- Split View
-
Views
-
Cite
Cite
Zhiyong Mi, Hongtao Guo, Philip Y. Wai, Chengjiang Gao, Paul C. Kuo, Integrin-linked kinase regulates osteopontin-dependent MMP-2 and uPA expression to convey metastatic function in murine mammary epithelial cancer cells, Carcinogenesis, Volume 27, Issue 6, June 2006, Pages 1134–1145, https://doi.org/10.1093/carcin/bgi352
- Share Icon Share
Abstract
Metastasis-supporting physiological alterations are regulated by cell signaling molecules, which target signal transduction pathways and gene expression. Osteopontin (OPN) overexpression may represent a key molecular event in cancer metastasis. In this study, using metastatic 4T1 and non-metastatic 4T07 murine mammary cancer cell lines, we demonstrate that 4T1 cells exhibit significantly increased OPN, integrin-linked kinase (ILK), matrix metalloproteinase-2 (MMP-2) and urokinase-type plasminogen activator (uPA) expression in contrast to 4T07 cells. Blockade of OPN binding to 4T1 cell-surface integrins by the competitive ligand inhibitor, RGD, or a blocking antibody to α v β 3 integrin decreases OPN, ILK, MMP-2 and uPA expression. Conversely, exposure of 4T07 cells to exogenous OPN increases ILK, MMP-2 and uPA levels. Further experiments demonstrate that OPN–α v β 3 integrin binding in 4T1 with subsequent activation of ILK results in binding of AP-1 to MMP-2 and uPA promoter and increased in vitro promoter activation, as measured by transient transfection assays using MMP-2 and uPA promoter-reporter constructs. AP-1 activity is ablated by co-transfection of DN-ILK or exposure to RGD. Finally, functional correlative assays demonstrate that inhibition of ILK activity or RGD-mediated blockade of α v β 3 integrin binding significantly inhibits in vitro invasion, migration and invasion properties of 4T1 cells. In addition, uPA and MMP-2 have overlapping contributions to 4T1 migration and invasion characteristics. However, OPN and ILK activities contribute to 4T1 adhesion activities via mechanisms that are independent of uPA and MMP-2. Our results indicate that binding of an RGD-bearing ligand, such as OPN, to integrin receptors in metastatic 4T1 cells transcriptionally mediates MMP-2, uPA and OPN expression through ILK-dependent AP-1 activity and significantly increases in vitro functional correlates of metastasis. In 4T1 murine mammary cancer cells, we conclude that OPN mediates metastatic behavior, in part, through upregulation of MMP-2 and uPA protein expression.
Introduction
Cancer progression depends on an accumulation of metastasis-supporting genetic modifications and physiological alterations. These physiological changes are regulated by cell signaling molecules, which target signal transduction pathways and, ultimately, gene expression. One such molecule, osteopontin (OPN), functions as both a cell attachment protein and a cytokine that signals through two cell adhesion molecules: α v β 3 -integrin and CD44 ( 1 – 3 ). Initially discovered as an inducible, tumor-promoter gene, OPN is an acidic hydrophilic glycophosphoprotein, is overexpressed in human tumors, is the major phosphoprotein secreted by malignant cells in patients with advanced metastatic cancer and has been implicated in tumor cell migration and metastasis. Data suggest that OPN overexpression represents a key molecular event in tumor progression and metastasis ( 4 – 11 ). Potential steps that may involve OPN include (i) tumor cell attachment to basement membrane through cell-surface adhesion molecules, (ii) proteolytic degradation of the extracellular matrix (ECM) by tumor-derived proteinases and (iii) tumor cell migration through the ECM ( 12 – 14 ). However, while a great deal of correlative data exists, the signal transduction pathway by which OPN may facilitate metastatic transformation is unknown.
In this regard, we have previously examined the differential transcriptional regulation of OPN in the murine mammary epithelial tumor cell lines, 4T1 and 4T07 ( 15 ). These are thioguanine-resistant sublines derived from the parental population of 410.4 cells from Balb/cfC3H mice ( 16 ). Although they share a common origin, these lines are phenotypically heterogeneous in their metastatic behavior. 4T1 hematogeneously metastasizes to the lung, liver, bone and brain, whereas 4T07 is highly tumorigenic but fails to metastasize. The tumor growth and metastatic spread of 4T1 cells closely mimics Stage IV human breast cancer. In this study, again comparing 4T1 and 4T07 cells, we demonstrate that OPN conveys metastatic function by upregulating expression of matrix metalloproteinase-2 (MMP-2) and urokinase-type plasminogen activator (uPA) through integrin-linked kinase (ILK) dependent AP-1 activation. This has not been described previously. Our results provide new insights into the signal transduction pathway by which OPN mediates metastatic behavior in malignancy.
Methods
Cell culture
Mouse mammary tumor cell lines 4T1 and 4T07 were cultured in Dulbecco's Modified Eagle's medium (DMEM) supplemented with 10% fetal calf serum, penicillin (100 U/ml) and streptomycin (100 µg/ml) and maintained at 37°C in a humidified atmosphere of 5% CO 2 . In selected instances, cell suspensions were cultured with OPN (R&D Systems, MN), ILK siRNA (Santa Cruz Biotechnology, CA), 2 µg 23C6 α v β 3 integrin Ab (Santa Cruz Biotechnology) and 100 nM RGD or RGE peptide (Sigma, MO) containing medium for 24 h. The ILK siRNA sequences for sc-35667 (Santa Cruz Biotechnology) are derived from GenBank accession number NM010562. This siRNA is a pool of three strands:
489-507
ACUUGUUACAGAUGCUCACtt
GUGAGCAUCUGUAACAAGUtt
668-686
UUUGAAGUCAAUACCGGAGtt
CUCCGGUAUUGACUUCAAAtt, and
1525-1543
UAACUUUGGAGAAGUGUACtt
GUACACUUCUCCAAAGUUAtt.
ILK siRNA mis-match (MM) control sequence is as follows:
5′-AGCGUGUACGUAUACCCTT
3′-TTUCGCACAUGCAUAUGGG.
Western blot analysis
Cells were lysed in buffer (0.8% NaCl, 0.02 KCl, 1% SDS, 10% Triton X-100, 0.5% sodium deoxycholic acid, 0.144% Na 2 HPO 4 and 0.024% KH 2 PO 4 , 2 mM phenylmethylsulfonyl fluoride, pH 7.4) and centrifuged at 12 000 × g for 10 min at 4°C. The protein concentration was determined by Bio-Rad protein assay kit (Bio-Rad, CA); the protein samples were separated by 4–20% sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS–PAGE) and electrotransferred onto polyvinylidene difluoride membranes (Amersham Bioscience, NJ) by semi-dry transfer (Bio-Rad). The membranes were probed with the following primary antibodies for 1 h at room temperature: goat anti-mouse OPN Ab (R&D Systems, Minneapolis, MN), rabbit MMP-2 polyclonal Ab (Santa Cruz Biotechnology), rabbit uPA polyclonal Ab (Santa Cruz Biotechnology), mouse CD44 monoclonal Ab (Santa Cruz Biotechnology), goat β-actin polyclonal Ab (Santa Cruz Biotechnology) or rabbit ILK polyclonal Ab (Upstate, NY). These antibodies were detected using the appropriate horseradish peroxidase-conjugated secondary antibody. The reactive proteins were visualized by chemiluminescence using the ECL kit (Amersham Bioscience, NJ), quantified by Alphaimager 3400 (AlphaInnotech, CA) and normalized to β-actin protein controls.
Reverse-transcription polymerase chain reaction (PCR) analysis
Total RNA was isolated using a TRIzol kit according to the manufacturer's instruction (Invitrogen, MD). cDNAs were synthesized from 1 µg RNAs using the cDNA synthesis kit manual (Bio-Rad). The oligonucleotide sequences of the primer sets used in this study are as follows:
β-actin, 5′-end primer: AGAGGGAAATCGTGCGTGACA. 3′-end primer: CAATAGTGATGACCTGGCCGT;
uPA, 5′-end primer: AGCGCCAATAGCATTACCCTCAGA. 3′-end primer: TCACGTGGAGCATCACATGGAAGA;
MMP-2, 5′-end primer: TGTTTACCATGGGTGGCAATGCAG. 3′-end primer: TGTTTGCAGATCTCCGGAGTGACA;
OPN, 5′-end primer: TCAGAGGAGAAGCTTTACAG, 3′-end primer: TGCAAAGTAAGGAACTGTGT;
ILK, 5′-end primer: ACTGGAAGGCCTTCGGCCTACC. 3′-end primer: ATGTTGGGAGACATCATGTGGC.
A log-linear dose–response curve was determined for each set of primers to determine the appropriate number of amplification cycles. Verification of the amplified PCR products was performed by automated DNA sequencing. The PCR products were visualized by UV illumination after electrophoresis through 1.0% agarose (UltraPure; Sigma Chemical Co.) and staining in TRIS borate–EDTA buffer containing ethidium bromide. DNA was visualized on a UV illuminator; gel photographs were then scanned and the area under the curve was normalized to β-actin mRNA controls.
Transient transfection assay
A non-targeting, kinase-deficient E359K ILK protein functioning as a dominant-negative form of ILK and wild-type ILK expression plasmids were a generous gift from Dr Christopher E. Turner (SUNY Upstate University, Syracuse, NY). 4T1 cell was transiently transfected with the above plasmids using Lipofectamine 2000 according to manufacturer's instruction (Invitrogen, MD). Briefly, 4 × 10 5 cells were seeded with antibiotic-free DMEM medium on each well of 12-well plates the day before transfection. Two micrograms of plasmid DNA and 4 µl Lipofectamine 2000, diluted with Opti-MEM medium, were mixed gently and incubated with cells. Culture medium was changed after 6 h transfection and incubated further at 37°C for 24 h. The control cells received Lipofectamine 2000 alone.
ILK kinase activity assay
The ILK immune complex kinase assay was carried out as described in Hannigan et al . ( 17 ) Briefly, cell lysis was performed directly on the plates with lysis buffer (150 mM NaCl, 1% NP-40, 0.5% sodium deoxycholate, 50 mM HEPES, pH 7.5, 1 µg/ml leupeptin, 1 µg/ml aprotinin, 3 mM phenylmethyl sulfonyl fluoride); the soluble lysate was collected by centrifugation at 13 000 × g for 15 min at 4°C. Protein concentration was determined by the Bio-Rad protein kit (Bio-Rad). Equivalent concentrations of lysate were precleared with protein G plus-agarose (Santa Cruz Biotechnology). Following rotation incubation with anti-ILK antibody (Upstate Biotechnology, NY) at 4°C overnight, protein G plus-agarose was added followed by rotation for 4 h at 4°C. ILK immune complexes were spun down at 1000 × g for 5 min at 4°C. After two washes with lysis buffer and two washes with kinase wash buffer (10 mM MgCl 2 , 10 mM MnCl 2 , 50 mM HEPES, pH 7.5, 0.1 mM sodium orthovanadate, 1 mM dithiothreitol), the pellet was suspended in 25 µl kinase assay buffer (10 mM MgCl 2 , 10 mM MnCl 2 , 50 mM HEPES, pH 7.5, 1 mM sodium orthovanadate, 2 mM NaF, 10 µci [γ-P32]-ATP[10 mci/ml], 5 µg myelin basic protein) and incubated for 30 min at 30°C. The reaction was terminated by adding an equal volume of 2× SDS loading buffer and boiled for 5 min. Phosphorylated myelin basic protein bands were separated by 4–20% SDS–PAGE gel, visualized on film (Kodak, NY) at −70°C and quantified by Alphaimager 3400 (AlphaInnotech, CA).
Promoter analysis
Plasmids and constructs were described previously ( 18 ). Briefly, promoter sequences for murine OPN (−107 to +79; Genbank M38399), uPA (−2620 to +44; NCBI murine chromosome 14 sequence) and MMP-2 (−2390 to +100; Genbank AB125668) were cloned into a pGL3-basic luciferase reporter plasmid (Promega, Madison, WI). The point mutants of the AP-1 binding site in each of the promoter constructs were constructed by two-step PCR; the AP-1 binding sites of OPN promoter TGACACA (−75 to −69), uPA promoter TGAGGTCA (−2407 to −2400) and MMP-2 promoter TGACTCA (−2307 to −2301) were all mutated to T C A T A T A and confirmed by DNA sequencing. pAP1-luc and deletion mutant plasmids were purchased from Stratagene, TX. Luciferase reporter assays were performed by Dual Luciferase Reporter Assay System (Promega, Madison, WI) after 36 h post-transfection. Following the manufacturer's instruction, cells were washed with phosphate-buffered saline (PBS) and lysed in lysis buffer. Cell debris was removed by centrifugation at 10 000 r.p.m. for 1 min. Twenty microliters of the supernatant were mixed with 100 µl of luciferase substrate and measured by luminometer (Turner Designs TD-20/20). Fire-fly luciferase activity was normalized for transfection efficiency using Renilla Luciferase activity (pRL-SV40).
Nuclear extracts and electrophoretic mobility shift assays (EMSA)
Nuclear extract preparation and EMSA were conducted as described previously ( 18 ). All buffers containing the protease inhibitors and dithiothreitol were prepared daily. The cell pellet collected by centrifugation was resuspended in 5× PCV (packed cell volume) in ice-cold buffer A containing 10 mM HEPES, pH 7.9, 10 mM KCl, 0.1 mM EDTA, 1.0 mM dithiothreitol, pepstatin A (2 µg/ml) and 0.5 mM phenylmethylsulfonyl fluoride, followed by incubation on ice for 20 min. Then, 10% Nonidet P-40 was added to 0.5% final concentration and vortexed briefly. After centrifugation at 150× g for 5 min, the cell pellets were resuspended in 2× PCV in ice-cold buffer C containing 20 mM HEPES, pH 7.9, 0.4 M NaCl, 1.0 mM EDTA, 1.0 mM EGTA, 1.0 mM DTT, pepstatin A (2 µg/ml) and 0.5 mM phenylmethylsulfonyl fluoride and vortexed for 20 min at 4°C. Insoluble material was removed by centrifugation at 14 000 r.p.m.; the supernatant containing the nuclear protein was stored at −80°C until use. Wild-type and mutant probes were end-labeled with [γ-P 32 ]ATP (2500 ci/mmol) using T4 polynucleotide kinase (Promega), followed by G-50 column purification. The reactions were resolved on 6% native acrylamide gel in 0.5× TBE buffer and visualized by autoradiography. In specific competitive binding assays, unlabeled oligonucleotides were added at 50-fold molar excess. In non-specific competitive binding assays, unlabeled poly(dI-dC) was used. Supershift assays were performed by preincubating nuclear extracts with rabbit anti-mouse c-Jun polyclonal Ab (Santa Cruz). The oligonucleotide sequences were as follows: probe, 5′AAA ACC TCA TGA CAC A TC AC; mutant, 5′ AAA ACC TCA T C A T A T A TC AC.
Chromatin immunoprecipitation assay
Chromatin immunoprecipitation was performed using the chromatin immunoprecipitation (ChIP) Assay kit (Upstate, New York). Two micrograms of rabbit c-Jun polyclonal Ab (Santa Cruz) was added to the supernatant of the 1% formaldehyde cross-linked cell lysates and incubated overnight at 4°C with rotation. For the negative control, preimmune serum replaced the antibody. The immune complexes were recovered by the addition of salmon sperm DNA/protein A agarose slurry. After reverse cross-linking through proteinase K treatment for 4 h at 50°C, DNA was recovered by phenol chloroform extraction and ethanol precipitation. PCR amplification was used to assay the immunoprecipitates by using primers specific for the various promoter regions containing AP-1 binding site. The PCR program was 95°C × 5 min, followed by 94°C × 45 s, 55°C × 45 s and 72°C × 45 s for a total of 20 cycles. The PCR primer sequences were as follows:
OPN Forward: 5′ CTG ATG CTC TTC CGG GAT TC, Reverse: 5′ TTC CTC CGA GAA TGC CTG CC; Product 305 bp.
uPA Forward: 5′-CTG TGA TGA GTT GGA GGC ACG AGG, Reverse: 5′-TCC CAG CAG GCT GTC GTG ATT CAC; Product 213 bp.
MMP-2 Forward: 5′-CCT CAC AGG ACC CTC ACC AG, Reverse: 5′-GGG CAA GGC CTC CTG TCT GT; Product 130 bp. PCR products were fractionated and visualized on 1.2% agarose gel stained with ethidium bromide.
Cell adhesion assay
4T1 cells were treated with RGD peptide as above or transfected with WT-ILK or DN-ILK plasmid DNA for 24 h. The in vitro adhesion assay was performed on 96-well microtiter plates coated with 10 µg/ml Matrigel (Collaborative Biomedical Products, MA); cells were trypsinized and resuspended in DMEM with 1% bovine serum albumen (BSA), 1 mM MgCl 2 , 0.5 mM CaCl 2 and 1 mM RGDS at a concentration of 1 × 10 6 cells/ml. 1 × 10 5 cells (100 µl) were added into each well and placed for 30 min at 37°C in 5% CO 2 humidified air incubation. Non-adhering cells were removed by gently washing the wells three times with PBS (with 1 mM MgCl 2 and 0.5 mM CaCl 2 ). Adherent cells were fixed with 3.7% paraformaldehyde for 10 min at room temperature, followed by rinsing with PBS, and stained with 0.4% crystal violet for 10 min. After extensive rinsing, the dye was released from the cells by addition of 30% acetic acid, and the microtiter plates were read in a microplate reader (Molecular Devices, CA) at 590 nm. The absorbance obtained for control and experimental groups were each divided by the absorbance obtained for controls and expressed as an index. Triplicate assays were performed for each group of cells.
Cell migration and invasion assay
The migration and invasiveness of 4T1 were evaluated in 24-well transwell chambers with upper and lower culture compartments separated by polycarbonate membranes with 8 µm sized pores (Costar 3422, Corning, NY). Cells were grown to 60% confluence, harvested using Cell-stripper (Cellgro, Herndon, VA), washed with 1× PBS and resuspended in DMEM–0.1% BSA. Before plating cells into the transwells, DMEM–0.1% BSA was incubated in the top chamber of each transwell at 37°C for 1 h to saturate non-specific binding sites and then subsequently removed. 5 × 10 4 cells suspended in 100 µl of DMEM–0.1% BSA were plated into the top chamber. DMEM–10% FBS was placed in the bottom chamber to act as a chemoattractant. After 24 h incubation at 37°C in 5% CO 2 humidified air, the cells remaining at the upper surface of the membrane were removed with a cotton swab. The cells that migrated through the 8 µm sized pores and adhered to the lower surface of the membrane were fixed with 3.7% paraformaldehyde, stained with 0.2% crystal violet and washed with 1× PBS three times. The dye was eluted using 30% acetic acid, and quantification of cell number was performed using colorimetric analysis with a microplate reader (absorbance at 590 nm). The absorbance units obtained for control and experimental clones were each divided by the absorbance units obtained for controls and expressed as a migration index. By definition, untreated 4T1 control cells were assigned an index of 1. In a similar fashion, the invasiveness of pS-OPN cells were evaluated in Matrigel (Collaborative Biomedical Products, MA) coated 24-well transwell chambers. Matrigel basement membrane matrix is a solubilized basement membrane preparation extracted from the Engelbreth–Holm–Swarm (EHS) mouse sarcoma and has been established for use in a reliable assay for in vitro cell invasiveness. Matrigel was used at a concentration of 0.4 mg/ml. Cells, media, experimental conditions and analysis performed were similar to those of migration assays. The absorbance obtained for control and experimental groups were each divided by the absorbance obtained for controls and expressed as a migration or invasion index. By definition, untreated 4T1 controls were assigned an index of 1. Triplicate assays were performed for each group of cells. Sense and antisense oligonucleotide sequences were as follows: uPA antisense: 5′-GGC TCG CCA GCC AGA CTT TCA T; MMP-2 antisense: 5′-TGT TTG CAG ATC TCC GGA GTG ACA; uPA sense: 5′-ATG AAA GTC TGG CTG GCG AGC C; and MMP-2 sense: 5′-TGT TTA CCA TGG GTG GCA ATG CAG.
Statistical analysis
Data are expressed as mean ± standard error of the mean. Statistical analysis was performed using the Student's t -test; P values less than 0.05 were considered significant.
Results
OPN–integrin binding and expression of ILK, MMP-2 and uPA proteins
To determine a potential relationship between OPN cell-surface binding and ILK protein, 4T1 and 4T07 cell lysates were assayed for OPN, MMP-2, uPA and ILK protein expression. ( Figure 1A ) Of note, 4T1 hematogeneously metastasizes, and 4T07 is highly tumorigenic but fails to metastasize. In the metastatic 4T1 cells, OPN and ILK protein expression are significantly greater than that noted in 4T07 cells. In addition, levels of MMP-2 and uPA protein are significantly upregulated in the 4T1 cells, consistent with its metastatic phenotype.
As OPN may bind to α v β 3 integrins and/or CD44 receptors, we next sought to determine whether protein expression of ILK, MMP-2 and uPA were related to OPN–integrin binding. ( Figure 1B ) To block OPN–integrin binding, 4T1 cells were exposed to RGD (100 nM), a competitive ligand inhibitor for OPN–integrin binding. Conversely, 4T07 cells were exposed to exogenous OPN (10 nM). In the presence of RGD, expression of ILK, MMP-2 and uPA protein in 4T1 cells was decreased by 8-, 6-, and 5-fold, respectively, in comparison with untreated 4T1 cells ( P < 0.02 4T1 + RGD versus 4T1 for ILK, MMP-2 and uPA). When 4T07 cells were stimulated with OPN, ILK, MMP-2 and uPA protein expression are increased by 5-, 8- and 8-fold, respectively, when compared with untreated 4T07 cells ( P < 0.01 4T07 + OPN versus 4T07 for ILK, MMP-2 and uPA). Additional experiments were then performed to rescue the 4T1 and 4T07 wild-type phenotypes. In 4T07 cells, ILK, MMP-2 and uPA expression were restored to original levels when binding of exogenous OPN to its integrin receptor was inhibited by concomitant addition of 20-fold excess RGD. Similarly, in 4T1 cells, ILK, MMP-2 and uPA expression were restored to baseline levels when OPN (200 nM) at 20-fold excess and RGD were both added to the culture medium. 4T1 + OPN and 4T07 + RGD treatment groups did not exhibit significantly altered expression of ILK, MMP-2 and uPA proteins when compared with their respective wild-type controls. (Data not shown.)
Additional control experiments were then performed ( Figure 1C ). Experiments were performed in 4T1 and 4T07 cells in which RGE was substituted for RGD, an α v β 3 integrin blocking Ab was utilized or siRNA inhibition of ILK mRNA expression was performed. In 4T1 cells, incubation with an indifferent peptide such as RGE did not alter ILK, MMP-2 and uPA protein expression in comparison with Control cells. Similarly, the ILK mis-match siRNA did not alter ILK, MMP-2 or uPA expression. In contrast, addition of a blocking antibody to α v β 3 integrin or ILK siRNA dramatically decreased ILK, MMP-2 and uPA protein to near-undetectable levels. Addition of OPN at 20-fold excess reversed the effects of the α v β 3 integrin antibody. In 4T07 cells, addition of RGE, α v β 3 integrin antibody or ILK siRNA did not alter the previously observed profile in Control 4T07 cells of minimal ILK, MMP-2 and uPA expression. (Data not shown.) These data indicate that OPN–integrin binding regulates, in part, ILK, MMP-2 and uPA protein expression. Blockade of OPN binding to α v β 3 integrin receptors in 4T1 cells abolishes ILK expression with subsequent ablation of downstream MMP-2 and uPA protein expression. Conversely, repletion of OPN in 4T07 cells significantly increases ILK, MMP-2 and uPA expression. These data demonstrate that OPN and its binding as a GRGDSP-containing ligand to its target α v β 3 integrin receptors mediate ILK, MMP-2 and uPA expression.
We next focused upon ILK kinase activity in 4T1 cells ( Figure 1D ). ILK activity is readily detectable in 4T1 cells. However, in the presence of RGD ligand blockade of OPN–integrin binding or α v β 3 integrin blocking antibody, ILK activity was significantly decreased by over 10-fold. ( P < 0.01 4T1 versus 4T1 + RGD and 4T1 + Integrin Ab) In both instances, 20-fold excess OPN was able to effectively compete for binding to α v β 3 integrin receptors and restore ILK activity to levels not significantly different than that of untreated wild-type 4T1 cells. RGE treatment did not alter ILK activity in 4T1 cells. Protein loading was equivalent among the groups as determined by Coomassie blue staining. This result suggests that blockade of OPN binding to α v β 3 integrin inhibits ILK phosphorylation activity.
Lastly, as OPN may bind to both integrin and CD44 cell-surface receptors, immunoblot analysis was performed to determine the extent of CD44 expression in 4T1 and 4T07 cells ( Figure 1E ). There was no CD44 expression noted. In this system, OPN mediates its effects exclusively through integrin binding.
OPN–integrin binding blockade and MMP-2/uPA expression
To determine the potential role of ILK in the expression of MMP-2 and uPA proteins, 4T1 cells were again incubated in the presence of RGD to block OPN–integrin binding. As our previous data have demonstrated that ILK protein expression and kinase activity are decreased in this setting, WT-ILK and DN-ILK expression vectors were transfected in selected instances. Immunoblots and reverse-transcriptase PCR for MMP-2 and uPA were first performed to assess levels of protein and steady-state mRNA. ( Figure 2A ) Inhibition of ILK activity in 4T1 cells by transfection of DN-ILK decreased MMP-2 and uPA protein levels by over 10-fold each. To confirm the role of OPN–integrin binding in ILK expression, further immunoblot studies were performed in 4T1 cells treated with RGD and/or WT-ILK ( Figure 2B ). Again, RGD treatment resulted in a significant decrease in MMP-2 and uPA protein expression ( P < 0.01 4T1 + DN-ILK versus 4T1 for MMP-2 and uPA). However, in the presence of both RGD and WT-ILK, transfection of the WT-ILK expression vector reversed the RGD-mediated decrease in MMP-2 and uPA protein levels in 4T1 cells. These data indicate that OPN–integrin binding regulates protein expression of MMP-2 and uPA through ILK. RT–PCR was then performed to determine the effect of OPN–integrin binding/blockade and ILK on steady-state levels of MMP-2 and uPA mRNA ( Figure 2C and D ). ILK mRNA levels are decreased in the presence of RGD ( P < 0.01 versus 4T1 alone). MMP-2 and uPA mRNA levels are also significantly decreased in 4T1 cells exposed to RGD or transfected with DN-ILK. Repletion of ILK in 4T1 + RGD cells by transfection with WT-ILK returns MMP-2 and uPA mRNA levels to those seen in wild-type 4T1 cells. Finally, addition of both OPN + RGD restores ILK, MMP-2 and uPA mRNA levels to levels equivalent to that of wild-type 4T1. These data suggest that OPN–integrin binding regulates ILK expression and upregulates levels of MMP-2 and uPA protein and mRNA.
OPN and ILK regulate AP-1 DNA binding activity
A well-described downstream signaling event associated with ILK is AP-1 activation ( 19 ). To determine the potential role of OPN–integrin binding in ILK-dependent AP-1 activation, a series of experiments utilizing EMSA assays with a consensus AP-1 binding oligonucleotide were performed. ( Figure 3 ) In Figure 3A and B , EMSA assays were performed with 4T1 nuclear extract and a labeled oligonucleotide with a consensus AP-1 binding site. Constitutive AP-1 DNA binding activity is present and is confirmed with the supershift assay using c-jun antibody; IgG serum did not shift the AP-1 band. The signal is ablated in specific competition assays using an increasing concentration of excess unlabelled target oligo; in contrast, a non-specific competitor (poly dI-dC) did not alter the AP-1 signal. No AP-1 binding was seen in the presence of the mutant AP-1 target oligonucleotide. (Data not shown.) In Figure 3C depicting another representative EMSA experiment, treatment of 4T1 cells with RGD or transfection with DN-ILK abolishes AP-1 oligonucleotide binding. When RGD-treated cells are transfected with WT-ILK or excess OPN is added with RGD, AP-1 DNA binding is restored. Finally, cellular c-jun protein levels are not altered by any of these treatments ( Figure 3D ). An AP-1 heterologous promoter-luciferase construct or a construct bearing a mutated AP-1 site was then transiently transfected into 4T1 cells ( Figure 3E ). In the presence of RGD treatment and DN-ILK co-transfection, AP-1-mediated transactivation as measured by the wild-type AP-1 luciferase construct, activity is decreased by 85 and 75%, respectively ( P < 0.01 Control versus RGD or DN-ILK). When WT-ILK is co-transfected in the setting of RGD treatment or RGD and excess OPN are added together, AP-1-mediated luciferase activation is restored to levels that were statistically equivalent to that of wild-type controls. Predictably, the mutated AP-1 luc construct did not exhibit any appreciable activity with any of the treatment conditions. These data suggest that binding of OPN as a GRGDSP-containing ligand to integrins and subsequent ILK activation are required for AP-1 DNA binding and transactivation in 4T1 cells.
ILK-mediated AP-1 activity regulates MMP-2 and uPA expression
To determine the role of integrin binding and ILK-dependent AP-1 activity in MMP-2 and uPA expression, promoter analysis and ChIP assays were performed. The murine uPA and MMP-2 promoters contain AP-1 binding sites at nt-2407 to nt-2400 and nt-2307 to nt-2301, respectively. The murine OPN promoter has an AP-1 site at nt-75 to nt-69. In both instances, ChIP assays demonstrate in vivo c-jun binding that is abolished by the presence of RGD or DN-ILK ( Figure 4 ). Conversely, WT-ILK-transfected 4T1 cells demonstrate c-jun binding in both the presence and the absence of RGD. The addition of both RGD and OPN restores c-jun binding to levels noted in wild-type 4T1 cells. Interestingly, in a further ChIP assay addressing c-jun binding in the OPN promoter, the same results were obtained. Input DNA was equivalent among all experimental groups for OPN, MMP-2 and uPA analysis. (Data not shown.) These results indicate that AP-1 binding to the MMP-2, uPA and OPN promoters is ILK-dependent and requires binding of OPN as a GRGDSP-containing ligand to its corresponding integrin receptor.
Transient transfection assays with accompanying mutation analysis were performed using MMP-2, uPA and OPN promoter-reporter constructs ( Figure 5 ). A wild-type promoter construct and an AP-1-mutated promoter construct were utilized for each protein. In the MMP-2 promoter, luciferase activity was decreased by 83 and 85% in the presence of RGD and DN-ILK, respectively ( P < 0.01 versus Control cells for both RGD and DN-ILK). Conversely, in RGD-treated cells that were transfected with WT-ILK and in the setting of OPN and RGD treatment, activity was restored to levels that were 78 and 88% of Control values, respectively ( P = NS versus Control cells for RGD + WT-ILK and RGD + OPN groups). Similarly, in the uPA and OPN promoters, RGD or DN-ILK was associated with large and significant decreases in activity that were reversed in RGD cells by concomitant transfection with WT-ILK and addition of the RGD + OPN combination. In all instances, mutation of the AP-1 binding site was associated with near-total ablation of luciferase activation in the MMP-2, uPA and OPN promoters. These data indicate that OPN, uPA and MMP-2 promoter activities are regulated by integrin binding of OPN as an RGD-containing ligand, ILK activation and AP-1 DNA binding/transactivation. In total, these results indicate that integrin binding of OPN to metastatic 4T1 cells transcriptionally regulates MMP-2, uPA and OPN expression through ILK-dependent AP-1 activity.
Functional in vitro correlates of metastatic behavior
To ascertain the functional correlates of these observations, adhesion, migration and invasion assays were performed ( Figure 6 ). By definition, control wild-type 4T1 cells are assigned a value of 1. In 4T1 cells, RGD treatment decreased adhesion, migration and invasion by 66, 91 and 82%, respectively ( P < 0.01 versus WT for all conditions). Similarly, transfection of DN-ILK decreased adhesion, migration and invasion by 60, 87 and 89%, respectively ( P < 0.01 versus WT for all conditions). In the WT-ILK, RGD + WT-ILK and RGD + OPN groups, adhesion, migration and invasion were not statistically different from those found in wild-type 4T1 controls. These results indicate that RGD-mediated inhibition of OPN–integrin binding or DN-ILK-dependent inhibition of ILK activity is associated with significantly decreased cell adhesion, migration and invasion. Conversely, restoration of OPN binding or ILK activity restores function to levels equivalent to that noted in wild-type cells. We conclude that integrin binding of OPN with associated ILK activity determines in vitro functional correlates of metastatic behavior.
To confirm a role for MMP-2 and uPA in OPN-dependent adhesion, migration and invasion characteristics of 4T1 cells, MMP-2 and uPA expression were inhibited by transfection with antisense oligonucleotides to MMP-2 and/or uPA and blocking Ab to α v β 3 integrin. ( Figure 7 ) Again, control wild-type 4T1 cells are assigned a value of 1. Adhesion of 4T1 cells was decreased by 60% in the presence of integrin blockade ( P < 0.01 versus IgG Control). However, in the presence of uPA antisense or MMP-2 antisense oligonucleotides, there was no change in 4T1 adhesion. In migration assays, integrin blocking antibody was associated with an 80% decrease in activity ( P < 0.01 versus IgG Control). However, in this setting of antisense oligos directed at uPA or MMP-2, there was a 75% and 75% decrease in migration, respectively ( P < 0.01 versus Sense Control for both MMP-2 and uPA). When antisense oligos for both uPA and MMP-2 were added together, there was an 82% decrease in adhesion ( P < 0.01 versus uPA + MMP-2 Sense Control). Finally, in invasion assays, integrin blocking antibody was associated with a 71% decrease in activity ( P < 0.01 versus IgG Control). When antisense oligos directed at uPA or MMP-2 are added, there was a 53 and 50% decrease in in vitro invasion, respectively ( P < 0.01 versus Sense Control for both MMP-2 and uPA). When antisense oligos for both uPA and MMP-2 were added together, there was an 82% decrease in invasion ( P < 0.01 versus uPA + MMP-2 Sense Control). However, these were not significantly different from that found with uPA antisense alone or MMP-2 antisense alone. These data indicate that 4T1 adhesion, migration and invasion requires intact α v β 3 integrin binding activity. In addition, uPA and MMP-2 have overlapping and, possibly, additive contributions to 4T1 migration and invasion characteristics. On the basis of our previous OPN and ILK experiments, as depicted in Figure 6 , it appears that OPN and ILK activities contribute to 4T1 adhesion via mechanisms that are independent of uPA and MMP-2.
Discussion
The signal transduction pathway by which OPN may facilitate metastatic transformation is unknown. In this study, using the murine mammary epithelial tumor cell lines, 4T1 and 4T07, we demonstrate that (i) differential expression of OPN with significantly greater levels of OPN, ILK, MMP-2 and uPA occurs in association with the highly metastatic 4T1 subclone, (ii) RGD and α v β 3 integrin Ab-mediated inhibition of OPN binding to its integrin receptors inhibits ILK, MMP-2 and uPA expression in 4T1 cells, (iii) α v β 3 integrin-associated ILK expression and activity are required for AP-1 transactivation of the OPN, MMP-2 and uPA promoters with ultimate expression and (iv) OPN-dependent ILK activity regulates in vitro functional correlates of metastatic behavior ( 15 ). These findings have not been described previously. Our results indicate that binding of OPN as a GRGDSP-containing ligand to α v β 3 integrin receptors activates ILK with subsequent upregulation of AP-1 and induction of MMP-2 and uPA synthesis. Interestingly, this pathway represents an autoregulatory process in which OPN synthesized by 4T1 cells may act in an autocrine or paracrine fashion to potentiate metastatic behavior as mediated by MMP-2 and uPA expression.
OPN is a secreted glycoprotein that is rich in aspartate and sialic-acid residues and contains functional domains for calcium-binding, phosphorylation, glycosylation and extracellular matrix adhesion ( 20 ). OPN appears to mediate cell–matrix interactions and cellular signaling through binding with integrin, primarily α v β 3 , and CD44 receptors. OPN is expressed in multiple species, including humans and rodents ( 21 ). Cells that express OPN include osteoclasts, osteoblasts, kidney, breast and skin epithelial cells, nerve cells, vascular smooth muscle cells and endothelial cells ( 20 , 22 – 25 ). Activated immune cells such as T-cells, NK cells, macrophages and Kupffer cells also express OPN. The secreted OPN protein is widely distributed in plasma, urine, milk and bile ( 26 – 28 ). Constitutive expression of OPN exists in several cell types, but induced expression is detected in T lymphocytes, epidermal cells, bone cells, macrophages and tumor cells in remodeling processes such as inflammation, ischemia-reperfusion, bone resorption and tumor progression ( 20 , 24 , 25 ). A variety of stimuli including phorbol 12-myristate 13-acetate (PMA), 1,25-dihydroxyvitamin D, basic fibroblast growth factor (bFGF), TNF-α, IL-1, IFN-γ and LPS appears to upregulate OPN expression ( 20 , 24 , 25 , 29 ). OPN has multiple molecular functions that mediate cell adhesion, chemotaxis, macrophage-directed IL-10 suppression, stress-dependent angiogenesis, prevention of apoptosis and anchorage-independent growth of tumor cells ( 20 , 24 , 25 , 29 ). Recently, a substantial body of data has linked OPN with the regulation of metastatic spread by tumor cells. However, the molecular mechanisms that define the role of OPN in tumor metastasis are incompletely understood.
With regard to breast cancer, OPN expression has been correlated with breast cancer metastasis. Fedarko et al . ( 9 ) found significantly increased serum levels of OPN in 20 breast cancer patients when compared with those of normal controls. When primary breast cancer tumors were compared with bone metastases, elevated cellular OPN expression, as detected by immunohistochemistry, was noted to be significantly higher in bone metastasis ( 30 ). In addition, using expression microarray analysis, Korkola et al . ( 31 ) have found that OPN is differentially expressed between lobular and ductal breast carcinomas. In contradistinction to the bulk of the correlative clinical data, Coppola et al . ( 7 ) recently surveyed immunohistochemical expression of OPN in a variety of human tumors. These investigators found little or no OPN expression in 23 out of 26 breast cancers, although increased expression of OPN was noted in a number of other cancer types. Although speculative, these divergent pieces of data would suggest a differential role for OPN in metastasis in contrast to tumorigenesis. However, little is known of the signal transduction pathways by which OPN mediates metastasis.
In melanoma cell lines, Kundu et al . ( 8 , 32 – 36 ) have addressed the OPN-dependent pathways by which MMP-2, MMP-9 and uPA undergo activation. These investigators show that MMP-2 plays a direct role in OPN-induced cell migration, invasion and tumor growth and that demonstrates that OPN-stimulated MMP-2 activation occurs through NF-κB-mediated induction of membrane type 1 MMP. With regard to MMP-9, they have demonstrated that nuclear factor-inducing kinase (NIK) plays a crucial role in OPN-induced mitogen-activated protein kinase/IκBα kinase-dependent NF-κB-mediated promatrix metalloproteinase-9 activation ( 32 ). OPN enhances JNK1-dependent/independent AP-1-mediated uPA secretion, uPA-dependent promatrix MMP-9 activation, cell motility and invasion. In addition, OPN induces NIK/MEKK1-mediated JNK1-dependent AP-1-mediated pro-MMP-9 activation and regulates the negative cross-talk between NIK/ERK1/2 and MEKK1/JNK1 pathways that ultimately controls the cell motility, invasiveness and tumor growth ( 33 ). With respect to uPA, they have also shown that OPN induces α v β 3 integrin-mediated AP-1 activity and uPA secretion by activating c-Src/EGFR/ERK signaling pathways and further demonstrate a functional molecular link between OPN-induced integrin/c-Src-dependent EGFR phosphorylation and ERK/AP-1-mediated uPA secretion ( 8 , 34 ). Clearly, OPN–α v β 3 integrin binding can regulate activation of metastatic mediators such as the MMPs and uPA, and conversely, mammary epithelial cell invasiveness induced by OPN requires uPA and MMPs ( 37 , 38 ). However, the above results notwithstanding, the link between OPN, ILK and AP-1-mediated MMP-2 and uPA gene transcription has not been addressed previously. In this study utilizing 4T1/4T07 murine breast cancer cell lines, we show that ILK activation by constitutive OPN expression in 4T1 is associated with AP-1-regulated expression of MMP-2 and uPA, two well-characterized markers of metastatic behavior.
ILK was originally discovered in 1996 as a B1 and B3 cytoplasmic domain interacting protein ( 17 ). It functions as a scaffold to form multi-protein complexes connecting integrins to signaling pathways ( 19 , 39 ). Its activity is stimulated by adhesion to the ECM and by growth factors in a PI3K-dependent fashion. ILK is involved with anchorage-dependent cell growth; cell-cycle progression, invasion and migration; cell motility and contraction; vascular development and tumor angiogenesis. In epithelial cells, overexpression of ILK induces epithelial–mesenchymal transition to result in a transformed tumorigenic phenotype. ILK expression and activity are increased in many types of cancer, such as prostate, colon, gastric and ovarian cancers ( 14 , 19 , 40 – 44 ). Specifically, in mammary epithelial cells, expression of ILK is associated with hyperplasia and tumor formation ( 43 , 45 ). Small molecule inhibitors of ILK activity have been found to decrease tumor growth, invasion and angiogenesis. In normal cells, ILK is thought to be transiently activated. However, in the setting of malignancy, ILK may be overexpressed or constitutively activated to effect multiple pathways, including RAC-GTP and CDC42-GTP, AKT with activation of mTOR/HIF-1, caspase-3, and NF-κB, GSK-3 phosphorylation and inhibition with subsequent AP-1 and β-catenin/TCF activation ( 19 , 39 , 40 , 44 , 46 , 47 ). Certainly, in the present study, ongoing OPN synthesis in 4T1 cells ‘constitutively’ activates ILK, resulting in AP-1-mediated expression of both MMP-2 and uPA.
With respect to metastatic behavior in breast cancer, the role of ILK has not been extensively studied. Novak et al . ( 48 ) demonstrated that overexpression of ILK in mammary epithelial cells results in an invasive phenotype characterized by downregulation of E-cadherin expression, translocation of beta-catenin to the nucleus and transactivation by a LEF-1/beta-catenin complex. Subsequently, Dedhar's group has found that overexpression of ILK in the SCP2 murine mammary epithelial cells is associated with a profound inhibition of anoikis, an important hallmark of oncogenic transformation and metastasis ( 46 ). In addition, inhibition of ILK activity induced anoikis in two human breast cancer cell lines. This group then demonstrated that ILK can induce an invasive phenotype in mammary epithelial cells by AP-1-dependent upregulation of MMP-9 ( 49 ). Recently, Dedhar et al . ( 47 ) have demonstrated that ILK is required for the activation of Rac and Cdc42 in epithelial cells. Ectopic expression of active ILK in mammary epithelial cells induces reorganization of the actin cytoskeleton and promotes rapid cell spreading on fibronectin. These effects are associated with constitutive activation of both Rac and Cdc42, but not Rho. Their data demonstrate an essential role of ILK kinase activity in Rac- and Cdc42-mediated actin cytoskeleton reorganization in epithelial cells, lending further weight to a role for ILK in the regulation of cancer cell motility and invasiveness. In order to determine if ILK overexpression can result in the formation of mammary tumors in vivo , Dedhar's group generated transgenic mice expressing ILK in the mammary epithelium ( 45 ). By the age of 6 months, female mice developed a hyperplastic mammary phenotype, which was accompanied by the constitutive phosphorylation of PKB/Akt, GSK-3β and MAP kinase. Focal mammary tumors subsequently appeared in 34% of the animals at an average age of 18 months. These results provide the first direct demonstration of a potential oncogenic role for ILK, which is upregulated in human tumors and tumor cell lines. The link between ILK and OPN and the metastatic phenotype in breast cancer has not been investigated previously.
OPN overexpression has been associated with tumor metastasis in a variety of settings. In this study, using the metastatic 4T1 and non-metastatic 4T07 cell lines, we demonstrate that 4T1 cells exhibit significantly increased OPN, ILK, MMP-2 and uPA expression in contrast to 4T07 cells. Blockade of OPN binding to 4T1 cell-surface integrins by the competitive ligand inhibitor, RGD, decreases OPN, ILK, MMP-2 and uPA expression. Conversely, exposure of 4T07 cells to exogenous OPN increases ILK, MMP-2 and uPA levels. In addition, OPN–integrin binding in 4T1 with subsequent activation of ILK results in binding of AP-1 to MMP-2 and uPA promoter DNA, as determined by both ChIP and EMSA assays, and increased in vitro promoter activation, as measured by transient transfection assays using MMP-2 and uPA promoter-reporter constructs. AP-1-mediated transactivation is ablated by transfection of DN-ILK or exposure to RGD. Our results indicate that binding of an RGD-bearing ligand, such as OPN, to integrin receptors in metastatic 4T1 cells transcriptionally mediates MMP-2, uPA and OPN expression through ILK-dependent AP-1 activity. Functionally, OPN-dependent ILK activity is associated with in vitro correlates of more aggressive metastatic behavior.
This work was supported by NIH grants AI44629, GM665113, GM069331 to P.C.K. and SUS-Ethicon Research Fellowship to P.Y.W.
Conflict of Interest Statement : None declared.
References
Denhardt,D.T., Lopez,C.A., Rollo,E.E., Hwang,S.M., An,X.R. and Walther,S.E. (
Denhardt,D.T., Giachelli,C.M. and Rittling,S.R. (
Weber,G.F., Ashkar,S. and Cantor,H. (
Denhardt,D.T. and Chambers,A.F. (
Brown,L.F., Papadopoulos-Sergiou,A., Berse,B., Manseau,E.J., Tognazzi,K., Perruzzi,C.A., Dvorak,H.F. and Senger,D.R. (
Agrawal,D., Chen,T., Irby,R., Quackenbush,J., Chambers,A.F., Szabo,M., Cantor,A., Coppola,D. and Yeatman,T.J. (
Coppola,D., Szabo,M., Boulware,D., Muraca,P., Alsarraj,M., Chambers,A.F. and Yeatman,T.J. (
Das,R., Mahabeleshwar,G.H. and Kundu,G.C. (
Fedarko,N.S., Jain,A., Karadag,A., Van Eman,M.R. and Fisher,L.W. (
Gotoh,M., Sakamoto,M., Kanetaka,K., Chuuma,M. and Hirohashi,S. (
Grano,M., Mori,G., Minielli,V., Colucci,S., Vaira,S., Giannelli,G., Martemucci,S., Giorgino,F. and Zallone,A.Z. (
Pantel,K. and Brakenhoff,R.H. (
Ahmad,S.A., Berman,R.S. and Ellis,L.M. (
Berman,R.S., Portera,C.A.,Jr and Ellis,L.M. (
Mi,Z., Guo,H., Wai,P.Y., Gao,C., Wei,J. and Kuo,P.C. (
Heppner,G.H., Miller,F.R. and Shekhar,P.M. (
Hannigan,G.E., Leung-Hagesteijn,C., Fitz-Gibbon,L., Coppolino,M.G., Radeva,G., Filmus,J., Bell,J.C. and Dedhar,S. (
Gao,C., Guo,H., Wei,J., Mi,Z., Wai,P. and Kuo,P.C. (
Hannigan,G., Troussard,A.A. and Dedhar,S. (
Denhardt,D.T., Noda,M., O' Regan,A.W., Pavlin,D. and Berman,J.S. (
Attur,M.G., Dave,M.N., Stuchin,S., Kowalski,A.J., Steiner,G., Abramson,S.B., Denhardt,D.T. and Amin,A.R. (
Craig,A.M. and Denhardt,D.T. (
Denhardt,D.T. and Guo,X.J. (
O'Regan,A. and Berman,J.S. (
Weber,G.F. (
Senger,D.R., Asch,B.B., Smith,B.D., Perruzzi,C.A. and Dvorak,H.F. (
Senger,D.R., Perruzzi,C.A., Gracey,C.F., Papadopoulos,A. and Tenen,D.G. (
Bautista,D.S., Xuan,J.W., Hota,C., Chambers,A.F. and Harris,J.F. (
Hijiya,N., Setoguchi,M., Matsuura,K., Higuchi,Y., Akizuki,S. and Yamamoto,S. (
Carlinfante,G., Vassilioul,D., Svensson,O., Wendel,M., Heinegard,D. and Andersson,G. (
Korkola,J.E., DeVries,S., Fridlyand,J., Hwang,E.S., Estep,A.L., Chen,Y.Y., Chew,K.L., Dairkee,S.H., Jensen,R.M. and Waldman,F.M. (
Rangaswami,H., Bulbule,A. and Kundu,G.C. (
Rangaswami,H., Bulbule,A. and Kundu,G.C. (
Das,R., Mahabeleshwar,G.H. and Kundu,G.C. (
Philip,S., Bulbule,A. and Kundu,G.C. (
Philip,S. and Kundu,G.C. (
Tuck,A.B., Hota,C. and Chambers,A.F. (
Tuck,A.B. and Chambers,A.F. (
Tan,C., Cruet-Hennequart,S., Troussard,A., Fazli,L., Costello,P., Sutton,K., Wheeler,J., Gleave,M., Sanghera,J. and Dedhar,S. (
Edwards,L.A., Thiessen,B., Dragowska,W.H., Daynard,T., Bally,M.B. and Dedhar,S. (
Marotta,A., Parhar,K., Owen,D., Dedhar,S. and Salh,B. (
Persad,S., Attwell,S., Gray,V., Delcommenne,M., Troussard,A., Sanghera,J. and Dedhar,S. (
Somasiri,A., Howarth,A., Goswami,D., Dedhar,S. and Roskelley,C.D. (
Tan,C., Costello,P., Sanghera,J., Dominguez,D., Baulida,J., de Herreros,A.G. and Dedhar,S. (
White,D.E., Cardiff,R.D., Dedhar,S. and Muller,W.J. (
Attwell,S., Roskelley,C. and Dedhar,S. (
Filipenko,N.R., Attwell,S., Roskelley,C. and Dedhar,S. (
Novak,A., Hsu,S.C., Leung-Hagesteijn,C., Radeva,G., Papkoff,J., Montesano,R., Roskelley,C., Grosschedl,R. and Dedhar,S. (