The neck of caveolae is a distinct plasma membrane subdomain that concentrates insulin receptors in 3T3-L1 adipocytes

Abstract

Insulin receptors (IRs) segregate on plasma membrane microvilli, but in cells devoid of microvilli, such as adipocytes, the localization of IRs is a matter of controversy. In the present study, we examined the distribution of IRs in the plasma membrane of 3T3-L1 adipocytes. Quantitative electron microscopy indicates that IRs are predominantly associated with the neck, but not the bulb, of caveolae. Caveola necks represent distinct microdomains of the plasma membrane. Indeed, as shown by freeze–fracture analysis, intramembrane particles are concentrated as necklaces around the craters of caveolae. In addition, subcellular fractionation suggests that the neck and the bulb of caveolae present a different resistance to detergent solubility. Finally, cytoskeletal components, including actin, are highly enriched in the membrane area underlying the neck part of caveolae. IRs coimmunoprecipitate with cytoskeletal components, and disruption of the actin cytoskeleton alters IRs expression, localization, and signaling, thus supporting the notion that caveola necks are involved in intracellular signaling by IRs. Together, these results suggest that cytoskeletal proteins anchor IRs to microdomains in the caveola necks of 3T3-L1 adipocytes. By homology with IR localization in other cell types, we suggest that the necks of caveolae may represent the counterpart of microvillar domains in cells poor in microvilli such as adipocytes and that they play an important role as signaling platforms.

The mechanisms that account for the specificity of insulin effects on target tissues remain mostly unknown. One way of achieving insulin-signaling specificity is to regulate this process tightly in time and space. In this respect, we have shown previously that insulin receptors (IRs) are compartmentalized in specific plasma membrane domains, and after insulin stimulation, they undergo a journey through different domains of the cell surface, then inside the cell, and finally back to the cell surface (1). Along this route, distinct intracellular pathways have been reported to be activated, which supports the concept of both a spatial and temporal control of IR signaling (2). Understanding the biological processes controlling IR localization and trafficking is thus mandatory to unravel factors determining the specificity of the insulin-signaling cascades.

In various cell types, including hepatocytes, lymphocytes, and fibroblasts, IRs are enriched in thin digitations of the cell surface called microvilli (1). Compartmentalization of IRs on microvilli might reflect and depend on their coupling to cytoskeleton elements, which are particularly abundant in these structures (3). In this regard, actin has been shown to bind EGF receptors (4). Recently, filamin, an actin-binding protein, has also been shown to interact directly with the IR (5), thus providing the first identified cytoskeletal partner for IRs susceptible to mediate their anchoring to microvilli.

By contrast to other major target cells for insulin, adipocytes develop few microvilli on their surface. Instead, the plasma membrane of mature adipocytes is studded with small invaginations, the caveolae (6). Caveolae are 50- to 100-nm flask-shaped invaginations of the plasma membrane enriched in cholesterol and sphingolipids covered on their cytoplasmic side with a characteristic striated coat consisting of oligomerized caveolin proteins (7). Through both qualitative morphological or biochemical analyses, recent studies have suggested an association of IRs with caveolae in adipocytes (8, 9), but these results have been challenged by others who failed to find any preferential association of IRs with purified caveolae (10, 11).

In the present work, we examined where and how IRs are concentrated in the plasma membrane of mouse 3T3-L1 adipocytes through both biochemical and quantitative morphological analyses at the ultrastructural level. Our results indicate that IRs are anchored to specific microdomains forming the neck of caveolae through interactions with the underlying cortical cytoskeleton network and that this localization is critical for mitogenic insulin signaling.

Results

IRs Associate with Caveolae in Adipocytes.

To examine whether IRs concentrate in specific microdomains of the plasma membrane in 3T3-L1 adipocytes, we labeled IRs with colloidal gold and localized them by EM on purified plasma membranes. IR-associated gold particles were observed by transparency through plasma membrane sheets adherent by their extracellular side to poly-l-lysine-coated EM grids (Fig. 1A). By this method, plasma membrane microdomains such as clathrin-coated pits/lattices and caveolae are morphologically clearly identified, and the presence of gold-labeled IRs within these structures can be accurately quantified. Approximately 50% of the gold particles were observed within caveolae or closely apposed to them (Fig. 1B). Other gold particles were dispersed throughout flat and uncoated plasma membrane domains, whereas <0.7% of the labeling was associated with clathrin-coated areas. Caveolae constituted 25.8% of the plasma membrane in 3T3L1 adipocytes as determined by morphometric analysis of Epon-embedded thin sections (Table 1). It thus appears that IR association with caveolar structures is highly preferential (>3 times more than in other areas) (Fig. 1B).

Fig. 1.

IRs associate with caveolar microdomains. (A) Electron micrograph of IR gold labeling associated with caveolae (arrows) on 3T3-L1 adipocyte plasma membrane sheets. (Scale bar, 0.1 μm.) (B) Quantification of IR association and enrichment in caveolae. Data are means ± SE of quantifications performed on 10 micrographs together totaling 41.4 μm² of membranes and 1,000 gold particles.

Table 1.

Percentage of plasma membrane constituted by distinct domains in 3T3-L1 adipocytes

	Flat membrane	Microvilli	Clathrin-coated pits	Caveolae
Control	69.0 ± 1.5	4.8 ± 1.0	0.3 ± 0.1	25.8 ± 1.4
+ Latrunculin	68.6 ± 1.5	3.8 ± 0.8	0.8 ± 0.2	26.7 ± 1.5

Results are derived from the analysis of 98 micrographs per 539.55 μm of membrane and 79 micrographs per 537.04 μm of membrane for control and latrunculin-treated cells, respectively.

IRs Associate Predominantly with the Neck of Caveolae.

We previously demonstrated that specific markers residing within caveolae, such as gangliosides, can be efficiently labeled by gold particles (up to 15 nm) on Epon-embedded thin sections of 3T3-L1 cells (12). By using immmunogold labeling of IRs, we found that IRs preferentially associated with caveolae (Fig. 2A and B). However, IRs were mostly present (≈60% of the gold labeling) at the neck of these invaginations with <8% of gold particles present in the bulb (Fig. 2D). Because the neck of caveolae was defined as the 20-nm length of plasma membrane on each side of the tip of the caveola neck curvature (Fig. 2E), these microdomains represent 17.5% ± 1.2% of the plasma membrane (quantifications derived from 539.55 μm of surface membranes analyzed presenting 2,260 caveolae). Accordingly, IRs are enriched 3.3 ± 0.1-fold at the neck of caveolae. Upon insulin stimulation, the percentage of IRs associated with the neck of caveolae increased (Fig. 2D), suggesting that activated IRs are recruited to the caveola necks. Potential crosslinking or steric hindrance effects of gold-conjugated secondary antibody probes on the IR distribution were ruled out because use of antibodies conjugated to gold particles of smaller size (5 nm) or gold-conjugated protein A did not affect the observed distribution of IRs (data not shown).

Fig. 2.

IRs associate with the neck of caveolae and microvilli. Electron micrographs show IR gold labeling on the necks of caveolae (A and B) and on microvilli (C) of 3T3-L1 adipocytes. (Scale bar, 0.25 μm.) (D) Quantification of IR association with distinct plasma membrane microdomains. Data are means ± SE of quantifications performed on 98–101 micrographs from two experiments together totaling 103 cells per 2,665 gold particles and 102 cells per 2,105 gold particles for unstimulated and insulin-stimulated cells, respectively. *, P < 0.05; **, P < 0.01. (E) Scheme of a caveola. Black dotted lines, caveola neck areas; gray line, caveola bulb.

Interestingly, although 3T3-L1 adipocytes are almost devoid of cytoskeleton-enriched membrane protrusions, the rare microvilli observed at the surface (i.e., 4.8% of the plasma membrane; Table 1) were also, like caveolae, heavily labeled by gold particles (Fig. 2C), suggesting that IRs have a high propensity to associate with cytoskeletal-rich areas of the plasma membrane.

Together, these results indicate that in 3T3-L1 adipocytes, IRs concentrate mainly at the neck of caveolae, which must thus represent a specific subdomain.

The Neck of Caveolae Represents a Specific Plasma Membrane Microdomain.

The characteristics of the caveola neck were further analyzed on freeze–fracture replicas, allowing visualization of the distribution of intramembrane particles, which represent transmembrane or membrane-associated proteins (13, 14). As illustrated in Fig. 3A, intramembrane particles are highly concentrated around the craters of the caveolae. Quantification of the intramembrane particle density as a function of the distance from the caveola necks indicated that intramembrane particles are ≈1.7-fold more concentrated in the neck area than in other regions of the plasma membrane (Fig. 3B). Immunogold labeling of IRs detected on freeze–fracture replicas further confirmed the localization of IRs in these microdomains (Fig. 3C).

Fig. 3.

The caveola necks concentrate intramembrane particles including the IRs. (A) Freeze–fracture replica of a 3T3-L1 adipocyte showing accumulation of intramembrane particles (arrows) around the neck of the caveola craters in the protoplasmic face of the membrane. (Scale bar, 0.1 μm.) (B) Quantification of intramembrane particle densities in concentric ring-like areas surrounding the caveola craters. Data are means ± SE of particle densities ≈50 caveolae. **, P < 0.01; ***, P < 0.001. (C) Freeze–fracture replica showing IR gold-associated particles (arrowheads) around the neck of the caveolae in the exoplasmic face of the membrane. (Scale bar, 0.2 μm.)

Detection of IRs in caveola-enriched subcellular fractions also suggested that the bulb and the neck of caveolae are structurally distinct microdomains. Indeed, caveolar microdomains are not soluble in various detergents at low temperature, and they can be isolated by floatation on sucrose density gradients. We thus isolated 3T3-L1 adipocyte detergent-resistant membranes by using Triton X-100, CHAPS, or Brij 98 detergents. In all of these conditions, IRs were never recovered in floating membranes highly enriched in caveolin or GM1, two common markers for caveolae (Fig. 4). On the contrary, IRs comigrated with caveolin/GM1-enriched fractions when caveolae were purified by an alternative procedure in the absence of any detergents (15, 16). However, in these conditions, transferrin receptors, which are classically excluded from caveolae, cofractionated with the caveolin/GM1 markers, indicating that in the absence of detergents pure caveola fractions cannot be resolved (Fig. 4).

Fig. 4.

IR cofractionation with caveolar membranes is detergent-sensitive. Typical distribution of IR, transferrin receptor (TfR), caveolin (Cav.), and GM1 in caveola-enriched subcellular fractions of 3T3-L1 adipocytes prepared either in the presence or absence of various detergents is shown. Data are representative of three independent experiments.

Together, these data indicate that the necks of caveolae are distinct microdomains where a high density of proteins, including IRs, are concentrated, and which differ from the caveola bulbs in terms of resistance to detergent solubility.

IRs Interact with Cytoskeletal Proteins at the Neck of Caveolae.

As previously described in other cell types (1), we found IRs more concentrated on cytoskeleton-enriched microvilli (≈1.2-fold enrichment) in adipocytes (Fig. 2). Caveolae and the actin cytoskeleton are also intimately linked as suggested by a recent report (17), which led us to hypothesize that the neck of caveolae and microvilli might display structural and functional similarities to concentrate IRs. In support of this assumption, microvilli and the neck of caveolae presented a typical electron-dense material reminiscent of the cortical cell web, and immunogold labeling of cytoskeletal proteins on ultrathin cryosections indicated the presence of actin both in the microvilli cores and at the neck of caveolae (Fig. 5 A and B). By contrast, the electron-dense coat is not surrounding the bulb of caveolae, which is also mostly devoid of labeling for cytoskeletal proteins (Fig. 5 B–D). Detection of ERM proteins and talin also evidenced the presence of these actin-binding proteins in the electron-dense material in close proximity to the caveola neck membranes (Fig. 5 C and D).

Fig. 5.

Microvilli and the neck of caveolae are enriched in cytoskeletal elements. Electron micrographs show immunogold labeling of cytoskeletal proteins in the cytoplasm underlying microvillar and caveolar neck microdomains. (A) High actin content of microvilli. (Scale bar, 0.2 μm.) Actin (B), ERM (C), and talin (D) are detected in the cytoplasmic area close to the neck curve of caveolae. (Scale bars, B, 0.1 μm; C and D, 0.2 μm.)

Whether anchoring of IRs to cytoskeleton-enriched microdomains is mediated by direct or by indirect interactions of the receptors with cytoskeletal elements is still unclear, but filamin, an actin-binding protein, has been shown to interact directly with the IR in HepG2 cells (5). We could not assess the localization of filamin at the ultrastructural level because of the lack of antibodies appropriated for immunocytochemistry. However, filamin was coimmunoprecipitated specifically with IRs in insulin-stimulated cells (Fig. 6A). Moreover, specific interactions of the IR with actin and moesin, a member of the ERM family, were also detected (Fig. 6A).

Fig. 6.

Cytoskeleton disruption affects IR association with caveolae and signaling. (A) Coimmunoprecipitation of IRs with actin, moesin, and filamin in 3T3-L1 adipocytes. Immunoprecipitation of the transferrin receptor (TfR) was used as a control. Data are representative of three to five independent experiments. (B) Effect of actin-disrupting drugs on IR association with caveolae. The micrograph shows that caveolae (arrowheads) are preserved in latrunculin-treated cells. The graph indicates the percentage of caveolar structures that are labeled by IR-associated gold particles in cells pretreated or not with latrunculin (45 μM, 3 h at 37°C) or cytochalasin D (20 μg/ml, 3 h at 37°C). Data are means ± SE of quantifications performed on 49–98 electron micrographs for each condition, totaling 52 cells per 549 gold particles per 1,447 caveolae, 79 cells per 573 gold particles per 1,512 caveolae, and 52 cells per 589 gold particles per 1,787 caveolae analyzed for control, latrunculin-, and cytochalasin D-treated cells, respectively. *, P < 0.05; ***, P < 0.001. (C) Akt and ERK1/2 phosphorylation by insulin. Cells were treated (L) or not (C) with latrunculin (45 μM, 3 h at 37°C) and stimulated with 10⁻⁸ M insulin for different times. A representative Western blot of phosphorylated Akt and ERK1/2 is shown as well as quantifications of the ratio of phosphorylated to total proteins. Results are the mean ± SE of three independent experiments. * P < 0.05.

Together, these results suggest that localization of IRs to the neck of caveolae is mediated by direct or indirect interactions of the receptors with the underlying cytoskeletal proteins enriched in these plasma membrane microdomains.

IR Expression, Localization, and Signaling Depend on Cytoskeleton Integrity.

To assess whether the integrity of the cortical cytoskeleton network is required for IR localization to the neck of caveolae, cells were treated with latrunculin, a drug known to disrupt the actin cytoskeleton, and association of IRs with caveolar domains was examined at the ultrastructural level. Latrunculin decreased the efficiency of IR gold labeling at the surface of 3T3-L1 adipocytes (3.77 ± 0.27 versus 1.28 ± 0.08 gold particles per μm of membrane in control and latrunculin-treated cells, respectively, P < 0.001), suggesting that surface expression of IRs is altered in latrunculin-treated cells. IR down-regulation was further supported by ¹²⁵I-insulin-binding experiments showing that 3T3-L1 adipocytes treated with latrunculin bind 30% ± 3% (P = 0.086) less ¹²⁵I-insulin than untreated cells (data not shown). The integrity and abundance of caveolae (Fig. 6B and Table 1) as well as the distribution pattern of gold-tagged IR between the bulb and the neck of caveolae (data not shown) were not altered in latrunculin-treated cells. However, the percentage of caveolae labeled for IRs relative to the total number of caveolar structures present at the membrane was decreased by half in cells treated with latrunculin or cytochalasin D, another actin-disrupting drug structurally unrelated to latrunculin (Fig. 6B). Thus, disruption (at least partial) of the cytoskeleton decreases IR surface expression and the proportion of neck microdomains containing IRs, thus outlining the importance of the cytoskeleton integrity for a correct IR expression and localization at the plasma membrane.

Functional relevance of IR association with cytoskeleton elements at the neck of caveolae was also assessed by analyzing the activation of downstream effectors of insulin, i.e., Akt and ERK1/2, which mediate, respectively, the metabolic and mitogenic effects of insulin. We found that latrunculin treatment did not affect insulin signaling through the PI3-kinase pathway as reflected by the unaltered phosphorylation of Akt. In contrast, insulin-induced ERK1/2 phosphorylation was inhibited by latrunculin (Fig. 6C), suggesting that insulin mitogenic signals are dependent on the cytoskeleton integrity and thus likely on IR localization to the cytoskeleton-enriched neck of caveolae.

Discussion

In this work, we demonstrate that in 3T3-L1 adipocytes, IRs are segregated to specific microdomains located at the neck of caveolae. These microdomains display characteristics that distinguish them from flat uncoated plasma membranes or the bulb of caveolar invaginations. Indeed, we show that the necks of caveolae are connected to cytoskeletal elements and concentrate intramembrane particles. Our morphological observations, together with biochemical data supporting the absence of IRs in caveola bulbs and a tight association of IRs with cytoskeletal proteins, indicate that IRs are associated with specific microdomains forming the neck of caveolae through interactions with the cortical cytoskeleton. Finally, we show that the integrity of the cytoskeleton is critical for IR expression, localization, and signaling.

The association of IRs with caveolae is controversial. IR association with caveolae was suggested by qualitative EM analyses of isolated 3T3-L1 adipocyte plasma membranes as well as by biochemical fractionation of caveolae with a detergent-free procedure (9, 16). However, other reports revealed a lack of IR association with caveolin-enriched and detergent-insoluble fractions or with immunopurified caveolae (10, 11). Our detailed localization of IRs on ultrathin sections and on fractured replicas of 3T3-L1 adipocytes now provides a possible explanation for these discrepancies. Indeed, IRs are not enclosed in the bulb of caveolar structures but instead anchored to microdomains forming the neck of cavolae. Previous morphological localization of IRs did not allow us to distinguish between the neck and the bulb of caveolae, and detergent-free procedures used to fractionate caveolae can likely not separate these two distinct microdomains. Regarding this last hypothesis, we provide evidence that IRs are excluded from detergent-insoluble caveolae in 3T3-L1 adipocytes, whereas they cofractionate with caveolin/GM1-positive fractions when caveola purification was performed in the absence of detergent. These data suggest that isolation of caveolae by using a detergent-free procedure copurifies also the plasma membrane neighboring the caveolae including the neck of these structures. This issue has been raised recently by others in a broad analysis of methods developed to isolate lipid raft microdomains (18).

Cholesterol-enriched lipid rafts or caveolae have been suggested to represent plasma membrane compartments involved in insulin signaling in 3T3-L1 adipocytes. Indeed, studies in which cells have been treated with cholesterol-sequestering agents indicate an important role for cholesterol-enriched microdomains in transducing PI3-kinase-dependent insulin signaling (8, 19). In addition, a new signaling pathway involving the Cbl–CAP–TC10 complex and activated by IRs has been described to be restricted to caveolae. Although this new pathway is initiated in caveolae and is cholesterol-dependent, these studies did not report a direct association of IRs with caveolae (19, 20). Our results are not contradictory with these studies because it is conceivable that cholesterol is needed for efficient IR signaling even if the IR is not localized within caveolae as has been shown for the EGF receptor (21). Alternatively, microdomains forming the neck of caveolae may also contain cholesterol even if they are distinct from the caveola core in terms of resistance to detergent solubility and protein composition, which may explain the sensitivity of insulin signaling to cholesterol-sequestering agents. Interestingly, Vainio et al. (22) have described an association of IRs with lipid rafts upon insulin stimulation in liver-derived cells devoid of caveolae. Other studies have also shown that the EGF receptor associates with caveolar-like domains for signaling, then exits these structures to be captured by clathrin-coated pits and internalized (23). Our data regarding the IRs support this concept because we found that activated IRs are recruited to the caveola neck microdomains upon insulin stimulation but are likely internalized through clathrin-coated pits as we previously demonstrated (24). Because many other lipid raft-dependent signaling pathways rely on raft aggregation, IR aggregation to the caveola neck microdomains might thus represent an important mechanism to mediate transduction of insulin signaling in 3T3-L1 adipocytes. Further investigations are now needed to verify the relevance of this hypothesis.

We previously reported that IRs are located on microvilli in several cell types (1, 25), and the rare microvilli forming at the surface of 3T3-L1 cells are also concentrating IRs, which led us to hypothesize that the necks of caveolae may represent the counterpart of microvillar domains in cells mostly devoid of microvilli such as adipocytes. In this regard, microvilli display structural similarities with caveola necks. Microvilli are enriched in cytoskeletal elements (3), and they are characterized by the presence, in freeze-etched replicas, of a high concentration of intramembrane particles (26). Similarly, and as previously observed in intestinal smooth muscle cells (27), the necks of caveolae concentrate also intramembrane particles in 3T3-L1 adipocytes. In addition, we showed that in 3T3-L1 ultrathin cryosections an electron-dense layer of material, reminiscent of the cytoskeletal network, is present at the necks of caveolae, but it does not surround the bulb. Cytoskeletal proteins, such as actin, ERM, and talin are also specifically present at the neck of caveolae. Interestingly, in smooth muscle cells, thin filaments of cytoskeleton were also shown to be anchored to the neck of caveolae (28). Finally, actin was found recently to localize in 3T3-L1 adipocytes along the inner circumference of ring-like caveola rosettes, suggesting that actin cytoskeleton is anchored to plasma membrane areas concentrating caveolae (17). Together, these observations support a structural and functional analogy in terms of insulin signaling between the caveola neck microdomains and plasma membrane microvilli, which both concentrate IRs. Noteworthy, we observed that insulin tends to relocate IRs from microvilli to the caveolar necks in 3T3-L1 cells, which might be because insulin can induce a shortening of microvilli and a preferential cytoskeleton polymerization at sites where IRs are mostly concentrated, thus partially relocating IRs from microvilli to caveolar necks (29, 30).

IR localization in cytoskeleton-enriched membrane domains, such as microvilli or the caveola necks, suggests also that IRs might interact with cytoskeletal proteins. In HepG2 cells, IRs were shown to interact constitutively with filamin, an actin-binding protein (5). In 3T3-L1 adipocytes, we found similarly that IRs associate with filamin. Because filamin is constitutively associated with small GTPases that control the actin cytoskeleton organization, interaction of activated IRs with filamin could be the primary step in the processes of insulin-induced cytoskeleton reorganization (31). However, association of IR with filamin does not likely represent the primary interaction responsible for anchoring IRs to the neck of caveolae because the two proteins interact only after insulin stimulation. By contrast, we identified actin and moesin as two other potential candidates connecting IRs to the cytoskeleton. It is thus tempting to speculate that moesin, which connects the actin network to membrane proteins, represents the cytoskeletal link allowing IRs to anchor either to microvilli or to the caveola necks. In this regard, we previously described that IR sequences mediating its anchoring to microvilli were contained within exon 17 and involve in particular two dileucine-like motifs (32, 33). Whether these sequences are involved in the binding of the receptor to moesin is currently under investigation.

The role of the cytoskeleton and caveolar domains in insulin signaling is still unclear. As previously observed by others (17), we found that latrunculin-induced actin depolymerization in 3T3-L1 adipocytes does not affect the caveola integrity. However, insulin signaling through the ERK1/2 pathways is significantly inhibited in latrunculin-treated cells, whereas PI3-kinase/Akt activation is unaffected. Studies performed in L6 myoblasts showed equally that actin disassembly prevents insulin-induced ERK1/2 but not PI3-kinase activation (34), and interaction of IRs with the cytoskeletal protein filamin was also shown to modulate ERK1/2 activation (5). In contrast, membrane cholesterol depletion, which disrupts the lipid organization of caveolar domains, was shown to prevent insulin activation of the PI3-kinase pathway but not ERK1/2 signaling (8). Together, these data suggest that a cholesterol-enriched lipidic microenvironment is critical to allow IRs to stimulate the PI3-kinase/Akt pathway, whereas ERK1/2 activation requires IR interaction with the cytoskeleton. Whether membranes constituting the necks of caveolae are rich in cholesterol as those forming the bulb of these invaginations has yet to be investigated. However, because caveola necks concentrate IRs and are connected to the cytoskeletal network, these specific domains likely provide an integrative signaling platform propagating both metabolic and mitogenic signals triggered by insulin.

Materials and Methods

Reagents and Cell Culture.

Cytochalasin D and horseradish peroxidase/cholera toxin were purchased from Sigma (St. Louis, MO), and latrunculin A was from Calbiochem (Luzern, Switzerland). Anti-actin was purchased from Chemicon International (Temecula, CA); anti-filamin was from Neomarkers (Fremont, CA); anti-moesin and anti-caveolin were from Transduction Laboratories (Lexington, KY); anti-ERM and talin were from Santa Cruz Biotechnology (Santa Cruz, CA); anti-transferrin receptor was from Zymed (San Francisco, CA); anti-Akt, anti-phospho-Akt (Ser-473), anti-ERK1/2, and anti-phospho-ERK1/2 were from Cell Signaling Technnology (Danvers, MA); gold-conjugated antibodies were from British Biocell International (Cardiff, U.K.). Anti-IR (clone 83-14) was provided by K. Siddle, Cambridge University, Cambridge, U.K; 3T3-L1 adipocytes were grown and differentiated as described previously (35).

EM Analyses.

IR immunogold labeling.

Cells serum-starved for 12 h were incubated 2 h at 4°C with 83-14 (1 μg/ml) in PBS/1% (vol/wt) BSA, washed, and incubated with secondary 10-nm gold-conjugated antibodies for 90 min at 4°C. Cells were then warmed at 37°C in the presence of 10⁻⁷ M insulin before being processed for EM analysis. Samples were examined on a CM10 TEM (Philips, Eindhoven, The Netherlands).

Epon-embedded ultrathin sections.

Immunogold labeled cells were fixed, dehydrated, and processed for EM as described previously (36).

Isolated plasma membranes.

The plasma membranes of immunogold-labeled cells were isolated and negative stained as described previously (36).

Freeze–fracture.

Cells were fixed in 0.1 M phosphate buffer/2.5% (vol/vol) glutaraldehyde (pH 7.4), infiltrated in 30% phosphate-buffered glycerol, and frozen in Freon 22 cooled with liquid nitrogen. Fracture and shadowing were carried out in a Balzers BAF301 apparatus (Bolgers, Fürstentum, Liechtenstein). Replicas were then washed successively in a sodium hypochlorite solution, in chloroform/methanol 2:3, and rinsed in distilled water before mounting on EM grids. With cells immunogold-labeled for IRs, the replicas were washed with 2.5% SDS/10 mM Tris·HCl (pH 9) instead of sodium hypochlorite and chloroform/methanol to preserve the labeling (37).

Ultrathin cryosections.

3T3-L1 adipocytes were fixed, scraped from the Petri dishes, and processed for cryosectioning as previously described (38). Frozen sections were then cut with a Leica FCS cryotome (Leica, Vienna, Austria) transferred to grids, and incubated with the indicated primary antibodies followed by gold-conjugated secondary antibodies.

Morphometric analyses and quantifications.

Quantification of the gold-labeled IR localization and morphometric analyses of plasma membrane differentiations were performed on electron micrographs obtained from either Epon-embedded ultrathin sections or negative-stained plasma membrane sheets by using a Wacom tablet coupled to the QWin software (Leica, Vienna, Austria). Quantification of intramembrane particle densities on freeze–fracture replicas were obtained by counting the number of intramembrane particles in concentric ring-like surface of known areas surrounding the caveola craters.

Statistical analysis.

Results are expressed as mean ± SE. Comparisons were made by using Student's t test. Differences were considered as significant when P < 0.05 (*), P < 0.01 (**), or P < 0.001 (***).

Subcellular Fractionation.

3T3-L1 adipocytes were scraped and lysed in homogenization buffer (50 mM Tris·HCl, pH 7.5/100 mM NaCl/2 mM EDTA/10 mM NaF/5 mM VO₄/protease inhibitors) with tight Dounce homogenizer before ultracentrifugation at 200,000 × g at 4°C to pellet microsomal membranes. Microsomes were then solubilized at 4°C for 20 min in 25 mM Mes/150 mM NaCl containing 1% Triton X-100, 1% Brij 98, or 2% CHAPS. Samples adjusted to 40% sucrose were overlayed with 2 ml of 30%, 25%, 15%, and of 5% sucrose solutions prepared in 25 mM Mes/150 mM NaCl. Samples were ultracentrifuged in a SW41Ti rotor (Beckman, Fullerton, CA) for 18 h at 200,000 × g. Collected fractions were trichloroacetic acid-precipitated and analyzed by Western blotting. Isolation of caveolae in the absence of detergents was performed as described previously (15, 16).

Immunoprecipitation.

Cells were lysed in ice-cold buffer A [20 mM Hepes, pH 7.2/50 mM NaCl/10% (vol/vol) glycerol/1% Triton X-100/protease inhibitors] for IR coimmunoprecipitation with actin, buffer B [25 mM Tris·HCl, pH 8/150 mM NaCl/1 mM EDTA/10% (vol/vol) glycerol, 1% Triton X-100/protease inhibitors] for IR coimmunoprecipitation with moesin and in buffer B + 2 mM DTT for IR coimmunoprecipitation with filamin. Lysates were precleared and immunoprecipitated with 1 μg of anti-IR 83-14 or anti-filamin. Immune complexes were then pulled down with protein A/G-Sepharose, resolved by SDS/PAGE, and analyzed by Western blotting with an ECL kit from Amersham (Piscataway, NJ).

Insulin Signaling.

Cells stimulated with 10⁻⁸ M insulin at 37°C were lysed in RIPA buffer including 10 mM NaF, 5 mM VO₄, and protease inhibitors, and equal amount of proteins were resolved by SDS/PAGE and analyzed by Western blotting with an ECL kit from Amersham. ECL signals were quantified by using ChemiDocXRS from Bio-Rad (Hercules, CA) and Quantity One software (Bio-Rad).

Abbreviations

IR

insulin receptor

PI3-kinase

phosphatidylinositol 3-kinase.