High resolution mapping of mast cell membranes reveals primary and secondary domains of Fc(epsilon)RI and LAT.

In mast cells, cross-linking the high-affinity IgE receptor (Fc(epsilon)RI) initiates the Lyn-mediated phosphorylation of receptor ITAMs, forming phospho-ITAM binding sites for Syk. Previous immunogold labeling of membrane sheets showed that resting Fc(epsilon)RI colocalize loosely with Lyn, whereas cross-linked Fc(epsilon)RI redistribute into specialized domains (osmiophilic patches) that exclude Lyn, accumulate Syk, and are often bordered by coated pits. Here, the distribution of Fc(epsilon)RI beta is mapped relative to linker for activation of T cells (LAT), Grb2-binding protein 2 (Gab2), two PLCgamma isoforms, and the p85 subunit of phosphatidylinositol 3-kinase (PI3-kinase), all implicated in the remodeling of membrane inositol phospholipids. Before activation, PLCgamma1 and Gab2 are not strongly membrane associated, LAT occurs in small membrane clusters separate from receptor, and PLCgamma2, that coprecipitates with LAT, occurs in clusters and along cytoskeletal cables. After activation, PLCgamma2, Gab2, and a portion of p85 colocalize with Fc(epsilon)RI beta in osmiophilic patches. LAT clusters enlarge within 30 s of receptor activation, forming elongated complexes that can intersect osmiophilic patches without mixing. PLCgamma1 and another portion of p85 associate preferentially with activated LAT. Supporting multiple distributions of PI3-kinase, Fc(epsilon)RI cross-linking increases PI3-kinase activity in anti-LAT, anti-Fc(epsilon)RIbeta, and anti-Gab2 immune complexes. We propose that activated mast cells propagate signals from primary domains organized around Fc(epsilon)RIbeta and from secondary domains, including one organized around LAT.

Introduction

In 1972, Singer and Nicolson introduced the fluid mosaic model of membrane structure. Integrating a large body of work, this model proposed that cellular membranes are essentially two-dimensional solutions composed of integral membrane proteins within a lipid matrix. Although this model is still used to describe membrane organization locally, exceptions to the random arrangement of membrane constituents were immediately described, including evidence that leukocyte membrane proteins can segregate into or out of membrane that is internalized during phagocytosis (Tsan and Berlin, 1971; Oliver et al., 1974) and that antibodies and plant lectins can induce the movement of membrane proteins into patches and caps in fibroblasts, lymphocytes, and erythrocytes (for review see Berlin et al., 1974, 1975). Continued work by many investigators made it clear that membrane proteins rarely undergo random diffusion over long distances. Many factors, including cytoskeletal tethers and motors, interactions with both cytoplasmic and matrix proteins, and segregation with membrane lipids into somewhat ordered domains, contribute to their limited and/or directed movement (for review see Jacobson and Dietrich, 1999).

Current interest in organized regions of membrane, variously called microdomains, rafts, detergent-resistant membranes, and glycosylphosphatidylinositol-enriched membranes, centers around their potential roles in signal propagation and membrane trafficking (Edidin, 1997; Simons and Ikonen, 1997; Anderson, 1998; Brown and London, 1998; Jacobson and Dietrich, 1999). Typically, these membranes are isolated by detergent extraction and sucrose gradient centifugation, yielding a light fraction that accumulates acylated Src-family kinase members and is also enriched for glycerophosphatidylinositol-linked proteins, glycosphingolipids, gangliosides, and cholesterol. Detergent-resistant microdomains on leukocyte surfaces are particularly implicated in signaling via multi-chain immune recognition receptors, including the TCR, BCR, several Fcγ receptors, and the high-affinity IgE receptor (FcεRI)* of mast cells and basophils (for reviews see Horejsi et al., 1999; Langlet et al., 2000; Dráber et al., 2001).

The high-affinity IgE receptor of mast cells and basophils is an αβγ2 tetramer with immunoreceptor tyrosine-based activation motifs (ITAMs) in both the β and γ subunit cytoplasmic tails. Cross-linking this receptor activates the Src-family kinase, Lyn, initiating a cascade of events that include the activation of Syk, PLCγ and phosphatidylinositol 3-kinase (PI3-kinase), the mobilization of Ca2+, the secretion of inflammatory mediators from granules, the production of Th2 cytokines, and other events including membrane ruffling, spreading and increased adhesive activity. From detergent extraction and sucrose gradient centrifugation studies, Field et al. (1995)(1997, 1999) suggested that cross-linked FcεRI moves into detergent-resistant microdomains to encounter Lyn. Stauffer and Meyer (1997) localized a fluorescent Syk-SH2 domain with aggregated FcεRI in ganglioside-enriched membrane patches. These and other recent experiments using fluorescent reporters (Teruel and Meyer, 2000) have provided new insights into dynamics of signal transduction both temporally and spatially but are limited by the resolution of the light microscope.

Recently, we used immunogold labeling of mast cell membrane sheets and analysis by transmission electron microscopy to document the locations of signaling proteins at higher resolution during FcεRI signaling in rat basophilic leukemia cell line 2H3 (RBL-2H3) mast cells. We showed that the sequential association of FcεRI with Lyn and Syk occurs in topographically distinct microdomains (Wilson et al., 2000). In resting cells, FcεRI are distributed as dispersed small aggregates that are often loosely colocalized with small Lyn aggregates. After cross-linking, FcεRI redistribute to membrane domains that stain more intensely than bulk membrane with osmium. These osmiophilic membrane patches exclude Lyn and recruit Syk. Receptors are ultimately internalized through coated pits that bud from the periphery of the patches.

Here, plasma membrane sheets from RBL-2H3 mast cells are used to map the distribution of FcεRI β in relation to a further subset of proteins in the signaling cascade, including PLCγ1, PLCγ2, and PI3-kinase implicated in the remodeling of membrane inositol phospholipids, as well as the scaffolding/adaptor proteins, linker for activation of T cells (LAT) and Grb2-binding protein 2 (Gab2). Our results suggest that activated mast cells may propagate signals from primary signaling domains organized around FcεRI β and from secondary signaling domains, including one organized around the transmembrane protein, LAT.

Results

Phospholipase Cγ isoforms have distinct distributions in resting and activated mast cells

Previously, we showed that RBL-2H3 cells contain ∼10-fold more PLCγ2 than PLCγ1 and that there are substantial differences in the distribution of these PLCγ isoforms in RBL-2H3 cells. In particular, the results of immunoelectron microscopy in LR-White–embedded cells indicated that a portion of PLCγ2 is inherently associated with the plasma membrane, whereas PLCγ1 is recruited to membrane ruffles in response to FcεRI cross-linking (Barker et al., 1998).

These results are confirmed and extended here. As shown in Fig. 1 A, PLCγ2 interacts strongly with the membrane of resting cells. Nevertheless, very little of this enzyme associates with FcεRI β (Fig. 1 A, Table I). Instead, PLCγ2 occurs in resting cells either in small dispersed clusters or along submembranous cables (Fig. 1 A). The cables can be labeled with gold conjugates of antimyosin antibodies (Fig. 2, A and B) and phalloidin (Fig. 2, B and C), identifying them as components of the cortical actomyosin cytoskeleton.

Figure 1.

PLCγ2 redistributes to FcεRI-rich osmiophilic patches in activated mast cells. Membrane sheets were prepared from RBL-2H3 cells before (A) or after (B and C) cross-linking the FcεRI with DNP-BSA. Sheets were double- labeled with anti-FcεRI β monoclonal antibody conjugated to 10-nm anti–mouse gold particles and anti-PLCγ2 antibody conjugated to 5-nm anti–rabbit gold particles. (A) 10-nm gold particles marking FcεRI are distributed in small clusters and singlets (boxed regions) that do not mix substantially with small clusters of 5-nm gold particles marking PLCγ2 on the membrane (circled regions) and on cytoskeletal cables (arrowheads). (B and C) 5-nm gold particles marking PLCγ2 (arrows) colocalize with 10-nm gold particles specific for FcεRI in osmiophilic membrane patches (boxed regions). *Marks a coated vesicle budding from the patch in B. PLCγ2 label also persists along cytoskeletal cables in activated cells (arrowheads). Bar, 0.1 μm.

Table I.

PLCγ distribution on membrane sheets

	Resting	2′ XL	Wortmannin + 2′XL
Total PLCγ2 gold particles	2,182	2,215	—
Fold increase PLCγ2 gold particles	—	—	—
% colocalized with FcεRI β gold particles and in osmiophilic patches	6.7%	33.2%	—
Total PLCγ1 gold particles	363	712	514
Fold increase PLCγ1 particles	—	2×	1.4×
% colocalized with FcεRI β gold particles and in osmiophilic patches	6.6%	10.8%	8%

Membrane sheets were prepared from IgE-primed resting or activated (2 min with 0.1 mg/ml DNP-BSA) cells and labeled for PLCγ2 or PLCγ1 and FcεRI β; where indicated, cells were treated with 10 nM Wortmannin before 2-min activation. Numbers of gold particles marking the PLCγ isoforms and their proximity to particles marking FcεRI β were counted for matching areas of membrane for each data set. For PLCγ2, gold particles were counted for 30 μm2 of membrane; for PLCγ1, gold particles were counted for 97.5 μm2 of membrane.

Figure 2.

Actin and myosin are components of the cytoskeletal fibers that associate with membrane sheets, coprecipitate with PLCγ2, and contribute to PLCγ2 activation in suspension cells. Membrane sheets were prepared from untreated RBL-2H3 cells and labeled with 5-nm antimyosin gold particles (A and B) and either 15- (B) or 5-nm (C) avidin-gold conjugated to biotinylated phalloidin to label filamentous actin. Arrowheads (A and B) point to examples of myosin label, typically found at branchpoints in the membrane-associated cytoskeletal meshwork. Arrows (B and C) point to examples of gold particles for actin that decorates, but does not uniformly label, the cytoskeletal cables. Coated vesicles (B and C) are marked (*). Bars, 0.1 μm. (D) PLCγ2 and myosin immune complexes were prepared from pooled fractions of sucrose density gradients, separated by SDS-PAGE, and probed for actin or PLCγ2.

Supporting the microscopic evidence that a portion of PLCγ2 associates directly with the cytoskeleton, we performed coprecipitation studies using detergent-lysed RBL-2H3 cells. Light, intermediate, and heavy fractions from sucrose gradient centrifugation of the Triton X-100 extracts of resting and activated RBL-2H3 cells were used as the starting material for these studies. PLCγ2 and myosin were absent from the lightest fractions of the gradient but were both immunoprecipitated from the intermediate and heavy fractions (data not shown). Coprecipitating proteins in the immune complexes were resolved by SDS-PAGE and immunoblotting. Anti-PLCγ2 immunoprecipitates contained actin; antimyosin immunoprecipitates contained PLCγ2 (Fig. 2 D). These results support an association in both resting and activated cells between PLCγ2 and the cytoskeletal polymers, actin and myosin.

In contrast to PLCγ2, very little PLCγ1, the minor isoform of PLCγ in RBL-2H3 cells, was bound to the membrane of unstimulated mast cells (Fig. 3 A; Table I).

Figure 3.

PLCγ1 membrane label increases after FcεRI activation, mostly outside of osmiophilic patches. RBL-2H3 membrane sheets were labeled with 5-nm anti-PLCγ1 gold particles and with 10 nm anti-FcεRI β gold particles. In resting cells (A), there are very few 5-nm gold particles marking PLCγ1 (circled regions) and no FcεRI β-PLCγ1 colocalization. In stimulated cells (B), there is an increase in 5-nm gold particles marking PLCγ1 (circled regions), but very few of these particles colocalize with 10-nm gold marking FcεRI in osmiophilic patches (boxed regions). Bar, 0.1 μm.

Total numbers of gold particles recognizing PLCγ2 at the plasma membrane did not change after a 2-min activation (Table I), consistent with earlier data obtained by counting anti-PLCγ2 gold label in ultrathin sections prepared from LR-White–embedded cells and analyzed by standard TEM (Barker et al., 1998). Nevertheless, FcεRI cross-linking induced a strong association of PLCγ2 with receptors in osmiophilic patches (Table I; Fig. 1, B and C). Although direct observation (Fig. 1 B) and gold particle counting (data not shown) indicated no significant drop in PLCγ2 along the lengths of cytoskeletal elements after activation, we cannot rule out the possibility that some PLCγ2 recruited to receptor patches may have originated from the cytoskeleton-associated pool.

The distribution of PLCγ1 on membrane sheets from activated cells is shown in Fig. 3 B and quantified in Table I. FcεRI cross-linking doubles the number of PLCγ1-gold particles on membrane sheets. These gold particles occur as singlets and small clusters that are only occasionally seen near FcεRI β or in osmiophilic patches (Fig. 3 B). Wortmannin treatment, shown previously to inhibit PLCγ1 translocation and tyrosine phosphorylation (Barker et al., 1998), reduces the FcεRI-mediated recruitment of PLCγ1 to membrane sheets (Table I).

Distribution of PI3-kinase in activated cells

Maximal PLCγ activation after FcεRI cross-linking is dependent on both tyrosine phosphorylation and on the activation of PI3-kinase to form phosphatidylinositol-3,4,5-P3 (Ptd-Ins[3,4,5]P3), implicating D-3 phosphoinositides in determining both the location and catalytic activity of tyrosine phosphorylated PLCγ (Barker, et al., 1999; Smith et al., 2001). In Fig. 4 , polyclonal antibodies to the noncatalytic p85 subunit were used to localize the heterodimeric class IA PI3-kinases on membrane sheets. The sheets were colabeled with monoclonal antibodies to FcεRI β or to PLCγ isoforms. Fig. 4 A shows the scattered distribution of p85 label on membrane sheets prepared from resting cells. After 2 min of antigen stimulation, a substantial portion of p85 is strongly colocalized with FcεRI in osmiophilic patches (Fig. 4 B, boxed regions).

Figure 4.

PI3-kinase label colocalizes both with FcεRI in osmiophilic patches and with PLCγ1 outside of these patches in activated cells. Membrane sheets were prepared from RBL-2H3 cells before (A) or after (B and C) FcεRI cross-linking. (A and B) The sheets were labeled with 5-nm anti-p85 gold particles and with 10-nm anti-FcεRI β gold particles. Resting cells have relatively little membrane-bound PI3-kinase (A, circled region). 5-nm gold particles marking p85 are strikingly colocalized with FcεRI β within the osmiophilic patches (boxed regions) of activated cells (B). (C) Sheets were labeled with 5-nm anti-PLCγ1 gold and 10-nm anti-p85 gold. PLCγ1 and p85 colocalize in activated cells in mixed clusters outside of osmiophilic patches (circled regions). Bar, 0.1 μm.

The p85 subunit of PI3-kinase can be detected in FcεRI immunoprecipitates (see Fig. 10 A); anti-FcεRI β immune complexes from activated cells have ∼3× as much PI3-kinase activity as anti-FcεRI β immune complexes from resting cells (Fig. 5 A). The specificity of the assay was confirmed by use of the PI3-kinase inhibitor, Ly294002. Relative to the levels of activity in anti-p85 immunoprecipitates, this represents <5% of total PI3-kinase activity (Fig. 5 D). The results suggest that a small but significant fraction of p85 forms detergent-stable complexes with FcεRI β, which becomes activated after FcεRI phosphorylation and movement into osmiophilic patches.

Figure 10.

Biochemical analyses of protein–protein interactions in activated mast cells. (A) Coprecipitation studies demonstrate associations between signaling proteins. Resting and activated RBL-2H3 cells were solubilized in 0.5% Triton X-100, and the clarified cell lysates were used for immunoprecipitation afterwards by immunoblotting as indicated. Results show constitutive associations of p85 with LAT, Gab2, and FcεRI in adherent cells. They also show constitutive association of PLCγ2 with LAT and an inducible association of PLCγ1 with LAT. (B) Analysis of proteins distributed across sucrose density fractions. RBL-2H3 cells were solubilized in 0.05% Triton X-100 before (left) or after (right) cross-linking the FcεRI for 2 min with DNP-BSA. Cell lysates were loaded onto an 80% sucrose cushion, followed by layers of 35, 25, 20%, 15, and 10% sucrose. After ultracentrifugation, fractions were collected from the top (fraction 1) to the bottom of the gradient (from the lightest to heaviest fractions). Aliquots of the fractions were solubilized in Laemmli buffer, and proteins were resolved by SDS-PAGE, followed by immunoblot analysis as indicated in the key. Results show very little correlation with the protein–protein interactions observed by microscopy.

Figure 5.

Receptor cross-linking leads to increases in PI3-kinase activity associated with FcεRI, Gab2, and LAT. Immune complexes were prepared from detergent lysates of RBL-2H3 cells before or after 2 min activation with DNP-BSA. To measure associated PI3- kinase activity, immune complexes were incubated with PtdIns and γ[32P]ATP for 30 min at 37°C. Lipids in the reaction were extracted and resolved by TLC, and formation of PtdIns(3)P measured by PhosphorImager analysis. Background levels were established by adding PI3-kinase inhibitors (10 nM Wortmannin or 10 μM LY2940002) to replicate samples 15 min before initiating the reaction. As shown in A–C, there is an increase in PI3-kinase activity associated with FcεRI, Gab2, and LAT after 2 min of activation. Data are representative of at least three separate experiments, each performed in duplicate, and error bars represent SEM. Units are arbitary PhosphorImager values per 1,000. A one-sample t test was performed using Prism software (Graph Pad); all results have significant p-values (<0.03). (D) Shows the comparative levels of PI3-kinase activity in FcεRI, Gab2, and LAT precipitates, relative to total PI3-kinase activity in p85 immune complexes prepared from replicate lysates of antigen-stimulated cells. Data for Gab2 is from seven separate experiments (p-value = 0.00005); data for LAT (p-value = 0.0004) and FcεRI (p-value = 0.0288) are for three separate experiments.

Clusters of p85 also occur outside the osmiophilic patches (Fig. 4, B and C, circled regions). The micrograph in Fig. 4 C shows that some of the p85 clusters outside of osmiophilic patches colocalize with PLCγ1. Despite this clear association seen by microscopy, we were unable to measure PI3-kinase activity in PLCγ1 immune complexes prepared from detergent lysates of activated RBL-2H3 cells (data not shown). The most plausible explanation, supported by the biochemical studies reported in Barker et al. (1999) and Smith et al. (2001), is that the association seen by microscopy principally reflects the recruitment of PLCγ1-PH to newly synthesized PtdIns(3,4,5)IP3, rather than a direct detergent-stable interaction between the two proteins.

Distinct localizations for two adaptor proteins

Tyrosine kinase cascades typically involve the phosphorylation and/or translocation of a series of tyrosine-phosphorylated adaptor proteins that serve as platforms for macromolecular assembly of downstream components. We hypoth-esized that PLCγ1 may be selectively segregated away from osmiophilic patches on membrane sheets from activated cells by its interaction with a scaffolding or adaptor protein. We selected Gab2 and LAT as candidates for analysis. Both of these proteins have been implicated in the coupling of receptor activation to changes in inositol phospholipid metabolism.

Grb2-associated binder 2 (Gab2) is a 100-kd cytoplasmic protein with a documented role in growth factor receptor coupling to PI3-kinase (Nishida et al., 1999). Results in Fig. 6 A show that Gab2 gold particles are relatively sparse on membrane sheets from resting cells. Nevertheless, a fraction of these gold particles (∼18%) are in close proximity to FcεRI β (Table II). Gab2 gold label is increased 2.6-fold after FcεRI cross-linking. Furthermore, >50% of Gab2 is found in the FcεRI-, p85- and PLCγ2-containing osmiophilic patches that are induced by FcεRI cross-linking (Fig. 6 B; Table II).

Figure 6.

Gab2 colocalizes with FcεRI in signaling patches in activated mast cells. Membrane sheets were prepared from RBL-2H3 cells before (A) or after (B) cross-linking the FcεRI for 2 min with DNP-BSA. Membranes were labeled with 5-nm anti-Gab2 gold particles and with 10-nm anti-FcεRI β gold particles. (A) Shows that resting membranes have rather few 5-nm gold particles marking Gab2 (circled regions) and only occasional FcεRI β–Gab2 colocalization (boxed region). (B) Gold particles marking Gab2 (arrows) are strikingly colocalized with FcεRI β in osmiophilic patches of activated cells (boxed regions). Bar, 0.1 μm.

Table II.

Gab2 distribution on membrane sheets

	Resting	2′ XL
Total Gab2 particles	630	1663
Fold increase in Gab2 particles	—	2.6×
% colocalized with FcεRI β gold particles and in osmiophilic patches	18.5%	51%

Membrane sheets were prepared from IgE-primed resting or activated (2 min with 0.1 mg/ml DNP-BSA) cells and labeled for Gab2 and FcεRI β. Numbers of gold particles marking Gab2 and their proximity to particles marking FcεRI β were counted for 82.5 μm2 of membrane.

We found recently that adhesion of RBL-2H3 cells leads to association of PI3-kinase with Gab2 (unpublished data; see Fig. 10 A). Despite this stable association, complementary biochemical assays showed that FcεRI cross-linking markedly increases PI3-kinase activity in anti-Gab2 immune complexes (Fig. 5 C), representing ∼13% of total PI3-kinase activity (Fig. 5 D). These results suggest that p85 forms detergent-stable complexes with Gab2 that may localize active PI3-kinases to osmiophilic patches. Notably, because FcεRI immunoprecipitates do not also contain Gab2 (data not shown), it is likely that the association of PI3-kinase with multiple proteins reflects redundant mechanisms to activate PI3-kinase.

LAT is an integral membrane protein whose cytoplasmic tail is palmitoylated and contains multiple tyrosine phosphorylation sites. LAT has been proposed to act as a scaffold for downstream signaling components and was recently shown to be a critical element upstream of PLCγ activation and calcium responses in antigen-stimulated mast cells (Saitoh et al., 2000; Kimura et al., 2001).

Counts of >10,000 LAT gold particles showed no differences in the density of LAT between membrane sheets from resting and cells that have stimulated for 2 min with antigen (data not shown). However, the organization of LAT is dramatically altered after short periods of FcεRI cross-linking. In Fig. 7 A, gold labeling for LAT identifies some singlets and numerous small clusters across membrane sheets prepared from unstimulated RBL-2H3 cells. In the majority of membrane sheets from resting cells, the number of gold particles per cluster is <20 (Fig. 8, A and B) . Very few (<5%) of the 5-nm gold particles labeling LAT colocalize with 10-nm gold particles marking FcεRI β in unstimulated cells. Within 30 s of FcεRI cross-linking, very large elongated LAT clusters are found on membrane sheets (Fig. 7, B–E). Multiple groups of 50–150 gold particles were documented in cells activated for 30 s or 1 or 2 min (Fig. 8, B–E). LAT aggregates often transect the osmiophilic signaling patches that accumulate FcεRI β. However, rather than uniform mixing of the two sizes of gold label, they remain separate from each other. This is particularly evident in the large osmiophilic membrane patch seen in the upper left half of the micrograph in Fig. 7 B. Numerous LAT clusters are also found in apparently unspecialized membrane remote from aggregated FcεRI (Fig. 7, B and C).

Figure 7.

LAT rafts are larger after FcεRI cross-linking and can intersect osmiophilic patches. Membrane sheets prepared before (A) or after (B-E) cross-linking the FcεRI for 2 min with DNP-BSA were labeled with 5-nm gold particles for LAT and 10 nm gold particles for FcεRI β. (A) Shows that the numerous clusters of LAT (circled regions) and FcεRI (boxed regions) rarely colocalize in resting membranes. (B–E) Osmiophilic patches that label with FcεRI after 2 min activation are boxed and LAT clusters are outlined. LAT clusters in activated cells are often large (>50 particles). Bar, 0.1 μm.

Figure 8.

FcεRI cross-linking leads to rapid increases in LAT cluster size. (A) Summary data representing LAT cluster sizes in resting cells (hatched bars) or cells stimulated for 2 min with antigen (solid bars) for four independent experiments. A minimum of 3,000 gold particles were counted for each experimental condition. (B–E) Gold particles marking LAT were scored for cluster size in membrane sheets prepared from resting cells (B) or cells stimulated for 30 s (C), 1 min (D), or 2 min (E) with DNP-BSA. Counts are taken from one of two similar experiments.

Because PLCγ1 also redistributes on activated cells to membrane ruffles (Barker et al., 1998) and is dependent on D-3 phosphophoinositides for membrane recruitment, tyrosine phosphorylation and activation (Barker et al., 1998, 1999), it seemed likely that LAT might nucleate secondary signaling domains that include PLCγ1 and PI3-kinase. Analyses of membrane sheets double labeled for p85 and LAT support this hypothesis. Fig. 9 A shows the intersection, but not mixing, of large LAT clusters with p85 in a particularly dramatic osmiophilic patch after 2 min of antigen stimulation. There are also gold particles marking LAT away from the osmiophilic patches. These LAT clusters are larger than those on resting membranes and are strongly colocalized with p85 (Fig. 9 A, triangle regions).

Figure 9.

LAT topography in relation to the distributions of PI3-kinase and PLCγ1. Membrane sheets were prepared from antigen-stimulated RBL-2H3 cells and labeled with 5-nm anti-LAT gold particles (A–C) and with 10-nm gold particles specific for either the p85 regulatory subunit of PI3-kinase (A) or PLCγ1 (B and C). (A) Examples of LAT-PI3- kinase proximity are marked by triangles. Note that several large LAT rafts (circled regions) in an osmiophilic patch in the lower right do not mix extensively with nearby p85 clusters. (B and C) Examples of LAT-PLCγ1 colocalization are marked with triangles; separate LAT rafts are outlined. Note that colocalization is principally outside of osmiophilic patches. Bar, 0.1 μm.

Supporting the microscopy, anti-LAT immune complexes generated from RBL-2H3 cells coprecipitate the p85 subunit of PI3-kinase (Fig. 10 A). Although the amounts of p85 recovered in LAT precipitates is roughly equivalent in resting and antigen-stimulated cells, PI3-kinase activity is increased approximately threefold after FcεRI cross-linking (Fig. 5 B). Like FcεRI, the amount of PI3-kinase activity in LAT precipitates is a small but significant portion of the total activity measured in anti-p85 precipitates (Fig. 5 D). This may be an underestimate since coprecipitation of PI3-kinase with LAT is exceptionally sensitive to the detergent used during lysis. We found little PI3-kinase associated with LAT if immunoprecipitates were prepared from lysates solubilized in Brij 96 rather than Triton X-100 (not shown).

Importantly, PLCγ1 colocalizes with LAT after 2 min of FcεRI cross-linking. Gold particles marking PLCγ1 are frequently found at the edge of large LAT clusters (Fig. 9, B and C, arrowheads) and mixed with LAT in smaller clusters (Fig. 9, B–C, triangle regions). The inducible association of PLCγ1 with LAT inferred from microscopy of native membranes could also be demonstrated biochemically. Fig. 10 A shows a dramatic increase in PLCγ1 coprecipitating with anti-LAT immunoprecipitates from activated cells. Conversely, anti-PLCγ1 immunoprecipitates from activated but not resting cells contain coprecipitating LAT. Coprecipitation studies showed additionally that a portion of PLCγ2 also associates with LAT. However, the extent of PLCγ2-LAT coprecipitation was similar from both resting and activated cells, suggesting that PLCγ2 associates constitutively with LAT (Fig. 10 A). (TEM confirmation of the inherent association of PLCγ2 with LAT was not obtained because the available antibodies to both molecules are rabbit polyclonals and thus unsuitable for double label studies.)

Biochemical analyses of isolated lipid rafts do not reproduce the protein interactions observed directly on native membrane sheets

Previous investigators have inferred properties of mast cell FcεRI signaling complexes from the composition of detergent-resistant membranes isolated by detergent extraction and sucrose density gradient centrifugation analysis of activated mast cells (Field et al., 1995, 1997, 1999; Surviladze, 1998). The results in Fig. 10 B reproduce published experiments showing that Lyn and LAT are both in the light fractions containing detergent-resistant membrane. A portion of FcεRI β is also recovered in these light fractions, with a modest increase seen in activated cells. This experiment is popularly interpreted as showing the activation-induced recruitment of cross-linked FcεRI to Lyn microdomains at the onset of signaling. However, it is difficult to reconcile with studies in native membranes showing that cross-linked FcεRI in fact segregate rapidly away from Lyn on native membranes (Wilson et al., 2000) and that receptor and LAT show little colocalization in resting cells and only transient colocalization in activated cells (this study).

The discrepancies are compounded when other signaling molecules are included in the analysis. A portion of PLCγ1 is redistributed to the detergent-resistant membrane, as might be expected if this membrane represents regions of active signaling. However, Syk is found separated from receptor in the heavy fractions of the sucrose gradient (Fig. 10 B). In contrast, Syk accumulates with cross-linked FcεRI in the osmiophilic patches of native membranes (Wilson et al., 2000). Similarly, Gab2, PLCγ isoforms, and PI3-kinase are all separated from both receptor and LAT in the heavy fractions of the gradient (Fig. 10 B), even though they can be found in close proximity to receptors and LAT in native membranes. In short, the results of sucrose gradient centrifugation analysis of detergent-extracted mast cells give a strikingly different impression of protein–protein interactions involved in FcεRI signaling than the impression obtained by direct microscopy on native membranes.

Discussion

We have used immunoelectron microscopy on native membrane sheets to map the topography of PLCγ1, PLCγ2, and the p85 subunit of PI3-kinase, all implicated in the remodeling of mast cell membrane inositol phospholipids in response to cross-linking the high-affinity IgE receptor, FcεRI. We have also examined the topography of the 38-kD transmembrane adaptor protein, LAT. LAT is a major substrate of ZAP-70 and Syk (Weber et al., 1998; Zhang et al., 1998) that, in its phosphorylated form, mediates the SH2-dependent recruitment of several additional signaling molecules, including Grb2, Cbl, Vav, and PLCγ1. Finally, we have examined the topography of the adaptor protein, Gab2, during FcεRI signaling. Members of the Gab family of scaffolding proteins are tyrosine-phosphorylated in response to ligand binding to several growth factor and cytokine receptors (Holgado-Madruga et al., 1996; Gu et al., 1998; Nishida et al., 1999; Rodrigues et al., 2000; Schaeper et al., 2000; Gadina et al., 2000). We have recently found that Gab2 is the major tyrosine phosphorylated protein associated with class 1A PI3-kinases in antigen-stimulated adherent RBL-2H3 cells (unpublished data).

Previously, we demonstrated that cross-linked FcεRI encounter Syk in characteristic regions of the mast cell membrane that are easily identified by their dark staining with osmium and by the frequent formation of clathrin-coated pits at the periphery of the patch. Here, we show that the osmiophilic patches also accumulate PLCγ2, apparently reflecting a redistribution from its intrinsic associations in part with bulk membrane and in part with components of the cortical cytoskeleton. Osmiophilic patches also attract a substantial portion of class IA PI3-kinase heterodimers, marked here by antibodies to the p85 regulatory subunit, that are translocated from the cytosol during mast cell activation. Additionally, these membrane patches recruit Gab2 that couples to Shc, Grb2, and SHP-2, as well as PI3-kinase (unpublished data). The accumulation of these signaling molecules further establishes the identity of the osmiophilic patches as sites of signaling from cross-linked FcεRI to downstream responses.

We show that FcεRI cross-linking also causes the formation of large elongated rafts that label with antibodies to the scaffolding protein, LAT. Although the LAT rafts often intersect osmiophilic patches containing FcεRI, they do not mix extensively with receptors in the patches. We hypothesize that the small LAT rafts characteristic of resting cells redistribute after FcεRI cross-linking into the osmiophilic patches, where they encounter and are phosphorylated by Syk. Modification of LAT may then promote the formation of larger, mobile LAT rafts. Our micrographs suggest that these rafts, which become prominent features outside of osmiophilic patches, may form separate domains for the docking of other proteins. At least two proteins are candidates for this docking activity: PLCγ1, whose association with LAT rafts is induced by FcεRI cross-linking (Fig. 10 A), and a portion of PI3-kinase that is apparently distinct from that associated with FcεRI or Gab2, based on coprecipitation studies (Fig. 5 B).

Several studies have implicated LAT, together with the SH2 domain–containing cytoplasmic adaptors in the SLP-76 family, in the activation of both PLCγ isoforms (Jevremovic et al., 1999; Pasquet et al., 1999; Ishiai et al., 2000; for reviews see Pivniouk and Geha, 2000; Myung et al., 2000). A direct association of LAT and PLCγ proteins is implied by coprecipitation studies (Fukazawa et al., 1995; Finco et al., 1998; Tridandapani et al., 2000; Martelli et al., 2000; Zhang et al., 2000) and by the isolation of LAT on immobilized PLCγ SH2(C) domains (Gross et al., 1999). Our results suggest different roles for LAT in the localization, and likely the activation, of the two PLCγ isoforms in mast cells. In RBL-2H3 cells, FcεRI cross-linking promotes the association of PLCγ1 with LAT domains, whereas a significant fraction of PLCγ2 appears to interact constitutively with LAT. Both results are consistent with previous evidence by immunogold labeling in thin section TEM that PLCγ1 is recruited to antigen-stimulated RBL membranes, whereas PLCγ2 is inherently at the membrane (Barker et al., 1998). The present results raise the possibility that PLCγ2 is inherently at the membrane as a result of its constitutive interaction with LAT and can redistribute to osmiophilic patches after FcεRI cross-linking. In contrast, PLCγ1 is transiently at the membrane when levels of D3-phosphoinositides produced by PI3-kinase in LAT domains are locally high.

We show that the topography of LAT on membrane sheets is markedly different from another class of adaptor, Gab2. Membrane-associated Gab2 was shown here to increase >2.5-fold in response to FcεRI cross-linking. Most of this Gab2 is found in the osmiophilic patches. Gab2 translocation is accompanied by a severalfold increase in associated PI3-kinase activity, identifying sites of Gab2 accumulation as potential sites of D-3 phosphoinositide synthesis.

Why would cells form multiple signaling domains? Because cross-linked FcεRI are internalized relatively rapidly by endocytosis through coated pits, whereas LAT is not internalized through coated pits, one prediction is that signaling domains organized by LAT may be more stable than signaling domains organized by FcεRI. In this case, one role for LAT domains could be to sustain and amplify signaling as receptor levels drop. Interestingly, although secretion from RBL-2H3 cells halts within seconds of the addition of monovalent haptens that disrupt FcεRI aggregates, calcium levels remain high for several minutes (Lee and Oliver, 1995). This could be explained if LAT continues to serve as a scaffold for propagating signals to PI3-kinase and PLCγ proteins for some time after the disruption of FcεRI signaling patches. Another prediction is that distinct arms of the FcεRI signaling cascade could be propagated in these separate domains. These distinct arms may be complementary or, alternatively, may have independent outcomes.

Importantly, the p85 regulatory subunit of PI3-kinase can be found in both FcεRI and LAT domains. D-3 phosphoinositides represent a small and transient fraction of total membrane phosphoinositides (Traynor-Kaplan et al., 1989). Thus, the recruitment of PI3-kinase to specific domains may serve to generate the locally high concentrations of Ptd-Ins(3,4,5)P3 required for full activation of both PLCγ isoforms (Barker et al., 1999; Smith et al., 2001) at their distinct membrane locales (Barker et al., 1998). If this is true, then location is likely to be a factor in the activation of other enzymes, such as Akt, that also require PtdIns(3,4,5)P3 or its metabolite PtdIns(3,4)P2 for activation.

Many questions remain regarding the topography and functions of the class IA PI3-kinases. We showed previously that tyrosine phosphorylated proteins coprecipitate with all three p110 PI3-kinase catalytic subunits and that microinjection of blocking antibodies to the p110δ and p110β isoforms, but not p110α, inhibits calcium responses in RBL-2H3 cells (Smith et al., 2001). Here, we have observed that FcεRI cross-linking leads to increases in PI3-kinase activity in distinct macromolecular complexes organized around LAT, FcεRI, and Gab (Fig. 5). Together, these data raise the possibility that the different class IA PI3-kinases, composed of unique p85-p110 heterodimers, may distribute to distinct membrane domains to regulate different cellular functions. In particular, the presence of locally high PI3-kinase activity may induce locally high concentrations of D3-phosphoinositides to serve as lipid anchors for the recruitment of signaling molecules like PLCγ1. As well as determining the topography of recruited proteins, the local remodeling of membrane inositol phospholipid composition by nonrandomly distributed PI3-kinase family members may help to create or maintain preferred environments for transmembrane proteins like acylated LAT.

Current models of membrane structure envision membranes as dynamic mixtures of more or less ordered lipids in association with distinct proteins. An important example is the newly described immunological synapse that forms between the T cell and an antigen-presenting cell during conjugation of TCR–MHC peptide complexes (for review see Dustin and Chan, 2000). When observed by sophisticated fluorescence microscopic techniques, engaged TCR form the center of a bull's eye (referred to as the central supramolecular activation cluster or cSMAC), surrounded by a ring of adhesion receptors (the peripheral supramolecular activation cluster or pSMAC). Signaling molecules such as PKCθ may associate stably with the synapse (Monks et al., 1998), whereas others such as CD45 may be conditionally or transiently excluded (Sperling et al., 1998; Leupin et al., 2000; Johnson et al., 2000). Thus, the new models developed around data derived in both the TCR and FcεRI systems readily accommodate the concept that biological membranes may include one or more domains that are compositionally distinct from bulk membrane and can form and disassemble in a highly dynamic fashion. They also accommodate the hypothesis presented here that the segregation may be initiated in part when enzymes remodel the membrane inositol phospholipid composition after activation by receptor-coupled signaling pathways.

The relationship of the rafts isolated biochemically to the signaling domains observed microscopically remains to be determined. The protein associations we observe in native membrane sheets do not always correlate with implied associations based upon analysis of detergent-solubilized sucrose density fractions. For example, using the sucrose density fractionation protocol, we are able to confirm the localization of acylated proteins such as Lyn kinase and LAT to the light fraction. We also find a portion of PLCγ1 in the light fraction after FcεRI cross-linking. However, none of Syk, p85, PLCγ2, or Gab2 is found in the light fraction, even though their close interaction with receptor and LAT is readily demonstrated by TEM of native membranes.

In contrast, results using light microscopic approaches to elucidate interactions of FcεRI with rafts are more compatible with our results on native membrane sheets. Stauffer and Meyer (1997) showed that GFP chimeric proteins integrating the tandem SH2 domains of Syk were recruited to punctate structures at the plasma membrane, consistent with our observations that Syk is recruited to osmiophilic patches with receptors (Wilson et al., 2000). Using membrane sheets, we demonstrated that Lyn segregates from receptors with the first 2 min of cross-linking. The stringed appearance of Lyn as it segregates from osmiophilic patches, and the colocalization of Lyn with the actin-based cytoskeleton, suggests that cytoskeleton plays a role in Lyn's dissociation from receptors (Wilson et al., 2000). Using a complementary fluorescence confocal microscopy approach, Holowka et al. (2000) showed cytochalasin treatment leads to prolonged associations between cross-linked FcεRI and Lyn. Thus, a combination of light and electron microscopic approaches are likely to ultimately yield the clearest insight into the different protein–protein and protein–lipid interactions involved in the FcεRI signaling cascade.

Materials and methods

Reagents and cell culture

RBL-2H3 cells were grown in MEM (GIBCO BRL) supplemented with 10% fetal calf serum, penicillin–streptomycin, and l-glutamine. Monoclonal anti-p85α antibodies were from Santa Cruz Biotechnology, Inc. Rabbit anti-Gab2, anti-LAT and panreactive anti-p85 antibodies were from Upstate Biotechnology. Mouse anti-FcεRI β monoclonal antibodies and rabbit antimyosin (PTH) were gifts from Dr. Juan Rivera and Dr. Robert Adelstein, respectively (National Institutes of Health, Bethesda, MD). Anti-PLCγ1 (1249) and anti-PLCγ2 (Q-20) polyclonal antibodies were from Santa Cruz Biotechnology, Inc. Monoclonal antibodies to PLCγ1 were a mixture of E-12 (Santa Cruz Biotechnology, Inc.) and B-6-4 (Upstate Biotechnology). Isoform-specific PLCγ antibodies were tested by immunoblotting for absence of cross-reactivity to the other γ isoform. Affinity-purified mouse anti-DNP IgE was prepared as described (Liu et al., 1980; Seagrave et al., 1991). Anti–mouse HRP-conjugated antibodies were from Transduction Laboratories, and anti–rabbit HRP-conjugated antibodies were from Jackson ImmunoResearch Laboratories. Colloidal gold particles (5–10 nm in diameter) conjugated with anti–rabbit IgG, anti–mouse IgG, and streptavidin were from Nanoprobes and Amersham Pharmacia Biotech. Biotin–phalloidin and dinitrophenol-conjugated BSA (DNP-BSA) were from Molecular Probes.

Cell activation and membrane labeling

RBL-2H3 cells were allowed to settle overnight onto 15-mm round clean glass coverslips in the presence of anti-DNP IgE (1 μg/ml) to prime cell surface FcεRI. After washing to remove excess IgE, FcεRI were cross-linked by incubation for 2 min at 37°C with DNP-BSA (1 μg/ml). Plasma membrane sheets were prepared and labeled with antibody- or phalloidin-conjugated colloidal gold particles as described in Wilson et al. (2000), using a modification of procedures developed by Sanan and Anderson (1991). Samples were examined and photographed using an Hitachi H600 transmission electron microscope.

Quantifying gold particle distributions

Methods for counting gold particle distributions were established previously (Wilson et al., 2000). Here, micrographs from 2–4 separate experiments were sorted into groups according to distinct treatment and labeling conditions. For determination of cluster size and codistribution, gold particles were counted for matching sets of micrographs. Gold particles per set range 500–3,000 and reflect the relative abundance of label for antigen in the micrographs. For measuring translocation, numbers of gold particles were counted for each experimental condition over equivalent areas of membrane (defined in μm2).

Sucrose gradient centrifugation and analysis of membrane fractions

IgE-primed RBL-2H3 cells (40 × 106 cells per treatment condition) were harvested from culture dishes with 1.5 mM EDTA in Hanks' buffered saline without divalent cations. Washed cells were resuspended in Hanks' buffered saline, divided into two aliquots, and held for 2 min at 37°C with or without DNP-BSA (1 μg/ml). Cells were collected by centrifugation at 4°C, cell pellets were resuspended in 750 μl ice-cold lysis buffer containing low concentrations of detergent (10 mM Tris/HCl, pH 8.0, 0.05% Triton X-100, 50 mM NaCl, 10 mM EDTA, 10 mM glycerophosphate, 1 mM NaV04 and 1× protease inhibitor cocktail from Roche Molecular Chemicals). Lysates were mixed with 750 μl 80% sucrose (prepared in 10 mM Tris-HCl, pH 8.5, 50 mM NaCl, 2 mM EDTA) and overlaid onto 0.5 ml 80% sucrose in polyallomer tubes (13 × 51 mm), followed by 0.5-ml layers of 35, 25, and 20% and 0.6-ml aliquots of 15 and 10%. The gradient was centrifuged in a SW 55 (Beckman) rotor at 200,000 g for 16 h at 4°C. Fractions (0.5 ml) were harvested sequentially from the top of the gradient. For analyses of protein composition, aliquots (35 μl) were mixed with equal volume of 2× SDS sample buffer, boiled for 5 min, and separated by 8 or 10% SDS-PAGE. Alternatively, designated fractions were diluted in 50 mM Tris, 150 mM NaCl, pH 7.4, 1 mM NaV04 to a total volume of 1 ml and rocked for 1 h with beads prebound to primary antibodies, followed by separation of proteins bound to washed beads by SDS-PAGE. Proteins were transferred to nitrocellulose using a semidry blotting system (Labconco). Blots were probed with primary antibodies (1 μg/ml), followed by HRP-conjugated secondary antibodies (anti–mouse HRP-conjugates were diluted 1:10,000, and anti–rabbit HRP-conjugates were diluted 1:40,000); immunolabeled proteins were visualized by fluorography (ECL, Pierce Chemical Co.).

PI3-kinase assays

IgE-primed RBL cells were incubated with or without DNP-BSA at 37°C, and reactions were stopped with ice-cold saline. Cells were sedimented by microcentrifugation and lysed in 50 mM Tris/HCl, pH 7.2, 150 mM NaCl, 1 mM NaV04, and protease inhibitor cocktail (Boehringer) containing 1% Brij 96 or 1% Triton X-100, as specified. Lysates were clarified by centrifugation (15,000 g) for 5 min and rocked 2 h at 4°C with 40 μl (per ml of lysate) protein A– and G–Sepharose bead mixture (1:1) (Amersham Pharmacia Biotech) prebound to specified antibodies (1 μg). Washed immune complexes were incubated in reaction buffer (20 mM Hepes, pH 7.4, 5 mM MgCl2, 0.25 mM EGTA), containing 10 μM ATP, 20 μCi γ[32P]ATP, 500 μg/ml sonicated PtdIns (50 μl total) for 30 min at 37°C. The reactions were stopped by addition of 3N HCl. The lipid fraction was isolated using chloroform–methanol partitioning (Jackson et al., 1992) and resolved by thin-layer chromatography in 1-propyl acetate:2-propanol:ethanol:6% aqueous ammonia (1:3:1:3) (vol/vol) on silica gel 60 plates (Merck) (Hegewald, 1996). Formation of 32P-labeled PtdIns(3)P was imaged on a STORM 860 PhosphorImager (Molecular Dynamics) and quantified by means of Image Quant software. Data shown are duplicates ± SEM and are representative of at least three experiments.