Nonsteroidal Anti-inflammatory Drugs Alter the Spatiotemporal Organization of Ras Proteins on the Plasma Membrane

Background: Ras proteins form structurally distinct nanoclusters on cell plasma membrane that are critical for signal transduction.

Results: Amphiphilic anti-inflammatory agents stabilize cholesterol-dependent Ras nanoclusters, perturb GTP-dependent lateral segregation, and compromise nanocluster separation.

Conclusion: Biological amphiphiles have wide-ranging effects on plasma membrane heterogeneity and protein nanoclustering.

Significance: This is a novel mechanism for altering membrane protein segregation that has important consequences for cell signaling.

Abstract

Ras proteins on the inner leaflet of the plasma membrane signal from transient nanoscale proteolipid assemblies called nanoclusters. Interactions between the Ras lipid anchors and plasma membrane phospholipids, cholesterol, and actin cytoskeleton contribute to the formation, stability, and dynamics of Ras nanoclusters. Many small biological molecules are amphiphilic and capable of intercalating into membranes and altering lipid immiscibility. In this study we systematically examined whether amphiphiles such as indomethacin influence Ras protein nanoclustering in intact plasma membrane. We found that indomethacin, a nonsteroidal anti-inflammatory drug, induced profound and complex effects on Ras spatial organization, all likely related to liquid-ordered domain stabilization. Indomethacin enhanced the clustering of H-Ras.GDP and N-Ras.GTP in cholesterol-dependent nanoclusters. Indomethacin also abrogated efficient GTP-dependent lateral segregation of H- and N-Ras between cholesterol-dependent and cholesterol-independent clusters, resulting in mixed heterotypic clusters of Ras proteins that normally are separated spatially. These heterotypic Ras nanoclusters showed impaired Raf recruitment and kinase activation resulting in significantly compromised MAPK signaling. All of the amphiphilic anti-inflammatory agents we tested had similar effects on Ras nanoclustering and signaling. The potency of these effects correlated with the membrane partition coefficients of the individual agents and was independent of COX inhibition. This study shows that biological amphiphiles have wide-ranging effects on plasma membrane heterogeneity and protein nanoclustering, revealing a novel mechanism of drug action that has important consequences for cell signaling.

Introduction

The small GTPase Ras regulates cell differentiation, proliferation, and survival (1, 2). The three major Ras isoforms that are ubiquitously expressed in mammalian cells, H-, K-, and N-Ras, contain a nearly identical G-domain, which binds guanine nucleotides and interacts with downstream effectors, but have significantly divergent C-terminal hypervariable regions (HVR).² H-, N-, and K-Ras are anchored to the inner surface of the plasma membrane by a C-terminal S-farnesyl cysteine carboxymethyl ester that operates in concert with a second signal in the HVR (3). The second signal is one or two palmitoyl chains in N- and H-Ras, respectively, and a polybasic domain in K-Ras. On the plasma membrane, H-, K-, and N-Ras proteins assemble into spatially distinct, highly dynamic nanoclusters with lifetimes of <1 s and radii of <14 nm that contain ∼7 Ras proteins (4, 5). Lateral segregation is also regulated by Ras activation, because GTP-loaded H-, N-, and K-Ras proteins form nanoclusters that are spatially and structurally distinct from those formed by the cognate GDP-loaded proteins (5–7). The formation of nanoclusters by Ras proteins is crucial to their signaling function (8–10). The molecular mechanisms that drive nanoclustering involve a complex interplay of lipid-lipid and protein-lipid interactions between the membrane anchor, flanking protein sequences, and lipids and proteins of the plasma membrane (4–6, 8, 11–13). Given the critical role of plasma membrane lipids in nanocluster formation, any perturbation of the phospholipid membrane could potentially affect the spatiotemporal dynamics of Ras nanoclustering and signaling. Many small biological molecules, such as nonsteroidal anti-inflammatory drugs (NSAIDs) and bile salts, are amphiphilic and are capable of strong association with phospholipid bilayers. Indeed, our previous studies (14, 15) demonstrated that a potent NSAID, indomethacin, markedly alters the heterogeneity of synthetic phospholipid membranes. In the current study, we therefore systematically examined the ability indomethacin and other amphiphiles to perturb plasma membrane nanostructure using multiple Ras isoforms as spatiotemporal probes.

EXPERIMENTAL PROCEDURES

Materials

Indomethacin, aspirin, salicylate, ibuprofen, and naproxen were purchased from Sigma-Aldrich and prepared as described previously (14, 15).

Fluorescence Lifetime Imaging Microscopy (FLIM)

Expression vectors for mGFP- and mRFP-linked tH, H-Ras, H-RasG12V, tK, K-Ras, K-RasG12V, N-Ras, and N-RasG12V have been described (8). Baby hamster kidney (BHK) cells cultured in DMEM containing 10% bovine calf serum (BCS) were transfected using Lipofectamine according to the manufacturer's instructions. Twenty-four hours after transfection, cells were treated with various concentrations of NSAIDs for 1 h in DMEM and then washed and fixed with 4% paraformaldehyde (PFA). The GFP fluorescence lifetime was measured using a Lambert Instruments (Roden, The Netherlands) FLIM attachment on a wide field Nikon microscope. BHK cells expressing combinations of mGFP- and mRFP-linked constructs were excited using a sinusoidally modulated 3-Watt 497-nm light-emitting diode (LED) at 40 MHz under epi-illumination. Fluorescein at 1 μm was used as a lifetime reference standard. Cells were imaged with a ×60 Plan-Apo/1.4 NA oil objective using an appropriate GFP filter set. The phase and modulation lifetimes were determined from a set of 12 phase settings using the manufacturer's software. Three independent experiments were performed for each treatment, and data from at least 60 cells were pooled to calculate the mean and S.E. Tests for significant differences from control were performed by one-way ANOVA. Fluorescence lifetime FRET measurements have certain advantages over intensity-based methods, because fluorescence lifetime is an intrinsic property of the fluorescent molecule that is not influenced by factors such as excitation source, detection gain, optical loss, and variation in fluorophore concentration and is much less sensitive to variation in the donor-acceptor ratio (assuming an excess of acceptor) (16).

Immuno-electron Microscopy (EM) and Spatial Analysis

Immuno-EM and spatial analysis were conducted exactly as described previously (5, 7). In brief, BHK cells were cultured on glass coverslips to ∼80% confluency and treated with 75–100 μm indomethacin for 1 h. Intact apical cell plasma membrane sheets were attached to EM grids, washed, and fixed in 4% PFA, 0.1% glutaraldehyde. Grids were incubated with anti-GFP antisera coupled directly to 4.5-nm gold, washed, stained in uranyl acetate, and embedded in methylcellulose. Digital images of the immunogold-labeled plasma membrane sheets were taken at ×100,000 magnification in an electron microscope. Intact 1-μm² areas of the plasma membrane sheet were identified, and the x and y coordinates of each gold particle were determined using ImageJ software. The K-functions and transformations (5, 7, 17) were then calculated according to Equations 1 and 2,

where K(r) is the univariate K-function for a pattern of n points in an area A; r = radius at which K(r) is calculated (here we evaluated 1 < r < 240 nm at 1-nm increments); ‖·‖ is Euclidean distance; 1(·) is the indicator function (this takes a value of 1 if ‖x_i − x_j‖ ≤ r and 0 otherwise); and w_ij⁻¹ is the proportion of the circumference of the circle with center x_i and radius ‖x_i − x_j‖ contained within A (this term builds in an unbiased edge correction for points at the edge of the study area). L(r) − r is a linear transformation of K(r), and we standardized L(r) − r on the 99% confidence interval estimated from Monte Carlo simulations. Under the null hypothesis of complete spatial randomness L(r) − r has an expected value of 0 for all values of r (18, 19) and thus can be used to quantify the extent of clustering of gold particles. An L(r) − r value greater than the 99% confidence interval for a random pattern indicates statistically significant clustering. The magnitude of the peak L(r) − r value (L_max) correlates directly with the extent of clustering, and the value of r at L_max correlates with the radius of the clusters. Bootstrap tests can be used to examine for statistical differences between replicated point patterns. These bootstrap tests were constructed exactly as described (20), and statistical significance was evaluated against 1000 bootstrap samples.

A related analysis, the bivariate K-function, was used to quantify the extent of co-clustering of two gold populations. Plasma membrane sheets prepared from cells co-expressing a GFP- and RFP-tagged protein were double labeled with anti-GFP directly coupled to 2-nm gold and anti-RFP directly coupled to 6-nm gold. EM images of the plasma membrane sheets were digitized as described above, and the x and y coordinates of the small (2 nm) and large (6 nm) gold particles were determined using ImageJ software. Bivariate K-functions and transformations (5, 7, 17) were then calculated according to Equations 3–6.

In this set of calculations, where an area A contains n_b 6-nm gold particles (b = big gold) and n_s 2-nm small gold particles (s = small gold), K_bs(r) is the bivariate K-function that maps the distribution of the big gold particles with respect to each small gold particle, whereas K_sb(r) maps the distribution of the small gold particles with respect to each big gold particle. Other notations are as shown in Equations 1 and 2. K_bs(r) and K_sb(r) are combined into a single bivariate estimator, K_biv(r). L_biv(r) − r is a linear transformation of K_biv(r), and we standardized L_biv(r) − r on the 95% confidence interval estimated from Monte Carlo simulations. Under the null hypothesis that there is no spatial interaction between the two populations of gold particles L_biv(r) − r has an expected value of 0 for all values of r and can therefore be used to quantify the extent of co-clustering of the two sets of gold particles. A value of L_biv(r) − r greater than the estimated 95% confidence interval indicates statistically significant co-clustering of the two gold particle populations at that value of r.

Western Blot-MAPK Pathway and Raf Kinase Activity

BHK cells were grown in serum-free DMEM for 2 h and then treated for 1 h with various NSAIDs dissolved in serum-free DMEM. Western blotting was used to quantify the level of pERK (6, 21). To measure Raf kinase activity, BHK cells stably expressing either constitutively active H-Ras.G12V or K-Ras.G12V were grown to 85–90% confluency and harvested. Membrane (P100) and cytoplasmic fractions (S100) were prepared by hypotonic lysis and high speed centrifugation, and Raf kinase activity was measured in P100 fractions in a coupled MEK/ERK kinase assay (21).

RESULTS

Indomethacin Enhances Cholesterol-dependent Nanoclusters in BHK Cell Plasma Membrane

We reported previously that indomethacin at relatively high concentrations enhances liquid-ordered, (L_o) domain formation in model membranes and also enhances the nanoclustering of GFP-tH (GFP targeted to the plasma membrane by the minimal membrane anchor of H-Ras) in intact cells (15). Here, we systematically examined the ability of indomethacin to alter nanocluster formation by full-length Ras proteins. GFP-tH nanoclusters are cholesterol-dependent, as are nanoclusters of GDP-loaded full-length H-Ras (H-Ras.GDP) and GTP-bound N-Ras (N-Ras.GTP) (7, 22). We first used EM spatial mapping to explore whether therapeutically relevant concentrations (23, 24) of indomethacin have similar effects on the organization of these three types of Ras nanocluster. Spatial point pattern analysis of immunogold-labeled plasma membrane sheets prepared from BHK cells expressing GFP-tH, GFP-H-Ras (H-Ras.GDP), and oncogenic mutant GFP-N-RasG12V (N-Ras.GTP) all showed a marked increase in nanoclustering after 1 h of treatment with 75–100 μm indomethacin as measured by Ripley's K-function (Fig. 1, a–c). Indomethacin induced an increase in the peak L(r) − r value (L_max) and also the value of r at which L_max occurs (r_max). Because r_max and L_max are correlated with the mean radius of the clusters and the mean number of proteins per cluster, respectively (5), we concluded that indomethacin increases the size and extent of nanoclustering of GFP-tH, H-Ras.GDP, and N-Ras.GTP proteins in their respective nanoclusters. This result is consistent with a general effect of the indomethacin on enhancing L_o domain stability in biological membranes. To complement the EM studies, we examined molecular interactions within nanoclusters using fluorescence lifetime imaging microscopy combined with fluorescence resonance energy transfer (FLIM-FRET). BHK cells were transiently transfected with GFP/RFP FRET pairs, and GFP lifetime was measured. FRET between GFP donor and RFP acceptor molecules within nanoclusters results in a reduction in donor lifetime (6, 15). We found that GFP-tH expressed alone in BHK cells had a fluorescence lifetime of ∼2.1 ns, which was subsequently decreased to ∼1.85 ns when BHK cells co-expressed both GFP-tH and RFP-tH (Fig. 1d). GFP-tH fluorescence lifetime in cells that co-expressed both GFP-tH and RFP-tH was further decreased to ∼1.75 ns when exposed to indomethacin at concentrations of 75–100 μm (Fig. 1d). GFP fluorescence lifetime was not affected by changes in the fluorophore environment, as the GFP fluorescence lifetime in cells expressing GFP-tH alone was not altered in the presence of indomethacin (data not shown). A similarly enhanced reduction in GFP lifetime on indomethacin treatment was also observed in BHK cells co-expressing GFP- and RFP-H-Ras or GFP- and RFP-N-RasG12V (Fig. 1d). These data are concordant with the EM experiments in suggesting a generalized effect of indomethacin on enhancing the formation or stability of cholesterol-dependent nanoclusters in the plasma membrane.

FIGURE 1.

Indomethacin enhances cholesterol-dependent nanoclustering. Plasma membrane sheets were generated from BHK cells expressing GFP-tH (a), GFP-H-RasWT (b), or GFP-N-RasG12V (c). Cells were left untreated or treated with 75–100 μm indomethacin for 1 h. Plasma membrane sheets were labeled with anti-GFP antibody conjugated to 4.5-nm gold. The plasma membrane sheets were imaged in an electron microscope, and the spatial distribution of the gold labeling was analyzed using Ripley's K-function (representative EM images are shown in supplemental material Fig. S1). Maximum L(r) − r values above the 99% CI for complete spatial randomness indicate clustering at that value of r. Univariate K-functions are weighted means (n ≥ 8) standardized on the 99% CI for a random pattern. Significant differences from the control pattern (Con) for indomethacin-treated cells were assessed using bootstrap tests. Treatment with indomethacin (Indo) significantly enhanced nanocluster formation by GFP-tH (p = 0.001), GFP-H-RasWT (p = 0.001), and GFP-N-RasG12V (p = 0.001). d, the fluorescence lifetime of GFP in BHK cells expressing GFP-tH, GFP-H-RasWT, or GFP-N-RasG12V alone (GFP-Ras) or co-expressing the cognate FRET acceptor RFP-tH, RFP-H-RasWT, or RFP-N-RasG12V in the absence (GFP-RFP) or presence of 75 and 100 μm indomethacin for 1 h was measured in a wide field FLIM microscope. Data were collected from multiple single cells. GFP fluorescence lifetime is shown as the mean ± S.E. (n ≥ 60 cells from three independent experiments). Differences between control, untreated cells expressing the GFP/RFP FRET pair and cells treated with indomethacin were examined by one-way ANOVA. Significant differences are shown (*, p ≪ 0.05). e, examples of pseudo-colored FLIM images of BHK cells with or without indomethacin treatment are shown: BHK cells expressing GFP-H-rasWT alone (i); BHK cells expressing the FRET pair GFP- and RFP-H-rasWT with no treatment (ii) or in the presence of 75 μm indomethacin (iii) or 100 μm indomethacin (iv).

Indomethacin Has Selective Effects on Cholesterol-independent Clusters

We next tested the effect of indomethacin on the organization of Ras isoforms and mutants that form cholesterol-independent nanoclusters. EM spatial mapping showed that indomethacin had no effect on the extent of clustering of oncogenic mutant GFP-K-RasG12V (K-Ras.GTP), which forms cholesterol-independent nanoclusters, or on GFP-tK (GFP targeted to the plasma membrane by the minimal membrane anchor of K-Ras), which co-localizes with GFP-K-RasG12V (Fig. 2, a and b). Indomethacin also had no effect on the clustering of GFP-K-RasWT (K-Ras.GDP) (data not shown). Similarly FLIM-FRET analysis of GFP/RFP pairs of the same proteins showed that indomethacin did not further reduce GFP fluorescence lifetime (Fig. 2e), indicating no detectable effect of indomethacin on molecular interactions among GFP-tK, K-Ras.GDP, or active K-Ras.GTP in their cognate cholesterol-independent nanoclusters.

FIGURE 2.

Indomethacin has variable effects on cholesterol-independent nanoclusters. EM spatial mapping was performed as described in the legend for Fig. 1 using BHK cells expressing GFP-tK (a), GFP-K-RasG12V (b), GFP-H-RasG12V (c), or GFP-N-RasWT (d). Univariate K-functions are weighted means (n ≥ 7) standardized on the 99% confidence interval for a random pattern. Significant differences were assessed using bootstrap tests. Exposure to indomethacin (Indo) at 75 μm for 1 h had no effect on the clustering of GFP-tK and GFP-K-RasG12V (p = 1 in both cases). In contrast, treatment with indomethacin at 75 μm for 1 h significantly decreased the nanoclustering of GFP-H-RasG12V (p = 0.001) and GFP-N-RasWT (p = 0.001). e, GFP fluorescence lifetime was measured in BHK cells expressing GFP-tK or GFP-K-RasG12V alone (GFP-Ras) or co-expressing the cognate FRET acceptor RFP-tH or RFP-K-RasG12V in the absence (GFP-RFP) and presence of 75 and 100 μm indomethacin. f, a similar FLIM experiment was performed using BHK cells expressing GFP-H-RasG12V or GFP-N-RasWT (GFP-Ras) or co-expressing the cognate FRET acceptor RFP-H-RasG12V or RFP-N-RasWT in the absence (GFP-RFP) or presence of 75 and 100 μm indomethacin. Data were collected from multiple single cells. GFP fluorescence lifetime is shown as mean ± S.E. (n ≥ 70 cells from three independent experiments). Differences between untreated and indomethacin-treated GFP/RFP-expressing cells were examined by one-way ANOVA. Con, control.

In contrast, H-RasG12V, the constitutively active and oncogenic form of H-Ras (H-Ras.GTP), displayed a different response to indomethacin. EM spatial analysis of GFP-H-RasG12V on intact plasma membrane sheets showed a reduction in L_max values when cells were treated with indomethacin (Fig. 2c), indicating a significant reduction of active H-Ras.GTP nanocluster formation in the plasma membrane. A concomitant reduction in r_max showed that indomethacin treatment also reduced the diameter of H-Ras.GTP nanoclusters. Inactive wild-type N-Ras.GDP also formed cholesterol-independent nanoclusters. EM spatial analysis of GFP-N-RasWT-expressing cells showed that indomethacin again significantly decreased L_max (Fig. 2d), similar to the effect observed on GFP-H-RasG12V. Paradoxically, FLIM-FRET imaging of cells expressing GFP/RFP FRET pairs of H-RasG12V or N-RasWT showed no increase in GFP fluorescence lifetime when cells were treated with indomethacin (Fig. 2f). Thus, although the EM analysis reported an indomethacin-induced reduction in nanoclustering, the FLIM-FRET analysis reported no overall change in the fraction of H-RasG12V or N-RasWT proteins within the FRET distance of each other in their respective nanoclusters. Taken together these data suggest that indomethacin induces a more complex reorganization of H-Ras.GTP and N-Ras.GDP nanoclusters than can be explained by a simple reduction in nanoclustering.

Indomethacin Induces Heterotypic H-Ras Nanocluster Formation

We showed previously that indomethacin significantly alters the lateral heterogeneity of model bilayer and cell plasma membranes (15). It is therefore conceivable that indomethacin induces an interaction between different nanoclusters that are normally spatially and structurally segregated. A commonality between H-Ras and N-Ras is that GTP/GDP loading regulates their lateral segregation between cholesterol-dependent and cholesterol-independent nanoclusters. We therefore tested whether indomethacin compromises this segregation by directly looking for heterotypic nanoclustering between GFP-tH (a surrogate marker of H-Ras.GDP) and H-RasG12V, which normally form laterally segregated cholesterol-dependent and cholesterol-independent nanoclusters, respectively. We first quantified the ability of indomethacin to drive heterotypic nanocluster formation between GFP-tH and RFP-H-RasG12V in the plasma membrane by immuno-EM. Plasma membrane sheets prepared from BHK cells co-expressing GFP-tH and RFP-H-RasG12V were labeled with 2-nm anti-GFP gold and 6-nm anti-RFP gold. The immuno-gold point patterns were evaluated using the bivariate K-function, L_biv(r) − r, which quantifies the extent of co-localization of the two populations of gold particles. For control cells, the L_biv(r) − r was within the 99% confidence interval (CI), indicating no detectable co-localization of GFP-tH and RFP-H-RasG12V (Fig. 3a). However, treatment with indomethacin caused a marked increase in L_biv(r) − r values above the 99% CI (Fig. 3a), suggesting that indomethacin significantly induced co-clustering between GFP-tH and RFP-H-RasG12V in the plasma membrane. Using FLIM-FRET imaging, we observed a slight decrease in GFP-tH fluorescence lifetime in control cells co-expressing RFP-H-RasG12V, indicating some degree of molecular proximity between GFP-tH and RFP-H-RasG12V molecules on the plasma membrane (Fig. 3, b and c), as reported previously (6). However, indomethacin induced a significant further decrease in GFP-tH fluorescence lifetime, indicating an enhanced interaction between GFP-tH and RFP-H-RasG12V and incorporation of both types of molecules within a single nanocluster, hence the formation of heterotypic nanoclusters.

FIGURE 3.

Indomethacin generates heterotypic nanoclusters containing GFP-tH and H-RasG12V. a, intact plasma membrane sheets from BHK cells expressing both GFP-tH and RFP-H-RasG12V were labeled with 2-nm anti-GFP gold and 6-nm anti-RFP gold. The gold patterns were analyzed using a bivariate K-function to assess the relative spatial relationship of the two gold populations. Bivariate K-functions (L_biv(r) − r) are weighted means (n ≥ 9) standardized on the 99% CI for a random pattern. L_biv(r) − r values above the 99% CI indicate significant co-localization of the two labeled proteins at that value of r. In untreated BHK cells the L_biv(r) − r curve does not leave the 99% CI (black curve). Exposure to indomethacin (Indo) at 75 and 100 μm for 1 h induced a significant increase in L_biv(r) − r values well above the 99% CI (blue and red curves). b, GFP fluorescence lifetime was measured in BHK cells expressing GFP-tH alone (GFP-tH) or co-expressing the FRET acceptor RFP-H-RasG12V in the absence (GFP-tH/RFP-HG12V) or presence of 75 and 100 μm indomethacin. Data were collected from multiple single cells. GFP fluorescence lifetime is shown as mean ± S.E. (n ≥ 188 cells from three independent experiments). Differences between untreated and indomethacin-treated GFP/RFP-expressing cells were examined by one-way ANOVA (*, p ≪ 0.05). c, examples of FLIM images of BHK cells with or without indomethacin treatment are shown: BHK cells expressing GFP-tH alone (i); BHK cells expressing GFP-tH and RFP-H-rasG12V with no treatment (ii) or incubated in the presence of 75 μm indomethacin (iii) or 100 μm indomethacin (iv). d, approximately 40% of H-Ras.GTP (H-RasG12V) forms cholesterol-independent clusters. Indomethacin enhances nanoclustering of GFP-tH (and H-Ras.GDP) by promoting assembly or inhibiting disassembly of GFP-tH and H-Ras.GDP nanoclusters. Indomethacin also decreases homotypic H-Ras.GTP clustering (the inhibition is shown as bars). In the presence of indomethacin there is missorting of H-Ras.GTP and GFP-tH molecules into common cholesterol-dependent heterotypic nanoclusters. Con, control.

A model that accounts for these results and the univariate EM and FLIM data of Fig. 2, c and e, is shown in Fig. 3d. We propose that H-Ras.GTP is localized predominantly to homotypic, cholesterol-independent clusters, with a population of freely diffusing monomeric H-Ras.GTP molecules. We further propose, based on the EM/FLIM data of Fig. 3, a and b, the existence of a small population of H-Ras.GTP in cholesterol-dependent nanoclusters containing GFP-tH molecules, called mixed, or heterotypic, nanoclusters. When cells are exposed to indomethacin, the drug stabilizes L_o domains, which enhances the formation or stability of both the GFP-tH homotypic clusters and GFP-tH/H-Ras.GTP heterotypic clusters. The stoichiometry of H-Ras.GTP in mixed nanoclusters must be lower than in H-Ras.GTP in homotypic nanoclusters, as heterotypic nanoclustering reduces the univariate L_max value (Fig. 2c), even though the overall fraction of GFP-/RFP-H-Ras.GTP molecules within FRET distance of each other in the membrane does not change (Fig. 2f). An identical model can account for the N-Ras.GDP data by substituting N-Ras.GDP for H-Ras.GTP (see Fig. 2d). In both cases, the effect of indomethacin on stabilizing the L_o domains has the consequence of impairing the GTP-dependent lateral segregation of H- and N-Ras, resulting in the formation of mixed clusters of proteins that are normally efficiently segregated to spatially separated clusters.

Indomethacin Drives Co-clustering of L_o Domains with K-Ras Nanoclusters

We next performed similar experiments to explore whether indomethacin would induce an interaction between GFP-tH and K-Ras nanoclusters. Plasma membrane sheets prepared from BHK cells co-expressing GFP-tH and RFP-K-RasG12V were labeled with anti-GFP 2-nm gold and anti-RFP 6-nm gold and imaged by EM. The bivariate K function, L_biv(r) − r, showed no evidence of co-clustering of GFP-tH and RFP-K-RasG12V under control conditions. Indomethacin at 75 and 100 μm, however, significantly increased L_biv(r) − r values (Fig. 4a). The r_max value for the indomethacin-treated L_biv(r) − r curve suggested a large domain radius (∼50–60 nm) for co-clustered GFP-tH and RFP-K-RasG12V proteins as compared with an r_max of ≤30 nm for homotypic nanoclusters of GFP-tH in the presence of indomethacin (Fig. 1a) and ≤20 nm for GFP-tH and RFP-K-RasG12V homotypic nanoclusters under control conditions. In addition, univariate analyses of the GFP-tH 2-nm gold pattern and RFP-K-RasG12V 6-nm gold pattern in the same experiment also returned r_max values up to 30 to 40 nm. Taken together these spatial mapping data suggest that indomethacin induced the co-association or aggregation of intact L_o nanoclusters (GFP-tH) with intact cholesterol-independent nanoclusters (RFP-K-RasG12V), with no mixing of the constituent proteins of the two types of nanoclusters. If this conclusion is correct and indomethacin is not driving the formation of homogeneously mixed nanoclusters between GFP-tH and RFP-K-RasG12V, then indomethacin treatment should not increase FRET between GFP-tH co-expressed with RFP-K-RasG12V. The FLIM-FRET data shown in Fig. 4, b and c confirm this prediction. Identical results were obtained in immuno-EM and FLIM-FRET experiments between GFP-tH and RFP-tK or GFP-tH and RFP-K-RasWT (data not shown). These results are summarized in the model shown in Fig. 4d. Note that in contrast to the indomethacin effect shown in Fig. 3, this effect of indomethacin is related to an induced failure of lateral separation of otherwise intact and normally assembled nanoclusters.

FIGURE 4.

Indomethacin induces aggregation of K-Ras and GFP-tH nanoclusters. a, plasma membrane sheets from BHK cells expressing GFP-tH and RFP-K-RasG12V were labeled with 2-nm anti-GFP gold and 6-nm anti-RFP gold particles. The gold patterns were analyzed using a bivariate K-function to assess the spatial relationship of the two gold populations as described in the legend for Fig. 3. Bivariate K-functions (L_biv(r) − r) are weighted means (n ≥ 12) standardized on the 99% CI for a random pattern. Note that the value of r at which L_biv(r) − r is maximal (r_max) is shifted significantly to a higher value in the presence of indomethacin, illustrating an increase in the size of the co-clusters. b, GFP fluorescence lifetime was measured in BHK cells expressing GFP-tH alone (GFP-tH) or co-expressing the FRET acceptor RFP-K-RasG12V in the absence (GFP-tH/RFP-KG12V) or presence of 100 μm indomethacin (Indo). Data were collected from multiple single cells. GFP fluorescence lifetime is shown as mean ± S.E. (n ≥ 94 cells from three independent experiments). Differences between untreated and indomethacin-treated GFP/RFP-expressing cells were examined by one-way ANOVA (*, p ≪ 0.05). c, examples of FLIM images: BHK cells expressing GFP-tH alone (i); BHK cells expressing the FRET pair GFP-tH and RFP-K-rasG12V with no treatment (ii) or in the presence of 100 μm indomethacin (iii). d, a model that accounts for the results shown in a and b. Indomethacin enhances nanoclustering of GFP-tH by promoting assembly or inhibiting disassembly of GFP-tH nanoclusters, as shown in Fig. 3. Approximately 40% of K-Ras.GTP (K-Ras.G12V) forms cholesterol-independent nanoclusters. Indomethacin has no effect on the actual formation of K-Ras.GTP nanoclusters; however, in the presence of indomethacin, K-Ras.GTP nanoclusters form immediately adjacent to cholesterol-dependent clusters. There is no actual mixing of the components of the cholesterol-dependent and cholesterol-independent nanoclusters. Con, control.

Indomethacin Inhibits Raf Activation in Ras-transformed Cells

We next explored whether the indomethacin-induced changes in Ras nanoclustering have significant effects on Ras signal transmission. Raf is activated exclusively in Ras.GTP nanoclusters on the plasma membrane (9, 10, 21), and therefore assays of Raf activation and Ras/Raf binding report directly on Ras nanocluster function. Fig. 5a shows that indomethacin significantly inhibited Raf kinase activation by H-RasG12V and K-RasG12V. Moreover, the FLIM experiments in Fig. 5b show that 100 μm indomethacin significantly increased (p < 0.05) the fluorescence lifetime of GFP-K-Ras.GTP or GFP-H-Ras.GTP in the presence of RFP-CRaf, indicating that indomethacin weakens the molecular association between GFP-Ras.GTP and RFP-CRaf. These results are consistent with earlier studies showing that cholesterol-dependent clusters are suboptimal for Raf activation (21, 25). Consistent with these results indomethacin pretreatment also significantly abrogated EGF-stimulated MAPK activation in BHK cells (Fig. 5, c and d).

FIGURE 5.

Indomethacin decreases Raf kinase activity in BHK cells expressing K-RasG12V or H-RasG12V. a, serum-starved BHK cells stably expressing either K-RasG12V or H-RasG12V were treated with 100 μm indomethacin for 1 h. Raf kinase activity in membrane fractions prepared from the cells was measured in a coupled EMK/ERK kinase assay. Data are shown as mean ± S.E. from three independent experiments. Statistical significance was calculated using one-way ANOVA (*, p < 0.05). b, GFP fluorescence lifetime was measured in BHK cells expressing GFP-K-RasG12V or GFP-H-RasG12V alone (GFP-Ras) or co-expressing the FRET acceptor RFP-CRaf in the absence (GFP-Ras/RFP-CRaf) or presence of 100 μm indomethacin (Indo). Data were collected from multiple single cells. GFP lifetime is shown as mean ± S.E. (n ≥ 92 cells from three independent experiments). Differences between control untreated cells expressing the GFP/RFP FRET pair and cells treated with indomethacin were examined by one-way ANOVA (*, p ≪ 0.05). c, serum-starved wild-type BHK cells were untreated or treated with 75 and 100 μm indomethacin for 1 h and stimulated with various concentrations of EGF for 5 min. ppERK levels were measured by quantitative Western blotting. A representative blot is shown. d, mean normalized ppERK levels from three independent experiments are plotted against EGF concentration. Statistical significance between the ppERK response in control and NSAID-treated cells at each EGF concentration was evaluated by one-way ANOVA (*, p < 0.05).

NSAIDs Perturb Ras Nanoclustering at Low Therapeutic Concentrations

Although indomethacin concentrations as high as 160 μm have been used therapeutically (23, 24), the more common indomethacin plasma concentrations are in the range of 5 to 10 μm. Using FLIM-FRET, we found that indomethacin at low doses of 5 and 10 μm significantly decreased the fluorescence lifetime of GFP-tH, GFP-H-Ras WT, and GFP-N-RasG12V in BHK cells co-expressing the cognate RFP-tagged FRET pair (Fig. 6a) while having no effect on GFP-K-RasG12V molecular interactions (Fig. 6b). Indomethacin at 5 and 10 μm also significantly decreased the fluorescence lifetime of GFP-tH in cells co-expressing RFP-H-RasG12V while having little effect on GFP-tH fluorescence lifetime in cells co-expressing GFP-tH and RFP-K-RasG12V (Fig. 6, c and d). These results therefore fully replicate the effects observed with higher doses of indomethacin shown in Figs. 1–5.

FIGURE 6.

Indomethacin at low therapeutic concentrations alters Ras nanoclustering. a, GFP fluorescence lifetime was measured in BHK cells expressing GFP-tH, GFP-H-RasWT, or GFP-N-RasG12V alone (GFP-Ras) or co-expressing the cognate FRET acceptor RFP-tH, RFP-H-RasWT, or RFP-N-RasG12V in the absence (GFP-RFP) or presence of 5 and 10 μm indomethacin. b–d, identical experiments were carried out in BHK cells expressing the cognate FRET pair for K-Ras.G12V (b) and the FRET pairs GFP-tH and RFP-H-RasG12V (c) and GFP-tH and RFP-K-RasG12V (d), each in the absence and presence of 5 and 10 μm indomethacin. Data were collected from multiple single cells. GFP lifetime is shown as mean ± S.E. (n ≥ 60 cells from three independent experiments). Differences between control untreated cells expressing the GFP/RFP FRET pair and cells treated with low dose indomethacin were examined by one-way ANOVA. Significant differences are shown (*, p ≪ 0.05).

Finally, to examine the generality of the effect of amphiphilic NSAIDs on Ras nanoclustering and signaling, we screened four other drugs: aspirin, salicylate, naproxen, and ibuprofen. We found that all of these NSAIDs at their corresponding therapeutic concentrations induced a significant decrease in the fluorescence lifetime of GFP in cells co-expressing the FRET pair of GFP-tH/RFP-tH (Fig. 7, a and e), similar to that seen with indomethacin. The same FLIM experiment was then performed using a fixed concentration of 25 μm for all tested NSAIDs to compare their efficacy. Whereas the induced changes in the fluorescence lifetime of GFP-tH in the FRET assay did not correlate with IC₅₀ values against COX-1/COX-2 (Fig. 7b), there was a strong correlation with the membrane partition coefficients of the NSAIDs induced (Fig. 7c). Taken together these results strongly suggest that NSAID-induced L_o domain stabilization is related to the biophysical properties of the drugs and not their ability to inhibit COX. Consistent with this interpretation, we found that salicylate, which did not inhibit COX at the tested concentration, also significantly inhibited EGF-stimulated activation of the MAPK cascade (Fig. 7d).

FIGURE 7.

NSAIDs alter Ras nanoclustering and MAPK signaling in a COX-independent manner. a, GFP fluorescence lifetime was measured in BHK cells expressing GFP-tH alone (GFP-tH) or co-expressing the FRET acceptor RFP-tH in the absence (GFP-RFP) or presence of various NSAIDs at therapeutic concentrations. The NSAIDs tested were aspirin (50 μm), salicylate (100 μm), ibuprofen (150 μm), and naproxen (25 μm). Data were collected from multiple single cells. GFP lifetime is shown as mean ± S.E. (n ≥ 60 cells from three independent experiments). Differences between untreated and indomethacin-treated GFP/RFP-expressing cells were examined by one-way ANOVA (*, p < 0.05). The FLIM experiment described in a was repeated using a fixed concentration of 25 μm aspirin, salicylate, indomethacin, and naproxen. The GFP lifetime values obtained were plotted against a logarithm of IC₅₀ values for each NSAID against COX (Log IC₅₀ COX) (b) and the logarithm of the membrane partition coefficient of each NSAID (Log K_w-NSAIDs) (c) of the corresponding NSAIDs. Linear correlation coefficients are shown as R² values. d, serum-starved wild-type BHK cells were untreated or treated with 100 μm salicylate (Sal) for 1 h. Cells were stimulated with various concentrations of EGF for 5 min, and ppERK levels were measured by quantitative Western blotting. The statistical significance between the ppERK response in control (Con) and NSAID-treated cells at each EGF concentration was evaluated by one-way ANOVA (*, p < 0.05). e, examples of FLIM images of BHK cells with or without indomethacin treatment are shown: BHK cells expressing GFP-tH and RFP-tH with no treatment (i); BHK cells expressing GFP-tH and RFP-tH in the presence of 5 μm indomethacin (ii), 50 μm aspirin (iii), 100 μm salicylate (iv), 150 μm ibuprofen (v), or 25 μm naproxen (vi).

DISCUSSION

We report here that biological amphiphiles, which include many anti-inflammatory agents, induce a wide-ranging reorganization of the plasma membrane nanostructure. Using different Ras isoforms as nanodomain markers, we have shown that indomethacin stabilizes cholesterol-dependent nanoclusters and induces mixing of Ras proteins that normally segregate efficiently to cholesterol-independent nanoclusters. The mixing of proteins in what we term heterotypic nanoclusters is of two types. The first class of heterotypic cluster results from a failure of GTP-dependent lateral segregation of H-Ras and N-Ras between cholesterol-dependent and cholesterol-independent nanoclusters, leading for example to mixed clusters of H-Ras.GTP and the lipid raft marker GFP-tH, which extensively co-clusters with H-Ras.GDP. The second class of heterotypic cluster is visualized as an aggregation of cholesterol-dependent and cholesterol-independent nanoclusters that are normally spatially separated, as for example the association of K-Ras.GTP nanoclusters with L_o clusters containing GFP-tH. In this second class of mixed clusters, there appears to be no actual mixing of the constituents of each type of nanocluster, which assemble normally but form in the same region of the membrane instead of segregating to spatially distinct areas. Signal transmission by both classes of heterotypic cluster is compromised, as evidenced by reduced Raf recruitment and activation by H-Ras.GTP or K-Ras.GTP clusters in indomethacin-treated cells. Ras-dependent Raf activation proceeds inefficiently in cholesterol-dependent nanoclusters (4, 21), and thus the presence of cholesterol within or near previously cholesterol-independent nanoclusters may account for the abrogated MAPK output. Although our detailed study focused on indomethacin, all other anti-inflammatory amphiphiles that we tested had similar effects on Ras nanoclustering and signal output.

We propose that indomethacin-induced changes in nanoclustering originate from the ability of the amphiphilic agent to increase lipid immiscibility within the plasma membrane. We showed previously that indomethacin enhances phase separation in synthetic and biological membranes by increasing the fluidity and polarity of liquid-disordered, (L_d) domains (14, 15). These results are consistent with studies of synthetic model bilayers and computer simulations, which show that amphiphilic drugs, detergents, or more polar lipids enhance or stabilize domain formation in gel phase, L_o phase, or raft-like bilayers (26–32). In the current study we observed a similar pattern of effects of amphiphilic NSAIDs on the phase heterogeneity of biological cell plasma membranes with stabilization of cholesterol-dependent or raft-like GFP-tH, H-Ras.GDP, and N-Ras.GTP nanoclusters. The same biophysical mechanism may also account for the ability of indomethacin to impair the GTP-dependent lateral segregation of H-Ras. A balance model has been used to describe how GTP loading reorganizes H-Ras membrane interactions (6, 8, 33). The β2-β3 loop and helix α5 of the H-Ras G-domain act as a guanine nucleotide-regulated switch (called switch III) that mediates the competing interactions of helix α4 and basic residues in the HVR with polar lipids in the plasma (6, 8, 11). When interactions of the HVR prevail, the palmitates of the C-terminal membrane anchor are highly ordered, and the GDP-loaded H-Ras protein is driven to L_o nanoclusters (11). When H-Ras is GTP-loaded, the interactions of helix α4 predominate, the palmitates are disordered, and there is a major change in the orientation of the G-domain with respect to the membrane, which allows binding of scaffold proteins and assembly into cholesterol-independent clusters (6, 8, 12). We propose that changes in the polarity of the plasma membrane induced by NSAIDs interfere with the operation of switch III, resulting in H-Ras.GTP retaining the orientation of H-Ras.GDP with the associated ordered palmitates of the C-teminal anchor; this in turn results in the co-assembly of H-Ras.GTP proteins into the L_o domains with H-Ras.GDP. Differential G-domain orientation also contributes to an increased rate of palmitate turnover on H-Ras.GTP compared with H-Ras.GDP. In the GDP orientation the H-Ras C-terminal anchor peptide is pulled deeper into the inner leaflet of the plasma membrane than when in the GTP orientation, rendering the palmitoyl thioester bonds of Cys-181 and Cys-184 less accessible to thioesterases (34–38). Furthermore at least one major thioesterase, APT1, is palmitoylated and localizes to the L_d membrane (38, 39). Thus the combined effects of NSAIDs in increasing the L_o localization of H-Ras.GTP, removing it from L_d-localized APT1 and potentially stabilizing H-Ras.GTP in an aberrant H-Ras.GDP orientation, may together reduce the rate of palmitate turnover on H-Ras.GTP.

Indomethacin has no effect on the homotypic nanoclustering of K-Ras. However, electrostatic interactions between the K-Ras polybasic domain and negatively charged polar lipids in the plasma membrane are critical for the formation of K-Ras nanoclusters (40–46). In this context, it is of interest that amphiphilic drugs, such as trifluoperazine, thioridazine, clozapine, chloropromazine, haloperidol, and cyclosporine, tend to preferentially partition into L_o/L_d domain boundaries (47, 48). If indomethacin partitions similarly, the presence of negatively charged indomethacin at the L_o/L_d domain boundary could account for the increased probability of K-Ras nanocluster formation at the L_o/L_d domain boundaries and close interaction with cholesterol-dependent clusters such as GFP-tH. Despite the proximity with L_o domains in indomethacin-treated cells, K-Ras proteins do not mix homogeneously because the branched and polyunsaturated farnesyl chain is efficiently excluded by the tightly packed, ordered lipids of the L_o domain (49, 50).

Clinical studies suggest that long-term daily intake of aspirin and other NSAIDs may significantly decrease the risk of death due to cancer by as much as 40% (51). Interestingly, aspirin is particularly effective against esophageal, colorectal, and lung cancers (51), all of which express oncogenic K-Ras with high frequency (52). Although it has been generally assumed that these chemopreventive effects of NSAIDs may be linked to their ability to inhibit COX, our findings here suggest that NSAIDs rather may act directly on the biophysical properties of the plasma membrane to modulate Ras signal transmission. More broadly, this study represents the first demonstration that Ras spatiotemporal dynamics may be a viable drug target for therapeutic intervention.