Research ArticleCell Biology

The inositol phosphatase SHIP2 enables sustained ERK activation downstream of FGF receptors by recruiting Src kinases

See allHide authors and affiliations

Science Signaling  18 Sep 2018:
Vol. 11, Issue 548, eaap8608
DOI: 10.1126/scisignal.aap8608

Converting transient to sustained signaling

Activation of fibroblast growth factor receptors (FGFRs) stimulates downstream signaling transiently because the receptors are endocytosed and degraded after activation. Nevertheless, FGFRs stimulate both sustained and transient ERK signaling. Fafilek et al. found that the inositol phosphatase SHIP2 was required for converting transient FGFR activation into sustained ERK signaling. The catalytic activity of SHIP2 was not required. Instead, SHIP2 acted as a scaffold that recruited Src family kinases to FGFR complexes, thus enhancing the phosphorylation of adaptor proteins that mediated signal relay from FGFRs to ERK. Because sustained ERK activation due to aberrant FGFR signaling is associated with oncogenesis and developmental disorders, SHIP2 may be a potential therapeutic target for these pathologies.


Sustained activation of extracellular signal–regulated kinase (ERK) drives pathologies caused by mutations in fibroblast growth factor receptors (FGFRs). We previously identified the inositol phosphatase SHIP2 (also known as INPPL1) as an FGFR-interacting protein and a target of the tyrosine kinase activities of FGFR1, FGFR3, and FGFR4. We report that loss of SHIP2 converted FGF-mediated sustained ERK activation into a transient signal and rescued cell phenotypes triggered by pathologic FGFR-ERK signaling. Mutant forms of SHIP2 lacking phosphoinositide phosphatase activity still associated with FGFRs and did not prevent FGF-induced sustained ERK activation, demonstrating that the adaptor rather than the catalytic activity of SHIP2 was required. SHIP2 recruited Src family kinases to the FGFRs, which promoted FGFR-mediated phosphorylation and assembly of protein complexes that relayed signaling to ERK. SHIP2 interacted with FGFRs, was phosphorylated by active FGFRs, and promoted FGFR-ERK signaling at the level of phosphorylation of the adaptor FRS2 and recruitment of the tyrosine phosphatase PTPN11. Thus, SHIP2 is an essential component of canonical FGF-FGFR signal transduction and a potential therapeutic target in FGFR-related disorders.


Maintenance of tissue homeostasis depends on complex intercellular signaling networks that govern basic cell functions. The fibroblast growth factor (FGF) signaling network is a major contributor to such cell-to-cell communications. The four mammalian FGF receptors (FGFR1 to FGFR4) are receptor tyrosine kinases (RTKs) that respond to extracellular signals delivered by at least 18 FGF ligands during development, postnatal life, and disease. The importance of FGFR signaling is emphasized by their role in pathological functions. Various disorders arise from mutations, gene fusions, increased copy numbers, and other lesions affecting FGFR genes. These include cancer, developmental defects, and bone and skin disorders. Activating mutations in FGFR3 alone are associated with five severe skeletal dysplasias, nine types of cancer, and two skin syndromes (14).

Signaling through the extracellular signal–regulated kinase (ERK), a mitogen-activated protein (MAP) kinase, is implicated in most cellular phenotypes affected by FGF signaling. In human embryonic stem cells (hESCs) and induced pluripotent cells, FGFR-mediated ERK activation is critical for maintenance of the undifferentiated phenotype (57). Activation of ERK is the predominant mechanism through which FGFRs trigger cell proliferation in fibroblasts, endothelial cells, myoblasts, mesenchymal stem cells, neural progenitors, and lung and lens epithelial cells (1). However, aberrant sustained ERK activation mediates FGFR oncogenic signaling in FGFR4-driven prostate cancer, BaF3 lymphocytes and leukemic KG1 cells that express cancer-associated FGFR3 and FGFR1 fusion proteins that result from chromosomal translocations, respectively, and KMS11 multiple myeloma cells overexpressing FGFR3 as a result of a chromosomal translocation (811). ERK drives much of the pathology underlying FGFR3-related skeletal dysplasias that include achondroplasia (ACH) and thanatophoric dysplasia (TD). Sustained ERK activation due to heterozygosity for FGFR3-activating mutations that cause ACH and TD results in decreased chondrocyte proliferation, loss of extracellular matrix (ECM) in the growth plate cartilage, and altered chondrocyte differentiation (12, 13). Similarly, increased ERK signaling causes premature fusion of synchondroses in the developing vertebrae, which results in a narrowing of the spinal canal at the foramen magnum that contributes to the skeletal complications in ACH (14). Expression of constitutively active MEK1, the MAP kinase kinase upstream of ERK, produces similar phenotypic consequences as ERK overactivation, leading to an ACH-like dwarfism in mice (15).

FGFRs stimulate ERK signaling through adaptor-mediated translocation of the RAS guanine nucleotide exchange factor SOS1 to the cell membrane. FGFR substrate 2 (FRS2) is a major adaptor involved in this process. FGFRs phosphorylate FRS2 on at least six tyrosine residues that serve as binding sites for SOS1 complexed with growth factor receptor–bound protein 2 (GRB2), directing a substantial amount of SOS1 to the cell membrane where it activates RAS (16, 17).

Sustained ERK activation downstream of FGFR signaling is normal in some contexts but pathogenic when it occurs inappropriately. In chondrocytes and hESCs, FGFR activation leads to strong FRS2 tyrosine phosphorylation and corresponding sustained ERK activation lasting for more than 8 hours (18). The molecular processes underlying sustained ERK activity in FGFR signaling are poorly characterized. Here, we report that FGFRs interacted with and phosphorylated SRC homology 2 (SH2) domain–containing inositol 5-phosphatase 2 [SHIP2; also known as inositol polyphosphate phosphatase–like 1 (INPPL1)], which, in turn, recruited Src family kinases (SFKs) to the FGFR signaling complex, where they promoted FGFR-mediated phosphorylation of the adaptors FRS2 and GRB2-associated binding protein 1 (GAB1) to relay the signal to the ERK pathway. Loss of SHIP2 converted sustained FGF-mediated ERK activation into a transient signal, demonstrating that SHIP2 is an essential component of FGF signaling and necessary for the maintenance of FGF-mediated ERK activation.


FGFRs interact with SHIP2 and target SHIP2 to focal adhesions

In a previous study, we used proteomics to identify proteins that were phosphorylated upon activation of endogenous Fgfr2 and Fgfr3 (19) signaling in cultured rat chondrosarcoma (RCS) cells. As detailed previously, the tyrosine phosphorylated proteins were purified from FGF2-treated cells by immunoprecipitation (IP) with an antibody that recognizes phosphorylated tyrosine residues and subjected to tandem mass spectrometry (19). The inositol phosphatase SHIP2 was among the most frequently identified proteins phosphorylated at tyrosine upon FGFR activation, being found in four of the six experiments we performed (19).

Both wild-type FGFR3 and the activating mutant FGFR3-K650M, which is associated with TD (20), coimmunoprecipitated with human SHIP2 when expressed in 293T cells (Fig. 1A). FGFR3 induced SHIP2 phosphorylation at Tyr986, Tyr987, and Tyr1135 in 293T cells (Fig. 1A) and in cell-free kinase assays using recombinant FGFR3 and recombinant SHIP2 (recSHIP2; Fig. 1B). Addition of recSHIP2 did not affect the activity of FGFR3 in a kinase assay, as demonstrated by no change in FGFR3 autophosphorylation in the presence of SHIP2 (Fig. 1C). Similarly, FGFR3 did not alter the phosphoinositide phosphatase (PS) activity of SHIP2, as demonstrated by a cell-free SHIP2 activity assay based on colorimetric detection of phosphate released during the hydrolysis of phosphatidylinositol (3,4,5)-trisphosphate [PtdIns(3,4,5)P3] to phosphatidylinositol (3,4,5)-bisphosphate [PtdIns(3,4)P2; Fig. 1D]. These experiments established that FGFR3 interacted with SHIP2 and acted as a SHIP2 kinase. The experiments also show that the FGFR3-SHIP2 interaction did not affect the catalytic activity of either protein.

Fig. 1 FGFRs interact with SHIP2 and phosphorylate SHIP2 in cells.

(A) 293T cells were transfected with V5-tagged wild-type (WT) FGFR3 or FGFR3-K650M together with FLAG-tagged SHIP2 and subjected to V5 or FLAG IP. Immunoprecipitates were separated by SDS–polyacrylamide gel electrophoresis (PAGE) and subjected to Western blotting (WB) with antibodies to detect FGFR3, FGFR3 phosphorylated on Tyr653 or Tyr654 (pFGFR3Y653/4), SHIP2, SHIP2-FLAG, and SHIP2 phosphorylated on Tyr986 or Tyr987 (pSHIP2Y986/7). Actin is a loading control. Note the SHIP2 phosphorylation in cells transfected with the activated FGFR3 mutant FGFR3-K650M. Immunoblotting for SHIP2 detects both endogenous and transgenically expressed SHIP2. (B) Cell-free kinase assays were carried out with purified recombinant FGFR3 and recSHIP2 proteins in the presence of adenosine 5′-triphosphate (ATP), and the reactions were separated by SDS-PAGE and subjected to Western blotting. The sample without ATP is a negative control. Phosphorylation of SHIP2 (pSHIP2) was detected with 4G10, an antibody that recognizes all phosphotyrosine residues. (C) Cell-free kinase assays were repeated in the presence and absence of the FGFR inhibitor (SU5402) and immunoblotted to show phosphorylated and nonphosphorylated forms of FGFR3 and SHIP2. Data in (A) to (C) are representative of three independent experiments. (D) Cell-free colorimetric SHIP2 phosphatase activity assay measuring SHIP2-mediated hydrolysis of phosphoinositide PtdIns(3,4,5)P3 (PIP3) to PtdIns(3,4)P2 (PIP2). Purified, recSHIP2, and recombinant FGFR3 were added to reaction buffer in the presence of absence of PIP3 as indicated. Samples without ATP are negative controls for FGFR3 activity. Significance was assessed using Welch’s t test. Each column represents three independent experiments with SD indicated. For each experiment, the values were calculated as averages from two technical duplicates, each measured three times. n.s., not significant.

SHIP2 contains three distinguished domains according to structural data available in the Protein Data Bank (PDB) database ( an N-terminal SH2 domain (PDB: 2MK2), a central PS domain (PDB: 4A9C), and a C-terminal sterile alpha motif (SAM) domain (PDB: 2K4P). In addition, we identified two proline-rich (PR) domains in SHIP2, located at residues 123 to 160 (PR1) and residues 935 to 1105 (PR2) in human SHIP2 (Fig. 2A). We generated a series of C-terminally V5-tagged truncated human SHIP2 variants, coexpressed each with wild-type FGFR3 in 293T cells, and determined whether each SHIP2 variant coimmunoprecipitated with FGFR3. Deletion of the PS or PR1 domain had no effect on SHIP2 coimmunoprecipitation with FGFR3 (Fig. 2, B and C). In contrast, deletion of the SH2 or SAM domain substantially impaired FGFR3-SHIP2 association [Fig. 2, D and E (arrows)], suggesting a bipartite SHIP2-FGFR3 interaction involving both the SH2 and SAM domains of SHIP2. Wild-type SHIP2 did not coimmunoprecipitate with the catalytically inactive K508M mutant form of FGFR3 (Fig. 2F), suggesting that SHIP2 associates preferentially with active FGFR3.

Fig. 2 SHIP2 interacts with FGFR3 through its SH2 and SAM domains.

(A) Schematic representation of truncated SHIP2 constructs. The V5 epitope was added to the C terminus of each construct. (B to E) The indicated SHIP2 variants were expressed with WT FGFR3 in 293T cells, immunoprecipitated (IP) using the Flag epitope tag, and analyzed by Western blotting (WB) for FGFR3 and SHIP2. Actin is a loading control. Input, total cell lysates used for IP. Data are representative of three independent experiments. IgH, immunoglobulin heavy chain. (F) Wild-type SHIP2 (SHIP2-WT)was coexpressed in 293T cells with WT FGFR3 (FGFR3-WT) or the catalytically inactive mutant FGFR3-K508M as indicated. Data are representative of three independent experiments.

Having found that SHIP2 interacted with and was phosphorylated by FGFR3, we asked whether SHIP2 also associated with other FGFRs. In 293T cells, SHIP2 coimmunoprecipitated with FGFR1 and FGFR4 but not with FGFR2 (Fig. 3A). We next used osteosarcoma U2OS cells stably expressing a C-terminally green fluorescent protein (GFP)–tagged FGFR1 to probe the association of endogenous SHIP2 with FGFR1. Proximity ligation assays (PLAs) using primary antibodies recognizing SHIP2 and GFP revealed that endogenous SHIP2 colocalized with transgenically expressed FGFR1 (Fig. 3, B and C). We obtained similar data using PLA with primary antibodies recognizing FGFR1 and SHIP2 (Fig. 3, B and D). We detected no substantial PLA signal in isogenic negative controls for both experiments: human osteosarcoma U2OS cells stably expressing FGFR1–BirA (FGFR1 fused to biotin ligase)–HA (hemagglutinin tag) (for GFP:SHIP2 PLA) and U2OS cells stably expressing FGFR4-GFP (for FGFR1:SHIP2 PLA).

Fig. 3 SHIP2 interacts with FGFR1 in cells.

(A) 293T cells were cotransfected with FLAG-tagged SHIP2 and V5-tagged WT FGFR1, FGFR2, or FGFR4, and lysates were subjected to FLAG IP and Western blotting (WB) for the FGFRs and SHIP2. Data are representative of three independent experiments. (B to D) U2OS cells stably expressing FGFR1-GFP were subjected to PLA using antibodies detecting SHIP2 and GFP (B and C) or SHIP2 and FGFR1 (B and D). U2OS cells stably expressing FGFR1-BirA-HA or FGFR4-GFP were used as negative controls in (B) and (C) and (B) and (D), respectively. n = 3 independent experiments. Data are means ± SEM. Statistical significance was determined by Welch’s t test with Bonferroni’s correction of P values (n.s., P > 0.05; ***P < 0.001). Scale bar, 20 μm. Numbers in columns (B) indicate the total number of cells scored across the three independent experiments.

Treatment of RCS cells with FGF2 triggered phosphorylation of endogenous Ship2 at Tyr986 and Tyr987 (Fig. 4A) and targeted Ship2 to the cell periphery, where it partially colocalized with the focal adhesion marker vinculin (Fig. 4, B and C). We observed similar translocation and association with peripheral focal adhesions for the signaling adaptor p130Cas (Fig. 4, D and E), which was previously found to be phosphorylated at multiple tyrosines in FGF2-treated RCS cells (19). Because Ship2 and p130Cas interacted with FGFR3 and are FGFR3 substrates, it is likely that they are phosphorylated by FGFR3 and transit to focal adhesions together. In support of this hypothesis, we previously reported that p130Cas and SHIP2 interact through the SH3 domain of p130Cas and localize to lamellipodia together (21).

Fig. 4 FGF signaling targets SHIP2 to focal adhesions.

(A) RCS cells were treated with FGF2 for the indicated times, and lysates were subjected to immunoblotting for Ship2 phosphorylated on Tyr986 or Tyr987 (pShip2Y986/7) and for total Ship2. Data are representative of three independent experiments. (B) Immunofluorescence images showing phosphorylated Ship2 (pShip2Y986/7) in RCS cells treated with FGF2 or not (control). Arrows indicate peripheral focal adhesions. DAPI, 4′,6-diamidino-2-phenylindole; DIC, differential interference contrast. (C) Immunofluorescence images showing pShip2Y986/7 and the focal adhesion marker vinculin (vinc.). Arrows indicate focal adhesions in RCS cells treated with FGF2. (D) Immunofluorescence images showing p130Cas phosphorylated at Tyr410 (pp130Cas) and vinculin in RCS cells treated with FGF2. Arrows indicate focal adhesions. Scale bars, 25 μm. (E) Higher magnification of boxed areas indicated in (C) and (D). Images are representative of three independent experiments. (F) Immunoblot showing Ship2 in the indicated Ship2Crispr RCS cell lines compared to WT RCS cells. Data are representative of three independent experiments. (G) Quantification of fetal bovine serum (FBS)– and FGF2-stimulated migration in WT RCS cells and the indicated Ship2Crispr cell lines. Data are means ± SEM. Statistical significance was determined by Welch’s t test with Bonferroni’s correction of P values, ***P < 0.0001. Data are representative of three independent experiments.

SHIP2 is part of the integrin adhesome and is known to interact with various cytoskeletal proteins (22). Previous data obtained in human fibroblasts show that general cell migration is substantially reduced in SHIP2-null fibroblasts derived from opsismodysplasia (OPS) patients (23). To determine whether Ship2 plays a role in FGF-mediated cell migration, we used the CRISPR (clustered regularly interspaced short palindromic repeat)/Cas9 (CRISPR-associated) system to disrupt the Ship2 gene in RCS cells. We selected four RCS clones with the Ship2 locus targeted by CRISPR/Cas9 (Ship2Crispr cells), analyzed their genotypes by DNA sequencing, and determined whether Ship2 was present in these cells by Western blotting (clone names and genotypes: Ship2c−/−, Ship2a+/−, Ship2g−/−, and Ship2 f+/−; Fig. 4F). Using fibronectin-coated glass bottom chambers, we compared the cell migration velocity of wild-type and Ship2Crispr cells. In the presence of 10% serum, migration was inhibited in two Ship2Crispr cell lines compared to wild-type cells (Fig. 4G). In contrast, Ship2 deletion significantly increased migration in cells treated with FGF2 in the absence of serum (Fig. 4G). Therefore, Ship2 acts as a negative regulator of FGF-mediated migration in RCS cells.

Ship2 deletion rescues cellular phenotypes caused by FGFR activation

RCS cells are a well-established chondrocyte model for studying pathologic increased FGFR3 signaling as seen in ACH or TD due to activating FGFR3 mutations. The cells respond to activation of endogenous FGFR signaling with complex changes in cell behavior including growth arrest, diminished production of collagen and proteoglycan ECM components, induction of premature senescence, and alteration in cell shape (12, 13, 2426). FGF2 stimulation induced less cell growth arrest in Ship2Crispr RCS cells compared to wild-type RCS cells (Fig. 5A). This phenotype was not due to clonal variation because three randomly selected RCS clones with wild-type Ship2 genotypes responded to FGF2 with growth arrest comparable to wild-type cells (Fig. 5, B and C). The Ship2Crispr cells also showed less FGF2-induced ECM loss, as determined by Western blotting for collagen 2 and Alcian blue staining for sulfated proteoglycans (Fig. 5, D and E, and fig. S1A). Finally, FGF-mediated induction of premature senescence was also less pronounced in Ship2Crispr cells compared to wild-type controls, as demonstrated by Western blotting for the presence of the senescence marker caveolin 1 (Fig. 5D) (12). These experiments established that Ship2 deletion rendered RCS cells less responsive to FGF stimulation with respect to multiple FGF-mediated cellular phenotypes. In contrast, Ship2 deletion did not affect FGF-induced changes in cellular shape, manifested as cell flattening and enlargement (Fig. 5F and fig. S1B) due to cytoskeletal remodeling with abundant formation of actin stress fibers (Fig. 5G) (18). The translocation of Ship2 to peripheral focal adhesions in response to FGF2 (Fig. 4, B and C) was therefore not essential for FGF-mediated cell spreading and cytoskeletal remodeling.

Fig. 5 Loss of SHIP2 rescues some FGF-induced cell phenotypes.

(A) Quantification of cell proliferation, as determined by crystal violet staining, in WT RCS cells and the four indicated Ship2Crispr RCS cell lines in the presence of increasing amounts of FGF2. (B) Quantification of cell proliferation in three randomly selected RCS clones in which the Ship2 was not disrupted by CRISPR/Cas9 (2E11, 1E11, and 2B10) and the three indicated Ship2Crispr RCS cell lines. Data in (A) and (B) represent averages from eight wells ± SD. Results are representative of three independent experiments. Statistically significant differences are highlighted. ***P < 0.001, Welch’s t test. (C) Immunoblot showing total Ship2 in the indicated RCS cell lines. Results are representative of three independent experiments. (D) Immunoblot showing the cartilaginous ECM marker collagen 2 and the senescence marker caveolin in two WT and two Ship2 knockout RCS cell lines in the presence and absence of FGF2. Blot is representative of three independent experiments. (E) Alcian blue staining showing sulfated proteoglycans characteristic of cartilaginous ECM in the indicated WT and Ship2 knockout RCS cell lines in the absence and presence of FGF2. Scale bar, 200 μm. (F) Evidence for spreading in cells treated with FGF2 for 72 hours, due to the formation of actin stress fibers, visualized by phalloidin staining. Scale bar, 200 μm. (G) Immunofluorescence showing phalloidin in WT and Ship2Crispr cells in the absence and presence of FGF2. Scale bars, 20 μm. Data are representative for three independent experiments.

Ship2 deletion converts FGF-mediated sustained ERK activation into a transient signal

It is possible that the lack of response to FGF2 in Ship2Crispr RCS cells may simply have been the result of reduced FGFR activity, but we found no differences between the amounts of Fgfr2 or Fgfr3 in wild-type and Ship2Crispr cells (fig. S2, A and B). Because sustained Erk activation accounts for most of cellular phenotypes triggered by FGF in RCS cells (27), we examined the kinetics of Erk activity in Ship2Crispr cells. All four Ship2Crispr cell lines showed markedly impaired FGF-mediated Erk activation, which progressed with time, resulting in diminished Erk signaling in cells treated with FGF2 for more than 4 hours (Fig. 6A and figs. S3, A to C, and S4, A to D). In contrast, wild-type cells maintained phosphorylated Erk for at least 10 hours. Thus, the Ship2Crispr cells initially responded to FGF-mediated Erk activation normally but failed to maintain Erk activity over prolonged periods of time. The defective maintenance of Erk activation accounts for the rescue of the FGF-mediated changes in cell behavior observed in Ship2Crispr cells.

Fig. 6 Loss of Ship2 impairs Fgfr-Erk signaling.

(A) Representative Western blot showing total and phosphorylated (p) forms of Frs2, Gab1, Akt, and Erk in WT and Ship2Crispr (Ship2a+/−) RCS cells treated with FGF2 for the indicated times. Actin is a loading control. Densitometry quantification and number of independent replicates are provided in fig. S4A. Additional analyses carried out in other Ship2Crispr cell lines are shown in figs. S3 (A to C) and S4 (C and D). (B) Quantification of Ship2 protein in three WT RCS cultures by Western blotting of lysates and serial dilutions of purified recombinant (rec) SHIP2, followed by densitometry to quantify the band intensities. The number of Ship2 molecules per cell was estimated on the basis of comparing the band intensities and number of cells in each lysate (lanes 1 to 3) with the band intensities of the purified protein samples (lanes 4 to 10). The numbers above lanes 1 to 3 indicate the number of cells represented by the lysates, and the numbers above lanes 4 to 10 represent the number of molecules of recSHIP2 per sample. On the basis of these calculations, we estimated there to be 287,546 ± 37,305 (means ± SEM) Ship2 molecules per cell. n = 9 independent RCS cultures. I.O.D., integrated optical density. (C) Ship2Crispr cells were microinjected with amounts of recombinant SHIP2 (recSHIP2) equal to about 1/10 (~25,000 molecules per cell) of the number of endogenous Ship2 molecules per cell along with a dTomato transcriptional reporter of ERK activity [pKrox24(MapErk)dTomato]. TexasRed conjugated to dextran was used as a marker for injection. The cells were then treated with FGF2 and assayed for dTomato expression. Ph2, phase contrast. Scale bar, 200 μm. (D) Quantification of dTomato expression in the two indicated Ship2Crispr cell lines treated as in (C). n = 3 independent experiments. ***P ˂ 0.001, **P ˂ 0.01, Student’s t test. (E) WT and Ship2a+/− RCS cells were treated with FGF2 for the indicated times, Frs2 was immunoprecipitated (IP) from the lysates, and the immunoprecipitates were subjected to Western blotting for Ptpn11. Blot is representative of three independent experiments.

We next asked whether the defect in FGF-mediated Erk activation in Ship2 null cells could be restored by microinjection of recSHIP2 protein. We used Western blotting to determine the amounts of endogenous Ship2 protein in wild-type RCS cells (means ± SEM, 287,546 ± 37,305 molecules per cell; Fig. 6B). Individual cells from two RCS Ship2 null cell lines (Ship2c−/− and Ship2 g−/−) were microinjected with an amount of purified recSHIP2 equal to about 1/10 of the endogenous amount of Ship2 (~25,000 molecules per cell), together with a dTomato transcriptional reporter for FGF-mediated ERK activation [pKrox24(MapErk)dTomato] (28). Cells were treated with FGF2, and the induction of dTomato expression was analyzed 24 hours later and plotted. Injection of recSHIP2 increased the ability of FGF to activate Erk in the Ship2Crispr cells (Fig. 6, C and D), demonstrating that addition of SHIP2 protein significantly increased the Erk activation by FGF signaling.

To test whether other RTKs also require Ship2 to activate Erk, we transfected RCS cells with constructs encoding full-length wild-type human tropomyosin receptor kinase A (TRKA) or human epidermal growth factor receptor (EGFR) and treated the cells with the TRKA ligand nerve growth factor (NGF) or the EGFR ligand EGF, respectively. We then compared TRKA- and EGFR-mediated Erk activation between wild-type cells and two Ship2Crispr cell lines. No substantial differences in Erk activation were found among wild-type and Ship2Crispr cells (fig. S5, A and B), suggesting that the requirement for Ship2 is specific to FGFRs.

FGFRs relay signaling to the ERK pathway through adaptors such as FRS2 and GAB1, both of which undergo tyrosine phosphorylation upon FGF treatment and form complexes with the tyrosine phosphatase and adaptor protein tyrosine–protein phosphatase non-receptor type 11 (PTPN11; also known as SHP2) to activate the RAS-ERK signaling module. FGF2 elicited abundant tyrosine phosphorylation of Frs2 at Tyr436 and Gab1 at Tyr627 in RCS cells (27), and this was impaired in Ship2Crispr cells (Fig. 6A and figs. S3, A to C, and S4, A to D). These data suggest that the defect in Erk activation in Ship2Crispr cells may stem from poor recruitment of Ptpn11 to underphosphorylated adaptors. Despite multiple attempts, we were unable to immunoprecipitate Gab1 from RCS cells due to technical difficulties. However, Frs2-Ptpn11 coimmunoprecipitation experiments showed impaired association of Frs2 with Ptpn11 upon FGF2 treatment of Ship2Crispr cells compared to wild-type cells (Fig. 6E and fig. S3D).

SHIP2 promotes association of SRC family kinases with FGFR3

The SFKs BLK, FGR, FYN, HCK, LCK, LYN, and YES were among the most frequently identified proteins in mass spectrometry analyses of the FGFR3 interactome (19), and coimmunoprecipitation experiments carried out in 293T cells confirmed the association of FGFR3 with LCK (Fig. 7A), FYN, LYN (fig. S6A), FGR, and BLK (fig. S6B). We used bimolecular fluorescence complementation (BiFC) to visualize the interaction of FGFR3 and LYN in RCS cells. For this analysis, LYN complementary DNA (cDNA) was C-terminally tagged with cDNA encoding the N-terminal 158 amino acid residues of the yellow fluorescent protein (YFP) variant Venus (LYN-V1) (29), and FGFR3 cDNA was fused to cDNA encoding amino acid residues 159 to 239 from the C terminus of Venus (FGFR3-V2). When brought into close proximity, the N- and C-terminal Venus fragments reconstitute fluorescence. BiFC assays demonstrated that FGFR3 and LYN were located in very close proximity to one another in RCS cells (Fig. 7, B and C).

Fig. 7 SHIP2 promotes the association of SFKs with FGFR3.

(A) 293T cells were transfected with vectors encoding FLAG-tagged WT FGFR3 or the K650E activating mutant (FGFR3-K650E) together with the SFK LCK. FGFR3 was immunoprecipitated (IP) from lysates using the FLAG tag, and the immunoprecipitates were analyzed for FGFR3 and LCK by Western blotting (WB). (B and C) Bimolecular complementation assay. (B) RCS cells were transfected with vectors encoding FGFR3-Venus2 (V2) and Lyn-Venus1 (V1), fixed, and counterstained by GFP antibody to visualize expression of either part of Venus protein. ICC, immunocytochemistry; fluor., fluorescence. (C) The percentage of transfected cells showing Venus activity was calculated and plotted. n = 3 independent experiments. Data represent averages ± SEM. Scale bar, 10 μm. (D and E) RCS cells (D) and 293T cells (E) were treated with FGF2 alone or in the presence of the SRC inhibitors (inh.) AZM475271 or A419259 and analyzed by Western blotting for total and phosphorylated (p) FRS2 and ERK. Data are representative of three independent experiments. (F) Western blot showing activation (phosphorylation) of WT FGFR3 overexpressed in RCS cells in the absence and presence of the inhibitor of FGFR3 catalytic activity AZD4547 (AZD). (G and H) Immunofluorescence images showing PLAs for FGFR3 and the SFK YES tagged with YFP (YES-YFP) in RCS cells. Scale bars, 10 μm. (I) Combined results of three independent PLA experiments assessing the colocalization of FGFR3 and YES in two Ship2Crispr cell lines (Ship2c−/− and Ship2a+/−) compared to WT RCS cells. Transfections containing empty vector and a vector expressing YFP only are negative controls for the PLA assay. Data represent averages ± SEM. *P < 0.05, ***P < 0.001, Welch’s t test with Bonferroni’s correction of P values. (J) Immunofluorescence images showing PLAs testing the association of plasmid-encoded FGFR1-GFP with the endogenous SFK LYN. (K) Western blot for SHIP2 in U2OS clones with the indicated SHIP2 genotypes. (L) Combined results of three independent PLA experiments assessing the colocalization of FGFR1 and LYN in two SHIP2Crispr U2OS cell lines (SHIP2g−/− and SHIP2b−/−) compared to SHIP2+/+ U2OS cells. U2OS cells expressing FGFR4-GFP were used as a negative control for the PLA assay. Data represent averages ± SEM (**P < 0.01, ***P < 0.001, Welch’s t test).

Inhibition of endogenous SFK catalytic activity by two unrelated chemical inhibitors (AZM475271 and A419259) abolished ERK activation in response to the addition of exogenous FGF2 in both RCS and 293T cells (Fig. 7, D and E). ERK inhibition was accompanied by the inhibition of FRS2 phosphorylation, suggesting that SFKs are recruited to FGFRs and participate in FGFR-mediated adaptor phosphorylation and ERK activation.

Having established that Ship2 deletion and SFK inhibition both impaired FGF-mediated ERK activation in human and rat cells, we speculated that SHIP2 promotes the association of SFKs with FGFRs. We used PLAs to probe the interaction of FGFR3 with the SFK YES in different Ship2Crispr RCS cell lines. Expression of FGFR3 in RCS cells caused it to become spontaneously activated (Fig. 7F) and associate with coexpressed YES at the cell membrane (Fig. 7, G and H). This association was reduced by 40 and 61% in two tested Ship2Crispr cell lines (Fig. 7I), suggesting that Ship2 facilitates the interaction of FGFR3 with SFKs. Next, we used the CRISPR/Cas9 system to disrupt the SHIP2 gene in U2OS cells expressing FGFR1-GFP and used PLA to test the association of FGFR1 with endogenous LYN. In the two SHIP2Crispr U2OS clones, the interaction of FGFR1 with LYN was inhibited by 47 and 66%, compared to wild-type cells (Fig. 7, J to L). Together, the PLA experiments demonstrated less association of SFKs with FGFRs in cells with deleted SHIP2.

The inositol phosphatase activity of SHIP2 is not necessary for FGF-mediated ERK activation

The truncated human SHIP2 variant lacking the entire inositol phosphatase domain (SHIP2-ΔPS) interacted normally with FGFR3 (Fig. 2B). Similarly, FRS2, LYN, LCK, and FGR coimmunoprecipitated normally with SHIP2-ΔPS or with the phosphatase-defective SHIP2 (SHIP2-PD) triple mutant (P686A/D690A/R691A; Fig. 8, A to C, and fig. S7) (30). Thus, the interaction with FGFR3 or with members of the FGFR3 signaling complex does not require SHIP2’s inositol phosphatase domain or its catalytic activity.

Fig. 8 The inositol phosphatase activity of SHIP2 is not necessary for its association with FGFR signaling complexes or FGF-mediated ERK activation.

(A to C) Representative Western blots showing LCK (A), LYN (B), and FRS2 (C) in SHIP2-V5 coimmunoprecipitates (IP) from 293T cells expressing the indicated combinations of V5-tagged WT SHIP2, the catalytically inactive SHIP2 triple mutant SHIP2-PD (P686A, D690A, and R691A), a SHIP2 mutant lacking the entire inositol phosphatase domain (SHIP2-ΔPS), LCK, LYN, and FRS2. Nontransfected cells or those transfected with vectors encoding only GFP serve as negative IP controls. Actin is a loading control. (D) RCS cells were treated with inhibitor of SHIP2 phosphatase activity AS1949490 (AS19) and purified FGF2 for the indicated times, and the lysates were analyzed by Western blotting for phosphorylated (p) forms of the indicated proteins. Actin and total protein are loading controls. Immuoblotting data are representative of three independent experiments. (E) Dual-luciferase assay. RCS cells were transfected with WT SHIP2 or SHIP2-ΔPS together with a firefly luciferase pKrox24 ERK reporter plasmid that is transactivated by FGF-ERK signaling and a Renilla luciferase pTK-RL control plasmid. Cells were treated with FGF2 for 24 hours before the dual-luciferase assay. Data are compiled from three independent experiments, with four biological and two technical replicates for each treatment. Bars are averages ± SD. **P < 0.01, ***P < 0.001, Welch’s t test with Bonferroni’s correction of P values. (F) Model illustrating the putative role of SHIP2 in canonical FGFR-ERK signaling. Upon being activated by binding to FGFs, FGFRs are activated and phosphorylate the adaptor protein FRS2. Although GAB1 phosphorylation can be also detected in FGF-treated cells, FRS2 is the major adaptor involved in signal relay from FGFR to the ERK pathway (17, 45, 81). Activated FRS2 and GAB1 recruit PTPN11 and SOS1 to the membrane. This activates RAS, which in turn activates ERK. Activated FGFRs also phosphorylate SHIP2 and recruit SHIP2 to the FGFR signaling complex at the cell membrane. SHIP2 recruits SFKs to the complex, which enhances FGFR-mediated phosphorylation of the adaptor proteins, activation of RAS, and ERK signaling. SHIP2 knockout effectively converts FGF-induced sustained ERK activation into the FGF-induced transient activation and rescues the cell phenotypes induced by sustained FGFR-ERK signaling (premature senescence, ECM degradation, and growth arrest). This is due to hypophosphorylation of FRS2 in SHIP2 knockout cells, resulting in diminished recruitment of PTPN11-SOS1 complexes that activate the RAS-ERK signaling module.

To test whether the catalytic function is required for the role of SHIP2 in FGF-ERK signaling, RCS cells were treated with a chemical inhibitor of SHIP2 inositol phosphatase activity (AS1949490) (31). Because the ability of SHIP2 to negatively regulate the phosphoinositide 3-kinase (PI3K)–AKT pathway depends on its catalytic activity, AS1949490 should increase AKT signaling. AS1949490 treatment produced no changes in basal phosphorylation of Akt at Ser473 in FGF2-naïve cells. However, AS1949490 prevented the inhibition of Akt activity by prolonged Fgfr signaling in RCS cells (Fig. 8D and fig. S8) (32), confirming that the inhibition of Ship2’s catalytic activity increased Akt activation. AS1949490 did not alter FGF-mediated FRS2 phosphorylation and, unexpectedly, enhanced rather than suppressed FGF-mediated ERK phosphorylation (Fig. 8D and fig. S8). Next, we monitored Erk activity in RCS cells using the pKrox24(MapErk)Luciferase reporter, which uses the promoter of a locus strongly induced by ERK signaling (EGR1) to drive expression of a luciferase reporter. pKrox24 was designed to record quantitative changes in ERK pathway activation by FGF signaling (28). FGF-mediated pKrox24 transactivation was enhanced in cells transfected with wild-type human SHIP2, compared to untransfected controls (~13-fold versus ~9-fold; Fig. 8E). The pKrox24 transactivation was enhanced up to ~17-fold in cells transfected with SHIP2-ΔPS, together suggesting that lack of SHIP2 catalytic activity does not account for diminished FRS2 phosphorylation and Erk activation in FGF-treated Ship2Crispr cells. Thus, the decrease in FGF-mediated Erk activation caused by loss of Ship2 does not result from the loss of Ship2 catalytic activity but likely reflects the role of Ship2 as an adaptor in the signaling cascade.


Most of the 20 established RTK families use the ERK pathway as a major effector of their signal transduction. The magnitude and duration of ERK activation appear critical for cell decisions during the physiological processes regulated by RTKs. In PC12 pheochromocytoma cells, for instance, the transient Erk activation elicited by Egfr leads to increased proliferation, whereas the sustained Erk activation elicited by Trk results in growth arrest and neuronal differentiation (33, 34). Under pathological conditions, sustained ERK activation is almost invariantly the major driver of cell phenotypes caused by RTK deregulation. The FGFRs are no exception from this paradigm, but the mechanisms associated with sustained ERK activation in FGFR signaling have been unclear. Deepening our understanding of the mechanisms associated with ERK activation downstream of FGFRs will aid in progress toward treatments for FGFR-related pathologies.

FGFRs relay their signal to ERK through adaptors such as FRS2. FGFRs phosphorylate FRS2 at several tyrosine residues that serve as binding sites for the GRB2-SOS1 (Tyr196, Tyr306, Tyr349, and Tyr392) and PTPN11-GRB2-SOS1 (Tyr436 and Tyr471) complexes, thus directing SOS1 guanine nucleotide exchange factor activity to the cell membrane where it activates RAS (16, 17). Adaptors not only relay the signal but also act as its amplifiers because the phosphorylation of several adaptor molecules by a single FGFR substantially increases the docking interface for cytoplasmic SOS1 at the cell membrane. Thus, RTK families that use adaptors, such as FGFR or TRK, tend to produce higher and longer ERK activation compared to RTKs that engage their downstream signaling targets directly (35).

Although FGFRs induce strong ERK activation in general, the magnitude and duration of ERK activity fluctuate substantially among different cell types, ranging between 1 and 2 hours in fibroblasts to persistent signal for at least 24 hours in chondrocytes (36). The mechanisms maintaining this ERK activity are poorly understood. According to the existing paradigm, activation of FGFRs leads to their rapid internalization and degradation. This is followed by a period during which a given cell is insensitive to FGF, before new FGFR molecules are present at the cell surface (4, 3739). In some pathological contexts, the continuous presence of FGFRs at the cell surface contributes to sustained ERK activity. These include gastric and lung cancers, in which FGFR1 and FGFR2 are frequently amplified, leading to constant gene expression and protein production, and multiple myeloma, in which FGFR3 is overproduced as a result of the t(4;14)(p16.3;q32) chromosomal translocation that places an immunoglobulin promoter upstream of the FGFR3 gene (4042). Although the high abundance of FGFR3 is part of the chondrocyte differentiation program, experimental evidence has demonstrated rapid FGFR3 down-regulation after its activation even in these cells (25). Thus, the persistence of FGFRs at the membrane is unlikely to account for sustained ERK activation in chondrocytes.

The adaptors that mediate ERK activation downstream of FGFRs may attenuate the ERK signal. FRS2 represents such a site of potent negative feedback. First, tyrosine-phosphorylated FRS2 mediates the interaction of the ubiquitin ligase cCBL with FGFR, leading to its ubiquitin-mediated degradation (43). Second, active ERK phosphorylates FRS2 at multiple threonine residues adjacent to the phosphorylated tyrosines, leading to diminished PTPN11-GRB2-SOS1 recruitment and attenuation of ERK activity (44, 45). The components of the ERK-FRS2 negative feedback loop exist in chondrocytes, yet there is no corresponding attenuation of ERK activity in these cells (27). Moreover, analyses of transcriptional changes triggered by FGFR activation in chondrocytes (18) demonstrated a potent induction of the ERK phosphatase dual specificity phosphatase 6 and members of the Sprouty family of RAS-ERK pathway inhibitors (46, 47). Thus, FGF-mediated ERK activation persists for many hours despite the induction of negative feedback mechanisms, suggesting the existence of other molecular processes capable of overriding established negative feedback mechanisms. Our study describes one such mechanistic pathway, linking the PS SHIP2 to the FGFR signaling complex for sustained ERK activation. We showed that SHIP2 interacted with FGFRs, was phosphorylated by active FGFRs, and promoted FGFR-ERK signaling at the level of adaptor phosphorylation and PTPN11 recruitment (Fig. 8F).

The SFKs are known to participate in a wide array of FGF-regulated events including cell proliferation, shape changes, migration, adhesion, and differentiation (4855). The mechanism of FGFR-mediated recruitment of SFKs is somewhat unclear, although earlier studies demonstrated direct interaction of SFKs with FGFRs, or recruitment of SFKs by FRS2 (5658). Similarly, although most of the cellular phenotypes caused by FGF-SRC signaling depend on ERK (48, 59), it is not clear how SFKs mechanistically integrate into the FGF-ERK pathway. We showed that SFKs are recruited to the FGFR signaling complex by SHIP2. Inhibition of SFK activity abolished FGF-mediated ERK activation and FRS2 phosphorylation, suggesting that SFKs assist FGFRs in adaptor phosphorylation and signal relay to the ERK pathway, together enabling a robust, negative feedback–resistant relay of the FGFR signal to the RAS-ERK pathway. Our experiments thus establish SHIP2 as an essential component of canonical FGFR signaling that enables sustained FGFR-mediated ERK activation by recruiting SFKs to the FGFR signaling complex (Fig. 8F).

The major physiological function attributed to SHIP2 is the regulation of insulin signaling. Ship2 knockout mice show increased sensitivity to insulin, causing hypoglycemia, deregulated expression of genes involved in liver gluconeogenesis, and perinatal death (60). A later study did not confirm such a severe phenotype but reported that Ship2−/− animals were resistant to weight gain due to the decreased cellular response to insulin (61). In addition to altered insulin signaling, the Ship2−/− mice are about half the size of their wild-type littermates, suggesting a potential role for SHIP2 in the growth of both the axial and appendicular skeletons. In humans, loss-of-function mutations in SHIP2 cause OPS, an autosomal recessive skeletal dysplasia characterized by severe shortening of all the appendicular bones with radiographic evidence of endochondral ossification delay, extremely short hands, poor bone mineralization, flattening of the spine, and severe midface hypoplasia (6264). Histologic analyses of cartilage growth plate morphology in patients with mutations in SHIP2 showed poor organization of the zone of proliferating chondrocytes, near the absence of the zone of hypertrophic chondrocytes, and increased vascular invasion (65, 66).

Molecular analyses have demonstrated that the negative role of SHIP2 in insulin signaling is caused by SHIP2-mediated inhibition of PI3K-AKT signaling (30, 61). SHIP2-mediated dephosphorylation of plasma membrane PtdIns(3,4,5)P3 to PtdIns(3,4)P2 reduces the amount of AKT that can be recruited to the cell membrane and subsequently activated by PDK1 (3-phosphoinositide-dependent protein kinase-1) (67). Similar to the role of SHIP2 in insulin signaling, the developmental defects in OPS result from the lack of SHIP2 catalytic activity because several missense mutations associated with OPS map to the catalytic domain (68, 69). Because insulin-like growth factor (IGF) signals primarily through the PI3K-AKT pathway (70), it is likely that the mutations in SHIP2 that cause OPS do so through deregulation of IGF-AKT mitogenic signaling. IGF signaling is a major growth-promoting signal for skeleton (71). Knockout models of IGF or its receptor IGF-1R show growth retardation with poor bone mineralization and an abnormal cartilage growth plate that shares phenotypic similarities with the human OPS cases characterized by disorganized proliferating chondrocytes and a shortened hypertrophic chondrocyte zone, supporting that the hypothesis abnormal IGF signaling affects the growth plate (72, 73). In addition to PI3K-AKT signaling, IGF-1R interacts with the parathyroid hormone–related peptide–Indian hedgehog pathway, adding additional complexity to direct and indirect effects of SHIP2 loss on the cellular behavior of chondrocytes that are also affected by dysregulation of FGFR signaling (73).

However, the SHIP2-mediated propagation of FGF signaling in chondrocytes appears to be independent of SHIP2 catalytic activity and the AKT pathway. First, deletion or reduction of Ship2 in Ship2Crispr RCS cells did not increase AKT signaling as would be expected whether the role of Ship2 is limited to the inhibition of the PI3K-AKT pathway. Second, chemical inhibition of Ship2 phosphatase activity increased AKT phosphorylation but did not reduce FGF-mediated ERK activation or adaptor phosphorylation. Third, genetic inactivation or deletion of the SHIP2 catalytic domain did not affect its association with the FGFR signaling complex. It thus appears that SHIP2’s catalytic function is important for mitogenic IGF signaling in growth plate cartilage, whereas its adaptor function is critical for the growth-inhibitory signaling of FGFR3 in the same tissue. Further experiments should unravel the interesting yet unknown biology behind how these two growth factor systems, IGF and FGF, integrate SHIP2 into their downstream signaling to achieve their opposing effects on cell proliferation. These studies should also illuminate the therapeutic potential of modulating SHIP2 activity in developmental pathologies caused by activating mutations in FGFRs.


Cell culture, vectors, transfection, and reagents

293T (American Type Culture Collection), U2OS, and RCS cells (obtained from B. de Crombrugghe) (74) were propagated in Dulbecco’s modified Eagle’s medium (DMEM), supplemented with 10% fetal bovine serum and antibiotics (Invitrogen). All expression vectors are listed in table S1. To obtain cells stably expressing FGFR4-GFP, U2OS cells were transfected with pEGFP-N1-FGFR4 (75) using FuGENE6 according to the manufacturer’s protocol (Promega). Clones were selected with geneticin (1 mg/ml), and their FGFR4-GFP expression amounts were analyzed by immunofluorescence and Western blot. A homogenous clone with moderate FGFR4-GFP expression was chosen for further studies. U2OS cells stably expressing FGFR1-GFP and FGFR1-BirA-HA have been described (76). For generation of truncated SHIP2 variants, the C-terminal Myc-DDK tag in pCMV6-hSHIP2 vector (OriGene) was replaced by V5-HIS tag using a NEBuilder HiFi DNA assembly kit (New England Biolabs). Truncated human SHIP2 constructs were generated by polymerase chain reaction (PCR) mutagenesis. Catalytically inactive 5′-PD SHIP2 was created by introducing P686A, D690A, and R691A substitutions into the inositol phosphatase domain by site-directed mutagenesis (Agilent). For cell migration assay, cells were plated on fibronectin-coated eight-chambered μ-slide wells and allowed to migrate for 15 hours as reported before (23, 77). Cells were analyzed with a Leica DM6000B microscope, using 10× magnification objective for live-cell imaging. Each cell was tracked over the period of time using the manual tracking plugin of the ImageJ software. FGF2, NGF, and EGF were from R&D Systems, heparin was from Sigma-Aldrich, AZM475271, A419259, and AS1949490 were from Tocris Bioscience.

Bimolecular fluorescence complementation

For the BiFC assays, cDNA encoding amino acid residues 2 to 158 of Venus was fused to the C terminus of LYN cDNA (LYN-V1) (29), and cDNA encoding amino acid residues 159 to 239 of Venus was fused to the C terminus of FGFR3 cDNA (FGFR3-V2). The ECM of RCS cells was degraded by 4-hour treatment with 0.3% bacterial collagenase 2 (Gibco) in no serum DMEM. Cells were transfected using FuGENE6 (1 μg of DNA per well of 24-well plate), fixed using 4% paraformaldehyde 24 hours later, and counterstained for transfection using GFP antibody (table S2), and the percentage of Venus complementation in transfected cells was calculated.

Dual-luciferase assays

For dual-luciferase assay, the pKrox24(2xD-E_inD)Luc firefly luciferase reporter (28) was transfected together with the pTK-RL vector (Promega) in 3:1 (microgram of DNA) ratio. Both plasmids were transfected using FuGENE6 reagent, 0.75 μg of plasmid DNA per well of a 24-well plate containing 2 × 105 RCS cells, in 1:3.2 ratio of DNA/FuGENE6 (micrograms per microliter). The cells were treated with FGF2 for 24 hours, and the luciferase signals were determined using dual-luciferase assay (Promega). Briefly, cells were lysed in passive lysis buffer (100 μl per well) for 15 min, and 10 μl of lysate was used for determination of the dual-luciferase signal using the manufacturer’s protocol and TriStar Multimode Luminometer (Berthold Technologies); 40 μl of each luciferase substrate was used per sample, and the measurement times were 5 to 10 s for firefly and 5 s for Renilla luciferase.


Cells were microinjected using a FemtoJet 4i microinjector with micromanipulator InjectMan 4 (Eppendorf) with 80 hPa injection pressure, 0.3 s injection time, and 10 hPa compensation pressure. Each cell was injected with 200 fl of phosphate-buffered saline containing ~25,000 molecules of recSHIP2 (SignalChem) together with fluorescence marker Dextran Alexa Fluor 647 (Thermo Fisher Scientific) and 100 molecules of pKrox24(MapErk)dTomato reporter. Cells were treated for 24 hours with FGF2 (10 ng/ml), fixed with 10% paraformaldehyde, and immediately imaged for pKrox24 transactivation, using the automatic microscope TissueFAXS i (TissueGnostics) with 20× air objective. We acquired images for red and far-red fluorescence channels showing pKrox24(MapErk)dTomato reporter and far-red counterstain marker Dextran Alexa Fluor 647. Percentages of injected cells positive for dTomato were determined by Fiji free software.

IP and Western blotting

About 5 × 106 of transfected cells were lysed for 30 min in 1 ml of IP buffer containing 50 mM tris-HCl (pH 7.4), 150 mM NaCl, 0.5% NP-40, 0.25% sodium deoxycholate, 2 mM EDTA, and 1 mM Na3VO4, supplemented with protease inhibitors. For IPs, lysates were cleared by centrifugation (14,000g for 30 min), and supernatants were incubated for 1 hour with antibodies on rotating platform at 4°C. Immunocomplexes were collected on 30 μl of A/G agarose (Santa Cruz Biotechnology) in overnight incubation, spun down, and washed four times in 400 μl of the IP buffer. Proteins attached to the beads were eluted to 70 μl of loading buffer. For Western blotting, 20 μl of cell lysates or immunoprecipitates were resolved by SDS-PAGE, transferred onto a polyvinylidene difluoride membrane, blocked with 5% (w/v) in tris-buffered saline (TBS)/Tween 20] nonfat milk, incubated with primary antibodies overnight at 4°C, washed 3× for 10 min in TBS/Tween 20, incubated with secondary antibodies at room temperature for 1 hour, washed 3× for 10 min TBS/Tween 20, and visualized by chemiluminiscence using Pierce ECL (Thermo Fisher Scientific), Immobilon Western (Millipore), or SuperSignal West Femto (Thermo Fisher Scientific) substrates. Table S2 lists all antibodies used in the study.

Kinase activity assays

Two hundred nanograms of recombinant FGFR3 was incubated for 30 min at 30°C with recSHIP2 (SignalChem) as a substrate in 50 μl of kinase buffer [60 mM Hepes (pH 7.5), 3 mM MgCl2, 3 mM MnCl2, 10 μM Na3VO4, and 1.2 mM dithiothreitol] supplemented with 10 μM ATP. For the SHIP2 activity assay, 300 ng of FGFR3 and 400 ng of SHIP2 were incubated with 50 μM PtdIns(3,4,5)P3 (Echelon Biosciences) for 30 min at 30°C. The reaction was then incubated with BIOMOL Green Reagent (Enzo Life Sciences), and optical density at 620 nm was determined by a spectrophotometer. Crystal violet staining was carried out as described (78).


SHIP2 ablation in RCS and U2OS cells using CRISPR/Cas9 technology was carried out as described (79). The CHOPCHOP tool was used to design single-guide RNAs for a pair of SpCas9n (D10A) nickases, which targeted 5′-GTGTGGGGCACCGAGTCCCG-3′ and 5′-GTCACGGTGATACCAGGCAG-3′ sites in the first exon of rat Ship2 and 5′-CGATGGCAGCTTCCTGGTCC-3′ and 5′-GCGCTCTGCGTCCTGTGAGT-3′ sites in the first exons of the human SHIP2 gene (80). Two of each Crispr/Cas9n plasmids (3 μg; Addgene plasmid #42335) were electroporated (Neon Transfection System, Thermo Fisher Scientific; 100 μl of tip, RCS cell protocol: 1600V, 10 ms, 3 pulses; U2OS cells: 1050 V, 30 ms, 2 pulses) together with a 0.5 μg of GFP-based CRISPR Reporter [designed GFP-based reporter plasmid for CRISPR (GREP), a gift from V. Korinek] where the targeted sequence was cloned within the GFP coding sequence. Successful targeting restored the GFP-reading frame, and GFP-positive cells were manually picked and expanded. Individual clones were screened by Western blot for SHIP2, and the targeted locus was PCR-amplified using 5′-agcctccactccaagcttcc-3′ and 5′-aaggtctccactcacggtgg-3′ primers for rat locus and 5′-GTACCACCGCGACCTGAG-3′ (forward) and 5′-cttggctttctcctgggtct-3′ (reverse) primers for human locus. PCR products were cloned into pGEM-T easy vector (Promega) and sequenced to determine the SHIP2 genotype.

Immunofluorescence and immunocytochemistry

Cells were fixed in 4% paraformaldehyde and incubated with primary antibodies at 4°C overnight. Secondary antibodies were conjugated with Alexa Fluor 488/594 (Life Technologies). Polymerized actin was visualized using Alexa Fluor 594–conjugated Phalloidin (Life Technologies) and vinculin-stained with vinculin–fluorescein isothiocyanate antibody according to the manufacturer’s protocol (Sigma-Aldrich). Images were taken using a confocal-inverted microscope Carl Zeiss LSM 700 with 63× oil immersion objective. Images shown in Figs. 3 (C and D), 4 (B to E), 5G, and 7 (B, H, and J) represent maximum projections of several acquired z sections. Table S2 lists all antibodies used in the study.

Proximity ligation assay

For Duolink PLA (Sigma-Aldrich), cells were fixed with 4% paraformaldehyde for 15 min and stained according to the manufacturer’s protocol. For PLA between overexpressed YES-YFP and FGFR3-Flag, RCS cells were transfected using FuGENE6 (1 μg of DNA per well of a 24-well plate), and the PLA events were calculated in three dimensions and normalized to YFP signal. Cells were analyzed by confocal microscopy (LSM700 laser scanning microscope, ZEISS) as described above.

Alcian blue staining

Cells were fixed with 4% paraformaldehyde for 10 min, stained with 1% Alcian blue 8GX (Sigma-Aldrich), thoroughly washed with distilled water, and mounted in glycerol. Brightfield images of Alcian blue samples (Fig. 5E and fig. S1A) were taken using the Olympus microscope IX71. Phase-contrast images were taken using a Carl ZEISS Axio Observer microscope. The contrast, brightness, and gamma were adjusted.

Densitometry and statistical analysis

Densitometry was conducted using Gel Analyzer function in Fiji software. The optical density of bands was measured four times to obtain mean and SD values, which were then normalized. Statistical significance of differences was evaluated by Welch’s t test. Where appropriate, Bonferroni’s correction of P values was used.


Fig. S1. Ship2 deletion or knockdown by CRISPR/Cas9 impairs Fgfr3-induced ECM degradation but not cell spreading in RCS cells.

Fig. S2. Ship2 deletion or knockdown by CRISPR/Cas9 does not affect Fgfr2 and Fgfr3 abundance in RCS cells.

Fig. S3. Deletion or knockdown of Ship2 impairs FGF2-mediated adaptor phosphorylation and activation of Erk.

Fig. S4. Densitometric quantifications of Western blot analyses.

Fig. S5. No substantial impairment of NGF- or EGF-induced changes in ERK activity in Ship2Crispr cells.

Fig. S6. FGFR3 interacts with the SFKs FYN, LYN, FGR, and BLK.

Fig. S7. FGR associates with both wild-type and catalytically inactive SHIP2.

Fig. S8. Densitometric quantifications of Western blot analyses.

Table S1. Expression vectors used in this study.

Table S2. Antibodies used in this study.


Acknowledgments: We thank P. Nemec for excellent technical assistance, V. Korinek for the GReP plasmid, and A. Wiedlocha for helpful discussions. Funding: This study was supported by Ministry of Education, Youth and Sports of the Czech Republic (KONTAKT II LH15231), Agency for Healthcare Research of the Czech Republic (15-33232A, 15-34405A, and NV18-08-00567), Czech Science Foundation (GA17-09525S), National Program of Sustainability II [MEYS CR (LQ1605) and CEITEC 2020 (LQ1601)]. This study was partially supported by the Research Council of Norway through its Centers of Excellence funding scheme (project number 262652). E.M.H. has a Career fellowship (project number 6842225) from the Norwegian Cancer Society. M.K.B. was supported by junior researcher funds from the Faculty of Medicine, Masaryk University. J.T.Z. was supported by Kohn fellowship at Masaryk University and National Institute of Arthritis and Musculoskeletal and Skin Diseases of the NIH (1F31AR066487). D.K. is supported by NIH grants R01-AR066124 and R01-AR062651. The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH. C.E. is supported by grants from the Université Libre de Bruxelles and of the Fonds de la Recherche Scientifique Médicale (J.0078.18). S.G. is supported by Hoguet, Fonds Lekime-Ropsy, and Télévie fellowships. Author contributions: P.K. designed the project. C.E., L.T., D.K., J.W., and P.K. designed experiments. B.F., L.B., M.K.B., M.V., A.N., T.G., I.G., J.K., S.G., M.P., L.J., N.H.C., J.T.Z., M.K., E.M.H., and P.K. performed experiments. B.F., M.K.B., C.E., J.W., D.K., and P.K. wrote the manuscript. Competing interests: The authors declare that they have no competing interests. Data and materials availability: All data needed to evaluate the conclusions in the paper are present in the paper or the Supplementary Materials.

Stay Connected to Science Signaling

Navigate This Article