Hyaluronan-mediated CD44 interaction with RhoGEF and Rho kinase promotes Grb2-associated binder-1 phosphorylation and phosphatidylinositol 3-kinase signaling leading to cytokine (macrophage-colony stimulating factor) production and breast tumor progression.

In this study we have examined CD44 (a hyaluronan (HA) receptor) interaction with a RhoA-specific guanine nucleotide exchange factor (p115RhoGEF) in human metastatic breast tumor cells (MDA-MB-231 cell line). Immunoprecipitation and immunoblot analyses indicate that both CD44 and p115RhoGEF are expressed in MDA-MB-231 cells and that these two proteins are physically associated as a complex in vivo. The binding of HA to MDA-MB-231 cells stimulates p115RhoGEF-mediated RhoA signaling and Rho kinase (ROK) activity, which, in turn, increases serine/threonine phosphorylation of the adaptor protein, Gab-1 (Grb2-associated binder-1). Phosphorylated Gab-1 promotes PI 3-kinase recruitment to CD44v3. Subsequently, PI 3-kinase is activated (in particular, alpha, beta, gamma forms but not the delta form of the p110 catalytic subunit), AKT signaling occurs, the cytokine (macrophage-colony stimulating factor (M-CSF)) is produced, and tumor cell-specific phenotypes (e.g. tumor cell growth, survival and invasion) are up-regulated. Our results also demonstrate that HA/CD44-mediated oncogenic events (e.g. AKT activation, M-CSF production and breast tumor cell-specific phenotypes) can be effectively blocked by a PI 3-kinase inhibitor (LY294002). Finally, we have found that overexpression of a dominant-negative form of ROK (by transfection of MBA-MD-231 cells with the Rho-binding domain cDNA of ROK) not only inhibits HA/CD44-mediated RhoA-ROK activation and Gab-1 phosphorylation but also down-regulates oncogenic signaling events (e.g. Gab-1.PI 3-kinase-CD44v3 association, PI 3-kinase-mediated AKT activation, and M-CSF production) and tumor cell behaviors (e.g. cell growth, survival, and invasion). Taken together, these findings strongly suggest that CD44 interaction with p115RhoGEF and ROK plays a pivotal role in promoting Gab-1 phosphorylation leading to Gab-1.PI 3-kinase membrane localization, AKT signaling, and cytokine (M-CSF) production during HA-mediated breast cancer progression.

Members of the Rho subclass of the Ras superfamily (small molecular weight GTPases, e.g. RhoA, Rac1, and Cdc42) are known to transduce signals regulating many cellular processes (25). The rationale for our focusing on RhoGTPase is based on previous reports suggesting that CD44-associated cytoskeletal proteins (e.g. ankyrin and ERM) and that tumor cell-specific phenotypes are dependent on RhoGTPase signaling events (16,22,26). Overexpression of certain RhoGTPases in human tumors often correlates with poor prognosis (28,29). In particular, coordinated RhoGTPase signaling is considered to be a possible mechanism underlying cell proliferation and motility, an obvious prerequisite for metastasis (16, 25, 26, 28 -31). Presently, very little information is available regarding how they are activated by cell surface receptors. To date, at least 30 different guanine nucleotide exchange factors (GEFs) have been identified (32). Our recent results indicate that the interaction between CD44v3 isoform and two GEFs, such as Tiam1 (24) and Vav2 (21), up-regulates Rac1 signaling and cytoskeleton-mediated metastatic tumor progression. As part of our continued effort to identify CD44 isoform-linked GEFs, which correlate with certain metastatic behaviors, a new candidate molecule, named p115RhoGEF, has been identified. p115RhoGEF, which is the human homolog of the mouse protein Lsc, is involved in RhoA activation (33); structurally, it contains numerous functional domains and structural motifs found in signal transduction proteins and oncoproteins. These motifs include N-terminal RGS (regulator of G protein signaling) domain, a dbl homology domain (DH), and a pleckstrin homology domain (PH) (33,34). In particular, the sequence between amino acids (aa) 421 and aa 635 contains significant sequence homology to the dbl homology domain (DH) of many proteins that exhibit GDP/GTP exchange activity for specific members of the Ras superfamily of GTP-binding proteins (32). The p115RhoGEF has been clearly shown to act as a GDP/GTP exchange protein for the Rho subfamily of GTPases, including RhoA (33). In addition, p115RhoGEF (similar to PDZ-RhoGEF) contains an RGS domain, which interacts with G␣ 13 and regulates the ability of DH to carry out GDP/GTP exchange activity for RhoA (33,34). In the C-terminal region, p115RhoGEF contains one PH domain that is commonly detected in signaling molecules and cytoskeletal proteins (35,36). Activation of RhoGTPases such as RhoA has been shown to produce specific structural changes in the plasma membrane associated with actin filament bundling, stress fiber formation, and acto-myosin-based cytoskeletal function (25).
Several different enzymes have been identified as possible downstream targets for RhoA signaling. One such enzyme is Rho kinase (ROK, also called Rho-associated kinase), which is a serine-threonine kinase known to interact with Rho in a GTP-dependent manner (37)(38)(39). ROK is composed of four functional domains, including a kinase domain (catalytic site), a coiled-coil domain, a Rho-binding (RB) domain, and a PH domain (40). This enzyme has been shown to regulate cytoskeleton function by phosphorylating several important cytoskeletal regulators, including myosin-binding subunit of myosin phosphatase (41), calponin (42), adducin (43), and Lin 11/Isl-1/ Mec-3 (LIM) kinase (44). ROK is also involved in the "crosstalk" between Ras and Rho signaling leading to cellular transformation (30). We have demonstrated that ROK phosphorylates the cytoplasmic domain of the CD44v 3, 8 -10 isoform and up-regulates the interaction between the CD44v 3, 8 -10 iso-form and the cytoskeletal protein, ankyrin, during HA/CD44regulated tumor cell migration (22). Most recently, the PH domain of ROK is found to be involved in the direct binding to CD44-and HA-mediated Ca 2ϩ signaling in endothelial cell function (45). Thus, ROK is clearly one of the important signaling molecules required for membrane-cytoskeleton interaction, Ca 2ϩ regulation, and HA/CD44-mediated cell function (22,45).
In the pathogenesis of cancer, Grb-2-associated binder-1 (Gab-1) and phosphatidylinositol 3-kinase (PI 3-kinase) are key mediators in regulating oncogenesis (for review see Ref. 46). Specifically, Gab-1 (a member of the insulin receptor substrate (IRS) family) functions as one of the major adapter molecules downstream of growth factor signaling (47,48). Gab-1 also possesses multiple phosphorylation sites that could act as docking sites for PI 3-kinase known to consist of a catalytic subunit p110 (␣, ␤, and ␦) and regulatory subunit p85 (␣, ␤, and p55␥) or the catalytic subunit p110␥ and the regulatory subunit p101 (for reviews see Refs. 46, 49 -51). One recent study found a positive link between HA-CD44 interaction and Gab-1-associated PI 3-kinase activation during the stimulation of cellular transformation by a Met-hepatocyte growth factor oncoprotein (Tpr-Met) (52). HA and CD44 also promote PI 3-kinase signaling in a tumor cell-specific manner. For example, PI 3-kinase participates in CD44-mediated survival pathway in colon carcinoma cells (53). HA activates PI 3 kinase-AKT pathways leading to cell motility and cell survival-signaling pathways (54). The active mutant of p110 subunit of PI 3-kinase exerts its action on the cleavage of CD44 during cancer cell migration (55). These findings suggest PI 3-kinase activation is closely coupled with HA-mediated CD44 signaling.
Because both Rho signaling and PI 3-kinase activation play an important role in regulating breast tumor progression, we have focused in this study on the relationship between these two signaling pathways during HA/CD44-mediated breast tumor progression. A unique mechanism is described concerning CD44 interaction with p115RhoGEF and ROK that stimulates Gab-1 phosphorylation/membrane localization and PI 3-kinase-AKT activation leading to HA/CD44-regulated cytokine production and tumor cell behaviors required for breast cancer progression.

MATERIALS AND METHODS
Cell Culture-The breast tumor cell line MDA-MB-231 was obtained from the American Type Culture Collection (ATCC) and grown in Eagle's minimum essential medium supplemented with Earle's salt solution, essential and non-essential amino acids, vitamins, and 10% fetal bovine serum.
Antibodies and Reagents-Monoclonal rat anti-human CD44 antibody (clone: 020; isotype: IgG 2b ; obtained from CMB-Tech, Inc., San Francisco, CA) used in this study recognizes a common determinant of the CD44 class of glycoproteins. Both rabbit anti-CD44v3 antibody and rabbit anti-Rho kinase (ROK) were prepared according to the procedures described previously (22). For the preparation of polyclonal rabbit anti-p115RhoGEF antibody, specific synthetic peptides (ϳ15-17 amino acids unique for the ROK or p115RhoGEF sequence) were prepared by the Peptide Laboratories using an Advanced Chemtech automatic synthesizer (model ACT350). All polyclonal antibodies were prepared using conventional DEAE-cellulose chromatography and tested to be monospecific (by immunoblot assays). Mouse monoclonal anti-green fluorescent protein (GFP) and mouse monoclonal anti-FLAG (M2) were purchased from BD Pharmingen and Sigma, respectively. Rabbit antiphospho-threonine antibody and rabbit anti-phospho-serine antibody were obtained from Zymed Laboratories Inc.. Monoclonal mouse anti-p110␣, mouse anti-p110␤, mouse anti-p110␥ and mouse anti-p110␦ were purchased from Santa Cruz Biotechnology. Several other immunoreagents, including mouse anti-AKT-1 (protein kinase B), rabbit antiphospho-AKT-1 (threonine 308), and rabbit anti-Gab-1, were purchased from Upstate Biotechnology, Inc. The specific inhibitor of PI 3-kinase (LY294002) was obtained from Calbiochem. Rooster comb hyaluronan (HA) was purchased from Sigma. High molecular mass HA polymers (ϳ10 6 Da) were purified by gel filtration column chromatography using a Sephacryl S1000 column as described previously (18). The purity of high molecular weight HA polymers used in our experiments was further verified by anion exchange high performance liquid chromatography. No small HA fragments was detected in these preparations.
Cloning, Expression, and Purification of CD44 Cytoplasmic Domain (CD44cyt) from E. coli-The cytoplasmic domain of human CD44 (CD44cyt) was cloned into pFLAG-AST using the PCR-based cloning strategy (15). Using human CD44 cDNA as template, one PCR primer pair (left, FLAG-EcoRI; right, FLAG-XbaI) was designed to amplify complete CD44 cytoplasmic domain. The amplified DNA fragments were one-step-cloned into a pCR2.1 vector and sequenced. Then, the DNA fragments were cut out by double digestion with EcoRI and XbaI and subcloned into EcoRI/XbaI double-digested pFLAG-AST (Eastman Kodak-IBI, Rochester, NY) to generate a FLAG-pCD44cyt construct. The nucleotide sequence of FLAG/CD44cyt junction was confirmed by sequencing. The recombinant plasmids were transformed to BL21-DE3 to produce a FLAG-CD44cyt fusion protein. The FLAG-CD44cyt fusion protein was further purified by using an anti-FLAG M2 affinity gel column (Eastman Kodak-IBI). The nucleotide sequence of primers used in this cloning protocol is as follows: FLAG-EcoRI, 5Ј-GAGAATTCGA-ACAGTCGAAGAAGGTGTCTCTTAAGC-3Ј; FLAG-XbaI, 5Ј-AGCTCT-AGATTACACCCCAATCTTCAT-3Ј.
Cloning and Expression of the PH Domain of Gab-1-The Gab-1 cDNA fragment containing the PH domain (bp 132-482; GenBank TM accession number NM_002039) was amplified by RT-PCR using PHspecific primers (upper primer containing a start codon) 5Ј-GCCATG-GGTGAAGTGGTCTGCTCCGGATG-3Ј and 5Ј-TGTTGGATTAAACCC-ACAGATGTCACA-3Ј. The Gab-1-PH cDNA fragment was cloned into p cDNA3.1-V5/His TOPO vector (Invitrogen). The inserted PH sequence was confirmed by nucleotide sequencing analyses (UCSF Biomolecular Resource Center). This V5/His-tagged Gab-1-PH fragment cDNA was then used for transient expression in COS-7 cells or MDA-MB-231 cells as described below. The expression of His-tagged Gab-1-PH domain in COS-7 or MDA-MB-231 cells was detected by SDS-PAGE and immunoblot/immunoprecipitation analyses.
In Vitro Binding of CD44cyt to the PH Fragment of Gab-1 and Intact Gab-1-Aliquots (0.5-1 ng of protein) of V5/His-tagged Gab-1 PH fragment (isolated from COS-7 cells)-conjugated Sepharose beads or V5/ His-conjugated beads were incubated in 0.5 ml of binding buffer (20 mM Tris-HCl (pH 7.4), 150 mM NaCl, 0.1% bovine serum albumin, and 0.05% Triton X-100) containing FLAG-CD44cyt fusion protein (as described above) at 4°C for 4 h. The binding equilibrium was found to be established when the in vitro CD44-Gab-1-PH binding assay was conducted at 4°C after 4 h. Following binding, the Gab-1-PH-V5/Hisconjugated beads or V5/His-conjugated beads were washed extensively in binding buffer followed by immunoblotting with anti-FLAG antibody. In some cases, 125 I-labeled CD44cyt (the cytoplasmic domain of CD44 fusion protein; ϳ0.35 nM protein, 2 ϫ 10 4 cpm/ng) was incubated with Sepharose beads containing intact Gab-1 phosphorylated by ROK plus activated RhoA (RhoA with GTP␥S bound (GTP␥S⅐RhoA)) or ROK plus unactivated RhoA (e.g. RhoA without GTP␥S bound) in the presence of various concentrations of unlabeled CD44cyt ranging from 10 Ϫ12 to 10 Ϫ6 M. The radioactivity associated with Gab-1 beads was analyzed by liquid scintillation counting.
Method for Preparing GFP-tagged Dominant-negative Form, Containing the RB Domain of ROK-The cDNA fragment encoding the dominant-negative form of ROK (2719 -3237 bp, containing the RB sequence) was amplified by reverse transcription-PCR using RB-specific primers linked with an enzyme (XhoI and HindIII) digestion site, 5Ј-CGATCTCGAGGGCCTTCTGGAGGAGAGTA-3Ј and 5Ј-CGATAAG-CTTCTGCATCTGAAGCTCATTCC-3Ј. PCR product digested with XhoI and HindIII was purified with a QIAquick PCR purification kit (Qiagen) (22). The ROK-RBcDNA fragments were cloned into pEGFPC1 vector (Clontech) digested with XhoI and HindIII. The inserted RB sequence was confirmed by nucleotide sequencing analyses. This GFP-ROK-RBcDNA was then used for a transient expression in MDA-MB-231 cells. The GFP-tagged ROK-RB (molecular mass, ϳ50 kDa) expressed in MDA-MB-231 cells was analyzed by SDS-PAGE and immunoblot as described below.
Cell Transfection-To establish a transient expression system, MDA-MB-231 cells were transfected with various plasmid DNAs (e.g. Histagged Gab-1-PH cDNA or GFP-tagged ROK-RBcDNA or vector alone) using LipofectAMINE 2000. Transfected cells were then grown in the culture medium containing G418 for at least 3 days. Various transfectants were analyzed for their protein expression (e.g. Gab-1-PH domain or ROK-RB domain) by immunoblot and functional assays as described below. In some experiments, COS-7 cells were transfected with Histagged Gab-1 fragment cDNA (or vector alone) using LipofectAMINE 2000. The expression of His-tagged Gab-1 PH domain in COS-7 was also detected by SDS-PAGE and anti-His-mediated immunoblot/immunoprecipitation analyses.
In some experiments, Nonidet P-40-solubilized cell lysate isolated from MDA-MB-231 cells (untransfected (treated with or without LY294002) or transfected with GFP-tagged ROK-RBcDNA or vector alone-treated with HA (50 g/ml) or pre-treated with anti-CD44 antibody followed by HA (50 g/ml) or without HA treatment) was analyzed by SDS-PAGE followed by immunoblotting with anti-GFP or anti-phospho-AKT-1 (50 g/ml) followed by re-blotting with anti-AKT-1 (50 g/ ml). These blots were then treated with peroxidase-conjugated goat anti-rabbit IgG and ECL chemiluminescence reagent.
Some MDA-MB-231 cells (untransfected (treated with or without LY294002) or transfected with ROK-RBcDNA or vector alone-treated with HA (50 g/ml) or pre-treated with anti-CD44 antibody followed by HA (50 g/ml) or without HA treatment) were also immunoprecipitated with anti-Gab-1 antibody followed by immunoblotting with anti-phospho-serine or anti-phospho-threonine, respectively. These blots were then developed using ECL chemiluminescence reagent according to the manufacturer's instructions. During these immunological analyses, an equal amount of cellular protein (50 g/ml) immunoprecipitated with the antibody was applied to SDS-PAGE followed by immunoblot analyses.
p115-mediated GDP/GTP Exchange for RhoA-Purified Escherichia coli-derived GST-tagged RhoA or GST alone (20 pmol) was preloaded with GDP (30 M) in 10 l of buffer containing 25 mM Tris-HCl (pH 8.0), 1 mM dithiothreitol, 4.7 mM EDTA, 0.16 mM MgCl 2 , and 200 g/ml bovine serum albumin at 37°C for 7 min. To terminate preloading procedures, additional MgCl 2 was then added to the solution (reaching a final concentration of 9.16 mM) as described previously (16,17,59). Subsequently, 2 pmol of p115RhoGEF (bound to anti-p115RhoGEFconjugated Sepharose beads), isolated from MDA-MB-231 cells grown in the presence or absence of HA (50 g/ml), was preincubated with 2.5 M [ 35 S]GTP␥S (ϳ1250 Ci/mmol) (in the presence or absence of 2.25 M GTP␥S) for 10 min followed by adding 2.5 pmol of GDP-loaded GSTtagged RhoA (or GDP-treated GST). At various time points, the reaction of each samples was terminated by adding ice-cold termination buffer containing 20 mM Tris-HCl (pH 8.0), 100 mM NaCl, and 10 mM MgCl 2 followed by filtering through nitrocellulose filters. The radioactivity associated with the filters was measured by scintillation fluid. The amount of [ 35 S]GTP␥S bound to p115RhoGEF (bound to anti-p115RhoGEF-Sepharose beads) or control sample (preimmune serumconjugated Sepharose beads) in the absence of GST-tagged RhoA was subtracted from the original values. In some cases, p115RhoGEF-catalyzed GDT/GTP exchange reactions were measured using MDA-MB-231 cells preincubated with monoclonal rat anti-CD44 antibody (or normal rat IgG) followed by HA treatment or fibronectin treatment (or no treatment). Data represent an average of triplicates from three to five experiments. The standard deviation was less than 5%.
Measurement of M-CSF Production-MDA-MB-231 cells (untransfected (treated with or without LY294002) or transfected with ROK-RBcDNA or vector alone) were washed three times with serumfree DMEM and incubated in 3 ml of serum-free DMEM containing various reagents (e.g. HA (50 g/ml) or anti-CD44 antibody plus HA (50 g/ml) or without HA treatment) for 24 h at 37°C in a 5% CO 2 humidified chamber. Subsequently, M-CSF concentrations in the conditioned medium and cells were determined using the Quantikine M-CSF immunoassay (R&D Systems). Statistical analysis was done using the Student t test. All data were expressed as the mean Ϯ S.D.
In Vitro Tumor Cell Growth Assays-MDA-MB-231 cells (untransfected (treated with or without LY294002) or transfected with ROK-RBcDNA or vector alone) (5 ϫ 10 3 cells/well) were treated with HA (50 g/ml) (or pre-treated with rat anti-CD44 followed by HA treatment (50 g/ml) or untreated). These cells were then plated in 96-well culture plates in 0.2 ml of Dulbecco's modified Eagle's medium/ F-12 medium supplement (Invitrogen, Grand Island, NY) containing either 0.5% fetal bovine serum or no serum for 24 h at 37°C in 5%CO 2 / 95% air. In each experiment, a total of five plates (10 wells/treatment; e.g. HA treatment or rat anti-CD44 plus HA treatment or no treatment/ plate) were used. Experiments were repeated three times. The in vitro growth of these cells was analyzed by measuring increases in cell number using MTT (3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide) assays (CellTiter 96® non-radioactive cell proliferation assay according to the procedures provided by Promega) (2). Subsequently, viable cell-mediated reaction products were recorded by using a Molecular Devices (Spectra Max 250) enzyme-linked immunosorbent assay reader at a wavelength of 450 nm.
Tumor Cell Invasion Assays-Twenty-four Transwell units were used for monitoring in vitro tumor cell invasion as described previously (56,57). Specifically, the 5-m porosity polycarbonate filters coated with the reconstituted basement membrane substance Matrigel (Collaborative Research, Lexington, MA) were used for the cell invasion assay (56,57). MDA-MB-231 cells (untransfected (treated with or without LY294002) or transfected with ROK-RBcDNA or vector alone) (1 ϫ 10 Ϫ4 cells/well in phosphate buffered saline, pH 7.2) were placed in the upper chamber of the Transwell unit in the presence or absence of HA. The growth medium, containing high glucose DMEM supplemented by 10% fetal bovine serum, was placed in the lower chamber of the Transwell unit. After 18-h incubation at 37°C in a humidified 95% air/5% CO 2 atmosphere, cells on the upper side of the filter were removed by wiping with a cotton swap. Cell invasion processes were determined by measuring the cells that migrate to the lower side of the polycarbonate filters by standard cell number counting methods as described previously (15,56,57). Each assay was set up in triplicate and repeated at least five times. All data were analyzed statistically by Student's t test, and statistical significance was set at p Ͻ 0.01.

Interaction of CD44v3 and p115RhoGEF in Breast Tumor Cells (MDA-MB-231 Cells)-
The level of CD44v3 isoform expression increases as the histological grade of the breast tumor progresses. In fact, there is a direct correlation between CD44v3 isoform expression and increased histological grade of the malignancy (22). Using anti-CD44v3 antibody and immunoblot analysis, we have confirmed the presence of the 85-kDa CD44v3 protein in MDA-MB-231 cells (Fig. 1, lane 1). A number of studies have shown that CD44v3 and RhoGTPases (in particular, RhoA) are structurally and functionally coupled in breast tumor cells (22). Although several RhoA-specific guanine nucleotide exchange factors (RhoGEFs) have been shown to be involved in RhoA signaling (33,34,58), identification of the specific RhoGEF protein that plays a direct role in regulating CD44v3-specific metastatic behaviors in breast tumor cells is addressed in this study.
Immunoblotting with anti-p115RhoGEF antibody indicates that a single polypeptide (molecular mass, ϳ115 kDa) is expressed in MDA-MD-231 cells (Fig. 1, lane 2). In addition, we have carried out anti-p115RhoGEF-mediated or anti-CD44v3mediated immunoprecipitation followed by anti-CD44v3 immunoblot (Fig. 1, lane 3) or anti-p115RhoGEF immunoblot (Fig. 1, lane 4), respectively. Our results clearly indicate that the CD44v3 is present in anti-p115RhoGEF-immunoprecipitated materials (Fig. 1, lane 3). Conversely, the p115RhoGEF band is also detected in the anti-CD44v3-immunoprecipitated materials (Fig. 1, lane 4). The results with control samples confirm the specificity of these immunological techniques. For example, very little p115RhoGEF is detected in anti-CD44v3mediated immunoprecipitated materials blotted by an anti-p115RhoGEF-free serum (anti-p115RhoGEF antibody pre-absorbed by an excess amount of p115RhoGEF) (Fig. 1, lane 5). Similarly, no CD44v3 is observed in anti-p115RhoGEF-mediated immunoprecipitated materials blotted by an anti-CD44v3free serum (anti-CD44v3 antibody pre-absorbed by an excess amount of CD44v3) (Fig. 1, lane 6). These findings indicate that MDA-MB-231 cells were solubilized by 1% Nonidet P-40 buffer followed by immunoprecipitation and/or immunoblotting by anti-CD44v3 antibody or anti-p115RhoGEF antibody, respectively, as described under "Materials and Methods." Lane 1, immunoblot of MDA-MB-231cells with anti-CD44v3 antibody. Lane 2, immunoblot of MDA-MB-231cells with anti-p115RhoGEF antibody. Lane 3, detection of CD44v3 in the complex by anti-p115RhoGEF-immunoprecipitation followed by immunoblotting with anti-CD44v3 antibody. Lane 4, detection of p115RhoGEF in the complex by anti-CD44v3-mediated immunoprecipitation followed by immunoblotting with anti-p115RhoGEF antibody. Lane 5, detection of anti-CD44v3-mediated immunoprecipitated materials blotted by an anti-p115RhoGEF-free serum (anti-p115RhoGEF antibody pre-absorbed by an excess amount of p115RhoGEF). Lane 6, detection of anti-p115RhoGEF-mediated immunoprecipitated materials blotted by an anti-CD44v3-free serum (anti-CD44v3 antibody pre-absorbed by an excess amount of CD44v3).
CD44v3 and p115RhoGEF are physically associated in breast tumor cells.
Analyses of p115RhoGEF-catalyzed RhoA Activation in MDA-MB-231 Cells-It has been reported that p115RhoGEF functions as an exchange factor for Rho-like GTPases such as RhoA in G-protein-coupled receptor activation (22,33,34). To confirm that the p115RhoGEF-like molecule in this study functions as a GDP/GTP exchange factor (or a GDP-dissociation stimulator protein) for RhoGTPases in breast tumor cells, we first isolated p115RhoGEF from MDA-MD-231 cells using anti-p115RhoGEF-conjugated Sepharose beads. As indicated in Fig.  2, the purified p115RhoGEF activates GDP/GTP exchange on GST-tagged RhoA (Fig. 2). The initial onset of the exchange reaction on GST-tagged RhoA occurs within 0.5-2 min after the addition of p115RhoGEF, and the reaction reaches its maximum ϳ6 min after p115RhoGEF addition (Fig. 2). Most importantly, we have observed that the addition of HA to CD44v3containing MDA-MB-231 cells stimulates both the rate of p115RhoGEF-mediated GDP/GTP exchange reaction and the maximal amount of bound [ 35 S]GTP␥S to GST-RhoA (at least a 1.75-fold increase) (Fig. 2, line a) as compared with the p115RhoGEF isolated from MDA-MB-231 cells without HA treatment (Fig. 2, line b). In control samples, a low level of [ 35 S]GTP␥S-bound material was detected in GST alone under same GDP/GTP exchange reaction using p115RhoGEF isolated from MDA-MB-231 cells in the presence (Fig. 2, line c) or absence (Fig. 2, line d) of HA treatment. Therefore, we conclude that p115RhoGEF in MDA-MB-231 cells functions as a GDP/ GTP exchange factor for Rho-like GTPases such as RhoAGTPase.
Identification of Gab-1 as a New Cellular Substrate for ROK-Several studies indicate that Rho kinase (ROK, also called Rho-associated kinase-a serine-threonine kinase) interacts with RhoA in a GTP-dependent manner during HA/CD44mediated cell activation (22,45). Using a ROK-specific anti-body, we have confirmed the presence of ROK (ϳ160-kDa polypeptide) in MDA-MB-231 cells (Fig. 3, lane 2). No ROKcontaining material was observed in control samples when pre-immune rabbit serum was used in these experiments (Fig.  3, lane 1). This RhoA-activated ROK has been shown to induce phosphorylation of a number of cytoskeletal proteins and signaling molecules (22,(41)(42)(43)(44)(45). A recent study indicates that overexpression of an active form of RhoA (RhoA v14 ) promotes ROK-mediated serine phosphorylation of insulin-receptor substrate-1 (IRS-1) during insulin signaling (59). In this study we have found that a member of the IRS family, Gab-1, is expressed in MDA-MD-231 cells (Fig. 4A, panel c). Furthermore, our data indicate that ROK isolated from MDA-MD-231 cells is capable of phosphorylating the Gab-1 at both serine (Fig. 4A, panel a, lane 3) and threonine residues (Fig. 4A, panel b, lane 3) (as detected by anti-phospho-serine and anti-phospho-threonine-mediated immunoblot, respectively, followed by reblotting with anti-Gab-1 antibody (Fig. 4A, panel c, lane 3) in the presence of activated GTP␥S⅐RhoA (Fig. 4A, panels a-c, lane 3). The level of serine/threonine phosphorylation of Gab-1 appears to be relatively low if ROK was incubated with unactivated RhoA (RhoA without GTP␥S bound) (Fig. 4A, panels a-c, lane 2). In the absence of ROK, no serine/threonine phosphorylation of Gab-1 was detected (Fig. 4A, panels a-c, lane 1). These results indicate that Gab-1 serves as one of the cellular substrates for the RhoA-dependent kinases such as ROK in vitro.
Furthermore, we have examined the effect of ROK-mediated Gab-1 phosphorylation on CD44 binding. Specifically, the highly phosphorylated form of Gab-1 (by ROK plus activated RhoA (RhoA with GTP␥S bound, GTP␥S⅐RhoA); as shown in Fig. 4A, lane 3) or the minimally phosphorylated form of Gab-1 (using ROK plus unactivated RhoA (e.g. RhoA without GTP␥S bound); as shown in Fig. 4A, lane 2) was incubated with various concentrations of 125 I-labeled CD44 cytoplasmic domain (FLAG-CD44cyt) under equilibrium binding conditions. Our results indicate that CD44 binds to either highly phosphorylated Gab-1 or minimally phosphorylated Gab-1 at a single site. Importantly, the highly phosphorylated Gab-1 displays at least 3-fold higher CD44 binding affinity (with an apparent K d of ϳ0.18 nM) (Fig. 4B, panel a) than the minimally phosphorylated Gab-1 (with an apparent K d of ϳ0.58 nM) (Fig. 4B, panel  b). These results clearly support the notion that Gab-1 phosphorylation by RhoA-activated ROK enhances its binding interaction with CD44, which may be required for Gab-1-mediated recruitment of signaling components into CD44associated membrane during HA activation. Other approaches, such as mutational analyses of the appropriate serine/threonine residues on Gab-1 (activated by RhoA-activated ROK) required for CD44 binding, are under investigation in our laboratory.
Previous studies determined that HA-CD44 interaction promotes ROK activation in breast tumor cells (22). Several cellular proteins, including the cytoplasmic domain of CD44 and IP 3 receptors, have been identified as ROK-specific cellular sub- strates during HA-CD44 signaling (45). Overexpression of a dominant-negative form of ROK (by transfecting cells with the RB domain of ROK (ROK-RB) cDNA) is capable of inhibiting endogenous ROK activity leading to down-regulation of HA/ CD44 signaling and tumor-specific phenotypes (22). To demonstrate ROK-mediated phosphorylation of Gab-1 in vivo, a dominant-negative form of ROK (ROK-RB) cDNA was cloned into a green fluorescent protein (GFP)-tagged expression vector (pEF-GPC1 vector) followed by a transient transfection of GFP-RBcDNA into MDA-MB-231 cells. Anti-GFP-mediated immunoblot analysis indicates that the RB fragment of ROK is expressed in MDA-MB-231 cells transfected with the RBcDNA of ROK (Fig. 5A, panel a, lanes 4 -6) but not in those cells transfected with vector alone (Fig. 5A, panel a, lanes 1-3). Using anti-ROK-mediated immunoblot, we have detected no significant difference in endogenous ROK expression between the ROK RBcDNA-transfected cells (Fig. 5A, panel a, lanes  4 -6) and vector-transfected cells (Fig. 5A, panel a, lanes 1-3) treated with HA or no HA (or pretreated with anti-CD44 followed by HA treatment). The question of whether Gab-1 phos-phorylation and Gab-1-mediated signaling events are affected by HA/CD44-activated ROK and/or the dominant-negative form of ROK (ROK-RB overexpression) in breast tumor cells is the focus of this study.
The results of our analyses indicate that the level of serine and threonine phosphorylation of Gab-1 (as detected by anti-Gab-1-mediated immunoprecipitation followed by immunoblotting with anti-serine (Fig. 5B, panel a) or anti-threonine (Fig.  5B, panel b) or anti-Gab-1 (Fig. 5B, panel c), respectively) is greatly enhanced in vector-transfected cells treated with HA (Fig. 5B, panel a, lane 2; panel b, lane 2). In contrast, Gab-1 serine and threonine phosphorylation is relatively low in vector-transfected cells without any HA treatment (Fig. 5B, panel  a, lane 1; panel b, lane 1) or those vector-transfected cells pre-treated with anti-CD44 followed by HA treatment (Fig. 5B,  panel a, lane 3; panel b, lane 3). These observations strongly support the conclusion that HA-mediated Gab-1 phosphorylation is CD44-dependent. It is also noted that transfection of MDA-MB-231 cells with RBcDNA of ROK does not alter the basal level of Gab-1 serine and threonine phosphorylation in transfectants without HA treatment (Fig. 5B, panel a, lane 4; panel b, lane 4) or transfectants pre-treated with anti-CD44 followed by HA treatment (Fig. 5B, panel a, lane 6; panel b,  lane 6). However, overexpression of the RB of ROK appears to greatly reduce the ability of Gab-1 to respond to HA-mediated serine (Fig. 5B, panel a, lane 5) and threonine phosphorylation (Fig. 5B, panel b, lane 5). These findings suggest that ROK is closely involved in the regulation of HA/CD44-mediated Gab-1 phosphorylation in vivo.
Many reports have shown that certain adaptor molecules such as Gab-1 are involved in bringing p110 to the plasma membrane during cellular signaling (46 -51). Gab-1 contains several functional domains, including the pleckstrin homology (PH) domain (Fig. 7A, panel a), which may mediate association with the submembrane region of the cell via protein-protein or protein-lipid interactions (35,36). To determine whether the PH domain of Gab-1 is involved in a direct interaction with CD44, we have prepared a purified recombinant Gab-1-PH fragment (conjugated to Sepharose beads) and FLAG-tagged cytoplasmic domain of CD44 (FLAG-CD44cyt) fusion protein for in vitro binding analysis. Our results show that FLAG-CD44cyt is tightly associated with Gab-1-PH-His fragmentconjugated beads (Fig. 7A, panel b, lane 2) but not with Hisconjugated beads (Fig. 7A, panel b, lane 1). These observations establish the fact that the PH domain of Gab-1 may be responsible for the recognition of CD44 in vitro.

FIG. 7. Interaction between the cytoplasmic domain of CD44 (CD44cyt) and Gab-1-PH fragment in vitro (A) and detection of CD44v3-Gab-1⅐PI 3-kinase complex in Gab-1-PH cDNA-transfected/vector-transfected cells (B). In
containing the PH domain acts as a potent competitive inhibitor for endogenous intact Gab-1 binding to CD44v3 in vivo; it also functions as a strong dominant-negative mutant for blocking CD44-Gab-1-mediated PI 3-kinase membrane localization.
In addition, we have investigated the effects of ROK-RB overexpression (by transfecting cells with the RBcDNA of ROK) on CD44v3 interaction with Gab-1 and PI 3-kinase. Using anti-CD44v3 immunoprecipitation of MDA-MB-231 cells followed by immunoblotting with anti-Gab-1, we have confirmed that both CD44v3 and Gab-1 are physically linked as a complex in vector-transfected cells without HA treatment (Fig. 8A, panels e and f, lane 1). However, HA treatment of vector-transfected cells appears to promote a significant recruitment of Gab-1 into CD44v3 (Fig. 8A, panels e and f, lane 2). Moreover, we have used anti-CD44v3-mediated immunoprecipitation followed by immunoblotting with various anti-PI 3-kinase antibodies (e.g. anti-␣, ␤, ␥, and ␦ forms of p110 catalytic subunit of PI 3-kinase). Our results show that a relatively low level of some PI 3-kinase (e.g. p110␥ and p110␦ forms but not p110␣ and p110␤) is co-precipitated with CD44v3 in vector-transfected cells without HA treatment (Fig. 8A, panels a-f, lane 1). However, HA treatment of vector-transfected cells appears to enhance the recruitment of PI 3-kinase (including p110␣, p110␤, p110␥, and p110␦) into CD44v3 (Fig. 8A, panels a-f, lane 2). We have also determined that the basal level of Gab-1⅐PI 3-kinase (p110␥ and p110␦) association with CD44v3 is greatly reduced in the RBcDNA of ROK-transfected cells (without HA treatment) (Fig. 8A, panels a-f, lane 3). Most noticeably, overexpression of ROK-RB in MDA-MB-231 transfectants abolishes the ability of HA to recruit Gab-1-linked PI 3-kinase (Fig. 8A, panels a-f, lane 4). It is therefore possible that HA/ CD44-activated ROK plays an important role in the recruitment of Gab-1⅐PI 3-kinase complexes to certain plasma membrane proteins such as CD44v3 in breast tumor cells.

Effects of RhoA-activated ROK and Gab-1-linked PI3 Kinase on AKT Activation, Cytokine (M-CSF) Production, and Breast
Tumor Cell Growth and Invasion-Activation of AKT by growth factors is known to be mediated by PI 3-kinase (60 -62). Most importantly, AKT acts as an important mediator of many cell survival-signaling pathways and is often overexpressed/ amplified in tumor progression (60 -62). In this study we have investigated the possible role of ROK and PI 3-kinase in regulating AKT-1 activation in breast tumor cells during HA-CD44 signaling. Specifically, we have determined that AKT-1 kinase is significantly stimulated by HA treatment in vector-transfected cells (detected by anti-phospho-AKT-1 (Fig. 9, panel a, lane 1) followed by reblotting with anti-AKT-1 antibody) (Fig.  9, panel b, lane 2) as compared with the level of AKT-1 phosphorylation in vector-transfected cells without any HA treatment (Fig. 9, panels a and b, lane 1). Moreover, treatment of vector-transfected cells with a PI 3-kinase inhibitor (LY294002) results in down-regulation of AKT-1 activation during the response to HA treatment (Fig. 9, panels a and b,   lanes 3 and 4). Furthermore, HA-activated AKT-1 phosphorylation is also greatly reduced in the RBcDNA of ROK-transfected cells (Fig. 9, panels a and b, lanes 5 and 6) comparing with AKT-1 phosphorylation in vector-transfected cells treated with HA treatment (Fig. 9, panels a and b, lanes 1 and 2). These observations strongly support the conclusion that both PI 3-kinase signaling and ROK activation are important upstream activators for AKT signaling required for HA/CD44-mediated tumor cell survival.
As part of our effort to identify new effector functions influenced by HA/CD44-mediated ROK and PI 3-kinase signaling in breast tumor cells, we decided to examine the production of M-CSF, which is a homodimeric cytokine in the colony-stimulating factor family (63). In breast cancers, M-CSF serum levels correlate with prognosis, activity, and invasiveness of the disease (64). Here, we have found that HA promotes M-CSF production in untransfected MDA-MB-231 cells (Table IA). It is also noted that the amount of M-CSF production detected in the media of untransfected MDA-MB-231 cells pre-treated with rat anti-CD44 (but not normal rat IgG) followed by HA treatment was significantly reduced (Table IA). These results support the notion that the stimulation of M-CSF production by HA is CD44-dependent. Furthermore, treatment of untransfected cells with a PI 3-kinase inhibitor, LY294002, blocked HA-mediated M-CSF production (Table IB). Moreover, our data show that HA/CD44-activated and PI 3-kinase-regulated M-CSF production are significantly inhibited in MDA-MD-231 cells transfected with the RBcDNA of ROK as compared with that detected in vector-transfected cells (Table IC). Together, these findings suggest that both RhoA-ROK and PI 3-kinase activation are closely involved in HA/CD44-mediated M-CSF production in breast tumor cells.
Furthermore, using in vitro growth and invasion assays, we have found that CD44v3-containing MDA-MB-231 cells undergo active cell growth (Table IIA) and invasion (Table  IIB). HA activates both breast tumor cell growth (Table IIA) and invasion (Table IIB). In contrast, MDA-MB-231 cells pretreated with rat anti-CD44 (but not normal rat IgG) followed by HA addition greatly reduced their ability to undergo growth and invasion (data not shown). These observations suggest that HA-mediated tumor cell growth and invasion is also CD44-dependent. Treatment of MDA-MB-231 cells with various agents such as the PI3 inhibitor (LY294002) causes a significant inhibition of HA-mediated tumor cell growth (Table IIA) and invasion (Table IIB). Importantly, transfection of MDA-MB-231 cells with the dominant-negative form of ROK (ROK-RB) cDNA (but not vector alone) also effectively blocked HA/CD44-mediated breast tumor cell growth (Table  IIA) and invasion (Table IIB). Together, these findings support the notion that HA-mediated CD44 signaling is functionally linked to RhoA-ROK and Gab-1-linked PI 3kinase activation required for metastatic breast tumor progression. DISCUSSION CD44 belongs to a family of transmembrane glycoproteins (e.g. CD44s (standard form), CD44E (epithelial form), and CD44v (variant isoforms)), which are often overexpressed in a variety of human solid neoplasms, including breast cancers (1)(2)(3)(4)(5)(6). In particular, the CD44v3-containing isoforms are expressed preferentially in highly malignant breast carcinoma tissue samples (1)(2)(3)(4)(5)(6). In breast tumor cells, CD44v3 has a heparin sulfate addition site in the membrane-proximal extracellular domain of the molecule that confers the ability to bind vascular epithelial growth factor, but not fibroblast growth factor-2 (13). The attachment of vascular epithelial growth factor to the heparin sulfate sites on CD44v3 may be respon-  (lanes 1-4) or ROK-RBcDNA (lanes 5 or 6) treated with HA (lanes 2, 4, and 6) or no HA (lanes 1, 3, and 5) in the presence (lanes 4 and 6) or the absence of LY294002 (lanes 1, 2, 3 sible for the onset of breast tumor-associated growth, invasion, and angiogenesis. The level of CD44v3 isoform expression often increases as the histological grade of each of the breast tumors progresses. In fact, there is a direct correlation between CD44v3 isoform expression and increased histological grade of the malignancy (1-6). These lines of evidence suggest that expression of certain CD44v3 isoform(s) may be an accurate predictor of eventual survival (including nodal status, tumor size, and grade) during breast cancer progression (4). It is now well established that the extracellular matrix component, hyaluronan (HA), is one of the primary ligands recognized by surface CD44 molecules, including CD44v3 (21)(22)(23)(24). Both HA and CD44v3 appear to be overexpressed at sites of tumor attachment and are known to be involved in tumor cell-specific properties (e.g. tumor cell growth, survival, migration, and invasion) (21)(22)(23)(24). In this study we have found that CD44v3 is expressed as a 85-kDa polypeptide in breast tumor cells (MDA-MB-231 cells (Fig. 1)). The question of which oncogenic pathways are directly involved in regulating HA-activated and CD44v3-specific breast tumor cell behaviors is addressed in this study.
The invasive phenotype of breast tumor cells, characterized by HA-CD44v3-mediated "invadopodia" formation (57), matrix metalloproteinase-9 activation (57), and tumor cell migration and invasion (21-24, 56, 57), has been linked to cytoskeletal function, a process in which the small GTP-binding proteins such as RhoA are known to play important roles (22). The activities of members of the RhoGTPases (including RhoA) are often regulated by guanine nucleotide exchange factors (GEFs) that contain a dbl homology (DH) domain (33,34). In recent years, a RhoA-specific GEF (p115RhoA) molecule containing several functional domains (an N-terminal RGS domain, a dbl homology (DH) domain, and a PH domain) have been identified and characterized (33,34). In particular, the DH domain of p115RhoGEF by itself exhibits GDP/GTP exchange activity for RhoA and plays an important role in RhoGTPase signaling (33,34). Unique to p115RhoGEF, compared with other GEFs for small GTPases, is the presence of an RGS domain. A number of laboratories have reported that the N-terminal RGS domain interacts with the ␣ subunit (␣ 12/13 ) of the G-protein (33,34). In fact, the RGS sequence has been shown to play an important role in regulating p115RhoGEF-G␣ 12/13 interactions during Gprotein-coupled receptor signaling (33,34). Our laboratory has demonstrated that p115RhoGEF is expressed in MDA-MB-231 cells (Fig. 1). In addition, we have presented new evidence for a close, physical interaction between p115RhoGEF and the transmembrane glycoproteins, CD44v3 isoform (Fig. 1). Most importantly, our results show that the binding of HA to CD44v3 promotes p115RhoGEF-catalyzed RhoA activation (Fig. 2). Therefore, it is likely that CD44v3 serves as an activator of p115RhoGEF function. Clearly, these two proteins, CD44v3 and p115RhoGEF, are not only structurally linked, but also functionally coupled. It is also noted that two other RGScontaining RhoGEFs (e.g. PDZ-RhoGEF and/or leukemiaassociated RhoGEF) have been shown to activate RhoA signaling (34,58). The relationship between PDZ-RhoGEF and/or leukemia-associated RhoGEF in HA/CD44-mediated cellular events remains to be elucidated.
Rho kinase (ROK) stimulated by activated RhoA (GTPbound form of RhoA) appears to play a pivotal role in promoting cytoskeletal function and a variety of other cellular activities (22, 40 -45). In a previous study (22) we demonstrated that ROK phosphorylates the cytoplasmic domain of CD44v 3, 8 -10 isoform and up-regulates the interaction between the CD44v 3, 8 -10 isoform and the cytoskeletal protein, ankyrin, during HA/CD44-regulated tumor cell migration. More recently, we have shown that the PH domain of ROK is the primary binding site for CD44 (45). Most importantly, our results indicate that CD44 interaction with ROK plays a piv-   54 58 otal role in the phosphorylation of inositol triphosphate receptor (IP 3 receptor) leading to an increase of IP 3 binding followed by a rapid stimulation of Ca 2ϩ mobilization during HA-mediated cell migration (45). Therefore, we believe that ROK plays an important role in HA/CD44-mediated functions (45). Structurally, ROK is composed of catalytic, coiled-coil, RB, and PH domains (37)(38)(39)(40). A number of substrates, including myosinbinding subunit of myosin phosphatase (41), calponin (42), adducin (43), Lin 11/Isl-1/Mec-3 (LIM) kinase (44), IP 3 receptors (45), and CD44 (22), have been shown to be phosphorylated by ROK. In MDA-MB-231 cells, we have identified a 160-kDa protein as a ROK molecule (Fig. 3). In an effort to identify possible new cellular substrates for ROK, we have focused on the multiple docking protein, Gab-1, that is often tyrosinephosphorylated by receptor tyrosine kinases such as hepatocyte growth factor receptor, c-Met (65), and epidermal growth factor receptors (66). In this study we have demonstrated that ROK (by binding to activated RhoA) is capable of inducing marked serine and threonine phosphorylation of Gab-1 (isolated from MDA-MB-231 cells) in vitro (Fig. 4). The ability of ROK to phosphorylate Gab-1 in the presence of unactivated RhoA appears to be greatly reduced (Fig. 4). These results clearly indicate that ROK acts as one of the downstream effectors of RhoA signaling and utilizes Gab-1 as one of its cellular targets in vitro. Overexpression of the RB domain of ROK has been found to function as a dominantnegative form of ROK, which inhibits ROK activity, ROKmediated protein phosphorylation, and HA/CD44-mediated breast tumor progression (22). This information has prompted us to examine whether ROK plays a role in regulating Gab-1 phosphorylation and Gab-1-mediated cell signaling in HA/CD44-mediated breast tumor cell activation in vivo. Transfection of MDA-MB-231 cells with the RB domain cDNA of ROK (Fig. 5), which effectively competes for endogenous activated RhoA binding to ROK and inhibits the ability of ROK to phosphorylate Gab-1 during HA-CD44 signaling (Fig. 5), strongly suggests that RB of ROK is a potent inhibitor of ROK-mediated Gab-1 phosphorylation in vivo. A number of studies have shown that tyrosine-phosphorylated Gab-1 is involved in PI 3-kinase binding, membrane translocation, and activation (47,48,67,68). The question of whether HA/CD44-activated RhoA-ROK serine/threonine phosphorylation of Gab-1 plays a role in regulating PI 3-kinase signaling is addressed in this study. Previous reports have indicated that co-expression of multiple PI 3-kinase isoforms (e.g. ␣, ␤, ␥, and ␦ forms of p110) occurs in many different cells (46, 49 -51). The p110␣, p110␤, and p110␥ forms are expressed in all tissue, whereas p110␦ is preferentially expressed in leukocytes (46, 49 -51). The ␣, ␤, and ␦ forms of the p110 catalytic subunit belong to the class IA PI 3-kinase known to transmit signals from tyrosine kinase-coupled receptors (66,67). These class IA PI 3-kinases (e.g. p110␣, p110␤, and p110␦) are often associated with certain regulatory units (e.g. p85, p55, or p50) that bind to specific phosphorylated tyrosine residues in receptor proteins or other adaptor molecules such as Gab-1 (66,67). The class IB PI 3-kinase (the ␥ form of p110 catalytic subunit) can be activated by G-protein-coupled receptors (46, 49 -51, 67). This p110␥ form interacts with a unique regulatory protein, p101 (but not p85, p55, or p50) (69). Here, we have found that multiple isoforms of PI 3-kinase (as detected by immunoblot with specific anti-␣, ␤, ␥, and ␦ forms of the p110 catalytic subunit) are also co-expressed in MDA-MB-231 cells (Fig. 6). Apparently, the status of Gab-1 serine/threonine phosphorylation (either highly phosphorylated or unphosphorylated form) does not affect Gab-1 association with any particular isoform of PI 3-kinase (e. g. p110␣, p110␤, p110␥, and p110␦). These re-sults suggest that the physical association between Gab-1 and PI 3-kinase is always present and serine/threonine phosphorylation of Gab-1 plays a small role (if any) in influencing Gab-1⅐PI 3-kinase complex formation (Fig. 6). The question of whether Gab-1 (phosphorylated or unphosphorylated form) directly binds to p110 subunits or selectively interacts with various regulatory domains (e.g. p85, p55, p50, or p101) in HA/CD44-mediated RhoA-ROK signaling awaits further analyses.
In addition, using two recombinant proteins (the PH domain of Gab-1 and the cytoplasmic domain of CD44), we have demonstrated that the PH of Gab-1 is responsible for the recognition of CD44 in vitro (Fig. 7A). We have also provided new evidence that phosphorylation of intact Gab-1 by RhoA-activated ROK enhances its binding to CD44 (Fig. 4B). Furthermore, we have found that transfection of MDA-MB-231 cells with Gab-1-PH cDNA (Fig. 7B) effectively blocks endogenous Gab-1 association with CD44v3 and the subsequent recruitment of signaling molecules (e.g. PI 3-kinase) to CD44v3-containing membrane (Fig. 7B). These findings further support our conclusion that Gab-1-PH acts as a potent competitive inhibitor, which is capable of interfering with endogenous Gab-1-CD44v3 interaction in vivo. These results are consistent with our previous study showing a sequence adjacent to the Nterminal region of PH domain (the PHn-CC-EX domain) of Tiam1 (an Rac1-specific guanine nucleotide exchange factor (GEF)), which is involved in the direct binding to CD44v3 isoform during HA-stimulated Rac1 signaling and cytoskeleton-mediated tumor cell migration (24). Our recent study also shows that the PH domain of ROK binds to CD44v10 in endothelial cells and this ROK-PH interaction with CD44v10 regulates HA-mediated Ca 2ϩ signaling and endothelial cell migration (45). It is therefore, apparent that the close interaction between CD44 isoforms and certain PH domain-containing molecules (e.g. Gab-1, Tiam1, and ROK) plays an important role in HA signaling. Moreover, we have observed that the Gab-1 (highly serine/threonine-phosphorylated form of Gab-1)⅐PI 3-kinase complex becomes preferentially recruited into CD44v3-containing membranes in vector-transfected cells treated with HA (Fig. 8A). In contrast, in other situations, the Gab-1 (with a minimal level of serine/threonine phosphorylation)⅐PI 3-kinase complex fails to become membrane-localized with CD44v3 in vector-transfected cells with no HA treatment (Fig. 8A) or in the RBcDNA of ROK-transfected cells with or without HA treatment (Fig. 8A). Thus, our results provide evidence for HA/CD44-activated RhoA-ROK to play a role in the membrane translocation of phosphorylated Gab-1⅐PI 3-kinase complexes into CD44v3. The recruitment of phosphorylated Gab-1⅐PI 3-kinase into CD44v3-associated membranes may bring the p110 catalytic subunits into close proximity to the inositol lipid substrates and thereby facilitate the onset of enzymatic activity.
Previously, PI 3-kinase has been shown to be activated by guanine-nucleotide-binding proteins (70,71). PI 3-kinase can also be activated via its binding to the active conformation of Ras during the membrane localization of p110⅐p85 complex (72,73). Our analyses indicate that ROK-mediated Gab-1 phosphorylation and Gab-1⅐PI 3-kinase membrane localization are accompanied by a robust stimulation of PI 3-kinase activity (in particular, ␣, ␤, and ␥ forms, but not the ␦ form of p110 catalytic subunit) in vector-transfected cells treated with HA (Fig. 8B). Under physiological conditions, hyaluronan often exists as a high molecular mass polymer (Ͼ10 6 Da). However, generation of hyaluronan fragments may occur during tumor progression. Both HA and HA fragments of certain size (3-25 disaccharide units) have been found to either promote up-regulation or down-regulation of CD44-specific biological events (18). Because we have used HA of large size (Ͼ10 6 Da) for all experiment described in this study, it is possible that large HA fragments preferentially activate p110␣, p110␤, and p110␥ forms but not the p110␦ form in MDA-MB-231 cells (Fig.  8B). The question of whether one or more certain small size HA fragments are capable of promoting p110␦ and/or the same p110␣, p110␤, and p110␥ forms as the native large molecular weight hyaluronan used in this study is currently undergoing investigation in our laboratory.
Furthermore, our results indicate that the reduction of Gab-1 threonine/serine phosphorylation and the lack of Gab-1⅐PI 3-kinase membrane localization in the RBcDNA of ROKtransfected cells (Fig. 7B) are closely associated with an inhibition of certain PI 3-kinase activity (e.g. ␣ and ␤ forms but not the ␥ form of p110 catalytic subunit) during HA treatment (Fig.  8). Apparently, the activation of both ␣ and ␤ forms of p110 catalytic subunits is ROK-dependent, whereas the stimulation of the ␥ form of p110 catalytic subunit is ROK-independent (Fig. 8). Moreover, our observations, showing that ROK is required for the recruitment of PI 3-kinase p110␥ but ROK is not required for HA-mediated activation of this form (in terms of PIP3 production), are noteworthy. Activation of the catalytic subunit p110␥ is known to be mediated by the regulatory subunit p101. It is possible that membrane localization of PI 3-kinase p110␥ is ROK-dependent, whereas p101-mediated activation of PI 3-kinase p110␥ is ROK-independent during HA-CD44 signaling. Characterization of p101-mediated activation of p110␥ (in a ROK-independent process) during HA-CD44 signaling awaits further investigation. The fact that HA-mediated PI 3-kinase p110␥ recruitment and activation can be readily blocked by anti-CD44 antibody pretreatment of vector and ROK-RBcDNA-transfected cells (data not shown) suggests that CD44 is involved in this HA signaling event. Of course, we cannot preclude the possibility that other HA receptors (e.g. receptor for HA-mediated motility (RHAMM)) are also capable of generating PI 3-kinase signaling in either a ROK-dependent or independent manner. The selective activation of ␣, ␤, and ␥ forms (but not the ␦ form of p110 catalytic subunit) may significantly influence the biological outcomes of HA-CD44mediated oncogenic events in breast tumor progression.
One of the well established effector molecules for PI 3-kinase is AKT, which is known as an important mediator of many cell survival-signaling pathways (60 -62). Recent evidence has suggested that AKT signaling plays a central role in human malignancy. AKT is a putative oncogene encoding a member of a subfamily of protein-serine/threonine that includes the protein kinase B family (60 -62). In humans, at least three AKT genes (designated AKT-1, AKT-2, and AKT-3) have been identified (62). It is still unclear how much functional redundancy exists among these three AKTs. In this study we have found that AKT-1 is expressed in MDA-MB-231 cells (Fig. 9), and HA-CD44 interaction promotes AKT-1 activation (Fig. 9). The facts that overexpression of ROK-RB (by transfecting MDA-MB-231 cells with ROK-RBcDNA) impairs HA-mediated AKT-1 activation and that a PI 3-kinase inhibitor (LY294002) also blocks AKT-1 activation (Fig. 9) suggest that both ROK and PI 3-kinase are located upstream of AKT-1 and are involved in regulating HA/CD44-mediated AKT signaling required for antiapoptotic processes and tumor cell survival. Our results are consistent with a previous report showing that HA-CD44 interaction is involved in PI 3-kinase/AKT cell survival signaling in human lung carcinoma cells (27). Perturbation of HA-CD44 binding effectively down-regulates PI 3-kinase/AKT activities and inhibits tumor cell growth (27).
A recent study indicates that breast tumor cells (e.g. MDA-MB-231 cells) are capable of inducing osteoclastogenesis by secreting M-CSF, which is a homodimeric cytokine of the colony-stimulating factor family (63). The addition of M-CSF to osteoclast precursors induces the expression of RANK (the receptor for RANKL, the receptor activator of nuclear factor B ligand), which in turn interacts with RANKL and induces differentiation of cells in the macrophage/osteoclast lineage (74 -78). In the presence of M-CSF, the RANK/RANKL/osteoprotegrin axis mediates osteoclast formation and activity and thereby bone resorption (63, 74 -78). These findings suggest that the stimulation of osteoclastogenesis by M-CSF-producing breast tumor cells is closely involved in osteolytic metastases. In MDA-MB-231 cells we have found that HA promotes M-CSF production in a CD44-dependent manner ( Table I). Inhibition of both ROK and PI 3-kinase results in significant reduction in M-CSF production (Table I). These findings suggest that these two signaling molecules (e.g. ROK and PI 3-kinase) are functionally linked to HA/CD44-mediated M-CSF production during breast tumor progression. We have also demonstrated that overexpression of ROK-RB (by transfecting MDA-MB-231 cells FIG. 10. A proposed model for the interaction between CD44v3-mediated RhoA-ROK activation and Gab-1-linked PI 3-kinase signaling during breast tumor progression. HA-CD44 interaction is tightly coupled with p115RhoGEF in a complex that can up-regulate RhoA signaling and Rho kinase (ROK) activity (Step 1). Activated ROK then phosphorylates certain cellular proteins, including the linker molecule, Gab-1 (Step 2). Most importantly, phosphorylation of Gab-1 by ROK promotes the membrane localization of Gab-1 and PI 3-kinase to CD44 and activates certain isoforms of PI 3-kinase to convert PtdIns(4,5)P 2 to PtdIns(3,4,5)P 3 (Step 3) leading to AKT activation, cytokine (M-CSF) production, and tumor cell behaviors (e.g. tumor cell growth, survival, and invasion) required for breast tumor progression.
with ROK-RBcDNA) exhibits a dominant-negative effect on both p110␣ and p110␤ forms of PI 3-kinase (but not the p110␥ or p110␦ forms of PI 3-kinase) (Fig. 8B) leading to an inhibition of HA-mediated M-CSF production in MDA-MB-231 cells (Table I). These findings suggest that both p110␣ and p110␤ forms of PI 3-kinase (and to a lesser extent the p110␥ or p110␦ forms of PI 3-kinase) are most likely involved in the regulation of HA-mediated M-CSF production. Of course, we cannot rule out the possibility that other PI 3-kinases (e.g. p110␥ or p110␦) may also be involved in M-CSF production in an HA-independent manner. The use of small interference RNA to selectively knock down PI 3-kinase (e.g. p110␣, p110␤, p110␥, or p110␦ forms) and thereby interfere with PI 3-kinase isoform-mediated functions and breast tumor progression is currently being investigated in our laboratory.
Finally, we have found that treatment of MDA-MB-231 cells with the PI 3-kinase inhibitor LY294002 (Table II) (Table II) induces reversal of tumor cell-specific phenotypes such as tumor cell growth and invasion (Table II). Together, these findings strongly suggest that HA/CD44-mediated ROK activation and PI 3-kinase signaling are closely coupled during the onset of tumor-specific behavior in MDA-MB-231 cells.
As summarized in Fig. 10, we propose that CD44 is tightly coupled with RhoGEF in a complex that can up-regulate RhoA signaling and Rho kinase (ROK) activity (Step 1). Activated ROK then phosphorylates certain cellular proteins, including the linker molecule, Gab-1 (Step 2). Most importantly, phosphorylation of Gab-1 by ROK promotes the membrane localization of Gab-1 and PI 3-kinase to CD44 and activates certain isoforms of PI 3-kinase to convert PtdIns(4,5)P 2 to PtdIns(3,4,5)P 3 (Step 3) leading to AKT activation, cytokine (M-CSF) production, and tumor cell behaviors (e.g. tumor cell growth, survival, and invasion) required for breast tumor progression.