Contact inhibition of locomotion (CIL) is the process by which cells stop the continual migration in the same direction after collision with another cell. Highly invasive malignant cells exhibit diminished CIL when they contact stromal cells, which allows invasion of the tissue by tumors. We show that Nm23-H1 is essential for the suppression of Rac1 through inactivation of Tiam1 at the sites of cell–cell contact, which plays a pivotal role in CIL. U87MG cells show CIL when they contact normal glia. In spheroid confrontation assays U87MG cells showed only limited invasion of the glial population, but reduction of Nm23-H1 expression in U87MG cells abrogated CIL resulting in invasion. In U87MG cells, Nm23-H1 is translocated to the sites of contact with glia through association with α-catenin and N-cadherin. Mutants of Nm23-H1, which lacked the binding ability with Tiam1, or α-catenin did not restore CIL. Moreover, the expression of ephrin-B1 in tumor cells disrupted CIL and promoted invasion. As one mechanism, ephrin-B1 inhibits the association of Nm23-H1 with Tiam1, which contributes for activation of Rac1. These results indicate a novel function of Nm23-H1 to control CIL, and its negative regulation by ephrin-B1.

To gain a better understanding of metastasis it is important to understand the mechanism of invasion of normal stroma tissues by malignant cells. One mechanism is activation of metalloproteinases, which degrade surrounding extracellular substrates. Another is contact inhibition of locomotion (CIL) is the process by which cells disrupt or change their direction of migration upon contact with another cell (Mayor and Carmona-Fontaine, 2010). Malignant cells with high invasion potential exhibit diminished CIL behavior, which allows them to invade stromal tissues. Although the concept of CIL was first reported in the 1950s, the molecular mechanism of CIL is not well understood (Abercrombie and Heaysman, 1953). In our attempts to monitor invasion of various glioblastoma cells into glia populations in cell culture, we observed that U87MG glioblastoma cells did not intermingle with normal glias, and a clear boundary was maintained between them. However, some other glioblastoma cell lines do invade glias. Therefore, we particularly focused on the invasion mechanisms of these glioblastoma cells. Regulation of Rac1 GTPase is essential for construction of membrane protrusions in the direction of migration. Among the regulators of Rac1 activity, Tiam1 is a guanine nucleotide exchange factor (GEF), which was originally identified in T cells as an invasion- and metastasis-inducing gene (Habets et al., 1994). Tiam1 is also expressed in various cancer cells, and we previously observed that the GEF activity of Tiam1 is regulated by physical interaction with Nm23-H1 and ephrin-B1 (Tanaka et al., 2004; Otsuki et al., 2001).

Nonmetastatic protein 23 (Nm23) is a nucleotide diphosphate (NDP) kinase that is expressed in a wide range of organisms from bacteria to mammals (Lacombe et al., 2000). In humans, there are eight members of the Nm23 family (Nm23-H1-8, respectively), of which Nm23-H1 and -H2 are most abundantly expressed (Boissan et al., 2009). Nm23-H1 was the first protein to be identified as a metastasis suppressor, and its decreased expression is associated with higher tumor grade and increased metastasis (Steeg et al., 2003). In contrast, exogenous expression of Nm23-H1 results in decreased migration, anchorage-independent growth and metastasis in various cancer cells (Suzuki et al., 2004; Jung et al., 2006; McDermott et al., 2008). The metastasis suppressive effect of Nm23-H1 was confirmed using Nm23 knockout mice, in which the rate of formation of hepatocellular carcinoma was unchanged; however, the knockout mice developed twofold more pulmonary metastases (Boissan et al., 2005). In addition to their nucleotide diphosphate kinase activity, members of this family have both histidine protein kinase and 3′ to 5′ exonuclease (DNase) activities (Wagner et al., 1997; Fan et al., 2003). The protein histidine kinase activity of Nm23-H1 mediates the antimotility function in vitro, possibly by serine phosphorylation of the kinase suppressor of ras (KSR) and suppression of ras signaling (Hartsough et al., 2002). Recently, 3′ to 5′ exonuclease activity was shown to be important for the Nm23-H1 metastasis suppressor function (Zhang et al., 2011). As one of the mechanisms by which Nm23-H1 suppresses cell migration, we previously observed that Nm23-H1 inhibits Rac1 activity through physical interaction with Tiam1 (Otsuki et al., 2001).

The members of the Eph receptor family can be classified into two groups based on their sequence similarity and their preferential binding to ligands tethered to the cell surface either by a glycosylphosphatidyl inositol-anchor (ephrin-A) or a transmembrane domain (ephrin-B) (Poliakov et al., 2004; Pasquale, 2008; Daar, 2012). Interaction of the Eph family of receptor protein tyrosine kinases and its ligand, ephrin family members, induces bi-directional signaling via cell–cell contacts. Overexpression of B-type ephrin in cancer cells correlates with high invasion and high vascularity of some tumors (Meyer et al., 2005; Castellvi et al., 2006), and elevated expression of ephrin-B1 is observed in poorly differentiated invasive tumor cells and other tumors with poor clinical prognosis (Kataoka et al., 2002; Varelias et al., 2002). Although investigation of the functions of Eph receptors and ephrins has focused on the development of the vascular and nervous systems, the roles of Eph–ephrin pathways in epithelial cells and cancers have also attracted interest (Batlle et al., 2005; Holmberg et al., 2006; Castaño et al., 2008; Vaught et al., 2008). We previously observed that signaling through ephrin-B1 is involved in the promotion of carcinomatous peritonitis of gastric cancer and pancreas cancer (Tanaka et al., 2007a; Tanaka et al., 2007b; Tanaka et al., 2010).

In this report, we show that Nm23-H1 is required for CIL as it suppresses Tiam1 activity at the sites of intercellular contacts between U87MG and glias. The C-terminus of Nm23-H1, which is required for association with α-catenin, was necessary for U87MG cells to maintain CIL in the presence of glia. In addition, ephrin-B1 inhibits the association of Nm23-H1 with Tiam1, which could also be involved in tumor cell invasion into glias by affecting CIL. These results suggest that the regulation of intracellular localization of Nm23-H1 and its interaction with Tiam1 affects the invasion of stromal tissue by some types of cancer cell.

Heterotypic CIL between U87MG and glia

In our attempt to study the mechanism of tumor cells invasion of mesenchymal tissues, spheroid confrontation assays, also called boundary assays were performed, where a co-culture is made using two different cell types with only a small gap separating the two cell fronts. As the cells divide and migrate, the two cellular fronts get closer to each other and finally collide. This allows the behavior of the two cellular populations to be analyzed at the boundary. Among several glioblastoma cell lines examined, U87MG showed a clear border between their leading edge and those of glias, therefore they did not intermingle with each other (Fig. 1A). Invasion by tumor cells was evaluated as an ‘invasion index’, which indicates the percentage of the area in glial cell population that has been invaded (supplementary material Fig. S1A). When spheroids of tumor cells and glias were confronted on agarose-coated plates, U87MG did not invade the glial sphere (Fig. 1B).

Fig. 1.

Heterotypic CIL between U87MG and glia. (A,B) Spheroid confrontation assay was performed using rat glia with U87MG on fibronectin-coated coverslips (A), or on agarose-coated 96-well plates (B) as described in Materials and Methods. The left panel in A is the bright-field image showing the boundary of glia and U87MG cells. Glia cells and tumor cells were labeled with DiO (green) and DiI (red), respectively. Scale bars: 20 µm. In B, computer-reconstructed vertical sections in the X-Z or Y-Z plane were combined. I is the invasion index (%), as described in Materials and Methods. (C,D) Representative time-lapse microscopy images of U87MG–glia collisions at the indicated times. (C) Single U87MG (marked with yellow and red asterisks) and glia (green asterisk) cells were cocultured (see supplementary material Movie 1). The arrows indicate the direction of movement of the U87MG cells marked with the same colored asterisks. Scale bars: 50 µm. (D) Spheroid confrontation assay was performed as in A (see supplementary material Movie 2). U87MG cells are colored red. The yellow arrow indicates the direction of movement of a U87MG cell. The dotted line indicates the periphery of the glia. Scale bars: 50 µm.

Fig. 1.

Heterotypic CIL between U87MG and glia. (A,B) Spheroid confrontation assay was performed using rat glia with U87MG on fibronectin-coated coverslips (A), or on agarose-coated 96-well plates (B) as described in Materials and Methods. The left panel in A is the bright-field image showing the boundary of glia and U87MG cells. Glia cells and tumor cells were labeled with DiO (green) and DiI (red), respectively. Scale bars: 20 µm. In B, computer-reconstructed vertical sections in the X-Z or Y-Z plane were combined. I is the invasion index (%), as described in Materials and Methods. (C,D) Representative time-lapse microscopy images of U87MG–glia collisions at the indicated times. (C) Single U87MG (marked with yellow and red asterisks) and glia (green asterisk) cells were cocultured (see supplementary material Movie 1). The arrows indicate the direction of movement of the U87MG cells marked with the same colored asterisks. Scale bars: 50 µm. (D) Spheroid confrontation assay was performed as in A (see supplementary material Movie 2). U87MG cells are colored red. The yellow arrow indicates the direction of movement of a U87MG cell. The dotted line indicates the periphery of the glia. Scale bars: 50 µm.

To further investigate the behavior of U87MG cells when they contact glia, we performed time-lapse analysis in cocultures. When a free-moving U87MG cell came into contact with a glia cell, its protrusion collapsed, the direction of its migration changed and it moved away from glia (Fig. 1C; supplementary material Movie 1). We also performed time-lapse analysis in the spheroid confrontation assay where the cellular outgrowth of U87MG and glia met. U87MG cells promptly retracted their protrusions or lamellipodia after contact with glia, and changed the direction of migration (Fig. 1D; supplementary material Movie 2). By tracking the migration of tumor cells, it was apparent that cell migration increased after collision (supplementary material Fig. S2A,B). These observations suggest that contact inhibition of locomotion (CIL) occurs in U87MG cells after they collide with normal glias, which could account for the limited invasion of glias by U87MG cells in the spheroid confrontation assay.

N-cadherin regulates CIL and invasion, and Nm23-H1 is accumulated in the N-cadherin complex at cell–cell contacts

In a recent analysis of developing embryos of Xenopus laevis, deprivation of N-cadherin abolished CIL in migrating neural crest cells (Theveneau et al., 2010). N-cadherin is a major cell adhesion molecule in glioblastomas and glia, and N-cadherin is located at the sites of contact between U87MG and glia cells (Fig. 2A). Therefore, we examined whether N-cadherin was necessary for heterotypic CIL between U87MG and glia. Reduction of N-cadherin expression in U87MG cells and glias by RNA interference (RNAi) allowed tumor cells to invade glias (Fig. 2B,C). In order to show different planes of Z-stacks, a pitched view of each X-Y plane was constructed, which also showed that intermingling of the two cell populations is induced by reduction of N-cadherin (supplementary material Fig. S3). From the time-lapse analysis, N-cadherin silencing in U87MG cells inhibited retraction of the protrusion after collision with glia (supplementary material Movie 3). These results suggest that N-cadherin is necessary for establishment of CIL and attenuates invasion by U87MG cells.

Fig. 2.

N-cadherin mediates CIL between U87MG and glia. (A) Glia expressing EGFP and U87MG cells were cocultured. Cells were immunostained with anti-N-cadherin antibody (red). Scale bar: 10 µm. Asterisk indicates the position of U87MG cell. (B) Expression of N-cadherin was reduced in U87MG and glia cells using RNAi of human or rat N-cadherin, respectively. Band intensities were quantified and relative expression level was calculated. (C) A spheroid 3D confrontation assay was performed with either control or N-cadherin-reduced U87MG and glia cells on agarose-coated 96-well plates. N-cadherin RNAi#1 indicates confrontation between U87MG expressing miRNA#1 and glia expressing siRNA#1. Glia cells and tumor cells were labeled with DiO (green) and DiI (red), respectively. Numbers below the panels are the invasion index, which is shown graphically at the bottom. Scale bar: 20 µm.

Fig. 2.

N-cadherin mediates CIL between U87MG and glia. (A) Glia expressing EGFP and U87MG cells were cocultured. Cells were immunostained with anti-N-cadherin antibody (red). Scale bar: 10 µm. Asterisk indicates the position of U87MG cell. (B) Expression of N-cadherin was reduced in U87MG and glia cells using RNAi of human or rat N-cadherin, respectively. Band intensities were quantified and relative expression level was calculated. (C) A spheroid 3D confrontation assay was performed with either control or N-cadherin-reduced U87MG and glia cells on agarose-coated 96-well plates. N-cadherin RNAi#1 indicates confrontation between U87MG expressing miRNA#1 and glia expressing siRNA#1. Glia cells and tumor cells were labeled with DiO (green) and DiI (red), respectively. Numbers below the panels are the invasion index, which is shown graphically at the bottom. Scale bar: 20 µm.

Next, to elucidate the mechanism of N-cadherin-mediated CIL and suppression of invasion, we focused on Nm23-H1. A recent study showed that Nm23-H1 makes a complex with N-cadherin through association with α-catenin (Aktary et al., 2010). In addition, we previously observed that Nm23-H1 physically interacts with Tiam1 and prohibits the GEF activity of Tiam1. As Rac1 GTPase is important for the formation of cell membrane protrusions necessary for the forward cell movement, its regulation by Nm23-H1 at N-cadherin-mediated cell–cell contacts may be important for establishment of CIL. Because Nm23-H1 and Tiam1 were expressed in U87MG and glia cells, we investigated the role of Nm23-H1 in the heterotypic CIL (supplementary material Fig. S4A). When localization of Nm23-H1 was examined in U87MG cells expressing EGFP-tagged Nm23-H1 under coculture with glias, it was found to accumulate at the leading edge of the cell membrane contacting the glia, and also at homotypic cell contacts between tumor cells (supplementary material Fig. S4B, left). In contrast, such accumulation of Nm23-H1 at the sites of cell–cell contact was scarcely observed in N-cadherin-deprived U87MG (supplementary material Fig. S4B, right). Furthermore, EGFP-tagged Nm23-H1 accumulated at the cell membrane of U87MG cells in contact with microbeads coated with the extracellular domain of N-cadherin, but not control beads coated with mouse IgGFc (supplementary material Fig. S4D). These results indicate that Nm23-H1 is translocated to the site of cell–cell contacts, which largely depends on an N-cadherin-mediated adhesion complex.

The region of Nm23-H1 responsible for binding to α-catenin was next examined using deletion mutants of Nm23-H1. The mammalian Nm23 family of proteins are 152-168 amino acids (aa) long and highly conserved. Diverse regions in the N-terminus and C-terminus are called the V1 and V2 region, respectively (Iwashita et al., 2004). Based on this information, deletion mutants were constructed (see Materials and Methods). Immunoprecipitation analysis identified that Nm23-H1-C (lacking aa 37-152), but not Nm23-H1-N (lacking aa 1-138) binds to α-catenin, which indicates that the C-terminal region of Nm23-H1 is required to interact with α-catenin (Fig. 3A,B). In contrast, the C-terminal region of Nm23-H1 was not involved in the physical association with Tiam1, because Nm23-H1-N, which lacks the C-terminal region, still binds to Tiam1 as well as wild-type Nm23-H1 (Fig. 3C). Moreover, the expression of Nm23-H1-N inhibited Tiam1-induced Rac1 activation in whole cell lysate examined by GTP–Rac1 pull down analysis (Fig. 3D). Therefore, these results indicate that Nm23-H1-N cannot bind to α-catenin, but can interact with Tiam1 and inhibits its GEF activity toward Rac1.

Fig. 3.

The C-terminus of Nm23-H1 physically associates with α-catenin. (A) A schematic diagram of wild-type and deletion mutants of Nm23-H1 used in this study. ‘RNAi’ indicates the position of miRNA used in Fig. 4. ‘121–141’ indicates the position of the sequence depicted in Fig. 5B. (B,C) Physical association of Nm23-H1 with α-catenin and Tiam1. Cos1 cells were transiently transfected with N-terminally myc-tagged wild-type (wt) or truncated Nm23-H1 together with N-terminally FLAG-tagged α-catenin (B) or Tiam1 (C), as indicated. (B) Nm23-H1 was immunoprecipitated (IP) with anti-myc antibody, and co-precipitated α-catenin was detected by anti-FLAG antibody. The asterisk indicates immunoglobulin H chain. (C) Tiam1 was immunoprecipitated by anti-Tiam1 antibody and co-precipitated Nm23-H1 was detected by anti-myc antibody. It should be noted that six copies of the myc epitope tagged to the Nm23-H1 cDNA increases the protein size by around 9 kDa. Arrowheads indicate co-precipitated wild-type or mutant of Nm23-H1. (D) Cos1 cells were transfected with Tiam1, with or without deletion mutants of Nm23-H1, as indicated. Cells were lysed for affinity precipitation with immobilized GST–PBD. Precipitated endogenous GTP-bound Rac1 was detected by immunoblotting. Band intensities were quantified, and normalized intensities were calculated relative to control cells (lane 1). Data are means ± s.d. (n = 3); *P<0.01 versus control (lane 1).

Fig. 3.

The C-terminus of Nm23-H1 physically associates with α-catenin. (A) A schematic diagram of wild-type and deletion mutants of Nm23-H1 used in this study. ‘RNAi’ indicates the position of miRNA used in Fig. 4. ‘121–141’ indicates the position of the sequence depicted in Fig. 5B. (B,C) Physical association of Nm23-H1 with α-catenin and Tiam1. Cos1 cells were transiently transfected with N-terminally myc-tagged wild-type (wt) or truncated Nm23-H1 together with N-terminally FLAG-tagged α-catenin (B) or Tiam1 (C), as indicated. (B) Nm23-H1 was immunoprecipitated (IP) with anti-myc antibody, and co-precipitated α-catenin was detected by anti-FLAG antibody. The asterisk indicates immunoglobulin H chain. (C) Tiam1 was immunoprecipitated by anti-Tiam1 antibody and co-precipitated Nm23-H1 was detected by anti-myc antibody. It should be noted that six copies of the myc epitope tagged to the Nm23-H1 cDNA increases the protein size by around 9 kDa. Arrowheads indicate co-precipitated wild-type or mutant of Nm23-H1. (D) Cos1 cells were transfected with Tiam1, with or without deletion mutants of Nm23-H1, as indicated. Cells were lysed for affinity precipitation with immobilized GST–PBD. Precipitated endogenous GTP-bound Rac1 was detected by immunoblotting. Band intensities were quantified, and normalized intensities were calculated relative to control cells (lane 1). Data are means ± s.d. (n = 3); *P<0.01 versus control (lane 1).

Association of Nm23-H1 with α-catenin is required for CIL and suppression of tumor invasion

We next examined whether recruitment of Nm23-H1 to the N-cadherin-mediated complex at cell–cell contacts is important for CIL between glioblastoma cells and glias. First, U87MG cells stably expressing Nm23-H1 micro RNA (miRNA) were established, and then, reconstituted with wild-type Nm23-H1 or Nm23-H1-N by transfection of the miR-resistant rat ortholog of Nm23-H1 (Fig. 4, top). The invasion of glia by U87MG cells was examined by 2D or 3D sphere confrontation assays. U87MG cells in which Nm23-H1 was reduced (Nm23H1 miR) invaded the glia region (Fig. 4). The clear border region between tumor cells and glia was restored after addition of wild-type Nm23-H1, but not Nm23-H1-N to Nm23H1 miR cells (Fig. 4; supplementary material Fig. 1B,C: Res Nm23-wt and -N, respectively). The intracellular localization of EGFP-tagged Nm23-H1-N was further examined in U87MG cells. In contrast to the accumulation of wild-type Nm23-H1 in the cell–cell contacts between U87MG cells and glia, EGFP–Nm23-H1-N remained diffusely distributed in the cytoplasm, and did not accumulate at intercellular contacts (supplementary material Fig. S4C). These results suggest that association with α-catenin is required for Nm23-H1 to accumulate to the site of cell–cell contact and suppress cell invasion.

Fig. 4.

Association of Nm23-H1 with α-catenin is required for CIL. Top: U87MG cells were stably transfected with control miR (cont) or Nm23-H1 miR (Nm23H1miR). ResNm23H1 wt and ResNm23H1-N are Nm23-H1 rescue clones which reconstituted in Nm23H1miR cells with untagged wild-type Nm23-H1 or Nm23-H1-N, respectively. Expression of Nm23-H1 in each cell line was examined by immunoblotting with anti-Nm23-H1 antibody. Bottom: Spheroid confrontation assay of DiO-labeled glias (green) and various U87MG cell lines labeled with DiI (red) was performed in 2D culture on fibronectin-coated glass or in 3D culture on agarose-coated plates, as described in Fig. 2. Scale bars: 20 µm. Numbers below the panels are the invasion index.

Fig. 4.

Association of Nm23-H1 with α-catenin is required for CIL. Top: U87MG cells were stably transfected with control miR (cont) or Nm23-H1 miR (Nm23H1miR). ResNm23H1 wt and ResNm23H1-N are Nm23-H1 rescue clones which reconstituted in Nm23H1miR cells with untagged wild-type Nm23-H1 or Nm23-H1-N, respectively. Expression of Nm23-H1 in each cell line was examined by immunoblotting with anti-Nm23-H1 antibody. Bottom: Spheroid confrontation assay of DiO-labeled glias (green) and various U87MG cell lines labeled with DiI (red) was performed in 2D culture on fibronectin-coated glass or in 3D culture on agarose-coated plates, as described in Fig. 2. Scale bars: 20 µm. Numbers below the panels are the invasion index.

To further evaluate the effects of Nm23-H1 on invasion by tumor cells, time-lapse analysis was performed. After contact with glia, U87MG Nm23-H1 miR cells continued to move forward, and did not move away from the glia (supplementary material Movie 4). In spheroid confrontation assays, migrating U87MG Nm23-H1 miR cells invaded the glial sphere (supplementary material Movie 5). The CIL-like movement of U87MG cells was restored by reconstitution of wild-type Nm23-H1, but not Nm23-H1-N, suggesting that association of Nm23-H1 with α-catenin is necessary to maintain CIL (supplementary material Movies 6, 7 and 8, respectively).

Nm23-H1 suppresses tumor invasion through association with Tiam1

Next, we examined whether interaction of Nm23-H1 with Tiam1 is important for regulation of tumor invasion. Recently, the unique acidic sequence motif necessary for the association of the PHCCEx region of Tiam1 and Tiam2 with other proteins was identified in several molecules including ephrin-B1 and CD44, whose interaction was also reported in our previous study (Terawaki et al., 2010; Tanaka et al., 2004). When the sequence of Nm23-H1 was compared with the Tiam1-binding motif of ephrin-B1295-313 and CD44713-730, Nm23-H1 aa 121-139 were found to be similar (Fig. 5B). The critical residues in the acidic motif ExxE/DxxxxL are conserved in Nm23-H1121-139, and the acidic residues glutamate and aspartic acid are frequently present within this region. Association of wild-type Nm23-H1 with PHCCEx of Tiam1 was detected by the immunoprecipitation analysis (Fig. 5C). Therefore, we prepared Nm23-H1 with mutations in two glutamates (E124,127K), whose corresponding residues in ephrin-B1 and CD44 are essential for binding with Tiam1. From immunoprecipitation analysis, the association of Nm23-H1E124, 127K (hereafter referred to as Nm23-H1-2EK) with Tiam1 was significantly decreased compared with wild-type Nm23-H1 (Fig. 5D). Binding of Nm23-H1 with Tiam1 is required for Nm23-H1 to suppress invasion by U87MG cells, because the reconstitution of Nm23-H1-2EK in U87MG Nm23-H1 miR cells did not inhibit invasion of glias by tumor cells, and failed to suppress Rac1 activation (Fig. 5E, middle; 5F, Res Nm23H1-2EK). In addition, reduction of Tiam1 expression suppressed Rac1 activation and invasion of Nm23-H1-deprived U87MG cells (Fig. 5E, right, 5F, Nm23H1 miR+Tiam1 miR). These results suggest that Nm23-H1 prevents invasion by U87MG cells through Tiam1 suppression of Rac1.

Fig. 5.

Association of Nm23-H1 with Tiam1 is required for CIL. (A) The domains in Tiam1. The PHCCEx region contains an N-terminal Pleckstrin homology (PHn) domain, a putative coiled-coil (CC) region and a conserved extra (Ex) region. PDZ, PSD-95/DlgA/ZO-1 domain; DH, Dbl homology domain; Phc, C-terminal PH domain. (B) Sequence alignment of the PHCCEx-binding region of human ephrin-B1 and CD44 with human or rat Nm23-H1121–141. Crucial residues for binding to PHCCEx are marked by asterisks. Acidic residues are in red; xxxxxxx residues are in blue. (C) Cos1 cells were transiently transfected with Tiam1 N-terminally FLAG-tagged 9in the PHCCEx region, and with or without C-terminally HA-tagged Nm23-H1. Nm23-H1 was immunoprecipitated with anti-HA, and PHCCEx was detected by anti-FLAG antibody. The asterisk indicates the immunoglobulin L chain. (D) Cos1 cells were transfected with HA-tagged wild-type Nm23-H1 (wt.) or Nm23-H1E124, 127K (2EK) together with or without Tiam1. Tiam1 was immunoprecipitated, and Nm23-H1 was detected by anti-HA antibody. (E) Spheroid 3D confrontation assay of DiO-labeled glia and DiI-labeled mutant Nm23-H1E124, 127K (2EK) U87MG cells or double knock down Nm23-H1 and Tiam1 (Nm23-H1 miR+Tiam1 miR) U87MG cells was performed. Scale bars: 20 µm. (F) Rac1 activation (GTP–Rac1) was assayed in various U87MG cell lines as indicated. Band intensities were quantified and normalized intensities were calculated relative to control cells (lane 1). Data are means ± s.d. (n = 3). *P<0.01 versus control.

Fig. 5.

Association of Nm23-H1 with Tiam1 is required for CIL. (A) The domains in Tiam1. The PHCCEx region contains an N-terminal Pleckstrin homology (PHn) domain, a putative coiled-coil (CC) region and a conserved extra (Ex) region. PDZ, PSD-95/DlgA/ZO-1 domain; DH, Dbl homology domain; Phc, C-terminal PH domain. (B) Sequence alignment of the PHCCEx-binding region of human ephrin-B1 and CD44 with human or rat Nm23-H1121–141. Crucial residues for binding to PHCCEx are marked by asterisks. Acidic residues are in red; xxxxxxx residues are in blue. (C) Cos1 cells were transiently transfected with Tiam1 N-terminally FLAG-tagged 9in the PHCCEx region, and with or without C-terminally HA-tagged Nm23-H1. Nm23-H1 was immunoprecipitated with anti-HA, and PHCCEx was detected by anti-FLAG antibody. The asterisk indicates the immunoglobulin L chain. (D) Cos1 cells were transfected with HA-tagged wild-type Nm23-H1 (wt.) or Nm23-H1E124, 127K (2EK) together with or without Tiam1. Tiam1 was immunoprecipitated, and Nm23-H1 was detected by anti-HA antibody. (E) Spheroid 3D confrontation assay of DiO-labeled glia and DiI-labeled mutant Nm23-H1E124, 127K (2EK) U87MG cells or double knock down Nm23-H1 and Tiam1 (Nm23-H1 miR+Tiam1 miR) U87MG cells was performed. Scale bars: 20 µm. (F) Rac1 activation (GTP–Rac1) was assayed in various U87MG cell lines as indicated. Band intensities were quantified and normalized intensities were calculated relative to control cells (lane 1). Data are means ± s.d. (n = 3). *P<0.01 versus control.

Ephrin-B1 attenuates CIL by blocking physical association of Nm23-H1 with Tiam1

We examined whether expression of ephrin-B1 affects CIL-like behavior, and increases the invasion by tumor cells. Because U87MG cells expressed very low level ephrin-B, ephrin-B1 was stably overexpressed in this cell line (U87MG ephrin-B1; Fig. 6D). In a the spheroid confrontation assay, U87MG cells expressing ephrin-B1 substantially invaded the area of glia cells (Fig. 6A; supplementary material Fig. S1B). Similar results were obtained in an assay in which spheroids of tumor cells and glias were three-dimensionally confronted on agarose-coated plates. Many U87MG ephrin-B1 cells invaded the sphere of glia cells (Fig. 6B; supplementary material Fig. S1C). Consistent with these results, time-lapse analysis revealed that U87MG ephrin-B1 cells did not show repulsive CIL in both single cell coculture and spheroid confrontation (supplementary material Movies 9 and 10, respectively). In addition, U87MG cells expressing the ephrin-B1 mutant, which lacked the cytoplasmic domain (Δcyt ephrin-B1), did not invade glias (Fig. 6C, left). Therefore, the cytoplasmic region of ephrin-B1 is required for cell invasion.

Fig. 6.

Heterotypic CIL between U87MG and glia was abolished by expression of ephrin-B1. (A,B) Spheroid confrontation assay was performed using glia (green) with U87MG ephrin-B1 cells (red) on fibronectin-coated coverslips (A), or on agarose-coated 96-well plates (B) as in Fig. 1. Scale bar: 20 µm. (C) Spheroid 3D confrontation assay of glia (green) and U87MG (red) cells stably expressing cytoplasmic-domain-deleted ephrin-B1 (Δcyt EFNB1), ephrin-B1E297K, D300K or ephrin-B1 together with Tiam1 miRNA (Ephrin-B1 Tiam1 miR) was performed. Expression of wild-type (wt.) and Δcyt ephrin-B1 was shown by western blotting using an antibody to the extracellular domain of ephrin-B1. Scale bar: 20 µm. (D) Expression level of ephrin-B proteins in various glioblastoma cell lines was examined by immunoblotting with anti-pan ephrin-B antibody as described in Materials and Methods. (E) Rac1 activation (GTP–Rac1) was assessed in various U87MG cell lines as indicated. Band intensities were quantified, and normalized intensities were calculated relative to control cells (lane 1). Data are means ± s.d. (n = 3); *P<0.01 versus parent U87MG (lane 1).

Fig. 6.

Heterotypic CIL between U87MG and glia was abolished by expression of ephrin-B1. (A,B) Spheroid confrontation assay was performed using glia (green) with U87MG ephrin-B1 cells (red) on fibronectin-coated coverslips (A), or on agarose-coated 96-well plates (B) as in Fig. 1. Scale bar: 20 µm. (C) Spheroid 3D confrontation assay of glia (green) and U87MG (red) cells stably expressing cytoplasmic-domain-deleted ephrin-B1 (Δcyt EFNB1), ephrin-B1E297K, D300K or ephrin-B1 together with Tiam1 miRNA (Ephrin-B1 Tiam1 miR) was performed. Expression of wild-type (wt.) and Δcyt ephrin-B1 was shown by western blotting using an antibody to the extracellular domain of ephrin-B1. Scale bar: 20 µm. (D) Expression level of ephrin-B proteins in various glioblastoma cell lines was examined by immunoblotting with anti-pan ephrin-B antibody as described in Materials and Methods. (E) Rac1 activation (GTP–Rac1) was assessed in various U87MG cell lines as indicated. Band intensities were quantified, and normalized intensities were calculated relative to control cells (lane 1). Data are means ± s.d. (n = 3); *P<0.01 versus parent U87MG (lane 1).

Rac1 activation of U87MG cells was increased by overexpression of either Tiam1 or ephrin-B1 (Fig. 6E). Expression of the ephrin-B1 mutant, ephrin-B1 E297K, D300K, that lacks the capacity to bind to Tiam1, did not affect Rac1 activation or invasion in spheroid confrontation assays (Fig. 6E,C, middle). CIL was not affected by expression of ephrin-B1 E297K, D300K, either as determined by time-lapse analysis (supplementary material Movie 11). These results suggest that association with Tiam1 is necessary for ephrin-B1-mediated tumor invasion and disruption of CIL. By contrast, knockdown of Tiam1 blocked Rac1 activation and invasion induced by ephrin-B1, suggesting that Rac1 activation following ephrin-B1 expression is mediated through Tiam1 (Fig. 6C right, E, EFNB1 Tiam1 miR).

We next examined whether association of ephrin-B1 with Tiam1 attenuates the Tiam1–Nm23-H1 interaction. Association of Nm23-H1 with Tiam1 was inhibited by the expression of ephrin-B1 (Fig. 7A). The association of endogenous Nm23-H1 with Tiam1 was also detected in U87MG cells, which was much reduced by overexpression of ephrin-B1 (Fig. 7B). Moreover, treatment of cells with the peptide corresponding to the binding motif of ephrin-B1 with Tiam1 (ephrin-B1 295-313) also blocked interaction of Tiam1 with Nm23-H1, and induced Rac1 activation and invasion by tumor cells (supplementary material Fig. S5A). Intracellular delivery of the peptides was confirmed by immunostaining of the cells with anti-TAT antibody (data not shown). These results suggest that ephrin-B1 and Nm23-H1 competitively associate with Tiam1, and overexpression of ephrin-B1 sequesters Tiam1 from Nm23-H1.

Fig. 7.

Effects of ephrin-B1 and Nm23-H1 on CIL are through regulation of Tiam1. (A) Cos1 cells were co-transfected with C-terminally HA-tagged Nm23-H1 and Tiam1 with or without ephrin-B1. Cells were lysed for IP with anti-Tiam1 and then subjected to immunoblotting (IB) with anti-HA to detect Nm23-H1. Band intensities (marked by arrowhead) were quantified, and the relative intensity ratio of precipitated Nm23-H1 was calculated. Data are means ± s.d. (n = 3); *P<0.01. (B) Lysate prepared from wild-type or ephrin-B1-overexpressing U87MG cells was subjected for immunoprecipitation with anti-Tiam1 antibody or control rabbit Iggy. Co-precipitated Nm23-H1 was detected by immunoblotting with anti-Nm23H1 antibody. (C) C6 glioblastoma cells invade glia cells. Top: spheroid confrontation assay on fibronectin-coated coverslips revealed significant intermingling of C6 and glia cells. Bottom: invasion by C6 cells was inhibited by reduction of ephrin-B1. Scale bars: 10 µm. *P<0.01 versus control. (D) Tiam1 was immunoprecipitated from lysate prepared from wild-type or ephrin-B1-reduced C6 cells. Co-precipitated Nm23-H1 was detected by immunoblotting with anti-Nm23H1 antibody. Data are means ± s.d. (n = 3); *P<0.01. (E) Rac1 activation was assessed in wild-type or ephrin-B1-reduced C6 cells. Band intensities were quantified, and normalized intensities were calculated relative to control cells (lane 1). Data are means ± s.d. (n = 3); *P<0.01.

Fig. 7.

Effects of ephrin-B1 and Nm23-H1 on CIL are through regulation of Tiam1. (A) Cos1 cells were co-transfected with C-terminally HA-tagged Nm23-H1 and Tiam1 with or without ephrin-B1. Cells were lysed for IP with anti-Tiam1 and then subjected to immunoblotting (IB) with anti-HA to detect Nm23-H1. Band intensities (marked by arrowhead) were quantified, and the relative intensity ratio of precipitated Nm23-H1 was calculated. Data are means ± s.d. (n = 3); *P<0.01. (B) Lysate prepared from wild-type or ephrin-B1-overexpressing U87MG cells was subjected for immunoprecipitation with anti-Tiam1 antibody or control rabbit Iggy. Co-precipitated Nm23-H1 was detected by immunoblotting with anti-Nm23H1 antibody. (C) C6 glioblastoma cells invade glia cells. Top: spheroid confrontation assay on fibronectin-coated coverslips revealed significant intermingling of C6 and glia cells. Bottom: invasion by C6 cells was inhibited by reduction of ephrin-B1. Scale bars: 10 µm. *P<0.01 versus control. (D) Tiam1 was immunoprecipitated from lysate prepared from wild-type or ephrin-B1-reduced C6 cells. Co-precipitated Nm23-H1 was detected by immunoblotting with anti-Nm23H1 antibody. Data are means ± s.d. (n = 3); *P<0.01. (E) Rac1 activation was assessed in wild-type or ephrin-B1-reduced C6 cells. Band intensities were quantified, and normalized intensities were calculated relative to control cells (lane 1). Data are means ± s.d. (n = 3); *P<0.01.

We further examined the stromal invasion by glioblastoma cells endogenously expressing ephrin-B1 at a high level, i.e. the C6 cell line, and revealed that C6 and glia cells diffusely intermingled (Fig. 7C, upper). When ephrin-B1 of C6 was reduced by miRNA, C6–glia intermingling was inhibited (Fig. 7C, bottom). In time-lapse analyses most of the wild-type C6 cells remained attach to glia after they made contact, whereas C6 ephrin-B1 miR cells changed direction and moved away from glia (supplementary material Movies 12, 13). In addition, silencing of ephrin-B1 in C6 cells increased the association of Nm23-H1 with Tiam1, and suppressed Rac1 activation (Fig. 7D,E). These results suggest that ephrin-B1 on the cell surface plays pivotal roles in tumor cell invasion by mechanisms including suppression of CIL.

Ephrin-B1 increases peripheral invasion by glioblastoma cells in vivo

To further examine whether ephrin-B1 affects tumor invasion in vivo, tumors of ephrin-B1-overexpressing U87MG cells or ephrin-B1-reduced C6 cells in mice brains were compared with those of normal U87MG and C6 cells. The tumors of control U87MG cells expanded in the brain with a relatively clear margin, whereas the margin of the tumors of ephrin-B1-overexpressing U87MG cells were irregularly shaped (Fig. 8A; Table 1). Irregularity of the tumor periphery was also confirmed by tracing and measuring the tumor margins (Materials and Methods; supplementary material Fig. S2C). In addition, control C6 cells invaded cerebral tissue with an irregular tumor periphery, whereas C6 ephrin-B1miR cells was less invasive with a relatively clear tumor margin (Fig. 8B). Expression of ephrin-B1 in these transplanted tumors was also confirmed by immunohistochemistry (Fig. 8A,B, bottom panels). Although we did not observe significant difference in tumor size between the wild-type and the ephrin-B1-overexpressing (U87MG) or knockdown (C6) cells, survival rate was decreased by expression of ephrin-B1 (Table 1). These results indicate the expression of ephrin-B1 promotes the peripheral invasion by some glioblastoma cells in vivo.

Fig. 8.

Ephrin-B1 increases peripheral invasion by glioblastoma cells in brain. (A,B) Control U87MG mock-transfected cells or U87MG ephrin-B1-overexpressing cells (A), or control C6 or C6 ephrin-B1miR cells (B) were injected intracranial into nude mice (2×105/mouse). Five mice were analyzed for each group and this was repeated twice. Representative histology of cerebra 12 days after injection of tumor cells is shown: Hematoxylin and Eosin staining (upper); immunohistochemistry with goat anti-ephrin-B1 antibody (bottom). T indicates the area occupied by tumor cells. Tumors in the same location were analyzed for the irregularity of the tumor margin, and the index (Ir) is shown below the images and in the bar graphs on the right. *P<0.01 versus control. Scale bars: 500 µm. Two clones of U87MG ephrin-B1-overexpressing and C6 ephrin-B1miR cells were examined and similar results were obtained.

Fig. 8.

Ephrin-B1 increases peripheral invasion by glioblastoma cells in brain. (A,B) Control U87MG mock-transfected cells or U87MG ephrin-B1-overexpressing cells (A), or control C6 or C6 ephrin-B1miR cells (B) were injected intracranial into nude mice (2×105/mouse). Five mice were analyzed for each group and this was repeated twice. Representative histology of cerebra 12 days after injection of tumor cells is shown: Hematoxylin and Eosin staining (upper); immunohistochemistry with goat anti-ephrin-B1 antibody (bottom). T indicates the area occupied by tumor cells. Tumors in the same location were analyzed for the irregularity of the tumor margin, and the index (Ir) is shown below the images and in the bar graphs on the right. *P<0.01 versus control. Scale bars: 500 µm. Two clones of U87MG ephrin-B1-overexpressing and C6 ephrin-B1miR cells were examined and similar results were obtained.

Table 1.

Incidence of mice bearing tumors with irregularly shaped margins

Tumor cell type No. with irregular margins/no. bearing tumorsa,b No. surviving/no. transplanteda No. surviving/no. transplantedc 
U87MG control 0/8 9/10 6/10 
U87MG ephrin-B1 8/8 8/10 2/10 
C6 control 8/8 8/10 1/10 
C6 ephrin-B1miR 3/10 10/10 6/10 
Tumor cell type No. with irregular margins/no. bearing tumorsa,b No. surviving/no. transplanteda No. surviving/no. transplantedc 
U87MG control 0/8 9/10 6/10 
U87MG ephrin-B1 8/8 8/10 2/10 
C6 control 8/8 8/10 1/10 
C6 ephrin-B1miR 3/10 10/10 6/10 
a

12 days after inoculation with tumor cells.

b

Number of mice containing tumors with irregularly shaped margins (Ir>3.0)/number of mice bearing tumors.

c

Number of mice surviving 21 days after inoculation with tumor cells.

Discussion

In this study, we focused on the invasion by brain tumor cells after they confront normal glias. In order to examine tumor cells invasion of normal cells, spheroid confrontation assays are often performed (Golembieski et al., 1999; Rosenzweig et al., 2006). Among the cell lines derived from glioblastoma, U87MG cells did not enter into the sphere of glias in a confrontation assay. The mild invasion of glias or normal fetal brain aggregates by U87MG cells was also observed by others using a similar method (Golembieski et al., 1999; Rosenzweig et al., 2006). Time-lapse imaging showed that the protrusion of U87MG cells was retracted after contact with glia, and new lamellipodia or protrusions were produced in the opposite direction, and then, the U87MG cells moved away from the glia. This behavior matches that of CIL, a process that stops the continual locomotion of a cell in the same direction after collision with another cell. Therefore, we used U87MG cells as a model of CIL of tumor cells in the presence of stromal cells. The parameters of CIL were analyzed in various time-lapse images of single cell cocultures, and plotted in supplementary material Fig. S2B. The frequency of CIL was also estimated by counting tumor cells, which moved away from glia during the hour following collision (Table 2). From these results, the degree of tumor cells invasion in spheroid confrontation assays was well correlated with disturbance of CIL.

Table 2.

The incidence of CIL

Tumor cell type Incidence of CIL (%)a nb 
U87MG   
Wild type 70.0 30 
N-cadherin miR (No. 1) 13.3 30 
Nm23-H1miR 12.5 32 
Res Nm23-H1 wt 70.0 30 
Res Nm23-H1-N 13.3 30 
Ephrin-B1 10.0 30 
Ephrin-B1E297K, D300K 66.7 30 
C6   
Wild type 6.7 30 
Ephrin-B1 miR 46.7 30 
Tumor cell type Incidence of CIL (%)a nb 
U87MG   
Wild type 70.0 30 
N-cadherin miR (No. 1) 13.3 30 
Nm23-H1miR 12.5 32 
Res Nm23-H1 wt 70.0 30 
Res Nm23-H1-N 13.3 30 
Ephrin-B1 10.0 30 
Ephrin-B1E297K, D300K 66.7 30 
C6   
Wild type 6.7 30 
Ephrin-B1 miR 46.7 30 
a

Percentage of tumor cells that moved back (θ<90°) and away from glia by 1 hr after the collision.

b

Number of cells analyzed.

We showed that Nm23-H1 plays a pivotal role in heterotypic CIL between parent U87MG cells and glia, which requires the interaction of Nm23-H1 with Tiam1 at the sites of intercellular contacts. As one of the mechanisms of recruitment of Nm23H1 to cell–cell contacts, it associates with α-catenin and N-cadherin, and inhibits local activation of Tiam1, which might suppress the formation of membrane protrusions toward glia (supplementary material Fig. S7A). Ephrin-B1 disrupts CIL through activation of Tiam1, which at least partly depends on the mechanism that attenuates Nm23-H1–Tiam1 interaction (supplementary material Fig. S7B).

The inhibitory effect of Nm23-H1 on cell migration and invasion has been reported in some tumors (Boissan et al., 2010). We also observed increased motility and Rac1 activation in whole cell lysate of Nm23-H1-deprived U87MG (supplementary material Fig. S5C and data not shown). Therefore, Nm23-H1 could be involved in the suppression of cell movements through downregulation of total Rac1 activation. However, because the mutant Nm23-H1-N, which is unable to associate with α-catenin, did not rescue CIL without losing its function as a Tiam1 inhibitor, it would seem that recruitment of Nm23-H1 to the site of cell–cell contact is required for Nm23-H1 to cause CIL, and suggests that localized Rac1 activation at intercellular contacts is important for regulation of CIL.

Nm23-H1 has at least three functions: as an NDP kinase, a histidine kinase and a 3′ to 5′ exonuclease. However, Nm23-H1-mediated CIL is unlikely to correlate with these functions. We observed that Nm23-H1 K12Q, which has recently been reported to abrogate all these three functions of Nm23-H1, did not apparently affect the CIL of U87MG toward glias (data not shown) (Zhang et al., 2011). In contrast, the C-terminal region of Nm23-H1 was required to maintain the CIL of U87MG cells. Therefore, establishment of CIL through the C-terminus of Nm23-H1 might be a novel mechanism of Nm23 to suppress tumor invasion. This C-terminal portion contains the V2 region, which is diverse in the Nm23 family. Among Nm23-H1 to -H8, the V2 region is conserved more in Nm23-H2 (76% amino acids identity), and less in Nm23-H3 (42% identity). Therefore, Nm23-H1 and -H2 may preferentially contribute to the establishment of CIL.

Recent studies of cell migration of cranial neural crest (CNC) cells, a highly migratory and multipotent embryonic cell population, demonstrated that the directional migration of CNC cells in vivo is regulated by CIL, and N-cadherin plays a pivotal role during CIL (Carmona-Fontaine et al., 2008; Theveneau et al., 2010). Loss of CIL by reduction of N-cadherin in U87MG cells and glias suggests that N-cadherin-mediated cell–cell communication is also important for the establishment of CIL between tumor cells and stromal cells. In addition, Nm23-H1 might interact with several intercellular adhesion molecules besides N-cadherin through α-catenin. For example, E-cadherin also forms a complex with α-catenin and Nm23-H1 (Aktary et al., 2010). It should be elucidated whether Nm23-H1 also makes a complex with other adhesion molecules, and is involved in regulation of the CIL of other types of cancer.

Cell repulsion is a well-known effect of Eph–ephrin interaction (Pasquale, 2008; Poliakov et al., 2004). When an Eph-expressing cell meets another cell expressing cognate ephrin, they move in the opposite direction, which is the same behavior as in CIL. A recent study showed that EphA and ephrin-A mediate homotypic CIL of prostate cancer cells, whereas the forward signaling of EphBs in cancer cells disrupts CIL between cancer cells and fibroblasts (Astin et al., 2010). Therefore, we examined expression and activation of several Eph receptors in confrontation assays of U87MG and glia. EphA4, EphA5 and EphB2 were detected as major Eph receptors in U87MG cells by RT-PCR and western blotting. We did not detect activation of any of these receptors in cocultures with glia, whereas activation of EphB2 was observed when ephrin-B1-overexpressing U87MG cells were cocultured with glia, as expected (supplementary material Fig. S6). Although there was no evidence of activation of Eph receptors, we cannot rule out activation of other Ephs or ephrins expressed at low levels in U87MG cells that could cause repulsive CIL after contact with glia. It should be further elucidated whether Nm23-H1 modifies Eph–ephrin-mediated cell repulsion by interacting with them.

Extracellular interaction of EphB receptors activates ephrin-B1 signaling in tumor cells, which might also affect CIL and invasion, as we observed EphB2 expression in glia. However, a membrane-permeable peptide of ephrin-B1, which binds to Tiam1 (ephrin-B1295–313) disrupted invasion by U87MG and activated Rac1 as well as wild-type ephrin-B1 (supplementary material Fig. S5A). Therefore, association with Tiam1 may be enough for ephrin-B1 to cause invasion, although some other mechanisms could also be involved. An increase of in vivo invasion by U87MG cells as a result of overexpression of ephrin-B2 was also reported (Nakada et al., 2010). As ephrin-B1, -B2 and -B3 share a conserved motif necessary for binding to Tiam1, all these B-type ephrins might affect interaction of Tiam1 with Nm23-H1 and -H2, and regulate CIL behavior in malignancies.

This study suggests that ephrin-B1 contributes to the local activation of Rac1 through sequestration of Tiam1 from its inhibitor Nm23-H1. In addition, certain possibilities of ephrin-B1-mediated Rac1 activation need to be considered. For example, physical association of ephrin-B1 might alter the conformation of Tiam1 and affect its activity, or ephrin-B1 might modify the interaction of N-cadherin with β- and α-catenin at cell–cell contacts. However, we do not have any evidence at present that ephrin-B1 directly activates the GEF activity of Tiam1, and overexpression of ephrin-B1 did not affect the physical interaction of N-cadherin with β- and α-catenin (data not shown). As one of the mechanisms of ephrin-B1-mediated enhancement of cancer cell invasion, we previously reported that the C-terminus of ephrin-B1 regulates the exocytosis of matrix metalloproteinase (MMP) through the activation of Arf1 GTPase (34,Tanaka et al., 2007a). We detected elevation of MMP-1 and slight increase in the invasion of collagen by U87MG cells overexpressing ephrin-B1 (supplementary material Fig. S5B,D). However, elevation of MMP-1 was also detected in U87MG cells expressing ephrin-B1 E297K, D300K, which did not affect CIL or invasion by U87MG cells (supplementary material Fig. S5D, Movie 11). Therefore, if activation of MMP-1 contributes to invasion by ephrin-B1-overexpressing U87MG cells, it does not seem to play a major role in this assay. On the other hand, the CIL frequency of C6 ephrin-B1 miR cells was low, and the angle (θ:0°) and persistence suggested relatively mild CIL was induced compared with that of U87MG cells. Therefore, some other mechanisms besides CIL probably modify invasion by tumor cells, especially C6 cells in the spheroid confrontation assay.

This study revealed a novel function of Nm23-H1 in the establishment of CIL. Although we focused on the CIL of heterotypic cells to investigate the mechanism of stromal invasion of tumors, Nm23-H1 could also play a central role in homotypic CIL in various normal cells. In addition to the known enzymatic activities, Nm23-H1 C-terminus-mediated regulation of CIL is considered to contribute to metastasis-suppressing activity.

Plasmids, antibodies and reagents

Plasmids encoding full-length cDNAs of human ephrin-B1 and Nm23-H1 were described previously (Tanaka et al., 2007a; Otsuki et al., 2001). Rat cDNA of Nm23-H1 was PCR amplified from cDNA synthesized from rat kidney. Mutant forms of Nm23-H1 lacking the N-terminus (aa 37-152; Nm23-H1-C) or carboxyl terminus (aa 1-138; Nm23-H1-N) were generated by PCR-based techniques, and cloned into pCS2 with or without six copies of the myc-epitope tag at the N-terminus. Ephrin-B1E297K, D300K and Nm23-H1E124, 127K were generated using a site-directed mutagenesis kit (Stratagene, Santa Clara, CA). The PHCCEx region of human Tiam1 was described previously (Tanaka et al., 2004). To generate the recombinant retrovirus, cDNA was subcloned into a pDON-AI vector (Takara, Tokyo, Japan). N-cadherin–Fc was constructed by fusing the extracellular region of murine N-cadherin1–725 with Fc region of mouse IgG2b. N-cadherin–Fc fusion protein was purified from the culture medium of COS1 cells transfected with plasmids encoding N-cadherin–Fc using a protein-G–Sepharose column as described previously (Tanaka et al., 2005). Rabbit polyclonal antibody that recognizes the conserved C-terminal region of ephrin-B1, -B2 and -B3 (C18) was purchased from Santa Cruz Biotechnology Inc. (Santa Cruz, CA) The goat polyclonal antibody against ephrin-B1, which reacts with the entire extracellular domain, was purchased from R&D Systems (Minneapolis, MN). Polyclonal antibodies for Nm23-H1 and MMP-1 were purchased from Santa Cruz and Millipore, respectively. Monoclonal antibody for N-cadherin, which reacts with the cytoplasmic region of N-cadherin was purchased from BD Biosciences. Monoclonal antibodies for N-cadherin, the HA epitope and the myc epitope (9E10) were purchased from BD Biosciences (Franklin Lakes, NJ), BAbCO (Richmond, CA, USA) and Santa Cruz, respectively. Ephrin-B1295–313 peptide was synthesized as a fusion peptide of the membrane-permeable domain of HIV-TAT as follows: GRKKRRQRRRPPQGGGAGTEPSDIIIPLRTTENNY. As a control, the scrambled sequence peptide that contains the same amino acid composition was synthesized: GRKKRRQRRRPPQGGGETLITIYERPGADNTSPIN.

Preparation of N-cadherin–Fc-coated beads

Latex-sulfate microspheres (5.2 µm diameter; International Dynamics Corporation) were coated with N-cadherin–Fc fusion protein according to the procedure described previously (Tanaka et al., 2005). Briefly, beads were suspended in 0.1 M borate buffer, pH 8.0, and incubated with goat anti-mouse IgG (Fc-specific) antibody (ICN, Aurora, OH). After the beads were washed with PBS, they were incubated with N-cadherin–Fc protein, washed three times with PBS, and resuspended in PBS containing 5 mg/ml bovine serum albumin (BSA).

Cell culture and transfection

Glioblastoma cell lines used in this study were obtained from the American Type Culture Collection (Rockville, MD, USA). Glias were isolated from the brains of embryonic day 14 Wistar rat fetuses. Briefly, brains were aseptically dissected, placed in dishes containing DMEM, and meningeal coverings were removed. Then, the brains were dissociated with 0.025% trypsin by pippeting. The single cells were transferred into DMEM containing 10% FBS, gentamycin (0.1 mg/ml), penicillin (100 U/ml) and streptomycin (100 mg/ml). The U87MG cell line and COS1 cells were cultured in DMEM supplemented with 10% fetal bovine serum. For transient expression assays, COS1 cells were transfected with plasmid DNA using FuGene6 reagent (Invitrogen, Carlsbad, CA). Recombinant retroviral plasmid, pDON-AI was co-transfected with pCL-10A1 retrovirus packaging vector (IMGENEX, San Diego CA) into 293gp cells to allow the production of retroviral particles (Tanaka et al., 2007a). U87MG cells stably overexpressing ephrin-B1 were established, after retroviral infection, through selection in medium containing G418 (600 µg/ml). In some experiments, peptides were added to the cultured cells at the final concentration of 5 µM.

Construction of siRNA and miR RNAi vectors

A system stably expressing miRNA was generated using the BLOCK-iT PolII miR RNAi expression vector kit (Invitrogen) according to the manufacturer's instructions. In the generation of the miR RNAi vectors for Nm23-H1, N-cadherin and Tiam1, the following forward primers were used. Nm23-H1, 5′-TGCTGAATGAAGGTACGCTCACAGTTGTTTTGGCCACTGACTGACAACTGTGAGTACCTTCATT-3′; N-cadherin#1, 5′-TGCTGTTCACCAGAAGCCTCTACAGAGTTTTGGCCACTGACTGACTCTGTAGACTTCTGGTGAA-3′. N-cadherin#2, 5′-TGCTGTGAAGATACCAGTTGGAGGCTGTTTTGGCCACTGACTGACAGCCTCCATGGTATCTTCA-3′. Tiam1, 5′-TGCTGAAAGCTCGCCGTCTCCATGAAGTTTTGGCCACTGACTGACTTCATGGACGGCGAGCTTT-3′.

U87MG cells stably expressing the miRNA vector for Nm23-H1 were established and cultured in medium containing blasticidin (Invitrogen) at a concentration of 10 µg/ml for 3 weeks. Small interfering RNAs (siRNA) of rat N-cadherin were synthesized as follows (CosmoBio, Tokyo, Japan). Rat N-cadherin sense no. 1: 5′-CCAUCAAACCUGUGGGAAU-3′; Rat N-cadherin sense no. 2: 5′-CCGCAAGAGCUUGUCAGAA-3′; The control siRNA (scramble II duplex) was purchased from Dharmacon. siRNAs were incorporated into rat glia cells with Lipofectamine RNAiMax according to the manufacturer's instructions (Invitrogen).

Spheroid confrontation assay

Spheroid confrontation assays were performed as described previously (Golembieski et al., 1999) with some modifications. Briefly, glias were labeled with green fluorescent 3,3′-dioctadecycloxacarbo-cyanine perchlorate (DiO) dye and tumor cells were labeled with the red fluorescent 1,1′-dioctadecyl-3,3,3′,3′-tetramethyllindo-carbocyanine perchlorate (DiI) dye according to the manufacturer's instructions (Invitrogen). The fluorescently labeled cells were detached with trypsin, and samples of 3×105 cells were collected in complete growth medium and cultured in six-well plates coated with 1% agarose gel overnight for the formation of spheroids of cell aggregates. For 3D confrontation assays, the glia and tumor aggregates were brought into contact by means of a sterile needle in 96-well plates coated with 1% agarose gel. After 72 hrs incubation, the fused aggregates were fixed in 4% paraformaldehyde, and sealed with a coverslip on a glass slide. When confrontation assays were performed in 2D culture, aggregates of glia and tumor cells were randomly plated on new 24-well plates coated with fibronectin (20 µg/ml), and incubation continued for 12 hrs until the cells grew out from each of the aggregates meet each other. The invasion of normal glia aggregates by tumor spheroids was observed by confocal microscopy (LSM510, Zeiss). From z-stack images of each sample, the single plane with maximal intermingling of glia and tumor cells was selected. In cases where no evident invasion was observed, a single plane of the middle z-position was chosen. In some cases, each X-Y plane of Z-stacks was pitched as a ‘3D cut view’ (LSM510, Zeiss) to show the images in different planes together (supplementary material Fig. S6).

Live-cell imaging

Prior to imaging, glias and tumor cells were labeled with DiO and DiI, respectively. Collisions of tumor cells and glia were analyzed by phase-contrast time-lapse microscopy with the 40× objective for up to 8 hrs (37°C, 5% CO2; BZ-9000, Keyence). Fluorescence was imaged prior to the start of imaging in order to distinguish glia and tumor cell. When time-lapse analysis was performed in the spheroid confrontation assay, images were taken by multi-acquisition phase-contrast time-lapse microscopy with the 20× objective at 10 min intervals for up to 24 hrs (37°C, 5% CO2). The images were analyzed by tracking migration of tumor cells using ImageJ software and the Manual tracking plugin. The positions of collisions (at t) between tumor cells and glial cells were determined manually, and the positions at the appropriate time-point before (t−Δt) and after (tt) the collision were also determined. Basically, Δt is 1 hr. The angle, θ, was calculated from these points. Persistence before and after collisions was determined from the direct distances between the points divided by the actual distances traveled during Δt. The nucleus was used as a marker, but the protrusion was tracked in some cases when the displaced distance of protrusion was larger than that of the nucleus. Both of them were correlated (e.g. supplementary material Movies 4, 6 and 12).

Invasion index and statistical analysis

To quantify the invasion of the glial cell (labeled green) region by the glioma cells (labeled red) the fluorescence images were analyzed using ImageJ software. The invasion index (I) was calculated as the ratio (percentage) of the overlapping glioma area (invaded red: iR) to the glial cell area (total green: tG). I (%) = iR/tG×100. At least three assays were performed for each sample, and the invasion index is shown as mean ± s.d.

Immunoprecipitation and immunoblotting

Cell lysates were prepared with protease inhibitors in PLC buffer [50 mM Hepes (pH 7.5), 150 mM NaCl, 1.5 mM MgCl2, 1 mM EGTA, 10% glycerol, 100 mM NaF, 1 mM Na3VO4 and 1% Triton X-100]. To precipitate the proteins, 1 µg of monoclonal or affinity purified polyclonal antibody was incubated with 500 µg of cell lysate for 2 hrs at 4°C, and then precipitated with protein-G–agarose for 1 hr at 4°C. Immunoprecipitates were extensively washed with PLC buffer, separated by SDS-PAGE and immunoblotted.

Cell migration assay

Migration assays were performed using Transwell chambers with a polycarbonate nucleopore membrane (BD Falcon, Franklin Lakes, NJ). Filters (8-µm pore size) were rehydrated with 100 µl of medium. Then, 1×104 cells in 200 µl of serum-free medium were seeded onto the upper part of each chamber, and the lower compartment were filled with 600 µl of the same medium with 10% fetal bovine serum (FBS). After incubation for 8 hrs at 37°C, non-migrated cells on the upper surface of the filter were removed with a cotton swab, and the migrated cells on the lower surface of the filter were fixed and stained with Giemsa's stain solution. Migration was determined by counting cells in five microscopic fields per well, and the extent of migration was expressed as the average number of cells per microscopic field. The assays were performed three times.

Rac1–GTP pull-down assay

The activation of Rac1 was examined by affinity precipitation of GTP-bound Rac1 with the glutathione S-transferase (GST) fusion protein of the p21-binding domain of PAK1 (GST–PBD) as described previously (Otsuki et al., 2001). Briefly, cell lysates were prepared in lysis buffer [50 mM Hepes pH 7.5, 150 mM NaCl, 10 mM MgCl2, 10% glycerol, 100 mM NaF, 1 mM Na3VO4 and 1% Triton X-100] and then incubated with gluthathione–Sepharose beads containing a GST–PBD fusion protein for 45 min at 4°C. Precipitates were washed four times in the same buffer, and the precipitated GTP-bound Rac1 was detected by immunoblotting. Band intensities were quantified using ImageJ software, and the amount of GTP–Rac1 was normalized to total Rac1. Experiments were repeated three times.

In vivo intracranial dissemination assay

All animal experimental protocols were approved by the Committee for Ethics of Animal Experimentation, and the experiments were conducted in accordance with the guidelines for Animal Experiments in Akita University. Intracranial dissemination of tumors was tested by intracranial injection of 2×105 tumor cells suspended in 30 µl DMEM into 6-week-old BALB/c nude mice (n = 10, each). The mice were sacrificed 12 days after injection and subjected to histopathological examination of the tumors. To examine survival rate, tumor cells were injected into another set of mice as above, and evaluated 21 days later. To quantify irregularity of tumor expansion in mice brains, HE-stained section was photographed, and the images were analyzed using ImageJ software. The boundaries were determined by thresholding of the color density of the tissues, and the edge points were manually tracked. We picked a few random points (B) and the fixed positions at the end of the image (A,C) on the invasive edge, and drew the rectangle and the circumscribed circle (CC) of it (supplementary material Fig. S1C). The diameter of CC (R) was calculated from the distances between three points (formula 1). The center angle of the minor arc between A and C was calculated by the length of AC and R (formula 2). The length of the minor arc (Lc) was determined by the diameter and the center angle (formula 3). We calculated the total length of the edge-line from the trajectory of the image (La). The irregular index (Ir) is defined as La/Lc. For instance, when any B is on the same CC, Ir becomes 1, which means the smooth boundary between tumor and the invaded tissues.

Immunocytochemical staining

Immunocytochemical staining was performed as previously described (Tanaka et al., 2005). In some experiments, the coverglasses were coated with fibronectin. For transfection, 5.0×104 cells were seeded on a glass and then fixed in 4% paraformaldehyde and stained. The staining was examined using a confocal microscopic system (Zeiss).

Immunohistochemical analysis

Tumor tissues in the brain of nude mice were fixed, and embedded in paraffin. Paraffin blocks were sectioned and subjected to immunohistochemical staining using the indirect polymer method with Envision reagent (DAKO, Carpinteria, CA). Antigen retrieval was performed by placing sections in citrate buffer and autoclaving as described in the manufacturer's instructions.

Funding

This work was supported by the Japan Society for the Promotion of Science (JSPS) [grant numbers 22300324, 23650590 to M.T. and 24700962 to S.K.]; the National Cancer Center Research and Development Fund [grant number 23-A-9 to M.T.]; the Uehara Memorial Foundation; The Naito Foundation; and The Yasuda Medical Foundation.

Abercrombie
M.
,
Heaysman
J. E. M.
(
1953
).
Observations on the social behaviour of cells in tissue culture. I. Speed of movement of chick heart fibroblasts in relation to their mutual contacts.
Exp. Cell Res.
5
,
111
131
.
Aktary
Z.
,
Chapman
K.
,
Lam
L.
,
Lo
A.
,
Ji
C.
,
Graham
K.
,
Cook
L.
,
Li
L.
,
Mackey
J. R.
,
Pasdar
M.
(
2010
).
Plakoglobin interacts with and increases the protein levels of metastasis suppressor Nm23-H2 and regulates the expression of Nm23-H1.
Oncogene
29
,
2118
2129
.
Astin
J. W.
,
Batson
J.
,
Kadir
S.
,
Charlet
J.
,
Persad
R. A.
,
Gillatt
D.
,
Oxley
J. D.
,
Nobes
C. D.
(
2010
).
Competition amongst Eph receptors regulates contact inhibition of locomotion and invasiveness in prostate cancer cells.
Nat. Cell Biol.
12
,
1194
1204
.
Batlle
E.
,
Bacani
J.
,
Begthel
H.
,
Jonkheer
S.
,
Gregorieff
A.
,
van de Born
M.
,
Malats
N.
,
Sancho
E.
,
Boon
E.
,
Pawson
T.
, et al. 
(
2005
).
EphB receptor activity suppresses colorectal cancer progression.
Nature
435
,
1126
1130
.
Boissan
M.
,
Wendum
D.
,
Arnaud–Dabernat
S.
,
Munier
A.
,
Debray
M.
,
Lascu
I.
,
Daniel
J. Y.
,
Lacombe
M. L.
(
2005
).
Increased lung metastasis in transgenic NM23-Null/SV40 mice with hepatocellular carcinoma.
J. Natl. Cancer Inst.
97
,
836
845
.
Boissan
M.
,
Dabernat
S.
,
Peuchant
E.
,
Schlattner
U.
,
Lascu
I.
,
Lacombe
M. L.
(
2009
).
The mammalian Nm23/NDPK family: from metastasis control to cilia movement.
Mol. Cell. Biochem.
329
,
51
62
.
Boissan
M.
,
De Wever
O.
,
Lizarraga
F.
,
Wendum
D.
,
Poincloux
R.
,
Chignard
N.
,
Desbois–Mouthon
C.
,
Dufour
S.
,
Nawrocki–Raby
B.
,
Birembaut
P.
, et al. 
(
2010
).
Implication of metastasis suppressor NM23-H1 in maintaining adherens junctions and limiting the invasive potential of human cancer cells.
Cancer Res.
70
,
7710
7722
.
Carmona–Fontaine
C.
,
Matthews
H. K.
,
Kuriyama
S.
,
Moreno
M.
,
Dunn
G. A.
,
Parsons
M.
,
Stern
C. D.
,
Mayor
R.
(
2008
).
Contact inhibition of locomotion in vivo controls neural crest directional migration.
Nature
456
,
957
961
.
Castaño
J.
,
Davalos
V.
,
Schwartz
S.
,
Jr and Arango
D.
(
2008
).
EPH receptors in cancer.
Histol. Histopathol.
23
,
1011
1023
.
Castellvi
J.
,
Garcia
A.
,
de la Torre
J.
,
Hernandez
J.
,
Gil
A.
,
Xercavins
J.
,
Ramón y Cajal
S.
(
2006
).
Ephrin B expression in epithelial ovarian neoplasms correlates with tumor differentiation and angiogenesis.
Hum. Pathol.
37
,
883
889
.
Daar
I.
(
2012
).
Non-SH2/PDZ reverse signaling by ephrins.
Semin. Cell Dev. Biol.
23
,
65
74
.
Fan
Z.
,
Beresford
P. J.
,
Oh
D. Y.
,
Zhang
D.
,
Lieberman
J.
(
2003
).
Tumor suppressor NM23-H1 is a granzyme A-activated DNase during CTL-mediated apoptosis, and the nucleosome assembly protein SET is its inhibitor.
Cell
112
,
659
672
.
Golembieski
W. A.
,
Ge
S.
,
Nelson
K.
,
Mikkelsen
T.
,
Rempel
S. A.
(
1999
).
Increased SPARC expression promotes U87 glioblastoma invasion in vitro.
Int. J. Dev. Neurosci.
17
,
463
472
.
Habets
G. G.
,
Scholtes
E. H.
,
Zuydgeest
D.
,
van der Kammen
R. A.
,
Stam
J. C.
,
Berns
A.
,
Collard
J. G.
(
1994
).
Identification of an invasion-inducing gene, Tiam-1, that encodes a protein with homology to GDP-GTP exchangers for Rho-like proteins.
Cell
77
,
537
549
.
Hartsough
M. T.
,
Morrison
D. K.
,
Salerno
M.
,
Palmieri
D.
,
Ouatas
T.
,
Mair
M.
,
Patrick
J.
,
Steeg
P. S.
(
2002
).
Nm23-H1 metastasis suppressor phosphorylation of kinase suppressor of Ras via a histidine protein kinase pathway.
J. Biol. Chem.
277
,
32389
32399
.
Holmberg
J.
,
Genander
M.
,
Halford
M. M.
,
Annerén
C.
,
Sondell
M.
,
Chumley
M. J.
,
Silvany
R. E.
,
Henkemeyer
M.
,
Frisén
J.
(
2006
).
EphB receptors coordinate migration and proliferation in the intestinal stem cell niche.
Cell
125
,
1151
1163
.
Iwashita
S.
,
Fujii
M.
,
Mukai
H.
,
Ono
Y.
,
Miyamoto
M.
(
2004
).
Lbc proto-oncogene product binds to and could be negatively regulated by metastasis suppressor nm23-H2.
Biochem. Biophys. Res. Commun.
320
,
1063
1068
.
Jung
S.
,
Paek
Y. W.
,
Moon
K. S.
,
Wee
S. C.
,
Ryu
H. H.
,
Jeong
Y. I.
,
Sun
H. S.
,
Jin
Y. H.
,
Kim
K. K.
,
Ahn
K. Y.
(
2006
).
Expression of Nm23 in gliomas and its effect on migration and invasion in vitro.
Anticancer Res.
26
,
249
258
.
Kataoka
H.
,
Tanaka
M.
,
Kanamori
M.
,
Yoshii
S.
,
Ihara
M.
,
Wang
Y. J.
,
Song
J. P.
,
Li
Z. Y.
,
Arai
H.
,
Otsuki
Y.
, et al. 
(
2002
).
Expression profile of EFNB1, EFNB2, two ligands of EPHB2 in human gastric cancer.
J. Cancer Res. Clin. Oncol.
128
,
343
348
.
Lacombe
M. L.
,
Milon
L.
,
Munier
A.
,
Mehus
J. G.
,
Lambeth
D. O.
(
2000
).
The human Nm23/nucleoside diphosphate kinases.
J. Bioenerg. Biomembr.
32
,
247
258
.
Mayor
R.
,
Carmona–Fontaine
C.
(
2010
).
Keeping in touch with contact inhibition of locomotion.
Trends Cell Biol.
20
,
319
328
.
McDermott
W. G.
,
Boissan
M.
,
Lacombe
M. L.
,
Steeg
P. S.
,
Horak
C. E.
(
2008
).
Nm23-H1 homologs suppress tumor cell motility and anchorage independent growth.
Clin. Exp. Metastasis
25
,
131
138
.
Meyer
S.
,
Hafner
C.
,
Guba
M.
,
Flegel
S.
,
Geissler
E. K.
,
Becker
B.
,
Koehl
G. E.
,
Orsó
E.
,
Landthaler
M.
,
Vogt
T.
(
2005
).
Ephrin-B2 overexpression enhances integrin-mediated ECM-attachment and migration of B16 melanoma cells.
Int. J. Oncol.
27
,
1197
1206
.
Nakada
M.
,
Anderson
E. M.
,
Demuth
T.
,
Nakada
S.
,
Reavie
L. B.
,
Drake
K. L.
,
Hoelzinger
D. B.
,
Berens
M. E.
(
2010
).
The phosphorylation of ephrin-B2 ligand promotes glioma cell migration and invasion.
Int. J. Cancer
126
,
1155
1165
.
Otsuki
Y.
,
Tanaka
M.
,
Yoshii
S.
,
Kawazoe
N.
,
Nakaya
K.
,
Sugimura
H.
(
2001
).
Tumor metastasis suppressor nm23H1 regulates Rac1 GTPase by interaction with Tiam1.
Proc. Natl. Acad. Sci. USA
98
,
4385
4390
.
Pasquale
E. B.
(
2008
).
Eph-ephrin bidirectional signaling in physiology and disease.
Cell
133
,
38
52
.
Poliakov
A.
,
Cotrina
M.
,
Wilkinson
D. G.
(
2004
).
Diverse roles of eph receptors and ephrins in the regulation of cell migration and tissue assembly.
Dev. Cell
7
,
465
480
.
Rosenzweig
T.
,
Ziv–Av
A.
,
Xiang
C.
,
Lu
W.
,
Cazacu
S.
,
Taler
D.
,
Miller
C. G.
,
Reich
R.
,
Shoshan
Y.
,
Anikster
Y.
, et al. 
(
2006
).
Related to testes-specific, vespid, and pathogenesis protein-1 (RTVP-1) is overexpressed in gliomas and regulates the growth, survival, and invasion of glioma cells.
Cancer Res.
66
,
4139
4148
.
Steeg
P. S.
,
Ouatas
T.
,
Halverson
D.
,
Palmieri
D.
,
Salerno
M.
(
2003
).
Metastasis suppressor genes: basic biology and potential clinical use.
Clin. Breast Cancer
4
,
51
62
.
Suzuki
E.
,
Ota
T.
,
Tsukuda
K.
,
Okita
A.
,
Matsuoka
K.
,
Murakami
M.
,
Doihara
H.
,
Shimizu
N.
(
2004
).
nm23-H1 reduces in vitro cell migration and the liver metastatic potential of colon cancer cells by regulating myosin light chain phosphorylation.
Int. J. Cancer
108
,
207
211
.
Tanaka
M.
,
Ohashi
R.
,
Nakamura
R.
,
Shinmura
K.
,
Kamo
T.
,
Sakai
R.
,
Sugimura
H.
(
2004
).
Tiam1 mediates neurite outgrowth induced by ephrin-B1 and EphA2.
EMBO J.
23
,
1075
1088
.
Tanaka
M.
,
Kamata
R.
,
Sakai
R.
(
2005
).
Phosphorylation of ephrin-B1 via the interaction with claudin following cell-cell contact formation.
EMBO J.
24
,
3700
3711
.
Tanaka
M.
,
Sasaki
K.
,
Kamata
R.
,
Sakai
R.
(
2007a
).
The C-terminus of ephrin-B1 regulates metalloproteinase secretion and invasion of cancer cells.
J. Cell Sci.
120
,
2179
2189
.
Tanaka
M.
,
Kamata
R.
,
Takigahira
M.
,
Yanagihara
K.
,
Sakai
R.
(
2007b
).
Phosphorylation of ephrin-B1 regulates dissemination of gastric scirrhous carcinoma.
Am. J. Pathol.
171
,
68
78
.
Tanaka
M.
,
Kamata
R.
,
Yanagihara
K.
,
Sakai
R.
(
2010
).
Suppression of gastric cancer dissemination by ephrin-B1-derived peptide.
Cancer Sci.
101
,
87
93
.
Terawaki
S.
,
Kitano
K.
,
Mori
T.
,
Zhai
Y.
,
Higuchi
Y.
,
Itoh
N.
,
Watanabe
T.
,
Kaibuchi
K.
,
Hakoshima
T.
(
2010
).
The PHCCEx domain of Tiam1/2 is a novel protein- and membrane-binding module.
EMBO J.
29
,
236
250
.
Theveneau
E.
,
Marchant
L.
,
Kuriyama
S.
,
Gull
M.
,
Moepps
B.
,
Parsons
M.
,
Mayor
R.
(
2010
).
Collective chemotaxis requires contact-dependent cell polarity.
Dev. Cell
19
,
39
53
.
Varelias
A.
,
Koblar
S. A.
,
Cowled
P. A.
,
Carter
C. D.
,
Clayer
M.
(
2002
).
Human osteosarcoma expresses specific ephrin profiles: implications for tumorigenicity and prognosis.
Cancer
95
,
862
869
.
Vaught
D.
,
Brantley–Sieders
D. M.
,
Chen
J.
(
2008
).
Eph receptors in breast cancer: roles in tumor promotion and tumor suppression.
Breast Cancer Res.
10
,
217
224
.
Wagner
P. D.
,
Steeg
P. S.
,
Vu
N. D.
(
1997
).
Two-component kinase-like activity of nm23 correlates with its motility-suppressing activity.
Proc. Natl. Acad. Sci. USA
94
,
9000
9005
.
Zhang
Q.
,
McCorkle
J. R.
,
Novak
M.
,
Yang
M.
,
Kaetzel
D. M.
(
2011
).
Metastasis suppressor function of NM23-H1 requires its 3′ to 5′ exonuclease activity.
Int. J. Cancer
128
,
40
50
.

Supplementary information