Address correspondence to Dr Joan M. Boggs, Department of Structural Biology and Biochemistry, Hospital for Sick Children, 555 University Avenue, Toronto, Ontario, M5G 1X8, Canada. E-mail: email@example.com
Two different types of low-density detergent-insoluble glycosphingolipid-enriched membrane domain (DIG) fractions were isolated from myelin by extraction with Triton X-100 (TX-100) in 50 mM sodium phosphate buffer at room temperature (20°C) (procedure 1), in contrast to a single low-density fraction obtained by extraction with TX-100 in Tris buffer containing 150 mM NaCl and 5 mM EDTA at 4°C (procedure 2). Procedure 1 has been used in the past by others for myelin extraction to preserve the cytoskeleton and/or radial component of oligodendrocytes and myelin, whereas procedure 2 is now more commonly used to isolate myelin DIG fractions. The two DIG fractions obtained by procedure 1 gave opaque bands, B1 and B2, at somewhat lower and higher sucrose density respectively than myelin itself. The single DIG fraction obtained by procedure 2 gave a single opaque band at a similar sucrose density to B1. Both B1 and B2 had characteristics of lipid rafts, i.e. high galactosylceramide and cholesterol content and enrichment in GPI-linked 120-kDa neural cell adhesion molecule (NCAM)120, as found by others for the single low-density DIG fraction obtained by procedure 2. However, B2 had most of the myelin GM1 and more of the sulfatide than B1, and they differed significantly in their protein composition. B2 contained 41% of the actin, 100% of the tubulin, and most of the flotillin-1 and caveolin in myelin, whereas B1 contained more NCAM120 and other proteins than B2. The single low-density DIG fraction obtained by procedure 2 contained only low amounts of actin and tubulin. B1 and B2 also had size-isoform selectivity for some proteins, suggesting specific interactions and different functions of the two membrane domains. We propose that B1 may come from non-caveolar raft domains whereas B2 may derive from caveolin-containing raft domains associated with cytoskeletal proteins. Some kinases present were active on myelin basic protein suggesting that the DIGs may come from signaling domains.
Extensive studies have demonstrated the existence of lateral membrane domains that contain a specific repertoire of lipids and proteins. These lipid-rich microdomains, or rafts, in the plasma membrane are thought to be formed by the tight packing of the long and highly saturated fatty acids of the sphingolipids and cholesterol and contain GPI-linked and acylated proteins, and certain integral membrane proteins (Brown and Rose 1992; Harder and Simons 1997; Simons and Ikonen 1997; Melkonian et al. 1999; London and Brown 2000). Glycosphingolipid (GSL)/cholesterol-enriched membrane domains can be isolated on the basis of their resistance to either high pH or non-ionic detergents, such as Triton X-100 (TX-100), and their low density (Parton and Simons 1995).
The myelin sheath is a unique tissue as it contains a high lipid to protein ratio and is enriched in the GSLs, galactosylceramide (GalC) and its sulfated form cerebroside sulfate (CBS), and cholesterol (Norton 1977). It has a complex structure as it contains several substructures with different molecular compositions (compact myelin and cytosol-containing paranodal loops, outer loop and periaxonal inner loop). Compact myelin also contains cytoskeletal proteins (Wilson and Brophy 1987; Pereyra et al. 1988; Gillespie et al. 1989; Cabrera et al. 2000; Marta et al. 2002) and a series of tight junctions called radial component which passes through many layers of compact myelin (Karthigasan et al. 1994). The potential presence of different plasma membrane microdomains is an added level of complexity to the organization of myelin membranes.
The low density and high content of GSLs and cholesterol in myelin are similar to those of DIGs from other membranes. Nevertheless, it can be fractionated further into DIGs of even lower density and increased GSL/phospholipid ratio by detergent extraction. Early detergent solubilization studies of myelin and/or oligodendrocytes (OLs) reported the isolation of a detergent-insoluble cytoskeletal fraction and/or radial component containing actin, tubulin, GalC/CBS, cholesterol, exon II+ isoforms of myelin basic protein (MBP), 2′,3′-cyclic nucleotide 3′-phosphodiesterase (CNP), myelin/OL basic protein and the tight junction protein claudin-11 (formerly called OSP) (Pereyra et al. 1988; Gillespie et al. 1989; Wilson and Brophy 1989; Karthigasan et al. 1994; Morita et al. 1999; Yamamoto et al. 1999). However, this insoluble material was not fractionated further. In more recent detergent solubilization studies of myelin and/or OLs, low-density DIGs were isolated by sucrose density centrifugation (Kramer et al. 1997, 1999; Kim and Pfeiffer 1999; Simons et al. 2000; Taylor et al. 2002). The presence of actin and tubulin in DIGs isolated from myelin has not been reported, but those isolated from OLs contained some actin and tubulin (Klein et al. 2002; Marta et al. 2003).
Our efforts to define specialized membrane compartments and protein–protein complexes within the myelin sheath began with characterization of the TX-100-soluble fraction of bovine brain myelin (Arvanitis et al. 2002). Here we report the identification and characterization of two buoyant fractions of different density that resulted from fractionating the TX-100-insoluble residue obtained by the extraction method of (Karthigasan et al. 1994) by sucrose density gradient centrifugation. Both fractions, in accordance with their detergent insolubility, densities and lipid composition, had characteristics of GSL/cholesterol-enriched membrane domains, and contained distinctive protein residents. The higher-density DIGs contained most of the cytoskeletal elements but were enriched in GSLs and were of much lower density than cytoskeletal proteins. They also contained most of the caveolin, which we recently reported was present in myelin (Arvanitis et al. 2004).
Materials and methods
Enhanced chemiluminescence (ECL) reagents were purchased from Amersham Pharmacia (Oakville, ON, Canada). SeeBlue prestained molecular mass (Mr) standards, Silver Xpress Kit, 10, 14 and 8–16% precast sodium dodecyl sulfate (SDS) polyacrylamide gradient gels were purchased from Helixx Technologies (Toronto, ON, Canada). Nitrocellulose (0.2 mm) was purchased from Bio-Rad Laboratories (Mississauga, ON, Canada). TX-100, urea, antipain, leupeptin and other common laboratory reagents were from Sigma Chemical Company (Oakville, ON, Canada). [32P]ATP (3000 Ci/mmol) was from Perkin Elmer Life Sciences (Boston, MA, USA). Recombinant p42 mitogen-activated protein kinase (MAPK) (Erk2) was purchased from New England Biolabs (Beverley, MA, USA).
Antibodies used for immunoblotting
Affinity-isolated rabbit polyclonal antibody against a peptide (C11) from the C-terminus of actin, mouse monoclonal anti-CNP antibody, IgG1, mouse monoclonal anti-GSK-3 (IgG1) and biotin-conjugated cholera toxin B subunit were purchased from Sigma-Aldrich Canada, Ltd (Oakville, ON, Canada). Rat monoclonal anti-tubulin antibody, IgG2a (tissue culture supernatant) was purchased from Serotec International (Raleigh, NC, USA). Mouse monoclonal anti-myelin-associated glycoprotein (MAG) antibody, IgG1 fraction, and affinity-purified rabbit anti-neural cell adhesion molecule (NCAM) were purchased from Chemicon International Inc. (Temecula, CA, USA). Mouse monoclonal anti-fyn was purchased from LabVision Corp. (Fremont, CA, USA). Rabbit polyclonal anti-bovine MBP antibody (E5), IgG fraction, was a gift from Dr E. Day (Boggs et al. 1985) and was used as described previously (Arvanitis et al. 2002). Polyclonal antisera against proteolipid protein (PLP)/DM20 were raised in rabbits by immunization with lyophilized human PLP (a generous gift from Dr M. A. Moscarello, Hospital for Sick Children, Toronto, Canada) in complete Freund's adjuvant containing bovine serum albumin, and IgG fractions were prepared (Arvanitis et al. 2002). Rat monoclonal anti-CD44 IgG2b was purchased from Calbiochem (San Diego, CA, USA). Affinity-purified rabbit polyclonal anti-MAPK IgG, anti-phospho-p44,42-MAPK (Thr202,Tyr204), anti-MEK and anti-Akt were purchased from Cell Signaling Technology (Beverly, MA, USA). Mouse monoclonal anti-flotillin-1 was purchased from Santa Cruz Biotechnology, Inc. (Santa Cruz, CA, USA). Affinity-purified rabbit polyclonal anti-phosphoserine and mouse monoclonal anti-phosphotyrosine were purchased from Zymed Laboratories (San Francisco, CA, USA). Mouse monoclonal anti-MOG IgG (prepared by Dr C. Linington, University of Aberdeen, UK) was kindly provided by Dr S. E. Pfeiffer (University of Connecticut Medical School, Farmington, CT, USA). Protein A-horseradish peroxidase (HRP) conjugate was purchased from Bio-Rad Laboratories. Rabbit polyclonal anti-caveolin (serum) was purchased from BD Biosciences (Mississauga, ON, USA), and mouse monoclonal anti-estrogen receptor IgG2aκ was purchased from Upstate Biotechnology (Lake Placid, NY, USA). HRP-conjugated streptavidin and a control non-immune rabbit IgG were from Jackson ImmunoResearch Laboratories, Inc. (West Grove, PA, USA). Affinipure rat IgG (H + L) HRP conjugate and affinity-purified goat anti-mouse IgG (H + L) HRP conjugate with minimal cross-reactivity to rat serum proteins were purchased from BioCan Scientific (Mississauga, ON, Canada).
TX-100 extraction of purified myelin
Bovine brain myelin was isolated by the procedure of Norton and Poduslo (1973). It was extracted with TX-100 by two different procedures used by others for TX-100 extraction of myelin or OLs. Procedure 1 was used by us previously to yield detergent-soluble and -insoluble fractions (Arvanitis et al. 2002). Briefly, myelin membranes were detergent extracted at room temperature at a concentration of 1 mg myelin protein per mL in 1 mL 1% TX-100 in extraction buffer (50 mm sodium phosphate buffer, pH 7.4; Karthigasan et al. 1994), containing 1 mm each of the protease inhibitors leupeptin and antipain. Myelin membranes were extracted for 30 min while rotating on a lab quake shaker. The samples were centrifuged (12 000 g, 10 min, 4°C) to separate them into a detergent-insoluble pellet and detergent-soluble supernatant fraction as described by Karthigasan et al. (1994). The soluble fraction was removed and the pellet was extracted once more with 1 mL extraction buffer containing 1% TX-100. Supernatants from both extractions were pooled and stored at − 20°C. The pellets were used fresh or after storage at − 20°C.
Myelin was also extracted by a detergent extraction protocol used more recently by several other groups (procedure 2) (Simons et al. 2000; Taylor et al. 2002). Briefly, purified myelin (40 µg protein) was extracted at 4°C in 1 mL 1% TX-100 in 25 mm Tris buffer containing 150 mm NaCl, 5 mm EDTA and 1 mm phenylmethylsulfonyl fluoride at pH 7.4 (TNE buffer) for 30 min. Samples were centrifuged (139 000 g, 20 min, 4°C) in a Beckman SW-41 rotor using a Beckman Optima L-90K ultracentrifuge (Fullerton, CA, USA) to separate the detergent-insoluble fraction from the detergent-soluble fraction.
Sucrose density gradient centrifugation of TX-100-insoluble fraction
Continuous sucrose density gradients (5–60% or 5–40%) in 50 mm phosphate buffer or TNE buffer were prepared in 12-mL SW-41 ultraclear centrifuge tubes. The Triton-insoluble pellet was resuspended in 200 µL extraction buffer. The resuspended Triton-insoluble pellets from each preparation (400–600 µg protein) were layered over their respective gradients and centrifuged at 178 065 g for 18–24 h at 4°C in a Beckman SW-41 rotor. Control sucrose density gradients containing unfractionated myelin were run in parallel. Following ultracentrifugation, 12 fractions of 1.0 mL each were collected from the top (fraction 1) to the bottom of the gradient. Any material at the bottom of the tube was washed out and collected as fraction 13. These 1-mL fractions and the supernatant containing the Triton-soluble fraction were analyzed for their total protein content by the Peterson assay (Peterson 1977), and for the presence of actin, tubulin and flotillin-1 by western blotting, and GM1 by dot blotting. For further analysis, the two visible, opaque bands obtained by procedure 1, termed band 1 (B1) for the lower-density band and band 2 (B2) for the higher-density band, and the single opaque band obtained by procedure 2, were each collected as 1.5-mL fractions. B1 and B2 from procedure 1 were diluted with 50 mm phosphate buffer, pH 7.4, and the band obtained by procedure 2 was diluted with TNE. They were centrifuged at 178 065 g for 18 h at 4°C in a Beckman SW-41 rotor to wash away excess sucrose. The resulting band pellets, as well as the original total insoluble and soluble fractions, were analyzed for their protein composition by one- and two-dimensional polyacrylamide gel electrophoresis (PAGE) and immunoblotting, and for their lipid composition by TLC. B1 and B2 were examined by transmission electron microscopy.
In some cases, the Triton-insoluble pellet from procedure 1 was resuspended in 1 mL 40% sucrose containing 50 mm phosphate buffer, loaded in the bottom of the tube, and the sucrose density gradient was layered on top of it. In addition, B1 and B2 fractions obtained by layering on top of the gradient and centrifuging were diluted to a final sucrose concentration of 40%, loaded at the bottom of the gradient in separate tubes, and recentrifuged. The sucrose solutions for this purpose were prepared with or without 1% TX-100.
Transmission electron microscopy of B1 and B2
Aliquots (5 µL) of B1 and B2 were deposited on a formvar-coated 300-mesh copper grid, negatively stained with a 2% solution of uranyl acetate, and air dried. The samples were analyzed in a JEM 1230 transmission electron microscope (JEOL USA, Inc., Peabody, MA, USA) operated at 80 kV. Digital images were acquired with a CCD camera (AMT Advantage HR camera system; AMT, Danvers, MA, USA) attached to the microscope.
Protein analysis of sucrose density gradient sedimentation fractions
The washed pellets were resuspended in 1% SDS buffer (100 µL) for dot blots, or into SDS–PAGE sample buffer (100 µL) for western blots. An equal volume of each suspension, except where noted, was loaded on to the gel, and the proteins were transferred to nitrocellulose and then analyzed by immunoblotting using antibodies for specific proteins. Stained spots were detected using ECL, and the densities were quantified using a FluorochemTM 8000 image analyzer (Alpha Innotech Corporation, San Leandro, CA, USA) and compared with the unfractionated Triton-insoluble pellet and soluble fraction to determine relative quantities.
Western blotting was performed to determine the ratios of PLP, L-MAG, p44 MAPK and NCAM140 to their smaller isoforms DM20, S-MAG, p42 MAPK and NCAM120 respectively. For determination of the ratio of L-MAG to S-MAG, 1-mL aliquots from the band fractions were deglycosylated according to a standard protocol (Technical Guide; Boehringer Mannheim Biochemica, Indianapolis, IN, USA). The deglycosylated samples were concentrated to a final volume of 50 mL using the UVS400 Speed Vacuum Plus (Savant Instruments Inc., Holbrook, NY, USA) and mixed with an equal volume of SDS sample buffer. Equal volumes were assayed by SDS–PAGE on 14% gels followed by western blotting. Although CNP occurs as two size isoforms, we were not able to resolve them on either gradient gels (8–16%) or linear gels (14%).
In-gel kinase assay
The in-gel kinase assay of renaturable kinases reactive with MBP was carried out as described by Bhat and Zhang (1996) with some modifications. Briefly, the samples (16 µg protein equivalents) were resolved on a 12% SDS gel that was polymerized in the presence of MBP (0.4 mg/mL). After separating proteins by electrophoresis, SDS was washed off the gel by soaking in 250 mL 50 mm Tris-HCl (pH 8.0) (buffer A) containing 20% 2-propanol for 1 h. The gel was then soaked in 250 mL buffer A containing 5 mm dithiothreitol (DTT) for 1 h. The proteins were denatured by incubating the gel in buffer A containing 6 m guanidine-HCl and 5 mm DTT. All the steps above were carried out at room temperature with gentle shaking. Then proteins were renatured by incubation of the gel in buffer A containing 5 mm DTT and 0.04% Tween-20 for 16 h at 4°C, with at least three changes of the renaturation solution. After renaturation of proteins, the gel was preincubated in MAPK assay buffer (40 mm HEPES-HCl, pH 8.0) containing 1 mm DTT, 0.5 mm EGTA, 2 mm MnCl2 and 10 mm MgCl2) for 30 min at 30°C. The reaction was initiated by addition of [32P]ATP at a concentration of 50 µCi/mL to the kinase buffer and incubation was continued for 1 h at 30°C. The gel was washed extensively with 5% trichloroacetic acid containing 1% sodium pyrophosphate, dried under vacuum and exposed to X-ray film. Kinase activities were visualized as radioactive bands of phosphorylated MBP.
Lipid analysis of sucrose density gradient sedimentation fractions
For TLC analysis, the washed pellets were lyophilized and dissolved in 30 µL chloroform/methanol (2 : 1). Aliquots (25 µL of B1 and 15.0 µL of B2) were spotted on to Silica Gel 60-precoated TLC plates (20 cm × 20 cm), which were activated for 1 h at 90°C. Lipid mixtures were resolved in a solvent chamber containing chloroform/methanol/water at a ratio of 14 : 6 : 10. To observe lipid bands, the plate was sprayed with 50% H2SO4 and heated at 100°C for 10 min. Unfractionated myelin was run in parallel on each plate for comparison. The densities of stained lipid bands were quantified using a FluorochemTM 8000 image analyzer (Alpha Innotech Corporation) and the percentage of the total density of all bands was calculated for each band for comparison of relative amounts of each lipid in myelin, B1 and B2. GM1 was quantified on dot blots. Aliquots of all gradient fractions and the supernatant were loaded on to polyvinylidene difluoride membranes, incubated with biotin-conjugated cholera toxin B subunit (5 µg/mL) in 5% milk followed by incubation with 1 µg/mL HRP-conjugated streptavidin in 5% milk. The signal was detected by ECL.
Two-dimensional gel electrophoresis, two-dimensional western blotting and mass spectrometry
For two-dimensional PAGE, B1 and B2 were prepared for isoelectric focusing as described in Arvanitis et al. (2004). Briefly, the samples were resuspended in a solution containing 7 m urea and 2 m thiourea, pH 8, and dialyzed against the same buffer overnight at room temperature. They were adjusted to contain a final concentration of 7 m urea, 2 m thiourea, 100 mm DTT, 4% CHAPS and 2% IPG buffer (equilibration buffer). Isoelectric focusing was performed with immobilized pH gradients (linear pH gradient from 3 to 10) of length 7 cm (Amersham Biosciences, Baie d‘urfe, Quebec, Canada). The gels were rehydrated overnight with the samples in equilibration buffer and run according to standard protocols (Berkelman and Stenstedt 1998). After isoelectric focusing, the IPG gel strips were incubated twice for 15 min in 50 mm Tris-HCl, pH 8, containing 6 m urea, 30% glycerol and 2% SDS. DTT (125 mm) was included in the first incubation and 125 mm iodoacetamide was included in the second. The gels were then rinsed in deionized distilled H2O. In the second dimension, SDS–PAGE (on 14% acrylamide, 7 cm × 7 cm, 1-mm thick gels) was carried out using the methods of Laemmli et al. (1970). The gels were either silver stained and processed for mass spectrometry, or the resolved proteins were transferred on to nitrocellulose membrane using a semidry appartatus and used for western blotting with anti-phosphoserine and anti-phosphotyrosine antibodies. Samples used for western blotting with these antibodies were supplemented with 0.1 mm sodium orthovanadate just before addition of sample buffer. An IgG and HRP-conjugated protein G control was used on the membranes before they were probed for the phospho-containing proteins. The membranes were blocked with 2% gelatin (Sigma) to prevent phosphatase activity. The immunolabeled proteins were detected using HRP-conjugated protein G and ECL reagents. Selected silver-stained protein spots residing in the B2 fraction were excised for in-gel protein digestion with sequencing grade modified trypsin (Promega Corp., Madison, WI, USA) and analyzed by mass spectrometry at the Advanced Protein Technology Center, Hospital for Sick Children.
We also compare the results of extraction by procedure 1 with those obtained by extraction at 4°C with TX-100 in 25 mm Tris containing 150 mm NaCl and 5 mm EDTA (procedure 2), as used more recently by others for myelin extraction (Kramer et al. 1997, 1999; Kim and Pfeiffer 1999; Simons et al. 2000; Taylor et al. 2002). Extraction by procedure 1 solubilized significantly less protein than procedure 2, leaving 48% protein in the insoluble pellets compared with 24.5% for procedure 2 (Table 1). About half of the actin and all of the tubulin remained in the TX-100-insoluble fraction obtained by procedure 1 in agreement with Cabrera et al. (2000) and Gillespie et al. (1989). Following extraction by procedure 2, more actin and less tubulin were retained in the insoluble fraction (Table 1). Furthermore, about half of the PLP, DM20 and MAG, and two-thirds of the MBP, also remained in the insoluble pellet obtained by procedure 1 (Table 1) in agreement with Karthigasan et al. (1994). Using somewhat different buffers, Cabrera et al. (2000) and Pereyra et al. (1988) obtained similar results. This contrasts with the results of others using procedure 2, which resulted in solubilization of most of the PLP, MAG and MBP (Kramer et al. 1997; Taylor et al. 2002). Pereyra et al. (1988) found that the amount of MBP, PLP and MAG extracted increased with addition of more monovalent or divalent salt to the buffer.
Table 1. Myelin protein distribution in TX-100-insoluble pellet and low-density opaque bands from sucrose density gradient
Values are mean ± SD of three determinations from three different extractions. aPercentage of total protein and of each protein in the Triton pellets (P), and bands from sucrose density gradient centrifugation of the Triton pellet, relative to the amount in myelin. b100% of the myelin tubulin was found in the Triton-insoluble pellet, and in B2. cTotal of large PLP and small DM20 isoforms. dGlycosylated (glycos) MAG (total of all isoforms). eMAG could not be completely deglycosylated in myelin although it could in the TX-100 pellet. The percentage of L- and S- MAG in the pellet, B1 and B2, could not therefore be determined, so ratios are presented instead. A, Peterson assay; B, slot blot; C, western blot using 10 or 14% SDS-gels; D, ratios of L-MAG to S-MAG and large PLP to DM20 determined by western blotting. nd, Not detected. *p < 0.05, **p < 0.025, ***p < 0.005 versus B2 (two-tailed Student's t-test).
Half of the CNP and MAPK and most of the NCAM120 also remained in the insoluble fraction from procedure 1. We recently reported that CD44, MEK, caveolin-1 and the membrane form of the estrogen receptor (mER) were also present in myelin (Arvanitis et al. 2002, 2004). Most of the CD44, MEK, caveolin and about half of the mER were present in the TX-100 insoluble fraction (Table 1). The present study defines further the heterogeneity and composition of the TX-100-insoluble myelin residue following fractionation by sucrose density gradient sedimentation to separate insoluble cytoskeletal proteins from DIGs.
Sucrose density sedimentation analysis of insoluble fraction
The sedimentation profile of the Triton-insoluble residue that resulted from extraction by procedure 1 revealed two subfractions visible as discrete well separated opaque bands. A lower-density band, B1, was isolated in fractions 2 and 3 (at approximately 10–14% sucrose) and a higher-density band, B2, containing twice as much protein (Table 1) and lipid (as indicated by the density of bands on a TLC plate) was isolated in fractions 5 and 6 (at approximately 24–29% sucrose) (Fig. 1). There was little protein in other gradient fractions. Extraction by procedure 2 resulted in only one opaque band, located in fraction 3 (at approximately 14% sucrose) and more protein was dispersed throughout the gradient (Fig. 1). Note that soluble proteins, which sediment to a higher sucrose density, had already been removed from both extracts. Unfractionated myelin sediments as a single band mainly in fraction 4 (approximately 18–20% sucrose) (Fig. 1).
Electron microscopy of negatively stained samples showed that both B1 and B2 contained vesicles and other structures of various sizes including large aggregated clumps of vesicles. However, B1 consisted mostly of vesicles, including many small ones of size 200–300 nm and some larger ones (Fig. 2a), whereas B2 consisted mostly of what appeared to be small membrane fragments with net-like structures and ribbon or tube-like structures attached (Fig. 2b).
If the Triton-insoluble pellet from procedure 1 was layered at the bottom of the gradient and centrifuged in the absence of TX-100, only a single band was obtained at a similar density as B2. If B1 and B2, obtained by layering on top of the gradient and centrifuging, were diluted to a final sucrose concentration of 40%, layered at the bottom of the gradient in separate tubes and recentrifuged, B2 floated to the same density as before but B1 floated to a density just below that of B2. If B1 was mixed vigorously in 40% sucrose, a more diffuse band was observed at even higher densities. However, if TX-100 was included in the sucrose density gradient to prevent resealing of the vesicles, then B1 and B2 floated up to the same low densities as when the pellet was layered on top of the gradient. It therefore seems likely that the vesicles in B1 in the absence of TX-100 lost water or took up sucrose when diluted with 40% sucrose, thus preventing them from floating as high as they did in the presence of TX-100. For the insoluble fraction obtained from procedure 2, a single low-density band was obtained regardless of whether the sample was layered at the bottom or the top of the gradient.
Distribution of proteins in low-density bands
The protein and lipid composition of B1 and B2 was characterized further. All fractions obtained by both procedures were analyzed for their actin and tubulin content. The distribution of a number of other proteins in B1 and B2 was also ascertained by immunoblotting. All the proteins listed in Table 1 were found in both B1 and B2, except for tubulin, which was exclusively located in B2. B1 contained significantly more CNP, MBP, MAG, GPI-linked NCAM120 and MOG than B2, but little actin and no tubulin. In contrast, B2 contained most of the insoluble actin (41% of the total myelin actin) and all of the myelin tubulin. Very little actin and tubulin was found in the other gradient fractions. In addition, most of the flotillin-1 (Fig. 3a), caveolin, MEK, fyn (Fig. 3b) and insoluble mER was found in B2 (Table 1). MOG, PLP (total of large and small isoforms), CD44 and MAPK were distributed more equally between B1 and B2. Akt and GSK were detected in myelin and in the soluble fraction, but not in B1 or B2 (data not shown).
The single low-density band that resulted from procedure 2 extraction contained only 9.5% of the actin and 18.5% of the tubulin, much less than in B2 (Table 1 and Figs 4b and c). Most of the insoluble cytoskeletal proteins from procedure 2 extraction were found throughout the gradient (Figs 4b and c). Actin was found at high density indicating that it was polymerized, but most of the tubulin was found in fractions 1 and 2, not associated with the lipid–protein band and not at high density, indicating that it was monomeric.
Protein isoform distribution in B1 and B2
As PLP, MAG, MAPK and NCAM each occur as two or more size isoforms, we determined the distribution of their size isoforms in B1 and B2 by western blotting (Table 1). The mean ± SD ratio of the large PLP isoform to the small isoform, DM20, in myelin was 1.46 ± 0.17 (mean ± S.D.), whereas the ratio was much lower in the TX-100-insoluble fraction (Table 1). B2 contained more of the large PLP isoform in the insoluble fraction whereas DM20 was distributed more equally between B1 and B2. Thus, two-thirds of the total PLP in B1 was the small DM20 isoform. MAG is heterogeneously glycosylated and gives a broad band around Mr 100 000. After enzymatic deglycosylation of the Triton-insoluble pellet and B1 and B2 with N-glycosidase F, two polypeptides of Mr 72 000 and 67 000 were detected. B1 contained more L-MAG than S-MAG whereas B2 contained more S-MAG than L-MAG (Table 1 and Fig. 5). OL membranes contain some of the 140- and 180-kDa isoforms of NCAM (Kramer et al. 1997), but we detected only NCAM120 in the Triton-insoluble pellet (P), B1 and B2 fractions of myelin (data not shown).
The MAPK isoforms exhibited striking selective distribution between the two bands. The MAPK in B1 was mainly the p44 MAPK isoform, whereas MAPK in B2 was mainly p42 MAPK (Table 1 and Fig. 6a). Blotting with anti-phospho-MAPK, which recognizes the activated form of MAPK, doubly phosphorylated on Thr 202 and Tyr 204, showed that both forms of MAPK were phosphorylated in B2 (Fig. 6b); neither was detected for B1.
Kinase activity in B1 and B2
An in-gel kinase assay showed that a band in B2 of similar Mr to recombinant p42 MAPK was active on MBP (Fig. 6c). Weak activity of a similar band in B1 was detected on longer exposure (Fig. 6d). Two kinases of Mr similar to those of p42 and p44 MAPK were active in myelin (Fig. 6d). On prolonged exposure of an autoradiogram, a broad band encompassing 36–44 kDa was also detected in B2, which might correspond to the Mr values for p38, p42 and p44 MAPK (Fig. 6e). Several higher molecular weight kinases of Mr approximately 50, 64, 75 and 100–200 kDa in B2 were also active on MBP. The only other renaturable kinases active on MBP in B1 were of 100–110 kDa. This is the first report of active kinases of similar Mr to MAPK in myelin, although an earlier study by Bhat and Zhang (1996) detected activity of kinases with a higher Mr in myelin.
Two-dimensional PAGE of B1 and B2
A two-dimensional gel of B2 showed that it was enriched in a series of proteins at approximately 28 kDa (Fig. 7d, arrow), and another series at around 40 kDa that were less abundant in B1 (Fig. 7c). The series of spots of equal Mr but different pI suggests that those at 28 kDa may be result from post-translational modifications of the same protein, which affect its charge. Western blotting of the two-dimensional gel with antibodies directed against phosphoserine (Fig. 8a) and phosphotyrosine (Fig. 8b) revealed that these proteins were phosphorylated on serine and tyrosine residues. Those of higher pI have more phosphoserine whereas those of lower pI have more phosphotyrosine. An attempt was made to identify these protein spots using in-gel digests followed by mass spectrometry. Main peptide peaks were used to search online databases, but no significant matches were found. MS/MS data yielded a peptide sequence that did not match any protein in the database. A few additional spots of various molecular weights were seen in the anti-phosphotyrosine western blots. These other spots correspond to an Mr of approximately 17, 33–36, 48–55, 104 and 220 kDa. No other phosphoserine-containing proteins were present other than those at 28 kDa. Although MBP can be isolated from myelin by acid extraction as a phosphoprotein, MBP in the myelin used for this study was not phosphorylated owing to phosphatase activity (Arvanitis et al. 2002). The presence of some phosphorylated proteins in B2 from isolated myelin, which were not subject to phosphatase activity under the conditions used, further indicates that B2 is a signaling domain.
Distribution of myelin lipids
TLC showed that both B1 and B2 contained all of the major lipids found in myelin but they contained a little more cholesterol and GalC and less phospholipid (with the exception of PE) than myelin. Dot blots showed that B2 contained most (88%) of the GM1 in myelin whereas B1 contained only 8% (data not shown). Only traces of GM1 were found in other gradient fractions. The only other significant difference in lipid composition between B1 and B2 was relative depletion of CBS in B1 (Table 2 and Fig. 9). We reported earlier that the Triton-soluble fraction contained an increased CBS to GalC ratio compared with the insoluble fraction (Arvanitis et al. 2002). Thus B1 and B2 had higher ratios of the major myelin GSL, GalC, and cholesterol to phospholipid than myelin, whereas only B2 had a higher ratio of GM1 and the other major myelin GSL, CBS, to phospholipid.
Table 2. Distribution of myelin lipids in B1 and B2 compared with myelin
Values from three different extracts were averaged and the mean ± SD is shown. aDensity of spot on TLC plate as a percentage of total densities of all spots. bRatio of densities of cholesterol (chol), GalC and CBS to phospholipids (PL) other than PE. *p < 0.05, **p < 0.025, ***p < 0.005 versus myelin; †p < 0.05, ††p < 0.025 versus B2 (two-tailed Student's t-test).
This is the first report that two different DIG fractions can be isolated from myelin by TX-100 extraction with sodium phosphate buffer lacking EDTA/EGTA, at room temperature. Both of these fractions have characteristics of GSL/cholesterol-enriched membrane domains based on their TX-100 insolubility, high ratios of cholesterol and GalC to phospholipid, enrichment in the GPI-linked protein NCAM120, and buoyancy on a sucrose density gradient. However, B2 also contained the cytoskeletal proteins actin and tubulin, and the raft markers flotillin-1, caveolin and GM1, and thus may consist of caveolae or other types of GSL/cholesterol-enriched membrane domain associated with cytoskeletal proteins. It does not contain non-membrane-associated cytoskeleton, which would sediment faster to a much higher density.
Although B1 did not contain the raft marker GM1, it contained more of the raft marker NCAM120 than B2. (Schnitzer et al. 1995b) similarly found that caveolae from endothelial cells contained GM1 but were not rich in GPI-linked proteins, whereas non-caveolar DIGs were enriched in GPI-linked proteins but lacked GM1. Other cell types have also been found to have several different microdomains of different composition (Madore et al. 1999; Roper et al. 2000). He and Meiri (2002) observed two visible bands on a sucrose density gradient after centrifuging a detergent extract of growth cones. The lower-density fraction contained a population enriched in NCAM120 that lacked caveolin and the higher-density one contained a population that was enriched in caveolin but lacked NCAM120. Two different sphingolipid/cholesterol-enriched DIG fractions from melanoma cells could be separated by immunoprecipitation (Iwabuchi et al. 1998). One was enriched in GM3 and lacked caveolin whereas the other lacked GM3 but contained caveolin.
Our ability to detect two distinct DIG fractions, and the lower solubility of several myelin proteins in TX-100 in contrast to other studies on myelin using extraction procedure 2, may be due to the absence of EDTA and/or higher extraction temperature. Although raft solubility in detergents usually increases with temperature, several studies have shown that some types of rafts persist in Brij98 and TX-100 at 37°C (Drevot et al. 2002; Braccia et al. 2003). Extraction by procedure 1 allowed retention of all of the tubulin and half of the myelin actin in B2. Microtubules are usually unstable at 4°C (Pirollet et al. 1992). Although we do not know whether the actin and tubulin in myelin are still polymerized or form a network, similar extraction procedures have been shown by microscopy to cause retention of a cytoskeletal network in OLs (Wilson and Brophy 1989) and of intact radial component in myelin (Karthigasan et al. 1994). Use of buffer containing EDTA and extraction at a low temperature resulted in dispersal of the tubulin and actin throughout the gradient and less association of these cytoskeletal proteins with a lipid-protein fraction (Fig. 4). More tubulin, but less actin was also found in the supernatant after procedure 2 extraction. Intact cytoskeletal elements are also Triton insoluble and would sediment faster to a high density if not associated with lipid. Most of the actin, but no tubulin, was found at high density after procedure 2 extraction. However, most of the tubulin was found in fractions 1 and 2, not associated with the opaque lipid-protein band. This probably represented tubulin monomers, which would have a slow sedimentation rate, as a result of tubulin depolymerization or dissociation from other proteins in the DIGs while on the gradient or during manipulation of the TX-100-insoluble pellet. The tubulin must have remained associated with the DIGs after the extraction or it would have gone into the soluble phase.
Association of cytoskeletal proteins with DIGs could also be affected by dissociation of Ca2+-binding proteins. EGTA was found to cause dissociation of the Ca2+-binding proteins annexins II and VI from higher-density DIGs from lung membranes, resulting in the conversion of two low-density DIGs to one (Parkin et al. 1996). Gillespie et al. (1989) extracted myelin at room temperature but included EGTA in their buffer and found only a single broad band on sucrose density centrifugation at 30–38% sucrose, a considerably higher density than B2. Our observation of two bands after extraction of myelin with buffer lacking EDTA is consistent with the report of Olive et al. (1995) of two types of DIGs isolated from cerebellar tissue by a procedure similar to that of procedure 2 but using buffer lacking EDTA. However, they did not examine the cytoskeletal protein content of their fractions.
The question of whether myelin proteins such as PLP, MAG and MBP are constituents of DIGs is important. We found more of these proteins in TX-100 DIGs by procedure 1 than has been found by procedure 2 (Kramer et al. 1997, 1999; van der Haar et al. 1998; Kim and Pfeiffer 1999; Simons et al. 2000; Taylor et al. 2002). Their solubility depends on the type of detergent and the salt concentrations used for extraction. They are solubilized more by TX-100 if 300 mm KCl or 25 mm MgCl2 is added to the buffer (Pereyra et al. 1988; Arvanitis 2004). However, PLP interacts with CHAPS-insoluble DIGs from myelin and OLs, although not with the DIGs isolated from PLP-transfected BHK cells (Simons et al. 2000). MAG is located in low-density buoyant Lubrol-insoluble membrane fractions isolated from whole brain and primary OLs (Vinson et al. 2003). On the other hand, MOG from myelin is partially TX-100 insoluble by either procedure 1 or 2 (Taylor et al. 2002). Thus PLP, MAG and MBP may be more loosely associated with GSL/cholesterol-enriched domains than other myelin proteins such as MOG.
The differences in protein and GSL composition of B1 and B2 suggest distinct functions of these two myelin microdomains. The compartmentalization of particular protein and lipid components may be important for cell and membrane functions such as polarization and signal transduction. Of the proteins in the Triton-insoluble fraction, B1 contained most of the p44 MAPK, two-thirds of the MAG (predominantly L-MAG), NCAM120, MBP and most of the CNP, whereas B2 contained most of the p42 MAPK in addition to the MAPK activator MEK. The MAG present in B2 was predominantly S-MAG. Fyn has been found in the single type of DIGs isolated from myelin using procedure 2 (Kramer et al. 1997). We also detected fyn in B2 but not Akt or GSK.
In addition to selectivity for size isoforms of MAG and MAPK, B1 and B2 also showed selectivity for specific size isoforms of PLP; the DM20 isoform was the major PLP isoform present in B1 whereas a greater amount of the larger PLP isoform was present in B2. The two isoforms of PLP and of MAG differ in their cytosolic domains, and differences in their interactions with cytosolic and membrane proteins or negatively charged lipids, and differences in membrane trafficking have been reported (Boggs et al. 1977; Horvath et al. 1990; Minuk and Braun 1996; Kursula et al. 2001; Arvanitis et al. 2002; Gudz et al. 2002). Differences in association of these size isoforms with different membrane domains may also be due to post-translational modifications such as cysteine palmitoylation of PLP (Messier and Bizzozero 2000) and MAG (Pedraza et al. 1990) and differential phosphorylation of S- and L-MAG (Bambrick and Braun 1991; Jaramillo et al. 1994). Previous studies have demonstrated that DM20 cannot functionally replace PLP in compact myelin, demonstrating that the PLP-specific peptide confers critical properties for the long-term stability and normal function of myelin (Stecca et al. 2000). L-MAG and S-MAG also appear to have different functions in myelin (Schachner and Bartsch 2000).
The higher content of CNP, MOG and NCAM120 in B1 compared with B2 supports its identification as a GSL/cholesterol-enriched membrane domain. CNP and MOG were previously characterized as constituents of myelin DIGs by Kim and Pfeiffer (1999) and GPI-linked NCAM120 is enriched in myelin and OL DIGs isolated by procedure 2 (Kramer et al. 1997). CNP undergoes phosphorylation and acylation (De Angelis and Braun 1994; O'Neill and Braun 2000) and contains consensus sequences similar to those found in G proteins (Morell and Quarles 1999). These intrinsic properties suggest an involvement in signaling. NCAM120 and MAG are associated with fyn in DIGs from OLs and/or myelin (Umemori et al. 1994; Kramer et al. 1999) and so are probably also involved in signaling, as for NCAM120 in neurons (Niethammer et al. 2002). Although fyn could not be detected in B1, other kinases were present, which might be involved in signaling. MAG and MOG are transmembrane proteins localized in non-compact membrane regions of myelin, i.e. adaxonal myelin and the outer surface respectively. MOG may transmit extracellular information to the myelin sheath (reviewed in Johns and Bernard 1999) and MAG may be involved in myelin–axon communication (Quarles 2002). Therefore, the enrichment of GalC, cholesterol, CNP, L-MAG, p44 MAPK and NCAM120, and presence of MOG in B1 is consistent with non-caveolar membrane domains that may be involved in signaling events in the myelin sheath.
Unlike B1, the low-density fraction B2 resembles a membrane domain associated with a cytoskeletal network or radial component of myelin. The cytoskeleton can control membrane domain organization in some cell types (Foger et al. 2000; Holowka et al. 2000; Gomez-Mouton et al. 2001). It appears to participate in the redistribution of MOG to detergent-insoluble microdomains in OLs upon ligation with anti-MOG antibodies (Marta et al. 2003). However, little is known about the role of cytoskeletal proteins found in myelin, or whether they are part of or distinct from the radial component in myelin. The higher content of GM1 and CBS in B2 compared with B1 is also distinctive. The enrichment of CBS in B2 relative to B1 may be a result of electrostatic interactions with proteins or a specific sorting mechanism. PLP and CBS, but not GalC, are transported together to the OL plasma membrane (Brown et al. 1993).
The presence of some kinases in B1 and B2, and some phosphorylated proteins in B2, further suggests that both membrane domains could be involved in signaling. The fyn in adult myelin has not been found to be active, although it is active in myelin from young animals and in OLs (Kramer et al. 1999). However, we showed for the first time that a kinase of similar Mr to the p42 MAPK isoform in myelin was active on MBP. In a previous study of kinases in myelin, no p42 or p44 MAPK activity was detected, although they were active in OLs (Bhat and Zhang 1996). However, Bhat and Zhang (1996) showed that some kinases of higher Mr in myelin were active on MBP. In contrast to B1, which contained primarily the p44 isoform of MAPK, B2 contained almost exclusively the p42 isoform along with MEK. The p42 isoform was active in both B1 and B2. Other higher molecular weight kinases in B1 and B2 were also active on MBP. Although p44 MAPK predominated in B1, no activity on MBP was detected. Lower molecular weight kinase activity on MBP, which might be p38 MAPK, was also detected in B2. The p38 isoform has been detected in myelin sheaths in situ in the brain (Maruyama et al. 2000).
The p44 and p42 isoforms of MAPK are thought to be functionally redundant as they share 90% amino acid identity in their catalytic domain (Boulton et al. 1991). In several studies, however, it has been observed that a selective activation of p42 MAPK can occur (Bading and Greenberg 1991). The p42 form was found to predominate and be activated by PMA and receptor agonists in OL progenitors (Bhat and Zhang 1996; Larocca and Almazan 1997; Liu et al. 1999), whereas the p44 form predominated in more mature OLs and was also activated by PMA (Stariha and Kim 2001). These results have led to the suggestion that p42 MAPK may be important for mitogenesis of progenitor OLs whereas p44 MAPK may be important for differentiation of OLs (Stariha and Kim 2001). Their role in myelin is not known, but MBP, a MAPK substrate, is phosphorylated in vivo in myelin at the same site phosphorylated by MAPK in vitro (Martenson et al. 1983; Erickson et al. 1990). Microtubule-associated protein MAP1B is also a substrate for MAPK and has been found in myelin (Cabrera et al. 2000).
We previously demonstrated the presence of a mER and caveolin-1 in the myelin sheath (Arvanitis et al. 2004) and found that they were mostly in B2. The mER has been identified in numerous cells and shown to be responsible for rapid, non-genomic, signaling responses initiated at the plasma membrane of these cells (reviewed in Levin 2002). The mER appears to localize partially to caveolae (Kim et al. 1999), but the mechanisms by which this small pool of estrogen receptor translocates to this site are currently unknown. Although caveolin was detected in B2, we do not know if caveolae structures are present. Other prominent protein spots, including some phosphorylated proteins, were detected in the two-dimensional gel of B2, some of which became enriched in this fraction compared to myelin or the TX-100-insoluble extract.
Although myelin rafts have been of interest in recent years, much is still unknown about their molecular organization, formation and function. The isolation of two low-density DIGs from myelin indicates that the myelin membrane is heterogeneous, as found for plasma membranes from other cell types (Madore et al. 1999; Oliferenko et al. 1999; Roper et al. 2000; Nebl et al. 2002). Advancement towards a more detailed model of myelin organization should help us to understand the mechanisms that regulate its biological processes.
A studentship from the Multiple Sclerosis Society of Canada to DNA and a fellowship from the Multiple Sclerosis Society of Canada to YG are gratefully acknowledged. We thank Dr S. E. Pfeiffer, University of Connecticut, for the anti-MOG antibody.