|Home | About | Journals | Submit | Contact Us | Français|
Dysregulation of the receptor tyrosine kinase fibroblast growth factor receptor 3 (FGFR3) plays a pathogenic role in a number of human hematopoietic malignancies and solid tumors. These include t(4;14) multiple myeloma associated with ectopic expression of FGFR3 and t(4;12)(p16;p13) acute myeloid leukemia associated with expression of a constitutively activated fusion tyrosine kinase, TEL-FGFR3. We recently reported that FGFR3 directly tyrosine phosphorylates RSK2 at Y529, which consequently regulates RSK2 activation. Here we identified Y707 as an additional tyrosine in RSK2 that is phosphorylated by FGFR3. Phosphorylation at Y707 contributes to RSK2 activation, through a putative disruption of the autoinhibitory αL-helix on the C terminus of RSK2, unlike Y529 phosphorylation, which facilitates ERK binding. Moreover, we found that FGFR3 interacts with RSK2 through residue W332 in the linker region of RSK2 and that this association is required for FGFR3-dependent phosphorylation of RSK2 at Y529 and Y707, as well as the subsequent RSK2 activation. Furthermore, in a murine bone marrow transplant assay, genetic deficiency in RSK2 resulted in a significantly delayed and attenuated myeloproliferative syndrome induced by TEL-FGFR3 as compared with wild-type cells, suggesting a critical role of RSK2 in FGFR3-induced hematopoietic transformation. Our current and previous findings represent a paradigm for tyrosine phosphorylation-dependent regulation of serine-threonine kinases.
Fibroblast growth factor (FGF) receptor 3 (FGFR3) belongs to a family of receptor tyrosine kinases (RTKs) responding to FGF with four members (FGFR1 to -4) that share a conserved structure and a high level of amino acid homology (56 to 71% overall identity) (15). Each FGFR is composed of an extracellular ligand-binding domain, a transmembrane domain, and a split cytoplasmic tyrosine kinase domain (17). FGFR3 is activated by oligomerization induced by ligand binding, followed by autophosphorylation at multiple tyrosine residues that are believed to provide docking sites for signaling factors through their respective Src homology 2 (SH2) phosphotyrosine binding domains. This, in turn, is required for stimulation of the intrinsic catalytic activity and activation of downstream signaling modules, including the phosphatidylinositol 3-kinase (PI3K)/AKT and phospholipase C-γ (PLC-γ) pathways (13, 32).
The t(4;14) translocation has been identified in approximately 15% of multiple myeloma (MM) patients (3, 4) and results in overexpression of wild-type (WT) FGFR3. MM is among the most common hematologic malignancies in patients over 65 years of age and is currently incurable. The t(4;14) MM is associated with a particularly poor clinical prognosis using conventional treatment strategies. In some t(4;14) MM cases, the translocated FGFR3 gene contains an activating mutation, K650E, that, when present in the germ line, causes thanatophoric dysplasia type II (TDII) (30). Moreover, expression of a constitutively activated fusion tyrosine kinase, TEL-FGFR3, is associated with t(4;12)(p16;p13) acute myeloid leukemia (33). Thus, the pathogenic role of FGFR3 makes it an attractive therapeutic target. We and others have demonstrated the therapeutic efficacy of small molecule tyrosine kinase inhibitors, including PKC412, PD173074, SU5402, and TKI258, which effectively inhibit FGFR3, in murine hematopoietic Ba/F3 cells; FGFR3-expressing t(4;14)-positive human MM cell lines (HMCLs), including KMS11, KMS18, and OPM-2; and as in bone marrow (BM) transplant (BMT) and xenograft murine models (2, 12, 23, 31).
FGFR3 has been demonstrated to activate multiple signaling components. Identification and characterization of critical downstream signaling effectors of FGFR3 will inform not only molecular mechanisms underlying FGFR3-induced transformation but also development of novel therapeutic strategies to treat FGFR3-associated human malignancies. We have performed mass spectrometry-based phospho-proteomics studies (18) to comprehensively identify potential downstream substrates/effectors that are tyrosine phosphorylated in hematopoietic cells transformed by oncogenic FGFR3 mutants. We identified p90 ribosomal S6 kinase 2 (RSK2) as a substrate and signaling effector of FGFR3. RSK family members are Ser/Thr kinases and substrates of the Ras/extracellular signal-regulated kinase (ERK) pathway. RSK plays an essential role in a number of cellular functions, including regulation of gene expression, cell cycle, and survival by phosphorylating downstream substrates/signaling effectors.
While the C-terminal kinase (CTK) domain (CTD) is believed to be responsible for autophosphorylation and the N-terminal kinase (NTK) domain phosphorylates exogenous RSK substrates (8), the precise mechanism of RSK activation remains elusive. The current model suggests that ERK-dependent activation of RSK contains a series of sequential events. First, inactive ERK binds to the C terminus of RSK in quiescent cells, and this interaction is an absolute requirement for activation of RSK (10, 25, 29). Upon mitogen stimulation, ERK becomes activated and phosphorylates RSK at Thr577 (murine RSK2 numbering) in the activation loop of the CTD and Ser369 and Thr365 in the linker region between the two kinase domains, leading to activation of the RSK CTD. Second, activation of the CTD results in autophosphorylation of S386 in the linker region, which provides a docking site for 3-phosphoinositide-dependent protein kinase 1 (PDK1) (9). Third, PDK1 in turn phosphorylates Ser227 in the activation loop of the NTK domain, allowing RSK to phosphorylate its downstream targets (16). Finally, the activated NTK domain autophosphorylates Ser749 at the RSK CTD, which results in dissociation of active ERK from RSK (25).
We recently proposed a novel two-step model in which leukemogenic FGFR3 activates RSK2 by both tyrosine phosphorylation of RSK2 and activation of the MEK/ERK pathway (18). The first step involves tyrosine phosphorylation at Y529 of RSK2 by FGFR3, which facilitates binding of the inactive form of ERK to RSK2 in the initial step of ERK-dependent RSK2 activation. This binding, which is required for phosphorylation and activation of RSK2 by ERK, in turn promotes the second step where ERK is activated via the Ras/Raf/MEK/mitogen-activated protein kinase (MAPK) pathway downstream of FGFR3, leading to phosphorylation and activation of RSK2 by ERK. We also demonstrated that phosphorylation at Y529 of RSK2 is not a specific requirement of FGFR3 signaling in hematopoietic cells and that it may represent a more general mechanism for RSK2 activation (19). We found that upon treatment of EGF, RSK2 is tyrosine phosphorylated at Y529 and activated in 293T and COS7 cells that do not express FGFR3. However, this phosphorylation was not mediated directly by activated receptor tyrosine kinase epidermal growth factor (EGF) receptor (EGFR), but by Src tyrosine kinase family members. Phosphorylation at Y529 by Src facilitates ERK binding to RSK2, which represents a general requirement for RSK2 activation by EGF through the MEK/ERK pathway.
In this paper, we identified an additional tyrosine site in RSK2, Y707, that when phosphorylated by FGFR3 contributes to RSK2 activation. Phosphorylation at Y707 may disrupt the autoinhibitory αL-helix in the C terminus of RSK2 to activate RSK2 CTD (21, 24, 28), unlike Y529 phosphorylation, which facilitates ERK binding (18). Moreover, we found that FGFR3 interacts with RSK2 and that this association is critical for FGFR3-dependent tyrosine phosphorylation at Y529 and Y707 of RSK2 as well as its subsequent activation. Furthermore, we demonstrated that RSK2 is important for FGFR3-induced hematopoietic transformation in vivo in our murine model of leukemia.
Phosphopeptides were prepared using the PhosphoScan kit (Cell Signaling Technology, Inc. [CST]), and studies were performed as previously described (18). Purified recombinant RSK2 was incubated with active recombinant FGFR3 in an in vitro kinase assay as described previously (18). Endogenous RSK2 was enriched from 2 × 106 to 3 × 106 OPM1 cells by immunoprecipitation using anti-RSK2 antibody. The RSK2 samples were applied to sodium dodecyl sulfate-polyacrylamide gel electrophoresis followed by Coomassie blue staining. RSK2 protein bands were excised and treated with trypsin, followed by analysis using liquid chromatography coupled with mass spectrometry with equipment from CST. Tandem mass spectra were collected in a data-dependent manner with an LTQ ion trap mass spectrometer (ThermoFinnigan).
The myc- and glutathione S-transferase (GST)-RSK2 constructs have been described previously (18). Point mutations of RSK2 were generated by using QuikChange-XL site-directed mutagenesis kit (Stratagene, La Jolla, CA). Ba/F3 cells were cultured in RPMI 1640 medium in the presence of 10% fetal bovine serum (FBS) and 1.0 ng/ml interleukin-3 (IL-3) (R & D Systems, Minneapolis, MN). Human myeloma cell lines (HMCLs) were cultured in RPMI 1640 medium with 10% FBS. 293T cells were cultured in Dulbecco's modified Eagle's medium with 10% FBS. RSK2-expressing Ba/F3 cell lines were generated by retroviral transduction as described previously (18).
Phospho-Tyr antibody pY99 and antibodies against RSK2 and FGFR3 were from Santa Cruz Biotechnology, Santa Cruz, CA; antibodies against myc, ERK, and phospho-RSK (Ser380) were from CST, Danvers, MA; antibodies against GST and β-actin were from Sigma, St. Louis, MO. Specific antibodies against phospho-RSK2 (Y529 and Y707) were generated by CST. Polyclonal antibodies were produced by immunizing rabbits with a synthetic phosphopeptide (keyhole limpet hemocyanin coupled) corresponding to residues surrounding Y529 or Y707 of human RSK2. The amino acid sequence of the peptide for the p-RSK2-Y529 antibody development was CTITKTVEyLHAQG, and that for the peptide to develop the p-RSK2-Y707 antibody was CGAMAATySALNR. Antibodies were purified by protein A and peptide affinity chromatography at CST.
Coimmunoprecipitation (co-IP) and GST pull-down assays have been previously described (18). The S6 peptide kinase assay was carried out according to the manufacturer's protocol (Upstate Biotechnology) using RSK2 immunoprecipitates as previously described (18). To determine the ability of FGFR3 to phosphorylate RSK2, 500 ng of purified recombinant RSK2 CTD variants was incubated with 500 ng of recombinant active FGFR3 (Invitrogen, Carlsbad, CA) in 10 mM HEPES (pH 7.5), 150 mM NaCl, 1 mM dithiothreitol, 0.01% Triton X-100, 10 mM MnCl2, and 200 μM ATP for 30 min at 30°C. Phosphorylation of Y707 RSK2 was detected by specific phospho-antibody.
C57BL/6 RSK2−/− mice (20) were generously provided by Karsenty and Hanauer. The Lin− Sca-I+ c-Kit+ subpopulations in BM cells of wild-type (WT) and RSK2−/− mice were analyzed with a fluorescence-activated cell sorter (FACS). In brief, BM cells were obtained from 6-week-old WT and RSK2−/− mice. Cells were then stained with Lin-phycoeryrthrin, ScaI-fluorescein isothiocyanate, c-Kit-allophycocyanin (BD Biosciences, San Jose, CA) in phosphate-buffered saline (PBS) containing 3% FBS and 0.1% NaN3 for 30 min, washed with PBS, and analyzed on a FACS (BD FACS CantoII). In the CFU assay, BM cells were prepared from 5-fluorouracil (150 mg/kg of body weight)-treated C57BL/6 mice in RPMI medium. Collected cells were infected with retrovirus carrying pMSCV-neomycinr-TEL-FGFR3 by spin infection (2) in RPMI 1640 medium containing 6 ng/ml recombinant murine IL-3 (rmIL-3; R&D Systems), rmSCF (10 ng/ml; R&D Systems), recombinant human IL-6 (10 ng/ml; R&D Systems), and 10% FBS. Approximately 3 × 104 transduced BM cells were added into 3 ml of Methocult M3231 (StemCell Technologies, Vancouver, Canada) containing neomycin (1 mg/ml; Invitrogen, Carlsbad, CA) and plated in 35-mm dishes in triplicate. The cells were cultured at 37°C for 7 days, and the number of colonies was scored. Murine BMT assays were performed as described previously (27). Animals were carefully monitored under the auspices of institutionally approved protocols for the humane care of animals. WT or RSK2−/− C57BL/6 mice were used as donors. Donor BM cells were transduced with retroviral supernatant carrying the MSCV2.2-Gateway-IRESGFP-TEL-FGFR3 construct (2). A total of 1 × 106 cells/0.5 ml were injected into the lateral tail veins of lethally irradiated (2 × 550 cGy) syngeneic C57BL/6 recipient mice. Diseased TEL-FGFR3 BMT mice were examined each day and sacrificed at the first signs of morbidity, including scruffy coat, lethargy, weight loss, leukocytosis, and splenomegaly palpable beyond the midline. White blood cell counts and weights of organs including spleen and liver were recorded at time of necropsy. The in vivo homing assay was performed as described previously (5, 6). In brief, BM cells from 6-week-old RSK2−/− and WT mice were isolated and transduced with retroviral supernatant carrying the MSCV2.2-Gateway-IRESGFP-TEL-FGFR3 construct (2). A total of 5 × 106 events were acquired for green fluorescent protein (GFP)-positive cells on a FACS (BD FACS CantoII) to determine the retroviral infection efficiency. A total of 0.8 × 106 infected cells were injected to each recipient mouse. After injection (14 to 24 h), BM cells were isolated from recipients and GFP-positive cells were analyzed on a FACS (BD FACS CantoII).
Histopathologic analyses were performed as described previously (26). Prior to flow cytometric analysis, cell samples of single-cell suspensions were washed in the staining buffer (PBS with 0.1% NaN3 and 0.1% bovine serum albumin) and stained for 20 min on ice with combinations of labeled monoclonal antibodies recognizing Gr-1 and Mac-1 (BD Biosciences, San Diego, CA). After washing, the cells were resuspended in staining buffer containing 0.5 μg/ml 7-amino-actinomycin D (BD Biosciences) to allow discrimination of nonviable cells, and flow cytometric analysis was done on a FACSCalibur cytometer (BD Biosciences). At least 10,000 events were acquired, and the data were analyzed using CellQuest software (version 3.3). The results are presented as dot plots of viable cells selected on the basis of scatter and 7-amino-actinomycin D staining.
We recently proposed a novel two-step model that leukemogenic FGFR3 activates RSK2 by both assisting inactive ERK binding via direct tyrosine phosphorylation of RSK2 at Y529 and activating the MEK/ERK pathway (18). We also found that another tyrosine residue, Y707, is phosphorylated in human t(4;14) MM OPM1 cells that overexpress the FGFR3 (K650E) TDII mutant by phospho-proteomics and mass spectrometry-based analysis (Fig. (Fig.1A).1A). Further in vitro kinase assay-based studies using recombinant RSK2 (rRSK2) and active FGFR3 identified Y707 as another major phosphorylation site of RSK2 that is directly phosphorylated by FGFR3 (Fig. (Fig.1A).1A). To better understand the role of Y707 in the signaling properties of leukemogenic FGFR3, we generated an antibody that specifically recognizes phospho-Y707 of RSK2. Using this antibody, we observed that GST-tagged WT RSK2 and the Y529F mutant, but not Y707F mutant, were specifically tyrosine phosphorylated at Y707 in 293T cells expressing the constitutively activated TEL-FGFR3 fusion (Fig. (Fig.1B).1B). We also incubated purified rRSK2 CTD proteins with the recombinant, activated FGFR3 kinase domain (Invitrogen) and assayed Y707 phosphorylation using our phospho-Y707-specific RSK2 antibody. As shown in Fig. Fig.1C,1C, the WT RSK2 CTD was tyrosine phosphorylated at Y707 by FGFR3, whereas Y707 phosphorylation was abolished in the RSK2 CTD Y707F mutant. Using a pan-tyrosine phosphorylation antibody, pY99, we observed comparable tyrosine phosphorylation levels of both the rRSK2 WT and Y707F mutant by FGFR3 (Fig. (Fig.1C).1C). This may suggest that FGFR3 phosphorylates RSK2 at multiple sites, including Y707 and Y529 (18), while Y707 may not be a major phosphorylation site of RSK2 by FGFR3.
Moreover, we observed that endogenous RSK2 was phosphorylated at Y707 in not only 293T cells expressing active FGFR3 TDII (K650E) or TEL-FGFR3 mutants (Fig. (Fig.1D)1D) but also FGFR3-expressing, human t(4;14) OPM1 myeloma cells (Fig. (Fig.1E).1E). Furthermore, FGFR3-dependent Y707 phosphorylation was eliminated upon the treatment of OPM1 cells with the FGFR3 inhibitor TKI258, which effectively decreased FGFR3 kinase activation (Fig. (Fig.1E).1E). These data demonstrated that FGFR3-dependent RSK2 Y707 phosphorylation physiologically occurs in t(4;14) myeloma cells and depends on FGFR3 kinase activity. Consistent with these results, phosphorylation of RSK2 Y707 is also observed in 293T cells expressing active FGFR3 TDII or TEL-FGFR3, but not in cells expressing the kinase-dead forms of FGFR3, including the FGFR3 TDII FF4F mutant and TEL-FGFR3 K508R mutant (Fig. (Fig.1F1F).
We previously reported that EGF stimulation activates Src family members, including Src and Fyn, to phosphorylate RSK2 at Y529 and Y707 (19). To determine whether FGFR3 may activate Src to phosphorylate RSK2 at Y529 and Y707, we treated 293T and Ba/F3 cells expressing TEL-FGFR3 with either the FGFR3 inhibitor TKI258 or the Src inhibitor PP2 (Fig. (Fig.1G).1G). We found that treatment with TKI258, but not PP2, resulted in marked reduction of phosphorylation levels of Y529 and Y707 in RSK2 in cells transformed by TEL-FGFR3, suggesting that Src is not required to mediate FGFR3-dependent tyrosine phosphorylation of RSK2.
To further elucidate the role of tyrosine phosphorylation at Y707 induced by FGFR3 in RSK2 activation, we characterized the RSK2 mutants with single Y→A and Y→F substitutions at Y707. Retroviral vectors encoding distinct myc-tagged RSK2 mutants with a puromycin resistance gene were stably transduced into Ba/F3 cells that already stably expressed FGFR3 TDII. myc-RSK2 proteins were immunoprecipitated and assayed for specific phosphorylation at S386 as a measure of RSK2 activation. As shown in Fig. Fig.2A,2A, WT myc-RSK2 was phosphorylated at S386 in cells expressing FGFR3 TDII in the presence of ligand aFGF, whereas S386 phosphorylation was elevated in the RSK2 Y707A mutant that was reported to be constitutively activated (24). In contrast, phosphorylation at S386 was completely abolished in the control myc-RSK2 ΔC20 mutant that does not bind ERK, while myc-RSK2 Y707F demonstrated decreased phosphorylation levels of S386, suggesting that substitution at Y707 attenuates activation of RSK2 induced by FGFR3 TDII (Fig. (Fig.2A2A).
We also tested the kinase activity of the RSK2 Y707F mutant in an in vitro kinase assay. myc-RSK2 variants were immunoprecipitated from cell lysates of their respective Ba/F3 cell lines stably coexpressing FGFR3 TDII. The immunocomplexes were incubated with a specific exogenous S6 peptide substrate in the presence of [γ-32P]ATP. The myc-RSK2 Y707F mutant incorporated significantly less 32P into S6 peptide than did WT myc-RSK2, whereas the negative control myc-RSK2 ΔC20 mutant lost the ability to phosphorylate S6 peptide (Fig. (Fig.2B).2B). As reported previously, RSK2 Y707A demonstrated increased kinase activity (24). These data correlate with our observations of these RSK2 variants for S386 phosphorylation (Fig. (Fig.2A2A).
Inactive ERK interacts with RSK2 in quiescent cells, which occurs prior to and is required for ERK-dependent phosphorylation and activation of RSK2 (25). We previously demonstrated that tyrosine phosphorylation at Y529 by FGFR3 regulates RSK2 activation by facilitating inactive ERK binding (18). Thus, we next tested whether FGFR3-induced phosphorylation at Y707 may regulate RSK2/ERK interaction in a similar way. Ba/F3 cell lines stably expressing FGFR3 TDII and respective myc-RSK2 variants were treated with the MEK1 inhibitor U0126, since active ERK readily dissociates from RSK2 (25). As shown in Fig. Fig.2C,2C, the co-IP results demonstrated that substitution at Y707 in myc-RSK2 does not attenuate inactive ERK binding to RSK2. In contrast, substitution at Y529 results in a decreased ability of RSK2 to interact with inactive ERK. Phosphorylation at Y707 may alternatively regulate RSK2 activation by affecting the structure of the autoinhibitory C-terminal domain of RSK2. As discussed below, we hypothesize that phosphorylation of Y707 may result in disruption of the Y707-S603 hydrogen bond (Fig. (Fig.2D),2D), which was suggested to be essential to stabilize the autoinhibitory αL-helix in the substrate-binding groove of the RSK2 CTD (21).
To further understand the mechanisms underlying FGFR3-dependent phosphorylation of RSK2, we tested whether FGFR3 interacts with RSK2. We performed co-IP experiments in Ba/F3 cells stably expressing FGFR3 TDII or TEL-FGFR3. As shown in Fig. Fig.3A,3A, endogenous RSK2 was detected in immunocomplexes isolated using an FGFR3 antibody. The binding between FGFR3 and RSK2 was further confirmed in successive co-IP experiments using cell lysates from Ba/F3 cells coexpressing myc-tagged RSK2 and FGFR3 TDII or TEL-FGFR3. A myc-tagged truncated PI3K p85 subunit was included as a negative control. FGFR3 TDII and TEL-FGFR3 were found in myc immunocomplexes of RSK2 but not control protein (Fig. (Fig.3B).3B). Moreover, we confirmed interaction between FGFR3 and RSK2 in a GST pull-down assay. GST control or GST-tagged RSK2 was pulled down by beads from transfected 293T cells with coexpression of FGFR3 TDII or TEL-FGFR3. FGFR3 was detected in the complex of bead-bound GST-RSK2 but not the GST control (Fig. (Fig.3C).3C). These three lines of data together demonstrate that FGFR3 associates with RSK2.
Furthermore, we tested whether FGFR3 interacts with RSK2 in the absence of experimental manipulations. We isolated the endogenous RSK2 protein complexes from a group of HMCLs, and FGFR3 was detected in t(4;14)-positive FGFR3-expressing KMS11 and OPM1 cells, but not in control t(4;14)-negative ANBL6 cells that do not express FGFR3 (Fig. (Fig.3D).3D). These data further confirm that the FGFR3-RSK2 association occurs under the physiological conditions in hematopoietic cells transformed by FGFR3.
We next mapped the region of RSK2 that mediates FGFR3 binding. We generated a spectrum of truncated RSK2 mutants, as shown in Fig. Fig.4A.4A. We performed the co-IP experiments using cell lysates from Ba/F3 cells stably expressing TEL-FGFR3 and distinct RSK2 variants. As shown in Fig. Fig.4B,4B, FGFR3 was found in myc immunoprecipitates of WT RSK2 and the truncated mutant RSK2-NL that contains the NTK domain and the linker region. In contrast, no FGFR3 was detected in immunocomplexes of myc-tagged RSK2 NTK (RSK2-N) or CTK (RSK2-C). These data suggest that RSK2 requires the linker region to interact with TEL-FGFR3.
We then identified the minimal region of RSK2 that is required for FGFR3 and RSK2 association. We generated more truncated RSK2-NL mutants with further deletion of the linker region (Fig. (Fig.4A).4A). 293T cells were cotransfected with these truncated RSK2 mutants and TEL-FGFR3. Co-IP experiments demonstrated that FGFR3 interacts with WT RSK2 and RSK2-NL, whereas binding is drastically decreased upon deletion of amino acids 334 to 421. In contrast, FGFR3-RSK2 association was completely abolished when five additional amino acids were further deleted, including T329, I330, D331, W332, and N333 (Fig. (Fig.4C).4C). These data suggest that FGFR3 may bind to a minimal region including the five residues at positions 329 to 333 of the linker of RSK2.
We next examined whether these five residues are required for FGFR3 binding. 293T cells were cotransfected with FGFR3 TDII along with an RSK2 Δ329-333 mutant with a deletion of residues from T329 to N333 (Fig. (Fig.5A).5A). The co-IP results showed that deletion of these five amino acids in RSK2 abolished binding of FGFR3 TDII (Fig. (Fig.5B),5B), whereas deletion of the 20 amino acids that mediate ERK binding in the control truncated mutant RSK2 ΔC20 did not affect FGFR3 binding (Fig. (Fig.5B).5B). These results are consistent with our previous observation using truncated RSK2 constructs.
We next tested whether FGFR3 binding is crucial for RSK2 activation in the cells expressing FGFR3. Using 293T cells coexpressing TEL-FGFR3 and various RSK2 constructs, we observed that WT RSK2 was phosphorylated at S386 and activated, whereas the S386 phosphorylation was abolished in the RSK2 Δ329-333 mutant that does not interact with TEL-FGFR3 (Fig. (Fig.5C).5C). This result suggests that deletion of the residues at 329 to 333 in RSK2 linker region attenuates TEL-FGFR3 interaction as well as RSK2 activation.
We further determined which amino acid is critical to mediate FGFR3 binding, which might consequently lead to RSK2 activation. We generated a series of RSK2 mutants harboring distinct alanine substitutions at each of the five residues, including T329A, I330A, D331A, W332A, and N333A. 293T cells transfected with TEL-FGFR3 and RSK2 mutants harboring distinct point mutations were cultured in media in the absence of serum for 4 h prior to harvest, followed by co-IP and Western blotting using specific antibodies that exclusively recognize phospho-S386, phospho-Y529, or phospho-Y707 of RSK2. As shown in Fig. Fig.5D,5D, we found that WT RSK2 interacts with FGFR3 and is phosphorylated at Y529, Y707, and S386. In contrast, substitution at W322 and deletion of the five amino acids from T329 to N333 abolished phosphorylation at Y529 and Y707, as well as S386 phosphorylation of RSK2, an index of RSK2 activation. Substitutions at I330, D331, and N333 also resulted in decreased interaction between RSK2 and FGFR3, accompanied with decreased phosphorylation at Y707 and S386, whereas phosphorylation of Y529 appeared not affected in I330A, D331A, and N333A mutants. In contrast, mutation at T329 did not affect phosphorylation at Y529, Y707, or S386. To determine whether mutation of W332 specifically disrupts FGFR3-mediated RSK2 activation, we treated 293T cells expressing WT myc-RSK2 or W332A with EGF that activates RSK2 independent of FGFR3 (19). EGF stimulation activated RSK2 W332A to a comparable level to WT RSK2 as assessed by the phosphorylation level of Ser386 (Fig. (Fig.5E).5E). This supports our observation that W322 is specifically required for FGFR3 binding to RSK2 and mediates RSK2 activation by FGFR3.
Consistent with these observations, in the in vitro kinase assay, we observed that substitution at W322 and deletion of the five residues from T329 to N333 resulted in the greatest reduction in RSK2 activation (Fig. (Fig.5F).5F). In addition, mutations at I330 and D331 also resulted in marked decrease in RSK2 activation, whereas substitutions at T329 and N333 had minimal effect on RSK2 activation in this in vitro RSK2 kinase assay (Fig. (Fig.5F).5F). These data together suggest that FGFR3-dependent phosphorylation and activation of RSK2 may involve several sequential events and that binding of FGFR3 may be the initial step prior to phosphorylation at Y529 and Y707 that subsequently leads to S386 phosphorylation and activation of RSK2. Phosphorylation at either Y529 or Y707 appears to contribute to RSK2 activation and S386 phosphorylation to a certain level. Substitution at W332 resulted in complete loss of FGFR3-RSK2 interaction as well as phosphorylation at Y529 and Y707, which may subsequently attenuate RSK2 activation.
We next examined whether RSK2 is required for the in vitro transforming activity of FGFR3 in primary hematopoietic cells. We performed a myeloid CFU assay using the TEL-FGFR3 fusion tyrosine kinase, which was identified in acute myeloid leukemia harboring a chromosomal translocation t(4;12)(p16;p13) (33). Primary BM cells from WT C57BL/6 mice were transduced by retroviruses containing constructs encoding TEL-FGFR3, with a neomycin-resistant gene as a selection marker. Cells were cultured in methylcellulose containing neomycin in the presence or absence of RSK inhibitor fmk (6 μM), and the numbers of individual myeloid colonies were scored after 7 days. As shown in Fig. Fig.6A,6A, cultured progenitor cells transduced with TEL-FGFR3 formed individual colonies, and no significant alteration was observed in the numbers of colonies formed by cells cultured in the presence or absence of fmk treatment. However, inhibition of RSK2 by fmk effectively decreased the sizes of colonies compared with the sizes of the colonies formed by cells without fmk treatment. Similar results were obtained using TEL-FGFR3-transformed BM cells from WT or RSK2−/− C57BL/6 mice; knockout of RSK2 affects the sizes of colonies but not the colony numbers (Fig. (Fig.6B).6B). Together, these data suggest that RSK2 is probably required for proliferation of TEL-FGFR3-transformed hematopoietic progenitors in myeloid CFU assays but may be dispensable for initiation of TEL-FGFR3-induced transformation in myeloid cells.
In order to examine the role of RSK2 in TEL-FGFR3-induced hematopoietic transformation in vivo, we next performed a BMT assay using TEL-FGFR3. TEL-FGFR3 was retrovirally transduced into donor BM cells from either WT C57BL/6 mice or mice that are genetically deficient of RSK2 (RSK2−/−), and the transduced cells were subsequently injected into lethally irradiated syngeneic WT C57BL/6 recipient mice. As shown in Fig. Fig.7A,7A, RSK2 knockout does not affect cell numbers of the hematopoietic stem cell subpopulation characterized as Lin− c-Kit+ Sca-1+. We observed that the infection efficiencies of the retrovirus-carrying pMSCV-IRESGFP-TEL-FGFR3 construct are similar between WT and RSK2-null BM cells (Fig. (Fig.7B).7B). We also determined the initial homing efficiency of the TEL-FGFR3-expressing WT and RSK2 BM cells, and both groups of BM cells showed similar homing efficiencies in the BMT recipient mice (Fig. (Fig.7C7C).
As we previously reported (2), all of the mice receiving WT BM cells transduced by TEL-FGFR3 developed a rapidly fatal myeloproliferative disease (MPD) characterized by marked splenomegaly and a peripheral blood leukocytosis comprised predominantly of mature granulocytes (n = 7) (Fig. 7D to I). Mice receiving RSK2-deficient (RSK2−/−) BM cells transduced by TEL-FGFR3 also developed signs of myeloproliferation; however, these mice had a statistically significant prolongation in survival, compared with mice receiving WT BM cells expressing TEL-FGFR3 (Fig. (Fig.7D).7D). There was a significant decrease in spleen weight in the RSK2−/− cohort (Fig. (Fig.7E),7E), indicative of an attenuated MPD state in these animals, compared with WT BMT mice. This notion was further confirmed by the flow cytometric analysis that showed reduced numbers of mature neutrophils that were positive for the late myeloid markers Gr-1 and Mac-1 in spleen samples of representative mice transplanted with TEL-FGFR3-transformed RSK2−/− BM cells (50.6%), compared with TEL-FGFR3-expressing WT BM-transplanted animals (78.9%) (Fig. (Fig.7F7F).
Histopathologic examination of tissue samples from TEL-FGFR3 BM-transplanted mice demonstrated markedly hypercellular BM with a predominance of mature myeloid forms and frequent number of admixed histiocytes and macrophages (Fig. (Fig.7G),7G), a perturbation of normal splenic architecture with loss of white pulp and expansion of the red pulp by a prominent population of maturing myeloid forms (Fig. (Fig.7H),7H), and extensive myeloid cell infiltration in livers (Fig. (Fig.7I).7I). In contrast, although histologic evidence of myeloproliferation was evident in BM, spleen, and liver, the extent and degree of MPD were significantly reduced in these organs from TEL-FGFR3-expressing RSK2−/− BM-transplanted animals.
Our data support a multistep model by which FGFR3 activates RSK2 and mediates transformation signals in hematopoietic cells. The initial step involves FGFR3 interacting with RSK2, followed by tyrosine phosphorylation at multiple tyrosine residues, including Y529 and Y707 of RSK2 by FGFR3, which contribute to RSK2 activation. These modifications in turn promote the final step that FGFR3-activated ERK phosphorylates and actives RSK2 as we reported previously (18).
Moreover, our in vivo murine BMT assay demonstrated that RSK2 plays an important role in leukemogenic TEL-FGFR3-induced MPD. Our findings suggest that RSK2 may be involved in FGFR3-induced pathogenesis and disease progression in related hematopoietic malignancies. Furthermore, our data also suggest that targeting RSK2 may attenuate leukemogenic FGFR3-induced hematopoietic transformation in vivo. Because activating mutations of FGFR3 have also been identified in human bladder and cervical carcinomas (1), our findings may have therapeutic implications with regard to solid tumors associated with dysregulation of FGFR3.
RSK2−/− mice have decreased bone mass due to the critical role of RSK2 in osteoblast differentiation (7, 34). However, RSK2−/− mice have a normal life span and no histologic or metabolic evidence of internal organ dysfunction (7, 20, 34). Recently, Lin et al. (20) demonstrated that RSK2 is dispensable for homeostatic proliferation of normal Gr-1+ cells and Mac-1+ cells. We also observed that genetic deficiency of RSK2 does not affect the stem cell subpopulation in RSK2-null mice compared with WT mice (Fig. (Fig.7A).7A). Therefore, the less aggressive disease phenotype in TEL-FGFR3-induced MPD using RSK2-deficient BM cells in BMT mice is most likely due to impairment of RSK2-mediated signal transduction rather than abnormalities in the target cell populations. Such animal models provide a microenvironment with complete depletion of RSK2, which has advantages over other techniques, such as expression of endogenous inhibitors or dominant-negative mutants. The role of RSK2 in TEL-FGFR3-induced MPD is more likely to be associated with disease development and progression than with disease initiation. Knockout of RSK2 does not affect the TEL-FGFR3-induced MPD initiation but significantly extended latency of the TEL-FGFR3-transplanted mice and resulted in attenuated MPD burden in these mice (Fig. (Fig.7).7). Consistent with these observations, in the CFU experiments, the numbers of myeloid colonies were not affected using TEL-FGFR3-transduced hematopoietic progenitors with either knockout of RSK2 or inhibition of RSK2 by fmk treatment, compared with WT BM cells. However, knockout or inhibition of RSK2 effectively decreased the sizes of colonies (Fig. (Fig.6).6). Together, these data suggest that RSK2 is more likely to be involved in the proliferation of TEL-FGFR3-transformed myeloid cells than the initiation of TEL-FGFR3-dependent myeloid transformation in vitro and in vivo.
Tyrosine phosphorylation at Y529 may provide an additional docking site to promote the binding of inactive ERK to the C terminus of RSK2 (18). Future detailed structural studies would illuminate this process. Y707 is localized at the C-terminal tail of RSK2 (697HLVKGAMAATYSALNR712). This region represents a conserved putative autoinhibitory α-helix, which has been identified in calmodulin-dependent protein kinase 1 (CaMK1) to interact with the substrate-binding groove of the catalytic domain and inhibit substrate binding, although not in the classical pseudosubstrate mode of autoinhibition (11, 22, 28). The secondary structure prediction and alignment revealed that RSK2 Y707 is similar to the position of F298 in CaMK1 that is buried in the hydrophobic pocket of the substrate-binding groove. In CaMK1, this residue must be removed from the hydrophobic pocket to allow the proper orientation of the substrate. Calmodulin binding likely disrupts the interaction between the autoinhibitory α helix and the substrate-binding groove, reducing the ability of the α-helix to compete for substrate binding. Truncation of the autoinhibitory α-helix to remove F298 resulted in constitutively active CaMK1 (22). Interestingly, mutation of Y707 to alanine or truncation of the α-helix in RSK2 similarly resulted in significant autophosphorylation of S386 (24, 28).
Recently, structural studies of the CTD of RSK2 crystal revealed that disrupting the Y707-S603 hydrogen bond promotes displacement of the autoinhibitory αL-helix from the catalytic groove and leads to CTK activation (21). The authors proposed that ERK docking to the C terminus of RSK2 may result in disruption of the Y707-S603 hydrogen bond and displace the αL-helix from its inhibitory position. It is not inconceivable that phosphorylation of Y707 could have a similar destabilizing effect on the Y707-S603 hydrogen bond, with much the same rationale, resulting in alteration of the structure of the autoinhibitory αL-helix and relieving the substrate-binding groove. Therefore, our findings suggest that FGFR3-dependent phosphorylation at Y529 and Y707 may regulate RSK2 activation due to different mechanisms, where Y529 phosphorylation facilitates inactive ERK binding (18, 19) while phosphorylation at Y707 may disrupt the autoinhibitory αL-helix.
As shown in Fig. Fig.2D,2D, in addition to the Y707-S603 interaction, Y707 also packs against K541. We hypothesize that such hydrophobic contact may stabilize the autoinhibitory αL-helix in the substrate-binding groove. Mutation of Y707 to alanine may abolish not only the hydrogen bond between Y707 and S603 but also the hydrophobic contact between Y707 and K541. While mutation of Y707 to phenylalanine will remove the hydrogen bond, in contrast to Y707A, the hydrophobic packing to the aliphatic region of K541 is probably not lost. This may explain the reduced activity of the RSK2 Y707F mutant compared with WT RSK2 and the Y707A mutant.
Phosphorylation at Y707 of RSK2 has also been identified by mass spectrometry in human 293 cells with overexpression of FGFR1 (14), as well as in EGF-stimulated 293T cells that do not express FGFR1 or FGFR3 (19). The latter involves EGF-dependent activation of Src family members including Src and Fyn, which directly phosphorylate RSK2 at Y529 and Y707 (19), whereas FGFR3 directly phosphorylates RSK2 at these two sites independent of Src (Fig. (Fig.1G).1G). In summary, phosphorylation at Y529 and Y707 might be a general requirement for RSK2 activation through the ERK/MAPK pathway. Thus, our current and previous findings represent a paradigm for novel tyrosine phosphorylation-dependent regulation of serine-threonine kinases.
This work was supported in part by NIH grant CA120272 (J. Chen) and the Multiple Myeloma Research Foundation (J. Chen and S. Lonial). S. Kang is a special fellow of the Leukemia and Lymphoma Society. J. Chen is an American Cancer Society Basic Research Scholar and a Georgia Cancer Coalition Distinguished Cancer Scholar.
Published ahead of print on 17 February 2009.