Skip to main content

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

Regulation of sarcoma cell migration, invasion and invadopodia formation by AFAP1L1 through a phosphotyrosine-dependent pathway

Abstract

Invasion and metastasis are controlled by the invadopodia, which delivers matrix-degrading enzymes to the invasion interface permitting cancer cell penetration and spread into healthy tissue. We have identified a novel pathway that directs Lyn/Src family tyrosine kinase signals to the invadopodia to regulate sarcoma cell invasion via the molecule AFAP-1-like-1 (AFAP1L1), a new member of the AFAP (actin filament-associated protein) family. We show that AFAP1L1 can transform cells, promote migration and co-expression with active Lyn profoundly influences cell morphology and movement. AFAP1L1 intersects several invadopodia pathway components through its multiple domains and motifs, including the following (i) pleckstrin homology domains that bind phospholipids generated at the plasma membrane by phosphoinositide 3-kinase, (ii) a direct filamentous-actin binding domain and (iii) phospho-tyrosine motifs (pY136 and pY566) that specifically bind Vav2 and Nck2 SH2 domains, respectively. These phosphotyrosine motifs are essential for AFAP1L1-mediated cytoskeleton regulation. Through its interaction with Vav2, AFAP1L1 regulates Rac activity and downstream control of PAK1/2/3 (p21-activated kinases) phosphorylation of myosin light chain (MLC) kinase and MLC2. AFAP1L1 interaction with Nck2 recruits actin-nucleating complexes. Significantly, in osteosarcoma cell lines, knockdown of AFAP1L1 inhibits phosphorylated MLC2 recruitment to filamentous-actin structures, disrupts invadopodia formation, cell attachment, migration and invasion. These data define a novel pathway that directs Lyn/Src family tyrosine kinase signals to sarcoma cell invadopodia through specific recruitment of Vav2 and Nck2 to phosphorylated AFAP1L1, to control cell migration and invasion.

Introduction

The invasive cell phenotype is mediated by invadopodia, which are actin-rich matrix-degrading protrusive subcellular structures that form on the ventral cell surface and are critically controlled by Src family tyrosine kinases (SFKs).1, 2, 3, 4, 5 Through invadopodia-mediated delivery of metalloprotease-containing vesicles via the vesicle-tethering exocyst complex,6 they facilitate basement membrane and extracellular matrix degradation.7 Several actin-binding adaptor proteins are used in the initiation, assembly and maturation of invadopodia, including the recently characterized actin filament-associated protein (AFAP) family.8, 9 As with the actin-related protein complex-2/3 and cortactin, the neuronal Wiskott–Aldrich syndrome protein (N-WASp) binding adaptor,10 AFAP family members are SFK substrates that localize to invadopodia.9, 11 SFK regulate invadopodia initiation and assembly downstream of receptor tyrosine kinases and integrins by stimulating GTPases such as Rac1 and cdc42,12 as well as via phosphorylating substrates/scaffolds such as tyrosine kinase substrate-5 (Tks5).13 In addition, cortactin brings together actin nucleation regulators including actin-related protein complex-2/3, N-WASp, WASp-interacting protein (WIP) and cofilin.10, 14

AFAP family proteins also intersect protein kinase C-mediated invadopodia regulation.15 However, the molecular pathways modulated by SFK phosphorylation of AFAP proteins and their consequences have not been delineated. To address this important question, we analyzed the interactions, pathways and biological consequences of SFK tyrosine phosphorylation of AFAP-1-like-1 (AFAP1L1) in sarcoma cell invadopodia. The SFK member Lyn binds and phosphorylates AFAP1L1 at two tyrosine residues, Y136 and Y566, mediating specific interaction with Vav2, a Rac-guanidine nucleotide exchange factor, and Nck2, the N-WASp/WIP/cortactin-binding adaptor, respectively. AFAP1L1 localizes to invadopodia in sarcoma cells and mutation of its Vav2- and Nck2-binding motifs (Y136 and Y566) mitigates its ability to regulate cytoskeletal changes through Nck2 and Vav2. Knockdown of AFAP1L1 in sarcoma cells inhibits invadopodia re-formation, cell attachment, migration and invasion. Conversely, ectopic expression of AFAP1L1 alters cell movement and promotes mitogen gradient-directed cell migration and an invasive phenotype. These data illustrate a novel pathway that directs Lyn/SFK tyrosine kinase signals to the invadopodia in sarcoma cells through specific pY-motif-mediated interaction of AFAP1L1 with Vav2 and Nck2 to regulate cell migration and invasion.

Results

AFAP1L1 localizes to invadopodia in sarcoma cells and binds F-actin

We used an AFAP1L1-specific antibody (Supplementary Figure S1) to assess AFAP1L1 expression in human osteosarcoma cell lines. Immunoblot analysis detected robust expression of AFAP1L1 in the invadopodia-forming U2OS and MG-63 osteosarcoma cells lines (Figure 1a). Intriguingly, endogenous AFAP1L1 localized to the filamentous actin (F-actin)-rich subnuclear ventral invadopodia structures in U2OS cell (Figure 1b). To facilitate live-cell imaging and manipulation of AFAP1L1 in sarcoma cells (U2OS and MG-63), we generated stable lines expressing enhanced green fluorescent protein (eGFP)-tagged AFAP1L1, selecting cells that expressed levels of the tagged AFAP1L1 similar to endogenous levels. Immunoflourescent analysis of these lines showed that the tagged AFAP1L1 also localized to the F-actin-rich sub/proximal-nuclear ventral invadopodia structures in these sarcoma cells (Figures 1c and d, and Supplementary Movies S1 and S2). Further, when grown in three-dimensional solid extracellular matrix cultures, AFAP1L1 localized to the typical comet-like F-actin-rich invadopodia structures that projected deep into the extracellular matrix (Figure 1e). AFAP1L1 also strongly co-localized with the invadopodia marker cortactin in sarcoma cells U2OS (Figure 1f) and MG-63 (Supplementary Figure S2).

Figure 1
figure1

AFAP1L1 localizes to invadopodia in sarcoma cells, which requires its F-actin-binding/dimerization domain. (a) Immunoblot analysis of AFAP1L1 expression in osteosarcoma (U2OS and MG-63), pancreatic carcinoma (PANC-1) and lung carcinoma (A549) cell lines, relative to β-actin levels. (b) Subcellular localization of AFAP1L1 in U2OS osteosarcoma cells detected with an AFAP1L1-specific antibody. Co-localization of AFAP1L1 (green)- and F-actin (red)-rich sub-nuclear invadopodia is indicated (arrow head), nuclei stained with Hoechst 33342 (blue). Maximum image projection in xy plane and reconstructed image slices at point of strongest co-localization (arrow head) in yz and xz planes. Scale bar=10 μm. (c) Subcellular localization of AFAP1L1 in U2OS osteosarcoma cells stably expressing eGFP-tagged AFAP1L1. Co-localization of eGFP-AFAP1L1 (green)- and F-actin (red)-rich sub-nuclear invadopodia is indicated (arrow head), nuclei stained with Hoechst 33342 (blue). Maximum image projection in xy plane and reconstructed image slices at point of strongest co-localization (arrow head) in yz and xz planes. Scale bar=10 μm. (d) Subcellular localization of AFAP1L1 in MG-63 osteosarcoma cells stably expressing eGFP-tagged AFAP1L1. Co-localization of eGFP-AFAP1L1 (green)- and F-actin (red)-rich sub-nuclear invadopodia is indicated (arrow head), nuclei stained with Hoechst 33342 (blue). Maximum image projection in xy plane and reconstructed image slices at point of strongest co-localization (arrow head) in yz and xz planes. Scale bar=10 μm. (e) Subcellular localization of AFAP1L1 in U2OS osteosarcoma cells stably expressing eGFP-tagged AFAP1L1 cultured in three-dimensional (3D) extracellular matrix (egg white). Co-localization of eGFP-AFAP1L1 (green)- and F-actin (red)-rich sub-nuclear invadopodia is indicated (arrow heads), nuclei stained with Hoechst 33342 (blue). Maximum image projection in xy plane and reconstructed image slices at point of strongest co-localization (arrow head) in yz and xz planes. Scale bar=10 μm. (f) Subcellular localization of AFAP1L1, cortactin and F-actin in U2OS osteosarcoma cells stably expressing eGFP-tagged AFAP1L1. Co-localization of eGFP-AFAP1L1 (green (ii))-, cortactin (amber (iv))- and F-actin (red (iii))-rich sub-nuclear invadopodia is indicated (arrow heads), nuclei were stained with Hoechst 33342 (blue, (i)). Maximum image projection in xy plane and regions of co-localization indicated (arrow head). Merged image (v) shows location of line scan (vi) for quantification of co-localization. Scale bar=15 μm. (g) Subcellular localization of eGFP-AFAP1L1 in COS7 kidney epithelia cells. Co-localization of eGFP-AFAP1L1 (green)- and F-actin (red)-rich structures, nuclei stained with Hoechst 33342 (blue). Maximum image projection in xy plane shown. Scale bar=10 μm. (h) Analysis of direct interaction of the C-terminal region of AFAP1L1 (G-A-CT; aa 630–777) with F-actin by co-sedimentation analysis. Coomassie-stained SDS–polyacrylamide gel electrophoresis (PAGE) gel of high-speed (100 000 g) precipitate (ppt) and supernatants (SN) from F-actin incubated with GST or GST-AFAP1L1-CT (G-A-CT). (i) Y2H analysis of AFAP1L1 self-association. Yeast reporter activation of full-length, C-terminal-deleted (aa 1–627) and N-terminal-deleted (aa 630–777) LexA fusions of AFAP1L1 co-expressing full-length AFAP1L1 as a VP16 fusion. (j) Analysis of domains directing the subcellular localization of AFAP1L1 in U2OS sarcoma cells. The C-terminal leucine-zipper (L) and α-helical (H) domains are both required for localization of AFAP1L1 to actin-rich structures in sarcoma cells. Localization of eGFP-AFAP1L1 full-length (i) and deletion constructs (ii–vi) (green), F-actin (red)-rich structures and nuclei (Hoechst 33342; blue) in transiently transfected U2OS cells. Maximum image projection in xy plane shown. Scale bar=10 μm.

In COS7 cells, which do not readily form invadopodia, AFAP1L1 co-localized with F-actin structures (Figure 1g). Bioinformatic analysis of the AFAP1L1 protein sequence identified a potential F-actin-binding region in the carboxyl leucine-rich/α-helical 620–777 amino acids. To directly test this region for F-actin binding, a co-sedimentation assay was performed using polymerized F-actin and a purified glutathione S-transferase (GST) fusion of the 620–777aa region of AFAP1L1. This region of AFAP1L1 did indeed show strong direct binding to F-actin (Figure 1h). Interestingly, this region of AFAP1L1 also showed the capacity to multimerize when tested by yeast two-hybrid (Y2H) analysis (Figure 1i). We then assessed the ability of this region to direct AFAP1L1 localization to invadopodia in U2OS sarcoma cells (Figure 1j). Importantly, deletion of the C-terminal region (630–777aa) of AFAP1L1 completely eliminated co-localization with all F-actin structures and resulted in localization of AFAP1L1 to the nucleus in U2OS sarcoma cells (Figures 1j (i and ii)). Expression of the C-terminal 620–777aa region alone showed that this region directed a strong co-localization with F-actin (Figure 1j (iii)). However, compared with the full-length molecule, little localization to invadopodia could be detected, indicating that although this region promotes strong F-actin localization, other regions are required for localization to the invadopodia F-actin fraction. Deletion of the individual C-terminal homology domains, the coiled-coil (iv), leucine-rich (v) and α-helical region (vi) showed that removal of the coiled-coil region disrupted localization to invadopodia, although it could still co-localize with other F-actin structures. However, deletion of the leucine-rich or α-helical domains fully disrupted localization to all F-actin structures and the mutated AFAP1L1 now localized to nuclear puncta. Further, the coiled-coil mutant of AFAP1L1 also localized to non-F-actin-containing punctate structures close to the F-actin-containing invadopodia that may be sites of invadopodia formation before localized F-actin assembly.

Downregulation of AFAP1L1 in sarcoma cells inhibits invadopodia formation, cell attachment, migration and invasion

To define the importance of AFAP1L1 to sarcoma cells and invadopodia, we undertook RNA interference (RNAi)-mediated knockdown of AFAP1L1 in U2OS and MG-63 sarcoma cells, and analyzed the biological consequences. Two independent small interfering RNA oligonucleotides (S3 and S4) could efficiently mediate strong reduction in AFAP1L1 protein in both U2OS and MG-63 sarcoma cells (Figure 2a). Knockdown of AFAP1L1 did not significantly affect proliferation of U2OS cells (Supplementary Figure S3). The ability of U2OS cells to attach and spread on tissue culture dishes in the absence of serum was greatly diminished with knockdown of AFAP1L1 as measured by cell footprint area (Figure 2b) and cellular impedance (xCELLigence E-plates) (Figure 2c). Knockdown of AFAP1L1 also reduced the ability of these cells to migrate in response to a serum gradient (on serum-coated surfaces to mitigate their spreading defect in the absence of serum), as measured by Boyden chamber assays (Figure 2d), and their ability to form invadopodia (Figure 2e). In addition, in MG-63 cells there were also significant effects on the cytoskeleton, in particular their ability to reform invadopodia with vanadate stimulation after serum starvation (Figure 2f). Importantly, AFAP1L1 was required for efficient invasion through extracellular matrix coated with serum for U2OS cells as measured in Matrigel-coated Boyden chamber assays (Figure 2g).

Figure 2
figure2

RNAi knockdown of AFAP1L1 in sarcoma cells inhibits cell spreading, migration, invasion and invadopodia formation. (a) Immunoblot analysis of AFAP1L1 protein levels after 72 h of RNAi-mediated knockdown with oligonucleotides S43843 (S3), S43844 (S4) or non-targeting control RNAi (NT) in U2OS and MG-63 cells. (b) Analysis of cell spreading in serum-free and serum-containing cultures at 4 h post plating of U2OS cells with (S3 and S4) or without (NT) RNAi-mediated knockdown of AFAP1L1. Results are mean±s.d., n=3, **P<0.01. (c) Real-time assay for cell attachment and spreading using cell impedance measurements with E-plates on an xCELLigence (ACEA Biosciences Inc.). U2OS cells plated in serum-free and serum-containing media with (S3 and S4) or without (NT) RNAi-mediated knockdown of AFAP1L1. Results are mean±s.d., n=3. (d) Modified Boyden chamber migration assay. U2OS cells with (S3 and S4) or without (NT) RNAi-mediated knockdown of AFAP1L1 were seeded in serum-free media (top chamber) and assayed for migration to serum-containing media (bottom chamber), enumerated after 16 h by counting Hoechst 33342-stained nuclei on the bottom chamber side of the membrane (8 μm pore size). Results are mean±s.d., n=3, **P<0.01. (e) RNAi-mediated knockdown of AFAP1L1 in U2OS sarcoma cells reduces the number of cells showing clusters of invadopodia. U2OS cells after 72 h knockdown of AFAP1L1 (S3 and S4) or non-targeting control (NT) were assayed for invadopodia clusters by F-actin staining. Results are mean±s.d., n=3, *P<0.05, **P<0.01. (f) RNAi-mediated knockdown of AFAP1L1 in MG-63 sarcoma cells alters the clustering of invadopodia and their ability to reform with stimulation (serum+vanadate) after serum starvation. MG-63 cells after 72 h knockdown of AFAP1L1 (S3 and S4) or non-targeting control (NT) were assayed for invadopodia clusters by F-actin staining in growing cells (top panels, quantitation in right graph), and serum starvation (middle panels) and re-stimulation with serum/vanadate (bottom panels, quantitation in right graph). Results are mean±s.d., n=3, *P<0.05. (g) RNAi-mediated knockdown of AFAP1L1 in U2OS and MG-63 sarcoma cells reduces their invasion in modified Boyden chamber assays with a Matrigel matrix coating. U2OS and MG-63 cells with (S3 and S4) or without (NT) RNAi-mediated knockdown of AFAP1L1 were seeded in serum-free media (top chamber) and assayed for invasion through Matrigel matrix-coated filters (8 μm pore size) in to serum-containing media (bottom chamber, 10% fetal bovine serum), enumerated after 48 h by counting Hoechst 33342-stained nuclei on the bottom chamber side of the membrane. Results are mean±s.d., n=3, *P<0.05.

As AFAP1L1 knockdown in U2OS cells impeded invadopodia formation, we tested whether re-expression of wild-type AFAP1L1 or mutants of AFAP1L1 could rescue this phenotype. Indeed, U2OS cells treated with the S3 RNAi and transfected with mouse AFAP1L1 reformed invadopodia structures (Supplementary Figures S4 (i and ii)). However, expression of the C-terminal-deleted mutant (AA 1–630) or the C-terminal region alone (AA 620–777) failed to re-form invadopodia structures (Supplementary Figures S4 (iii and iv)).

AFAP1L1 interacts with and is phosphorylated by the SFK Lyn, and stimulation of tyrosine phosphorylation regulates total AFAP1L1 localization to invadopodia

Through a Y2H screen, AFAP1L1 was isolated as a binding partner of the SFK member Lyn (Figure 3a). This assay showed that Lyn bound to AFAP1L1 through its SH3 domain (Figure 3a). This interaction was supported through co-immunoprecipitation and co-localization of Lyn and AFAP1L1 when ectopically expressed in COS-7 cells (Figures 3b and c). Interestingly, co-expression with Lyn increased the level of tyrosine phosphorylation of AFAP1L1, suggesting that Lyn might phosphorylate AFAP1L1 (Figure 3b). This possibility was tested directly with kinase assays using immunoprecipitates of inactive (Y397F) or hyperactive (Y508F) Lyn and a purified GST fusion of AFAP1L1. Significantly, these experiments demonstrate that active Lyn could interact with and directly phosphorylate AFAP1L1 (Figure 3d).

Figure 3
figure3

AFAP1L1 binds and is phosphorylated by Lyn, and tyrosine kinase activation promotes total AFAP1L1 localization to invadopodia. Interaction of Lyn with AFAP1L1 and the ability of the kinase to phosphorylate AFAP1L1. (a) Using the Y2H technology, the SH3 domain of Lyn binds to AFAL1L1. (b) Co-immunoprecipitation of Lyn and AFAP1L1 from transiently transfected COS7 cells. Maximum co-immunoprecipitation occurs with hyperactive Lyn (LynY508F mutant) that also induces a maximal phosphorylation of AFAP1L1. (c) Co-localization of Lyn and AFAP1L1 in actin-rich structures (white arrowheads) in transiently transfected COS7 cells. Scale bar=10 μm. (d) Lyn can directly phosphorylate AFAP1L1. Inactive (Y397F) and hyperactive (Y508F) Lyn were immunoprecipitated from transiently transfected COS7 cells and using in vitro kinase assays with purified recombinant AFAP1L1 (as a GST fusion). (e) Subcellular localization of eGFP-AFAP1L1 (green) in live cell cultures of U2OS osteosarcoma cells during vanadate stimulation. Sub-nuclear invadopodia are indicated (arrow heads), nuclei stained with Hoechst 33342 (blue). Maximum image projection in xy plane. Scale bar=10 μm.

When U2OS (expressing eGFP-AFAP1L1) sarcoma cells were starved overnight and then stimulated with vanadate and visualized by live-cell imaging, AFAP1L1 localized to invadopodia within 5 min of stimulation and continued to accumulate in invadopodia over a 30-min time course (Figure 3e and Supplementary Movies S3 and S4).

Heterologous expression of AFAP1L1 promotes anchorage-independent growth, serum-directed migration and an invasive phenotype

Stable ectopic expression of AFAP1L1 was undertaken in NIH3T3 cells, as they did not express any detectable endogenous AFAP1L1 and are commonly used as a non-transformed cell line that are sensitive to oncogenic transformation, especially by SFK pathway oncogenes. This showed enforced expression of AFAP1L1 resulted in anchorage-independent growth (cellular transformation) as demonstrated by markedly enhanced colony formation in soft-agar assays (Figure 4a). This observation indicates that raising the level of AFAP1L1 may promote oncogenic transformation. In NIH3T3 cells expressing AFAP1L1, increasing Lyn activity via wild-type Lyn expression resulted in altered cell morphology. This was accentuated further with hyperactive Lyn (LynY508F) (Figure 4b). Interestingly, these cells also displayed a more ‘transformed’ spindle-shaped phenotype with reduced spreading on tissue culture dishes (Figure 4b). In scratch/wound-healing assays, elevated expression of AFAP1L1 led to a decrease in migration rate of NIH3T3 cells (Figure 4c).

Figure 4
figure4

AFAP1L1 has oncogenic potential and significantly alters the cytoskeleton, cell attachment/spreading and migration/invasion. (a) Soft agar anchorage-independent growth assay of NIH3T3 cells expressing AFAP1L1. Phase-contrast images of vector-only control (Con) and cells expressing AFAP1L1 after 7 days growth in soft agar (i), with enumeration of colonies formed (ii); results are mean ±s.d., n=3, **P<0.01. (b) Effect of expression of AFAP1L1 alone or in combination with wild-type Lyn (LynWT), inactive Lyn (LynY397F) or hyperactive Lyn (LynY508F) on NIH3T3 cell morphology. Phase-contrast images are shown. Scale bar=20 μm. (c) Wound-healing/scratch assay of NIH3T3 cells expressing AFAP1L1 and control cells. The number of cells in quadrants that had migrated towards the scratch was enumerated at 16 h post initiation of scratch assay. Results are mean±s.d., n=8, **P<0.01. (d) Real-time assay for cell attachment and spreading using cell impedance measurements with E-plates on an xCELLigence (ACEA Biosciences Inc.). HEK293 cells expressing AFAP1L1 or control (Con) cell plates in the presence or the absence of serum as indicated. Results are mean, n=3. (e) Analysis of spontaneous cell movement in low confluence cultures of HEK293 cells expressing AFAP1L1 and control cells (Con). Individual cell movement (n=30) for each cell type was analyzed at 1-min intervals over 16 min using Gradientech Tracking Tool (v1.70, Gradientech). (f) Morphological analysis of HEK293 cells expressing eGFP-tagged AFAP1L1 or vector-only control (Con) grown on solid extracellular matrix (egg white) for 7 days. Fluorescent images are shown. Scale bar=50 μm. (g) Modified Boyden chamber migration assay. HEK293 cells expressing AFAP1L1 or control (Con) cells in serum-free media (top chamber) migrating to serum-containing media (bottom chamber) were enumerated after 16 h. Results are mean±s.d., n=3, **P<0.01. Results are mean, n=3. (h) Real-time assay for cell migration using cell impedance measurements with CIM plates (8 μm pore size) on an xCELLigence (ACEA Biosciences Inc.). HEK293 cells expressing AFAP1L1 or control (Con) cells in serum-free media migrating to serum-containing media. Results are mean, n=3.

We also tested the effect of overexpression of AFAP1L1 in HEK293 cells, which express low levels of AFAP1L1, but are also non-transformed similar to NIH3T3 cells. In these cells, overexpression of AFAP1L1 did not significantly influence proliferation of the cells (Supplementary Figure S5). We found significant influence of AFAP1L1 expression on cell attachment, spreading, movement, migration and response to extracellular matrix (Figures 4d–f and Supplementary Figures S6a and b). Elevated AFAP1L1 reduced the ability of HEK293 cells to attach and spread (Figure 4d). Further, tracking the cell movement in low confluence cultures showed that AFAP1L1-overexpressing cells had a significantly reduced propensity for stochastic movement (Figure 5e), similar to the reduced migration (in response to release from contact inhibition) of NIH3T3 cells in the scratch/wound-healing assay (Figure 4c). Interestingly, these cells also had a dramatic response to extracellular matrix—elevated AFAP1L1 induced an invasive phenotype where cells grew into the solid matrix (Figure 4f). Further, HEK293 cells expressing AFAP1L1 grew to larger and more pleomorphic colonies in liquid cultures supplemented with the extracellular matrix (Supplementary Figures S6a and b). Elevated expression of AFAP1L1 significantly increased migration in response to a serum gradient in Boyden chamber assays (Figures 4g and h).

Figure 5
figure5

AFAP1L1 phosphorylated by Lyn generates specific pY motifs for binding Vav2 (Y136) and Nck2 (Y566). (a) Interaction of Vav2 with AFAP1L1 through both its SH3 and SH2 domains. (i) Y2H assay of Vav2 SH3 domain interacting with a proline-rich sequence in AFAP1L1. (ii) Phospho-tyrosine Y2H assay of Vav2 SH2 domain binding the phosphorylated Y136 motif of AFAP1L1. (iii) Co-immunoprecipitation of Vav2 and AFAP1L1 from transiently transfected COS7 cells. Maximum co-immunoprecipitation occurs with vanadate stimulation to induce phosphorylation of AFAP1L1, the Y136F mutation of AFAP1L1 significantly reduces co-immunoprecipitation of Vav2 and AFAP1L1. (iv) Co-localization of Vav2 and AFAP1L1 in actin-rich structures (white arrowheads) in transiently transfected COS7 cells stimulated with vanadate. Scale bar=10 μm. (v) Direct binding of labeled Vav2-SH2 domain via protein blotting of lysates from transiently transfected COS7 cells expressing hyperactive Lyn (Y508F) and either wild-type AFAP1L1 or the Y136F mutant. (b) Interaction of Nck2 with AFAP1L1 through both its SH2 domain. (i) Phospho-tyrosine Y2H assay of Nck2 SH2 domain binding the phosphorylated Y566 motif of AFAP1L1. (ii) Co-immunoprecipitation of Nck2 and AFAP1L1 from transiently transfected COS7 cells. Co-immunoprecipitation occurs with vanadate stimulation to induce phosphorylation of AFAP1L1; the Y566F mutation of AFAP1L1 eliminates detectable co-immunoprecipitation of Nck2 and AFAP1L1. (iii) Co-localization of Nck2 and AFAP1L1 in actin-rich structures (white arrowheads) in transiently transfected COS7 cells stimulated with vanadate. Scale bar=10 μm. (iv) Direct binding of labeled Nck2-SH2 domain via protein blotting of lysates from transiently transfected COS7 cells expressing hyperactive Lyn (Y508F) and either wild-type AFAP1L1 of the Y566F mutant. (c) In transiently transfected COS7 cells, phosphorylation of AFAP1L1 by LynY508F induces significant cytoskeletal changes that are dependent on both the Vav2- and Nck-binding Y136 and Y566 motifs. Phosphorylation of the AFAP family member AFAP1 by LynY508F induces podosomes in COS7 cells. (d) Immunoblot analysis of AFAP1L1 immunoprecipitated from U2OS cells stimulated with vanadate (VO4) probed for tyrosine phosphorylation (pY) and compartmentalization of AFAP1L1 to cytosolic/membrane (Cyt) and F-actin (Fil) subcellular fraction. (e) Immunoblot analysis of AFAP1L1 immunoprecipitated for association of Vav2 from U2OS cells stimulated with epidermal growth factor and fibronectin (EGF/FN) for the times (min) indicated. (f) Immunoblot analysis of Nck2 and AFAP1L1 immunoprecipitated from U2OS and MG-63 cells stimulated with vanadate (VO4). (g) Immunoblot analysis of Vav2 and AFAP1L1 immunoprecipitated from U2OS and MG-63 cells stimulated with vanadate (VO4), compared with total levels in lysates (L). (h) Subcellular localization of AFAP1L1 (green) and ((i) left panel) Vav2 (red) or ((ii) right panel) Nck2 in U2OS osteosarcoma cells after vanadate stimulation (1 h). Sub-nuclear invadopodia are indicated (arrow heads), nuclei stained with Hoechst 33342 (blue). Maximum image projection in xy plane. Scale bar=10 μm.

Phosphorylation of AFAP1L1 by Lyn/SFK links directly to Vav2 and Nck2, and the AFAP1L1 PH domains mediate PIP binding

We used our phosphotyrosine (pY) Y2H system16 using AFAP1L1 as the bait during co-expression with the active Lyn kinase domain and tested against a Y2H cDNA library to uncover what pathways phosphorylated AFAP1L1 intersected. This screen revealed that the SH2 domains of Nck2 and Vav2 could bind to AFAP1L1 in a phosphotyrosine-dependent manner (Figures 5a and b). Clones encoding Vav2 from the pY-Y2H screen encompassed both its C-terminal SH3 and SH2 domains. When tested independently, both domains bound to AFAP1L1, with the SH3 domain of Vav2 binding a classical PxPxxP motif within AFAP1L1 close to the tyrosine motif (Y136) responsible for specific binding to the SH2 domain of Vav2 (Figures 5a (i and ii)). Importantly, this interaction of Vav2 and AFAP1L1 was recapitulated in co-immunoprecipitation and co-localization experiments in COS7 cells expressing Vav2 and AFAP1L1 (Figures 5a (iii and iv)). Further, a purified NusA fusion of the SH2 domain of Vav2 (tagged with a fluorescent label) bound to the AFAP1L1 band in lysates from COS7 cells co-expressing AFAP1L1 and hyperactive Lyn (Y508F), whereas the Y136F mutant of AFAP1L1 did not show any binding (Figure 5a (v)).

Clones encoding Nck2 from the pY-Y2H screen only encompassed its C-terminal SH2 domain and mutagenesis revealed that this binding was dependent on tyrosine 566 of AFAP1L1 (Figure 5b (i)) and this interaction was confirmed in co-immunoprecipitation and co-localization experiments in COS7 cells (Figures 5b (ii and iii)). In addition, a purified NusA fusion of the SH2 domain of Nck2 (tagged with a fluorescent label) also bound to the AFAP1L1 band in lysates from COS7 cells co-expressing AFAP1L1 and hyperactive Lyn (Y508F), but not with the Y566F mutant of AFAP1L1 (Figure 5b (iv)). Collectively, these data demonstrate that Nck2 can bind AFAP1L1 via the C-terminal SH2 domain interacting specifically with the pY566 motif of AFAP1L1.

Interestingly, COS7 cells showed major alterations to their morphology when AFAP1L1 and hyperactive Lyn were co-expressed—the cells displayed a small rounded appearance, which was not observed when kinase inactive Lyn was co-expressed with AFAP1L1 (Figures 5c (i and ii)). Significantly, only when both sites (Y136 and Y566) were mutated did the COS7 cell morphology revert to their normal appearance (Figures 5c (iii and v)). Interestingly, in COS7 cells the related AFAP1L1 molecule AFAP1 mediated the formation of peripheral podosomes when co-expressed with hyperactive Lyn (Figure 5c (vi)).

Having established a molecular pathway regulated by phosphorylation of AFAP1L1 by Lyn/SFK through specific recruitment of Vav2 and Nck2 using heterologous systems, we now sought to determine whether this pathway held true in sarcoma cells that endogenously express AFAP1L1. Interestingly, eGFP-tagged AFAP1L1 in sarcoma cells showed highly dynamic recruitment into invadopodia and movement along stress fibers (Figure 3e and Supplementary Movies S3 and S4). Further, vanadate stimulation promoted tyrosine phosphorylation of endogenous AFAP1L1 in U2OS cells (Figure 5d). This phosphorylation was sensitive to SFK inhibition by the inhibitor PP2, using an antibody that specifically recognizes the pY136 site of AFAP1L1 (Supplementary Figure S7). Consequently, we then looked to see whether Vav2 and Nck2 associated with AFAP1L1 after vanadate stimulation, and found that this was indeed the case with both Vav2 and Nck2 co-immunoprecipitating with AFAP1L1 in U2OS and MG-63 sarcoma cells (Figures 5f and g). Importantly, growth factor/integrin stimulation of U2OS cells with EGF and fibronectin also induced Vav2 co-immunoprecipitation with AFAP1L1 (Figure 5e). Further, we also observed significant co-localization of both Vav2 and Nck2 with AFAP1L1 in stimulated U2OS cells (Figures 5h (i and ii)). The importance of the Y136 and Y566 sites for AFAP1L1 involvement in invadopodia formation was tested on U2OS cells treated with the S3 RNAi, to knock down endogenous AFAP1L1, and then transfected with mouse AFAP1L1 with the Y136 and Y566 sites mutated to phenylalanine. This showed that the Y136F/Y566F mutant of AFAP1L1 was unable to re-form invadopodia structures unlike the wild-type AFAP1L1 protein (Supplementary Figures S4 (i and vi)). These data provide support for a novel pathway of AFAP1L1 leading to Vav2 and Nck2 being present in U2OS and MG-63 sarcoma cells, localizing to invadopodia, which is activated when cells are stimulated through vanadate or growth factor/integrin ligation.

AFAP1L1 contains two pleckstrin homology (PH) domains (Figure 1i) that could each directly bind to phospholipids when assayed using PIP strips (Echelon, Salt Lake City, UT, USA) (Supplementary Figure S8a). Moreover, the PH domains could mediate plasma membrane localization in platelet derived growth factor-stimulated NIH3T3 cells, demonstrating that they are functionally capable of binding phosphatidyl inositol lipids (PIPs) generated by phosphoinositide 3-kinase (Supplementary Figure S8b). In addition, expression of mouse AFAP1L1 with its PH domains deleted (ΔPH1/2) in U2OS cells that had endogenous AFAP1L1 knocked down with the S3 RNAi were unable to re-form invadopodia structures unlike cells transfected with wild-type AFAP1L1 (Supplementary Figures S4 (i and v)). Together, these results show AFAP1L1 can use its PH domains to mediate localization to the plasma membrane of cells and are an important component of its signaling to invadopodia.

AFAP1L1 signals to Rac and PAK/myosin light chain kinase pathways

The interaction of AFAP1L1 with Vav2, which has Rho family guanine nucleotide exchange activity, suggested AFAP1L1 might regulate downstream pathways from Vav2, including Rac1 and p21-activated kinase (PAK) networks to mediate its effects on the cytoskeleton. Significantly, ectopic expression of AFAP1L1 in COS7 cells increased the levels of active Rac1 but not RhoA in PAK1-PBD and Rhotekin-PBD pull-down assays (Figure 6a). Further, on extended vanadate stimulation (1 h) or co-expression with hyperactive Lyn (Y508F), a greatly diminished pool of active Rac1 was observed that was not mediated by vanadate stimulation or LynY508F alone. This observation suggests that high levels of phospho-AFAP1L1 can result in feedback inhibition and consequently downregulation of active Rac1.

Figure 6
figure6

Signaling through AFAP1L1 regulates Rac, PAK and MLC2. (a) Rac1 and RhoA activation assays from COS7 cells transiently transfected with hemagglutinin (HA)-tagged AFAP1L1 and inactive Lyn (LynY397F) or hyperactive Lyn (LynY508F), or stimulated with vanadate (VO4). PAK1-PBD beads were used to pull down active Rac1, whereas Rhotekin-PBD beads were used to pull down active RhoA and immunoblots probed for Rac1 and RhoA, respectively. Relative Rac1 and RhoA activity is measured as a ratio of Rac1/RhoA in the pull down relative to total Rac1/RhoA in the lysate. Assays performed in duplicate (mean±s.e.m.). (b) Alterations to phosphorylation status of p21-activated kinases PAK1/2/3 and MLC2 in cells expressing AFAP1L1 and stimulated with vanadate. Immunoblot analysis of COS7 cells transiently transfected with or without HA-tagged AFAP1L1 and stimulated with vanadate (VO4) for the times indicated. The level of phosphorylated PAK to total PAK was quantitated by densitometry (graph). (c) RNAi-mediated knockdown of AFAP1L1 in U2OS sarcoma cells regulates PAK activity. Immunoblot analysis of U2OS cell lysates after 72 h of RNAi-mediated knockdown with oligonucleotides S43843 (S3), S43844 (S4) or non-targeting control RNAi (NT), stimulated with or without vanadate (VO4) for 1 h. (d) RNAi-mediated knockdown of AFAP1L1 in U2OS sarcoma cells regulates PAK activity and MLC2 phosphorylation, and subcellular compartmentalization. Immunoblot analysis of cytosolic (Cyt) and F-actin containing (Fil) subcellular fractions of U2OS cell lysates after 72 h of RNAi-mediated knockdown with oligonucleotides S43843 (S3), S43844 (S4) or non-targeting control RNAi (NT), stimulated with or without vanadate (VO4) for the times indicated.

Downstream of Rac1 are the PAK1/2/3.17 Therefore, we assayed their activation status in COS7 cells expressing AFAP1L1 during a time course of vanadate stimulation. Here we found that AFAP1L1 expression led to significant activation of PAK1/3 (with modest PAK2 activation), which was further enhanced by vanadate stimulation up to 30 min, and then subsequently downregulated by 1 h of treatment (Figure 6b). In contrast, cells not ectopically expressing AFAP1L1 only showed minimal PAK1/3 activation but pronounced PAK2 activation (Figure 6b). The transient PAK activation (with subsequent downregulation at 1 h) of vanadate-stimulated AFAP1L1-expressing cells correlates with the observed downregulation of Rac1 activity at this time point (Figure 6a). Downstream of the PAKs are several pathways, including the phosphorylation and inhibition of myosin light chain (MLC) kinase (which phosphorylates MLC), which is directed toward cytoskeletal reorganization.18 Consequently, we assessed the levels of MLC2 phosphoryaltion in cells expressing AFAP1L1 and found that indeed the increased PAK1/3 levels correlated with decreased phospho-MLC2 levels, and that although vanadate stimulated an increase of pMLC2 in control cells, those ectopically expressing AFAP1L1 showed significant inhibition of this response (Figure 6b).

With the significant effects of AFAP1L1 on cytoskeletal function and that ectopic expression of AFAP1L1 affects PAK/MLC2 we then looked and the effects of knockdown of AFAP1L1 on the signaling to PAK/MLC2. Importantly, knockdown of AFAP1L1 reduced the activation of PAK1/3, while increasing that of PAK2, when U2OS cells were stimulated with vanadate (Figures 6c and d). Commensurate with these changes in PAK isoform activation were also significant alterations to the level of phosphorylation and subcellular distribution of MLC2, with knockdown of AFAP1L1 causing less pMLC2 and total MLC2 being recruited to the F-actin compartment, resulting in more pMLC2/MLC2 remaining in the cytoplasm (Figure 6d). These data illustrate that AFAP1L1 provides a major signaling mechanism to regulate major cytoskeletal remodeling in sarcoma cells, especially through modulating PAK activity and MLC2 loading onto F-actin, potentially primarily through its interaction with Vav2.

Discussion

We report that the scaffolding protein AFAP1L1 directly links Vav2 and Nck2 pathways to regulate invadopodia formation and function in sarcoma cells. This is through our novel findings that two specific pY motifs phosphorylated by Lyn/SFKs in AFAP1L1, pY136 and pY566, bind to the SH2 domains of Vav2 and Nck2, respectively. This provides a regulatory mechanism for linking the downstream pathways of Vav2 through Rac1-PAK1/3 and Nck2 regulation of F-actin via the WIP/cortactin/Tks5//N-WASP/actin-related protein complex-2/3 complex,19, 20, 21, 22, 23, 24, 25 within the sarcoma cell invadopodia (summarized in Figure 7). These results provide an important mechanistic explanation for the regulation of invadopodia by AFAP1L1 in sarcoma cells. Our novel findings also significantly expand the recent findings on AFAP1L1 interacting with vinculin at invadopodia and its regulation of disease progression in colorectal cancers,26 and provide an impetus to delineate whether the Vav2/Nck2 connection to AFAP1L1 also mediates its links to colorectal cancer progression.

Figure 7
figure7

Schematic of AFAP1L1 signaling network in sarcoma cells. Schematic model of pathways intersected by AFAP1L1 in sarcoma cells to regulate invadopodia and invasion/migration. Activation through ligation of growth factors and extracellular matrix (GF/ECM) activate Lyn/Src family tyrosine kinases (Lyn/SFK), the phosphoinositide 3 (PI3) kinase pathway and protein kinase C (PKC), leading to phosphorylation of AFAP1L1 on serines (S) in the inter-PH domain region, tyrosines (Y) at Y136 and Y566, and promotion of location to phospholipid-enriched membranes. AFAP1L1 also has strong constitutive interaction with F-actin via C-terminal leucine-rich (L) and α-helical (H) regions that also mediate multimerization of AFAP1L1, which may facilitate cross-linking (X-linking) of F-actin fibers. The coiled-coil (C) region of AFAP1L1 may provide additional signals for localization to invadopodia-localized F-actin. A proline-rich N-terminal region (P) of AFAP1L1 can bind the SH3 domains of Src family kinases (SFK), Vav2, cortactin and Tks5. The SH2 domain of Vav2 binds to the phosphorylate Y136 motif and can lead to activation of a Rac1 (or possibly cdc42)–Pak–MLC kinase (MLCK) pathway. Further, the SH2 domain of Nck2 binds to the phosphorylated Y566 motif and with Nck2 forming a complex with cortactin, N-WASP, actin-related protein complex-2/3 (Arp2/3) and Tks5. Consequently, Vav2 and Nck2 interaction with AFAP1L1 facilitates close connection of multiple cytoskeletal-modulating proteins to bring about regulation of migration and invasion in sarcoma cells.

AFAP1L1 contains a direct F-actin binding and multimerization region in its C-terminus and two functional central PH domains that bind PIPs generated by growth factor-stimulated phosphoinositide 3-kinase, providing direct links to both the F-actin cytoskeleton (including the ability to cross-link the cytoskeletal network) and tethering to growth factor-activated plasma membrane loci. In addition, as AFAP1L1 is able to form dimers (and potentially higher-order multimers), and through its proline-rich motifs that interact with several SH3 domains, that is, Lyn, Vav2 and cortactin,8 this could facilitate simultaneous co-localization of multiple distinct AFAP1L1 complexes within the same tethered space. AFAP1L1 could also be directly interacting with Tks5 and p190RhoGAP within invadopodia via their SH3 domains through binding to PXP motifs within AFAP1L1, similar to that reported for AFAP1.27 These observations together with the ability of AFAP1L1 to directly bind the invadopodia/podosome marker cortactin8, associate with vinculin (found at focal adhesions as well as invadopodia)26 and the fact that AFAP1L1 encompasses a homologous site to that which is phosphorylated by protein kinase C in AFAP1 (which regulates podosome and potentially invadopodia lifespan),11 we now have significant evidence for placing AFAP1L1 as a crucial regulator of invadpodia in sarcoma cells, and that it directly intersects multiple critical invadopodia regulatory networks. The strongest evidence for the importance of AFAP1L1 for invadopodia regulation, from other investigators8, 9, 26 and as we report here, relates to the U2OS cell line. This suggests there are important cell type-specific ancillary components that may be direct interacting partners of AFAP1L1, which help mediate this cell-specific phenotype.

The regulation of invasion and metastasis by the invadopodia2, 4, 5, 28 is complex and involves many factors that are also involved in other cytoskeletal functions, for example, attachment and migration,2, 4, 5, 6, 7, 12, 29, 30, 31, 32, 33 and this includes the AFAP1 family (encompassing AFAP1, AFAP1L1 and AFAP1L2/XB130), which are scaffold-forming proteins providing direct network connections that facilitate cytoskeletal regulation.9, 11, 27, 34, 35 The initiation of invadopodia assembly involves cdc42 activity to facilitate targeting and activation of actin nucleation complexes (for example, actin-related protein complex-2/3, N-WASp and WIP) and tyrosine phosphorylation of substrates by SFKs (for example, cortactin, N-WASp, paxillin and Tks5) to enhance invadosome assembly.1, 2, 4, 5, 28, 36 Placing AFAP1L1, with its direct binding to Nck2 and Vav2 when phosphorylated by SFKs(Lyn), within the invadopodia provides further links between their other known direct partners (that is, for Nck2, N-WASp, WIP, cortactin and Tks5) and the Rac1/PAK pathway via Vav2 within the invadosome.10, 21, 23, 24, 25, 31, 37, 38 Although Nck and its partners are central factors in formation and maintenance of invadopodia,19, 24, 39 the involvement of Rac1 activators such as Vav2 has been less well characterized and, indeed, a recently identified Trio-Rac1-Pak1 pathway appears strongly associated with the disassembly of invadopodia.23 Intriguingly, there is a link between Rac1 and Nck in promoting actin nucleation via causing disassociation of the WASP-family verprolin homologous protein (WAVE1) complex,40 and potentially through AFAP1L1, bringing together the Nck and Rac1 (via Vav2) complexes, it could also regulate the actin polymerization dynamics within invadopodia. The phosphorylation status of AFAP1L1, and thus its recruitment/regulation of Nck2 and Vav2 pathways, may also regulate the type of cell-matrix complexes that a cell forms, that is, focal adhesions or invadopodia/podosomes, as several of its binding partners (for example, vinculin, F-actin and SFKs) can be found in both types of structures. This may also facilitate the much higher temporal dynamics of invadopodia/pososomes compared with more stabilized actin structures as occur in focal adhesions.

Interestingly, although overexpression of AFAP1L1 in HEK293 cells led to overall increases in Rac1 and PAK1/3 activity, after sustained stimulation (1 h vanadate, co-expression dominant active LynY508F, Figures 4a and b), Rac1 and PAK1/3 activity was decreased, showing that depending on the duration of SFK activity AFAP1L1 can up- and downregulate Rac1/PAK1/3 activity. Manipulating AFAP1L1 levels also had profound effects on the differential activation of PAK isoforms, PAK1/3 and PAK2, and differential compartmentalization of the PAK substrate MLC2 (cytosolic and F-action subcellular fractions). Potentially, AFAP1L1 could be both promoting assembly through Nck2 pathways and regulating disassembly through Vav2/Rac1 pathways within invadopodia, providing dynamic invadopodia turnover, which is important for invasion of cancer cells, as increased invadopodia turnover through Rac1 enhances this process.41

AFAP1L1 also had significant enhancing effects on mitogen gradient-directed migration, which is also intimately controlled by Vav2/Rac1/Pak1.22, 42, 43, 44 Consequently, AFAP1L1 could be providing an integration of both migration/locomotion control and invasion/invadopodia regulation, thus linking these two important aspects of cancer cell invasion and metastasis. Indeed, the pathways feeding into the AFAP1L1 complex formation and localization, that is, SFKs and phosphoinositide 3-kinase, are both intimately linked to promoting migration as well as invasion.1, 45, 46, 47

In addition, our data show that signaling from AFAP1L1 is modulated by a cell’s contact with the extracellular matrix, which could explain the differential effects on cell growth in two-dimensional monolayer cultures (which show no major effects on cell proliferation) compared with three-dimensional and in vivo cultures (which show significant effects on proliferation/survival) of cells over/underexpressing AFAP1L1.8, 26

Although high expression of total AFAP1L1 is associated with sarcomas (and colorectal cancer progression),8, 26 it will be important to determine whether the activation status (that is, phosphorylation of Y136 and Y566 sites, for which we have generated specific antibodies) also correlate with disease status and clinical outcomes. Important clinical applications of the molecular understanding we now have of the AFAP1L1 pathway need to be addressed through the use of both advanced cancer mouse models with altered AFAP1L1 (for example, sarcomas bearing wild-type or Y136/566F mutant AFAP1L1) and identifying therapeutic avenues to disrupt the AFAP1L1 pathway in sarcoma and potentially other (for example, colorectal cancer) patients.

Materials and methods

Cell culture, stable cell line generation and in vivo assays

U2OS, MG-63, A549, PANC-1, HEK293, NIH3T3 and COS7 cells (sourced from ATCC (Manassas, VA, USA) and mycoplasma free) were maintained in Dulbecco’s modified Eagle’s medium supplemented with penicillin and streptomycin (Invitrogen, Carlsbad, CA, USA), 10% fetal bovine serum (Gibco, Carlsbad, CA, USA) and cultured at 37 °C. Stable cell lines expressing eGFP-tagged mouse AFAP1L1 were generated by selecting eGFP-positive cells after 7 days of culture by fluorescence-activated cell sorting (BD FACS Aria II flow cytometer, Beckman-Coulter, Palo-Alto, CA, USA). NIH3T3 cell lines expressing different Lyn mutants were generated by retroviral infection (pMSCV2.2neo-Lyn; wild-type, kinase-inactive Y397F, or dominant-active Y508F Lyn) and selecting lines using neomycin (1 mg/ml) for 10 days. Egg whites were used as a source of extracellular matrix for three-dimensional cultures of cells on coverslips, as described.48 For transient expression of plasmids, cells were transfected using Lipofectamine2000 as per the manufacturer’s instructions (Life Technologies, Carlsbad, CA, USA). Cell proliferation assays were performed using the IncuCyte Zoom (Essen Biosciences Inc., Ann Arbor, MI, USA) system, the cell screen (Innovatis Inc., Salzburg, Germany) system or the xCELLigence (ACEA Biosciences Inc., San Diego, CA, USA) system, according to the manufacturer’s protocol.

Soft agar cultures of NIH3T3 clones stably expressing untagged mouse AFAP1L1 or control vector (pMSCV2.2neo) were performed essentially as described.49

Spontaneous cell movement in sub-confluent cultures was analyzed on timed series (1 min intervals) phase-contrast images of live cell cultures using Gradientech Tracking Tool (v1.07, Gradientech, Upsala, Sweden).

Wound-healing/scratch assays were performed on NIH3T3 cells by growing cells to confluence in six-well trays and then dragging a 200-μl pipette tip across the monolayer to remove a track of cells.50 Phase-contrast images were then taken at time points over a 16-h period and the degree of scratch closure quantified.

Cell attachment and spreading analysis was performed by plating cells on untreated tissue-culture plastic in serum-free media or media supplemented with 10% fetal bovine serum, phase-contrast images captured at 4 h post plating and the percentage of cells attaching and spreading quantified.

Modified Boyden chamber cell migration and invasion assays were performed in 24-well plates (8 μm pore size, BD Biosciences, Franklin Lakes, NJ, USA) coated with fetal bovine serum before adding cells (2.0 × 104) in the top chamber in serum-free media. For invasion assays, chambers were coated with a thin layer of Matrigel (BD Biosciences), 10 μl of Matrigel (1 mg/ml) was layered and allowed to set on the filter chamber membrane, then air-dried overnight. The lower chamber was filled with media containing 10% fetal bovine serum and cells that had migrated after 16 h to the underside of the filter were fixed, visualized and enumerated.

Real-time cell attachment, spreading, proliferation and migration assays were also performed using an xCELLigence RTCA DP system with E-plate-16 (for attachment/spreading assays) or CIM-plate-16 (for migration assays, 8 μm pore size), according to the manufacturer’s instructions (ACEA Biosciences Inc.), and analyzed using RTCS Software (v1.2, ACEA Biosciences Inc.).

Plasmid construction and in vitro assays

All plasmid constructs were generated by subcloning and/or site-directed mutagenesis using oligonucleotides and were confirmed by Sanger sequencing. Expression constructs for Lyn and the standard and phospho-tyrosine-specific Y2H assays were as previously described.16, 51, 52, 53, 54, 55, 56, 57

The SH2 domains of Nck2 and Vav2 were expressed as NusA fusions and purified using specific affinity columns (GE Healthcare, Little Chalfont, UK) and gel filtration using Profinia (Bio-rad, Hercules, CA, USA) and BioLogic DuoFlow chromatography (Bio-rad) instruments. The NusA-fusions were labeled with fluorescent probes for use as direct immunoblotting reagents using the IRDye-800CW protein labeling kit (LI-COR Biosciences, Lincoln, NE, USA). The purified C-terminal GST fusion of AFAP1L1 was also used in F-actin co-sedimentation assays using the Actin Binding Protein Biochem Kit (BK013, Cytoskeleton, Denver, CO, USA).

Kinase assays were performed using Lyn immunoprecipitated from COS7 cells transfected with inactive Lyn (LynY397F) or hyperactive Lyn (LynY508F) using anti-Lyn antibodies (sc-15, Santa Cruz Biotechnology, Dallas, TX, USA) and subsequent incubation with [γ-32P]-ATP in the presence or the absence of substrate (GST or GST-AFAP1L1) essentially as described previously.55

Rac1 and RhoA activation status was measured using the Rac1 and RhoA activation assay kits (BK035, BK036, Cytoskeleton), according to the manufacturer’s instructions, on COS7 cells transiently transfected with combinations of the hemagglutinin-tagged AFAP1L1, inactive Lyn (LynY397F) and hyperactive Lyn (LynY508F) mammalian expression plasmids at 48 h post transfections, either with or without vanadate stimulation for 1h.

Antibodies and reagents

Polyclonal rabbit antibodies were raised against AFAP1L1 using the purified C-terminal region (aa 627–777) of murine AFAP1L1 fused to GST as the antigen and high-titer serum was purified on protein-A beads (antibody specificity is detailed in Supplementary Figure S1). Additional antibodies used for immunoblotting and/or immunoprecipitation were anti-pY (PY100; 9411), anti-PAK1/2/3 (2604), anti-MLC2 (3672), anti-pMLC2 (pT18/pS19, 3674), pPAK1/2 (pS144/pS141, 2606) (Cell Signaling Technology, Danvers, MA, USA); anti-β-actin (AC-15, ab6276), anti-Nck2 (ab109239), anti-Vav2 (ab52640) (Abcam, Cambridge, UK); anti-Lyn (sc-15), anti-AFAP1L1 (sc-162497), anti-cortactin (H-191, sc-11408), anti-GFP(FL, sc-8834) (Santa Cruz Biotechnology); and anti-pY-HRP (4G10), anti-Myc-tag (clone 4A6), anti-hemagglutinin-tag (HA.11 clone) (Merck-Millipore, Billerica, MA, USA). TRITC-labeled phalloidin (P1591, Sigma-Aldrich, St Louis, MO, USA) was used to visualize F-actin. RNAi knockdown of AFAP1L1 was achieved using individual siGENOME oligonucleotides (S43843 and S43844, Dharmacon, Thermoscientific, Waltham, MA, USA).

Cell lysis, fractionation, immunoblotting and immunoprecipitation

Cells were generally lysed in raft buffer as previously described.16 For cytosolic/cytoskeletal fractionation studies, cells were lysed as previously described.54 The protein concentration of lysates was measured by the Bradford method (Bio-rad). For immunoprecipitation, 0.5–2 mg of total protein from clarified cell lysates were incubated with specific antibodies (0.5–5 μg) for 2 h at 4 °C, collected with protein G-Sepharose beads (Sigma-Aldrich) for 16 h (4 °C) before washing in lysis buffer, SDS–polyacrylamide gel electrophoresis and analysis by immunoblotting. After probing with specific primary antibodies, proteins were revealed using either secondary antibody coupled to horseradish peroxidase (GE Healthcare or Cell Signaling Technology) and detection by enhanced chemiluminescence (GE Healthcare), or with fluorescently labeled secondary antibodies and an Odyssey scanner (LI-COR Biosciences).

Microscopy and immunofluorescence

Glass coverslips (No. 1.5H, Marienfeld, Lauda-Konigshofen, Germany) were either uncoated or coated with either poly-L-lysine (Sigma-Aldrich) or serum before seeding cells. After specific experimentation treatments/time points, coverslips were gently washed with phosphate-buffered saline (PBS, 37 °C) then fixed in 4% paraformaldehyde/PBS for 15 min at 37 °C, followed by permeabilization with 0.5% Triton X-100 in PBS (5 min), then blocked for 2 h at 25 °C in 3% bovine serum albumin/PBS/Tween-20 (0.1%). Coverslips were incubated with primary antibody diluted in blocking buffer at 4 °C for 16 h, then washed three times in PBS/Tween-20 (15 min) before incubation with fluorophore-conjugated (Alexa-Fluor-labeled-488/594, Life Technologies) secondary antibodies (with Hoechst 33342 at 0.01 μg/ml) for 1 h at 25 °C, then washed again in PBS/Tween-20 before mounting in Vectashield (H-1000, Vector Labs, Burlingame, CA, USA). Fluorescent microscopy was performed using an × 60 plan-Apo 1.40 oil objective on an inverted Elcips-Ti microscope (Nikon, Tokyo, sJapan), fitted with a CoolSNAP HQ2 CCD camera (Photometrics, Tucson, AZ, USA). Images were captured, then z-stacks deconvolved and analyzed using NIS Elements 4.x (Nikon), or on a DeltaVision Elite system and analyzed with softWoRx suite2.0 (GE Healthcare).

Statistical analysis

For general statistical analysis, experiments were repeated as three biological replicates and analyzed by Student’s t-test or analysis of variance with two-tailed analysis, with orthogonal comparisons.

Abbreviations

SFK:

Src family kinase

References

  1. 1

    Chen WT, Chen JM, Parsons SJ, Parsons JT . Local degradation of fibronectin at sites of expression of the transforming gene product pp60src. Nature 1985; 316: 156–158.

    CAS  Article  Google Scholar 

  2. 2

    Murphy DA, Courtneidge SA . The ‘ins’ and ‘outs’ of podosomes and invadopodia: characteristics, formation and function. Nat Rev Mol Cell Biol 2011; 12: 413–426.

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  3. 3

    Ridley AJ . Life at the leading edge. Cell 2011; 145: 1012–1022.

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  4. 4

    Stylli SS, Kaye AH, Lock P . Invadopodia: at the cutting edge of tumour invasion. J Clin Neurosci 2008; 15: 725–737.

    CAS  Article  PubMed  Google Scholar 

  5. 5

    Weaver AM . Invadopodia: specialized cell structures for cancer invasion. Clin Exp Metastasis 2006; 23: 97–105.

    Article  PubMed  PubMed Central  Google Scholar 

  6. 6

    Poincloux R, Lizarraga F, Chavrier P . Matrix invasion by tumour cells: a focus on MT1-MMP trafficking to invadopodia. J Cell Sci 2009; 122: 3015–3024.

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  7. 7

    Schoumacher M, Goldman RD, Louvard D, Vignjevic DM . Actin, microtubules, and vimentin intermediate filaments cooperate for elongation of invadopodia. J Cell Biol 2010; 189: 541–556.

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  8. 8

    Furu M, Kajita Y, Nagayama S, Ishibe T, Shima Y, Nishijo K et al. Identification of AFAP1L1 as a prognostic marker for spindle cell sarcomas. Oncogene 2011; 30: 4015–4025.

    CAS  Article  PubMed  Google Scholar 

  9. 9

    Snyder BN, Cho Y, Qian Y, Coad JE, Flynn DC, Cunnick JM . AFAP1L1 is a novel adaptor protein of the AFAP family that interacts with cortactin and localizes to invadosomes. Eur J Cell Biol 2011; 90: 376–389.

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  10. 10

    Oser M, Yamaguchi H, Mader CC, Bravo-Cordero JJ, Arias M, Chen X et al. Cortactin regulates cofilin and N-WASp activities to control the stages of invadopodium assembly and maturation. J Cell Biol 2009; 186: 571–587.

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  11. 11

    Dorfleutner A, Cho Y, Vincent D, Cunnick J, Lin H, Weed SA et al. Phosphorylation of AFAP-110 affects podosome lifespan in A7r5 cells. J Cell Sci 2008; 121: 2394–2405.

    CAS  Article  PubMed  Google Scholar 

  12. 12

    Nakahara H, Otani T, Sasaki T, Miura Y, Takai Y, Kogo M . Involvement of Cdc42 and Rac small G proteins in invadopodia formation of RPMI7951 cells. Genes Cells 2003; 8: 1019–1027.

    CAS  Article  PubMed  Google Scholar 

  13. 13

    Lock P, Abram CL, Gibson T, Courtneidge SA . A new method for isolating tyrosine kinase substrates used to identify fish, an SH3 and PX domain-containing protein, and Src substrate. EMBO J 1998; 17: 4346–4357.

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  14. 14

    Seals DF, Azucena EF Jr, Pass I, Tesfay L, Gordon R, Woodrow M et al. The adaptor protein Tks5/Fish is required for podosome formation and function, and for the protease-driven invasion of cancer cells. Cancer Cell 2005; 7: 155–165.

    CAS  Article  Google Scholar 

  15. 15

    Gatesman A, Walker VG, Baisden JM, Weed SA, Flynn DC . Protein kinase Calpha activates c-Src and induces podosome formation via AFAP-110. Mol Cel Biol 2004; 24: 7578–7597.

    CAS  Article  Google Scholar 

  16. 16

    Ingley E, Schneider JR, Payne CJ, McCarthy DJ, Harder KW, Hibbs ML et al. Csk-binding protein mediates sequential enzymatic down-regulation and degradation of Lyn in erythropoietin-stimulated cells. J Biol Chem 2006; 281: 31920–31929.

    CAS  Article  PubMed  Google Scholar 

  17. 17

    Knaus UG, Bokoch GM . The p21Rac/Cdc42-activated kinases (PAKs). Int J Biochem Cell Biol 1998; 30: 857–862.

    CAS  Article  PubMed  Google Scholar 

  18. 18

    Sanders LC, Matsumura F, Bokoch GM, de Lanerolle P . Inhibition of myosin light chain kinase by p21-activated kinase. Science 1999; 283: 2083–2085.

    CAS  Article  PubMed  Google Scholar 

  19. 19

    Buday L, Wunderlich L, Tamas P . The Nck family of adapter proteins: Regulators of actin cytoskeleton. Cell Signal 2002; 14: 723–731.

    CAS  Article  PubMed  Google Scholar 

  20. 20

    Bustelo XR . Vav proteins, adaptors and cell signaling. Oncogene 2001; 20: 6372–6381.

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  21. 21

    Heo J, Thapar R, Campbell SL . Recognition and activation of Rho GTPases by Vav1 and Vav2 guanine nucleotide exchange factors. Biochemistry 2005; 44: 6573–6585.

    CAS  Article  PubMed  Google Scholar 

  22. 22

    Liu BP, Burridge K . Vav2 activates Rac1, Cdc42, and RhoA downstream from growth factor receptors but not beta1 integrins. Mol Cell Biol 2000; 20: 7160–7169.

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  23. 23

    Moshfegh Y, Bravo-Cordero JJ, Miskolci V, Condeelis J, Hodgson L . A Trio-Rac1-Pak1 signalling axis drives invadopodia disassembly. Nat Cell Biol 2014; 16: 574–586.

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  24. 24

    Stylli SS, Stacey TT, Verhagen AM, Xu SS, Pass I, Courtneidge SA et al. Nck adaptor proteins link Tks5 to invadopodia actin regulation and ECM degradation. J Cell Sci 2009; 122: 2727–2740.

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  25. 25

    Wells CM, Bhavsar PJ, Evans IR, Vigorito E, Turner M, Tybulewicz V et al. Vav1 and Vav2 play different roles in macrophage migration and cytoskeletal organization. Exp Cell Res 2005; 310: 303–310.

    CAS  Article  PubMed  Google Scholar 

  26. 26

    Takahashi R, Nagayama S, Furu M, Kajita Y, Jin Y, Kato T et al. AFAP1L1, a novel associating partner with vinculin, modulates cellular morphology and motility, and promotes the progression of colorectal cancers. Cancer Med 2014; 3: 759–774.

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  27. 27

    Crimaldi L, Courtneidge SA, Gimona M . Tks5 recruits AFAP-110, p190RhoGAP, and cortactin for podosome formation. Exp Cell Res 2009; 315: 2581–2592.

    CAS  Google Scholar 

  28. 28

    Nurnberg A, Kitzing T, Grosse R . Nucleating actin for invasion. Nat Rev Cancer 2011; 11: 177–187.

    Article  PubMed  PubMed Central  Google Scholar 

  29. 29

    Albiges-Rizo C, Destaing O, Fourcade B, Planus E, Block MR . Actin machinery and mechanosensitivity in invadopodia, podosomes and focal adhesions. J Cell Sci 2009; 122: 3037–3049.

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  30. 30

    Badowski C, Pawlak G, Grichine A, Chabadel A, Oddou C, Jurdic P et al. Paxillin phosphorylation controls invadopodia/podosomes spatiotemporal organization. Mol Biol Cell 2008; 19: 633–645.

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  31. 31

    Courtneidge SA, Azucena EF, Pass I, Seals DF, Tesfay L . The SRC substrate Tks5, podosomes (invadopodia), and cancer cell invasion. Cold Spring Harb Symp Quant Biol 2005; 70: 167–171.

    CAS  Article  Google Scholar 

  32. 32

    Li A, Dawson JC, Forero-Vargas M, Spence HJ, Yu X, Konig I et al. The actin-bundling protein fascin stabilizes actin in invadopodia and potentiates protrusive invasion. Curr Biol 2010; 20: 339–345.

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  33. 33

    Mader CC, Oser M, Magalhaes MA, Bravo-Cordero JJ, Condeelis J, Koleske AJ et al. An EGFR-Src-Arg-cortactin pathway mediates functional maturation of invadopodia and breast cancer cell invasion. Cancer Res 2011; 71: 1730–1741.

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  34. 34

    Lodyga M, Bai XH, Kapus A, Liu M . Adaptor protein XB130 is a Rac-controlled component of lamellipodia that regulates cell motility and invasion. J Cell Sci 2010; 123: 4156–4169.

    CAS  Article  PubMed  Google Scholar 

  35. 35

    Xu J, Bai XH, Lodyga M, Han B, Xiao H, Keshavjee S et al. XB130, a novel adaptor protein for signal transduction. J Biol Chem 2007; 282: 16401–16412.

    CAS  Article  PubMed  Google Scholar 

  36. 36

    Tehrani S, Tomasevic N, Weed S, Sakowicz R, Cooper JA . Src phosphorylation of cortactin enhances actin assembly. Proc Natl Acad Sci USA 2007; 104: 11933–11938.

    CAS  Article  PubMed  Google Scholar 

  37. 37

    Abe K, Rossman KL, Liu B, Ritola KD, Chiang D, Campbell SL et al. Vav2 is an activator of Cdc42, Rac1, and RhoA. J Biol Chem 2000; 275: 10141–10149.

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  38. 38

    Desmarais V, Yamaguchi H, Oser M, Soon L, Mouneimne G, Sarmiento C et al. N-WASP and cortactin are involved in invadopodium-dependent chemotaxis to EGF in breast tumor cells. Cell Motil Cytoskeleton 2009; 66: 303–316.

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  39. 39

    Rivera GM, Antoku S, Gelkop S, Shin NY, Hanks SK, Pawson T et al. Requirement of Nck adaptors for actin dynamics and cell migration stimulated by platelet-derived growth factor B. Proc Natl Acad Sci USA 2006; 103: 9536–9541.

    CAS  Article  PubMed  Google Scholar 

  40. 40

    Eden S, Rohatgi R, Podtelejnikov AV, Mann M, Kirschner MW . Mechanism of regulation of WAVE1-induced actin nucleation by Rac1 and Nck. Nature 2002; 418: 790–793.

    CAS  Article  PubMed  Google Scholar 

  41. 41

    Barrio-Real L, Kazanietz MG . Rho GEFs and cancer: linking gene expression and metastatic dissemination. Sci Signal 2012; 5: pe43.

    Article  PubMed  Google Scholar 

  42. 42

    Dang I, Gorelik R, Sousa-Blin C, Derivery E, Guerin C, Linkner J et al. Inhibitory signalling to the Arp2/3 complex steers cell migration. Nature 2013; 503: 281–284.

    CAS  Article  PubMed  Google Scholar 

  43. 43

    Delorme-Walker VD, Peterson JR, Chernoff J, Waterman CM, Danuser G, DerMardirossian C et al. Pak1 regulates focal adhesion strength, myosin IIA distribution, and actin dynamics to optimize cell migration. J Cell Biol 2011; 193: 1289–1303.

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  44. 44

    Huang CH, Tang M, Shi C, Iglesias PA, Devreotes PN . An excitable signal integrator couples to an idling cytoskeletal oscillator to drive cell migration. Nat Cell Biol 2013; 15: 1307–1316.

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  45. 45

    Frame MC . Newest findings on the oldest oncogene; how activated src does it. J Cell Sci 2004; 117: 989–998.

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  46. 46

    Timpson P, Jones GE, Frame MC, Brunton VG . Coordination of cell polarization and migration by the Rho family GTPases requires Src tyrosine kinase activity. Curr Biol 2001; 11: 1836–1846.

    CAS  Article  PubMed  Google Scholar 

  47. 47

    Yamaguchi H, Yoshida S, Muroi E, Yoshida N, Kawamura M, Kouchi Z et al. Phosphoinositide 3-kinase signaling pathway mediated by p110alpha regulates invadopodia formation. J Cell Biol 2011; 193: 1275–1288.

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  48. 48

    Kaipparettu BA, Kuiatse I, Tak-Yee Chan B, Benny Kaipparettu M, Lee AV, Oesterreich S . Novel egg white-based 3-D cell culture system. Biotechniques 2008; 45: 165–168 170-171.

    CAS  Article  PubMed  Google Scholar 

  49. 49

    Chauhan AK, Li YS, Deuel TF . Pleiotrophin transforms NIH 3T3 cells and induces tumors in nude mice. Proc Natl Acad Sci USA 1993; 90: 679–682.

    CAS  Article  PubMed  Google Scholar 

  50. 50

    Liang CC, Park AY, Guan JL . In vitro scratch assay: a convenient and inexpensive method for analysis of cell migration in vitro. Nat Protoc 2007; 2: 329–333.

    CAS  Article  Google Scholar 

  51. 51

    Ingley E, Williams JH, Walker CE, Tsai S, Colley S, Sayer MS et al. A novel ADP-ribosylation like factor (ARL-6), interacts with the protein-conducting channel SEC61beta subunit. FEBS Lett 1999; 459: 69–74.

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  52. 52

    Ingley E, Sarna MK, Beaumont JG, Tilbrook PA, Tsai S, Takemoto Y et al. HS1 interacts with Lyn and is critical for erythropoietin-induced differentiation of erythroid cells. J Biol Chem 2000; 275: 7887–7893.

    CAS  Article  PubMed  Google Scholar 

  53. 53

    Ingley E, Chappell D, Poon SY, Sarna MK, Beaumont JG, Williams JH et al. Thyroid hormone receptor-interacting protein 1 modulates cytokine and nuclear hormone signaling in erythroid cells. J Biol Chem 2001; 276: 43428–43434.

    CAS  Article  PubMed  Google Scholar 

  54. 54

    Samuels AL, Klinken SP, Ingley E . Liar, a novel Lyn-binding nuclear/cytoplasmic shuttling protein that influences erythropoietin-induced differentiation. Blood 2009; 113: 3845–3856.

    CAS  Article  PubMed  Google Scholar 

  55. 55

    Tilbrook PA, Ingley E, Williams JH, Hibbs ML, Klinken SP . Lyn tyrosine kinase is essential for erythropoietin-induced differentiation of J2E erythroid cells. EMBO J 1997; 16: 1610–1619.

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  56. 56

    Tsai S, Bartelmez S, Sitnicka E, Collins S . Lymphohematopoietic progenitors immortalized by a retroviral vector harboring a dominant-negative retinoic acid receptor can recapitulate lymphoid, myeloid, and erythroid development. Genes Dev 1994; 8: 2831–2841.

    CAS  Article  PubMed  Google Scholar 

  57. 57

    Vojtek AB, Hollenberg SM, Cooper JA . Mammalian Ras interacts directly with the serine/threonine kinase Raf. Cell 1993; 74: 205–214.

    CAS  Article  PubMed  Google Scholar 

Download references

Acknowledgements

We thank Janice Lam, Matt Lee, Rebecca Shapiro, Morgane Davies, Irma Larma and Kevin Li for technical assistance. This work was supported by grants from the National Health and Medical Research Council (513714 and 634352), the Medical Research Foundation of Royal Perth Hospital and the Cancer Council of Western Australia. EI received support from the Cancer Council of Western Australia, The Harry Perkins Institute of Medical Research, Sock-it-to-Sarcoma and the Hollywood Private Hospital Research Foundation.

Author contributions

DJMcC and SRT contributed equally to the manuscript and designed, supported and performed experiments, and analyzed data. TSK, AL, CL, JS and NK designed and performed experiments, and analyzed data. MP designed and analyzed experiments, and contributed to writing the manuscript. EI designed and supported the research, designed and undertook experiments, analyzed data and contributed to writing the manuscript.

Author information

Affiliations

Authors

Corresponding author

Correspondence to E Ingley.

Ethics declarations

Competing interests

The authors declare no conflict of interest.

Additional information

Supplementary Information accompanies this paper on the Oncogene website

Supplementary information

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Tie, S., McCarthy, D., Kendrick, T. et al. Regulation of sarcoma cell migration, invasion and invadopodia formation by AFAP1L1 through a phosphotyrosine-dependent pathway. Oncogene 35, 2098–2111 (2016). https://doi.org/10.1038/onc.2015.272

Download citation

Search

Quick links