Faciogenital Dysplasia Protein Fgd1 Regulates Invadopodia Biogenesis and Extracellular Matrix Degradation and Is Up-regulated in Prostate and Breast Cancer

Invasive tumoral or transformed cells grown on an extracellular matrix (ECM) substratum, form proteolytically active protrusions termed invadopodia that produce the focal degradation of the ECM. In recent years, many molecular players have been defined (reviewed in refs. 1, 2); among these, the Rho GTPase Cdc42 is considered a master regulator of the core actin polymerization machinery at invadopodia (3).

Cdc42, together with Rho and Rac, belongs to the Rho GTPase family that mediates the regulation of the actin cytoskeleton in response to extracellular signals and has been implicated in the control of cell polarity, motility, gene expression, mitosis, trafficking, transformation, and metastasis (reviewed in ref. 4). Rho GTPases cycle between inactive GDP-bound and active GTP-bound states under the control of three classes of accessory proteins: guanine nucleotide exchange factors (GEF), GTPase-activating proteins (GAP), and guanine nucleotide dissociation inhibitors. It is thought that the GEFs are critical mediators of Rho GTPase activation (5).

The FGD1 gene encodes the Cdc42-specific GEF Fgd1 (6). Expression-abrogating FGD1 mutations cause faciogenital dysplasia, an X-linked developmental disorder that affects skeletal elements resulting in facial, skeletal, and urogenital anomalies and in some patients, mental retardation (7). Like most Rho GEFs, Fgd1 contains a Dbl homology domain adjacent to a pleckstrin homology (PH) domain that catalyzes the exchange of GDP for GTP on Cdc42, an NH₂-terminal proline-rich domain (PRD), a cysteine-rich zinc-finger FYVE domain, and a second COOH-terminal PH domain. The PRD negatively regulates enzymatic activity and contains two putative Src-homology 3 (SH3)-binding domains, whose known interactors are cortactin and actin-binding protein 1 (8). Interestingly, direct binding of Fgd1 to cortactin stimulates actin nucleation, independently of its exchange activity (9).

Here, we show that Fgd1 is transiently associated with invadopodia and required for their formation and function in ECM degradation and acts by direct modulation of Cdc42 activation. We also found that in vivo, Fgd1 is expressed in human prostate and breast cancer as opposed to normal tissue and that expression levels correlated with tumor aggressiveness.

Reagents and basic procedures. Fgd1 constructs were provided by J. L. Gorski (University of Michigan Medical School, Ann Arbor, MI). Anti-Fgd1 antibodies were raised against a COOH-terminal fragment (amino acids 732-961). Anti–dynamin 2 was provided by M. McNiven (Mayo Clinic, Rochester, MN). Antibodies against Cdc42 and phosphotyrosine (Santa Cruz Biotechnology), cortactin and Rac1 (Upstate Biotechnology), α-tubulin (Sigma-Aldrich), and glyceraldehyde-3-phosphate dehydrogenase (GAPDH; Biogenesis) were also used. Secramine A was provided by T. Kirchhausen (Harvard Medical School, Boston, MA). The Smart pool reagent for Fgd1 was from Dharmacon.

Cell culture. Human melanoma A375MM cells were cultured as previously described (10, 11). Human breast carcinoma MDA-MB-231 and prostate cancer PC3 cells were maintained with DMEM and RPMI, respectively, supplemented with 10% FCS. Cells were grown at 37°C in a humidified atmosphere containing 5% CO₂. Transfection, invadopodia synchronization, ECM degradation assays, and RNAi were performed as described (10). Quantification of GTP-Cdc42/GTP-Rac was carried out with the PAK1-binding assay as previously described (12).

Microscopy. Immunofluorescence studies were carried out as previously described (10). For live videomicroscopy experiments, A375MM cells expressing GFP-WT-Fgd1, were plated with BB94 on TRITC-gelatin–coated glass-bottomed 35-mm dishes (MatTek Corporation). After an overnight incubation, BB94 was washed out and cells observed at different times. Finally, cells were fixed and labeled with Alexa 633-phalloidin. Experiments were performed with a T.I.L.L. Photonics setup (T.I.L.L. Photonics GmbH) on an Olympus IX70 microscope.

Immunohistochemistry. Samples from 42 human prostate cancers and 30 human breast ductal infiltrating adenocarcinomas were provided by J. Boniver (Department of Pathology, University Hospital of Liège, Liège, Belgium). Specimens were fixed in formalin, embedded in paraffin, and cut into 5-μm sections. Immunoperoxidase staining was performed using anti-Fgd1 and ABC Vectastain kit (Vector Laboratories). Sections were subclassified for staining intensity and extent scores according to arbitrary scales that ranged from 0 to 3: 0 stands for negative staining; 1+, 2+, and 3+ for weak, intermediate, or moderate and strong staining, respectively. When tumors showed heterogeneous intensity, scoring was based on the staining of the most positive tumor cells, when their estimated percentage represented at least 30% of the total positive tumor cell area. The lesions were evaluated by expert pathologists (Department of Pathology, University Hospital of Liège, Liège, Belgium). The immunohistochemically stained sections were reviewed in a blind fashion and scored by two independent observers.

Fgd1 is a transient component of invadopodia. Invadopodia are recognized by the colocalization of actin, cortactin, dynamin 2, phosphotyrosine, and other proteins, with areas of matrix degradation (Fig. 1A; refs. 11, 13). Fgd1-GFP, when expressed in A375MM cells, can be found at the perinuclear area, cytosol, plasma membrane, and nucleus, in accord with its reported distribution profile (8, 14, 15). Notably, invadopodia presented obvious Fgd1-GFP accumulation (Fig. 1B).

Next, we investigated whether A375MM cells endogenously expressed Fgd1. First, by RT-PCR from purified poly(A)-RNA, we found that Fgd1 mRNA was present in A375MM cells (data not shown). Next, immunofluorescence studies showed that endogenous Fgd1 colocalized with actin at some but not all invadopodia (Supplementary Figs. S1 and S2). Hence, in A375MM cells Fgd1 is expressed and present at invadopodia, perhaps transiently. To investigate this, we plated A375MM cells directly on fluorophore-conjugated gelatin (hereafter called “gelatin”) in the presence of BB94, a synthetic competitive and reversible broad-spectrum metalloprotease inhibitor that functions by fitting in the active site of the enzyme (16). In these conditions, invadopodia do not form and thus cells do not degrade the matrix (10). After BB94 washout to allow for invadopodia formation, cells were either directly fixed or live-imaged at different time points. Fixed cells were labeled with anti-Fgd1 and phalloidin. For live imaging experiments, cells were fixed and labeled with phalloidin at the end of the experiment.

Fifteen minutes after BB94 washout, we observed local accumulations of Fgd1, often with, but sometimes without actin and not associated with matrix degradation, consistent with nascent invadopodia. At longer times, when ECM degradation became detectable, not all actin-positive invadopodia at sites of degradation contained Fgd1 (Supplementary Fig. S1). Similar results were obtained by labeling cortactin in lieu of actin (data not shown). In live imaging experiments, at 60 minutes, we typically observe Fgd1 puncta without underlying degradation (Supplementary Fig. S2, red arrow). These might be nascent invadopodia. In the example shown, at 100 minutes, the very same spot acquires filamentous (F)-actin and features underlying degradation, thus is a fully formed invadopodium. At 60 minutes, one can also observe Fgd1/F-actin–positive invadopodia (Supplementary Fig. S2, blue arrows) with underlying degradation from which Fgd1, but not F-actin, disappears at 100 minutes. Finally, one can also observe the appearance at 100 minutes of new Fgd1-positive invadopodia (red arrowheads). The simultaneous presence of actin and ECM degradation define them as active “early” invadopodia.

In summary, Fgd1-positive, F-actin–negative puncta with no underlying matrix degradation (51% of the total) observed at earlier times might be nascent invadopodia where Fgd1 appears ahead of actin. At later times, structures where Fgd1 colocalizes with F-actin and underlying degradation are active early invadopodia (68% of the total number). Whereas Fgd1-negative, F-actin–positive invadopodia with underlying degradation (the remaining 32%) would be fully active “late” invadopodia, wherefrom Fgd1 has moved away. These observations, obtained by scoring 113 cells over 26 fields in 2 independent experiments, are in accord with the proposed stepwise mechanism for invadopodia biogenesis (17).

Fgd1 regulates invadopodia formation and function. To verify its role, we depleted endogenous Fgd1 in A375MM cells by RNAi. Cells were cultured on gelatin in the presence of BB94 and transfected 24 hours later with a siRNA pool specific for human Fgd1. After further 48 hours, BB94 was washed out to allow invadopodia formation and ECM degradation for 3 hours. Finally, cells were fixed and labeled. Fgd1 was significantly depleted (80%) in siRNA-treated cells compared with mock (Fig. 2A). Fgd1-depleted cells exhibited very few invadopodia and degraded 80% less than control cells (Fig. 3A).

Next, we used two GFP-tagged Fgd1 mutants. One (2DBDEL) lacks the FYVE, COOH-terminal PH and SH3-binding domains, and the PRD residues that negatively regulate exchange activity, thus behaving as a constitutively active GEF. The other (AS) is a naturally occurring splice variant without GEF activity, therefore constitutively inactive (Fig. 2B; ref. 18). Cells were transfected, plated on gelatin, and fixed after an overnight incubation. Wild-type (WT) and constitutively active Fgd1-expressing cells were able to degrade twice to thrice more than control cells. In contrast, constitutively inactive Fgd1-expressing cells degraded 60% less than control (Fig. 3B and C).

To exclude off-target effects, we transfected Fgd1-depleted cells with murine WT, dominant-positive, and negative forms of Fgd1 (Fig. 3D). We found that exogenous expression of WT and dominant positive forms partially reverted (50–60%) to the invasive phenotype when compared with control cells. In contrast, Fgd1-depleted cells, when transfected with the dominant-negative form, decreased drastically their ability to degrade the matrix.

Fgd1 behaved similarly in other cancer cells, as Fgd1 knockdown produced a 70% and 60% decrease in the ability of breast carcinoma MDA-MB-231 and prostate cancer PC3 cells to degrade the matrix, respectively (Fig. 2C and D).

Fgd1 regulates invadopodia formation and function through the activation of Cdc42. Cdc42 is essential for invadopodia formation in rat MTLn3 mammary adenocarcinoma cells as shown by RNAi (3). We obtained similar findings in A375MM cells (data not shown). To further confirm this, we used the novel Cdc42 inhibitor secramine A, a reversible and fast-acting compound that inhibits Cdc42 activation by specifically preventing GTP loading (19). We observed a 60% reduction in the ability of the cells to form invadopodia and degrade the ECM when compared with vehicle-treated cells (Fig. 4A). To test whether Rho and Rac played a role in invadopodia biogenesis in A375MM cells, we also transiently transfected constitutively active and inactive forms and found no significant effects on invadopodia biogenesis or ECM degradation (data not shown). These results confirm the role of Cdc42 as a regulator of invadopodia biogenesis, whereas excluding, at least in this model, a role for other Rho GTPases.

Wide variations have been found in the specificity of individual GEFs for Rho GTPases. Some act on a broader range of targets, whereas few have activity toward a single GTPase, such as Fgd1. As discrepancies exist between in vitro and in vivo specificities of some GEFs (reviewed in ref. 5), we sought to resolve this issue by assaying activated Cdc42 in control conditions and after Fgd1 knockdown. We found that in Fgd1-depleted cells, the level of GTP-Cdc42 was 40% lower when compared with control cells, whereas GTP-Rac1 was below detection regardless of Fgd1 expression (Fig. 4B). Hence, invadopodia-dependent ECM degradation is regulated by Fgd1-mediated activation of Cdc42.

Fgd1 may have Cdc42-independent functions in some signaling pathways (14) and actin remodeling (8, 9). Our results, however, imply that Fgd1 functions at invadopodia through its exchange activity toward Cdc42, as supported by two lines of evidence: (a) the constitutively active Fgd1 mutant that increases invadopodia formation and ECM degradation, lacks the two SH3-binding domains required for cortactin interaction; (b) Fgd1 knockdown inhibits Cdc42 activation, and this per se would impair invadopodia formation. We conclude that Fgd1 is a master regulator of Cdc42 in invadopodia biogenesis, although we cannot yet exclude the contribution of other GEFs. Considering the above and previous findings (10), we suggest a model whereby a signaling cascade triggered by integrin activation activates Fgd1 (perhaps by Src-family kinases) to stimulate Cdc42 that in turn activates its downstream targets among which cortactin and the actin polymerization regulator N-WASP.

Fgd1 is highly expressed in human prostate and breast cancers. Many GEFs were originally identified as oncogenes after transfection of immortalized fibroblast cell lines (reviewed in ref. 20). Because the functional effect of GEF overexpression is to elevate cellular levels of activated Rho GTPases, deregulated GEF expression might lead to aberrant growth, invasiveness, and/or metastatic potential.

Considering the above and our present findings, we investigated Fgd1 expression in human cancer. First, we examined samples obtained from 42 patients who had undergone a radical prostatectomy for localized prostate cancer. Normal prostatic glands were consistently found negative for Fgd1 (Supplementary Fig. S3A), whereas it was detectable in most cancer cells (Supplementary Fig. S3B–H). We also found that staining for Fgd1 was highest in the more aggressive lesions. Of the 15 least aggressive prostate adenocarcinomas (Gleason score between 4 and 5), 27% were negative, 40% were scored 1+, 33% were scored 2+, and none were scored 3+. Of the 18 lesions of intermediate aggressiveness, 11% were negative, 39% scored 1+, 33% scored 2+, and 17% scored 3+. Finally, of the 9 most aggressive (Gleason score 8 and 9), none were negative for Fgd1, 22% were 1+, 33% were 2+, and 45% were 3+.

We also examined 30 infiltrating ductal breast carcinoma samples for Fgd1 expression. Normal epithelial cells from both glandular lobules and ductus were consistently negative for Fgd1 (Supplementary Fig. S4A and B, respectively). Fgd1 was instead detectable in all cancer lesions examined (Supplementary Fig. S4C–G). As in prostate cancer cells, staining was mainly cytosolic. In general, infiltrating and poorly differentiated breast cancer cells expressed the highest levels of Fgd1. In the well- and intermediately differentiated infiltrating ductal carcinomas (18 specimens), 55% of the lesions were scored 1+, 4% scored 2+, and none scored 3+, whereas of the 12 poorly differentiated cancers examined, 20% scored 1+, 30% scored 2+, and 50% scored 3+. Of note, in some lesions, it seemed that cells at the tumor periphery expressed levels of Fdg1 higher than those found within the tumor mass (data not shown).

Our findings suggest a central role for Fgd1 in regulating focal degradation of the ECM at invadopodia. Furthermore, the correlation between levels of Fgd1 expression and tumor aggressiveness suggests, for the first time, a connection between Fgd1 and at least prostate and breast cancer. These observations imply that Fgd1, not expressed in normal tissues, could play a role during tumor progression and might be a potential new target for therapeutic intervention and/or as a prognostic indicator.

The etiopathology of faciogenital dysplasia indicates that Fgd1 is an important regulator of bone development. In fact, in the developing mouse embryo, the induction of Fgd1 expression coincides with the onset of ossification, whereas after birth, it is expressed more broadly in skeletal tissue (7). Can we generate a unifying hypothesis to place the pathogenesis of faciogenital dysplasia in the framework of our findings on invadopodia formation and tumor progression? ECM remodeling is critical in pathologic processes such as cancer, but also in physiologic events such as immune response, wound healing, embryonic morphogenesis, and differentiation. Fgd1 might thus play a physiologic role in matrix remodeling during development as exemplified by the severe clinical outcome of Fgd1 loss. In the adult, instead, aberrant expression could be linked to the pathologic events associated with tumorigenesis.

No potential conflicts of interest were disclosed.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

We thank Mario Gimona for insightful discussions and Giuseppe Di Tullio for expert assistance in antibody preparation. R. Buccione was supported by the Italian Association for Cancer Research, the European Commission (contract LSHC-CT-2004-503049), the Fondazione Cassa di Risparmio della Provincia di Teramo, and the Ministero della Salute (Art. 12 bis D.Lvo 502/92). V. Castronovo acknowledges support from NFSR (TELEVIE), Belgium. G. Caldieri was supported by fellowships from the Italian Foundation for Cancer Research and the Calogero Musarra Foundation. We acknowledge the GIGA-Cancer tissue bank and Jacques Boniver for the tissue specimens and Pascale Heneaux for expert immunohistochemistry technical assistance.