Akt1 Induces Extracellular Matrix Invasion and Matrix Metalloproteinase-2 Activity in Mouse Mammary Epithelial Cells¹

The serine/threonine-protein kinase, Akt/PKB,³ originally identified by its homology to protein kinases A and C (1, 2), is the transforming protein expressed by the murine AKT8 retrovirus (3, 4). Akt is composed of an NH₂-terminal pleckstrin homology domain that binds to membrane-associated phosphatidylinositol-3,4,5-trisphosphate and phosphatidylinositol-3,4-bisphosphate generated by phosphoinositide 3-kinase (5, 6), a kinase catalytic domain and a COOH-terminal regulatory domain. Akt1 is one of three mammalian isoforms and is activated by several growth factor pathways that elicit phosphorylation at T308 in the activation loop of the catalytic domain and at S473 in the COOH terminus (7). Akt1 is a downstream target of phosphoinositide 3-kinase (8) and translocates to the plasma membrane upon growth factor stimulation (9), where it binds to phosphatidylinositol-3,4-bisphosphate and phosphatidylinositol-3,4,5-trisphosphate. This results in its activation by transphosphorylation at T308 by phosphoinositide-dependent kinase 1 (7), and either autophosphorylation (10) or transphosphorylation at S473 by phosphoinositide-dependent kinase 1 (11, 12), integrin-linked kinase (13), and protein kinase C (14). Akt1 has been implicated in a variety of cellular functions, such as survival, transcription, and translation (8, 15). Overexpression of Akt blocks apoptosis in vitro and is overexpressed in ovarian, pancreatic, and breast cancer (16, 17, 18, 19).

Cancer cell invasion through the ECM and basement membranes is mediated by degradation of collagen IV, the major structural component of the basement membrane (20). Matrix metalloproteinases play a critical role in tumor progression, growth and metastasis (21). The association between the activation of MMP-2 (type IV collagenase/gelatinase A) and ECM invasion in vitro constitutes a model for metastatic progression in human breast cancer (22). Similar to other secreted MMPs, the latent proenzyme of MMP-2 is activated by its association with TIMP-2 (23, 24) and cell surface-anchored MT1-MMP (25, 26), and a stoichiometric imbalance in this ternary complex may be associated with tumor progression. Overexpression of MT1-MMP induces abnormalities and tumor formation in the mammary glands of transgenic mice (27) and increases the invasiveness of prostate cancer cells (28).

In this study, we sought to examine the role of Akt1 in the transformation and invasion of mammary epithelial cells and to determine whether these processes were controlled by the same determinants. In the course of these studies, we discovered that the induction of invasion and transformation were independent events, and that the invasive activity of Akt1-expressing cells correlated with increased MMP-2 expression and activation.

Antibodies and reagents were obtained from the following sources: monoclonal (Ab-3) and polyclonal (Ab-7) anti-MMP-2 antibodies (Oncogene Science, Boston, MA); rabbit polyclonal anti-Akt1 antibody (06-558; Upstate Biotechnology Lake Placid, NY); mouse monoclonal anti-Akt1 antibody (610860; BD Transduction Laboratories, Lexington, KY); rhodamine-conjugated goat antirabbit IgG and Texas Red-conjugated goat antimouse IgG (KPL, Gaithersburg, MD); BB-94 (Batimastat; British Biotech, Oxford, United Kingdom); APMA (Aldrich-Sigma Chemical Co. Chemical, St. Louis, MO); and lactacystin (Calbiochem, San Diego, CA).

COMMA-1D cells (29) were engineered to express the gp800 tv-a cDNA (Ref. 30; kindly provided by Dr. Yi Li, Memorial-Sloan Kettering Cancer Institute, New York, NY). tv-a was cloned into the pSRαMSVtkneo retroviral vector (31) and cotransfected into 293T cells with the pSV-ψ⁻-E-MLV ecotropic env vector by calcium phosphate precipitation. COMMA-1D cells were infected with virus-containing medium and selected in 600 μg/ml G418 for 2 weeks. The expression and the proper function of tv-a in G418-selected COMMA-1D cells were confirmed by immunostaining with a rabbit polyclonal anti-tv-a antibody (generously provided by Dr. Andrew Leavitt, University of California, San Francisco, CA) and by transduction with a recombinant ALV encoding the GFP gene (see below). High-titer ALV was generated by transfecting DF-1 chicken embryo fibroblast cells (Ref. 32; kindly provided by Dr. Stephen Hughes, National Cancer Institute, Bethesda, MD) growing in DMEM containing 10% FCS, 10% tryptose phosphate broth, and 1% chicken serum, with ALV generated with RCAS/Akt1 and RCAS/v-akt (Ref. 33; generously provided by Drs. Peter Vogt and Masahiro Aoki, The Scripps Research Institute, La Jolla, CA) using calcium phosphate precipitation (34). After 6 days, the medium was centrifuged at 3000 × g for 5 min, and the virus-containing supernatant was added to COMMA-1D/tv-a cells and incubated overnight. Cells were transduced by three sequential rounds of infection over 2 days. The medium was then aspirated, and the COMMA-1D/tv-a cell lines were cultured as monolayers in IMEM containing 2.5% FCS, epidermal growth factor (Upstate Biotechnology, Lake Placid, NY; 10 ng/ml), and insulin (Aldrich-Sigma Chemical Co., St. Louis, MO; 5 μg/ml).

COMMA-1D/Akt1 or COMMA-1D/v-akt cells (1 × 10⁵) were suspended in 2 ml of 0.5% (wt/vol) SeaPrep agar (BMA, Rockland, ME) dissolved in IMEM containing 10% FCS and gentamicin (50 μg/ml) and overlaid on 1% agar dissolved in IMEM containing 2% FCS in six-well plates. Cultures were fed once a week for 5 weeks.

A rabbit polyclonal Akt1 antibody (Upstate Biotechnology) was adsorbed on 50 μl of protein A/G-agarose (Santa Cruz Biotechnology, Santa Cruz, CA), and incubated with 250 μg of cell lysate. Immunoprecipitates were incubated in 30 μl of kinase buffer containing 10 μCi of [γ-³²P]ATP, 5 μm ATP, 75 mm MgCl₂, 20 mm 4-morpholinepropanesulfonic acid (pH 7.2), 25 mm β-glycerol phosphate (pH 7.0), 1 mm Na₃VO₄, 1 mm DTT, 17 μm protein kinase A inhibitor peptide, and 30 μm histone H2B as substrate (16). The relative amounts of incorporated radioactivity were determined by autoradiography and densitometry.

Monolayer cultures were grown in 75-cm² plastic flasks for 48 h, washed three times with PBS, and then incubated for 18 h in serum-free IMEM. The medium was collected and concentrated using a Centricon 10 centrifugal filter with a M_r 3000 cutoff (Amicon, Bedford, MA).

Whole cell lysates were prepared by lysing cells in a buffer containing 50 mm Tris-HCl (pH 7.4), 1% NP40, 150 mm NaCl, 1 mm EGTA, 1 mm phenylmethylsulfonyl fluoride, 1 mm Na₃VO₄, 1 mm NaF and Complete protease inhibitor cocktail (Boehringer Mannheim, Mannheim, Germany). Lysates (20 μg of protein) were mixed with Laemmli sample buffer, boiled for 5 min, and separated in 10% PAGEr Gold precast gels (BMA, Rockland, ME) by SDS-PAGE. Resolved proteins were electrophoretically transferred to a nitrocellulose membrane (Protran; Schleicher & Schuell, Keene, NH) and blocked for 1 h in 5% nonfat dry milk in TBST buffer [10 mm Tris-HCl (pH 7.2), 50 mm NaCl, and 0.2% Tween 20]. Blots were incubated overnight with primary antibody at 4°C and then for 1 h at room temperature with horseradish peroxidase-conjugated secondary antibody diluted 1:5000. Labeled proteins were visualized with the ECL detection system (Amersham, Arlington Heights, IL) according to the manufacturer’s instructions.

Assays were carried out with untreated cells or cells treated overnight with either 100 nm BB-94 or 1 μm lactacystin. Cells were then plated on Matrigel-coated Biocoat Cell Culture Insert chambers (Becton Dickinson Labware, Bedford, MA) containing 23-mm diameter, 8-μm pore size filters. Matrigel was diluted with cold, serum-free IMEM and 850 μg of Matrigel were applied to each filter, dried in a hood, and incubated at 37°C for 30 min. Cells (1 × 10⁵) were suspended in serum-free IMEM and added to the Matrigel-coated chamber, and conditioned medium was placed in the lower chamber as a chemoattractant. After incubation for 72 h at 37°C in 5% CO₂, the Matrigel-coated side of the filter was removed with a cotton swab, and cells on the opposite side of the filter were fixed in 10% formaldehyde, stained with H&E, and counted. Each assay was performed in triplicate.

Two μg of protein from concentrated cell-conditioned medium were loaded onto 10% acrylamide-0.1% gelatin gels (Novex, San Diego, CA) and separated by electrophoresis under nonreducing conditions. After electrophoresis, gels were renatured by soaking for 30 min at room temperature in 2.5% Triton X-100 with gentle agitation. Zymogram gels were then equilibrated for 30 min at room temperature in developing buffer [50 mm Tris, 40 mm HCl (pH 7.4), 200 mm NaCl, 5 mm CaCl₂, and 0.02% Brij35] and incubated overnight at 37°C. Transparent bands of gelatinolytic activity were visualized by staining with 0.5% Coomassie Blue R250.

Invasion assays were performed as described above, and after 60 h, attached cells were fixed for 30 min at 4°C with 2% paraformaldehyde and blocked for 30 min at room temperature in 10% normal goat serum, 1% gelatin in PBS. Cells attached to Matrigel were incubated for 2 h at room temperature with a polyclonal anti-MMP-2 antibody diluted 1:100 in blocking buffer and washed for 5 min with PBS three times. The sections were stained for 1 h with rhodamine-conjugated goat antirabbit IgG diluted 1:200, followed by washing with PBS.

For subcellular localization of Akt, chamber-mounted glass slides (Nalge Nunc International, Naperville, IL) were coated with a concentrated cell suspension and fixed for 10 min at room temperature by addition of an equal volume of 10% formaldehyde to the growth medium. The supernatant was removed, and cells were permeabilized with 0.05% Triton X-100 in PBS. Slides were washed with PBS three times and blocked for 1 h at room temperature with 20% goat serum in PBS. Cells were incubated for 1 h at room temperature with a monoclonal anti-Akt1 antibody diluted 1:100 in 20% goat serum/PBS. Slides were washed with PBS three times and then incubated for 1 h with Texas Red-conjugated goat antimouse antibody diluted 1:200, followed by washing with PBS. Slides were treated with ProLong Antifade reagent (Molecular Probes, Inc.), and stained sections were visualized by confocal microscopy.

To examine gene expression in mammary epithelial cells, COMMA-1D cells were engineered to express the ALV subtype A receptor, tv-a, to make them permissive to infection by RCAS viruses. COMMA-1D/tv-a cells were then transduced with RCAS/GFP to allow selection by FACS (Fig. 1). A single clone sorted from the high tv-a-expressing cells was expanded and used in all subsequent studies.

COMMA-1D cells transduced with RCAS-Akt1, and particularly RCAS–v-akt, exhibited high kinase activity compared with the parental cell line expressing GFP alone, as measured by immune-complex kinase assay (Fig. 2). To measure transformation potential, anchorage-independent growth of these cell lines in soft agar was determined (Fig. 3 A). After 5 weeks in culture, cells transduced with v-akt, but not Akt1, formed colonies, which appeared as early as 1 week after infection.

To determine whether these cells also possessed invasive potential, Boyden chamber assays were performed with Matrigel as a substrate (Fig. 3 B). Both COMMA-1D/Akt1 and COMMA-1D/v-akt cells exhibited 25–30% of the invasive potential of MDA-MB-231 breast carcinoma cells, in contrast to parental cells, which were not invasive.

No morphological differences were observed between control and Akt-transduced cells when grown on a plastic substratum (Fig. 4, A, C, and E); however, when cells were grown for 72 h on Matrigel, COMMA-1D/tv-a cells formed simple, spherical colonies (Fig. 4,B), in contrast to the more dense and branched lattices of cells formed by the Akt1- and v-akt-transduced cells (Fig. 4, D and F).

To determine whether secreted protease activity was associated with invasion activity, cell-conditioned medium was assayed by gelatin zymography (Fig. 5). Several bands of activity were revealed under nondenaturing, nonreducing conditions (Fig. 5,A), and immunoblotting of the concentrated cell-conditioned medium from COMMA-1D/Akt1 and COMMA-1D/v-akt cells indicated the presence of an M_r 85,000 form of MMP-2 (Fig. 5,B), the activity of which was confirmed by immunoprecipitation and gelatin zymography (Fig. 5,C). To determine whether the M_r 85,000 form of MMP-2 was proMMP-2, samples were incubated with the cysteine-modifying agent, APMA, to induce autocatalysis (Fig. 6 A). APMA induced the autocatalytic processing of M_r 85,000 MMP-2 to M_r 50,000–72,000 species, suggesting that the high molecular weight MMP-2 complex consisted predominantly of proMMP-2.

To further characterize MMP-2 activity, conditioned medium from Akt-transduced cells was incubated with the pan MMP inhibitor, BB-94 (Fig. 6,B). BB-94 inhibited the gelatinolytic activity of the M_r ∼85,000 form as well as lower molecular weight forms of MMP-2. Treatment of COMMA-1D/Akt1 and COMMA-1D/v-akt cells for 24 h with 100 nm BB-94, a noncytotoxic concentration, inhibited Matrigel invasion by 70% (Fig. 6 C), suggesting a causal relationship between MMP-2 expression and invasion.

To determine whether MMP-2 was localized to the cell surface, MMP-2 expression was measured in COMMA-1D/Akt1 cells by immunohistochemical staining (Fig. 7,A). MMP-2 was present in cell aggregates growing on Matrigel (Fig. 7A, left panel) but was barely visible in control cells (Fig. 7A, left panel). Immunohistochemical staining for Akt1 and v-akt indicated that both enzymes were concentrated in the plasma membrane and cytosol to the same degree (Fig. 7 B).

One mechanism by which Akt might regulate MMP-2 expression is by the modulation of proteasome activity either indirectly though GSK-3β inhibition (35, 36) or directly through MMP-2 phosphorylation. Therefore, cells were treated for 24 h with 1 μm lactacystin, a noncytotoxic concentration, and MMP-2 levels in the conditioned medium were determined (Fig. 8 A). Lactacystin treatment dramatically increased MMP-2 levels in control cells but had only a marginal effect on MMP-2 levels in Akt1- and v-akt-transduced cells. Treatment of COMMA-1D/tv-a cells with lactacystin also resulted in a dramatic increase in their invasive activity but only a slight increase in the activity of COMMA-1D/Akt1 and COMMA-1D/v-akt cells (Fig. 8B).

The present study has examined the ability of Akt1 and its oncogenic homologue, v-akt, to activate ECM invasion in mouse mammary epithelial cells. To achieve this objective, we developed a cell line that is amenable to transduction by one or more recombinant avian retroviruses (37). This system is based on the introduction of the ALV subtype A receptor, tv-a, into mammalian cells making them permissive to infection by recombinant ALV (38). The mouse mammary epithelial cell line, COMMA-1D, was used for this retrovirus delivery system because these cells have been shown to differentiate in vitro and to give rise to normal gland structures in the cleared fat pad of the BALB/c mouse (39, 40). Although these immortalized mammary epithelial cells contain one mutated p53 allele (41), they do not exhibit a transformed phenotype either in vitro or in vivo (39, 40).

Transduction of COMMA-1D/tv-a cells with Akt1 or v-akt led to an interesting dichotomy between their ability to induce transformation versus their ability to induce invasion. Although it is believed that the constitutive activity and plasma membrane localization of v-akt are responsible for its oncogenic activity (3, 33, 42), mammary epithelial cells were found to be susceptible to transformation by v-akt, but not by Akt1, despite little difference in membrane localization of these proteins (Fig. 7 B); however, v-akt did exhibit 2.5-fold greater activity, which may account, in part, for its transforming activity. Because v-akt and Akt1 induced similar invasion activity and MMP-2 expression, differences other than absolute activity and membrane localization appear to be associated with MMP-2 expression. One clue to the mechanism for this effect may be deduced from the lactacystin experiments. Control cells treated with this proteasome inhibitor exhibited a dramatic increase in MMP-2 levels and invasion activity in comparison with its marginal effect on Akt- and v-akt-expressing cells. This suggests that the mechanism of action of Akt1 and v-akt is similar to that of lactacystin, i.e., to block proteasome activity. This effect may occur indirectly though GSK-3β inhibition by Akt (35, 36) or directly through MMP-2 phosphorylation, although the latter possibility has not been reported.

Another way by which Akt may increase MMP-2 expression is by transcriptional activation. The MMP-2 promoter contains several cis-acting regulatory elements including cyclic AMP-responsive element binding, Sp1, Ets-1 and AP-2 (43). Akt is known to up-regulate ets-2 (44) and cyclic AMP-responsive element binding (45), and we are currently examining whether this is a potential regulatory pathway for MMP-2 expression by Akt1 in mammary epithelial cells.

ECM invasion depends on the cooperative processes of cell adhesion and membrane-associated proteolysis (46). MMP-2 has been implicated in ECM invasion in breast carcinoma cell lines (47, 48, 49) as well as in mammary epithelial cells transduced with H-Ras or stimulated by nitric oxide activators (50, 51). Invasion correlated with phosphoinositide 3-kinase and Akt activation in instances where ECM invasion was induced by epidermal growth factor or IGF-1 (52, 53). Akt activation has also been implicated in ECM adhesion and H-Ras-mediated transformation by blocking anoikis-induced cell death (54, 55), and Akt has been shown to localize to sites of epithelial cell-matrix contact (56). MMP-2 localizes to the surface of invading tumor and endothelial cells and interacts through its COOH-terminal hemopexin-like domain with αvβ3 integrin (57, 58, 59). Cleavage of the ECM by MMP-2 induces migratory activity of breast epithelial cells (60), suggesting a close relationship between cell adhesion and cell invasion in tissue remodeling. Because αvβ3-mediated cell migration in tumor cells was dependent on Akt activation (61), Akt and MMP-2 appear to be critically linked to cell adhesion.

MMP-2 exists as a M_r 100,000 complex of proMMP-2 and TIMP-2, which results in multiple forms of secreted MMP-2 (23). Incubation of this complex with the cysteine-reactive compound, APMA, promoted the autocatalytic processing of proMMP-2 to active M_r 72,000 to M_r 50,000 species (Fig. 6 B; Refs. 23, 24, 62). On the cell surface, proMMP-2 is activated by MT1-MMP after its association with TIMP-2 via its hemopexin domain (63). Formation of a complex of proMMP-2 with MT1-MMP correlated with ECM invasion where the complex localizes to invadopodia (64) and to sites of collagenolytic activity (25). Although MT1-MMP-dependent invasion was independent of proMMP-2 activation, both enzymes were required to be membrane-anchored to the cell surface to function as pericellular collagenases (25). Therefore, these studies suggest that the concentration of MMP-2 along basement membranes at sites of tissue remodeling and at the leading edge of invading cells (65, 66) is a major factor contributing to tumor progression and invasion (67, 68, 69). The present study further delineates that the Akt1 signaling pathway is involved in up-regulating this process and suggests a mechanism by which tumors with high Akt1 expression may be more invasive and metastatic.

We thank Drs. Peter Vogt and Masahiro Aoki (Scripps Research Institute) for providing RCAS/Akt1 and RCAS/v-akt, Dr. Stephen Hughes (National Cancer Institute) for providing the RCAS vector and DF-1 cells, Dr. Yi Li (Memorial Sloan-Kettering Cancer Institute) for providing the tv-a cDNA, and Dr. Andrew Leavitt (University of California, San Francisco, CA) for providing the polyclonal anti-tv-a antibody.