• We are sorry, but NCBI web applications do not support your browser and may not function properly. More information
Logo of pnasPNASInfo for AuthorsSubscriptionsAboutThis Article
Proc Natl Acad Sci U S A. Oct 25, 2005; 102(43): 15500–15505.
Published online Oct 13, 2005. doi:  10.1073/pnas.0504114102
PMCID: PMC1266090
Cell Biology

H-ras, K-ras, and inner plasma membrane raft proteins operate in nanoclusters with differential dependence on the actin cytoskeleton

Abstract

Plasma membrane compartmentalization imposes lateral segregation on membrane proteins that is important for regulating signal transduction. We use computational modeling of immunogold spatial point patterns on intact plasma membrane sheets to test different models of inner plasma membrane organization. We find compartmentalization at the nanoscale level but show that a classical raft model of preexisting stable domains into which lipid raft proteins partition is incompatible with the spatial point patterns generated by the immunogold labeling of a palmitoylated raft marker protein. Rather, ≈30% of the raft protein exists in cholesterol-dependent nanoclusters, with ≈70% distributed as monomers. The cluster/monomer ratio (number of proteins in clusters/number of proteins outside clusters) is independent of expression level. H-rasG12V and K-rasG12V proteins also operate in nanoclusters with fixed cluster/monomer ratios that are independent of expression level. Detailed calibration of the immunogold imaging protocol suggests that radii of raft and RasG12V protein nanoclusters may be as small as 11 and 6 nm, respectively, and shows that the nanoclusters contain small numbers (6.0-7.7) of proteins. Raft nanoclusters do not form if the actin cytoskeleton is disassembled. The formation of K-rasG12V but not H-rasG12V nanoclusters also is actin-dependent. K-rasG12V but not H-rasG12V signaling is abrogated by actin cytoskeleton disassembly, which shows that nanoclustering is critical for Ras function. These findings argue against stable preexisting domains on the inner plasma membrane in favor of dynamic actively regulated nanoclusters similar to those proposed for the outer plasma membrane. RasG12V nanoclusters may facilitate the assembly of essential signal transduction complexes.

Keywords: lipid rafts, microdomains, modeling

The plasma membrane is a complex, heterogeneous, lipid bilayer that is compartmentalized by combinations of lipid-lipid, lipid-protein, and actin-cytoskeleton interactions. This heterogeneity is biologically important because it offers an attractive mechanism to regulate the lateral segregation and output of membrane-anchored signaling proteins (1). Cholesterol-enriched, liquid-ordered domains (lipid rafts) that phase separate from surrounding liquid-disordered membrane in model in vitro systems have been postulated also to exist in biological membranes and to serve important roles in assembling cell-signaling complexes on the inner surface of the plasma membrane. The validity of this hypothesis continues to be debated (2). With the exception of lipid rafts aggregated by caveolin into caveolae, this uncertainty is due to their lack of detectable ultrastructure, the differences in size estimates reported, and the limitations of the techniques used to characterize these domains (3). Rafts were originally defined operationally, through their insolubility in various detergents and the ability of cholesterol depletion to disrupt their formation (4). However, there are problems with these approaches because detergent extraction has been shown to cause the formation of domains (5) and cholesterol depletion has additional effects on the actin cytoskeleton (6).

More sophisticated techniques have investigated raft domains in intact cells: fluorescence recovery after photobleaching, immunoEM, single-fluorophore tracking microscopy, photonic force microscopy, and FRET. Each of these techniques has different spatial and temporal resolution. Consequently, several basic models have emerged that aim to describe the characteristics of lipid rafts. In their classical representation, lipid rafts are considered relatively large structures (≈50 nm) enriched with cholesterol and sphingolipid that diffuse as stable entities within the fluid bilayer, into which proteins are selectively included or excluded (4, 7). Alternatively, the lipid shell hypothesis envisages cholesterol-sphingolipid-rich shells containing 80 lipid molecules ≈7 nm in diameter that exist as mobile entities in the plasma membrane. The shells have an affinity for preexisting caveolae/rafts and target the lipid-anchored or transmembrane protein they encase specifically to these membrane domains (8). A variation of this lipid shell model proposes an actively generated spatial and temporal organization of raft components in which lipid assemblies are small and dynamic and coexist with monomers (9, 10). Most studies have examined the outer plasma membrane; thus, it remains unclear whether raft domains on the inner and outer plasma membrane are linked and whether their formation is governed by similar principles (3).

Ras proteins are key regulators of signal transduction that are also useful markers to explore the microorganization of the inner plasma membrane. H-ras, K-ras, and N-ras are highly homologous proteins that differ significantly only in their C-terminal sequences, yet they generate distinct signal outputs (11). This biological diversity flows from the different membrane anchors used by the Ras isoforms to interact with the plasma membrane. EM and fluorescence recovery after photobleaching analyses show that H-ras has transient interactions with lipid rafts when bound to GDP but clusters in cholesterol-insensitive, galectin-1-dependent, nonraft domains when bound to GTP. K-ras also clusters in cholesterol-insensitive, nonraft domains that are spatially distinct from the activated H-ras microdomains (12-14). These conclusions are confirmed in part by FRET microscopy, which shows that different lipid anchors segregate peptides to different domains on the inner surface of the plasma membrane (15).

Here we use mathematical realizations of different lipid raft models to predict the surface distributions of an inner leaflet raft protein that would be expected were the protein labeled and detected by immunoEM. We compare the predicted and experimentally observed immunogold patterns to discriminate between the models and determine domain size and the number of proteins per cluster. We then extend the study to identify whether similar models can account for the plasma membrane distributions of activated H-ras and K-ras. Finally, we examine the contribution of the actin cytoskeleton to the formation of Ras microdomains and lipid rafts. These observations provide new insights into the formation of raft and nonraft domains on the inner leaflet of the plasma membrane and the complex interplay between plasma membrane organization and Ras signaling.

Materials and Methods

Cell Culture. BHK cells cultured at 37°C were transfected by using Lipofectamine (Invitrogen) as described in ref. 16. After 16 h, the cells were processed for EM analysis or Western blotting. Expression plasmids for GFP-tH (tH, plasma membrane-targeting domain of H-ras), GFP-H-rasG12V and GFP-K-rasG12V are described in ref. 13. The expression plasmid for monomeric red fluorescent protein (mRFP)-tH was constructed from an mRFP1 cDNA (17) kindly provided by R. Tsien (University of California at San Diego, La Jolla). For actin perturbation experiments, cells were treated with 50 nM or 1 μM latrunculin for 5 min before processing for EM.

Western Blotting. Whole-cell lysate (20 μg) was immunoblotted for GFP (Roche) and phosphorylated extracellular signal-regulated kinase (ERK) (Cell Signaling Technology, Beverly, MA). Blots were developed with enhanced chemiluminescence and quantified in a LumiImager (Roche Molecular Biochemicals).

EM. Apical plasma membrane sheets were prepared, fixed with 4% paraformaldehyde, 0.1% glutaraldehyde, and labeled with affinity-purified anti-GFP or anti-mRFP anti-sera coupled directly to 4-nm gold as described in ref. 13. Digital images of the immunogold-labeled plasma membrane sheets were taken at ×100,000 magnification. Intact 1-μm2 areas of the plasma membrane sheet were identified with image j, and the (x,y) coordinates of the gold particles were determined as described in ref. 13.

Sizing Microdomains. Colloidal gold (10- to 22-nm) was prepared according to standard procedures and adjusted to pH 7.0. Recombinant GFP prepared in Escherichia coli was added to the gold solution (2.5 μg of GFP per ml of gold) and stabilized with BSA (0.1% final concentration). The gold was centrifuged, and the loose pellet applied to a polylysine-coated EM grid, which was then labeled with 4-nm anti-GFP-gold. Large GFP-gold particles surrounded by anti-GFP-gold labeling were analyzed as described above.

Statistical Analysis. The K function (18, 19) was calculated according to Eqs. 1 and 2 and standardized on a 99% confidence interval (13). Bootstrap tests for differences between replicated point patterns were constructed as described in ref. 20, and statistical significance was evaluated against 1,000 bootstrap samples.

equation M1
[1]

K(r) is the K function for a pattern of n points in an area A. r is the radius at which K(r) is calculated (1-240 nm at 1-nm increments), ||. || is Euclidean distance, and 1(·) is the indicator function that takes a value of 1 if ||xi - xj||r and 0 otherwise. wij-1 is the proportion of the circumference of a circle with center xi and radius ||xi - xj|| contained within A.

equation M2
[2]

L(r)-r is a linear transformation of K(r). Under the null hypothesis of complete spatial randomness, L(r)-r has an expected value of 0 for all values of r.

Mathematical Models. Two models were used to realize immunogold point patterns that would be generated under a classical or alternative lipid raft model (described in Fig. 1).

Fig. 1.
In classical raft models, a fixed number of rafts accommodate a fixed fraction of raft proteins. As the expression levels of a raft partitioning protein increases (indicated by the arrow) more proteins must be accommodated in the rafts. The extent of ...

Fixed parameters for both models.

  1. The size of the study area, A, is 1,000 nm × 1,000 nm.
  2. The radius of the raft/cluster is the radius at which we observe maximal deflection of the L(r)-r curve in K function analyses. To model GFP-tH and mRFP-tH surface distributions, r was set to 22 nm. To model H-rasG12V and K-rasG12V surface distributions, r was set to 16 nm; the value of r does not vary with expression level of GFP-tH, mRFP-tH, H-rasG12V, and K-rasG12V. How r relates to the actual size of GFP-tH, mRFP-tH, H-rasG12V and K-rasG12V clusters is tackled later in this paper.

Variable parameters for both models.

  1. The number of raft proteins in the study area A, λ.
  2. The fraction of proteins that are localized to rafts/clusters, [var phi].
  3. The fraction of proteins that are randomly distributed over the study area, 1 - [var phi].

The classical raft model algorithm.

  1. Calculates the number of rafts (Rn) to be generated and randomly distributes Rn raft centers over the study area: Rn = μAAr2, where μA is the fractional area of A designated lipid raft (μA is a variable parameter for this model).
  2. Randomly selects a cluster center and places a protein at a random distance in the range (0, r) and in a random direction (0,360°) until [var phi]λ proteins have been allocated.
  3. Places (1 - [var phi])λ proteins randomly over the study area.
  4. Calculates the K function of the resulting point pattern.
  5. Repeats steps 1-5 for 20 or 100 simulations.
  6. Calculates a mean K function [K(r)] from the simulations.
  7. Derives L(r) - r and identifies Lmax.
  8. Derives a 99% confidence interval for a single observation.

As a summary statistic, we use the maximum observed value of the mean L(r) - r function (Lmax). Other summary statistics, such as ∫r-2K(r)dr or ∫K(r)dr, were also evaluated but yielded conclusions identical to Lmax (data not shown).

The alternative clustering algorithm.

  1. Calculates the number of nanoclusters (Rn) to be generated and randomly distributes Rn nanocluster centers over the study area: Rn = λ[var phi]C, where μC is mean number of proteins per cluster (μC is a variable parameter for this model).
  2. Randomly selects a cluster center and places a protein at a random distance in the range (0,r) and in a random direction (0,360°) until [var phi]λ proteins have been allocated.
  3. Places (1 - [var phi])λ proteins randomly over the study area.
  4. Calculates the K(r) of the resulting point pattern.
  5. Repeats steps 1-5 for 20 or 100 simulations.
  6. Derives L(r) - r and identifies Lmax.
  7. Derives a 99% confidence interval for a single observation.

Both models assume that gold particles have a diameter of 4 nm and, because of hardcore repulsion, no gold particle can be centered closer than 4 nm to a gold particle already on the pattern. If such a placement is attempted, it is rejected and a new randomization occurs. Extraction of contour lines for Lmax from (μA, [var phi], Lmax) or (μC, [var phi], Lmax) data sets was performed with gnuplot-4 in an X11 interface. Original data sets were reinterrogated to obtain a frequency distribution of cluster sizes at the r for Lmax. These observed frequencies were compared systematically with the predicted frequencies obtained under different realizations of the alternative model to obtain the best fit with the experimental data.

Results and Discussion

Mathematical Realizations of Different Plasma Membrane Raft Models. The classical lipid raft hypothesis postulates preexisting rafts on the plasma membrane. Proteins with lipid anchors able to partition into a liquid-ordered, cholesterol-enriched structure are sequestered or concentrated in these rafts. A generalization of this hypothesis suggests that proteins dynamically partition into and out of rafts such that only a fraction of proteins are clustered in the rafts. An alternative hypothesis is that raft proteins form dynamic, cholesterol-dependent nanoclusters, perhaps driven by a direct interaction of the lipid anchor with cholesterol and lipids in the plasma membrane. This model accounts for the surface distribution of GFP-glycosylphosphatidylinositol (GPI) in living cells imaged using homo-FRET (10). These alternative models are shown diagrammatically in Fig. 1. Although, these are only two possible models of many that could be constructed, they are worth studying because they capture many of the current concepts of cell-surface organization. These models intuitively suggest different outcomes if the expression of a raft protein is varied. In the classical raft model, upon increasing the level of expression of a raft marker, the number of raft proteins that must be accommodated in the preexisting rafts increases. Therefore, the number of proteins per raft also increases. Conversely, in the alternative model, the average number of proteins per cluster remains constant so that, as expression level increases, the number of clusters increases (Fig. 1). These differences should be reflected in the resulting equilibrium distributions of the raft proteins on the plasma membrane. The spatial distribution of plasma membrane proteins can be visualized by immunogold labeling of two-dimensional sheets of plasma membrane and the spatial structure of the labeling described by Ripley's K function. We therefore constructed mathematical models and used Monte Carlo simulations to determine the expected distributions of raft proteins that would be detected by immunogold labeling of plasma membrane sheets according to the two different hypotheses.

The raft model generates lipid rafts in a study area of 1 × 1 μm. The area of the plasma membrane occupied by rafts (μA), the fraction of raft proteins that occupy these lipid rafts ([var phi]), and the total number of raft proteins (λ) expressed in the study area were varied over wide ranges. Fig. 2 shows the results from 1,350 parameter combinations, with rafts occupying between 6% and 30% of the cell surface (0.06 ≤ μA ≤ 0.3) and accommodating between 10% and 90% of raft proteins (0.1 ≤ [var phi] ≤ 0.9) at three expression levels (λ = 250, 500, or 1,000). Lmax was used as a summary statistic of the K function: the greater the value of Lmax the greater the extent of clustering of the resulting pattern. An example of the generated point patterns is shown in Fig. 7, which is published as supporting information on the PNAS web site. The results of these simulations are shown in Fig. 2 A-C and can be used to extract relationships between Lmax and λ for any combination of μA and [var phi]. Three examples are shown in Fig. 2D, and all show linear relationships through the origin. The important overall conclusion to be drawn from these simulations is that for any set of combinations of μA and [var phi] in a classical raft model, Lmax is directly proportional to λ.

Fig. 2.
Monte Carlo simulations of the raft model described in Fig. 1. (A-C) All combinations of μA (in increments of 0.1) and [var phi] (in increments of 0.05) were simulated at three different densities (A, λ = 250 μm-2; B, λ ...

The alternative model generates nanoclusters in a study area of 1 × 1 μm. The mean number of proteins per cluster (μC), the fraction of raft proteins that occupy the nanoclusters ([var phi]), and the total number of raft proteins (λ) in the study area were again varied over wide ranges. An example of the generated point patterns is shown in Fig. 7. Fig. 3 shows the results from 1,350 parameter combinations for mean cluster size (1.75 ≤ μC ≤ 7.75), and clustered fraction (0.1 ≤ [var phi] ≤ 0.9) at three expression levels (λ = 250, 500, or 1,000). The data in Fig. 3 A-C can be used to extract the relationship between Lmax and λ for any combination of μC and [var phi]. Three examples are shown in Fig. 3D. In contrast to the raft model, the alternative model demonstrates very weak/no dependence of Lmax on λ for any set of combinations of μC and [var phi].

Fig. 3.
Monte Carlo simulations of the alternative clustering model described in Fig. 1. (A-C) All combinations of μC (in increments of 0.1) and [var phi] (in increments of 0.05) were simulated at three different densities (A, λ = 250 μm ...

Observed EM Immunogold Point Patterns Are Incompatible with a Classical Raft Model. We evaluated these models by analyzing a large number (n = 135) of immunogold-labeled plasma membrane sheets from cells expressing different levels of raft protein GFP-tH or mRFP-tH. The scatter plot of the experimental data (Fig. 4A) shows that Lmax is unrelated to the expression level (λ), suggesting that a classical raft model cannot explain the observed labeling patterns. To further explore this finding multiple examples from the parameter sets in Fig. 2 were simulated over the full λ range of the experimental data; two are shown in Fig. 4B. The results are plotted as confidence intervals that would be expected to capture 99% of the observed data if the underlying hypothesis were valid. One parameter set captures some of the data points at higher expression levels, and the other set captures some of the data at lower expression levels, but neither is a good fit for the experimental data of the complete λ range. Taking these results together with those in Fig. 2, we can infer that no realization of a classical raft model can account for the observed point patterns. In contrast, it is straightforward to identify a parameter set from the alternative model that captures close to all of the experimental data. For example, the parameter set (μC, [var phi]) = (2.5, 0.45) simulated over the full λ range of the experimental data is shown in Fig. 4C, together with a parameter set that clearly does not fit the data.

Fig. 4.
Surface distributions of GFP-tH and RFP-tH fit a realization of the alternative clustering model but no realization of the classical raft model. (A) Measurements of Lmax from 135 individual plasma membrane sheets prepared from BHK cells expressing different ...

At this point in the analysis, we need to consider whether a unique solution of (μC, [var phi]) for the experimental data can be identified. This issue is important because there is a range of solutions that approximately follow the Lmax = 2.5 isoline in Fig. 3B and, when simulated as in Fig. 4, yield similar confidence intervals. The GFP-tH and mRFP-tH point patterns were therefore reinterrogated to collect cluster size frequency data from which mean cluster size could be estimated. This analysis gives a value of μC = 2.51 ± 0.15 (mean ± SEM). Fitting a Poisson distribution to this result reveals that ≈9% of the “clusters” generated by the model comprise single gold particles; thus, the corrected value of [var phi] is 0.36 for the (μC, [var phi]) = (2.5, 0.45) solution. Therefore, we conclude that the distributions of GFP-tH and RFP-tH on the inner surface of the plasma membrane are fully accounted for if 36% of the gold-labeled proteins are present in cholesterol-dependent nanoclusters comprising small numbers of proteins, with the remaining 64% randomly arrayed over the plasma membrane. The extent of clustering and the number of gold-labeled proteins per cluster are independent of the density of the pattern. One implication of this result is that the area of the plasma membrane occupied by GFP-tH nanoclusters will increase with expression level. We reported previously that at very high levels of expression GFP-tH domains could occupy a significant area of the plasma membrane (13), although these earlier estimates were based on the assumptions inherent in a simplified classical raft model.

The experimental and modeled data consider the number of gold-labeled proteins per cluster. To convert this value into actual numbers of proteins, we measured the capture ratio of the anti-GFP 4-nm gold antibody in the EM protocol. Across multiple experiments, the capture ratio was 42% and was independent of GFP pattern density (Fig 9, which is published as supporting information on the PNAS web site). This gives an estimate of 5.9 for the mean number of proteins in a GFP-tH cluster. Together, these results are strikingly similar to a recent analysis of GFP-GPI anchored proteins (GFP-GPI-AP) in live cells using homo-FRET, which reported that 20-40% of GFP-GPI-AP exist in clusters of a similar size on the outer leaflet of the plasma membrane (10). The fraction of monomeric GFP-GPI-AP did not change with expression level, and exactly the same behavior is seen here with GFP-tH. Together, these results suggest that the mechanism(s) controlling segregation of proteins into cholesterol-sensitive clusters on the inner and outer leaflet of the plasma membrane may be analogous. The fixed monomer/cluster ratio observed here and by Sharma et al. (10) is interesting because it seemingly violates the law of mass action, which predicts that the fraction of clustered proteins should increase with increasing expression level. There are several possible explanations for this phenomenon. For example, the law describes a system that is homogenous and at equilibrium, conditions unlikely to apply to proteins diffusing laterally in the plasma membrane. In addition, preferential removal of clustered proteins is possible because of intrinsic instability of the nanoclusters, selective internalization, or some other energy-dependent biophysical mechanism that actively regulates the extent of clustering.

Calibrating the Size of Inner-Leaflet Microdomains. The average distance r for Lmax of the GFP-tH immunogold point patterns is 22 nm (13). However, the antigen (GFP) being labeled is spaced from the gold particle by the length of the antibody molecule in three dimensions. Thus, the gold could lie directly over or be at some distance from the antigen (Fig. 5A) (21), meaning that an accurate assessment of the actual size of the nanocluster is hard to infer. To tackle this problem, we generated a model system with GFP in a defined domain that could be labeled with the same gold probes used for the plasma membrane sheet analysis. Gold particles 14 nm in diameter were coated with recombinant GFP and applied to an EM grid. The grids were labeled with small (4 nm) anti-GFP gold and imaged exactly as for plasma membrane sheet analysis (Fig. 5B). The distribution of the 4-nm gold particles associated with a large number of such large GFP-gold particles was analyzed by using the K function and the size of the large GFP-gold particles measured directly. We repeated this analysis for a range of GFP-gold domain sizes (diameters of 10-22 nm) to obtain a calibration plot that relates GFP domain radius to the radius of the small gold pattern returned from the K function analysis (Fig. 5C). If we assume that a similar spacing of gold particle from antigen occurs when GFP is arrayed on plasma membrane sheets, Fig. 5C suggests that the 20- to 22-nm-radius gold clusters detected by immunoEM for GFP-tH are, in fact, representative of a protein cluster with a radius of ≈10-12 nm. However, for technical reasons we have used a spherical model domain to array GFP rather than the planar domains that will be present on the plasma membrane, so it is probably safer to consider the calibrated estimates of domain size obtained from Fig. 5C as a possible lower bound rather than a definitive size.

Fig. 5.
Analysis of a model system to estimate domain size. (A) Schematic representation of the model system showing a 14-nm gold particle coated with GFP and labeled with anti-GFP 4-nm gold. (B) Examples of 14-nm GFP-coated gold particles labeled with anti-GFP ...

The lower-bound estimate from EM for the diameter of cholesterol-dependent nanoclusters occupied by GFP-tH of ≈20 nm is larger than the ≈5-nm estimate for GFP-GPI-AP (10). This discrepancy could be related to the greater number of GFP-tH proteins per cluster compared with GFP-GPI-AP, or the technical limitation of FRET to report only on length scales ≤5 nm compared to EM, which reports on all length scales ≤200 nm.

Ras Proteins Exhibit Nanoclustering. To examine whether other inner plasma membrane proteins exhibit nanoclustering, we carried out identical analyses to those described in Figs. Figs.1,1, ,2,2, ,3,3, ,44 for large data sets of GFP-H-rasG12V (n = 64) and GFP-K-rasG12V (n = 52). In both cases the point patterns of the experimental data sets fit the alternative model shown in Fig. 1. Moreover, the distributions of H-rasG12V and K-rasG12V had similar characteristics to GFP-tH. The cluster/monomer ratio of both Ras proteins was stable over a wide range of expression: 44% of K-rasG12V and 40% of H-rasG12V was clustered. The mean number of gold particles per K-ras G12V cluster was 3.25, corresponding to ≈7.7 Ras proteins, and the mean number of gold particles per H-rasG12V cluster was 2.5, corresponding to ≈6 Ras proteins. The average distance r for Lmax of the GFP-H-rasG12V and K-rasG12V gold clusters is 16 nm (13), so applying the relationship in Fig. 5 gives a lower bound for the radius of these protein clusters of 6 nm (Fig. 5 C and D). The smaller diameter of the H- and K-ras clusters implies that the density of H-rasG12V and K-rasG12V proteins in their respective nanoclusters is substantially greater (1.9-fold and 2.45-fold, respectively) than the raft protein GFP-tH in cholesterol-dependent nanoclusters. This increased density probably reflects the different mechanisms driving Ras cluster formation and may be required for the K-rasG12V and H-rasG12V clusters to operate as signaling platforms.

EM is a powerful tool for the analysis of inner plasma membrane, but it only provides a snapshot of protein distributions with no time dimension. It is important, therefore, to compare the high spatial resolution data from EM with time-resolved, lower spatial resolution data sets. Murakoshi et al. (22), using single-fluorophore tracking microscopy to visualize Ras proteins on the inner leaflet of the plasma membrane reported that K-rasG12V and H-rasG12V molecules exhibit alternating immobile (≈30%) and mobile (≈70%) periods, values that are similar to the proportion of clustered and random proteins that we observe. We propose, therefore, that the K-rasG12V and H-rasG12V nanoclusters detected by EM are the sites where Ras proteins are transiently immobile and generate their signal output. Importantly, through the application of the computational modeling developed here to interpret immunoEM point patterns and the size and capture ratio calibrations of the EM imaging, we are able to accurately predict the size and number of Ras proteins in nanoclusters. In addition, because Ras proteins do not partition into preexisting domains, we propose that nanoclustering of activated Ras proteins facilitates assembly of a transient plasma membrane microenvironment that is required for the scaffolding of signal transduction complexes.

Role of the Actin Cytoskeleton in Maintaining Different Plasma Membrane Nanoclusters. Finally, we examined whether an intact actin cytoskeleton is required to generate inner leaflet protein clusters. Cells expressing GFP-tH were treated with latrunculin for 5 min immediately before plasma membrane sheets were prepared for EM analysis. Fig. 6 shows that 50 nM and 1 μM latrunculin significantly reduced and completely abolished clustering of GFP-tH, respectively. There was no loss of GFP-tH from the plasma membrane because the overall labeling density was unchanged by latrunculin treatment. This result suggests that formation of cholesterol-dependent clusters on the inner leaflet of the plasma membrane requires an intact cytoskeleton and could imply that actin has affinity for, or can stabilize, liquid-ordered domains. In this context, it is worth noting that depleting cells of cholesterol has been shown to perturb the actin cytoskeleton (6), a reciprocal effect to that described here.

Fig. 6.
Actin dependence of plasma membrane nanoclusters. (A-D) Plasma membrane sheets were prepared from cells expressing GFP-tH, H-rasG12V, K-rasG12V, or H-ras that were either untreated or treated for 5 min with 50 nM or 1 μM latrunculin. Sheets were ...

In contrast to GFP-tH, the clustering of GFP-H-rasG12V was unaffected by 50 nM or 1 μM latrunculin. Disassembling actin had an intermediate effect on GFP-K-rasG12V: 50 nM latrunculin significantly reduced but did not abolish clustering of GFP-K-rasG12V, and no further effect was seen with 1 μM latrunculin (Fig. 6). The significance of these effects on clustering for Ras signal output was assayed by measuring activation of the Raf/mitogen-activated protein kinase kinase/Erk cascade. Latrunculin treatment significantly reduced Erk activation by K-rasG12V but had no effect on Erk activation by H-rasG12V (Fig. 6). This result is further evidence that the formation of K-Ras nanoclusters is critical for signal transduction. Previous evidence has come almost exclusively from studies with H-ras. Galectin-1 knockdown mutations within the H-Ras membrane anchor or flanking sequences selectively abolish H-Ras but not K-ras clustering with a corresponding selective abrogation of H-ras signal output (11-13). In addition, changing the stoichiometry of H-ras palmitoylation drives H-rasG12V into aberrant cholesterol-dependent nanoclusters rather than cholesterol-independent nanoclusters and abrogates signal output (23).

The differential effect of latrunculin on H-rasG12V and GFP-tH nanoclusters clusters offers a test of the hypothesis that wild-type H-ras is in a GTP-regulated equilibrium between these two types of domain (12). Fig. 6 shows two effects of latrunculin on GFP-H-ras clustering: first, a significant reduction of clustering and, second, a decrease in the r value for Lmax from 22 nm, which is indicative of GFP-tH clusters, to 16 nm, which is indicative of GFP-H-rasG12V clusters. These effects are exactly what would be expected if nanoclustering of H-ras in cholesterol-dependent structures is prevented by latrunculin, as observed with GFP-tH, whereas nanoclustering in spatially distinct and activated H-ras domains is unaffected, as observed with GFP-H-rasG12V.

In conclusion, our findings do not support the existence of stable preexisting rafts but suggest that lipidated proteins on the inner plasma membrane are able to drive the formation of nanoclusters. Ras nanoclustering is likely driven by the complex biophysics of lipid-anchor-plasma-membrane-lipid interactions and transiently stabilized by protein-protein or protein-lipid interactions to form signaling platforms (23). These concepts represent converging models of organization of the inner and outer leaflet of the plasma membrane.

Supplementary Material

Supporting Figures:

Acknowledgments

The Institute for Molecular Bioscience at the University of Queensland is a Special Research Center of the Australian Research Council. This work was supported by National Institutes of Health Grant GM-066717.

Notes

Author contributions: S.J.P., R.G.P., and J.F.H. designed research; S.J.P., C.M., R.G.P., and J.F.H. performed research; S.J.P., C.M., R.G.P., and J.F.H. analyzed data; and S.J.P., R.G.P., and J.F.H. wrote the paper.

This paper was submitted directly (Track II) to the PNAS office.

Abbreviations: RFP, red fluorescent protein; mRFP, monomeric RFP; ERK, extracellular signal-regulated kinase; GPI, glycosylphosphatidylinositol; AP, anchored protein; tH, plasma membrane-targeting domain of H-ras.

References

1. Simons, K. & Toomre, D. (2000) Nat. Rev. Mol. Cell Biol. 1, 31-39. [PubMed]
2. Munro, S. (2003) Cell 115, 377-388. [PubMed]
3. Edidin, M. (2003) Annu. Rev. Biophys. Biomol. Struct. 32, 257-283. [PubMed]
4. Simons, K. & Ikonen, E. (1997) Nature 387, 569-572. [PubMed]
5. Heerklotz, H. (2002) Biophys. J. 83, 2693-2701. [PMC free article] [PubMed]
6. Kwik, J., Boyle, S., Fooksman, D., Margolis, L., Sheetz, M. P. & Edidin, M. (2003) Proc. Natl. Acad. Sci. USA 100, 13964-13969. [PMC free article] [PubMed]
7. Pralle, A., Keller, P., Florin, E. L., Simons, K. & Horber, J. K. (2000) J. Cell Biol. 148, 997-1008. [PMC free article] [PubMed]
8. Anderson, R. G. & Jacobson, K. (2002) Science 296, 1821-1825. [PubMed]
9. Varma, R. & Mayor, S. (1998) Nature 394, 798-801. [PubMed]
10. Sharma, P., Varma, R., Sarasij, R. C., Ira, Gousset, K., Krishnamoorthy, G., Rao, M. & Mayor, S. (2004) Cell 116, 577-589. [PubMed]
11. Hancock, J. F. (2003) Nat. Rev. Mol. Cell Biol. 4, 373-384. [PubMed]
12. Prior, I. A., Harding, A., Yan, J., Sluimer, J., Parton, R. G. & Hancock, J. F. (2001) Nat. Cell Biol. 3, 368-375. [PubMed]
13. Prior, I. A., Muncke, C., Parton, R. G. & Hancock, J. F. (2003) J. Cell Biol. 160, 165-170. [PMC free article] [PubMed]
14. Niv, H., Gutman, O., Kloog, Y. & Henis, Y. I. (2002) J. Cell Biol. 157, 865-872. [PMC free article] [PubMed]
15. Zacharias, D. A., Violin, J. D., Newton, A. C. & Tsien, R. Y. (2002) Science 296, 913-916. [PubMed]
16. Roy, S., Luetterforst, R., Harding, A., Apolloni, A., Etheridge, M., Stang, E., Rolls, B., Hancock, J. F. & Parton, R. G. (1999) Nat. Cell Biol. 1, 98-105. [PubMed]
17. Campbell, R. E., Tour, O., Palmer, A. E., Steinbach, P. A., Baird, G. S., Zacharias, D. A. & Tsien, R. Y. (2002) Proc. Natl. Acad. Sci. USA 99, 7877-7882. [PMC free article] [PubMed]
18. Besag, J. E. (1977) J. R. Stat. Soc. B 39, 193-195.
19. Ripley, B. D. (1977) J. R. Stat. Soc. B 39, 172-192.
20. Diggle, P. J., Mateu, J. & Clough, H. E. (2000) Adv. Appl. Probab. 32, 331-343.
21. Parton, R. G. & Hancock, J. F. (2004) Trends Cell Biol. 14, 141-147. [PubMed]
22. Murakoshi, H., Iino, R., Kobayashi, T., Fujiwara, T., Ohshima, C., Yoshimura, A. & Kusumi, A. (2004) Proc. Natl. Acad. Sci. USA 101, 7317-7322. [PMC free article] [PubMed]
23. Hancock, J.F. & Parton, R. G. (2005) Biochem. J. 389, 1-11. [PMC free article] [PubMed]

Articles from Proceedings of the National Academy of Sciences of the United States of America are provided here courtesy of National Academy of Sciences
PubReader format: click here to try

Formats:

Related citations in PubMed

See reviews...See all...

Cited by other articles in PMC

See all...

Links

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...