Assembly of Glycoprotein-80 Adhesion Complexes in Dictyostelium RECEPTOR COMPARTMENTALIZATION AND OLIGOMERIZATION IN MEMBRANE RAFTS*

The phospholipid-anchored membrane glycoprotein (gp)-80 mediates cell-cell adhesion through a homophilic trans -interaction mechanism during Dictyostelium development and is enriched in a Triton X-100-insoluble floating fraction. To elucidate how gp80 adhesion complexes assemble in the plasma membrane, gp80-gp80 and gp80-raft interactions were investigated. A low density raft-like membrane fraction was isolated using a deter-gent-free method. It was enriched in sterols, the phos-pholipid-anchored proteins gp80, gp138, and ponticulin, as well as DdCD36 and actin, corresponding to components found in the Triton X-100-insoluble floating fraction. Chemical cross-linking revealed that gp80 oligomers were enriched in the raft-like membrane fraction, implicating stable oligomer-raft interactions. However, gp80 oligomers resisted sterol sequestration and were partially dissociated with Triton X-100, suggesting that compartmentalization in rafts was not solely responsible for their formation. The trans -dimer known to mediate adhesion was identified, but cis -oli-gomerization predominated and displayed greater accumulation during development. In fact, oligomerization was dependent on the level of gp80 expression and occurred among isolated gp80 extracellular domains, indicating that it

Integral membrane proteins and lipids have been predicted to typically display long-range random distributions, and maintenance of any nonrandom distributions would require specific mechanisms (1). In recent years, many nonrandom distributions have been discovered. For individual proteins, long-range lateral diffusion can be restricted by corrals established by the underlying cytoskeleton (2). The cytoskeleton can also tether and actively redistribute transmembrane proteins (3). Discrete membrane structures, such as cell-cell and cell-substratum adhesion complexes, also form through interactions between multiple membrane components (4). Modulation of the structure and function of these complexes may influence many cellular and developmental events.
Glycosylphosphatidylinositol (GPI) 1 -anchored proteins are unique, since they lack the direct cytoskeleton interactions that affect the distributions and complexes of their transmembrane counterparts. Nevertheless, they can be clustered into rafts by lipid interactions (5,6). Rafts are membrane microdomains that form from the close packing of sterols and saturated lipids into a liquid ordered structure. Model membranes with this structure are insoluble in cold Triton X-100. Triton X-100insoluble floating fractions (TIFF) have been isolated from many different cell types and contain the lipid components of rafts, plus GPI-anchored and other acylated proteins (7)(8)(9)(10). A growing number of GPI-anchored cell adhesion molecules have been discovered in TIFF (11)(12)(13)(14), suggesting that they are components of membrane rafts. This potential compartmentalization could possibly effect receptor interactions, cytoskeleton associations, and signaling during the assembly of GPI-anchored cell adhesion molecule complexes.
We have used Dictyostelium as a model system to examine the involvement of TIFF in cell-cell adhesion mediated by GPIanchored receptors (14). Dictyostelium has a simple and well defined life cycle that permits the biochemical analysis of dynamic processes over developmental time (15). During the aggregation stage of development, single cells aggregate chemotactically and undergo intercellular adhesion to form stable aggregates (16). Cell-cell contacts are formed first by the Ca 2ϩdependent adhesion molecule DdCAD-1 (17,18) and distinct Mg 2ϩ -dependent adhesion sites (19). Subsequently, two Ca 2ϩ / Mg 2ϩ -independent adhesion molecules, gp80 and gp150/LagC, are expressed (20 -22).
Gp80 is a phospholipid-anchored receptor (23,24) which mediates cell-cell adhesion through a trans-homophilic binding mechanism (25). Gp80 is required for strong adhesion that maintains multicellularity during development (26,27). However, gp80 adhesion complexes must also be dynamic structures, since cells are constantly breaking and re-making contacts as they migrate. What mechanisms underlie the assembly and disassembly of gp80 adhesion complexes? Like many vertebrate GPI-anchored adhesion molecules (11)(12)(13), gp80 is a main component of TIFF (14). Additionally, Triton X-100-insoluble contact regions (28) were found to be a cytoskeleton-associated form of TIFF and large domains containing TIFF proteins and lipids localize to gp80-mediated cell-cell contacts, thus implicating a role for TIFF in gp80mediated adhesion. In this paper, we have extended our studies to investigate the assembly and organization of gp80 molecules in raft-like membrane domains as gp80-mediated cell-cell contacts form during development. Low density plasma membrane fragments were isolated using a detergent-free method. Their similarity to TIFF confirms the existence of raft-like domains in the Dictyostelium plasma membrane. Gp80 was found to oligomerize preferentially in these domains by direct cis-interactions between their extracellular domains. These results suggest that raft-like domains may provide optimal microenvironments for the assembly of gp80 cis-oligomers and their subsequent incorporation into adhesion complexes.

EXPERIMENTAL PROCEDURES
Cell Growth and Development-Both the wild-type axenic strain AX2 and the csaA-null strain GT10 (26) of Dictyostelium discoideum, were cultured with Klebsiella aerogenes (29). For development, cells were collected, washed, and resuspended at 2 ϫ 10 7 cells/ml in 17 mM sodium phosphate buffer (pH 6.1), shaken at 180 rpm and pulsed with cAMP every 7 min at a final concentration of 2 ϫ 10 Ϫ8 M.
Isolation of TIFF and Low Density Plasma Membrane Fragments-TIFF was isolated as previously described (14). Low density plasma membrane fragments were isolated according to an established protocol (30). Cell pellets were frozen, thawed, resuspended at 1.5 ϫ 10 9 cells/ml in 1 mM ZnCl 2 , and homogenized using a Dounce homogenizer. Microscopic examination revealed minimal breakage of nuclei. Cell particulate was collected by centrifugation at 3,000 ϫ g and plasma membranes were isolated using the aqueous two-phase polymer system (31). After washing, isolated plasma membranes were resuspended in 8.5% (w/v) sucrose, 1 mM EDTA, 20 mM phosphate buffer (pH 7.6), and sonicated with 12 10-s pulses of a Sonifier cell disruptor (Branson, Danbury, CT) with a duty cycle of 40% and an output of 4, on ice. The suspension was mixed 1:2 with 60% (w/w) sucrose in 20 mM phosphate buffer (pH 7.6) for a final sucrose concentration of ϳ45% (w/w). This mixture was overlaid with either a continuous (30 -45%, w/w) or a discontinuous (20/38%, w/w) sucrose gradient in 20 mM sodium phosphate buffer (pH 7.6) and centrifuged at 120,000 ϫ g for 15-17 h at 2°C, using the Beckman SW40 rotor. One-ml fractions were collected from the top for further analyses.
Immunoprecipitation of Membranes-Equivalent amounts of isolated low density plasma membrane fragments were incubated with antibodies as previously described (14). Samples were incubated with 0.1 mg/ml anti-gp80 IgG or 0.1 mg/ml of an irrelevant mouse IgG in TBS plus 0.1% (w/v) bovine serum albumin. After washing, samples were incubated with goat anti-mouse antibodies conjugated with 10-nm gold (Sigma) at 1:10 dilution in TBS plus 0.1% bovine serum albumin. After washing, membrane pellets were resuspended in ϳ40 l of sodium phosphate buffer (pH 7.6) and then raised to ϳ57% (w/w) sucrose, overlaid with a continuous gradient of 34 -45% (w/w) sucrose and centrifuged at 120,000 ϫ g for 15-17 h at 2°C. One-ml fractions were collected from the top of the gradient and proteins were collected by chloroform-methanol precipitation.
Preparation of Gp80 Extracellular Domains-The extracellular domains of gp80 were isolated from culture medium using a reported method (23). Cells were developed in suspension for 10 h and then pelleted at 3,000 ϫ g. The supernatant was centrifuged at 100,000 ϫ g for 1 h at 4°C. The resulting S-100 supernatant was adjusted to pH 7.6 and then concentrated using the Centricon centrifugal filtration device (Millipore, Bedford, MA).
Chemical Cross-linking-Samples were cross-linked with DSP, DSS, or BS 3 (Pierce, Rockford, IL). In a typical experiment, samples of equal protein concentration were incubated with 0.1 mM cross-linker in 20 mM sodium phosphate buffer (pH 7.6) for 30 min at room temperature. Unreacted cross-linker was quenched by the addition of 1 M Tris-HCl (pH 7.6) to a final concentration of 100 mM for 10 min. Cells were cross-linked at 2.0 ϫ 10 7 cells/ml in 17 mM sodium phosphate buffer (pH 7.6) with rotation at 60 rpm.
Gel Electrophoresis-Protein concentrations were determined using the bicinchoninic acid protein assay kit (Pierce). Protein samples were separated by SDS-PAGE (32). For immunoblot analysis, bound antibodies were detected using horseradish peroxidase-conjugated secondary antibodies and the enhanced chemiluminescence kit (Amersham Bioscience, Inc., Piscataway, NJ). Immunoblots were imaged and quantified using the Fluor-S Max multiimager system (Bio-Rad, Mississauga, ON). Two-dimensional SDS-PAGE was performed by first separating samples in a nonreducing 6% acrylamide slab gel. Separation in the second dimension was carried out in a 10% slab gel. Before the second dimension, stacking gel buffer containing 10% ␤-mercaptoethanol was applied to the stacking gel for 15 min and then removed. The lane from the first gel was excised, placed on top of the stacking gel of the second gel, and encased in molten 1% (w/v) agarose, 2% ␤-mercaptoethanol in stacking gel buffer.
Protein Identification and Lipid Analysis-For protein identification by mass spectrometry, silver-stained gel bands were excised, macerated, reduced, alkylated, and then digested with trypsin as previously described (14). MALDI-TOF mass spectrometry was performed by the Mass Spectrometry Laboratory of the Molecular Medicine Research Center, University of Toronto. Monoisotopic peptide masses were submitted to the ProFound search engine (available on the world wide web) for matches. Unless otherwise stated, search parameters were held constant and included all Dictyostelium proteins in the NCBI nonredundant or SWISS-PROT data bases, tolerance for peptide mass error of 300 ppm, and no missed cut sites per peptide. For lipid analysis, membrane samples were subjected to gas-liquid chromatography as described previously (14).
Immunofluorescent Staining-Cells were collected from suspension cultures and deposited on poly-L-lysine-coated coverslips for 10 min. Membrane samples were deposited on coated coverslips for 30 -45 min. The samples were then fixed, stained, and mounted on slides as previously described (33). Samples were incubated with antibodies against gp80 followed by Alexa 488-conjugated secondary antibodies (Molecular Probes, Eugene, OR) at 1:300 dilution. Laser scanning confocal microscopy was performed using a Leica DM IRBE inverted microscope equipped with a Leica TCS SP confocal system. Detection was maintained within the range of the gray scale to prevent signal saturation. Capping of gp80 was carried out using the anti-gp80 mAb 80L5C4 in the absence of secondary antibody using an established protocol (14). Cell aggregates were dispersed and deposited on coverslips for 10 min. Subsequently, cells were incubated with the mAb for 30 min, and then washed and incubated until 55 min after the initial addition of the mAb. The cells were then fixed and prepared for confocal microscopy.

Isolation of Low Density Plasma Membrane Fragments Enriched in Gp80 -
The phospholipid-anchored cell adhesion molecule gp80 is enriched in TIFF (14). To demonstrate that TIFF originated from discrete raft-like domains in the plasma membrane, a detergent-free isolation method was developed based on the one used for caveolae isolation (30). Plasma membranes were purified, sonicated, and subjected to flotation centrifugation into a continuous sucrose gradient. Without prior sonication, the plasma membrane preparation formed a single prominent band at a density of 1.19 g/ml. The band was collected in fraction 6 (termed fraction 6 hereafter) (Fig. 1A), which contained gp80 and most of the membrane protein. Proteins were not detected in the lower density fractions. With prior sonication, the main band shifted to a higher density region at 1.20 g/ml and was collected in fraction 7 (designated fraction s7 hereafter) (Fig. 1B). Additionally, proteins of 139, 105, 82, 41, and 16 kDa floated into the lower density fractions with maximum levels detected in fraction 4 (termed fraction s4 hereafter). Fraction s4 had a density of 1.17 g/ml, indicating the association of these proteins with lipid-rich domains of the plasma membrane. Relative to total protein, gp80 was enriched by 3-5-fold in fraction s4 over fraction s7.
The association of gp80 with these fractions was also examined by confocal microscopy. Fixed membrane samples were stained for gp80. The nonsonicated membranes in fraction 6 displayed a heterogeneous gp80 distribution, with gp80-rich structures of Ͻ0.5 m in diameter embedded in the bulk of the large membrane fragments (Fig. 1C). Fraction s4 contained an abundance of gp80-stained small membrane fragments (Fig.  1D), which might represent the gp80-rich regions observed in fraction 6.
Protein and Lipid Components of the Low Density Plasma Membrane Fragments-Next, we analyzed the protein and lipid compositions of fraction s4. Fifteen bands were excised from a silver-stained gel and then analyzed by MALDI-TOF mass spectrometry. Eight produced relatively strong mass spectra, and six of them were identified as the Dictyostelium proteins gp138, gp138B, DdCD36, gp80, actin, and ponticulin (Fig. 1E). The identifications of gp80 and actin were confirmed by immunoblot analysis. Post-source decay analysis was used to confirm the other identifications, and 18/20, 9/10, 19/21, and 49/54 of derived peptide fragments matched (error Ͻ1 Da) expected fragments of peptides 326 -338 of gp138, 484 -495 of gp138B, 635-656 of DdCD36, and 66 -89 of ponticulin, respectively. All the proteins identified here have been found in TIFF (14).
Total lipids were extracted and quantified (Table I). Lipidto-protein ratios were highest for the low density plasma membranes fragments and lowest for fraction s7, consistent with their apparent densities in the sucrose gradient. The phospholipid and sterol contents were analyzed. Relative to fraction 6, the sterol-to-phospholipid ratio of fraction s7 was reduced by ϳ40%, whereas fraction s4 showed an increase of ϳ36%. A 2-fold higher sterol-to-phospholipid ratio was observed in fraction s4 relative to fraction s7. The protein composition of fraction s4 and its high sterol level indicated that the low density membrane fragments had raft-like properties.
Recovery of Raft-like Membrane Fragments from gp80-null Cells-To investigate the role of membrane rafts in the assem-FIG. 1. Isolation and characterization of gp80-rich low density plasma membrane fragments. Plasma membrane preparations were floated into 30 -45% (w/w) sucrose gradients and 1-ml fractions were collected from the top of the gradient. A, nonsonicated plasma membranes. Top, relative amounts of total membrane protein (ࡗ) and gp80 (Ⅺ) in gradient fractions. Sucrose densities were also determined for each fraction ("). Middle, silver-stained gel profiles for the sucrose gradient fractions. Molecular weight markers are indicated on the left. Bottom, immunoblots depicting gp80 distribution along the sucrose gradient. B, sonicated plasma membranes. Top, distribution of membrane protein (ࡗ) and gp80 (Ⅺ) in gradient fractions. In addition to the main peak, a low density peak of gp80 appeared (arrow). Middle, silver-stained gel profiles of the gradient fractions. Gp80 (82-kDa band) and several other protein species were present in the lower density fractions (arrowheads). Bottom, immunoblots of gp80. Membrane fractions were deposited on coverslips, fixed, stained for gp80, and imaged by confocal microscopy. C, fraction 6 derived from the nonsonicated sample; D, fraction s4 from the sonicated sample. Bars ϭ 5 m. E, identification of proteins in fraction s4 by MALDI-TOF mass spectrometry. Proteins were separated in a 10% gel and silver-stained. The identified proteins are indicated. For DdCD36, the match was made to a protein termed lysosomal integral membrane protein II that was found to have Ͼ99% sequence identity with the partial DdCD36 sequence in the data base. Footnote a, gp138 and gp138B are highly glycosylated and have apparent molecular mass of 138 kDa, but the search engine considered only their primary sequences of 83.02 and 81.62 kDa, respectively. By using mass windows of 93.3 Ϯ 25% and 82 kDa Ϯ 25% for the 155-and 139-kDa bands, respectively, significant Z scores resulted. Footnote b, for ponticulin, a significant Z score resulted after correction for the cleaved N-terminal signal sequence and allowance of pyrrolidone carboxylic acid modification of the resulting N-terminal glutamine residue.
bly of gp80 adhesion complexes, we first determined whether the raft-like membrane domains required gp80 for their formation. Plasma membranes were isolated from csaA-null cells where the gene encoding gp80 was inactivated (26). Membrane samples were subjected to flotation into a continuous sucrose gradient with or without prior sonication (Fig. 2). The protein profiles of the gradient fractions were similar to those obtained with wild-type cells (see Fig. 1). With sonication, the bulk of the membrane moved to fraction 7 from fraction 6 where it banded without sonication. Although gp80 was absent, proteins corresponding to the gp138 species, DdCD36, actin, and ponticulin floated into fractions 3 and 4 only after sonication. Total lipids were extracted and quantified from these membrane fractions. Similar to the samples isolated from wild-type cells, both the lipid-to-protein and sterol-to-phospholipid ratios were highest for the low density plasma membranes fragments, intermediate for fraction 6, and lowest for fraction s7 (Table I). These results thus indicate that raft-like domains exist in the absence of gp80.
Co-fractionation of the Low Density Membrane Components with Gp80 -Since the low density membrane fraction might contain several types of raft-like domains, we assessed the relationship between the major protein species in the gp80enriched raft-like structures using a variation of the co-immunoprecipitation approach. The low density membrane fragments were incubated with either anti-gp80 mAb alone or mAb followed by colloidal gold-conjugated goat anti-mouse antibodies. Samples were floated into a 34 -45% (w/w) sucrose gradient to separate membrane fragments that contained immune complexes of gp80 from fragments devoid of gp80. The binding of colloidal gold was expected to shift the gp80-rich fragments to a higher density region. Anti-gp80 antibodies alone shifted the gp138 species, DdCD36 and ponticulin together with gp80 to a denser region of the gradient. Incubation of membrane fragments with anti-gp80 mAb and colloidal gold-conjugated secondary antibodies shifted these bands further in a region between fractions 5 to 13 (Fig. 3). This wide spread probably resulted from the formation of large immune complexes. Bands corresponding to the gp138 species, DdCD36, gp80, and ponticulin displayed a similar distribution pattern in all cases. Bands between 45 and 70 kDa were detected mainly in fraction 13 and were likely due to bovine serum albumin and protein aggregates in the antibody solutions. Neither bovine serum albumin nor control mAb produced the density shifts. The co-shifting of the major protein species with gp80 is consistent with them being directly or indirectly associated with gp80 in the same membrane rafts.
Chemical Cross-linking Reveals Close Clustering of Gp80 Molecules-To investigate interactions between gp80 and other proteins in the rafts, membrane samples were cross-linked with the amine-reactive chemical cross-linkers DSS (spacer length of 1.14 nm) or DSP (spacer length of 1.2 nm), a cleavable analog of DSS. First, TIFF was cross-linked with DSS. When protein blots were probed with anti-gp80 mAb, gp80-positive species of ϳ150, ϳ160, and ϳ215 kDa were detected (Fig. 4A). The 150-and 160-kDa species might represent two dimer conformations and the ϳ215-kDa species might represent gp80 trimers. DSP produced gp80-positive species of ϳ155 and ϳ215 kDa (Fig. 4A). The putative dimer band was more diffuse and may represent multiple dimer conformations. To confirm that the cross-linked species were gp80 oligomers, DSP cross-linked samples were analyzed by two-dimensional SDS-PAGE (Fig.  4B). Silver staining revealed that putative cross-linked gp80 dimers, trimers, and higher molecular weight oligomers were indeed comprised of gp80 monomers. Cross-linking of gp80 with other TIFF proteins was not observed.
DSP cross-linking of the low density membrane fragments (fraction s4) also revealed the two gp80 dimer species in addition to trimers. To assess whether gp80 oligomerization occurred to a greater extent in fraction s4 relative to the bulk of the membrane, fractions 6 and s7 were cross-linked. With equal amounts of protein, fractions 6 and s4 produced similar levels of the gp80 oligomers, whereas fraction s7 produced less (Fig. 4C). Quantitative analysis showed that the ratio between the level of dimers and total gp80 was 3-fold higher in fractions s4 and 6 than in fraction s7 (Fig. 4D), indicating an enrichment of gp80 oligomers in the raft-like membrane fraction. Additionally, the lack of additional cross-linked species in the bulk plasma membrane samples suggested that the gp80 oligomers were segregated from most membrane proteins, consistent with their localization to raft-like domains.
Detection of Gp80 Oligomers on Live Cells-To demonstrate the presence of gp80 oligomers under normal physiological conditions, cell aggregates at 10 h of development were crosslinked. DSS produced gp80 species corresponding to the dimers and trimers (Fig. 5A). Cross-linking with BS 3 (spacer length of 1.14 nm), a membrane-impermeable DSS analog, produced the same cross-linked species, indicating that gp80 oligomers were present on the cell surface.  2. Isolation of raft-like membrane fragments from csaAnull cells. Plasma membranes preparations from 10-h csaA-null cells were floated into 30 -45% (w/w) sucrose gradients. One-ml fractions were collected from the top. Nonsonicated membranes formed a sharp band collected in fraction 6. Sonicated membranes formed a sharp band collected in fraction 7. Equal proportions of the fractions were separated in a 10% gel. Silver staining revealed protein species with molecular weights of 139,000, 105,000, 41,000, and 16,000, corresponding to the gp138 family members, DdCD36, actin, and ponticulin, respectively, that floated into fractions 3-6 (arrowheads).
The specificity of the chemical cross-linking reaction was assessed by incubating 10-h cells with 0.01-1.0 mM DSS (Fig.  5B). Over this concentration range, the proportion of mono-meric gp80 decreased substantially, with a corresponding increase in the level of the cross-linked species. From 0.03-1.0 mM DSS, large gp80 complexes were observed in the stacking FIG. 3. Co-immunoprecipitation of the low density membrane components with gp80. Low density membrane fragments were isolated after gradient centrifugation and then incubated with: A, control mouse IgG alone; B, control mouse IgG plus 10-nm colloidal goldconjugated goat anti-mouse antibodies; C, anti-gp80 mAb alone; or D, anti-gp80 mAb plus gold-conjugated secondary antibodies. Following these incubations, membrane pellets were resuspended and adjusted to ϳ57% (w/w) sucrose in 1 ml, and then floated into 34 -45% (w/w) sucrose gradients. One-ml fractions were collected from the top and proteins were separated by SDS-PAGE and the gels were subjected to silver staining. The 139-, 105-, 82-, and 16-kDa bands identified previously as the gp138 family members, DdCD36, gp80, and ponticulin coshifted with gp80 (arrowheads). The 53and 25-kDa bands corresponding to the heavy and light chains of anti-gp80 IgG molecules are marked with circles.

FIG. 4. Detection of gp80 oligomers in raft-like membrane fractions. A,
TIFF was prepared and cross-linked with 0.1 mM DSP or DSS for 30 min, separated in a 5% gel, blotted, and probed for gp80. Both reactions produced putative gp80 dimer (d) and trimer (t) bands. Two dimer species of 150 and 160 kDa were clearly resolved with DSP and designated d1 and d2, respectively. B, two-dimensional gel analysis of gp80 oligomers. DSP crosslinked TIFF was separated in the first dimension by electrophoresis in a nonreducing 6% gel, followed by second dimension electrophoresis in a reducing 10% gel, and then silver stained. In addition to monomers (m), dimers (d), and trimers (t), higher molecular weight oligomers (hmw) of gp80 were detected. C, samples of fraction 6 isolated without sonication (6) and fractions 4 and 7 isolated with sonication (s4 and s7) were prepared. Aliquots of equal protein amounts were cross-linked with 0.1 mM DSS. Species with molecular weights expected for gp80 monomers (m), dimers (d1 and d2), and trimers (t) were detected. D, the Western blots in C were quantified by dividing the intensity of both cross-linked dimers (a) by that of the noncross-linked monomers (b). Data represent the mean values of three different experiments.
gel. Quantitative analysis showed only a minor reduction in the amount of monomeric gp80 between 0.1 and 1.0 mM DSS. The small change in gp80 cross-linking in response to a 10-fold increase in cross-linker concentration suggested that nonspecific cross-linking was minimal. When cells were cross-linked with 0.1 mM DSS for 15-60 min, the proportion of total gp80 remaining as monomers was constant at 55-60% regardless of the length of the reaction time (Fig. 5C). The results suggest that the incorporation of additional gp80 monomers into oligomers was minimal over a 60-min period and that the gp80 oligomers present on live cells were relatively stable.
Preferential Association of Gp80 Oligomers with Detergentinsoluble Membranes-An increase in the clustering of GPIanchored proteins can be induced by cold nonionic detergents, possibly due to the coalescence of associated rafts (5). We investigated this effect by extracting cells with cold 0.2% Triton X-100. Unexpectedly, extraction of cells prior to cross-linking decreased the level of cross-linked gp80 species and increased the amount of monomers (Fig. 6A). These results indicate that gp80-gp80 interactions were sensitive to cold Triton X-100.
To determine whether the stabilized oligomers were associated with raft-like structures, the extracted and cross-linked sample was separated into Triton X-100-soluble and -insoluble fractions. The cross-linked dimers and trimers were preferentially associated with the Triton-insoluble fraction, whereas the supernatant contained predominantly gp80 monomers (Fig. 6A). When cells were cross-linked first and then extracted, the levels of dimers, trimers, and species trapped in the stacking gel were ϳ2.3-, ϳ3-, and 9-fold higher, respectively, in the Triton X-100-insoluble fraction versus the supernatant (Fig.  6B). In contrast, the gp80 monomers were 1.2-fold higher in the supernatant. Therefore, gp80 oligomerization apparently enhanced its partitioning into the raft-like domains.

Effects of Sterol Sequestration and Cleavage of the Phospholipid Anchor on Gp80
Oligomerization-To assess the relationship between gp80 oligomerization and raft localization, we performed sterol sequestration treatments that were previously shown to disperse clusters of other GPI-anchored proteins (34). Cell aggregates were treated with low levels of digitonin or filipin to sequester sterols and then cross-linked with DSS. Although sterol sequestration induced cell rounding, neither digitonin nor filipin affected gp80 cross-linking, whereas Triton X-100 partially dissociated the oligomers (Fig.  7, A-C). Immunostaining of these cells revealed discrete patches of gp80 at both cell-cell contacts and noncontact surfaces, further suggesting that gp80-gp80 interactions resisted sterol sequestration.
The above observation suggested that sterol-gp80 interactions might not be a major contributor to the stability of gp80 oligomers. To assess whether direct interactions between extracellular domains could mediate gp80 oligomerization in the absence of membrane association, we analyzed gp80 molecules released from the cell surface through cleavage of their phospholipid anchors by an endogenous phospholipase (35). Gp80 molecules released into the medium were isolated in an S-100 fraction, which represented ϳ15% of the cellular gp80. After concentration, proteins were cross-linked with either DSP or DSS. Dimer and trimer species of gp80 were detected in both cases (Fig. 7D). Direct interactions between the extracellular domains of gp80 were evidently sufficient for oligomerization.
Gp80 Oligomerization Is Mediated Predominantly by cis-Interactions-The gp80 homophilic binding site that mediates trans-interactions has been mapped (25) and is recognized by the mAb 80L5C4 (20). We investigated the contribution of this site to gp80 oligomerization by incubating 10-h cells with equivalent amounts of either gp80 Fab or control Fab, followed by cross-linking. Anti-gp80 Fab consistently reduced the upper dimer band to the background level (Fig. 8A). However, the other cross-linked species were unchanged, suggesting that the lower dimer (d1) and the other oligomers were maintained by cis-interactions, whereas the upper dimer (d2) was formed by trans-interactions.
To assess gp80 oligomerization on single cells, cells were dissociated with high shear forces in 10 mM EDTA. The crosslinked gp80 species persisted after the dissociation (Fig. 8B).
Even with the addition of latrunculin B, the lower dimer (d1) and other oligomers remained, whereas the upper dimer (d2) was substantially reduced.
The presence of such gp80 cis-oligomers should permit gp80 capping through bivalent mAb cross-linking alone. Indeed, exposure of live single cells to anti-gp80 mAb alone readily induced the redistribution of gp80 into caps (Fig. 8C). Gp80 displayed an even membrane distribution on fixed and untreated single cells (Fig. 8D). Together, these results indicate that gp80 oligomers occurred predominantly in cis.
Elevated Gp80 Expression Is Accompanied by Increases in Gp80 Oligomerization-Clustering of GPI-anchored proteins into rafts through lipid interactions with their anchors was found to be independent of protein expression (5,6). However, if oligomerization is mediated by direct interactions between the extracellular domains of gp80, it may show a dependence on the level of gp80 expression. To distinguish between these possibilities, we cross-linked cells with 0.1 mM DSS from 0 to 12 h of development, when gp80 expression increases from background to maximal levels (36). From 4 to 12 h, total gp80 expression increased ϳ10-fold. The cross-linked species also increased, whereas the level of gp80 monomers formed a plateau from 8 to 12 h (Fig. 9, A and B). Assuming that the difference between total gp80 and monomeric gp80 corresponded to the amount of gp80 oligomers, the percentage of gp80 oligomers increased linearly from 20 to 55% over the 10-fold range of gp80 expression between 4 and 12 h (Fig. 9C). From 6 to 12 h, the putative trans-dimers increased by ϳ1.8fold, but the putative cis-dimers and trimers each increased by ϳ3-fold (Fig. 9D). Similar results were obtained after crosslinking with 1 mM DSS (data not shown). The data indicate that direct interactions between gp80 molecules are important for gp80 oligomerization associated with cell-cell adhesion.

DISCUSSION
In this paper, we have isolated and analyzed low density plasma membrane fragments that display characteristics of membrane rafts. Additionally, the phospholipid-anchored cell adhesion molecule gp80 was found to form stable oligomers that partitioned preferentially into raft-like membranes. Our results reveal the coordination of multiple molecular interactions as gp80 assembles into adhesion complexes during Dictyostelium development. A model is proposed for this process (Fig. 10).
The existence of rafts in Dictyostelium plasma membrane is evident from the isolation of similar raft-like complexes by distinct methods. In this study, low density plasma membrane fragments, isolated with a detergent-free method, have been shown to share many properties with Dictyostelium TIFF (14). Both complexes contain high lipid contents and are especially enriched in sterols. Although they exclude most membrane protein species, they both contain gp80, gp138 family members, DdCD36, actin, and ponticulin. Gp138 and ponticulin were identified in TIFF after performing the molecular weight corrections described in the legend to Fig. 1. The main proteins in   FIG. 7. Detection of gp80 oligomers after sterol sequestration and cleavage of the gp80 phospholipid anchor. A, cells were collected at 10 h of development and treated with either Me 2 SO carrier, 0.001% (v/v) Triton X-100, 0.001% (w/v) digitonin, or 0.0004% (w/v) filipin for 20 min prior to cross-linking with 0.1 mM DSS. Proteins were separated in a 5% gel, blotted, and probed for gp80. B, quantification of the relative amounts of monomers after cross-linking (n ϭ 3). C, confocal images of treated cells. Following the treatments described in A, cells were deposited on coverslips, fixed, and stained for gp80. Bars ϭ 5 m. D, S-100 fractions containing gp80 extracellular domains were concentrated ϳ15-fold and probed for gp80 oligomers after cross-linking with 0.1 mM DSP or DSS. Monomer (m), dimer (d), and trimer (t) bands are indicated.
the membrane fragments are phospholipid-anchored, as such anchorage has been shown for gp80 (23,24) and ponticulin (37) and predicted for gp138 family members (38). The enrichment of sterols and phospholipid-anchored proteins in these membranes, together with their insolubility in cold Triton X-100, suggest that they have the liquid-ordered structure characteristic of rafts (7,8).
The raft-like domains likely play a key role in the assembly and function of the cell adhesion molecule gp80. However, raft-like fragments can be isolated from gp80-null cells, indi- cating that the formation of rafts is independent of gp80. During development, gp80 molecules are recruited into these membrane domains as depicted in step 2 of Fig. 10, since the other components of the rafts co-migrate with gp80 immune complexes in flotation gradients (Fig. 3).
Many GPI-anchored proteins are clustered through compartmentalization in raft microdomains alone (5,6). The same proportion of these proteins can be chemically cross-linked regardless of protein levels. Clustering of these raft proteins was abrogated by removal of the GPI-anchor or by sterol sequestration, but clustering was elevated by detergent extraction that coalesces rafts (5). However, gp80 behaves in a distinct manner. Gp80 oligomers are resistant to sterol sequestration by digitonin or filipin, while gp80-gp80 interactions are sensitive to Triton X-100 extraction, consistent with what we have observed in in vitro binding studies (25). These results suggest that while lipid interactions may be important in the recruitment of gp80 molecules into membrane rafts, additional forces stabilize the oligomers. Analysis of soluble gp80 shed from cells during development indicates that direct gp80-gp80 interactions are sufficient for oligomerization of gp80 molecules that lack the phospholipid anchor. Moreover, the increased oligomerization observed over development was likely due to elevated gp80 expression driving direct gp80-gp80 interactions.
Our chemical cross-linking studies detected relatively stable gp80 oligomers on aggregating cells during development. Cis-interactions between gp80 molecules are probably the major force stabilizing these oligomers as depicted in step 3 of our model. Gp80 is known to mediate adhesion through transdimerization at a single homophilic binding site that we have mapped (25). Exposure to antibodies that block this site diminishes one of the two detected gp80 dimers, thus identifying it as the gp80 trans-dimer. The resistance of the other oligomers to both this treatment and cell dissociation indicates that they can be maintained by cis-interactions alone. These cis-oligomers also display independence from the trans-dimer during development, as they accumulate to a much greater extent. Moreover, immunoelectron microscopy detected clusters of gp80 on the surface of single cells (39). Most clusters have diameters from 30 to 70 nm, values remarkably close to sizes ascribed to vertebrate rafts (6,40). Consistent with this observation, gp80 redistribution into caps can be induced by exposure of live single cells to mAb alone. Since capping of membrane proteins normally requires multivalent cross-linking (41), capping in response to a bivalent cross-linker implicates the prior existence of gp80 clusters.
The gp80 oligomers are partitioned preferentially within rafts. Relative to monomers, the oligomers are enriched in rafts isolated either with or without detergent. Moreover, ϳ55% of gp80 molecules form oligomers at a stage when 50 -55% of gp80 is resistant to Triton X-100 (14). Since gp80 oligomerization was not concomitant with its recruitment into rafts, the pref- 10. A model for the assembly of gp80 adhesion complexes. The schematic drawing depicts the gp80-raft (shaded box) and gp80-gp80 interactions leading to the assembly of adhesion complexes. 1) Membrane rafts enriched with sterols and phospholipid-anchored proteins exist prior to gp80 expression. 2) Gp80 molecules are recruited into rafts through interactions between their phospholipid anchor and the sterol-rich domains. 3) As the overall level of gp80 increases, the elevated concentration of gp80 molecules in rafts leads to local cisinteractions via their extracellular domains. 4) The affinity of gp80 for rafts is increased upon oligomerization. 5) gp80 cis-oligomers accumulate as development proceeds. 6) The presentation of cis-oligomers in rafts facilitates trans-dimerization required for cell-cell adhesion. The combination of trans-and cis-interactions promotes the cross-linking and coalescence of the adhering rafts. 7) Cytoskeleton association with the enlarged rafts contributes to the further expansion of the gp80 adhesion complexes. erential compartmentalization of oligomers within the domains may arise from a combination of secondary effects. The general ability of rafts to recruit certain molecules and exclude others (10) may facilitate direct gp80-gp80 interactions. During development, accumulation of monomeric gp80 in rafts could expedite local gp80-gp80 interactions with minimal interference from other membrane proteins. Once formed, the gp80 oligomers may be retained preferentially in the domains, since their multiple phospholipid anchors could strengthen their affinity for rafts (see step 4 in Fig. 10).
Gp80 oligomerization is likely central to the assembly of gp80 adhesion complexes, since oligomerization reaches maximal levels when gp80-mediated adhesion is at its peak (36). The cis-oligomers in rafts would be favorable precursors to the adhesion complexes. The oligomers could mediate avid transinteractions, and the associated rafts could recruit additional components while minimizing interference from other membrane proteins (see steps 5 and 6 in Fig. 10). Adhering rafts on apposing cells may coalesce into the extended sterol-rich and cytoskeleton-associated domains found at gp80-mediated contacts (14). The large gp80-positive complexes trapped in the stacking gels after chemical cross-linking may correspond to such assemblies. Consistent with a role for gp80-gp80 interactions during such enlargement, gp80 complexes have been found to coalesce within contacts between model membranes with the addition of greater amounts of gp80 alone (42). Gp80 trans-interactions may cross-link gp80 cis-oligomers and facilitate the rapid formation of adhesion complexes at contacts, as likely occurs when gp80 is capped by mAb alone.
On cells, cross-linking of GPI-anchored proteins can produce signals, such as tyrosine phosphorylation, and recruit the cytoskeleton (8,(43)(44)(45). Similarly, the formation of large gp80 assemblies at cell-cell contacts may stabilize the associated rafts and recruit the cytoskeleton (see step 7 of Fig. 10). The cytoskeleton could attach directly to the raft component ponticulin, which is the major high affinity link between the membrane and the actin cytoskeleton (46). Sterols may promote lipid-lipid interactions potentially linking gp80 to ponticulin. Comitin (14) and the 30-kDa actin bundling protein (47) are potential components of peripheral cytoskeleton complexes. Such interactions are currently under investigation.
It is increasingly evident that raft-like membrane domains can have many forms and functions (10). Since raft-like domains can be isolated in the absence of gp80, they may also contribute to cell-cell interactions mediated by the gp138 family members during sexual fusion of Dictyostelium cells (38) or to steps in phagocytosis that may involve DdCD36 (48). Also, different types of rafts have been detected on many cell types (49,50). In Dictyostelium, a CHAPS-insoluble floating fraction has been isolated. Similar to TIFF, this membrane fraction is sterol-rich but its main protein component is the cAMP receptor cAR1, whereas gp80 is present at background levels (51). In contrast, cAR1 is not a major component of either TIFF or the raft-like membrane fragments. Thus, gp80 and cAR1 are likely segregated into different membrane domains. This compartmentalization may enhance the coordination of cell-cell adhesion and chemotaxis during Dictyostelium aggregation.
Compartmentalization of adhesion molecules within particular domains of the plasma membrane likely influences multiple aspects of cell adhesion. During the initial stages of cell-cell contact formation, restriction of different adhesion molecules to distinct domains could optimize conditions for individual complex formation and thereby maximize overall adhesion. The mobilization of rafts to cell-cell contacts may exclude pre-existing adhesion molecules that are incompatible with the lipid environment. Such segregation of adhesion complexes may facilitate the reorganization of adhesion complexes and modulate cell-cell adhesion during physiological and developmental processes. For instance, domains of various receptors, including GPI-anchored LFA-3, exchange positions during formation of the immunological synapse (52). NCAM and cadherin have been found to segregate into distinct patches of single contact regions (53), and VE-cadherin and PECAM-1 exhibit distinct patterns within individual endothelial junctions (54). Also, gp80 replaces DdCAD-1 at cell-cell contacts during Dictyostelium aggregation (55), and GPI-anchored NCAM-120 likely excludes cadherins from cell-cell contacts during pancreatic development (56). Rafts may facilitate the segregation of adhesion receptors, thus allowing signaling and regulatory events to be differentiated among various membrane complexes.