Mechanobiology June 26th - June 2nd 2016

Mechanobiology: June 26th  - June 2nd 2016

Src-mediated phosphorylation of mammalian Abp1 (DBNL) regulates podosome rosette formation in transformed fibroblasts
Lindsy R. Boateng, Christa L. Cortesio, Anna Huttenlocher

Summary

Podosomes are dynamic actin-based structures that mediate adhesion to the extracellular matrix and localize matrix degradation to facilitate cell motility and invasion. Drebrin-like protein (DBNL), which is homologous to yeast mAbp1 and is therefore known as mammalian actin-binding protein 1 (mAbp1), has been implicated in receptor-mediated endocytosis, vesicle recycling and dorsal ruffle formation. However, it is not known whether mAbp1 regulates podosome formation or cell invasion. In this study, we found that mAbp1 localizes to podosomes and is necessary for the formation of podosome rosettes in Src-transformed fibroblasts. Despite their structural similarity, mAbp1 and cortactin play distinct roles in podosome regulation. Cortactin was necessary for the formation of podosome dots, whereas mAbp1 was necessary for the formation of organized podosome rosettes in Src-transformed cells. We identified specific Src phosphorylation sites, Tyr337 and Tyr347 of mAbp1, which mediate the formation of podosome rosettes and degradation of the ECM. In contrast to dorsal ruffles, the interaction of mAbp1 with WASP-interacting protein (WIP) was not necessary for the formation of podosome rosettes. Finally, we showed that depletion of mAbp1 increased invasive cell migration, suggesting that mAbp1 differentially regulates matrix degradation and cell invasion. Collectively, our findings identify a role for mAbp1 in podosome rosette formation and cell invasion downstream of Src.

Introduction

Efficient cell migration and invasion are fundamental to a variety of cellular processes including developmental cell movements, leukocyte trafficking in immunity and fibroblast movement during wound healing. Defects in cell motility can contribute to the pathogenesis of diverse diseases including Wiskott–Aldrich syndrome (Jones et al., 2002; Zicha et al., 1998) and congenital neutropenias such as WHIM syndrome (Walters et al., 2010; Wetzler et al., 1990). Moreover, increased cell migration and invasion are hallmarks of invasive and metastatic cancer. Cell migration and invasion require the dynamic interaction between a cell and its surrounding matrix as well as the capacity to degrade the extracellular matrix. Some highly motile and invasive cells form organized actin-based structures known as podosomes or invadopodia (reviewed by Albiges-Rizo et al., 2009). Podosomes are integrin-mediated adhesive structures that also function to promote localized matrix degradation and are characteristically found in leukocytes and Src-transformed fibroblasts (Linder and Kopp, 2005; Gimona et al., 2008). By contrast, invadopodia are generally described as the invasive actin-based protrusive structures formed in some metastatic cancer cells (Weaver, 2006).

Dynamic regulation of podosomes is probably crucial for rapid cell motility because podosomes adhere to and degrade the extracellular matrix (ECM), and must assemble and disassemble for cell migration to occur (Adams, 2002; Linder and Aepfelbacher, 2003). Podosomes comprise an outer ring of integrin-associated proteins such as vinculin, talin and paxillin, and an inner core with actin regulatory proteins such as Arp2/3, WIP (WASP-interacting protein) and WASP (Linder and Kopp, 2005; Buccione et al., 2004; Linder and Aepfelbacher, 2003). Src is a non-receptor tyrosine kinase that plays a crucial role in regulating podosome and invadopodia formation and turnover (Tarone et al., 1985; Marchisio et al., 1987; Soriano et al., 1991; Cortesio et al., 2008; Chan et al., 2009). Many Src substrates such as cortactin, Tks5 and paxillin are essential for the dynamic regulation of podosomes and invadopodia (Badowski et al., 2008; Courtneidge et al., 2005; Mader et al., 2011). The scaffolding protein cortactin regulates actin polymerization and formation of podosomes through its interaction with Arp2/3 and N-WASP, and regulates secretion of matrix metalloproteinases at podosomes and invadopodia (Daly, 2004; Webb et al., 2006; Ayala et al., 2008; Clark and Weaver, 2008; Oser et al., 2009; Artym et al., 2006; Bowden et al., 2006; Bañón-Rodríguez et al., 2011; Uruno et al., 2001).

Mammalian actin-binding protein 1 (mAbp1, also known as drebrin-like protein DBNL, Hip-55, SH3P7) is an F-actin binding protein that has high structural similarity to cortactin. Both mAbp1 and cortactin bind F-actin with their N-terminal actin binding domains and mediate protein–protein interactions through their C-terminal proline-rich and SH3 domains (Fig. 1A). mAbp1 was initially identified in a phage display screen for SH3-domain-containing proteins (Sparks et al., 1996), and two tyrosine residues in the proline-rich region were identified as the Src-phospho-acceptor sites (Larbolette, 1999; Lock et al., 1998). The SH3 domain has been shown to interact with proteins involved in diverse functions including synaptogenesis, endocytosis and cell motility (Pinyol, 2007; Fenster et al., 2003; Han et al., 2003; Kessels et al., 2001; Hou et al., 2003; Cortesio et al., 2010). Although mAbp1-knockout mice are viable, they lack motor coordination and display behavioral defects, which are in part due to aberrant synaptic vesicle recycling (Connert et al., 2006). Several studies also suggest an important role for mAbp1 in immune function through its regulation of T-cell receptor endocytosis at the immune synapse (Le Bras et al., 2004), cytokine production (Han et al., 2005) and B cell receptor processing (Onabajo et al., 2008). In addition, we recently reported that mAbp1 localizes to growth factor-induced dorsal ruffles and regulates dorsal ruffle formation through its interaction with WIP (Cortesio et al., 2010). mAbp1 has also been reported to regulate cytoskeletal rearrangements in neurons through regulation of N-WASP and Arp2/3 (Pinyol et al., 2007). Although mAbp1 has been implicated in regulating actin dynamics and is similar to cortactin, its role in regulating podosome formation or dynamics has not been previously addressed.

Fig. 1.

mAbp1 localizes to podosomes in Src(Y527F)-transformed fibroblasts. (A) Schematic of the domain structures of mAbp1 and cortactin. mAbp1 uses a conserved actin-depolymerizing factor homology domain (ADF-H) and cortactin uses a 6.5 tandem repeat region to bind actin at the N-terminus. Both mAbp1 and cortactin have a proline-rich (PR) domain and Src-homology 3 (SH3) domain at the C-terminus. Cortactin and mAbp1 are tyrosine (Y) phosphorylated by Src in the proline-rich region. Specific binding partners are shown. (BD) SrcY527F-transformed NIH3T3 fibroblasts were plated on fibronectin (FN)-coated coverslips for 3 hours, fixed and co-stained with Rhodamine phalloidin and antibodies against mAbp1, cortactin or vinculin. mAbp1 colocalizes with cortactin and vinculin at sites of actin-rich podosome rosettes. Solid and hollow arrows represent podosome dots and rosettes, respectively. Scale bar: 20 μm.

Src-mediated transformation of fibroblasts [v-Src or Src(Y527F)] induces an invasive phenotype and the formation of two types of podosomes – dots and rosettes (Marchisio et al., 1987; Tarone et al., 1985). Dot-like podosomes in Src-transformed cells, migratory dendritic cells, osteoclasts and macrophages resemble invadopodia in cancer cells (Calle et al., 2004; Linder and Aepfelbacher, 2003; Linder and Kopp, 2005). Both podosome dots and invadopodia can degrade the extracellular matrix at distinct foci, and their formation is important for invasion through tissues (Buccione et al., 2004; Yamaguchi et al., 2006). In addition to dot-like podosomes, Src-transformed fibroblasts, some cancer cells, endothelial cells, and osteoclasts also form highly organized adhesive structures known as podosome rosettes. These organized structures mediate adhesion and migration of pancreatic cancer cells through the basal lamina (Kocher et al., 2009) and have also been implicated in mediating force transmission and mechanotransduction in a similar manner to focal adhesions (Collin et al., 2008). In addition, podosome rosettes in osteoclasts can mature into sealing rings that mediate degradation of the bone matrix (Teti et al., 1991; Destaing et al., 2003).

In this study, we identify a role for mAbp1 in podosome formation and invasive migration of SrcY527F-transformed NIH3T3 fibroblasts. We show that mAbp1 is dispensable for formation of podosome dots but is necessary for the formation of organized podosome rosettes in Src-transformed cells. We also found that Tyr350 of mAbp1 is crucial for the formation of podosome rosettes downstream of Src. Finally, we show that mAbp1-deficient cells are more invasive despite forming fewer podosome rosettes, suggesting that formation of podosome rosettes, under some conditions, can limit cell invasion. Taken together, our findings identify a role for mAbp1 in the regulation of podosome formation and invasion downstream of Src.

Results

Localization of mAbp1 and cortactin at podosomes in Src-transformed fibroblasts

mAbp1 is structurally similar to cortactin, and has overlapping and unique interacting partners through its SH3 domain (Fig. 1A). To determine whether mAbp1 and cortactin colocalize at podosomes, we immunolabeled endogenous mAbp1 and cortactin in SrcY527F-transformed NIH3T3 fibroblasts. Src-transformed fibroblasts form both organized podosome rosettes and dot-like structures on a fibronectin matrix. Actin and cortactin are well-established core components of podosomes. We found that endogenous mAbp1 colocalized with actin (Fig. 1B) and cortactin (Fig. 1C) at both podosome dots and rosettes. Although all podosomes contained cortactin, some podosomes did not have mAbp1, suggesting that mAbp1 localizes to only a subset of podosomes. To further characterize the actin-based dot and rosette structures formed in Src-transformed fibroblasts, we immunolabeled endogenous Arp2/3, Tks5, vinculin and paxillin in Src-transformed cells (supplementary material Fig. S1). As expected, Arp2/3 and Tks5 localized to both dot and rosette structures. By contrast, paxillin and vinculin localized to both focal contacts and podosome rosettes, but not dot podosomes (supplementary material Fig. S1). Accordingly, mAbp1 colocalized with vinculin at podosome rosettes (Fig. 1D). Interestingly, we found that Src-transformed cells that formed prominent podosome rosettes, had few focal adhesions, which is consistent with the idea that podosome rosettes are the primary adhesive structure in some Src-transformed fibroblasts (supplementary material Fig. S2) (Oikawa et al., 2008; Gimona et al., 2008; Tarone et al., 1985).

Fig. 2.

Effects of mAbp1 and cortactin on formation of podosome rosettes and dot podosomes. (A) Representative western blot of Src, mAbp1, cortactin and AKT expression in Src(Y527F)-transformed NIH3T3 cells stably expressing control siRNA, or siRNA to knock down mAbp1 or cortactin. (B) Relative expression levels of mAbp1 and cortactin in control and knockdown cell lines determined by densitometry. (C) Src(Y527F)-transformed fibroblasts stably expressing siRNA were plated on FN-coated coverslips for 3 hours, fixed and stained with antibody against mAbp1 and Rhodamine phalloidin. Solid and hollow arrows represent podosome dots and rosettes, respectively. (D) The percentage of Src-transformed cells producing podosome rosettes and (E) the average number of podosome dots per cell was quantified by counting at least 50 cells per condition (control, mAbp1 or cortactin siRNA). Each bar represents the average value from three independent experiments. Error bars, s.e.m; **P<0.01; ***P<0.001; one-way ANOVA with Tukey’s post hoc test. Scale bar: 20 μm.

Role of mAbp1 in the formation of podosome rosettes and dot podosomes

To determine whether mAbp1 regulates the formation of podosomes, we expressed activated Src (SrcY527F) in control and mAbp1-deficient NIH3T3 cells and quantified podosomes using Rhodamine phalloidin to label actin. mAbp1-deficient cells had a knockdown efficiency of approximately 80% compared with the control (Fig. 2A,B). Ectopic expression of SrcY527F induced the formation of both podosome rosettes and dots in control cells (Fig. 2C). Depletion of mAbp1 impaired the formation of rosettes induced by Src-transformation approximately threefold compared with control cells (Fig. 2D). By contrast, mAbp1-deficient cells had a twofold increase in the average number of podosome dots formed per cell compared with control cells (Fig. 2E). However, depletion of mAbp1 did not affect the localization of cortactin to podosome dots (supplementary material Fig. S3). These results suggest that mAbp1 differentially regulates the formation of podosome rosettes and dots. In contrast to mAbp1, cortactin-deficient cells formed fewer podosome dots than was observed in control cells, but there was no statistically significant difference in the formation of podosome rosettes (Fig. 2D,E). Taken together, the findings suggest that mAbp1, but not cortactin, is necessary for the formation of podosome rosettes.

We found that mAbp1-deficient cells formed more podosome dots (Fig. 2C,E); however, the mechanism of this effect was not clear. To further characterize the effects of mAbp1 on podosome dots, we analyzed actin turnover at dot podosomes in control and mAbp1-deficient cells by live-cell imaging and FRAP (supplementary material Fig. S4A–D). In mAbp1-deficient cells, dot podosomes were longer-lived (supplementary material Fig. S4A,D); however, the actin dynamics were not significantly different between control and mAbp1-deficient cells (supplementary material Fig. S4B). Consistent with previous reports, live imaging analysis suggested that the podosome rosettes are formed by the maturation of pre-existing podosome dots (Kocher et al., 2009; Poincloux et al., 2005; Gavazzi et al., 1989), and this maturation was impaired in mAbp1-deficient cells (Fig. 2D,E and supplementary material Fig. S4E). Taken together, these findings suggest that the defect in formation of podosome rosettes in mAbp1-deficient cells might be due to impaired maturation rather than a change in the actin dynamics at dot podosomes.

mAbp1 is necessary for podosome rosette formation

To determine whether expression of mAbp1 is sufficient to rescue formation of podosome rosettes, we stably expressed mCherry–mAbp1 in mAbp1-deficient Src-transformed cells (Fig. 3). mCherry–mAbp1 expression was approximately fourfold higher than endogenous mAbp1 levels in control cells and, as previously reported, mAbp1 was susceptible to proteolysis (Fig. 3A) (Cortesio et al., 2010; Chen et al., 2001; Kessels et al., 2000). As expected, wild-type mCherry–mAbp1 localized to podosomes and was sufficient to rescue formation of podosome rosettes in mAbp1-deficient cells (Fig. 3B,C). Interestingly, ectopic expression of mCherry–mAbp1 also reduced the number of podosome dots formed, supporting the idea that mAbp1 regulates the maturation of podosome dots into rosettes (Fig. 3D). In addition, ectopic expression of mCherry–cortactin in cortactin-deficient cells rescued the formation of dot podosomes, but did not significantly affect the formation of podosome rosettes (supplementary material Fig. S5, and data not shown).

Role of mAbp1–WIP interaction in the formation of podosome rosettes

We had previously reported that PDGF induces the interaction between mAbp1 and WIP at dorsal ruffles and that this interaction is required for formation of dorsal ruffles (Cortesio et al., 2010). To determine whether the mAbp1–WIP interaction is important for podosome rosette formation, we expressed a WIP-binding mutant, mCherry–mAbp1(W415K), in mAbp1-deficient cells. In contrast to its effects on dorsal ruffle formation, we found that the mCherry–mAbp1(W415K) was sufficient to rescue formation of podosome rosettes in mAbp1-deficient cells (Fig. 3B–D). These findings suggest that the interaction between mAbp1 and WIP is not necessary for the formation of podosome rosettes. To determine whether mAbp1 interacts with WIP in Src-transformed cells, we transiently expressed FLAG–WIP in wild-type NIH3T3 cells in the presence or absence of PDGF stimulation and in Src-transformed cells. As expected, FLAG–WIP co-immunoprecipitated mAbp1 in PDGF-stimulated fibroblasts with a fivefold increase in the interaction; however, there was no significant increase in the association between mAbp1 and WIP in Src-transformed fibroblasts compared with the control (Fig. 3E). Taken together, these findings suggest that the mAbp1–WIP interaction is not necessary for the formation of podosome rosettes in Src-transformed cells.

Tyrosine phosphorylation of mAbp1 in Src-transformed fibroblasts

Both mAbp1 and cortactin are phosphorylated by Src in the proline-rich region (Larbolette et al., 1999; Kessels et al., 2000; Wu et al., 1991). The role of Src-mediated cortactin phosphorylation in invadopodia formation has been studied extensively (Luxenburg et al., 2006; Zhou et al., 2006; Tehrani et al., 2007; Mader et al., 2011). However, the functional significance of Src phosphorylation of mAbp1 is not known. To confirm that mAbp1 and cortactin are phosphorylated in Src-transformed fibroblasts, we pulled down phosphorylated proteins using a total phosphotyrosine antibody (4G10) and immunoblotted for endogenous mAbp1 or cortactin (Fig. 4A). We found that both cortactin and mAbp1 were phosphorylated in Src-transformed fibroblasts (Fig. 4B). Tyr337 and Tyr347 had previously been identified as the Src phosphorylation sites on mAbp1 (Larbolette et al., 1999; Lock et al., 1998). To determine whether Y337 and Y347 are phosphorylated in Src-transformed fibroblasts, we performed site-directed mutagenesis (Fig. 4C) to generate both phospho-mutant (Y337F and Y347F) and phospho-mimetic constructs (Y337E and Y347E). We found that the phospho-mutant constructs had significantly impaired phosphorylation compared with levels in wild-type mAbp1 in Src-transformed fibroblasts (Fig. 4D). Taken together, the findings suggest that mAbp1 is phosphorylated at Y337 and Y347 in Src-transformed fibroblasts.

Effects of Src-mediated phosphorylation of mAbp1 on the formation of podosome rosettes

To determine whether phosphorylation of mAbp1 is necessary for formation of podosome rosettes, we analyzed podosomes in mAbp1-deficient cells stably expressing the phospho-mimetic mCherry–mAbp1(Y340E,Y350E) or phospho-mutant mCherry–mAbp1(Y340F,Y350F) (Fig. 5A,B). Expression of wild-type mCherry–mAbp1 and mCherry–mAbp1(Y340E,Y350E) rescued podosome rosette formation in mAbp1-deficient cells. By contrast, the phospho-mutant mCherry–mAbp1(Y340E,Y350E) did not rescue podosome rosettes (Fig. 5C). In addition, dot podosome formation inversely correlated with the formation of podosome rosettes (Fig. 5C,D). To confirm that Src phosphorylation of mAbp1 is necessary for formation of podosome rosettes, we treated Src(Y527F)-transformed fibroblasts with the Src-inhibitor, PP2. As expected, the formation of podosome rosettes was impaired by PP2 treatment (supplementary material Fig. S6). These findings suggest that Src-mediated phosphorylation of mAbp1 is necessary for the formation of podosome rosettes.

To identify the specific phosphorylation site that mediates the formation of podosome rosettes, we examined the effects of expressing single phospho-mutants, mCherry–mAbp1(Y340F) or mCherry–mAbp1(Y350F), in mAbp1-deficient cells (Fig. 6A,B). Ectopic expression of the Y340F mutation rescued podosome rosettes, but the Y350F mutation was not sufficient to rescue podosome rosettes (Fig. 6C). However, surprisingly the increase in dot podosomes in mAbp1-deficient cells was not rescued by expression of mCherry–mAbp1(Y350F) (Fig. 6D), indicating that the single phospho-mutants (Y340F or Y350F) are not sufficient to regulate podosome dot formation. These findings suggest that Y350 is crucial for mAbp1 effects on the formation of podosome rosettes and that both tyrosine sites are probably necessary for the inhibitory effects of mAbp1 on the formation of podosome dots in Src-transformed cells.

Fig. 3.

mAbp1 and mAbp1(W415K), which is deficient in WIP binding, rescue podosome rosette formation in mAbp1-deficient cells. Src(Y527F)-transformed control or mAbp1-deficient cells were stably transfected with mCherry–mAbp1, mCherry–mAbp1(W415K) or control. (A) Representative western blot showing mCherry–mAbp1 overexpression (solid arrow) compared with endogenous expression of mAbp1 or degradation products of mCherry–mAbp1 (hollow arrow) in cell lysates. Blot was probed for mAbp1 and vinculin as a loading control. (B) Cells were plated on FN-coated coverslips for 3 hours, fixed, and stained with anti-cortactin antibody to visualize podosomes. mCherry–mAbp1 and mCherry–mAbp1(W415K) colocalized with cortactin at podosomes. Scale bar: 20 μm. (C) The percentage of cells producing podosome rosettes and (D) the average number of podosome dots per cell was quantified from at least 50 cells per condition; each bar represents the average value from three independent experiments. Error bars, s.e.m.; *P<0.05; ***P<0.001; one-way ANOVA with Tukey’s post hoc test. (E) Representative western blot of co-immunoprecipitation experiments with transient FLAG–WIP expression. NIH3T3 cells transiently transfected with FLAG–WIP were untreated (control) or serum-starved overnight and treated with 50 ng/ml PDGF for 5 minutes to stimulate dorsal ruffle formation (PDGF). In parallel, untreated SrcY527F-transformed fibroblasts were transiently transfected with FLAG–WIP and control cells were transiently transfected with FLAG alone. Cell lysates were incubated with anti-FLAG Sepharose beads, separated by SDS-PAGE and probed with anti-mAbp1 and anti-FLAG antibodies. Scale bar: 20 μm.

Effects of mAbp1 on degradation of the extracellular matrix

Both podosome dots and rosettes are capable of degrading the ECM (Mukhopadhyay et al., 2009; Linder, 2007; Seals et al., 2005). To determine whether mAbp1-deficient cells retain their capacity to degrade the ECM, we plated cells on coverslips coated with a mixed substrate of fibronectin and Alexa-Fluor-568-conjugated gelatin. In control cells, podosome rosettes were associated with robust degradation of the ECM. By contrast, mAbp1-deficient cells showed a significant reduction in matrix degradation as a result of the loss of podosome rosettes (Fig. 7A,B; red bars). Despite the increased numbers of podosome dots, mAbp1-deficient cells also had reduced matrix degradation at podosome dots (Fig. 7B; blue bars). However, when degradation area was analyzed as a percentage of individual dot podosome area, we found that there was no difference in the degradation efficiency of podosome dots in mAbp1-deficient cells (Fig. 7C). To determine whether the reduced total degradation area of podosome dots in mAbp1-deficient cells was due to a smaller size of the dots, we analyzed the average area of podosome dots in mAbp1-deficient cells compared with the control (Fig. 7D). Indeed, the average size of podosome dots was significantly reduced in mAbp1-deficient cells; supporting the idea that mAbp1 regulates the maturation of podosomes. These results suggest that the total area of degradation correlates with the total area of podosomes, and that the dot podosomes in mAbp1-deficient cells are smaller, but functional. Accordingly, we found that matrix degradation was rescued with expression of wild-type mAbp1 and that the effects of the phospho-mutants correlated with the numbers of podosome rosettes that were formed. Specifically, the Y337E,Y347E and Y337F mutants restored matrix degradation of mAbp1-deficient cells, whereas the Y337F,Y347F and Y347F mutants did not (Fig. 7E,F).

Fig. 4.

mAbp1 is phosphorylated in the proline-rich region in Src(Y527F)-transformed NIH3T3 fibroblasts. (A) Control and Src(Y527F)-transformed NIH3T3 cells were lysed and probed for phosphorylated proteins using anti-phosphotyrosine antibody (4G10). The right panels were probed with antibodies against mAbp1 and cortactin after immunoprecipitation with the total phospho-tyrosine antibody. (B) Parental NIH3T3 (‘None’) and Src-transformed siRNA knockdown cells were lysed for anti-phosphotyrosine antibody immunoprecipitation. Phosphorylated proteins were probed with anti-mAbp1 and anti-cortactin antibodies and with anti-vinculin for loading control. (C) Proline-rich regions of mAbp1 mutants. Arrows indicate Src phosphorylation sites, Tyr337 and Tyr347. (D) Src-transformed fibroblasts expressing mCherry–mAbp1 constructs were lysed and phosphotyrosine immunoprecipitation experiments were performed as described. Cell lysates were probed with anti-mAbp1 and anti-vinculin to show expression levels and loading, respectively.

Fig. 5.

mAbp1-mediated podosome rosette formation is dependent on tyrosine phosphorylation of the proline-rich region. (A) Cell lysates from control and mAbp1-knockdown cell lines expressing mCherry empty vector (EV) or rescue constructs were probed with anti-mAbp1 and anti-vinculin antibody to show mCherry–mAbp1 overexpression (solid arrow) compared with endogenous expression of mAbp1 or degradation products of mCherry–mAbp1 (hollow arrow) in cell lysates. (B) Control or mAbp1-deficient cells were stably transfected with mCherry vector alone, mCherry–mAbp1 phospho-mimetic (EE) or phospho-mutant (FF) were plated on FN-coated coverslips for 3 hours, fixed, and stained with anti-cortactin antibody to identify podosomes. (C) The percentage of cells producing podosome rosettes and (D) the average number of podosome dots per cell was quantified from >50 cells per condition. Each bar represents the average value from three independent experiments. Error bars, s.e.m.; *P<0.05; **P<0.01; ***P<0.001; one-way ANOVA with Tukey’s post hoc test. Scale bar: 20 μm.

Fig. 6.

Tyr347 of mAbp1 mediates podosome rosette formation. (A) Src(Y527F)-transformed control or mAbp1-deficient cells were stably transfected with mCherry vector alone or mCherry–mAbp1 single phospho-mutant constructs (Y337F or Y347F), lysed and probed with anti-mAbp1 and anti-vinculin antibody to show mCherry–mAbp1 overexpression (solid arrow) compared with endogenous expression of mAbp1 or degradation products of mCherry–mAbp1 (hollow arrow) in cell lysates. (B) Control and mAbp1-deficient cells expressing mCherry or rescue constructs were plated on FN-coated coverslips for 3 hours, fixed, and stained with anti-cortactin antibody. The percentage of cells producing podosome rosettes (C) and the average number of podosome dots per cell (D) was quantified from >50 cells per condition. Each condition represents the average value from three independent experiments. Error bars, s.e.m.; **P<0.01; ***P<0.001; one-way ANOVA with Tukey’s post hoc test. Scale bar: 20 μm.

Invasive migration of mAbp1-deficient cells

Formation of podosomes is generally correlated with invasive migration (Mukhopadhyay et al., 2009; Linder, 2007; Seals et al., 2005). However, the relationship between podosome rosettes and invasion has not been defined in Src-transformed cells. To determine whether mAbp1 regulates invasive cell migration, we tested the effects of mAbp1 depletion on invasion through Matrigel-coated membranes (Fig. 8). Interestingly, mAbp1-deficient cells demonstrated increased invasion through Matrigel-coated membranes compared with control cells. Moreover, ectopic expression of wild-type mAbp1 or the Y337E,Y347E or Y337F mutants, but not the phosphomutants Y337F,Y347F and Y347F, were able to rescue invasion to control levels (Fig. 8C,D). These findings suggest that formation of podosome rosettes might be inhibitory to invasive cell migration and also identifies mAbp1 as a negative regulator of invasive migration of Src-transformed fibroblasts (Fig. 9).

Discussion

In this study, we identify a role for the actin binding protein mAbp1 in the formation of podosome rosettes and regulation of invasive cell migration. Substantial evidence supports an important role for cortactin and its phosphorylation downstream of Src kinases during invasive cell migration and formation of degradative structures such as podosomes (Webb et al., 2006; Ayala et al., 2008; Clark and Weaver, 2008; Oser et al., 2009; Bañón-Rodríguez et al., 2011). mAbp1 shares significant structural homology with cortactin; however, its role in podosome formation and invasive migration had not been previously defined. Here, we showed that both mAbp1 and cortactin colocalize to podosomes in Src-transformed fibroblasts but play distinct roles in regulating podosome formation and invasion. Cortactin mediates the formation of podosome dots but not rosettes (Fig. 2) and is known to mediate invasive migration (Weaver, 2008). By contrast, mAbp1 is necessary for the formation of podosome rosettes and impairs invasive cell migration. The identification of a specific phosphorylation site on mAbp1 that mediates podosome regulation and invasion highlights the importance of mAbp1 downstream of Src signaling.

Fig. 7.

Effect of mAbp1 on matrix degradation. (A) Control or mAbp1-deficient Src-transformed fibroblasts were plated on Alexa-Fluor-488-conjugated gelatin and FN-coated coverslips for 6 hours, fixed, and stained with anti-cortactin antibody to visualize podosomes. Degradation areas align with cortactin-positive podosomes. (B) Degradation area of Src-transformed control and mAbp1-deficient cells was quantified as a percentage of total cell area. Bars represent average total cell area degradation of at least 50 cells per condition. Percentage degradation mediated by podosome rosettes is represented in red and podosome dots are represented in blue. (C) Analysis of the degradation area of individual podosome dots as a percentage of podosome dot area in Src-transformed control and mAbp1-deficient cells. Podosome dot area was determined by area of cortactin staining. A minimum of 70 dots were analyzed per condition. (D) Analysis of average podosome dot area in control and mAbp1-knockdown cells. (E) Analysis of total cell area degradation (podosome dots and rosettes) as a percentage of total cell area in control and mAbp1-deficient cells with stable mCherry empty vector or mCherry–mAbp1 rescue expression. Error bars represent s.e.m. of three independent experiments. *P<0.02; **P<0.01; two-tailed t-test. (F) Control and mAbp1-deficient cells expressing mCherry alone or mCherry–mAbp1 rescue constructs were plated on Alexa-Fluor-568-conjugated gelatin and FN-coated coverslips for 6 hours, fixed and stained with anti-cortactin antibody to visualize podosomes. Scale bars: 20 μm.

Podosomes are complex structures that contain many adaptor and actin regulatory proteins. Cortactin activates and stabilizes actin polymerization through its interactions with Arp2/3, N-Wasp, and WIP and has been implicated in the formation of podosomes and invadopodia (Webb et al., 2006; Ayala et al., 2008; Clark and Weaver, 2008; Oser et al., 2009; Bañón-Rodríguez et al., 2011). A WIP-cortactin interaction is required for podosome formation and ECM degradation by dendritic cells (Bañón-Rodríguez et al., 2011). The interaction between mAbp1 and WIP is also required for dorsal ruffle formation downstream of PDGF (Cortesio et al., 2010). Here we show that mAbp1 effects on podosome architecture do not require binding to WIP. In contrast to the effects of PDGF stimulation, we did not observe increased association between WIP and mAbp1 in Src-transformed cells. These results suggest that mAbp1 regulates podosome rosette formation through a mechanism that is independent of WIP. Previous work has shown that mAbp1 does not affect actin polymerization at lamellipodia (Kessels et al., 2000). Accordingly, we found that depletion of mAbp1 had minimal effects on actin dynamics at podosome dots (supplementary material Fig. S2). Moreover, in control cells, podosome dots matured into podosome rosettes; however, this maturation was impaired in mAbp1-deficient cells (supplementary material Fig. S2 and data not shown). Our findings suggest that mAbp1 might serve as a crucial link that regulates the maturation of podosome dots into podosome rosettes that may function to limit cell invasion.

Fig. 8.

mAbp1 inhibits invasive cell migration. (A) Podosome rosettes form on membrane surface of matrigel invasion chamber and within pores. Src-transformed fibroblasts were cultured in matrigel invasion chamber for 6 hours, fixed, and stained with Rhodamine phalloidin. Membranes autofluoresce blue. (B) Confocal Z-stack images of Src-transformed control and mAbp1-knockdown fibroblasts cultured on matrigel invasion chambers. A representative control cell with a podosome rosette is shown on the left and representative mAbp1-deficient cell with podosome dots is shown on the right. Z-stack lateral view is shown in bottom panel. (C) Src-transformed control and mAbp1-deficient cells were cultured in matrigel invasion chambers for 24 hours, fixed, and stained with a HEMA staining kit. The number of invading cells was averaged from six 10× field-of-view images. Error bars represent s.e.m. for three independent experiments. **P<0.01; two-tailed t-test. Control and mAbp1-knockdown cells were cultured in matrigel invasion chambers for 18 hours, fixed and stained with HEMA for counting. (D) Quantification of the average number of cells which invaded on six 20× field-of-view images. Error bars represent s.e.m. of three independent experiments. **P<0.01; two-tailed t-test. (E) Representative images of stained cells that invade through matrigel invasion chambers for each condition. Scale bars: 10 μm.

Although WIP binding was not essential for podosome rosette formation, our work reveals that Src-phosphorylation of mAbp1 is crucial for its function. We show that two previously identified Src-phospho-acceptor sites (Larbolette et al., 1999; Lock et al., 1998) are important for mAbp1 regulation of podosome architecture and we identified an essential role for a specific tyrosine residue Y350 in the proline-rich region for podosome regulation and invasion. It is not known whether phosphorylation at these sites affects the binding of key interacting proteins that are crucial for podosome regulation. To our knowledge, no specific interacting proteins have been identified that bind to these residues. One attractive hypothesis is that mAbp1 regulates podosomes through its interaction with Rho GTPase regulatory proteins such as Fgd-1, a Cdc42 exchange factor (Hou et al., 2003). Expression of constitutively active Cdc42 (V12Cdc42) induces podosome formation and stimulates epithelial to mesenchymal transition of invasive cells (Bakin et al., 2000; Moreau et al., 2003). Moreover, active Rho localizes to podosome rosettes and regulates their formation (Berdeaux et al., 2004). Further work will be required to identify key interacting partners that mediate the effects of mAbp1 on podosomes.

An intriguing aspect of this study is the suggestion that formation of podosome rosettes may be inhibitory to invasion of Src-transformed cells. Cell invasion is dependent on the ability of cells to rapidly form and turnover adhesions and to efficiently degrade the matrix. Previous studies have suggested an inverse correlation between the formation of focal adhesions and cell migration (Cortesio et al., 2011; Barker et al., 2004; Huttenlocher et al., 1996; Huttenlocher et al., 1995). Podosome rosettes have been reported to have similar adhesive functions as focal adhesions, and mediate transduction of force, in contrast to invadopodia or dot-like podosomes (Collin et al., 2008), suggesting that formation of podosome rosettes might be inhibitory to cell invasion through their effects on adhesion. However, podosome rosettes in Src-transformed cells are associated with robust matrix degradation (Fig. 7), which is generally correlated with increased cell invasiveness (Mukhopadhyay et al., 2009; Linder, 2007; Seals et al., 2005). Depletion of mAbp1 impaired matrix degradation, suggesting that in some cases, increased matrix degradation might not be correlated with cell invasion. In support of this idea, FAK-deficient cancer cells have enhanced matrix degradation but reduced invasion because of impaired turnover of focal adhesions (Chan et al., 2009).

Fig. 9.

Model of mAbp1 regulation of podosome formation and invasive migration. mAbp1 binds to filamentous actin and is phosphorylated by Src at Y337 and Y347 in the proline-rich region (oval). Phosphorylation of mAbp1 at Y347 is required for podosome rosette formation and the inhibition of invasive migration. Y347 might be the crucial residue that mediates maturation of podosome dots to rosettes. Conversely, Src phosphorylation of both Y337 and Y347 impairs formation of podosome dots.

Cortactin expression is increased in many cancers and has been implicated in invasion downstream of Src kinase activation (Mader et al., 2011; Kelley et al., 2010; Oser et al., 2009). In this report, we have shown that mAbp1, in contrast to cortactin, is a negative regulator of cell invasion. Moreover, we have identified a specific tyrosine residue, Tyr350, in mAbp1 that functions downstream of Src to inhibit invasion of Src-transformed cells. These findings suggest that cortactin and mAbp1 play opposing roles in regulating cell invasion downstream of Src. These findings raise the intriguing possibility that Src might differentially regulate invasion depending on the substrate that is being targeted – Src-mediated phosphorylation of cortactin might promote invasion, whereas Src-mediated phosphorylation of mAbp1 might impair cell invasion. More specifically, the findings suggest that Src-mediated formation of podosome dots through cortactin may be pro-invasive, whereas Src-mediated formation of podosome rosettes through mAbp1 might function to limit cell invasion. A challenge for future investigation will be to identify the key effector pathways that regulate Src targeting of specific pathways and the switch between the formation of podosome dots and rosettes that modulate invasive cell migration.

Materials and Methods

Reagents and antibodies

Ham’s F12 medium, Opti-MEM, Alexa Fluor 350 phalloidin and Rhodamine phalloidin were purchased from Invitrogen (Carlsbad, CA). Fibronectin was purified as described previously (Ruoslahti et al., 1982). PDGF-BB was purchased from R&D Systems (Minneapolis, MN). Alexa Fluor 568 protein labeling kit (Invitrogen) and gelatin from porcine skin (Sigma) were used for gelatin degradation assays. The following mouse monoclonal antibodies were used: anti-vinculin (clone h-VIN1; Sigma, St Louis, MO), anti-cortactin (clone 4F11; Upstate Biotechnology, Lake Placid, NY), anti-Src (clone GD11), anti-paxillin (clone 349), and anti-phosphotyrosine (4G10; Millipore). The following rabbit polyclonal antibodies were used: anti-vinculin (Sigma), anti-Tks5/FISH (M300; Santa Cruz), anti-Arc32 (Millipore; Temecula, CA), and anti-mAbp1 from the University of Wisconsin Antibody facility as described previously (Cortesio et al., 2008). Secondary antibodies used were Marina blue goat anti-mouse IgG (Invitrogen), Rhodamine red-X goat anti-mouse IgG, TRITC donkey anti-rabbit IgG (Jackson ImmunoResearch Laboratories), and FITC donkey anti-rabbit IgG. Immunoprecipitation control IgGs used were mouse IgG (whole molecule; Chroma Pure) and rabbit IgG (Jackson ImmunoResearch Laboratories). Glutathione–Sepharose was purchased from Amersham Biosciences (Piscataway, NJ), PP2 and anti-FLAG M2 agarose was purchased from Sigma.

Cell lines and transfections

NIH3T3 cells were obtained from ATCC (Manassas, VA) and cultured in DMEM (Manassas, VA) supplemented with 10% BGS (Hyclone), non-essential amino acids (Sigma), and antibiotics (Cellgro; Mediatech) according to the manufacturer’s instructions. Stable siRNA knockdown cell lines have been previously described (Cortesio et al., 2010). Src transformation of NIH3T3 cells was performed with pMX-c-SrcY527F-Ires-GFP plasmid (Chan et al., 2009) using retroviral transfection by methods previously described (Franco et al., 2004). Cells positive for Src(Y527F) were identified by GFP expression. Generation of stable rescue lines was performed by subsequent retrovirus infection using empty vector pMX-mCherry-C1, pMX-mCherry-C1-Cortactin or pMX-mCherry-C1-mAbp1 constructs. Transient transfections were performed with Lipofectamine 2000 (Invitrogen) according to the manufacturer’s instructions.

DNA and siRNA constructs

Production of stable NIH3T3 knockdown lines using the pSUPER.retro RNAi system (OligoEngine, Seattle, WA) has been previously described (Cortesio et al., 2010). The generation of Src(Y527F) has also been described (Chan et al., 2009). mAbp1 was PCR amplified from GFP–mAbp1 and GFP–mAbp1(W415K) (Cortesio et al., 2010) using a forward primer containing an EcoRI site, 5′-GTC ACG GAA TTC GAT GGC GGT GAA CCT GAG CCG GAA C-3′ and BamHI-containing reverse primer, 5′-CGA TCG CGG ATC CTC ACT CTA TGA GCT CCA CGT AGT TGG CAG G-3′. PCR-amplified mAbp1 and mAbp1(W415K) was cloned into the EcoRI and BamHI sites of pmCherry-C1 (Lokuta et al., 2007). mCherry-C1–mAbp1 and mCherry-C1–Abp(W415K) were subcloned into pMX-Ires-GFP vector (Clive Svendsen, University of Wisconsin, Madison, WI) with the Ires–GFP portion excised. mCherry–mAbp1 was PCR amplified using a primer containing a BamHI site, 5′-GAC TTA GGA TCC TAT GGT GAG CAA GGG CGA GGA GGA TAA-3′ and reverse primer containing a SalI site, 5′-ATA GCC TGT AGG CAT TAC GTC GAA GTC GAC TCA CTC TAT GAG CTC CAC GTA GTT GGC AGG-3′. Similarly, pMX-mCherry-Cortactin was generated by PCR amplifying mouse cortactin from GFP–cortactin with a forward primer containing a XhoI site, 5′-AG GCC GCT CGA GGC ATG TGG AAA GCC TCT GCA GGC CAT GCT GTG TCC ATC-3′ and a reverse primer with a BamHI site, 5′-CTC CGC GGA TCC GCG CTA CTG CCG CAG CTC CAC ATA GTT GGC TGG GAA GAG-3′ and ligating into the pmCherry-C1 vector. Subsequently, mCherry–Cortactin was PCR amplified with a forward containing a BamHI site, 5′-GAC TTA GGA TCC TAA CCT ATG GTG AGC AAG GGC GAG GAG GAT AAC ATG GCC ATC ATC-3′ and a reverse primer with a SalI site, 5′-TTC GAC GTA ATC CCT ACA GGC TAT GTC GAC CTA CTG CCG CAG CTC CAC ATA GTT GGC TGG-3′ and cloned into pMX-Ires-GFP vector with the Ires–GFP portion excised.

Phospho-mutant and -mimetic PMX-mCherry-mAbp1 were generated with a Quikchange site-directed mutagenesis kit (Agilent Technologies) with the following primers: Y337E,Y347E (forward) 5′-GGA GCC AAC AGA AGA AGT ACC CCC AGA GCA GGA CAC CCT CGA AGA GGA ACC AC-3′ and (reverse) 5′-GTG GTT CCT CTT CGA GGG TGT CCT GCT CTG GGG GTA CTT CTT CTG TTG GCT CC-3′; Y337F,Y347F (forward) 5′-GGA GCC AAC ATT TGA AGT ACC CCC AGA GCA GGA CAC CCT CTT TGA GGA ACC AC-3′ and (reverse) 5′-GTG GTT CCT CAA AGA GGG TGT CCT GCT CTG GGG GTA CTT CAA ATG TTG GCT CC-3′; Y337F (forward) 5′-GGA GCC AAC ATT TGA AGT ACC CCC AGA GCA GGA CAC CCT CTA TGA GGA ACC AC-3′ and (reverse) 5′-GTG GTT CCT CAT AGA GGG TGT CCT GCT CTG GGG GTA CTT CAA ATG TTG GCT CC-3′; Y347F (forward) 5′-GGA GCC AAC ATA TGA AGT ACC CCC AGA GCA GGA CAC CCT CTT TGA GGA ACC AC-3′ and (reverse) 5′-GTG GTT CCT CAA AGA GGG TGT CCT GCT CTG GGG GTA CTT CAT ATG TTG GCT CC-3′.

Immunoblotting and immunoprecipitation

Cells were scraped into lysis buffer [20 mM HEPES, 50 mM KCl, 1 mM EDTA, 1% NP40, 0.2 mM phenylmethylsulfonyl fluoride (PMSF), 1 μg/ml pepstatin, 2 μg/ml aprotinin, and 1 μg/ml leupeptin] on ice, and clarified by centrifugation. Protein concentrations were determined with a bicinchoninic acid protein assay kit (Thermo Fisher Scientific) according to the manufacturer’s instructions. Equal protein levels were denatured in SDS and loaded on 4–20% SDS polyacrylamide gradient gels. Separated proteins were transferred to nitrocellulose, probed with the appropriate antibodies and imaged with an Odyssey Infrared Imaging System (Li-Cor Biosciences, Lincoln, NE). For phospho-tyrosine immunoprecipitation experiments, cells were cultured to approximately 80% confluency on 10 cm plates, washed once with PBS, and lysed in lysis buffer. Lysates were clarified by centrifugation and incubated with 5 μg mouse-anti-phospho-tyrosine antibody (4G10) or 5 μg mouse IgG. Immune complexes were captured on Gammabind G-Sepharose beads (GE-Healthcare), washed three times in lysis buffer and analyzed by immunoblotting. The FLAG–WIP immunoprecipitation experiments were performed as described previously (Cortesio et al., 2010), with a few modifications; in brief, mouse fibroblasts were transiently transfected with FLAG–WIP or FLAG alone, serum-starved, and treated with PDGF to stimulate dorsal ruffle formation and detect a positive interaction between mAbp1–FLAG and WIP. In parallel, we transiently transfected FLAG–WIP into SrcY527F-transformed fibroblasts and performed co-immunoprecipitations with anti-FLAG Sepharose beads.

Immunofluorescence

Coverslips were acid-washed, ethanol sterilized, coated with gelatin as previously described (Artym et al., 2006), and were subsequently coated with 10 μg/ml fibronectin overnight at 4°C. 3.0×104 cells were plated in DMEM medium on coverslips for 3 hours at 37°C and 5% CO2. Cells were fixed with 3.7% formaldehyde for 10 minutes, quenched with 0.15 M glycine for 10 minutes, permeabilized with 0.2% Triton X-100 for 10 minutes, washed with PBS three times, and blocked with 5% goat serum for 1 hour or overnight at 4°C. Cells were incubated with primary antibody and/or Rhodamine phalloidin for 1 hour, washed, then incubated with fluorophore-conjugated secondary antibodies for 1 hour. Images were captured with a Coolsnap fx cooled CCD camera (Photometrics, Huntington Beach, CA) on an Olympus 1X-70 inverted microscope (Olympus America, Melville, NY) using a 60×/1.4 oil-immersion objective. Images were processed with Metavue imaging software v6.2 (Universal Imaging, Downingtown, PA). Podosomes were quantified from cells stably expressing GFP–Src(Y527F). Podosome dots were quantified as average numbers formed per cell because over 95% of the transformed cells produced podosome dots but the numbers of dots formed per cell changed under different conditions. Podosome rosettes were quantified as the percentage of cells forming rosettes because fewer cells formed rosettes and the numbers formed ranged from 0–2 for most cells. Owing to the greater fluorescent signal of podosome rosettes, it is possible that podosome dots have been underestimated in some cells given the signal-to-noise ratio. Confocal images were acquired with a Fluoview, FV-1000 laser scanning confocal microscope (Olympus) using a 60× Plan Apo/1.45 oil-immersion objective with a 2–3 zoom factor on Fluoview software (FV10-ASW version 01.07; Olympus).

Fluorescence gelatin degradation assay

Alexa-Fluor-568-conjugated gelatin was prepared using the Alexa Fluor 568 protein labeling kit according to the manufacturer’s instructions (Invitrogen), and gelatin-coated coverslips were prepared as described previously (Artym et al., 2006). Fluorescent gelatin coverslips were coated with 10 μg/ml fibronectin overnight at 4°C. 3.0×104 cells were cultured on coverslips for 6 hours, fixed with 3.7% paraformaldehyde, and stained with anti-cortactin antibody. 40–60 GFP-positive cells (expressing Src) were imaged, the cell body was traced, and total area of degradation was analyzed as a percentage of the cell area covered using Metamorph Imaging Software.

Invasion assays

Cell invasion experiments were performed using 8μm porous chambers coated with Matrigel (BD), and cells were allowed to invade for 24 hours as previously described (Chan et al., 2009). Cells were removed from top the chamber and cells remaining on the bottom were stained with a Hema-3 stain kit (Thermo Fisher Scientific). Cells were counted from 6–12 representative 20× or 40× fields per condition from at least three independent experiments.

Time-lapse fluorescence microscopy

Live imaging of podosome dynamics was performed on an Olympus 1X-70 inverted microscope using a 60×/1.4 oil-immersion objective and was housed in a closed system to maintain temperature at 37°C. 24 hours before imaging, cells were transfected with RFP-actin using Lipofectamine 2000. Before imaging, 2.0×105 cells were plated on glass-bottom dishes (35 mm) pre-coated with 10 μg/ml fibronectin, and cells were allowed to adhere for 1 hour in DMEM complete medium. DMEM media was replaced with imaging media (Ham’s F12, 10% BGS, 25 mM HEPES) and topped with mineral oil for image acquisition. 6–8 cells were imaged simultaneously per day of experiments, and images were captured at 2 minute intervals over a 3 hour period using the Metavue v6.2 imaging software.

Fluorescence recovery after photobleaching (FRAP)

FRAP analysis of actin dynamics was performed with a Fluoview, FV-1000 laser scanning confocal microscope using a 60× Plan Apo/1.45 oil-immersion objective and a 3× zoom factor on Fluoview software. Cells were transfected with RFP-actin 24 hours before imaging, and 2×105 cells were plated on glass-bottom dishes (35 mm) in DMEM complete medium and allowed to adhere 1 hour before imaging. DMEM medium was replaced with imaging medium and topped with mineral oil before imaging. The confocal microscope was housed in a closed unit to maintain temperature at 37°C. One podosome dot per cell was selected for photobleaching and >10 podosome dots were analyzed per cell line per day. Photobleaching was performed with the 543 laser set at 100% intensity for 800 mseconds. 10 frames were imaged pre-bleach and fluorescence recovery was measured for 90 frames post-bleach with continuous scanning. Only podosome dots that were bleached to 30–50% their original intensity were used for analysis and background actin intensity levels were subtracted from each frame to account for general photo-bleaching over time.

Statistical analyses

The two-tailed t-test or one-way analysis of variance (ANOVA) test was used and P<0.05 was considered significant.

Acknowledgements

We thank members of the Huttenlocher lab for critical review of the manuscript.

Footnotes

  • Funding

    This work was supported by the National Institutes of Health [grant number R01 CA085862-08] to A.H and by the Department of Defense predoctoral fellowship [grant number BC073295] to C.C. Deposited in PMC for release after 12 months.

  • Supplementary material available online at http://jcs.biologists.org/lookup/suppl/doi:10.1242/jcs.096529/-/DC1

  • Accepted October 19, 2011.

References

View Abstract