The CD20 calcium channel is localized to microvilli and constitutively associated with membrane rafts: antibody binding increases the affinity of the association through an epitope-dependent cross-linking-independent mechanism.

CD20 is a B cell-specific membrane protein that functions in store-operated calcium entry and serves as a useful target for antibody-mediated therapeutic depletion of B cells. Antibody binding to CD20 induces a diversity of biological effects, some of which are dependent on lipid rafts. Rafts are isolated as low density detergent-resistant membranes, initially characterized using Triton X-100. We have previously reported that CD20 is soluble in 1% Triton but that antibodies induce the association of CD20 with Triton-resistant rafts. However, by using several other detergents to isolate rafts and by microscopic co-localization with a glycosylphosphatidylinositol-linked protein, we show in this report that CD20 is constitutively raft-associated. CD20 was distributed in a punctate pattern on the cell surface as visualized by fluorescence imaging and was also localized to microvilli by electron microscopy. The mechanism underlying antibody-induced association of CD20 with Triton-resistant rafts was investigated and found not to require cellular ATP, kinase activity, actin polymerization, or antibody cross-linking but was dependent on the epitope recognized. Thus, antibody-induced insolubility in 1% Triton most likely reflects a transition from relatively weak to strong raft association that occurs as a result of a conformational change in the CD20 protein.

CD20 is a B cell-specific tetraspan protein that assembles into oligomeric complexes and forms or regulates a store-operated calcium entry channel that is responsive to B cell receptor (BCR) 1 signaling (1). CD20 is also an effective target for in vivo depletion of malignant or autoimmune B cells using monoclonal antibodies (mAbs), which can activate apoptotic signaling pathways and mediate complement-mediated cytoxicity potentially through mechanisms involving cholesterol-and sphingolipid-rich membrane microdomains known as lipid rafts (2)(3)(4). Rafts are thought to function in part as platforms for signaling from those receptors with properties that allow their access to the tightly packed lipid raft environment, which otherwise excludes most membrane proteins (5)(6)(7). The operational criteria for assigning raft association of a protein are insolubility in nonionic detergents and buoyancy on density gradients. The detergent best characterized for raft isolation is Triton X-100, and we showed previously that antibodies induce translocation of CD20 from the soluble fraction of Triton X-100 cell lysates into the buoyant insoluble fraction, consistent with its induced association with rafts (8). However, raft association of some proteins can only be demonstrated using very low concentrations of Triton X-100 or other nonionic detergents (9 -11), and we recently found that unligated CD20, although soluble in 1% Triton X-100, was insoluble in 1% Brij 58 (1). Brij 58-insoluble CD20 localized to cholesterol-dependent, buoyant fractions on sucrose density gradients. Importantly, deletion of a short membrane-proximal cytoplasmic sequence, previously shown to be essential for efficient translocation into Tritonresistant rafts (12), also prevented the constitutive association of CD20 with Brij 58 buoyant fractions. CD20 thus appears to be an example of a raft-associated protein that is Triton-soluble. What, then, is the meaning of antibody-induced translocation into Triton-resistant rafts? Here, we first extend our observations on the constitutive nature of CD20-raft association and then explore the effects of antibody ligation on cell surface distribution and mechanisms of antibody-induced Triton-insolubility of CD20.
Our results show that antibody binding does not induce a detectable alteration in cell surface distribution of CD20, which in the absence of antibody binding is already distributed unevenly on the B cell surface. Transmission electron microscopy (EM) demonstrated that CD20 was localized to membrane protrusions, or microvilli. Anti-CD20 mAbs differ dramatically in their ability to induce Triton insolubility of CD20, and we show here that these differences cannot be attributed to the antibody isotype or to the amount of antibody bound per cell but are epitope-dependent. Translocation into Triton-resistant rafts is shown to be independent of the availability of ATP, kinase activity, actin polymerization, and the cross-linking effects of antibodies. Together, these data are consistent with the interpretation that CD20 is constitutively associated with lipid rafts on microvilli and that antibody engagement increases the affinity of the association through an intrinsic mechanism, such as a conformational change in the CD20 protein.

EXPERIMENTAL PROCEDURES
Cells-Ramos and BJAB B cells were maintained in culture in RPMI 1640 plus 7.5% fetal bovine serum. Peripheral blood mononuclear cells (PBMC) were isolated from whole blood using lymphocyte separation medium (ICN Biomedicals, Aurora, OH). Stable transfectants of BJAB cells expressing green fluorescence protein (GFP)-CD20 were generated by electroporation at 250 V and 500 microfarads (Gene Pulser II; Bio-Rad) with 30 g of DNA. GFP was fused to the amino terminus of CD20 by inserting human CD20 cDNA into the cloning site of the pEGFP expression vector (Clontech) using XhoI/SacII. CD20-positive cells were sorted by flow cytometry (FACStar cytometer; BD Biosciences) and maintained with Geneticin (Invitrogen) at 1 mg/ml.
Fab fragments of 2H7 were produced using the ImmunoPure Fab preparation kit (Pierce). Purity was assessed by SDS-PAGE separation of titrated 2H7 mAb, both intact and digested, followed by Western blot using biotin-conjugated anti-mouse IgG2b detected with avidin-horseradish peroxidase (both reagents from Southern Biotechnology Associates). The limit of detection of intact 2H7 mAb was 1 ng. The initial Fab preparation was contaminated with ϳ1 ng of intact mAb/g of protein; however, after two additional clearance steps with protein A-Sepharose, there was no detectable intact mAb in 5 g of Fab protein.
Triton X-100 and CHAPS were purchased from Pierce. Thesit, Brij 35, Brij 58, and Brij 96 were purchased from Sigma. Glutaraldehyde and osmium tetroxide were from Electron Microscopy Sciences (Fort Washington, PA) and paraformaldehyde (PFA) from BDH Chemicals.
Rafts Isolations and Immunoblots-Cell lysis in detergents (all used at 1%, except Triton X-100, which was also used at 0.05%) and sucrose density gradient centrifugation were done with slight modifications of the protocol previously described (8). Briefly, 10 8 cells were washed and then lysed for 15 min in ice-cold lysis buffer of detergent in MBS (25 mM MES, 150 mM NaCl) containing enzyme inhibitors (1 g/ml aprotinin, 1 g/ml leupeptin, 1 mM NaVO 4 , 1 mM NaMoO 4 , 1 mM phenylmethylsulfonyl fluoride, 1 mM EDTA). Lysates were mixed with an equal volume of 80% sucrose in MBS plus inhibitors, overlayered with 5 ml of 30% and 5 ml of 5% sucrose in MBS plus inhibitors, and then centrifuged at 235,000 ϫ g for 17 h. From the top of the gradients, eight 1.5ml fractions were collected on ice. Sample preparation for CD20 translocation experiments was as described (8). Briefly, after incubation with antibodies, the cells were lysed in 1% Triton X-100 and separated into soluble and insoluble fractions by centrifugation for 15 min at 14,000 ϫ g.
For GM1 dot blots, nitrocellulose membranes (Schleicher & Schuell) were spotted with equal cell equivalents from each gradient fraction. For immunoblots, equal cell equivalents from each gradient fraction were mixed with SDS sample buffer, heated to 95°C for 5 min, separated by SDS-PAGE under reducing conditions, and transferred to Immobilon P (Millipore Corp., Bedford, MA). All membranes were blocked with 5% bovine serum albumin. CTB-horseradish peroxidase (Sigma) was used to detect GM1. Other blots were performed as described (8). All blots were developed using SuperSignal chemiluminescent substrate (Pierce) and visualized using the Fluor-S Max (Bio-Rad) imaging system. Quantity One software was used to quantitate the signal.
Flow Cytometry-Cells were incubated with anti-CD20 mAb, followed by fluorescein isothiocyanate-conjugated anti-mouse IgG. The data were acquired using a Becton Dickinson FACScan (BD Biosciences) and analyzed using the FlowJo program (Tree Star, Inc., San Carlos, CA).
Immunofluorescence Imaging-Cells were fixed in 1% PFA at room temperature for 5 min and then incubated with antibody as indicated or incubated with antibody prior to fixation. Fixed and unfixed PBMC (10 6 ) were incubated with 1 g of 2H7-A488 in 100 l of PBS at 37 C for 15 min. Fixed Ramos B cells were incubated with B1-or 2H7-A488 (1 g/10 6 cells) for 30 min at 37°C; unfixed Ramos cells were incubated with the same antibodies for 10 min at 37°C, washed, and fixed. For colocalization studies, 10 6 BJAB (GFP-CD20) cells were fixed in 1% PFA; incubated with anti-CD59, anti-CD20, or anti-CD45; washed; and further incubated with anti-mouse IgG-Cy3.
Fluorescence imaging was done with a Leica DM RXA microscope attached to a 14-bit cooled CCD camera (Princeton Instruments, Monmouth Junction, NJ). Digital deconvolution was performed using the MicroTome software (VayTek, Fairfield, IA). In some experiments, im-aging of PBMC was with DeltaVision Image Restoration Microscopy System (Applied Precision, Issaquah, WA).
Electron Microscopy-For transmission EM, Ramos cells (1 ϫ 10 7 ) were incubated with mouse IgG anti-CD20 (15 g in 500 l of PBS), washed in PBS, and then incubated in colloidal gold (6 nm)-conjugated goat anti-mouse IgG (Electron Microscopy Sciences). After washing, the cells were fixed in cold 2.5% glutaraldehyde in PBS with rotation at 4°C for 1 h. After washing three times (20 min each) with the same buffer, the cells were pelleted and embedded in 3% agar. The agarized pellet was cut into 1-mm slices and immersed in 1% osmium tetroxide in PBS for 1 h at room temperature. The pellets were rinsed in distilled water, dehydrated in graded ethanol series, and embedded in Epon 812. Thin sections were cut with a diamond knife, mounted on naked copper grids, and viewed with a Hitachi (Schaumburg, IL) H-7000 transmission electron microscope.
For scanning EM, Ramos cells (1 ϫ 10 7 ) were fixed in cold 1% glutaraldehyde/PBS with rotation at 4°C for 1 h, washed three times in PBS, and dehydrated in graded ethanol series. The cell pellet was resuspended in 10 l of 100% ethanol and transferred to a poly-L-lysinecoated coverslip for critical point drying. Coverslips were then sputter-FIG. 2. Cholesterol dependence of CD20-raft association. A, BJAB cells were surface-labeled with biotin and incubated in the absence or presence of 2% MBC for 30 min before lysis in Brij 58. Lysates were fractionated by sucrose density gradient centrifugation, and fractions were analyzed by blotting with avidin-horseradish peroxidase. Similar data were obtained using Ramos cells. B, the percentage of biotinylated proteins in each fraction was estimated using Quantity One software. Data are the average Ϯ S.E. from three experiments. C, cells were incubated in the absence or presence of MBC. Samples after MBC treatment were followed by further incubation with 2% cholesterol-loaded MBC for 30 min, as indicated. The cells were then lysed in Brij 58 and fractionated on sucrose density gradients. Fractions were immunoblotted for CD20, G␣ i , and GM1 as indicated. Data shown are representative of two experiments. coated with gold palladium, dried, and mounted on the sample stage of a Philips XL30 scanning electron microscope (FEI Co., Hillsboro, OR).

CD20 Is Constitutively Associated with Membrane Rafts-To
test whether the use of detergents other than Triton X-100 might reveal a significant presence of CD20 in rafts in the absence of antibody engagement, Ramos B cells were lysed in the series of detergents indicated in Fig. 1. After sucrose density gradient centrifugation, fractions were tested by immunoblot for the presence of CD20. As controls, the gradient fractions were also blotted for actin, raft markers GM1 and G␣ i , and nonraft marker CD45. CD20 was solubilized by Triton X-100 at 1%, as expected. However, in 0.05% Triton and in all other detergents except Brij 96, 30 -80% of CD20 redistributed to the buoyant raft fractions 3-5. Glycosphinglipid GM1 localized to the raft fractions in all detergents. The acylated hetero- three-dimensional reconstructions of all deconvolved sections through the cells, as indicated. B, Ramos cells were fixed and then labeled with A488-conjugated Fab fragments of 2H7 or A488 -2H7 mAb as indicated. C, Ramos cells were fixed with 1% PFA, stained with B1-A488 or 2H7-A488 (left panels), or incubated at 37°C with the same antibodies and then fixed (right panels). D, BJAB cells stably transfected with GFP-CD20 were lysed in 1% Brij 58 and fractionated by sucrose density gradient centrifugation. Fractions were probed by anti-CD20 immunoblot. E, GFP-CD20-transfected BJAB cells were incubated at 37°C without antibody or with 2H7 or B1 mAbs as indicated. F, GFP-CD20 BJAB cells were incubated with anti-CD59 or anti-CD45 antibodies, and bound antibodies were visualized with Cy3-conjugated secondary antibody. Images separately acquired using GFP ( ex / em ϭ 470/525) and Cy3 ( ex / em ϭ 535/610) filter combinations were deconvolved and merged. Yellow indicates areas of overlapping distribution. meric G protein subunit G␣ i was found predominantly in the raft fractions in all detergents except Brij 96. The distribution of actin was similar in all gradients. CD45, an abundant nonraft protein, was not detected in the raft fractions in any detergent, indicating selectivity of the isolations.
Brij 58 and Brij 35 extracted the most CD20 into raft fractions. To further assess the distribution of plasma membraneassociated proteins in Brij 58 gradients, Ramos or BJAB B cells were surface-labeled with biotin before lysis, and fractions from the gradients were probed by Western blot with avidin-horseradish peroxidase. Results from BJAB cells are shown in Fig. 2; similar data were obtained with Ramos. Proteins found in raft fractions 3-5 appeared to include a distinct subset and represented ϳ30% of all surface-labeled proteins, as estimated by densitometry analysis (Fig. 2, A, left panel, and B). After pretreatment with methyl-␤-cyclodextrin (MBC), Ͻ13% of surfacelabeled proteins were detected in fractions 3-5 (Fig. 2, A, right  panel and B), demonstrating the cholesterol dependence expected of most protein-raft associations. The buoyancy of both CD20 and G␣ i on Brij 58 gradients was sensitive to cellular pretreatment with MBC and was restored by subsequent incubation with cholesterol-loaded MBC (Fig. 2C). Interestingly, the buoyancy of GM1 was less dependent on cholesterol. Quantitative analysis showed that the amount of GM1 in rafts (fractions 3-5) was reduced less than 20% by MBC treatment compared with more than 95 and 90% reduction for CD20 and G␣ i , respectively.
Cell Surface Distribution of CD20 and Colocalization with Raft Marker CD59 -Since constitutive raft association of CD20 was shown in the above experiments, we examined its cell surface distribution before and after ligation with mAb 2H7, which induces 95% of CD20 protein to become insoluble in 1% Triton X-100 (8,12). For these experiments, we used peripheral blood B cells (Fig. 3A), Ramos cells (Fig. 3, B and C), and BJAB cells transfected with a GFP-CD20 construct (Fig. 3, E and F), and results obtained with all three B cell sources were concordant. The pattern of CD20 staining on cells that were fixed prior to labeling was similar to that observed when cells were fixed after incubation with the 2H7 mAb (Fig. 3A). There was no apparent large scale change in the distribution of CD20 after antibody binding. CD20 was not evenly distributed around the membrane but was somewhat punctate. Similar data were obtained when cells were fixed with higher concentrations of PFA and/or in the presence of glutaraldehyde (data not shown) and also when CD20 was labeled with fluorochrome-conjugated highly purified Fab fragments of 2H7 to exclude cross-linking effects (Fig. 3B). Cell surface distribution of CD20 was not different when CD20 was ligated and labeled with B1 mAb, which induces minimal (Ͻ5%) Triton insolubility of CD20 (8, 15) (see below), compared with 2H7 (Fig. 3C). We expressed a GFP-CD20 construct in BJAB cells in order to observe the distribution of CD20 in the absence of antibody binding. The GFP-CD20 chimera distributed similarly to endogenous CD20 on sucrose density gradients prepared from Brij 58 lysates (Fig.  3D) and the distribution pattern of GFP-CD20 at the cell surface was similar whether the cells were untreated or pretreated with 2H7 or B1 mAb (Fig. 3E). To further assess the constitutive localization of CD20 in lipid rafts, we used fluorescence microscopy to examine the relative distribution of the glycosylphosphatidylinositol-linked raft protein, CD59. BJAB cells expressing GFP-CD20 were fixed and then labeled with anti-CD59 and Cy3-conjugated anti-mouse IgG. When the GFP signal was merged with the Cy3 signal, yellow color indicated areas of colocalization (Fig. 3F). In contrast, no colocalization with CD45 was evident.
CD20 Is Localized to Microvilli-In three-dimensional reconstructions of deconvolved optical sections such as those shown in fig. 3A, it appeared that CD20 staining was clustered on FIG. 5. Antibody-induced insolubility of CD20 in 1% Triton. A, untreated, MBC, or cholesterol/MBC-treated Ramos cells were incubated with isotype control or 2H7 mAb before lysis in 1% Triton X-100, fractionated on sucrose density gradients, and immunoblotted for CD20 or Lyn. B, Ramos cells were pretreated with MBC for the times indicated, and then 2H7 anti-CD20 was added for an additional 15 min before lysis in 1% Triton X-100. The soluble and insoluble fractions were collected and probed by immunoblot for CD20. Results are representative of three experiments in A and two experiments in B. For these and similar experiments shown in later figures, equal loading was demonstrated by actin blot or Coomassie Blue staining, but it is not shown for clarity of presentation. membrane protrusions. To examine this at higher resolution, Ramos B cells were labeled with CD20 mAb and gold-conjugated anti-mouse IgG and prepared for transmission EM. A grid was placed over each of 39 images derived from different cells in two experiments, and gold particles associated with membrane protrusions or with the flat continuous part of the plasma membrane were enumerated. Among a total of 656 particles counted, 591 (Ͼ90%) were associated with membrane protrusions. A representative image is shown in Fig. 4; low magnification of transmission and scanning EM images of Ramos B cells in the lower panels of Fig. 4 illustrate the gross plasma membrane architecture of these cells, which is similar to that reported previously for primary B cells (16 -19). No gold particles were observed in isotype control samples (data not shown).
Antibody-induced Insolubility in 1% Triton-Although unligated CD20 is soluble in 1% Triton, antibody binding can induce almost complete insolubility of CD20 in this detergent (8) (see Fig. 5A). Antibody-ligated Triton-insoluble CD20 was found in the low density region of sucrose gradients and not in the high density pellet (Fig. 5A, fraction P). The buoyancy of ligated Triton-insoluble CD20 was sensitive to cholesterol depletion and restored by subsequent incubation with cholesterol-loaded MBC (Fig. 5A). The distribution of the Src-family kinase Lyn on these gradients is shown for comparison. In light of the constitutive CD20-raft association demonstrated earlier, these data suggest that induced Triton insolubility reflects increased strength of the association with rafts. We then investigated the mechanism underlying this effect using a rapid and sensitive procedure, described previously (12), in which the loss of CD20 from the soluble fraction of Triton lysates caused by antibody binding is monitored as well as the corresponding gain of CD20 in the insoluble material obtained after microcentrifugation. To confirm the cholesterol dependence of CD20 insolubility using this procedure, Ramos B cells were pretreated with MBC before the addition of 2H7 anti-CD20 mAb. Both soluble and insoluble fractions were collected and probed for the presence of CD20 (Fig. 5B). As expected, antibodyinduced loss of CD20 from the soluble fractions (lane 5) was prevented by preincubation of the cells with MBC (lanes 6 -8).
The pellet fractions showed the reverse results (i.e. 2H7-induced CD20 Triton-insolubility was prevented by MBC pretreatment).
Antibody-induced association of CD20 with 1% Triton-resistant rafts could potentially be mediated by the actin cytoskeleton, by signaling events or post-translational modifications, by clustering on a scale that is not obvious at the level of light microscopy, or by a CD20-intrinsic mechanism such as a change in conformation. Cytoskeleton-associated proteins are among the most abundant proteins in membrane rafts (20). The membrane skeleton has also been suggested to regulate the size of lipid rafts (21). As shown in Fig. 6A, pretreating Ramos cells with 1 M cytochalasin D for up to 16 h did not affect CD20 Triton insolubility induced by 2H7, as compared with control cells. Similar results were obtained with 10 M cytochalasin D and with cytochalasin E (data not shown). Cytochalasin D at 1 M effectively disrupted the actin cytoskeleton as assessed by inhibition of antibody-induced BCR capping and cellular aggregation (data not shown), which require an intact cytoskeleton (22,23).
To test whether antibody-induced Triton resistance of CD20raft association is an active process that requires energy, we used sodium azide (NaN 3 ) and deoxyglucose to inhibit ATP generation before antibody addition. As shown in Fig. 6B, induced CD20 Triton insolubility occurred even at the highest concentrations used. The treatments effectively inhibited ATP generation as shown by the disappearance of the upper band, which represents the more heavily phosphorylated form of CD20 (24), and by inhibition of homotypic aggregation (data not shown).
Phosphorylation can mediate protein-protein and proteinlipid interactions. CD20 itself is phosphorylated at multiple sites by serine/threonine kinases (25) and activates tyrosine kinases after antibody cross-linking (26). Therefore, we tested whether inhibitors of serine/threonine kinases (staurosporine) and Src family kinases (PP1) have any effect on antibodyinduced Triton resistance of CD20-raft association. As shown in Fig. 6, C and D, neither inhibitor prevented the induction of CD20 Triton insolubility. Reduced presence of the upper CD20 band after exposure to staurosporine indicates that the treatment was effective (Fig. 6C). The effectiveness of the PP1 treatment was confirmed by its ability to inhibit BCR-stimulated tyrosine kinase-dependent calcium mobilization (data not shown). Together with the results in Fig. 6B, these data indi- FIG. 7. Induced CD20 Triton insolubility is epitope-dependent, but isotype-and cross-linking-independent. Ramos cells were incubated with B1, 2H7, or isotype switch variants of NK1 anti-CD20 mAbs (A); with B1, Bly1, 2H7, AT80, or Rituximab (B) that recognize different epitopes of CD20; or with 2H7 or Fab fragments of 2H7 (C) at different concentrations for 15 min at 37°C and then lysed in 1% Triton X-100. Soluble and insoluble fractions were obtained by microcentrifugation and probed for the presence of CD20 by immunoblot. Results are representative of three or more experiments. D and E, Ramos cells were labeled with anti-CD20 mAbs or Fab fragments of 2H7, followed by fluorescein isothiocyanate-conjugated secondary antibody. Fluorescence was acquired on a FACScan cytometer. Mean fluorescence intensity is indicated at the top right-hand corner of each panel.
cate that antibody-induced insolubility of CD20 in 1% Triton occurs independently of phosphorylation events.
Antibody-induced Triton Resistance of CD20-Raft Association Is Epitope-dependent and Cross-linking-independent-Anti-CD20 mAbs show marked differences in the amount of CD20 Triton resistance they induce (8,15) (see below). A potential cause of the variability could lie in isotype differences among CD20 mAbs. However, heavy chain isotype switch variants IgG1, IgG2a, and IgG2b of a single antibody specificity NK1-B20 (13) induced equivalent amounts of CD20 Triton insolubility, thus eliminating isotype differences as an underlying cause of the variability observed ( fig. 7A).
Recently, we demonstrated that B1 and 2H7 recognize distinct epitopes on CD20 and provided evidence strongly suggesting that the anti-CD20 mAb, Bly1, recognizes the same epitope as B1 (15). AT80 and Rituximab are two of several CD20 mAbs sharing similar fine specificity with 2H7. Here, we show that Bly1, like B1, induced minimal CD20 Triton insolubility, whereas AT80 and Rituximab, like 2H7, have major effects (Fig. 7B). The altered ratio of the upper/lower CD20 bands in lane 4 was not observed in all experiments. B1, Bly1, 2H7, and AT80 mAbs all bound strongly to B cells, as determined by indirect staining using fluorochrome-conjugated anti-mouse IgG and analysis by flow cytometry (Fig. 7D). Rituximab was not included in this analysis, because it is a chimeric mAb with human Fc regions and could not be compared directly with the other mAbs. These data strongly suggest that the inability of B1 and Bly1 mAbs to induce Triton resistance of CD20-raft association is an epitope-dependent effect.
The inability of divalent mAbs B1 and Bly1 to induce Triton resistance of CD20-raft association indicated that cross-linking per se was insufficient and raised the question of whether cross-linking was even necessary. To test this, highly purified Fab fragments of 2H7 were used to treat Ramos cells. We found that Fab-2H7 induced CD20 translocation to the detergentinsoluble fraction (Fig. 7C), although not to the same extent as intact antibody. The reduced level of CD20 Triton insolubility induced by monovalent Fab is consistent with reduced binding to CD20, relative to whole 2H7 mAb, as assessed by flow cytometry (Fig. 7E). Cross-linking is therefore neither sufficient (since B1 and Bly1 are ineffective) or required to induce Triton resistance of CD20-raft association. DISCUSSION Our conclusion that CD20 is constitutively localized to lipid rafts is supported by evidence from both biochemical and cellular studies. Unligated CD20 localized to low density detergent-resistant membranes isolated using Brij 35, Brij 58, Thesit, CHAPS, or a low concentration of Triton X-100. When isolating rafts using detergents that may be less stringent than 1% Triton, a significant concern is that nonraft membrane regions are inadequately solubilized. We controlled for this possibility using CD45, which is abundantly expressed on B cells, as it is on all leukocytes, and is not generally found in lipid rafts. The low density membranes isolated in our experiments did not include CD45. Localization of CD20 to rafts was cholesterol-dependent as expected and, as shown previously, prevented by deletion of 7 residues in a membrane proximal cytoplasmic region of CD20 (1,12). At the cell surface, unligated CD20 co-localized with rafts, as indicated by overlapping distribution of GFP-CD20 with a Cy3-labeled glycosylphosphatidylinositol-linked protein, CD59. Co-localization was not complete, however, suggesting that CD20 and CD59 may also occupy distinct rafts. Heterogeneity among rafts in B cells has been documented previously (14). In contrast, no co-localization between CD20 and CD45 was evident.
We compared the structures of the detergents used in this study with their ability to retain CD20, GM1, and G␣ i in low density raft fractions. Among the nonionic detergents, only Triton X-100 has a bulky p-isooctyl phenyl group (Table I), perhaps accounting for the ability of high concentrations of Triton to solubilize some raft proteins like CD20. Brij 96 behaved most similarly to Triton. Although Brij 96 has no phenyl group, there is a double bond in its oleyl chain not found in Brij 35, Brij 58, or Thesit. This and/or other distinctive features of Brij 96, might confer, like Triton, a greater ability to destabilize protein-raft associations. In contrast, the interactions of glycosphingolipid GM1 with other raft lipids were resistant to both Triton and Brij 96. Following a similar trend, the lipid-lipid raft interactions of G␣ i were resistant to Triton X-100 and to some extent also to Brij 96. This analysis suggests that CD20 is unlikely to associate with rafts via modification with long chain fatty acids. Although there are potential sites for palmitoylation at Cys 110 and Cys 220 , mutation of either or both sites did not prevent CD20 raft association (12) (data not shown). CHAPS is considered a stringent detergent in isolating raftassociated proteins (11), yet it retained most of CD20 in rafts. CHAPS is a steroid-based zwitterionic detergent frequently used to extract transmembrane proteins in their functional configuration, presumably by preserving protein-lipid interactions while disrupting protein-protein interactions. Thus, the resistance of CD20 to CHAPS may indicate that its constitutive raft localization is mediated by interactions with lipids rather than proteins. Consistent with this notion, the conformation of CD20 was recently shown to be sensitive to the level of membrane cholesterol, suggesting a potentially direct CD20-cholesterol interaction (27).
The restricted distribution of CD20 on microvilli is consistent with its co-localization with the BCR (14), which is also found on microvilli, presumably to facilitate interaction with antigen (28,29). In T cells, L-selectin also has microvillar localization and associates with rafts (30,31). Interestingly, the pentaspan microvilli-localized protein prominin is also associated with cholesterol-dependent lipid rafts that are soluble in Triton X-100 but insoluble in other nonionic detergents (32). In cells at the intestinal brush border, several transmembrane digestive enzymes and galectin 4 localize to Triton-sensitive rafts on microvilli (33,34). Thus, microvillar rafts may be a special subset with characteristics that attract proteins that associate with relatively weak affinity.
Antibody-induced effects on CD20-raft association are particularly important in the context of the therapeutic use of anti-CD20 antibodies (2). The ability of anti-CD20 antibodies to fix complement has been correlated with their ability to induce Triton-resistant raft association (4), and the activation of Src family kinases by CD20 cross-linking, which can lead to apoptotic cell death, is likely also a raft-dependent effect (2). Antibody cross-linking can stabilize weak protein-raft interactions or cause partitioning of proteins into rafts. In the case of CD20, cross-linking was shown here to be neither sufficient nor required to induce a transition to a high affinity raft-associated state resistant to Triton extraction. The antibody effect was epitope-dependent and inducible by highly purified Fab fragments. Together with our data excluding involvement of energydependent processes, these observations point to an intrinsic mechanism underlying a transition to high affinity raft association.
The epitope-dependent property of CD20 antibody engagement described here raises the question of possible physiological extracellular interactions that could mediate a similar effect. One consequence of such an interaction could be in-creased sensitivity of the store-operated calcium channel that is formed or regulated by CD20. Calcium influx in BCR-stimulated cells is significantly reduced by cholesterol depletion as well as by down-regulation of CD20, suggesting that the raft environment of CD20 may be important for channel function. Deletion of the sequence in CD20 that controls raft association also abolished calcium entry (1). Thus, increased strength of the association with rafts, mediated by a conformational change induced by an extracellular stimulus, could result in enhanced or prolonged calcium entry. Indeed, anti-CD20 mAbs have been shown to enhance calcium entry, with 1F5 performing better than B1 in this regard (35). The 1F5 mAb was not used in the current study but was shown previously to induce the association of CD20 with Triton-resistant rafts (8), suggesting a correlation between antibody-enhanced calcium entry and high affinity raft association.