Raf Kinase Inhibitor Protein RKIP Enhances Signaling by Glycogen Synthase Kinase-3β Free

Since the discovery of the Raf kinase inhibitory protein (RKIP) 26 years ago by Jones and colleagues (1), and later by Bernier and colleagues (2), there has been considerable amount of published data linking RKIP to various intracellular signaling networks that control cellular growth (3–6), motility (4, 7, 8), epithelial to mesenchymal transition (EMT; refs. 9, 10), and differentiation (11–13). RKIP is widely expressed in human tissues, indicating a cardinal role in various physiologic processes (14, 15). This small, evolutionary, conserved 21-kDa protein has a conserved ligand-binding pocket that is required for its inhibitory activity of the c-RAF kinase to bind and activate MAP/ERK kinase (MEK; refs. 16, 17). Therefore, RKIP was identified as the first physiologic inhibitor of the RAF-MEK-ERK pathway (18). RKIP phosphorylation at S153 by protein kinase C reduces the affinity of RKIP for c-RAF and enhances RKIP binding to G-protein–coupled receptor kinase-2 (19). In addition, RKIP phosphorylated at S153 associates with centrosomes and the kinetochore to control cell-cycle progression and genomic stability (20, 21). Moreover, RKIP has been shown to downregulate IκB kinase (IKK) activity directly and by interfering with IKK activators, namely, TAK1 and NIK, thus dampening the effects of tumor necrosis factor α and interleukin 1β on the activation of nuclear factor κB (NF-κB; ref. 22). Loss of RKIP expression in cancers leads to transcriptional activation of NF-κB, resulting in a dramatic inhibition of apoptosis and the development of chemoresistance (23–25). RKIP expression is negatively regulated by SNAIL, thus permitting enhanced NF-κB signaling resulting in a circuitry that regulates both the metastatic cascade and resistance to apoptosis by cytotoxic drugs (26).

Loss of or reduced RKIP expression in clinical samples has been associated with many types of aggressive and metastatic cancers including colorectal cancer (CRC), prostate cancer, and breast cancer (15, 27–29). These studies also suggested that the biological activities of RKIP could not be explained by its effects on a single pathway, prompting us to search for new RKIP targets.

Purified RKIP-HA fusion proteins (wild-type and D70A mutant) at 10 μmol/L were used to probe the ProtoArray Human Protein Microarray V.4 from Invitrogen. Each array contains approximately 8,000 individual protein species, with each protein expressed as an N-terminal GST fusion protein and printed in duplicate on a nitrocellulose-coated glass slide. Arrays were blocked for 1 hour in PBS containing 1% bovine serum albumin (BSA) and 0.1% Tween 20 before probing. Arrays were then washed three times in a dedicated Cyto-Tec slide mailer (Sakura Finetek) containing probing buffer (PBS containing 5 mmol/L MgCl₂; 0.5 mmol/L DTT; 0.05% Triton X-100; 5% glycerol, and 1% BSA) at 10-minute intervals on a shaker. A sandwich assay (anti-HA followed by anti-mouse-Alexa 647) was used for the detection of protein–protein interactions. Arrays were probed in primary antibody solution for 1 hour. Antibody solution was prepared with probing buffer to a final ratio of 1:1.562, as recommended by Invitrogen. Arrays were washed again as previously, following probing with primary and again following the application of the secondary antibody. Each ProtoArray was then scanned using Scan Array Express to visualize interactions. The RKIP interactions are scored and interpreted into a Confidence Flag value using the BlueFuse software (BlueGnome Limited). BlueFuse uses a proprietary algorithm that takes into account the variability of the experimental data (e.g., spot intensity, spot shape, and uniformity) to estimate a range of possible results, which is then used to provide an assessment of spot quality and generate a degree of confidence that should be placed in each result. A Confidence Flag is assigned to each interaction, using a Bayesian Framework which is based on an evaluation of spot shape, intensity, and integrity.

Data sets (n = 3) were compiled together, with an additional scoring system allowed for semiquantitation of data. These additional scores were used as additional threshold/cutoff for the further refinement of the list of potential interactors so that biological validation will be pursued only on the most confident potential interactors.

The target sequence of 5′-CCGCTATGTCTGGCTGGTTTA-3′ was cloned into pcDNA6.2-GW/EmGFPmiR to generate the miRNA vector pcDNA6.2-GW/EmGFPmiR-PBP_miR_499. This vector was transfected into HEK-293 cells, using an Amaxa Nucleofector (Amaxa). For this, solution V and program Q-001 were used according to the manufacturer's protocol. Cells were selected with 5 μg/mL blasticidin for 1 week. Then, the level of RKIP downregulation was checked with an anti-RKIP antibody (Upstate). HEK-293 cells and their transfected counterparts were maintained in Dulbecco's minimal essential medium (DMEM) with 10% FBS and 2 mmol/L glutamine (all from Invitrogen). The doxycycline-inducible cells are based on Flp-In T-Rex-293 from Invitrogen. A total of 1 × 10⁶ Flp-In T-Rex-293 cells were transfected with 1 μg of pcDNA5/FRT-FLAG-RKIP and 9 μg of pOG44 (Invitrogen; V6005-20). Forty-eight hours after transfection, cells were selected for 3 days with 150 μg/mL hygromycin. Afterward, clones were picked, expanded, and analyzed for RKIP silencing/expression as shown in Supplementary Figure S1. Flp-In T-Rex-293 cells were originally purchased from Invitrogen and authenticated by them (http://tools.invitrogen.com/Content/SFS/ProductNotes/F_051025_Flp-In-TS-TL-MKT-HL.pdf). HEK-293 cell lines were originally purchased from CRUK and were authenticated by the European Collection of Cell Cultures in September 2009, using microsatellite genotyping (PCR-based). These cells were used to construct the aforementioned HEK-499 cell line. All cells were used for experiments for 6 to 8 weeks before they were replaced with fresh stocks, which are stored in liquid nitrogen.

Western blotting experiments were done using standard described procedure. A list of antibodies utilized by this study is given in Supplementary Table 3. Detection was done by chemiluminescence, using a commercial kit (Super Signal West Pico chemiluminescent substrate; Pierce). Densitometric analyses were done using the Quantity one software of the GS-800 calibrated densitometer (Bio-Rad). Results were expressed as the ratio of the protein of interest normalized to a control protein. If reprobing was required, the membranes were stripped in stripping buffer (100 mmol/L 2-mercaptoethanol, 2% SDS, 62.5 mmol/L Tris-HCl, pH 6.7) at 50°C for 30 minutes with occasional shaking.

Cells were lysed in cell lysis buffer (Cell Signaling) containing 20 mmol/L Tris-HCl, pH 7.5, 150 mmol/L NaCl, 1 mmol/L EDTA 1 mmol/L EGTA, 1% Triton X-100, 2.5 mmol/L sodium pyrophosphate, 1 mmol/L β-glycerophosphate, 1 mmol/L sodium vanadate, and 1 μmol/L leupeptin and cleared of debris by centrifugation at 20,800 × g for 10 minutes. Cleared lysates were adjusted to a protein concentration of 0.5 mg/mL. Five microliters of anti-RKIP antibody (Santa Cruz) and 50 μL of agarose beads (protein A/G plus; Santa Cruz) were added and incubated at 4°C overnight with shaking. After the incubation, the mixture was centrifuged at 20,800 × g in Eppendorf microcentrifuge for 30 seconds. The slurry was washed three times with PBS. Then, the pellet was resuspended in SDS-PAGE loading buffer and proteins were separated by SDS-PAGE. Western blotting was done using anti-GSK3β, GSK3α antibodies (Cell Signaling), and anti-RKIP antibody (Santa Cruz). Signals were visualized as described in the Western blotting method.

Autoclaved glass coverslips were placed in 6-well tissue culture plates. One to 2 mL of cell suspension containing 5 × 10⁴ cells was added and allowed to grow at 37°C in a humidified CO₂ incubator until 70% confluency was reached. Cells were washed twice in PBS at room temperature and fixed with 4% paraformaldehyde in PBS, pH 7.4, for 15 minutes at room temperature. After 3 washes with PBS, cells were permeabilized by incubation in PBS containing 0.1% Triton X-100 for 15 minutes at room temperature, followed by 3 washes in PBS. Cells were then blocked by incubation in PBS containing 1% BSA for 1 hour at room temperature. The appropriate primary antibodies (Supplementary Table 3) were diluted 1:100 in PBS with 1% BSA. The cells were then incubated with a mixture of two different primary antibodies derived from 2 different species at 4°C overnight. After washing in PBS three times for 5 minutes, the appropriate species-specific fluorophore-conjugated secondary antibodies (1:500 dilution) were added for 1 hour at room temperature. After washing three times in PBS, 5 minutes each, coverslips were analyzed using a LSM 510 META confocal microscope (Carl Zeiss).

Cover slips seeded with the required cell lines were incubated with 5 μmol/L dihydroethidium (DHE), a redox-sensitive probe that is oxidized by superoxide and then intercalates into nuclear DNA (Molecular Probes). In parallel, MitoSox (Molecular Probes), a mitochondrial superoxide indicator, was used at a concentration of 5 μmol/L for 30 minutes at 37°C. Cells were then rinsed with 0.1% BSA/DMEM and analyzed by confocal fluorescence microscopy, using an excitation wavelength of 488 nm and emission wavelength of 650 nm. DHE or MitoSox alone did not produce any fluorescence. Images were collected and the fluorescent intensity was normalized by dividing the image average intensity by cell number.

H₂O₂ produced by cells in the media was measured using the Amplex Red Hydrogen Peroxide/Peroxidase Assay Kit (Invitrogen), as described by the manufacturer. In brief, Amplex Red reagent (10-acetyl-3,7-dihydroxyphenoxazine) reacts with H₂O₂ in a 1:1 stoichiometry to produce the red fluorescent oxidation product resorufin. Resorufin has excitation and emission maxima of 571 nm and 585 nm, respectively, and can be measured fluorometrically.

Cells were homogenized in ice-cold buffer containing 50 mmol/L Tris-HCl, pH 7.5, 5 mmol/L EDTA, and 1 mmol/L DTT. After centrifugation at 10,000 × g for 15 minutes at 4°C, the supernatants were collected and stored on ice. Thirty-three micrograms of protein extract from each sample was added to 50 μL cosubstrate mix containing 0.2 mmol/L NADPH, 2 mmol/L reduced glutathione, and 2 units/mL glutathione reductase. Twenty microliters of t-butyl hydroperoxide was added to the mix and agitated for a few seconds after which absorbance was read at 340 nm every minute to obtain five time points. Superoxide dismutase (SOD) activities were measured using Dojindo's assay kit (Dojindo) according to the manufacturer's instructions.

Sections from formalin-fixed, paraffin-embedded human colon cancer samples or transgenic mice tissues were cut into 4-μm thick sections. The downstream processing of slides for immunohistochemistry was done using standard procedure as previously described (15). Rabbit polyclonal antibody against RKIP protein (dilution 1:900), and rabbit monoclonal antibody against GSK3β (clone 27C10, dilution 1:100; Cell Signaling) were used.

The RKIP knockout mouse has been described previously (30). Mice were housed under standard conditions, and all experiments were done in accordance with UK Home Office guidelines and local ethical approval.

SPSS software version 17 was used for statistical calculations. Data are expressed as mean ± SD. Two-sampled Student's t test assuming unequal variances was used to compare mean data between 2 groups. A level of P ≤ 0.05 was considered to be significant.

To identify possible RKIP binding proteins, we utilized a protein array (ProtoArray Human Protein Microarray V.4)-based screen. Each array contains around 8,000 individual human proteins, with each protein expressed as an N-terminal GST fusion protein and printed in duplicate on a nitrocellulose-coated glass slide. Wild-type and D70A loss-of-function mutant RKIP-HA fusion proteins were used to probe the microarray. Wild-type RKIP binds to both GSK3α and GSK3β proteins (Fig. 1A). The RKIP D70A mutant, which does not bind to c-RAF and MEK (16, 17) and was used as a control for the specificity of interactions, had a strongly reduced binding to GSK3α and approximately 50% reduction in binding to GSK3β (Fig. 1A). To confirm this binding specificity further with endogenous proteins, we immunoprecipitated endogenous RKIP from HEK-293 cells and then immunoblotted for endogenous GSK3β and GSK3α. RKIP coprecipitated both GSK3β and GSK3α proteins (Fig. 1B). We then tested RKIP/GSK3β spatial distribution in HEK-293 cells, using confocal microscopy. We noted a significant colocalization of RKIP and GSK3β proteins in the cytoplasm, around the plasma membrane and in cellular protrusions (Fig. 1C). In summary, these data confirm that RKIP specifically binds GSK3β protein and that D70, although required for GSK3α binding, is involved but not essential for GSK3β binding.

To test the consequence of RKIP modulation on GSK3 proteins, we silenced RKIP expression in HEK 293 cells, using miRNA antisense technology. We generated 4 cell lines with reduced RKIP expression (Supplementary Fig. S1A) and utilized a stably transfected cell line, HEK-499, with low RKIP protein level for further studies (Supplementary Fig. S1B and C). In addition, using the Flp-In T-Rex-293 cells, we constructed a stable cell line harboring a doxycycline-inducible FLAG-RKIP transgene. In these cells, the overexpression of FLAG-RKIP can be upregulated to approximately 30% to 50% over endogenous level (Supplementary Figs. S1B and C), which is well in the range of physiologic variations of RKIP expression (23).

In RKIP-silenced HEK-499 cells, GSK3β protein levels were significantly downregulated both by Western blotting and by immunofluorescence (Fig. 2A and Supplementary Fig. S3A), whereas RKIP overexpression tended to upregulate GSK3β protein levels (Fig. 2A). In contrast, GSK3α expression was not affected by RKIP-level manipulation (Supplementary Fig. S2), indicating a specific role for RKIP in GSK3β stability. Therefore, we focused further analysis on GSK3β.

GSK3β is constitutively active due to cotranslational autophosphorylation on Y216 in the activation loop (31) and is inhibited by AKT-mediated phosphorylation of S9 (32). RKIP downregulation caused a reduction of the active Y216 phosphorylated form of GSK3β (Fig. 2B) although the inactive S9 phosphorylated form of GSK3β was maintained (Fig. 2C). It is worth noting that the normalized ratio of Y216 p-GSK3β to GSK3β did not change significantly in HEK-499 cells, indicating that the apparent reduction of the Y216 phosphorylated form of GSK3β was due to changes in the total level of GSK3β (Fig. 2B). Conversely, inducing RKIP expression in Flp-In T-Rex-293 cells had little effect on the expression level of GSK3β or its phosphorylation on Y216 (Fig. 2A and B). We confirmed these data, using confocal immunofluorescence imaging of inactive (pS9), active (pY216), and total GSK3β proteins (Supplementary Fig. S3A). GSK3β mRNA levels were not different between the different cell lines and their corresponding controls (Supplementary Fig. S3B), indicating that the modification of GSK3β levels mediated by RKIP loss occurred posttranscriptionally. These data show that RKIP is required for the posttranscriptional maintenance of GSK3β in its active form and that RKIP loss induced GSK3β destabilization and inactivation. Next, we set out to determine the mechanism.

As RKIP did not regulate the inhibitory S9 phosphorylation, we examined phosphorylation of T390, which was recently described as an inhibitory site targeted by p38 MAPK (33). We immunoprecipitated GSK3β and found an approximately 5-fold increase in T390 phosphorylation in GSK3β in HEK-499 compared with control cells after adjustment for GSK3β levels (Figs. 3A and B). These data were confirmed by confocal immunofluorescence (Supplementary Fig. S3C). In contrast, doxycycline-induced Flp-In T-Rex-293 cells had significantly lower levels of T390 p-GSK3β than uninduced cells after normalizing for GSK3β levels (Fig. 3B).

Although Akt/PKB (protein kinase B) is the best characterized kinase that inactivates GSK3β (32, 34), a recent finding has shown that GSK3β can also be equally inactivated through phosphorylation at T390 residue by p38 MAPK (33). Therefore, we assessed p38 expression and activation. We found that the phosphorylated and active form of p38 was significantly increased in RKIP-depleted HEK-499 cells, whereas p38-normalized phospho-p38 levels were significantly reduced in RKIP-induced Flp-In T-Rex compared with uninduced cells (Fig. 3C). Moreover, inhibition of p38 activation in HEK-499 cells, using the chemical inhibitor SB 203580, increased GSK3β stability and reduced its T390 phosphorylation (Fig. 3D). Surprisingly, however, the T390 phosphorylation was only marginally reduced after inhibiting p38 activity. These data suggest that p38 activation, and another yet to be identified factor, may be critical for RKIP-mediated regulation of GSK3β.

Because oxidative stress is a major mechanism known to activate p38α (35), we examined the hypothesis that loss or reduction of RKIP induces oxidative stress. We assessed reactive oxygen species (ROS) levels in HEK-499 cells by determining oxidation of the redox-sensitive dye DHE and MitoSox, a specific probe for detecting mitochondrial superoxide generation. RKIP downregulation augmented both general and mitochondrial ROS production, whereas RKIP overexpression suppressed ROS levels (Figs. 3E and F). In addition, key indices of oxidative stress including H₂O₂ (Supplementary Table S1) were elevated in response to RKIP silencing, and vice versa. As reactive antioxidant capacity exemplified by Cu-Zn-SOD, MN-SOD, and glutathione peroxidase were also increased in HEK-499 cells, whereas overexpression of RKIP induced the opposite effects (Supplementary Table 1). However, the increased antioxidant capacity observed in HEK-499 cells was insufficient to neutralize the intense oxidative stress in these cells. The connection between RKIP and ROS modulation is without precedent, and the exact mechanism/signaling pathway responsible for the induction of ROS by RKIP depletion remains under extensive study in our laboratory, using immortalized and cancer cell lines.

GSK3β is involved in a multitude of cellular functions such as cell division, proliferation, apoptosis, adhesion, motility, and differentiation (36). To achieve such a wide functionality, active GSK3β phosphorylates a variety of substrates inducing their ubiquitination and degradation. The most widely studied substrates include glycogen synthase (which is a target of insulin signaling; ref. 37), SNAIL (38), SLUG (which mediate EMT; ref. 39), cyclin D1 (which regulates cell-cycle progression and is often overexpressed in cancer; ref. 40), and β-catenin (which is a critical component in the Wnt pathway and unduly stabilized in several cancers, most frequently in CRCs; ref. 41). Therefore, we examined the consequences of RKIP on these GSK3β substrates. RKIP downregulation elevated the protein levels of β-catenin, SNAIL, SLUG, and cyclin D1 (Fig. 4). These results may explain the EMT phenotype and the aggressive invasive nature of cancer cells lacking RKIP expression.

To relate these in vitro findings to animal and disease models, we examined the expression of RKIP and GSK3β proteins in RKIP knockout mice and human CRCs, using immunohistochemistry. In wild-type mice, RKIP and GSK3β were heterogeneously expressed in tubuloalveolar prostate gland epithelium but showed a strong positive correlation of coexpression (Fig. 5). In RKIP knockout mice, GSK3β expression was lost in prostate epithelium and bile ducts (Fig. 6). Interestingly, brain and colon tissues expressed GSK3β regardless of RKIP status (Fig. 6). These data indicate that in normal mouse tissues, GSK3β protein expression and stability are controlled by RKIP in a tissue-specific fashion.

Because RKIP loss or reduced expression have been shown to enhance CRC aggressiveness (15, 42), we examined the correlation between RKIP and GSK3β expression in human CRC tissue (Supplementary Fig. S4 and Supplementary Table S2). We observed a significant positive correlation (P = 0.0001) between the expression of RKIP and GSK3β proteins (Supplementary Table S2). This correlation is strikingly apparent in CRCs with heterogeneous RKIP expression (Supplementary Fig. S5). In these cases, the areas of RKIP and GSK3β expression coincided (bottom of slide) whereas other areas lacked the expression of both proteins (top of the slide). These data not only confirm the relationship between RKIP and GSK3β expression protein observed in vitro in both human and mouse tissues but also suggest that other factors can regulate GSK3β protein levels in vivo in a tissue-specific manner.

In summary, our results reveal a pathway how RKIP regulates GSK3β and its substrates through ROS-activated p38 MAPK. Interestingly, this regulation resulted in RKIP enhancing GSK3β activity whereas RKIP inhibited kinases in all previous examples (18, 19, 21, 43). RKIP-related kinase inhibition usually was mediated by disrupting interactions with substrates rather than catalytic activities (44, 45). As we show here, RKIP can also regulate kinases by controlling oxidative stress. This adds another dimension to the pleiotropic actions of RKIP. Oxidative stress is involved in a variety of diseases including cancer and many degenerative and inflammatory syndromes. Therefore, the biological implications of our findings are not limited to cancer but could have far-reaching ramifications and applications to other diseases such as diabetes and Alzheimer's disease in which GSK3β is known to play major regulatory roles (36).

No potential conflicts of interest were disclosed.

We thank the staff of the at the Health Sciences Center, Kuwait University, for their excellent support and technical assistance, Mrs. Betty Thomas in the Proteomics Unit and Sunitha Pramod for their help with the Western blotting, confocal microscopy, and tissue culture, and Drs. J. Klysik and J. Sedivy for provision of the RKIP knockout mice. All authors have seen and approved the manuscript contents.

This study was supported by funds from the Kuwait Foundation for the Advancement of Sciences grant number (2006-1302-07) and Kuwait University grant (MG02/08); the RASOR project (BBSRC and EPSRC); Cancer Research UK; and Science Foundation Ireland grant no. 06/CE/B1129. Research Core Facility is supported by GM01/01 and GM01/05 grants from Research Administration of Kuwait University.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.