Oncogenic K-ras segregates at spatially distinct plasma membrane signaling platforms according to its phosphorylation status

Somatic mutations that activate Ras are detected in ∼15–20% of all human malignancies, highlighting the importance of Ras-GTPase-mediated signaling pathways in oncogenesis. These mutations, which give rise to a protein that is defective in GTP hydrolysis and, therefore, remains constitutively active in a GTP-bound form, have been detected in each of the three closely related human Ras genes (HRAS, NRAS and KRAS). However, the vast majority of mutations detected in human cancers arise in the KRAS (Prior et al., 2012; Pylayeva-Gupta et al., 2011; Schubbert et al., 2007).

H- and N-Ras achieve high-affinity hydrophobic membrane binding mainly through lipid modifications. By contrast, K-Ras has, adjacent to the farnesylated cysteine Cys185, a stretch of lysine residues – known as the polybasic domain – that promotes an electrostatic interaction with the negatively charged phospholipids (Hancock et al., 1989; Silvius, 2002), which confines K-Ras almost entirely to non-raft microdomains within the plasma membrane (Prior et al., 2001).

The different membrane anchors interact with lipids and proteins of the plasma membrane and, together with the hypervariable region (HVR), drive the Ras isoforms into spatially and structurally distinct nanodomains, of which each then contains a cluster of molecules (nanocluster) (Abankwa et al., 2008; Hancock and Parton, 2005). Importantly, the nanodomains that are occupied by the three isoforms of Ras do not show any overlap. Interestingly, not only are the different Ras isoforms laterally segregated, but inactive GDP-loaded Ras occupies nanodomains that are spatially distinct from those occupied by the active GTP-loaded form. This indicates that the globular domain of the protein also regulates its interaction with the distinct membrane nanodomains. Formation of these nanoclusters is essential for activation of mitogen-activated protein kinase (MAPK), because they constitute exclusive sites in the plasma membrane for Raf-1 recruitment and ERK activation (Kholodenko et al., 2010; Plowman et al., 2005; Tian et al., 2007).

Because electrostatic interactions control the process of membrane interaction for K-Ras, membrane affinity can be modulated by changes in the overall charge of the polybasic domain through phosphorylation of Ser181 (Ahearn et al., 2012; Ballester et al., 1987; Bivona et al., 2006; Plowman et al., 2008). Recently, we have demonstrated that Ser181 phosphorylation regulates the functions of both wild-type and oncogenic K-Ras. Growth without contact inhibition, mobility and apoptosis resistance upon adriamycin treatment of cells that express oncogenic non-phosphorylatable K-Ras were highly compromised, correlating with decreased activation of the main downstream effectors ERK and AKT. Therefore, in our model, phosphorylation of K-Ras is essential to ensure the correct activation of ERK and AKT signaling pathways with an important functional relevance (Alvarez-Moya et al., 2010; Alvarez-Moya et al., 2011).

Understanding how phosphorylation modulates oncogenic K-Ras activity is of outstanding interest in order to design new therapeutic strategies to treat human carcinomas that harbour oncogenic K-Ras. Here we show, by using cell fractionation and immunoelectron microscopy that – depending on the status of its phosphorylation – oncogenic K-Ras segregates at the plasma membrane, thereby potentially influencing the activation of its effectors. We propose that this different localization is responsible for the distinct functionality of phosphorylated versus non-phosphorylated K-Ras.

We have shown previously that Ser181 phosphorylation of the oncogenic K-RasG12V mutant (always GTP-loaded) positively modulates the activation of ERK and AKT, especially under stress conditions. However, the in vitro affinity between oncogenic K-RasG12V and Raf-1 was not affected by this post-translational modification of K-Ras (Alvarez-Moya et al., 2010) and, consequently, it could not explain the differences in ERK activation. Similarly, and in agreement with Plowman et al. (Plowman et al., 2008), FRET analysis did not show an increase in the association of the phosphomimetic Ser181 to Asp mutation of oncogenic K-Ras (hereafter referred to as K-RasG12V-S181D) and Raf-1 when compared with the non-phosphorylatable Ser181 to Ala mutation of oncogenic K-Ras (hereafter referred to as K-RasG12V-S181A) (Fig. 1A,B). Interestingly, although interaction of Raf-1 with K-Ras was not affected by the phosphorylation status of K-Ras, immunofluorescence analysis showed a higher proportion of active Raf-1 colocalizing with K-RasG12V-S181D, as shown by Raf-1 phosphorylation at Ser338 (Fig. 1C,D).

Coimmunoprecipitation analysis using HeLa cells transiently expressing HA-tagged oncogenic K-Ras mutants corroborated these results. Although the amount of Raf-1 that coimmunoprecipitated with non-phosphorylatable K-Ras or with phosphomimetic K-Ras was not significantly different, K-RasG12V-S181D showed increased ability to co-immunoprecipitate with active Raf-1 (Ser338-P). Finally, another K-Ras effector, the catalytic subunit of PI 3-kinase, p110α, also showed higher coimmunoprecipitation with K-RasG12V-S181D than with K-RasG12V-S181A (Fig. 1E). Thus, we hypothesize that Ser181 phosphorylation of oncogenic K-Ras favors activation or retention of activated effectors through a yet undefined mechanism.

The negative charges introduced in the HVR by phosphorylation might modify the affinity of K-Ras to the plasma membrane and alter its localization (Yeung et al., 2006). In fact, certain groups have reported that phosphomimetic K-Ras is internalized to intracellular membranes, such as those in mitochondria, the Golgi complex, the endoplasmic reticulum or in endosomes (Bivona et al., 2006; Fivaz and Meyer, 2005). However, in agreement with previous observations (Lopez-Alcalá et al., 2008; Plowman et al., 2008), we show here that, after simultaneous transfection of YFP-K-RasG12V-S181A and mCherry-K-RasG12V-S181D into HEK293 cells, both phospho-mutants are mainly located at the plasma membrane (Fig. 2A). We aimed to analyze whether different localization of phosphorylated versus non-phosphorylated K-Ras within the plasma membrane is the basis for the observed different recruitment of active Raf-1 and PI 3-kinase.

We, therefore, tested whether Ser181 phosphorylation of K-Ras induces its segregation into different plasma membrane domains by performing a cell fractionation in a density gradient. K-Ras, in contrast to N- and H-Ras, is mainly localized in disordered non-raft detergent-sensitive plasma membrane domains (Prior et al., 2001). This constitutes a drawback when studying its distribution by using the classic cell fractionation procedures. In our study, membranes from HEK293 cells were fractionated into a detergent free method to prevent the disruption of K-Ras domains (Macdonald and Pike, 2005). As cells were co-transfected with the two K-Ras mutants (YFP-K-RasG12V-S181A and mCherry-K-RasG12V-S181D), a single fractionation and single gel electrophoresis were performed per experiment, avoiding possible variability in gradient generation or gel loading.

The early endosome marker EEA1, the endoplasmic reticulum marker Sec61α and the caveolar lipid raft marker caveolin-1, were confined in the higher-density fractions; by contrast, Na⁺/K⁺ ATPase distribution was extended throughout many fractions (Fig. 2B). Endogenous K-Ras (wild type) was always found between fractions 4 and 6 (Fig. 2B,C,D,E). Furthermore, both exogenous non-oncogenic (wild-type) YFP-K-Ras and mCherry-K-Ras were found mainly in the same fractions as endogenous wild-type K-Ras (supplementary material Fig. S1). When analyzing the exogenous coexpressed oncogenic K-Ras phospho-mutants – although a more widespread distribution of the protein was observed compared with endogenous K-Ras – we reproducibly observed that the non-phosphorylatable K-Ras mutant (K-RasG12V-S181A) peaked between fractions 5 and 7, while the peak of phosphomimetic K-Ras (K-RasG12V-S181D) shifted towards the higher-density fractions 8–10 (Fig. 2B–D). A putative tag-artifact in the differential fractionation was dismissed because, when the reversely tagged proteins were used (mChery-K-RasG12V-S181A and YFP-K-RasG12V-S181D), differences of fractionation according to the phosphorylation status were maintained (Fig. 2D; supplementary material Fig. S1). In agreement with data shown in Fig. 1B, both PI 3-kinase (p110α) and Raf-1 phosphorylated at Ser 338 (Raf-1-Ser338-P) exhibited higher co-fractionation with K-RasG12V-S181D than with the non-phosphorylatable K-Ras (Fig. 2B). This reinforces the concept of a segregated membrane domain for K-Ras-Ser181-P that constitutes a preferential signaling platform.

We next analyzed the localization of the phosphorylatable oncogenic K-Ras (K-RasG12V-S181). As shown in Fig. 2E, the majority of K-RasG12V-S181 was found at the beginning of the gradient, whereas a certain amount fractionated together with K-RasG12V-S181D, suggesting that a proportion of K-RasG12V-S181 is phosphorylated under these conditions.

To further determine the presence of distinctly segregated K-Ras-Ser181-P nanodomains that ensure preferential signaling platforms, we attempted to analyze the distribution of our oncogenic K-Ras phospho-mutants at nanoscale level. Distribution analysis of gold-labeled protein by using a combined immune-EM-statistics approach allows the characterization of K-Ras nanoclusters in otherwise morphologically featureless plasma membrane (Prior et al., 2003a). It has also been shown previously that both phosphorylated and non-phosphorylatable K-Ras are able to form such nanoclusters (Plowman et al., 2008). Since our fractionation experiments indicate segregation of phosphorylated and non-phosphorylatable K-Ras, and the inner leaflet of the plasma membrane consists of a mosaic of different nanoclusters (Prior et al., 2003a), we wanted to directly compare the nanocluster distribution of our K-Ras variants. To this end, cells were co-transfected with both oncogenic K-Ras phospho-mutants fused to either YFP or mCherry. Intact 2D sheets of apical plasma membrane of adherent cells, were ripped off directly onto electron microscopy grids and were immunogold labeled using anti-GFP antibodies conjugated directly to 5 nm gold particles and anti-RFP antibodies directly conjugated to 10 nm gold particles. To estimate the degree of co-clustering of our K-RasG12V species, Ripley's bivariate K-function was used [L(Biv)-r]. As a positive control, co-transfection of the same mutant with different tags was performed to assess whether co-clustering can be observed. As expected, YFP-K-RasG12V-S181D co-clustered with mCherryK-RasG12V-S181D, thus dismissing the possibility of a tag-effect. By contrast, if cells were co-transfected with YFP-K-RasG12V-S181D and mCherryK-RasG12V-S181A no co-clustering was observed, providing striking evidence that confirms our initial conception of the existence of a spatially segregated cluster of phospho-K-RasG12V (Fig. 3A,D). In agreement with Plowman et al. (Plowman et al., 2008), the analysis of clusters, by using Ripley's univariate K-function in the same samples, showed that both non-phosphorylatable K-Ras and phosphomimetic K-Ras were able to form clusters (data not shown).

Protein kinase C (PKC) can phosphorylate K-Ras in vitro (Ballester et al., 1987), and it has been shown in vivo that phosphorylation of K-Ras at Ser181 is induced when PKC is activated (using PMA) and CaM is inhibited (using W13), whereas it is reduced after treating cells with PKC inhibitors (e.g. BIM) (Alvarez-Moya et al., 2010). To conclusively demonstrate that Ser181 phosphorylation regulates the localization of oncogenic K-Ras in different nanoclusters at the plasma membrane, co-clustering of K-RasG12V-S181 with K-RasG12V-S181D and with K-RasG12V-S181A was analyzed after phosphorylation was induced (using PMA+W13) or phosphorylation was inhibited (using BIM) by using the immuno-EM-statistics approach indicated above (see Materials and Methods) (Fig. 3B,D).

To analyze the functional significance of the different clustering of oncogenic K-Ras according to its phosphorylation status, the immuno-EM-statistics approach was used again. Cells were co-transfected with mGFP-p110α and mCherryK-RasG12V-S181D or mCherryK-RasG12V-S181A, and processed as indicated in the previous section. Ripley's bivariate K-function showed strong co-clustering of mGFP-p110α and the phosphomimetic K-Ras mutant, whereas mutually exclusive distribution was observed with the non-phophorylatable K-Ras. Finally, co-clustering analysis of mGFP-p110α with phosphorylatable K-Ras after induction (PMA+W13) or inhibition (BIM) of phosphorylation, conclusively demonstrated that PI 3-kinase (p110α) is efficiently recruited to the segregated clusters of phoshorylated K-Ras (Fig. 3C,D).

Through an integrated approach of density gradient fractionation and immuno-EM-statistics, we have found striking evidence of a lateral segregation of oncogenic K-Ras to the inner leaflet of the plasma membrane when it is phosphorylated at Ser181, which is of functional significance. We also demonstrated that phosphorylated K-Ras associates more with active Raf-1 and PI 3-kinase (p110α) and, at nanoscale, clusters with PI 3-kinase (p110α). Because nanoclusters operate as temporary signaling platforms at the plasma membrane and contain certain mixtures of kinases, phosphatases and other signaling proteins (Inder et al., 2008; Prior et al., 2003a), we expect to find that the molecular environment can facilitate the activation of main K-Ras effectors within the phospho-K-Ras platforms (Fig. 4). Our findings provide a new answer on how oncogenic K-Ras – which is always GTP-loaded and, thus, presumably always active – exhibits distinctive signaling activity after phosphorylation.

Human epithelial kidney cell lines (HEK293T) and HeLa cells were grown in Dulbecco's modified Eagle's medium (DMEM) containing 10% FBS (Biological Industries), pyruvic acid, antibiotics and glutamine.

Primary antibodies: anti-Raf-1 (#610152; BD Transduction Laboratories, San Jose, CA); anti-phospho-Ser338-Raf-1 (#05534; Millipore, Billerica, MA); anti-PI 3-kinase (p110α) (clone C73F8) (#4249; Cell Signalling, Danvers, MA); anti-HA (clone HA-7) (#A2095; Sigma, St Louis, MO); rabbit anti-RFP (#A01388-40; GenScript, Piscataway, NJ); mouse anti-GFP (#ADI-SAB-500-E; Stressgene); anti-K-Ras (Ab-1) (#OP24; Calbiochem); anti-H-Ras (C-20) (#sc-520, Santa Cruz, Santa Cruz, CA), anti-Na⁺/K⁺-ATPase (clone C464.6) (#05-369X-555; Millipore); anti-EEA1 (#610457; BD Bioscience, Franklin Lakes, NJ); anti-Sec61α (#07-204; Millipore); anti-Cav1 (#610407; BD Bioscience).

Phorbol-12-myristate-13-acetate (PMA) and N-(4-aminobutyl)-5-cloro-2-naphtalensulphonamide (W13) were from Sigma and bisindolylmaleimide (BIM) from Calbiochem.

HeLa cells (10-cm dish), co-transfected with Myc-Raf-1 and either HA-K-RasG12V-S181A or HA-K-RasG12V-S181D by using the X-tremeGENE HP DNA Transfection Reagent (Roche) were lysed in Ras extraction buffer as described before (Alvarez-Moya et al., 2010). The lysate was incubated for 3 hours at 4°C with anti-HA-tag antibody crosslinked to Dynabeads, washed three times and eluted with glycine 200 mM pH 2.5.

HeLa cells were used for confocal microscopy studies. To determine colocalization of mCherry-RasG12V-S181D and YFP-K-RasG12V-S181A, YFP and mCherry images were acquired sequentially using 514 nm and 561 nm laser lines, emission detection ranges are 525–573 nm and 580–700 nm, respectively, and the confocal pinhole was set at 1 Airy unit. Images were acquired at 400 Hz in a 1024×1024 pixel format, zoom 4 and pixel size of 60×60 nm, using Ras mutants fused to celurian fluorescent protein and Raf-1 fused to YFP.

FRET measurements were carried out on the basis of the acceptor photobleaching method and performed as previously decribed (Vilá de Muga et al., 2009), using Ras mutants fused to celurian fluorescent protein and Raf-1 fused to YFP.

HEK293T cells (seven 15-cm dishes) were transfected using the calcium phosphate method and, after 24–48 hours, fractionation of the cell membrane was performed into a continuous OptiPrep density gradient as previously described (Macdonald and Pike, 2005) except for the densities used [sample at 20% and continuous gradient from 15–0% (v/v)]. Gradients were fractionated into 670 µL aliquots.

HeLa cells were grown on coverslips at low density and co-transfected either with YFP- or mCherry- K-RasG12V, K-RasG12V-S181A, K-RasG12V-S181D; or with GFP-p110α and mCherry- K-RasG12V, K-RasG12V-S181A or K-RasG12V-S181D by using GeneJuice (Merck Millipore) according to the manufacturer's specifications. After 24 hours of transfection cells were fixed in 4% paraformaldehyde plus 0.1% glutaraldehyde, and labeled with gradient-purified 2 nm-gold- and 5 nm-gold-conjugated antibodies as previously described (Prior et al., 2003b). Plasma membrane sheets were imaged using an FEI 120 kV Tecnai transmission electron microscope obtaining images at a magnification of ×87,000. Image processing and Ripley's bivariate K-function analysis were performed to examine whether either gold particles, at a radius r, clustered around each other. Significance of Ripley's bivariate K-function was established by using the Monte Carlo method as described (Prior et al., 2003b).

We are grateful to R. Marais (UK Cancer Research Centre, London, UK) for kindly providing the pEF-HA-K-RasG12V and Myc-Raf-1 plasmids, to J. F. Hancock (University of Texas, Houston, USA) for the pEGFP-p110α plasmid and to the advanced optical microscopy unit of CCiT-UB for technical assistance.