Research ArticleCancer

Synaptojanin 2 is a druggable mediator of metastasis and the gene is overexpressed and amplified in breast cancer

See allHide authors and affiliations

Science Signaling  20 Jan 2015:
Vol. 8, Issue 360, pp. ra7
DOI: 10.1126/scisignal.2005537

Blocking Receptor Recycling to Prevent Metastasis

Blocking cancer cell metastasis can prolong patient survival. Ben-Chetrit et al. found that many patients with aggressive breast cancer have tumors with increased expression of SYNJ2, which encodes the lipid phosphatase synaptojanin 2. In cultured breast cancer cells, epidermal growth factor (EGF) triggered the localization of SYNJ2 to lamellipodia and invadopodia, which are cellular protrusions associated with invasive behavior. Knocking down SYNJ2 inhibited recycling of the EGF receptor to the cell surface and decreased the invasive behavior of cultured breast cancer cells. Expressing a phosphatase-deficient mutant of SYNJ2 in xenografted breast cancer cells suppressed tumor growth and lung metastasis in mice. A chemical screen identified SYNJ2 inhibitors that reduced cell invasion through a 3D matrix, suggesting that targeting SYNJ2 may prevent metastasis in breast cancer patients.


Amplified HER2, which encodes a member of the epidermal growth factor receptor (EGFR) family, is a target of effective therapies against breast cancer. In search for similarly targetable genomic aberrations, we identified copy number gains in SYNJ2, which encodes the 5′-inositol lipid phosphatase synaptojanin 2, as well as overexpression in a small fraction of human breast tumors. Copy gain and overexpression correlated with shorter patient survival and a low abundance of the tumor suppressor microRNA miR-31. SYNJ2 promoted cell migration and invasion in culture and lung metastasis of breast tumor xenografts in mice. Knocking down SYNJ2 impaired the endocytic recycling of EGFR and the formation of cellular lamellipodia and invadopodia. Screening compound libraries identified SYNJ2-specific inhibitors that prevented cell migration but did not affect the related neural protein SYNJ1, suggesting that SYNJ2 is a potentially druggable target to block cancer cell migration.


Despite progress in early detection and treatment, breast cancer remains a leading cause of cancer-related death in women. Like other carcinomas, tumors of the mammary gland carry somatic mutations, but only a fraction of these is causally implicated in oncogenesis (13). One frequent abnormality is copy number aberrations (4). For example, deletions of PTEN and INPP4B, which encode phosphoinositol (PI) lipid phosphatases, are detected in many breast tumors (5, 6). Conversely, amplification of HER2, which encodes a receptor tyrosine kinase (RTK) related to the epidermal growth factor receptor (EGFR), occurs in about 15% of breast cancers (7, 8). Antibodies and kinase inhibitors that inhibit HER2 are widely used to treat HER2-overexpressing breast cancers (9, 10). This exemplifies the therapeutic potential offered by the identification of oncogenic copy number aberrations.

Several reports recently identified copy number gains in genes that encode proteins involved in vesicular trafficking (11). For example, the gene that encodes cezanne-1 is amplified in a fraction of breast tumors (12). The encoded protein is a deubiquitination enzyme that enhances EGFR recycling. Likewise, chromosome 8p11–12 is frequently amplified in breast cancer (13). Encoded within this region is RAB-coupling protein (RCP), which cooperates with mutant p53 to coordinate the trafficking of integrins and RTKs (14). Another oncogenic copy number gain is found in the gene that encodes RAB25, a guanosine triphosphatase (GTPase) that controls vesicle recycling (15). Likewise, recurrent amplifications of the gene encoding RAB23 increase invasion by accelerating vesicular trafficking (16). These examples suggest that copy number aberrations might deregulate trafficking of RTKs and other receptors (17, 18).

Along with RAB family GTPases, PIs play pivotal roles in vesicular trafficking and cellular motility (19). For example, phosphorylation of the D3 position of phosphatidylinositol 4,5-bisphosphate [PI(4,5)P2] by phosphatidylinositol 3-kinase (PI3K) generates phosphatidylinositol 3,4,5-trisphosphate [PI(3,4,5)P3], which is necessary for the formation of both lamellipodia (20) and invadopodia (21). Moreover, growth factors enhance invadopodia formation (22, 23). Specifically, dephosphorylation of PI(3,4,5)P3 [generating PI(3,4)P2] causes the recruitment of an adaptor necessary for invasive growth, TKS5 [also called FISH (five SH3 domain–containing protein)], to the plasma membrane (24, 25). The present study was motivated by the identification of a copy number gain at chromosome 6q25, which affects a group of genes that encode endocytic proteins, including SYNJ2. Synaptojanin 2 (SYNJ2) is an effector of the Rho family GTPase Rac1 (26) and a homolog to SYNJ1, a 5-phosphatase that regulates vesicle recycling and availability at nerve terminals (27). Previous observations linked SYNJ2 to glioma cell invasion (26, 28). Here, we used clinical specimens, animal models, and in vitro assays to investigate whether aberrant expression of SYNJ2 is a potentially druggable driver of breast cancer.


Copy number gain or overexpression of SYNJ2 and diminished suppression by miR-31 correlate with shorter survival of breast cancer patients

Using the database of the Molecular Taxonomy of Breast Cancer International Consortium (METABRIC) (4), we observed gain of an about 1- to 2-megabase region centered at chromosome 6q24 that contains a cluster of genes, some of which have been implicated in vesicle trafficking (such as SNX9, TULP4, and SYNJ2; fig. S1A). The present study concentrates on SYNJ2 because its gain in 4% of breast cancer patients (76 of 1980) correlated with shorter survival (Fig. 1A). In line with promoting tumor aggressiveness, the expression of SYNJ2 correlated with shorter survival of estrogen receptor (ER)–positive patients (Fig. 1B). Not surprisingly, among ER-positive lesions, the luminal B subtype had the highest abundance (fig. S1B). In addition, tumors expressing high SYNJ2 had a predilection to metastasize to bones, pleura, and lungs (fig. S1C). These observations were confirmed at the protein level: a survey of 331 mammary specimens found stronger staining in malignant cells versus normal tissues (Fig. 1C). Likewise, SYNJ2 protein abundance correlated with overexpression of HER2, high tumor grades, and cell proliferation (Fig. 1D). In conclusion, copy number gain as well as increased mRNA and protein abundance of SYNJ2 correlated with poor prognosis and aggressive subtypes of breast cancer.

Fig. 1 Copy number gain and miR-31 affect SYNJ2 abundance and patient survival rates.

(A) Survival curves of 1980 breast cancer patients stratified according to SYNJ2 copy number. Death rates are indicated in parentheses. (B) Survival curves of ER-positive breast cancer patients stratified according to SYNJ2 abundance. (C) Representative images of SYNJ2 immunostaining (magnified, right) of the indicated tumor subtypes. Asterisk: vessel. (D) Immunohistochemical analysis of tissue microarrays from 331 invasive breast tumors, analyzed for various markers. (E) Schematic of the 3′UTR of SYNJ2 and the putative hsa-miR-31 binding sites. (F) Correlative analysis of hsa-miR-31 and SYNJ2 observed in breast cancer patients. (G) PCR and immunoblotting for mRNA and protein abundance of SYNJ2 in MDA-MB-231 and MCF10A cells transfected with mimic-miR-31 for 48 hours and normalized to control cells. Data are means ± SD from three experiments. (H) MCF-7 cells were cotransfected with the indicated mimic-miRNAs, control empty vector, and a reporter plasmid containing the 3′UTR of SYNJ2. Signals were first normalized to firefly luciferase reads and then to the average of two control miRNAs. Data are means ± SD of six biological repeats. (I) MCF-7 cells were cotransfected with wild-type (WT) or single (mut1 or mut2) or double (mut1+2) mutants of the 3′UTR reporter of SYNJ2, and treated as in (H). Data are means ± SD from three experiments.

Although we verified that, in general, copy number gain drove the higher SYNJ2 expression (fig. S1A), there were many cases in which overexpression was not driven by copy number gains. Hence, we considered regulation by microRNAs (miRNAs). Target prediction algorithms suggested that miR-31, an miRNA that inhibits metastasis by repressing the expression of RhoA and integrins (29), recognizes two sites within the 3′ untranslated region (3′UTR) of SYNJ2 (Fig. 1E) and could suppress SYNJ2 expression. Consistent with this, high abundance of miR-31 was associated with good prognosis in patients with ER-positive breast cancer (fig. S1D), the putative sites in the 3′UTR are highly conserved, and our analysis of a data set of mammary tumor miRNAs (30) indicated that SYNJ2 abundance was inversely related to the abundance in tumors of both miR-31 and miR-31*, the secondary molecule transcribed from the opposite arm of the precursor (Fig. 1F). As expected, transfection of mimic-miR-31 decreased the abundance of SYNJ2 at both the mRNA and protein levels (Fig. 1G). To exclude indirect effects, we cloned the 3′UTR of SYNJ2 into a luciferase reporter, and found that overexpression of miR-31 reduced the luciferase signals (Fig. 1H). In addition, mutating the putative binding sites within the 3′UTR of SYNJ2 reduced the inhibitory effect of mimic-miR-31 (Fig. 1I).

In conclusion, increased SYNJ2 abundance in breast tumors results from either copy number gain or decreased miR-31 abundance. To examine relevance to other tumor types, we compiled data from 1404 lung cancer patients and found that high SYNJ2 abundance correlated with shorter patient survival (fig. S1E). In a sample of brain tumors, we observed an inverse correlation between SYNJ2 and miR-31 abundance and identified SYNJ2high/miR-31low as a marker of poorer prognosis (fig. S1F). Thus, SYNJ2 might enhance progression of several types of tumors.

Growth factors increase SYNJ2 expression in association with increased cell invasion

To explore outcomes of increased SYNJ2 abundance, we examined MCF10A mammary cells (31) because they acquire an invasive phenotype after stimulation with different EGFR ligands, as shown by increased migration and invasion through Matrigel (fig. S2, A and B). Using polymerase chain reaction (PCR), DNA arrays, and immunoblotting, we noted that this involved increased abundance of SYNJ2 (Fig. 2, A and B). As expected, SYNJ2-overexpressing cells displayed enhanced migration (Fig. 2C and fig. S2C). Conversely, SYNJ2 knockdown reduced migration (Fig. 2D and fig. S2, D and E). Thus, in addition to copy gain and loss of miR-31, growth factors involved in breast cancer metastasis can increase the abundance of SYNJ2 (32, 33), and SYNJ2 seems to regulate both migration and invasion through barriers that represent those within tissues. Notably, in addition to increased protein abundance, growth factors likely stimulate the enzymatic function of SYNJ2 through activation of the kinase SRC (34).

Fig. 2 Transcriptional induction of SYNJ2 by growth factors and requirement for the catalytic activity for invasiveness.

(A and B) MCF10A cells were stimulated with EGF (20 ng/ml) or calf serum (5%) and assessed for mRNA (A) or protein (B) abundance of SYNJ2. Data are representative of three experiments. (C and D) Migration and invasion assays assessed at 18 hours in MCF10A cells either (C) overexpressing GFP-SYNJ2 (SYNJ2-OX) or LacZ (Ctrl) or (D) transfected with control or SYNJ2 siRNA, and either untreated (NT) or stimulated with EGF (10 ng/ml). Images are representative of five experiments. Data are means ± SD from three experiments. (E) Western blotting in shSYNJ2-expressing MDA-MB-231 cells reconstituted with either WT SYNJ2 (shSYNJ2+SYNJ2WT) or a catalytically deficient mutant (shSYNJ2+SYNJ2CD). (F) Invasion by MDA-MB-231 cells expressing shCtrl or shSYNJ2 through Matrigel over 4 days. Insets, magnified views of framed areas. The fraction of spheroids that disseminated cells into the matrix was quantified. Data are means ± SD from three experiments. Scale bar, 30 μm. (G) Matrigel invasion chamber assays using the indicated stable derivatives of MDA-MB-231 cells over 18 hours. Data are means ± SD from three experiments.

The phosphatase activity of SYNJ2 enhances tumor growth and metastasis in mice

Next, we surveyed publically available data sets of 56 human breast cancer cells to identify a cell line suitable for animal studies. This analysis found that MDA-MB-231, a basal B-like line, is the second and eighth best in terms of SYNJ2 and EGFR expression, respectively (fig. S3A). In addition, migration assays of representative cell lines of different disease subtypes confirmed that this line is highly migratory (fig. S3B). Moreover, depletion of EGFR using specific small interfering RNAs (siRNAs) substantially inhibited the relatively high migration and invasion of these cells (fig. S3C). This further motivated us to use MDA-MB-231 cells for our SYNJ2 functional assays. We first established clones that overexpressed SYNJ2 (fig. S4A) and validated that overexpression increased invasion (fig. S4B), whereas knockdown using siRNA markedly decreased invasion of MDA-MB-231 cells (fig. S4, C and D).

To stably deplete SYNJ2, we screened several different short hairpin RNA (shRNA) particles and selected one, number 37, for the establishment of a derivative of MDA-MB-231 cells (fig. S4E and Fig. 2E). When seeded in three-dimensional (3D) basement membranes, control cells formed clusters that disseminated elongated cells, but SYNJ2 depletion inhibited dissemination (Fig. 2F). This uncovered yet another function of SYNJ2 that might be relevant to tumor progression. Because SYNJ2 acts as both an enzyme and a scaffold, we addressed the requirement for the catalytic function. To this end, we reinfected shSYNJ2 cells with lentiviral particles encoding wild-type SYNJ2 or a catalytically deficient form (D388A/D726A, herein SYNJ2CD) containing mutations in the WXGDXN(F/Y)R motifs (35). Unlike wild-type SYNJ2, reexpression of the mutant failed to restore invasiveness (Fig. 2, E and G), indicating that the catalytic activity of SYNJ2 is essential for motility. Next, we implanted the cells into mammary fat pads of mice and assessed both tumor size (Fig. 3A) and metastases (Fig. 3, B and C). Primary tumors developed faster in mice implanted with control (shCtrl) or reconstituted (shSYNJ2+SYNJ2WT) cells compared with those implanted with knockdown (shSYNJ2) and “inactive rescue” (shSYNJ2+SYNJ2CD) cells. In addition, the shSYNJ2 and the “inactive rescue” groups displayed a statistically significant reduction in metastasis to lymph nodes. Moreover, the lungs of mice implanted with the shSYNJ2 cells or the “inactive rescue” cells showed fewer metastases (Fig. 3C). In conclusion, the phosphatase function of SYNJ2 contributes to tumor growth and metastasis in animals.

Fig. 3 The catalytic activity of SYNJ2 supports both metastatic spread and tumorigenic growth of human cancer xenografts in mice.

(A) Volume of tumors formed by the indicated derivatives of red fluorescent protein (RFP)–expressing MDA-MB-231 cells (2 × 106 per mouse) at 2 and 6 weeks after implantation into the fat pad of female severe combined immunodeficient (SCID) mice. Data are means ± SD from 10 to 11 mice per group; *P < 0.05, **P < 0.01, and ***P < 0.001. (B and C) Metastases that appeared 6 weeks after implantation in axillary and distant lymph nodes were quantified, and the lungs were photographed and quantified for small metastases. Data in (A) to (C) are representative of six experiments. (D to G) MDA-MB-231–RFP colonization of the lungs or intravasation into the blood 4 weeks after injection into the tail vein (1.5 × 105 cells per mouse) or mammary fat pad (2.5 × 106 cells per mouse), respectively, in 5-week-old female SCID mice. Cells expressed either control or SYNJ2 shRNA (D and E) or the control LacZ vector or a SYNJ2 expression vector (F and G). RFP-positive circulating tumor cells were scored per 1 × 106 readings obtained by cell sorting and normalized to tumor weight (E and G). Statistical parameters are indicated. Each dot represents an animal.

To circumvent effects on tumor volume, we separately examined intravasation and extravasation and normalized the results to primary tumor volumes. Cells transfected with either control shRNA or shSYNJ2 or expressing LacZ (control vector) or SYNJ2 were directly injected into the tail vein and scored for lung colonization (extravasation), or they were implanted into the fat pad and the blood was analyzed for circulating tumor cells (intravasation). SYNJ2 was necessary for both intravasation and extravasation, independently of tumor volume (Fig. 3, D to G). Notably, the two intravasation experiments, as well as the extravasation experiment using shSYNJ2, reached statistical significance, but the ability of SYNJ2-overexpressing cells to better extravasate did not, implying that the strong effects of SYNJ2 on metastasis were primarily due to enhanced intravasation. Notably, cells depleted of SYNJ2 formed smaller primary tumors (Fig. 3A), yet SYNJ2 overexpression weakly decreased xenograft growth (Fig. 3G). To try and resolve this, we used two in vitro cell proliferation assays. There was weak, if any, effect of depleting SYNJ2 on cell proliferation in culture (fig. S4, F and G). Nevertheless, when injected into the fat pad of female mice, SYNJ2-depleted cells formed statistically smaller tumors (fig. S4H). In conclusion, SYNJ2 not only accelerates metastasis in animal models but also positively influences tumor growth, although the latter was undetectable in the shorter-term in vitro experiments we performed.

SYNJ2 localizes to cellular protrusions involved in migration and matrix invasion

Using electron and fluorescence microscopy, we found that knocking down SYNJ2 transformed flat and adherent cells into weakly attached cells that displayed rudimentary lamellipodia and abnormal actin patches (fig. S5, A and B), in line with phenotypes of Saccharomyces cerevisiae after loss of synaptojanin-like proteins (36). Time-lapse microscopy confirmed abnormalities in lamellipodia and related the patches to large vesicles (movie S1), implying impaired vesicular trafficking. Assuming that these phenotypes would relate to the localization of SYNJ2, we obtained time-lapse images of green fluorescent protein (GFP)–tagged SYNJ2. These excluded an overlap between clathrin- and SYNJ2-containing puncta (Fig. 4A) and unveiled two locations (fig. S5C): small and dynamic assemblies at leading edges and larger, perinuclear assemblies. Notably, shortly after stimulation with a growth factor, peripheral SYNJ2 marked nascent lamellipodia. As discussed here and later, we infer that the peripheral SYNJ2 assemblies are recruited to nascent lamellipodia, whereas the perinuclear puncta represent invadopodia, actin-filled invasive protrusions (37). Accordingly, the assembly of peripheral puncta temporally overlapped focal formation of lamellipodia (movie S2), but unlike these highly dynamic puncta (fig. S6A, upper panel), the perinuclear clusters recruited actin (visualized using lifeACT-RUBY) and persisted for ~30 min (fig. S6A, lower panel, and movie S3). Moreover, by plating cells on fluorescent gelatin, we noted gradually increasing matrix degradation at the perinuclear SYNJ2 sites (Fig. 4B), consistent with a previous report in glioma cells (28). Together, these observations unveil dynamic translocations of SYNJ2 as well as implicate the phosphatase in regulating cell shape and actin protrusions.

Fig. 4 SYNJ2 localizes to lamellipodia and invadopodia and regulates endocytosis of EGFRs.

(A) MDA-MB-231 cells expressing GFP-SYNJ2 were transfected with an RFP-clathrin plasmid and plated on fibronectin and imaged every 5 s. Arrowheads mark recruitment of SYNJ2 to a newly formed leading edge. Scale bars, 5 μm. (B) Cells expressing GFP-SYNJ2 were plated on a fluorescently labeled gelatin. Photos at 10-s intervals were collected 5 hours later. Arrowheads mark colocalization of SYNJ2 and areas of degraded gelatin. Bottom, enlarged views of framed areas. Scale bar, 5 μm. (C) Cells were grown on fibronectin and stained for EGFR and F-actin. Scale bar, 20 μm. Insets, enlarged views of framed areas. (D) Immunoblotting (IB) of extracts from MDA-MB-231 cells transfected with the indicated siRNAs. (E) Cell sorting (left) and 125I-EGF binding (right) to surface EGFRs of the indicated derivatives of MDA-MB-231 cells. (F) Rose plots of migratory tracks of MDA-MB-231 cells after exposure to an EGF gradient. Red tracks mark migration toward greater EGF. (G) Immunoprecipitation (IP) for EGFR and then immunoblotting in lysates from MDA-MB-231 derivatives treated with EGF (10 ng/ml). Data either are representative or are means ± SD of three experiments.

SYNJ2 promotes recycling of EGFR at lamellipodia

The mostly bimodal compartmentalization of GFP-SYNJ2 was reinforced by the synchronous appearance and disappearance of fluorescence signals in experiments using both epifluorescence (red; relatively insensitive to changes in the z dimension) and total internal reflection microscopy [TIRF (total internal reflection fluorescence), green; limited to an approximately 200-nm depth]. Because puncta appeared yellow throughout their lifetime (fig. S6B), we concluded that SYNJ2 assembles and rapidly disassembles within the plane of the ventral membrane of lamellipodia. In line with this, an inhibitor of dynamin [the GTPase permitting invagination and facilitating migration (38)] inhibited the disassembly of SYNJ2’s peripheral puncta (fig. S6C and movie S4), indicating that recruitment to nascent lamellipodia depends on dynamin. Thus, a dynamin-dependent but clathrin-independent process mediates SYNJ2 trafficking to and from lamellipodia.

EGF-mediated migration entails clearance of PI(4,5)P2 from the leading edge, potentially by phospholipases and phosphatases like SYNJ2, and similarly, actin-regulated loading of PI(4,5)P2-depleted endosomes with EGFR molecules is coordinated by lamellipodin and endophilin, a SYNJ2 binder (39, 40). Accordingly, EGFR localized to lamellipodia in control cells, but accumulated in shSYNJ2 cells in abnormal intracellular vesicles surrounded by F-actin (Fig. 4C). This might be due to an inability to disassemble PI(4,5)P2-binding proteins from the vesicle’s coat or from actin comet tails (41). Consistent with intracellular trapping of EGFR, relatively high receptor abundance was detected in extracts of siSYNJ2-transfected cells (Fig. 4D), but two independent methods indicated reduced, rather than enhanced, surface abundance (Fig. 4E). Vesicular trapping bears functional consequences: shSYNJ2 cells severely lost the ability to migrate upward a gradient of EGF (Fig. 4F), which suggests impaired chemotaxis, in line with reports linking EGFR signaling and trafficking to the regulation of cofilin and cortactin in actin-filled protrusions (42, 43). Abnormal vesicular accumulation of EGFR could reflect impaired recycling or impaired sorting for degradation, a process regulated by ubiquitination (44). Indeed, SYNJ2 knockdown suppressed EGF-induced ubiquitination of EGFR (Fig. 4G). Furthermore, despite the fact that EGFR was tagged for degradation through phosphorylation of its Tyr1045 residue (fig. S6D), its degradation in shSYNJ2 cells was impaired (fig. S6E). To assess recycling, we monitored both EGFR and transferrin receptor. Although transferrin internalized normally, recycling was markedly decreased in shSYNJ2 cells and, contrariwise, markedly accelerated in SYNJ2-overexpressing cells (fig. S6F). Likewise, flow cytometry indicated defective recycling of internalized EGF in SYNJ2-depleted cells (fig. S6G). Thus, in line with ablation of SYNJ1, which leads to delayed vesicle reavailability (27), SYNJ2 seems essential both for EGFR recycling and for the tightly coupled process of lamellipodia formation (40).

The observed defective vesicular transport of EGFR in SYNJ2-depleted cells was complemented by results obtained from cells ectopically expressing SYNJ2 (fig. S7A). Using a radioactive form of EGF, we observed more rapid recycling of EGFR in SYNJ2-overexpressing cells (fig. S7B). As expected, increased recycling translated to receptor stabilization and more sustained AKT signaling (fig. S7C). Next, we asked if SYNJ2-mediated recycling applies to two other receptors, which are widely implicated in cell migration, namely, MET (hepatocyte growth factor receptor) and integrin β1. Similar to that of EGFR, the surface localization of MET in control cells was decreased, replaced by localization to inflated endosomes that were surrounded by actin patches (fig. S8A). Also in similarity to EGFR, mature MET, unlike the larger pro-MET precursor, was stabilized in shSYNJ2 cells, and both MET autophosphorylation and AKT transphosphorylation were strongly diminished (fig. S8B). The pattern assumed by integrin β1 molecules in shSYNJ2 cells was similarly characterized by large perinuclear aggregates that costained with both F-actin (fig. S8C) and phosphorylated EGFR (fig. S8D). Because the appearance of large intracellular aggregates of signaling and adhesion receptors was induced by depleting SYNJ2, we attempted rectifying it by introducing either the wild-type or a catalytically defective form of SYNJ2 (fig. S8E). To confirm that SYNJ2 reexpression was due to functional rescue, we examined EGFR accumulation and validated that wild-type but not mutant SYNJ2 decreased intracellular trapping of EGFR (fig. S8F). Thus, SYNJ2 controls recycling, as well as sorting of several surface receptors for degradation, in a way that might affect their involvement in cell migration.

SYNJ2 contributes to invadopodia formation

To resolve SYNJ2-mediated invasion, we examined matrix proteases. Zymography assays demonstrated defective secretion of matrix metalloproteinase 9 (MMP-9) when SYNJ2 was knocked down (fig. S9A). Conversely, overexpression of SYNJ2 increased both MMP9 mRNA and MMP-9 activity (fig. S9B). Consistent with our data on SYNJ2 in lamellipodia (Fig. 4B), matrix proteolysis corresponded to ventral actin- and SYNJ2-containing puncta (Fig. 5A). Notably, SYNJ2 overexpression increased whereas knockdown reduced the incidence of invadopodia (Fig. 5B). Correspondingly, we observed physical associations and colocalization of SYNJ2 and cortactin, a marker of invadopodia (fig. S9, C and D). Probing endogenous TKS5, a marker of invadopodia and a binder of PI(3,4)P2 (24, 25, 45), we confirmed its localization to sites of matrix degradation in control cells; however, siSYNJ2 cells displayed diffuse TKS5 and weaker matrix degradation (Fig. 5C). Furthermore, because invadopodial TKS5 anchors, via a PX domain, at ventral PI(3,4)P2, we used a cognate domain, the pleckstrin homology (PH) domain of Tapp1, a tandem PH domain–containing adapter. As expected, the PH domains colocalized with TKS5 in control but not siSYNJ2 cells (Fig. 5D). In conclusion, SYNJ2 appears necessary for a function preceding the recruitment of TKS5 to nascent invadopodia.

Fig. 5 SYNJ2-depleted cells display defective localization of PI(3,4)P2 and aberrant invadopodia.

(A) MDA-MB-231 cells stably expressing GFP-SYNJ2 were plated onto fluorescent gelatin-coated coverslips. Three hours later, cells were probed for GFP and F-actin, and invadopodial structures were detected (arrowheads). Scale bars, 10 μm. (B) Invadopodial structures in cells overexpressing SYNJ2 or transfected with control or SYNJ2 siRNA cultured on fluorescent gelatin-coated coverslips. Data are means ± SD from three experiments. (C) Invadopodial structures of MDA-MB-231 cells treated with the indicated siRNAs were detected by gelatin degradation, as well as by staining for F-actin and TKS5. Arrowheads (z-axis images) mark invadopodia. Scale bar, 10 μm. (D) Confocal microscopy of the codistribution of F-actin, TKS5, and PI(3,4)P2 (Tapp1) in the indicated cell derivatives transfected with a plasmid encoding a myc-tagged PH domain of Tapp1 and plated on fluorescent gelatin. Scale bar, 10 μm. (E) Colocalization of phalloidin and EGFR phosphorylated at Tyr1068 in MDA-MB-231 cells transfected with control or SYNJ2 siRNA plated on gelatin-coated coverslips. Scale bar, 10 μm. Images are representative of three experiments.

Because local activation of PI3K by RTKs is essential for invadopodia formation (21), and the generated PI(3,4,5)P3 molecules serve as substrates for SYNJ2, we expected that activated EGFR would localize to invadopodia. Localization of activated EGFR to sites of matrix proteolysis was indeed confirmed (Fig. 5E). Next, we tested a model proposing that focal processing of EGFR ligands by a complex comprising CD44 activates PI3K (46). Congruently, colocalization of CD44 in the cores of invadopodia was confirmed (fig. S9E), as previously reported (47), and we also found that surface CD44 was decreased in shSYNJ2 cells (fig. S9F). Recruitment of the matrix metalloproteinase MT1-MMP is yet another critical step in invadopodia maturation (48, 49). Accordingly, we detected MT1-MMP at invadopodia in control cells, but MT1-MMP accumulated in large, actin-decorated vesicles in siSYNJ2 cells (fig. S9G), which might correspond to MT1-MMP–positive late endosomes, as recently reported (48). Presumably, EGFR-mediated generation of PI(3,4,5)P2 and its dephosphorylation to PI(3,4)P2 by SYNJ2 instigates nascent invadopodia, which later mature to proteolytically active protrusions. In line with this model, EGF induced an increase in the number and size of invadopodia in MCF10A cells, but both parameters were significantly reduced when cells were pretreated with an EGFR-specific kinase inhibitor, gefitinib (fig. S10). In conclusion, growth factor–induced abundance and activation of SYNJ2 might contribute to the assembly and proteolytic activities of invadopodia.

Specific inhibitors of the 5′-phosphatase activity of SYNJ2 reduce cellular invasion

SYNJ2 and its brain-enriched kin, SYNJ1, belong to the dual-function class of inositol lipid phosphatases. This class has an N-terminal SAC-like domain, which encodes a polyphosphoinositide phosphatase activity, and a central 5′-phosphatase domain. The crystal structure of the 5′-phosphatase domain of yeast synaptojanin revealed that the enzyme adopts the fold of nucleases with two sheets forming an internal “sandwich” (50). These observations, as well as local amino acid sequence variations presented by SYNJ1 (51), suggested that small compounds might dock at the active site and inhibit SYNJ2 while sparing the critical action of SYNJ1 in synapses.

A homogeneous assay suitable for automation was established on the basis of the ability of SYNJ2 to dephosphorylate PI(3,4,5)P3 and produce PI(3,4)P2. To monitor this reaction, we used the PH domain of Tapp1 as a detector and a fluorescent PI(3,4)P2 as a probe. Polarization signals decreased when PI(3,4,5)P3 was incubated with a recombinant SYNJ2 (Fig. 6A). Several compound libraries (in total containing 53,540 molecules) were screened. To ensure selectivity toward SYNJ2, the inhibitory compounds were reassayed using a recombinant SYNJ1. This identified four selective inhibitors (Fig. 6, B and C). To test effects on cellular invasiveness, we applied a matrix invasion assay that clearly reflected SYNJ2 activity (Fig. 6D). As expected, all four compounds were found to inhibit invasion (at 10 μM; Fig. 6E). Future studies will test derivatives of these compounds in animal models, as a prelude for clinical development.

Fig. 6 Selection of compounds able to specifically inhibit the 5-phosphatase activity of SYNJ2 and attenuate cellular invasion.

(A) Fluorescence polarization signals to assess the 5-phosphatase activity of a purified SYNJ2 in vitro in the presence of the indicated reagents after 8 min of incubation at 33°C. Probe: fluorescent PI(3,4)P2; detector: a recombinant PH domain of Tapp1. (B and C) Chemical structures and median inhibitory concentration (IC50) values (B) and response curves (C) of selected SYNJ2-specific compounds tested against either purified SYNJ1 or SYNJ2. Data in (A) and (C) are means ± SD from four experiments. (D and E) 3D invasion assay of naïve MDA-MB-231 cells (E) or those expressing control or SYNJ2 shRNA (D) cultured for 72 hours in a basement membrane extract and then overlaid with an invasion matrix containing the indicated compounds listed in (B). Photos were taken 6 days later using ImageJ. Data are means and ranges from three experiments. Dimethyl sulfoxide (DMSO) was used as a solvent.


This study was motivated by the prediction that yet unknown gene copy number gains might contribute to aggressiveness of mammary tumors. Like in the case of the HER2-centered amplicon, such aberrations might identify patients suitable for treatment using molecular targeted drugs analogous to trastuzumab and lapatinib, the main HER2-blocking antibodies (52). In line with this prediction, SYNJ2 emerges from our study as a genetically aberrant and potentially druggable driver of tumor progression. Presumably, SYNJ2’s oncogenic activity relates to its ability to dephosphorylate critical phosphoinositides acting as signposts of both invadopodia and lamellipodia (19). Accordingly, SYNJ2 localizes to these actin-filled protrusions, and animal studies attributed essential roles in tumor growth and metastasis to the catalytic activity of SYNJ2. For example, SYNJ2-generated PI(3,4)P2 locally binds TKS5 and nucleates a cortactin-centered complex that enables cofilin to generate actin barbed ends within invadopodia (53). A similar mechanism might occur in the leading edge: locally generated PI(3,4)P2 likely binds lamellipodin and recruits Ena/VASP, an effector of the actin cytoskeleton (54). Our results highlight yet another key function of SYNJ2, namely, regulation of vesicular trafficking, in similarity to other lipid phosphatases (55). Although incompletely understood, we propose that the localization of SYNJ2 at the leading edge depends on dynamin and RAC1, although their distribution is distinct from that of caveolin-1 and clathrin. Hence, we assume that SYNJ2 controls variants of the clathrin-independent carriers, known to sustain membrane turnover at the leading edge (56).

Beyond a therapeutic scenario that selects patients for anti-SYNJ2 therapy on the basis of either SYNJ2 copy number or the SYNJ2/miR-31 ratio in the tumor, our findings suggest that carcinoma progression is propelled by successive processing of phosphoinositides by PI3K and SYNJ2 (25). Along with depleting PI(4,5)P2, which regulates endocytosis and the actin cytoskeleton, SYNJ2 dephosphorylates PI(3,4,5)P3, the product of PI3K, thereby generating PI(3,4)P2. Conceivably, two tumor suppressor phosphatases, PTEN [a 3-phosphatase that depletes both PI(3,4,5)P3 and PI(3,4)P2] and INPP4B [a 4-phosphatase that depletes PI(3,4)P2; (5)], normally balance the oncogenic alliance formed by PI3K and SYNJ2 (Fig. 7). Along with PI3K activating mutations, the triad of phosphatases is altered in cancer: deletions of PTEN and INPP4B frequently occur in tumors, and, as shown here, SYNJ2 copy gain and low miR-31 abundance are found in carcinomas and glioblastomas. Hence, compounds that inhibit SYNJ2, such as those we identified, might effectively block progression of tumors, especially those lacking the catalytic functions of PTEN and INPP4B. This prediction, along with the possibility that SYNJ2-generated inositol lipids can enhance tumor growth by directly or indirectly activating apoptosis-inhibitory kinases, like AKT and PDK, requires further investigation.

Fig. 7 A model depicting mechanistic aspects of the oncogenic activity of SYNJ2 within its two major locales, invadopodia and lamellipodia.

SYNJ2 dephosphorylates carbon 5 of the inositol ring. One of its product is PI(3,4)P2, but two phosphatases negate the action of SYNJ2: PTEN dephosphorylates the 3′ phosphate of PIs, such as PI(3,4,5)P3, and INPP4B dephosphorylates the 4′ phosphate of PI(3,4)P2 and other PIs. According to the model, active RTKs stimulate class 1 PI3Ks, which phosphorylate carbon 3 of the inositol ring, to generate several PIs, including the second messenger PI(3,4,5)P3. The latter can be dephosphorylated by SYNJ2 (or SHIP family members), thereby increasing the PI(3,4)P2 pool. Once locally formed, PI(3,4)P2 recruits TKS5, which anchors cortactin, nucleates actin polymerization, and instigates new invadopodia. In analogy, by dephosphorylating PI(3,4,5)P3 and enabling recruitment of the PI(3,4)P2-binding protein called lamellipodin, SYNJ2 might enable formation of lamellipodia (54). Both INPP4B and PTEN act as tumor suppressors, whereas PI3K is an established oncogene. We propose that the concerted action of PI3K and SYNJ2 is normally balanced by INPP4B and PTEN. In tumors, however, genetic aberrations affecting either PI3K or one of the three PI-specific phosphatases might support malignant transformation.


Reagents, antibodies, and compound libraries

Unless indicated, human recombinant growth factors and other materials were purchased from Sigma, and antibodies were from Cell Signaling Technology. Plates for wound healing assays were from ibidi. Glass-bottom dishes (35 mm) for time-lapse imaging were purchased from MaTek. An antibody against EGFR was purchased from Alexis. Antibodies to TKS5, Ras-GAP, AKT, and ERK (extracellular signal–regulated kinase) were from Santa Cruz Biotechnology. Fluorescein isothiocyanate–conjugated antibodies to CD44 were from BD Transduction Laboratories. Antibodies to phosphorylated EGFR (pTyr1068) and pAKT were from Cell Signaling Technology. Antibodies to EGFR (pTyr1068) for immunofluorescence and CD44 were from Epitomics. Antibodies against MMPs were from Millipore. A monoclonal antibody against SYNJ2 was from Abnova. Secondary antibodies were from Jackson ImmunoResearch Laboratories. siRNAs were from Dharmacon. Duo-set kits for growth factor assays were purchased from R&D Systems. Alexa Fluor 488 transferrin and goat anti-mouse Alexa Fluor 488, Alexa Fluor 555, and Alexa Fluor 647 secondary antibodies were from Invitrogen. Compound collections were purchased from Sigma (LOPAC 1280), Prestwick, Analyticon (MEGxp), and Chembridge (DIVERSet CL). Basement membrane extract was purchased from Trivigen. LifeACT-Ruby (from B. Shilo, Weizmann Institute) was used to visualize F-actin. The Dual Luciferase Reporter Assay System (including psiCHECK2 and a Renilla luciferase) from Promega was used for miRNA assays. Anti-GFP beads were purchased from Chromotek.

Cell lines, transfections, and RNA interference

MCF10A cells were grown in Dulbecco’s modified Eagle’s medium/F12 (1:1) supplemented with antibiotics, insulin (10 μg/ml), cholera toxin (0.1 μg/ml), hydrocortisone (0.5 μg/ml), heat-inactivated horse serum (5%, v/v), and EGF (10 ng/ml). Human mammary MDA-MB-231 cells were grown in RPMI-1640 (Gibco BRL) supplemented with 10% heat-inactivated fetal calf serum (Gibco), 1 mM sodium pyruvate, and a penicillin-streptomycin mixture (100 U/ml; 0.1 mg/ml; Beit Haemek Ltd.). The MDA-MB-231–RFP stable cell line was a gift from H. Degani (Weizmann Institute of Science, Israel). Plasmid transfections were performed using Lipofectamine 2000 according to the manufacturer’s guidelines (Roche). Alternatively, for transient mRNA knockdown experiments using siRNA oligonucleotides, cells were transfected with Oligofectamine (Invitrogen). The following siRNA sequences were used to deplete SYNJ2’s mRNA: GAAGAAACAUCCCUUUGAU and GGACAGCACUGCAGGUGUU. For control, we used siControl ON-TARGET plus (from Dharmacon). The following shRNA sequences (from Sigma) were used to deplete SYNJ2’s mRNA: (i) CCGGCCTACGATACAAGCGACAAATCTCGAAGATTTGTCGCTTGTATCGTAGGTTTTTG, (ii) CCGGCGAGAGGAGATCATTCGGAAACTCGAGTTTCCGAATGATCTCCTCTCGTTTTTG, and (iii) CCGGCCGGAAGAACAGTTTGAGCAACTCGAGTTGCTCAAACTGTTCTTCCGGTTTTTG.

Cell migration, invasion, and chemotaxis assays

Cells were plated in triplicates in the upper compartments of a Transwell tray (BD Biosciences) and allowed to migrate through the intervening membrane for 18 hours. Thereafter, cells were fixed in paraformaldehyde (3%), permeabilized in Triton X-100 (0.05%), and stained with methyl violet (0.02%). Cells growing on the upper side of the filter were removed, and migrating cells were photographed. Invasion assays were performed using BioCoat Matrigel chambers. For chemotaxis, we used chambers from ibidi. 3D spheroid cell invasion kits were from Trivigen. Briefly, cells (3000) were plated in basement membrane extract and cultured for 72 hours. Once spheroids were formed, invasion matrix was added to induce invasion, and images were taken after 6 days.

Gelatin zymography

To detect MMP activity, samples were separated electrophoretically on 10% polyacrylamide/0.1% gelatin–embedded gels. The gels were washed in 2.5% Triton X-100 and incubated at 37°C for 36 hours in 50 mM tris-HCl (pH 7.5) containing 0.2 M NaCl, 5 mM CaCl2, 1 μM ZnCl2, 0.02% Brij 35, and 1 mM p-aminophenylmercuric acetate.

Quantification of circulating tumor cells

Blood samples were purified on a Ficoll gradient. The resulting middle layer that contains mononuclear cells, along with the RFP-positive circulating tumor cells, was scored per 1 × 106 FACS (fluorescence-activated cell sorting) readings and normalized to tumor weight.

Metastasis tests in animals

Female CB-17 SCID mice (Harlan Laboratories; 15 per group) were implanted in the fat pad with MDA-MB-231 cells (1.4 × 106 cells per mouse). After 2 and 6 weeks, mice were anesthetized, tumor sizes were measured, and metastases in lymph nodes were visualized using a fluorescent binocular. For lung metastasis, lungs were removed and washed, and images were acquired using a fluorescent binocular.

Lentiviral vectors and virus production

Nontargeted shRNAs (control) and shRNAs directed against human SYNJ2 were produced in human embryonic kidney 293T cells following the manufacturer’s guidelines (Sigma). Target cells were infected with shRNA-encoding lentiviruses supplemented with polybrene (8 μg/ml) and cultured in the presence of puromycin (2 μg/ml) for 4 days. Stable gene-specific delivery of human SYNJ2 was performed using the ViraPower lentiviral expression system (Invitrogen) following the manufacturer’s guidelines.

Immunofluorescence and image processing

Cells were grown on fibronectin-coated coverslips for 48 hours. After treatments, cells were washed, permeabilized using 0.02% Triton X-100 and 3% paraformaldehyde, and fixed for 20 min. Confocal microscopy was performed using either a Zeiss LSM-710 microscope or a spinning disc microscope (numerical aperture, 1.45; Yokogawa CSU-22; Zeiss, fully automated inverted 200 M; Photometrics HQ-CCD camera) and solid-state lasers (473, 561, and 660 nm; exposure times: 0.25 to 1 s), under the command of SlideBook. 3D image stacks were acquired every 70 to 300 ms along the z axis by varying the position of the piezoelectrically controlled stage (step size: 0.1 to 0.4 μm). Alternatively, live cell fluorescence microscopy was carried out using the DeltaVision system (Applied Precision), and images were processed using Prism software.

Radiolabeling of EGF

Human recombinant EGF was labeled as follows: EGF (5 μg) was mixed in an Iodogen-coated tube (1 mg of reagent) with Na125I (1 mCi). After 15 min of incubation at 23°C, albumin was added to a final concentration of 0.1 mg/ml, and the mixture was separated on an Excellulose GF-5 column.

Determination of surface EGFR

Cells (2 × 104 per well) were seeded in triplicates in 24-well plates, with an additional well plated for control. Thereafter, cells were incubated with radiolabeled EGF for 1.5 hours at 4°C and rinsed with binding buffer. The control well was incubated with radiolabeled EGF and an excess of unlabeled EGF. Finally, cells were lysed in 1 M NaOH solution, and radioactivity was determined.

Immunoblotting analysis

Cells were washed briefly with ice-cold saline and scraped in a buffered detergent solution [25 mM Hepes (pH 7.5), 150 mM NaCl, 0.5% Na-deoxycholate, 1% NP-40, 0.1% SDS, 1 mM EDTA, 1 mM EGTA, 0.2 mM Na3VO4, and a protease inhibitor cocktail diluted at 1:1000]. For equal gel loading, protein concentrations were determined by using the bicinchoninic acid (Pierce) reagent. After gel electrophoresis, proteins were transferred onto a nitrocellulose membrane. The membrane was blocked in TBST buffer [0.02 M tris-HCl (pH 7.5), 0.15 M NaCl, and 0.05% Tween 20] containing 10% low-fat milk, blotted with a primary antibody for 60 min, washed with TBST, and incubated for 30 min with a secondary antibody conjugated to horseradish peroxidase.

Wound healing (scratch) assays

Cells were trypsinized and resuspended in EGF-deprived medium (7.0 × 105 cells/ml), and 70 μl was plated into specific wells (ibidi), resulting in confluent layers within 24 hours. Thereafter, culture inserts were removed by using sterile tweezers, and cells were allowed to migrate for 15 hours.

Electron microscopy

Cells were fixed in saline supplemented with 4% paraformaldehyde and 2% sucrose. Samples were washed and subjected to a second fixative (3% paraformaldehyde and 2.5% glutaraldehyde in 0.1 M cacodylate buffer supplemented with 1% sucrose and 5 mM CaCl2, pH 7.4). Cells were washed in 0.1 M cacodylate buffer and postfixed for 60 min with 1% osmium tetroxide in cacodylate buffer. For scanning electron microscopy, the postfixed samples were washed twice and treated with 1% tannic acid for 5 min followed by another wash and treatment with 1% uranyl acetate for 30 min. Samples were dehydrated in graded ethanol and made conductive by sputtering with a gold-palladium film. The samples were photographed using a scanning electron microscope (Leo Supra 55/Vp Zeiss).

Ligand recycling assays

MDA-MB-231 cells were preincubated for 30 min at 37°C with Alexa Fluor 488–transferrin (25 μg/ml in serum-free medium) or for 10 min with Alexa Fluor 488–EGF (40 ng/ml). Surface-bound ligands were detached by incubation for 30 min at 4°C in an acidic buffer (150 mM NaCl, 1 mM MgCl2, 0.125 mM CaCl2, 0.1 M glycine), before transfer to 37°C for the indicated time intervals, to allow for recycling of the internalized ligands. Cells were analyzed either by imaging or by FACS.

Immunohistochemical analyses of clinical specimens

The work presented is in accordance with the Portuguese National Regulatory Law of Tumor Bank Accession. Formalin-fixed, paraffin-embedded breast tumors were retrieved from the histopathology files of IPATIMUP and Hospital de Săo Joăo in Porto, Portugal. Analysis was performed using the Envision Detection System (DakoCytomation). Antigen retrieval was performed by using an EDTA solution (pH 9.0) at 98°C for 20 min. The SYNJ2 mouse monoclonal antibody was incubated overnight at 4°C. After immunostaining, slides were counterstained with Mayer’s hematoxylin. Two pathologists independently scored for staining intensity. Statistical analysis of the data was done using the SPSS suite.

Fluorescence polarization assays

Recombinant SYNJ2 and SYNJ1 were purchased from OriGene (TP315160 and TP315278, respectively). The PH domain of Tapp1 was produced in the Israel Structural Proteomics Centre. PI(3,4,5)P3, PI(3,4)P2, and PI(3,4)P2–tetramethylrhodamine (TMR) were purchased from Echelon Biosciences. All reagents were prepared in saline containing a lipid cocktail: SOPS (0.01 mg/ml), cholesterol (0.001 mg/ml), and C12E8 (0.005 mg/ml). The reaction mixture contained SYNJ2 or SYNJ1 (0.8 ng per reaction), MgCl2 (2 mM), dithiothreitol (5 mM), and PI(3,4,5)P3 (2 μM), and it was incubated at 33°C for 8 min. Reactions were terminated by the addition of a detector solution containing EDTA (2 mM), PI(3,4)P2-TMR, and the PH domain of Tapp1. Signals were determined using BMG PHERAstar FS, with a filter set of 540 and 560 nm for parallel and perpendicular emissions. Signals were then transformed to millipolarization (mP) units.

High-throughput screens

The fluorescence polarization assay was miniaturized to a total volume of 24 μl. The SYNJ2 solution was dispensed to black low-volume 384-well plates with a BioTek EL406 automated dispenser, and compounds from chemical libraries were transferred to the plates using a pin tool for approximate final concentrations of 15 μM. The reaction was initiated by the addition of 2 μM PI(3,4,5)P3 and 2 mM MgCl2 using a Bravo liquid handler (Agilent). Plates were incubated as above in an automated incubator (Liconic STX 220), and then the reaction was terminated by the addition of a detector solution with automated dispenser. Hit compounds were selected and retested in duplicate five-point dose-response assays to eliminate false-positives arising from machine errors. Active compounds were then obtained as dry powders, dissolved in DMSO, and tested in duplicate 10-point dose-response. For selectivity, the fluorescence polarization assay was performed using SYNJ1 instead of SYNJ2. Data representation and curve fitting were performed using the Genedata Screener software.

Statistical analyses

Two-sided Fisher’s exact test was used for analysis of lymph nodes. Tumor growth measurements used the Exact-sig (2 × 1–tailed) Mann-Whitney test. Other experiments were analyzed using one-way analysis of variance.


Fig. S1. SYNJ2 copy number gain and decreased miR-31 expression in human cancer.

Fig. S2. EGF family ligands enhance migration of mammary cells, but SYNJ2 depletion retards migration.

Fig. S3. SYNJ2 expression in and migratory behavior of various molecular subtypes of breast cancer cells.

Fig. S4. SYNJ2 abundance influences migration in vitro and cell proliferation in vivo.

Fig. S5. SYNJ2 knockdown influences cytoskeletal organization and cellular morphology.

Fig. S6. Subcellular localization and effects of SYNJ2 on receptor endocytosis.

Fig. S7. EGFR recycling and signaling to AKT are enhanced by SYNJ2.

Fig. S8. SYNJ2 depletion perturbs vesicular transport of the hepatocyte growth factor receptor MET and integrin subunit β1.

Fig. S9. SYJN2 is involved in matrix degradation and invadopodia formation.

Fig. S10. SYNJ2 facilitates EGFR-induced invadopodia assembly and maturation in MCF10A cells.

Movie S1. SYNJ2 knockdown impairs motility in MDA-MB-231 cells.

Movie S2. SYNJ2 colocalizes with sites of lamellipodia formation.

Movie S3. SYNJ2 localizes to sites of actin polymerization.

Movie S4. Dynamin recruits SYNJ2 and its pinching activity is necessary for disassembly of SYNJ2 puncta.


Acknowledgments: This work was performed at the Marvin Tanner Laboratory for Cancer Research. We thank L. Rameh, T. Takenawa, S. Lavi, M. Katz, I. Amit, A. Citri, Y. Peleg, S. Albeck, Y. Jacobovitch, A. Plotnikov, C. Wirth, and E. Muenstermann for their kind help. Funding: Our research is supported by the National Cancer Institute, the European Research Council, the Seventh Framework Program of the European Commission, the German-Israeli Project Cooperation (DIP), the Israel Cancer Research Fund, and the Dr. Miriam and Sheldon G. Adelson Medical Research Foundation. Y.Y. is the incumbent of the Harold and Zelda Goldenberg Professorial Chair. Author contributions: N.B.-C., D.C., M.E., and Y.Y. conceived the study. N.B.-C., D.C., C.K., M.M., A.A.-H., T.I., S.C., H.C.-D., W.J.K., K.S., M. Lauriola, M.K., M. Lindzen, Z.S., H.B., D.S., D.A.F., F.P., and F.M. performed experiments. R.R., R.B.-H., S.E., F.S., and C.C. analyzed patient data. H.G.-H., T.L., R.A., S.W., M.E., and Y.Y. supervised experiments. M.S. provided reagents. N.B.-C., M.E., and Y.Y. wrote the manuscript. Competing interests: The authors declare that they have no competing interests. Data and materials availability: The results of our screen of compounds are available at PubChem (ID: 22074).

Stay Connected to Science Signaling

Navigate This Article