected in fractions 10–12, indicating that cholesterol depletion leads to their solubilization. GM 1 and caveolin-1 do not behave identically to Pgp, probably because the critical level of cholesterol required to maintain their raft association is different, so their sensitivity to cho- lesterol depletion varies. Possible associations between Pgp, GM 1 and caveo- lin-1 were investigated by confocal fluorescence micros- copy. Clear colocalization was seen between caveolin-1 and GM 1 , consistent with fact that caveolar fractions are enriched in GM 1 [39]. However, we did not observe colocalization between Pgp and caveolin-1, or Pgp and GM 1 . Our findings agree with those of Hinrichs et al. who reported that the ABC transporter MRP1 does not colocalize with caveolin-1 and is enriched in non- caveolar detergent-insoluble domains [35]. However, Demeule et al. reported coimmunoprecipitation of Pgp and caveolin-1 in CH R C5 cells and brain endothelial cells [33,34]. We were unable to see any interaction between Pgp and caveolin-1 by coimmunoprecipitation under conditions where the raft vesicles are solubilized by detergent. It is therefore possible that Pgp and cave- olin-1 are localized in neighbouring raft domains at the plasma membrane and copartition into the same DRM after detergent extraction, but there is no direct, strong association between them. Alternatively, preservation of their interaction is highly dependent on the ratio of cell lipid ⁄ protein : detergent employed during extrac- tion. The association between Pgp and caveolin-1 may be cell type-specific, but the CH R B30 cell line used in this study was derived from CH R C5, so this seems unlikely. Lipid raft microdomains are proposed to exist in the more highly ordered l o phase, compared to the bulk membrane lipids, which are in the liquid-disordered l d phase. Work with the fluorescent probe, merocyanine 540, showed that increasing Pgp expression in MDR cells correlated with an increase in the packing density of the plasma membrane outer leaflet, relative to that of the drug-sensitive parent [47], perhaps reflecting larger numbers of raft microdomains containing Pgp. Unlike many membrane transporters, which often cease to function in rigid gel phase bilayers, the rate of Pgp- mediated drug transport remained high in the gel phase [26], suggesting that ordered microdomains may be help- ful to the function of the protein. Pgp-mediated ATP hydrolysis was also efficient in the gel phase, with a lower activation energy, E act , than in the liquid-crystal- line phase [17]. The intermediate density microdomains in which Pgp is located may therefore provide a suitable environment for the protein to function optimally. Experimental procedures Materials The anti-Pgp monoclonal antibody, C219, was supplied by ID Laboratories (London, ON, Canada). Anti-Lyn, anti- Yes, anti-Lck, anti-caveolin-1 and anti-CD71 (transferrin receptor) mouse monoclonal antibodies were purchased from Santa Cruz Biotechnology (Santa Cruz, CA, USA). CTB–HRP, MbCD, Protein A–agarose, Protein G–agarose, phenylmethanesulfonyl fluoride, Nonidet P-40, DNase, pepstatin A, n-octylglucoside and leupeptin were purchased from Sigma Chemical Co. (St Louis, MO, USA). HRP-labelled goat anti-rabbit and goat anti-mouse immu- noglobulin were purchased from Jackson Immunoresearch Laboratories (Mississauga, ON, Canada). Triton X-100 was supplied by Roche Diagnostics (Laval, QC, Canada), Brij- 96 was obtained from Fluka (Oakville, ON, Canada), and SDS was purchased from Fisher Scientific (Whitby, ON, Canada). Cells The highly MDR Chinese hamster ovary cell line, CH R B30, and a glycosylation deficient lectin-resistant variant, CH R PHA R , were as described previously [48]. Cells were grown as monolayers in a-minimal essential medium (a-MEM) containing 10% (v ⁄ v) fetal bovine serum (Hy- clone, Logan, UT, USA) supplemented with 2 mm glutamine and 2 mm penicillin ⁄ streptomycin, at 37 °C in a humidified atmosphere of 5% (v ⁄ v) CO 2 in the presence of 30 l g Æ mL )1 colchicine. Typically, cells were harvested using 0.25% (w ⁄ v) trypsin or 5 mm EDTA in phosphate-buffered saline (NaCl ⁄ P i , pH 7.4). The RBL-2H3 cell line was cultured as described previously [37]. Jurkat cells were grown using the same culture medium and conditions as CH R B30 cells. Isolation of lipid raft microdomains using sucrose gradient centrifugation Lipid rafts were isolated from either freshly harvested or frozen cells, using Triton X-100 or Brij-96, as described pre- viously for RBL-2H3 cells [37]. About 5–10 · 10 8 cells (200–250 lL cell pellet) were washed twice in NaCl ⁄ P i , pH 7.4 or Tris-buffered saline (TBS; 25 mm Tris ⁄ HCl, 140 mm NaCl, pH 7.5) and then treated on ice with 1 mL G. Radeva et al. P-glycoprotein in intermediate-density microdomains FEBS Journal 272 (2005) 4924–4937 ª 2005 FEBS 4933 [...]... Demeule M, Jodoin J, Gingras D & Beliveau R (2000) P-glycoprotein is localized in caveolae in resistant cells and in brain capillaries FEBS Lett 466, 219–224 Jodoin J, Demeule M, Fenart L, Cecchelli R, Farmer S, ´ Linton KJ, Higgins CF & Beliveau R (2003) P-glycoprotein in blood–brain barrier endothelial cells: interaction and oligomerization with caveolins J Neurochem 87, 1010–1023 Hinrichs JWJ, Klappe... (2004) Lipid rafts and plasma membrane microorganization: insights from Ras Trends Cell Biol 14, 141–147 46 Hansen GH, Immerdal L, Thorsen E, Niels-Christiansen LL, Nystrom BT, Demant EJF & Danielsen EM (2001) FEBS Journal 272 (2005) 4924–4937 ª 2005 FEBS P-glycoprotein in intermediate-density microdomains 47 48 49 50 51 52 53 Lipid rafts exist as stable cholesterol-independent microdomains in the... GJ, Schindler H & Schmitz G (2002) Apo AI ⁄ ABCA1-dependent and HDL3-mediated lipid efflux from compositionally distinct cholesterolbased microdomains Traffic 3, 268–278 42 Slimane TA, Trugnan G, Van Ijzendoorn SCD & Hoekstra D (2003) Raft-mediated trafficking of apical resident proteins occurs in both direct and transcytotic pathways in polarized hepatic cells: Role of distinct lipid microdomains Mol.. .P-glycoprotein in intermediate-density microdomains of lysis buffer consisting of 0.5% (w ⁄ v) Brij-96 or 1% (w ⁄ v) Triton X-100 in 25 mm Tris ⁄ HCl, 140 mm NaCl, pH 7.5 Detergent concentrations were chosen based on the fact that Triton X-100 rafts are most often isolated using a 1% concentration, and the upper limit of the solubility of Brij-96 at 4 °C is 0.5% Each lysis buffer contained the... gradient using a Density Gradient Fractionator (Brandel, Gaithersburg, MD, USA) Immunoblot analysis was performed to confirm which fractions contained lipid raft microdomains Protein quantification The bicinchoninic acid protein assay [49] was performed on aliquots of the sucrose fractions, using BSA (crystallized and lyophilized; Sigma) as a standard SDS/PAGE and western immunoblot analysis Equal volumes... glycine for 2–5 min at room temperature The plasma membrane was permeabilized using 0.1% (v ⁄ v) Triton X-100 in NaCl ⁄ Pi for 10 min and the cells were then washed twice in NaCl ⁄ Pi Blocking was carried out overnight in 5% (w ⁄ v) skim milk in NaCl ⁄ Pi at 4 °C Cells were incubated with primary antibodies in NaCl ⁄ Pi for 1.5 h (1 : 100 dilution) and for 45 min with secondary antibodies in blocking... Peihua Lu and Joseph Chu for providing CHRB30 cells This work was supported by a Discovery Grant from the Natural Sciences and Engineering Research Council of Canada, and by a Research Grant from the Canadian Cancer Society References 1 Simons K & Ikonen E (1997) Functional rafts in cell membranes Nature 387, 569–572 2 Pike LJ (2003) Lipid rafts: bringing order to chaos J Lipid Res 44, 655–667 3 Brown DA... ATP-binding cassette transporters are enriched in noncaveolar detergent-insoluble glycosphingolipid-enriched membrane domains (DIGs) in human multidrug-resistant cancer cells J Biol Chem 279, 5734–5738 Bacso Z, Nagy H, Goda K, Bene L, Fenyvesi F, Matko J & Szabo G (2004) Raft and cytoskeleton associations of an ABC transporter: P-glycoprotein Cytometry 61A, 105–116 Radeva G & Sharom FJ (2004) Isolation... Cholesterol interaction with the daunorubicin binding site of P-glycoprotein Biochem Biophys Res Commun 276, 909–916 24 Troost J, Lindenmaier H, Haefeli WE & Weiss J (2004) Modulation of cellular cholesterol alters P-glycoprotein activity in multidrug-resistant cells Mol Pharmacol 66, 1332–1339 25 Modok S, Heyward C & Callaghan R (2004) P-glycoprotein retains function when reconstituted into a 4936... and enantiomeric cholesterol on P-glycoprotein localization and function in low-density membrane domains Biochemistry 39, 7651–7661 22 Rothnie A, Theron D, Soceneantu L, Martin C, Traikia M, Berridge G, Higgins CF, Devaux PF & Callaghan R (2001) The importance of cholesterol in maintenance of P-glycoprotein activity and its membrane perturbing in uence Eur Biophys J Biophys Lett 30, 430–442 23 Wang EJ, . P-Glycoprotein is localized in intermediate-density membrane microdomains distinct from classical lipid rafts and caveolar domains Galina Radeva,. mem- brane microdomains that are distinct from classical lipid rafts containing Src-family tyrosine kinases. The mem- brane microdomains containing Pgp displayed