Met-Independent Hepatocyte Growth Factor-mediated regulation of cell adhesion in human prostate cancer cells

Prostate epithelial and stromal cell lines were maintained in T-media with 5% fetal bovine serum, at 37°C with 5% CO₂. Primary cultures of prostate stromal cells were derived from the tissue surrounding prostatic adenocarcinomas, as described by Ozen et al. [79]. Conditioned media were prepared by adding fresh media without serum when cells reached 80% confluence and removing it 48 h later. For signaling assays, cells were starved in RPMI-1640 phenol red-free medium (Life Technologies, Inc.) un-supplemented with serum. Laminin-1 (a kind gift of Roy Ogle at the University of Virginia) was purified from Engelbreth-Holme-Swarm (EHS) tumors according to the method of Davis et al., based on the protocol of Kleinman et al. [80, 81]. Hepatocyte Growth Factor (HGF) and all other chemicals were purchased from Sigma (St. Louis, MO). Anti-FAK and HGF antibodies were from Sigma (St. Louis, MO) and Santa Cruz Biotechnology (Santa Cruz, CA). Antibodies to nucleolin were purchased from Santa Cruz Biotechnology or received as a gift from Dr. Deng at Pittsburgh Medical Center. Met antibodies, as well as cdc-42 and Rac immuno-precipitation reagent were purchased from both UpState Biotechnology (Lake Placid, NY) and Transduction Laboratories (Lexington, KY). Phospho-tyrosine antibodies were from Transduction Laboratories (Lexington, KY). Phospho-Akt (Ser 473) antibody and Akt antibody were purchased from Cell Signaling Technology Inc. (Beverly, MA). Anti-β actin antibody was from Abcam Inc. (Cambridge, MA). All secondary-conjugated antibodies were from Jackson Immunochemicals (West Grove, PA).

Total cellular RNA was isolated with RNA-STAT (BioTec-X, Houston, TX). 5 μg of RNA was reverse transcribed using the OmniScript RT Kit (Qiagen, Inc., Valencia, CA). The primers used for met amplification were as follow: F-met 5'-GGTTGCTGATTTTGGTCAT-3' and B-met 5'-TTCGGGTTGTGGAGTCTT-3'.

Attachment assays were performed as previously described in Edlund et al., and Vafa et al., [68, 82]. Cell lines were grown to confluence, trypsinized, and re-plated (1:8) on tissue culture dishes, where they were allowed to grow for another two days before being lifted after a brief treatment with 10 mM EDTA, 20 mM Hepes buffer in T-media. After neutralizing the EDTA with CaCl2 and MgSO4, the cells were washed with T-media containing 0.1% BSA. Cells were placed on laminin1-coated dishes, allowed to adhere for 30 to 90 min with or without addition of either HGF or function blocking integrin antibodies, and then fixed in para-formaldehyde (3.8%). The percentage of spread cells was scored for each cell line, based on cell membrane protrusion (lamellipodia and/or filopodia), and all values were normalized to control cells treated identically except for being subjected to conditioned media or growth factors. This normalization step was necessary because of the differences in speed of attachment between the cell lines. Cell growth was quantified using MTT [83, 84].

ELISAs were performed as suggested in Pharmigen Research Products Catalog, 1999. Briefly, rabbit anti-human HGF antibodies were diluted to a concentration of 1 μg/ml in 0.1 M Na₂HPO₄ and 0.1 M NaH₂PO, pH 9.0. Wells of a 96 well ELISA plate (Costar) were filled with 50 μl, sealed with parafilm, and incubated overnight at 4°C. Plates were then brought to room temperature and antibodies captured and removed. 200 μl of blocking buffer (10% FBS in PBS) was added to each well and the plate incubated at room temperature for 2 hours. The plate was washed three times with PBS/Tween 20 (0.05%) (Sigma). 100 μl of standard or sample diluted in blocking buffer/Tween 20 (0.05%) was added to each plate, which was then sealed with parafilm and incubated at 4°C overnight. The following day, the plate was washed four times with PBS/Tween. 100 ul of the detection antibody, Goat anti-human HGF antibody (Sigma), diluted to a concentration of 0.01 μg/ml in blocking buffer/Tween 20, was added to each well. After one hour incubation at room temperature, and 4 washes with PBS/Tween 20, 100 μl of avidin-horseradish peroxidase conjugated mouse anti-goat IgG (Jackson ImmunoReseach) diluted 1:1000 in blocking buffer/Tween 20 was added to each well. After 30 minutes incubation at room temperature and five washes with PBS/Tween 20, 200 μl of Sigma Fast OPD solution (Sigma) was added to each well. After 20 minutes, the plate was read at a wavelength of 405 nm using a microplate reader (Molecular Devices, Sunnyvale, CA). Data were analyzed using the Molecular Devices SOFTmax program.

Cell migration assays were preformed according to product instructions (CSM Inc. Phoenix, AZ). Chilled cell manifolds were placed on Teflon-printed, precoated microscope slides, subdivided into 10 wells, and filled with ice-cold media. One μl of cell suspension (2500 cells) was added to each chamber and allowed to precipitate by gravity and adhere to the coverslip. After two hours, the manifold was moved to a cell incubator (37°C under 5% CO₂) and allowed to reach growth temperature for 4 hours, at which point the cell sedimentation manifold was removed and the coverslip submerged in media, to which HGF was added. The area covered by cells was recorded at 2 hrs and subsequently every 24 hours. The 2 hr area was used as a reference point for all succeeding measurements. Results are presented as increases relative to this area.

The HGF and major co-immunoprecipitated product from C4-2 cell lysates were excised from the gel and transferred to a siliconized tube, washed and destained in 50% methanol overnight. The gel pieces were dehydrated in aceto-nitrile, rehydrated in 10 mM dithiothreitol (DTT) in 0.1 M ammonium bicarbonate, and reduced at room temperature for 30 minutes. The DTT solution was removed and the samples alkylated in 50 mM iodoacetamide, in 0.1 M ammonium bicarbonate, for 30 minutes at room temperature. Samples were then dehydrated again in aceto-nitrile, rehydrated in 0.1 M ammonium bicarbonate, dehydrated in aceto-nitrile and completely dried by vacuum centrifugation. Finally, samples were rehydrated for 10 minutes in 20 ng/ml trypsin in 50 mM ammonium bicarbonate on ice. Any excess trypsin solution was removed, and 50 mM ammonium bicarbonate added. Samples were digested overnight at 37°C and the sequences of generated peptides were identified by mass specectrometry.

All cells were lysed in ice-cold lysis buffer (20 mMTris PH7.4, 40 mM NaCl, 20 mM beta-glycerophosphate, 2 mM EGTA, 1 mM sodium orthovanadate, 2 mM DTT, 2 mM PMSF, 1μg/ml aprotinin, and 1μg/ml aprotinin). The Map kinase assay was done with an assay kit (Upstate, Lake Placid, NY). The kinase reaction was started by addition of kinase reaction buffer that contains 2 mg/ml dephosphorylated myelin basic protein for each substrate, 20 mM MOPS (pH 7.2), 25 mM beta-Glycerophosphate, 5 mM EGTA, 0.4 mM MnCl₂, 1 mM sodium orthovanadate, 1 mM dithiothreitol, 75 mM MgCl₂, and 500 μM ATP. To prevent effects from other unknown kinases in the lysate, 20 μM PKC inhibitor peptide, 2 μM PKA inhibitor peptide, and 20 mM R24571 compound were added to the kinase reaction buffer. The reaction was incubated for 20 min at 30°C, terminated by the addition of the LDS sample buffer and loaded as aliquots for SDS-PAGE and immunoblot analyses. Membrane enriched fractions were purified as previously described [85].

After centrifugation for 15 min at 15000 rpm in 4°C, the lysate supernatant was collected. Protein concentration was determined by BRC assay (BioRad, Hercules, CA). Immunoblotting was performed using the NOVEX (Invitrogen, Carlsbad, CA) system. Briefly, 7.5 μg of cell extracts and Erk kinase assay products were separated on 4–12% Tris glycine PAGE gels and transferred onto a PVDF membrane (Immoblin-P, Millipore, Billerica, MA). The membrane was blocked 1 h at room temperature with TBST blocking buffer [50 mM Tris-HCl (pH 8.0), 150 mM NaCl, 0.05% Tween 20, and 5% nonfat milk]. The membrane was incubated for 1 hr at room temperature with primary anti-phosph MBP polyclonal antibody, anti-Erk 1,2 antibody, anti-PKB antibody, and anti-phosph PKB antibody (Ser 473) in PBSN blocking buffer. A secondary antibody (horseradish peroxidase-anti-rabbit or mouse antibody) (Amersham Bioscience, Inc., Piscataway, NJ) at a 1:5000 dilution was used in PBSN blocking buffer and incubated for 1 hr at room temperature. ECL plus (Amersham Bioscience, Inc.) reagent was used for detection.

Results were analyzed for statistical significance using the nonparametric Mann-Whitney U test, with significance at P < 0.05.

Previously, our laboratory characterized cell responses to media conditioned by primary prostate stromal cells, and found differences between the responses of parental, non-metastatic, human prostate LNCaP cells and those of its lineage-derived, metastatic C4-2 subline [86]. Prostate SCM collected in serum-free conditions had little effect on LNCaP cells, but increased cell spreading of C4-2 cells on laminin-1 substrates by 150–200% (Figure 1A) [68]. C4-2 cells were not the only prostate cancer cells to respond this way to SCM; cell spreading also increased following SCM treatment of DU145 (brain metastatic human prostate cell line), PC3, and PC3M (a bone metastatic human prostate cell line of shared lineage). The SCM collected from five different primary cultures and three different hosts all had similar effects when incubated with C4-2 cells (data not shown). Many growth factors stimulate focal adhesion assembly and influence integrin activity [87], but we report here that these cell spreading effects of SCM can be ameliorated by anti-HGF antibodies (Figure 1B). Thus, the HGF in the SCM is a major regulator of cell adhesion and cell spreading.

To measure HGF amounts present in the SCM, we used ELISA's and found HGF to range from 14 to 24 ng/ml (Table 1). Cell spreading responses to HGF are concentration-dependent for both PC3 and C4-2 prostate cancer cell lines, but an effective dose 50 (EC50) for PC3 cells is achieved at a lower concentration (5 ng/ml) than for C4-2 cells (30 ng/ml) (Figure 2A). These differences in maximal stimulation could be due to synergism between HGF and other growth factors in the SCM, to the combined presence of both met and nucleolin in PC3 cells, and/or to differences in nucleolin and met affinities for HGF. HGF-enhanced C4-2 cell spreading was substrate dependent and observed only on laminin-1 (LM) substrate (Figure 2B), not on fibronectin (FN), collagen-1 (Col-1) or vitronectin (VN). C4-2 cell spreading and cell migration respond inversely to HGF treatment; that is, cell spreading is enhanced, whereas cell locomotion is decreased on laminin. The inhibitory effects of HGF on cell migration were observed at 10 ng/ml (Figure 3A and 3B). Like the cell spreading effects, migration inhibition was seen only on a laminin-1 substrate, not on FN. HGF migration responses could not be explained by cell proliferation, because C4-2 cell cycle progression appeared unaffected by HGF (data not shown). Furthermore, all cellular adhesion could be block by either α₆ or β₁ integrin function blocking antibodies and no changes in surface expression of laminin binding integrins were observed (data not shown).

Using RT-PCR, we assess met expression levels in prostate cancer cell lines of different metastatic potential (LNCaP, C4-2, PC3 and DU145). Met receptor transcript (262 bp) was present in DU145, PC3, PC3M and HeLa cells, but absent or at very low levels in both LNCaP and C4-2 cells (Figure 4A). These results were further confirmed by Western blot analyses. Doublets of c-Met were observed, with similar results for both the N-terminal and C-terminal Met antibodies, used to check for possible Met isoforms differing in their cytoplasmic tails (Figure 4B and 4C).

The HGF-Met signaling cascade is well described and known to regulate invasion and metastasis, as well as cell proliferation, survival, differentiation and branching morphogenesis. Interactions between active HGF and Met result in αβ heterodimer formation, trans-autophosphorylation, and the recruitment of signaling intermediates [21, 33]. We searched for Akt and Erk phosphorylation in both PC3 and C4-2 cells, and confirmed that C4-2 cells do not respond to HGF with a functional Met signaling system. Following HGF stimulation, Met is phosophorylated and activated only in PC3 cells, not C4-2 (Figure 5A and 5B). Furthermore, both Akt and Erk are activated in a time-dependent manner in PC3 cells, but not in C4-2 cells (Figure 5C). These findings disagree with a previous report of a weak response to HGF, which may have been due to activation of the PI3-kinase/Akt pathway by cell-substrate adhesion that was further enhanced by HGF/SF stimulation [88]. Our static stimulation of serum-starved cells reveals clear enhancement of both Akt and Erk in PC3, but not C4-2 cells. Since both RT-PCR and Western blot were negative for met/Met in the LNCaP progression model, their dose-dependent responses to HGF stimulation suggest a Met-independent mechanism. This is further shown by the activation of Rac upon HGF stimulation in the C4-2 cells (Figure 5D and 5E). Together, these results correlate well with our cell-spreading and increased membrane ruffling data and further argue for a Met-independent pathway for HGF response in C4-2 cells.

Immunoprecipitation of PC3 and C4-2 total cell lysates further confirms that Met is present in HGF-stimulated PC3 cells and absent in C4-2 cells (Figure. 6A). HGF was immobilized on agarose beads and incubated with cell lysates. Immunoprecipitated product was then separated and stained with an antibody against the Met cytoplasmic domain. Control beads pulled down no detectable Met protein from either PC3 or C4-2 lysates, as assayed by both silver staining and immunoblotting (Figure 6A and 6B). Note that the silver-stained, immunoprecipitated products of C4-2 cells contained two major bands of approximately 60 kDa and 100 kDa. The lower band represents HGF itself, as identified by Western blot (data not shown), whereas the higher band was micro-sequenced and identified as nucleolin, with an expected size of 98 kDa (Table 2).

Of the 98 kDa band isolated from total C4-2 cell lysates (by columns of HGF-coated agarose beads), a total of 24 peptides were sequenced, corresponding to approximately 30% of the protein sequence of nucleolin, with 100% identity match to the previously published protein sequence (NCBI#4885511; Table 2). The identity of the 100 kDa protein was further confirmed by Western blotting to nucleolin (Figure. 6C). No other sequence homologies were found, suggesting that nucleolin is the major protein in the 100 kDa band, and that nucleolin is the major binding partner for HGF in C4-2 cells. Most compellingly, antibodies against nucleolin were able to block the cell spreading phenotype observed in HGF-treated cells in our cell attachment assay system (Figure 6D and 6E); in other words, Met-negative C4-2 cells increased cell spreading behavior when treated with HGF, and this response could be abolished by the addition of antibodies against the nucleolin protein (or HGF or specific integrin subunits, data not shown). Although we do not see any changes in the total nucleolin expression levels in the cells, we do see an increase in the membrane-associated nucleolin in both the LNCaP and PC3 progression models (Figure 7). This strengthens the link between nucleolin and HGF function, at the same time that it argues for cell surface localization of the nucleolin protein during cancer progression.

A Met-independent pathway acting in the absence of Met and/or alongside Met could increase the variety of possible cell responses to HGF. Classically, the binding of HGF to Met induces receptor dimerization and phosphorylation of two conserved tyrosine sites; when these sites are mutated, mice show similar phenotypes to those with HGF or Met knockouts [89]. But C4-2 cells, which lack this signaling pair, could be revealing spreading behaviors due to low-affinity HGF sites that allow response to higher levels of HGF (compared to Met-expressing PC3 cells; Figure 2). Past research on HGF affinity has identified some binding sites in hepatocytes with 10-fold lower affinity [90]. HGF interactions with proteoglycans and CD44 is linked to enhanced Met signaling and possible Met autophosphorylation [91, 92]. Pollack et al. [93] suggest that in Madin Darby canine kidney (MDCK) cells, low-affinity binding of HGF in the presence of the Met receptor may alter downstream responses. Thus, we are certainly not the first to suggest either additional HGF receptors, or HGF binding to heparin and oligosaccharide signaling systems [94, 95]. Our study is unique, however, in our identification of the HGF binding partner nucleolin.

Nucleolin, originally called C23 [96], was first characterized as an abundant nuclear protein with a role in ribosome biogenesis [69, 70]. Because nucleolin synthesis correlates positively with cell growth, the especially high expression of nucleolin in tumor cells is not surprising [97]. Nucleolin is further a major regulatory and phosphorylation target following androgen treatment [98, 99]. Nucleolin's in vivo localization to the cell surface in aggressive tumor cells [77, 78, 99, 100] agrees with our in vitro finding of membrane localization in both LNCaP and PC3 progression model cell lines (Figure 7). This membrane localization has recently generated interest in the use of nucleolin as a tumor marker and therapeutic target. A nucleolin functional inhibitor called AGRO100 (a guanine-rich oligonucletide) has been successful in Phase I clinical trials, and Phase II trials are underway [77, 78, 101]. Further evidence of nucleolin's role in cancer progression comes from our unrelated, in vivo bio-panning study. While bio-panning with a 12-mer peptide phage-display library, a peptide (designated L13) was found to bind with high specificity to murine bone marrow. When human bone marrow endothelial cell lysates were passed across an L13 column, the major species eluted was found to be nucleolin ([102]and unpublished observations). Indeed, a tumor homing peptide, F3, that binds specifically to tumor endothelial cells was described by Christian et al. [75] as interacting directly with nucleolin. Together, there is now clear evidence of nucleolin's involvement in cancer cell behavior and response to HGF, although this protein's direct and/or indirect functions remain to be discovered.

Nucleolin is named for the fact that it makes up as much as 10% of the total nucleolar protein [103], but it also functions as a shuttle protein between the cytoplasm and the nucleus, is found in clusters associated with the actin cytoskeleton [104‐106], and is sensitive to cytochalasin D [104]. Nucleolin is present on the cell surfaces of a variety of cell types [71, 107, 108], and surface-expressed nucleolin appears to interact with an array of other proteins, including viral proteins during infections [109, 110], βFGF and Midkine [111], and laminin-1 [112, 113].

A three-way interaction between nucleolin, HGF, and integrins is suggested by our ability, in short-term assays, to inhibit HGF's cell spreading effects using anti-nucleolin antibodies (Figure 6D); Nucleolin inhibition, alone, however, in the absence of HGF stimulation, does not alter C4-2 laminin adhesion (data not shown). Also, nucleolin by itself is not able to sustain laminin adhesion, as we were able to completely block cell spreading under stimulated and un-stimulated conditions, using function-blocking antibodies to either β₁ or α₆ integrin (data not shown). Yu et al. [114] studied the cellular distribution of nucleolin after stimulation with ECM proteins, and report that, following laminin stimulation, nucleolin translocates from the cytoplasm to the nucleus and stimulates cell proliferation. Following HGF stimulation, we do not detect changes in laminin-binding integrin expression profiles (data not shown), but we have previously seen changes in integrin function (not expression) in these same prostate cells [68]. Also possible is the movement of additional nucleolin protein to the cell surface, providing direct increase in laminin adhesion [72, 112, 114]. Nucleolin may function as a generic GF-HS binding receptor, linking HGF and integrin function. Like syndecan, nucleolin has been found to bind HGF's heparin-binding domain, thereby sequestering HGF on the cell surface. In our studies, nucleolin does not co-immunoprecipitate with other heparin-binding growth factors (data not shown), but HGF-Nucleolin interactions do respond to heparin competition (data not shown). We are currently searching for other components in nucleolin-HGF interactions using BIAcore. The findings we report here, together with the nucleolin-focused independent clinical trial underway [77, 78], introduce nucleolin as an attractive target for therapeutic regulation of prostate cancer, although the details of its position in cancer cell HGF signaling remain to be found.

We sought to begin placing nucleolin in or beside well-described HGF-met signaling pathways affecting integrin-based cell behaviors. Cell spreading and membrane ruffling behaviors seen in C4-2's met-independent HGF cell responses led us to look first at small GTPases (Figure 5). We found that Rac is activated by HGF treatment of met-negative C4-2 cells (Figure 5D,E). Rho, Rac and cdc42 are part of the Ras small GTPase superfamily, and play key roles in regulating cell shape, contraction, adhesion, motility, and proliferation [115‐117]. Members of this GTPase superfamily are also known molecular triggers to response cascades following growth factor stimulation ([118] and references within). Rac, in particular is a regulator of the superoxide generating NADPH oxidase in phagocytes [119]. The production of reactive oxygen species (ROS) mediates activation of NF-κB-dependent gene expression, essential for tumorigenesis and metastasis.