A New α5β1 Integrin-Dependent Survival Pathway Through GSK3β Activation in Leukemic Cells

Background Cell survival mediated by integrin engagement has been implicated in cell adhesion-mediated drug resistance. We have recently demonstrated that the activation of glycogen synthase kinase 3 β (GSK3β) is a new pathway supporting the chemoresistance of leukemic cells adhered to fibronectin. Methodology and Principal Findings We show here that in conditions of serum starvation, the fibronectin receptor α5β1 integrin, but not α4β1, induced activation of GSK3β through Ser-9 dephosphorylation in adherent U937 cells. The GSK3β-dependent survival pathway occurred in adherent leukemic cells from patients but not in the HL-60 and KG1 cell lines. In adhesion, activated GSK3β was found in the cytosol/plasma membrane compartment and was co-immunoprecipitated with α5 integrin, the phosphatase PP2A and the scaffolding protein RACK1. PP2A and its regulatory subunit B' regulated the Ser-9 phosphorylation of GSK3β. In adherent leukemic cells, α5β1 integrin but not α4β1 upregulated the resistance to TNFα-induced apoptosis. Both extrinsic and intrinsic apoptotic pathways were under the control of α5β1 and GSK3β. Conclusions and Significance Our data show that, upon serum starvation, α5β1 integrin engagement could regulate specific pro-survival functions through the activation of GSK3β.


Introduction
The glycogen synthase kinase 3b (GSK3b) is a serine/threonine protein kinase that is involved in many physiological processes, playing important roles in glucose metabolism, cell cycle division, cell adhesion and apoptosis. Deregulation of GSK3b activity is implicated in the pathogenesis of neurodegenerative and metabolic disorders, but also in cancer [1]. GSK3b is constitutively active under its Tyr-216 phosphorylated form and regulates many intracellular signaling pathways. At the post-translational level, the function of GSK3b is inhibited through phosphorylation of the Ser 9 residue by other protein kinases, including Akt, in response to insulin and growth factors [2].
Following integrin engagement, both inhibition and activation of GSK3b have been described. GSK-3b is inhibited by Ser-9 phosphorylation by the ILK/Akt and Cdc42/PKCf pathways to promote integrin-mediated cell proliferation or migration, respectively [3,4]. Conversely, cell adhesion to a 3D collagen matrix through a 2 b 1 engagement promotes activation of GSK3b as well as protein phosphatase 2A (PP2A) [5]. PP2A has been previously shown to reactivate GSK3b through dephosphorylation of Ser-9 [6,7]. However, no role has been ascribed to the activated form of GSK3b downstream of integrin engagement.
We have previously shown that GSK3b activation promotes the chemoresistance of adherent leukemic cells on fibronectin or on osteoblasts under serum starvation [8]. The endosteal niche supports chemoresistant leukemic stem cells [9] and is thought to be rich in fibronectin and hypoxic [10]. Adhesion of serum-starved leukemic cells to fibronectin through a 4 b 1 and a 5 b 1 engagement allows both Ser-9 dephosphorylation of GSK3b and NF-kB activation [8]. Others and we have demonstrated that GSK3b can upregulate cell survival through epigenetic and IkB-independent control of NF-kB activity [8,[11][12][13][14]. Strikingly, the anti-apoptotic role of GSK3b has been demonstrated in different tumors and may involve resistance to death receptor-induced apoptosis [15][16][17][18][19][20]. Recently, GSK3b was found associated with DDX3 and c-IAP-1 in a death antagonizing signaling complex at death receptors and the resistance to apoptosis was overcome by GSK3 inhibitors [21]. A mitochondrial-mediated cell death was also found regulated by GSK3 [22].
Adhesion to fibronectin through a 4 b 1 and a 5 b 1 engagement supports cell adhesion-mediated drug resistance (CAM-DR) of many tumors [23]. Different specific fibronectin domains are bound by a 4 b 1 and a 5 b 1 integrins and could each induce opposing effects on cell survival and proliferation [24]. The aim of our study was thus to determine the respective roles of a 4 b 1 and a 5 b 1 in GSK3b activation in serum-starved adherent leukemic cells. Our results demonstrate that a 5 b 1 but not a 4 b 1 regulates a signaling pathway leading to GSK3b activation and cell survival.

Antibodies and pharmacological inhibitors
Monoclonal antibodies against GSK3b, flotillin and RACK1 were from BD Transduction Laboratories. Monoclonal antibodies GSK3a/b, actin and integrin subunits (a 5 , P1D6; a 4 , P4G9) were purchased from Upstate or Biosource International (Camarillo, CA, USA), Sigma and Dako (Carpinteria, CA, USA), respectively. Monoclonal antibodies against a 5 subunit (clone JBS5), Akt and caspases were from Chemicon International, Santa Cruz Biotechnology (Santa Cruz, CA, USA) and Cell Signaling technology (Beverly, MA, USA), respectively. Polyclonal antibodies directed against PP2A-A (catalytic subunit of PP2A) and PP2A tyrosine phosphorylated at position 307 were from Santa Cruz Biotechnology, and those against integrin subunits (a 4 and a 5 ) came from Chemicon International. Polyclonal antibodies directed against PP2A-B' (regulatory subunit of PP2A), cytochrome C, GSK3a/ bserine phosphorylated at position 21/9 and Akt threonine phosphorylated at position 308 were from Cell Signaling Technology. Polyclonal antibody against p85 was from Upstate. Horseradish-peroxydase-conjugated secondary antibodies against mouse, rabbit or goat were from Cell Signalling Technology. Okadaic acid, a PP2A inhibitor, and the GSK3b inhibitor SB216763 were from Sigma. For Western blotting after immunoprecipitation, GSK3b (monoclonal from BD Transduction Laboratories) and P(ser9)GSK3b (polyclonal from Abcam) antibodies have been biotynylated in our laboratory.

Cells and cell culture
The human leukemic cell lines U937, HL-60 and KG1 were purchased from the German Collection of Microorganisms and Cell Cultures (Braunschweig, Germany). U937 and HL-60 cells were grown at 37uC in 5% CO 2 in RPMI-1640, containing 10% FCS, 50 mg/ml penicillin, and 50 mg/ml streptomycin. KG1 cells were grown in the same conditions in IMDM 20% FCS. Bone marrow leukemic cells from patients with acute myeloid leukemia (AML) were obtained upon informed consent and processed for their conservation as described previously [8]. Leukemic samples were characterized at the Hematology Department of Toulouse University Hospital (France), classified along French American British (FAB) classification (FAB0: undifferentiated AML; FAB1: myeloblastic AML; FAB2: myeloblastic with differentiation AML; FAB4: myelomonocytic AML; FAB5: monocytic AML). The samples contained more than 80% leukemic blasts after processing. After thawing, viable cells from patients were checked by blue trypan labelling, resuspended in IMDM, then washed once and quickly used for in vitro experiments.
Transfection of siRNA U937 cells were transfected using the Amaxa nucleofection technology (Amaxa, Koeln, Germany), as indicated in [8]. 6610 6 U937 cells in 100 ml solution V were mixed with 200 nM siRNA GSK3b, 100 nM siRNA PP2A-B' or with 100-200 nM nontargeting siRNA (Dharmacon Inc., Lafayette, CO, USA). For siRNA integrin, two sources have been used to target a 4 and a 5 subunits: Ambion (30 nM) and Qiagen (50 nM). Cells were immediately nucleofected with an Amaxa Nucleofector apparatus (Amaxa, program V01), then transferred into wells containing 37uC prewarmed culture medium in six-well plates. After transfection, cells were cultured from 24 to 96 h before analysing by Western blotting or FACS. Decrease of GSK3b and a 4 integrin subunit was maximal at 48 h and maintained at 72 h whereas decrease of PP2A-B' and a 5 integrin subunit was maximal at 72 h. Therefore, survival tests and Western blot analysis were performed at 72 h post-nucleofection.

Western blotting
For Western blotting, 0.5-1610 6 cells washed in Phosphate Buffer Saline (PBS) were denatured in Laemmli sample buffer. After sonication for 10 seconds and boiling for 10 min, proteins were resolved on polyacrylamide SDS gels (SDS-PAGE) and transferred to nitrocellulose (membrane Hybond-C super, Millipore). The membrane was blocked for 1 h at room temperature in Tris-buffered saline (TBS) containing 5% fat-free milk and then was probed overnight at 4uC with the appropriate monoclonal or polyclonal antibodies in TBS, 0.1% Tween, 3% fat-free milk and 3% Bovine Serum Albumin (BSA, Euromedex). After incubation for 1 h at room temperature with either anti-mouse or anti-rabbit IgG antibody coupled to horseradish peroxidase, or streptavidin-HRP, detection was achieved using a chemiluminescent substrate (SuperSignal, Amersham Pharmacia Biotech).

Survival and adhesion assays
We have previously set-up a protocol to study the survival pathway in leukemic cells strictly dependent on the integrin engagement without extrinsic growth factors that should not be found in the leukemic niche [8]. Thus, adhesion assays and treatments of cells were performed during 5 h of serum starvation followed by re-addition of serum to discard serum deprivationlinked cytotoxicity, and cell viability was measured at 24 h. Since adhesion under serum-starved conditions was required to trigger GSK3b activation downstream of integrin engagement (preliminary experiments not shown), we have thus measured the differential of GSK3b-linked cell survival between suspension and adhesion upon serum-starved conditions. Half of a 96 well microtiter plate (MaxiSorp Immuno Plate, Nunc, Denmark) was coated overnight at 4uC with 40 mg/ml of human fibronectin (Roche Molecular Biochemicals, Mannheim, Germany) in a final volume of 50 ml in PBS, and subsequently incubated with 1% fatty acid-free BSA in PBS to block non specific adhesion sites, 1 h at room temperature. Leukemic cells were diluted to 0.3610 6 /ml, left overnight (cell lines) or immediately processed after thawing (AML cells, protocol described below) and then serum-starved for 1 h, incubated or not with okadaic acid (100 nM) or SB216763 (10 mM). Then, cells were allowed to adhere on fibronectin-coated microtiter 96 well plates (0.8610 5 cells/well) for 1 h at 37uC or maintained in suspension. Where shown, leukemic cells in suspension or adhesion were treated with 10 ng/ml TNFa for 4 hours after which cells were washed and incubated in serumcontaining medium for 24 h at 37uC. Cell viability was then quantified by methyl thiazolyl tetrazolium (MTT) assay (Sigma). In some experiments, leukemic cells were allowed to adhere on a surface coated with a 4 -or a 5 -specific antibodies (clones P4G9, P1D6, JBS5: 1, 0.1 mg/ml and 0.2 mg/ml for optimal adhesion, respectively) as previously described in [8]. Assays were performed in triplicate. For the quantification of cell adhesion, adherent cells were washed in PBS, fixed with a Karnovsky solution and stained with 0.1% crystal violet solution.
For the apoptosis assay, 1610 6 U937 cells treated with control, GSK3b or integrin siRNAs were processed as described for the survival assay and at 5 h of incubation were washed with 1x PBS and then incubated 15 min at room temperature with a FITClabeled Annexin-V/propidium iodide solution (Sigma). These cells were directly analysed in a FACScan (Becton Dickinson) with a sample size of at least 10,000 cells gated on the basis of forward and side scatter. Storing and processing of data was accomplished using FACScan software and allowed the determination of the percentages of living, apoptotic and necrotic cells. Following the protocol described above including adhesion and TNF treatment upon serum starvation, the apoptotic effectors (caspases, cytochrome C) were detected by Western blot at 4 h after re-addition of serum.
In our experiments, we have only used samples from AML patients that did not display cell death after thawing over 10% as checked by blue trypan labeling. After thawing, cells were cultured in IMDM without serum for 2 h and then allowed to adhere on fibronectin. At the end of 4 h adhesion, culture medium was supplemented by SVF to maintain the cells until 24 h. Cell survival of AML blasts from patients was measured at 24 h by MTT labeling. Measurement of apoptosis in AML blasts treated or not by TNFa was realized by FACS using labeling by APO2.7 (APO2.7-PC5 monoclonal antibody from Immunotech, Marseille, France), a mitochondrial membrane protein expressed during the early stages of apoptosis in relation to the release of cytochrome C outside the mitochondria. APO2.7 labeling has been chosen since in our control apoptotic assays performed with daunorubicin we detected interference with the Annexin-V fluorescence (not shown). In our experimental conditions, spontaneous death after thawing was maximally 20% at 24 h (not shown).

Subcellular fractionation and a 5 immunoprecipitation
Culture dishes were coated overnight at 4uC with 40 mg/ml of human fibronectin in a final volume of 10 ml in PBS and subsequently blocked with 1% fatty acid-free BSA in PBS, 1 h at room temperature. 30610 6 cells were serum starved for 1 h at 37uC, then allowed to adhere to fibronectin-coated dishes for 1 h at 37uC, or maintained in suspension.
For immunoprecipitation of a 5 integrin, at the end of adhesion, cells were washed in PBS and lysed in a buffer containing 20 mM Tris HCl pH 8, 130 mM NaCl, 1% Triton X-100, 10% glycerol, orthovanadate and protease inhibitors. After sonication and centrifugation at 13500 g, supernatant was processed for protein quantification, preclearing and incubation overnight with 20 mL a 5 polyclonal antibodies. Then, immunoprecipitates were recovered with protein A sepharose, washed and analysed by Western blotting.

Statistical analysis
Student's t test was used for statistical analysis of n independent experiments realized in vitro.

Results
GSK3b, a 4 b 1 and a 5 b 1 integrins are implicated in cell survival of serum-starved adherent leukemic cells We have previously demonstrated that a 4 b 1 and a 5 b 1 integrins, and the kinase GSK3b, regulate the chemosensitivity of adherent leukemic cells onto fibronectin [8]. Using a siRNA approach, we show that survival of adherent U937 on fibronectin in serumstarved conditions involves both GSK3b and b 1 integrins (a 4 b 1 and a 5 b 1 ) (Fig. 1A). Decreased expression of GSK3b (70610%), of a 5 b 1 (33610%) and of a 4 b 1 (4765%) was assessed by Western blotting (Fig. 1B). None of these siRNA altered the adhesive capacities of leukemic cells (Fig. 1B). Viable cell recovery 24 hours after adhesion assay on fibronectin measured by MTT labeling was increased in adhesion conditions compared to suspension (Fig. 1A, 1866%, p,0.01). GSK3b and a 5 b 1 siRNA induced a 3065% decrease of cell recovery in adhesion (p,0.001), whereas a 4 b 1 siRNA was less potent (1865%, p,0.05). No significant changes occurred in suspension upon treatment with the different siRNAs. Of note, identical results were obtained with two sources of siRNA targeting different sequences of a 4 and a 5 integrin genes (not shown).
Since MTT measurement could be the result of both cell survival and proliferation, apoptotic cell death was assessed after adhesion assay by annexin labeling (Fig. 1A). Whereas MTT measurement at 24 hours may reflect both apoptotic and necrotic processes, annexin labeling was realized at 5 hours to check specifically for the occurrence of apoptosis. A higher level of apoptosis was detected in U937 electroporated with control siRNA comparatively to untreated cells (2062% versus #10%). Adhesion decreased the amount of annexin-positive cells compared to suspension (2464%, p,0.05). Downregulation of a 5 b 1 or a 4 b 1 expression in adherent U937 induced an increase of apoptotic cells compared to control siRNA in adhesion (6965% p,0.01 and 4466% p,0.05, respectively). Knockdown of GSK3b expression increased apoptosis of adherent U937 (8164%, p,0.01). Apoptosis in suspension was not significantly changed after treatment with integrin siRNAs.
Using the pharmacological GSK3b inhibitor SB216763 at a concentration (10 mM) without deleterious quantitative effect on cell adhesion (Fig. 1B), we have further demonstrated that, as well as in U937, the GSK3b-dependent survival pathway occurred in adherent leukemic cells from patients but not in the HL-60 and KG1 leukemic cell lines (Fig. 1C). Accordingly, the activated form of GSK3b (dephosphorylated Ser-9 GSK3b) was increased in adherent U937 and cells from AML patients, but not in adherent HL-60 and KG1 cells (Fig. 1D). Importantly, GSK3b-dependent cell survival was found in adherent AML cells classified along different FAB. However the pro-apoptotic response to SB216763 occurred in 50% AML samples of the cohort (n = 16: 3 FAB0, 4 FAB1, 2 FAB2, 3 FAB4 and 4 FAB5) and was more pronounced in AML from myelomonocytic FAB (Fig. 1C). Furthermore, most of the blasts underwent CAM-DR in vitro and SB216763 abolished it, as previously described [8].
Altogether, these results demonstrate that engagement of a 4 b 1 and a 5 b 1 integrins to fibronectin supports cell survival of serum-starved adherent leukemic cells, potentially through activation of GSK3b via its dephosphorylation.
The inhibitory Ser-9 phosphorylation of GSK3b is differentially regulated by a 4 b 1 and a 5 b 1 integrins To further demonstrate that GSK3b is involved in a 4 b 1 and a 5 b 1 -mediated prosurvival effect, we used siRNA to knockdown integrin expression (Fig. 2A). We have studied consequences of the decrease in expression of each integrin on GSK3b phosphoryla- tion. In serum-starved conditions, a 4 and a 5 siRNA had different effects on the inhibitory Ser-9 phosphorylation of GSK3b (Fig. 2B). Ser-9 phosphorylation of GSK3b was decreased (50%617, p,0.05) upon adhesion to fibronectin compared to suspension cells. Whereas a 5 siRNA abolished the Ser-9 dephosphorylation of GSK3b in adhered cells, a 4 siRNA had no significant effect on Ser-9 phosphorylation. Conversely, in suspension, none of integrin siRNA modified significantly Ser-9 phosphorylation of GSK3b. These results are in favor of GSK3b activation through Ser-9 dephosphorylation after a 5 b 1 engagement onto fibronectin.
Since we have previously shown that adhesion to surfaces coated with anti-a 4 (clone P4G9) or -a 5 (clone P1D6) stimulatory antibodies supported the chemoresistance of U937 cells [8], we studied GSK3b phosphorylation in U937 adhered (with the same efficiency) either to fibronectin or to a 4 or a 5 antibodies (Fig. 2C). Adhesion of U937 to coated anti-a 5 antibodies induced a decrease of Ser-9 phosphorylation of GSK3b compared to suspended cells (60%610, p,0.01), thus mimicking the observations after adhesion of leukemic cells on fibronectin. However, specific engagement of a 4 b 1 onto coated anti-a 4 antibodies did not significantly change the phosphorylation status of GSK3b compared to suspension (10%615). Interestingly, adhesion of U937 to coated anti-a 5 antibody clone JBS5 induced a strong adhesion without cell spreading by contrast with fibronectin and clone P1D6 (not shown and [25]) but was not efficient to trigger GSK3b dephosphorylation (Fig. 2C). Of note, adhesion of U937 on non-specific Ig did not induce changes in phosphorylation of GSK3b nor in cell survival (not shown and [8]).
These data unravel a differential control of Ser-9 phosphorylation of GSK3b by a 4 b 1 and a 5 b 1 integrins, allowing activation of the enzyme. However, upon adhesion of serum-starved leukemic cells on fibronectin, a 5 b 1 alone seems to support GSK3b activation.

Involvement of PP2A in a 5 b 1 -mediated GSK3b activation and cell survival
The phosphatase PP2A is a partner of b 1 integrins in the control of cell survival [26] and regulates Ser-9 phosphorylation of GSK3b [7]. As shown in Fig. 3A, the inhibitory Ser-9 phosphorylation of GSK3b in the cytosolic/membrane fraction was found to be strongly decreased in adhesion compared to suspension. Moreover, the active form of GSK3b (Ser-9 dephosphorylated GSK3b) was increased in a 5 integrin immunoprecipitate from adherent U937 comparatively to suspension and was found associated with PP2A (Fig. 3B). PP2A was poorly phosphorylated on its inhibitory site (Tyr-307) in a 5 immunoprecipitate from adherent U937. Interestingly, adhesion of U937 on fibronectin triggered the association of the scaffolding protein RACK1 with a 5 whereas it decreased the amount of the PI 3kinase regulatory subunit p85 associated with the integrin (Fig. 3B). These data show that GSK3b is co-localized with a 5 integrin in a molecular complex containing phosphatases and kinases potentially implicated in its regulation.
To further demonstrate that PP2A could play a role in GSK3b activation, we used okadaic acid (OA) to inhibit PP2A [26]. The inactive phosphorylated forms of PP2A and GSK3b (phosphoTyr-307 PP2A and phosphoSer-9 GSK3b, respectively) were decreased upon adhesion but restored upon treatment with OA ( Fig. 4A) showing that the activation of the two enzymes were correlated. Of note, under treatment by OA, a decrease of GSK3b expression was constantly measured (Fig. 4A) suggesting its Ser-9 phosphorylation-dependent proteasomal degradation as demonstrated previously [27]. The PP2A regulatory subunit B' (also called B56 or PR61) is responsible for the function of PP2A in cytoskeletal stability [28] and Akt regulation [29], potentially involved in the control of cell survival and GSK3b regulation. Indeed we have measured a decrease of the active form of Akt (Threonine 308 phosphorylated Akt) concomitantly with the activation of PP2A and GSK3b in adherent leukemic cells (Fig. 4A). As shown in Fig. 4B, a decrease of a 5 b 1 and PP2A-B' (40%65) expression by siRNA both triggered an increase in Ser-9 phosphorylation of GSK3b suggesting their roles in keeping GSK3b in an active state. The PP2A inhibitor OA decreased survival of adherent U937 (35%, p,0.05) to the same extent as a 5 b 1 knockdown (Fig. 4C). SiRNA directed against the PP2A regulatory subunit B' induced a moderate but significative survival decrease (14%, p,0.05 statistical apparied test) in adhesion (Fig. 4C). In suspension, leukemic cells were not significantly affected under these conditions (not shown). An increase of AML cell survival was measured upon adhesion on fibronectin (27%66) or a 5 antibody (clone P1D6, 46%63) but not a 4 antibody (clone P4G9, 10%67) compared to suspension. This improvement of AML cell survival upon adhesion was abolished by OA treatment (Fig. 4C). These results show a role for PP2A in GSK3b activation and cell survival of adherent leukemic cells.
Altogether, these data show that PP2A is activated in serumstarved adherent leukemic cells and cooperate with a 5 b 1 integrin in leukemic cell survival through the regulation of GSK3b. a 5 b 1 and GSK3b regulate TNFa resistance and both extrinsic and intrinsic apoptotic pathways in leukemic cells We have previously shown that both a 5 b 1 and a 4 b 1 integrins supported chemoresistance of U937 cells [8]. Since GSK3b activation is a 5 b 1 integrin-dependent in our conditions and has been shown to play a specific pro-survival role through the response to death receptor activation [21], we checked whether a 5 and a 4 integrins could be differentially involved in TNFa response. The incubation of U937 with TNFa in serum-starved suspension conditions induced a 44%63 (p,0.01) decrease in cell survival (Fig. 5A). Adherent U937 were more resistant to this treatment since a decrease of 24%64 (p,0.01) of cell survival was measured. a 5 siRNA as well as GSK3b siRNA, but not a 4 siRNA, abolished the adhesion-dependent resistance to TNFa (Fig. 5A).
The GSK3b-dependent TNFa resistance and NF-kB activation [8] in adherent leukemic cells suggest that both extrinsic and intrinsic apoptotic pathways could be regulated by a 5 b 1 integrin engagement. Indeed the treatment of U937 cells with siRNA directed to a 5 or GSK3b induced a cleavage of caspase 8 and an increase of cytosolic cytochrome C in favor of the activation of both apoptotic pathways in adherent leukemic cells treated by TNFa. As a result of both apoptotic pathways, caspase 3 was cleaved upon this treatment (Fig. 5B). Protection against TNFa confered by a 5 -dependent adhesion and activated GSK3b was confirmed in AML patient (Fig. 5B).
Thus, these data demonstrate that the engagement of a 5 b 1 integrin and GSK3b activation both support resistance to extrinsic and intrinsic pro-apoptotic pathways of serum-starved adherent leukemic cells.

Discussion
In this work, we have demonstrated that adhesion to fibronectin triggers a specific survival signaling pathway in U937 leukemic cells upon serum starvation. Importantly, this survival pathway occurs in leukemic blasts from AML patients and supports their chemoresistance [8]. The survival advantage conferred by adhesion to serum-starved leukemic cells requires the activation of GSK3b. A signaling cascade involving the a 5 b 1 integrin and the phosphatase PP2A is responsible for Ser-9 dephosphorylation and thus activation of GSK3b. GSK3b (P(S9)), and activated form of Akt (P(T308)), after adhesion of U937 on fibronectin 6 okadaic acid (OA, 100 nM) treatment. B-Consequences of a 5 or PP2A-B' siRNA on Ser-9 phosphorylation of GSK3b were studied. C-Adherent U937 transfected with a 5 siRNA or PP2A-B' siRNA underwent a survival assay (MTT) in serum-starved conditions as described for Fig. 1. Adhesion of cells from AML patients was performed on fibronectin or on a 5 antibody (P1D6) and a 4 antibody (P4G9) as described for Fig. 2 to discriminate the respective implication of both integrins in cell survival measured by APO2.7 labeling in the same conditions as for U937. OA (100 nM) was used to inhibit PP2A in U937 and AML blasts. A, B: Data are representative of three independent experiments. In C: U937 n = 3 (mean 6 S.E.M.), AML patients n = 2 (FAB5, mean 6 S.D.), comparison to cells in suspension: *P,0.05. doi:10.1371/journal.pone.0009807.g004 In adherent conditions, co-localization of GSK3b and PP2A with a 5 integrin in the membrane compartment correlated with Ser-9 dephosphorylation of GSK3b suggesting the activation of the enzyme under these conditions. Indeed, inhibition of a 5 b 1 and PP2A by siRNA or pharmacological drugs induced an increase of the Ser-9 phosphorylated inhibited form of GSK3b. The b 1 Figure 5. a 5 integrin and GSK3b regulate the response to TNFa and both extrinsic and intrinsic apoptotic pathways in adherent leukemic cells. A-U937 transfected with a 4 , a 5 or GSK3b siRNA in suspension or adhesion were treated with TNFa (10 ng/ml) for 4 hours and cell survival assays (MTT) were performed as described for Fig. 1. Mean 6 S.E.M. n = 3, *P,0.05 **P,0.01. Right panel shows the efficacy of siRNA directed to GSK3b, a 4 and a 5 integrins. B-Extrinsic (caspase-8) and intrinsic (cytochrome C) or both (caspase-3) apoptotic pathways were studied by Western blotting in TNFa-treated adherent U937 treated with siRNA against a 5 or GSK3b. SiC = Sicontrol, representative of three independent experiments. On the right side is shown cell apoptosis (APO2.7 labeling) measured in TNFa-treated blasts from AML patients in suspension or in adhesion on fibronectin or on P1D6 a 5 antibody as described in the legend of Fig. 2 (n = 2 FAB5, mean 6 S.D., *P,0.05). The treatment with the GSK3b inhibitor SB216763 was peformed as described for Fig. 1 integrin/PP2A pathway of GSK3b activation may be specific of some integrin heterodimers since it has been demonstrated downstream of a 2 b 1 [5], a 5 b 1 but not a 4 b 1 engagement. Moreover, in our experiments, serum starvation was required to trigger the integrin-dependent GSK3 activation (not shown). The a 5 b 1 /GSK3b pathway supports 30% of survival in adherent leukemic cells. Interestingly, we have previously implicated this survival pathway in cell adhesion-mediated drug resistance, where it was shown to allow a 30% increase in survival [8]. Our data also demonstrate that the a 5 b 1 /GSK3b pathway is involved in TNFa resistance, caspase 8 and cytochrome C regulation, and activation of the transcriptional factor NF-kB [8]. Thus, a 5 b 1 -mediated activation of GSK3b modulates both extrinsic and intrinsic apoptosis signaling pathways [15].
The a 5 b 1 /GSK3b pathway could modulate diverse signaling pathways controlling cell survival. ERK activation has been described to control cell survival that is linked to integrin engagement upon serum starvation [30] or Fas stimulation [26]. Moreover, GSK3b regulates MEKK1 demonstrating its implication in the stress-activated protein kinase pathway [31]. In our experiments, ERK or p38 inhibition did not influence cell survival (not shown). Interestingly, the activity of c-Jun N-terminal kinase (JNK), which is found constitutively activated in most patients with acute myeloid leukemia, has been correlated with a «multidrug» anthracycline resistance [32] and controls fibronectin survival signaling under serum-starvation conditions [33]. Our preliminary data suggest that JNK is activated in adherent U937. It remains to determine whether, in our experimental model, JNK activation is modulated by GSK3b.
Our data show that adhesion of U937 on fibronectin triggers the association of the scaffolding protein RACK1 with a 5 b 1 . Interestingly, RACK1 has been described as a signal integrator between growth factor receptor and b1 integrin [34]. PP2A and the PI 3-kinase regulatory subunit p85 are among the proteins whose recruitment and dissociation are modulated by RACK1. In our experiments, both an increase of activated GSK3b and a decrease of p85 were observed in a 5 immunoprecipitate from adherent U937. This result suggests that the integrin-dependent activation of GSK3b could result from both increased PP2A and decreased PI 3-kinase/Akt activities [5]. Accordingly we measured a concomitant decrease of the active form of Akt with GSK3b activation in adherent leukemic cells. Thus, survival of U937 in starved conditions may be linked to a quiescence status with decreased proliferative and migration capacities [35]. However, the organization of the cytoskeleton seems to play a key role for the a 5 b 1 -dependent activation of GSK3b since adhesion through the a 5 b 1 antibody clone JBS5 impaired specifically cell spreading and did not trigger GSK3b activation. Whether RACK1 directly regulates PP2A activity or targets GSK3b and PP2A catalytic/ regulatory subunits to specific locations is an open question. Interestingly, it has been shown that RACK1 is a component of the signaling pathway of the p55 TNF receptor [36] and is implicated in the resistance to apoptotic stimuli in hematopoietic cells [37]. Thus, RACK1 could regulate the death-antagonizing complex involving GSK3b at TNF receptor [21].
We have previously shown [8] that engagement of both a 5 b 1 and a 4 b 1 supported cell-adhesion mediated drug resistance of U937. However, a 5 b 1 alone was shown to activate GSK3b and TNFa resistance. Our unpublished data demonstrate that a 4 b 1 integrin controls U937 survival in adhesion through the tyrosine kinase Pyk2 activation. It suggests that a 5 b 1 /GSK3b and a 4 b 1 / Pyk2-dependent pro-survival pathways could cooperate through the activation of specific pathways of resistance to extrinsic and intrinsic pro-apoptotic pathways. Interestingly, our preliminary data show that the a 4 b 1 /Pyk2 cell survival pathway in adherent U937 involves PI 3-kinase activation and Bcl-xL expression. Since we have shown that the GSK3b-dependent survival pathway occurred in U937 but not in HL-60 and KG1 cell lines, it could be interesting to compare their a 5 or a 4 -dependent adhesive capacities. Notably, and by contrast with U937, RACK1 increase and PI 3-kinase subunit p85 decrease were not observed in a 5 immunoprecipitates from adherent HL-60 (not shown). However, we demonstrate that AML cells from patients classified along different FAB trigger mostly the GSK3b-dependent survival pathway upon adhesion onto fibronectin. Except a myelomonocytic phenotype, this survival pathway was not correlated with other clinico-biological parameters in patients.
In conclusion, we propose GSK3b activation as a new adhesion-dependent cell survival pathway that is regulated by the engagement of specific integrin heterodimers. It could control a pro-survival response to death receptors but also intrinsic prosurvival pathways under stress conditions. Importantly, the prosurvival GSK3b-dependent pathway may represent a new therapeutic target in cancer cells whose resistance to therapy is supported by cell adhesion.