CaOV3 and SkOV3 human ovarian cancer cells were grown in Dulbecco's modified Eagle medium (Invitrogen) containing 10% heat-inactivated fetal bovine serum (FBS; Hyclone). OVCA429 human ovarian cancer cells (gift of Dr. B. Vanderhyden, Ottawa Regional Cancer Centre) were grown in Minimal essential medium Eagle-alpha (Invitrogen) containing 10% FBS and supplemented with 0.1 mM non-essential amino acids (Invitrogen). Primary cultures of human OSE and EOC cells were isolated and maintained as previously described 25 . Treatment of OVCA429 cells with recombinant human BMP4 (R&D Systems) was performed as described previously 2 . Institutional approval for research with human materials was received prior to the initiation of these studies (QEII Health Sciences Centre, Research Ethics Committee, #QE-RS-99-016; IWK/Grace Hospital Research Ethics).

OVCA429 Tet-On cells (429T cells) were generated by transfecting OVCA429 cells with pTet-On vector (Clontech) using GeneJuice reagent (Novagen), followed by selection with 1 mg/mL Geneticin™ (Invitrogen). Of the 27 separate clones generated, 5 clones displayed detectable doxycycline-inducible activity, as assessed by transient transfection of the pTRE2-luc vector and followed by luciferase assays performed as previously described 2 . Doxycycline hyclate (Dox; Sigma) was used at a concentration of 2 μg/mL in all experiments, unless otherwise indicated. Dox-inducible ALK3^QD-expressing cells (429T-ALK3^QD) were generated by transfecting 429T cells with pTRE2-ALK3^QD-HA (original pCMV5-Alk3QDHA plasmid was a gift from Dr. L. Attisano, U. of Toronto), and selection with 250 μg/mL Hygromycin B (Invitrogen). Five different 429T-ALK3^QD cell lines possessing Dox-inducible ALK3^QD expression were identified by Western blot analysis, and two independent clones (clones 429T-A44 and -54) were chosen for further examination. Subsequent growth of 429T and 429T-ALK3^QD cell lines was maintained using 100 μg/mL Geneticin™ and 100 μg/mL Hygromycin B.

Northern analysis was performed as previously described 2 , using radiolabelled probes synthesized from human cDNA fragments of ALK3 (nucleotides 145-594), ALK6 (nucleotides 907-1256), BMPR2 (nucleotides 2561-3010), GAPDH (nucleotides 270-620), ID1 (nucleotides 189-549), and ID3 (nucleotides 356-750) as templates.

Protein isolation and subsequent Western analyses were performed as previously described 2 . ALK3^QD protein expression was detected using anti-HA-Peroxidase 3F10 monoclonal antibody (1:1000 dilution; Roche). The anti-actin polyclonal antibody (1:1000 dilution; Sigma) followed by incubation with horseradish peroxidase-conjugated sheep anti-rabbit IgG secondary antibody (1:5000 dilution; Chemicon) was used as a control for protein loading.

All studies conformed to the approved animal utilization protocol and Canadian Council for Animal Care guidelines. Eight- to 10-week-old female CD-1 nu/nu athymic nude mice (Charles River) were maintained in a sterile barrier animal facility. After approximately one week from receipt, 30 mice were started on a diet of gamma-irradiated rodent chow containing Dox at a concentration of 1000 ppm (Research Diets) for 2 d, and another group of 30 mice remained on the in-house rodent chow diet. Mice were supplied with food and water ad libitum. Each mouse was injected i.p. with a suspension of 5 × 10⁵ cells (either 429T-ALK3^QD or 429T control) in a volume of 0.2 mL sterile 1 × PBS, resulting in four groups of fifteen mice (429T-ALK3^QD or 429T controls, with or without Dox treatment). Mice were monitored for 90 d post-injection for signs of ascites formation or visible morbidity. At the experimental endpoint (90 d), mice were sacrificed and dissection was performed to assess and quantify ascites production and tumour formation.

Tissues harvested for histological analysis were fixed immediately in 4% paraformaldehyde/1 × PBS, paraffin-embedded and sectioned at 5 μm then stained with haematoxylin and eosin (tissue processing performed by Molecular Pathology, Robarts Research Institute, UWO). Microscopic images of stained tissue sections were captured using an Olympus IX70 inverted microscope and Image Pro 6.2 software, and subsequently adjusted for brightness/contrast and colour balance using Adobe Photoshop 7.0 software.

Confirmation of ALK3 ^QD transgene expression in tumours that formed in nude mice was performed by RT-PCR analysis of total RNA isolated from homogenized tissue using the Total RNA isolation kit (Sigma) as per manufacturer's instructions. Reverse transcription was performed with 2 μg of RNA reverse transcribed into cDNA using oligo-dT decamers as primers and Superscript II reverse transcriptase (Invitrogen) as per manufacturer's instructions. Subsequent PCR was carried out using oligonucleotides that specifically amplify the 3' end of the ALK3^QD-HA cDNA construct. Human GAPDH mRNA expression served as a control for tumour xenograft material in each sample.

For quantitative RT-PCR analysis, total RNA was isolated from cells treated with 2 μg/mL Dox for 2 days, or left untreated, and 2 μg of RNA was subsequently reverse transcribed into cDNA using oligo-dT decamers as primers and Superscript II reverse transcriptase (Invitrogen) as per manufacturer's instructions. PCR was performed using the Brilliant SYBR green QPCR Master Mix (Stratagene), and real-time measurement of the PCR reactions was recorded using the Mx3000P Real-time PCR System (Stratagene), and quantified using the 2^-ΔΔCt method 26 ; GAPDH expression was used for normalization, and the fold change in mRNA expression was calculated versus untreated cell samples.

All human gene-specific primer sequences used in RT-PCR are available upon request.

Cells were seeded at 2 × 10⁵ cells in 6-well dishes and 24 h later were treated with 2 μg/mL Dox, or left untreated. Cells were grown for 2-3 d until confluent monolayers were achieved, then a wound was created by scratching the wells with a sterile plastic pipette tip (~1 mm space). After several gentle washes with 1 × PBS, media was replaced (with or without Dox) and cells were monitored at multiple timepoints and photographed at 24 h. Photographic images were captured using a Nikon Coolpix digital camera and Nikon inverted phase contrast microscope at 100 × magnification.

For assessment of cell detachment, cells were seeded at 5 × 10⁵ cells in 6-well dishes, and 24 h later were treated with 2 μg/mL Dox, or left untreated. Detached cells were quantified as previously described 2 .

In adhesion experiments, cell lines were treated with 2 μg/mL Dox, or left untreated, and grown for 2-3 d until confluent monolayers were achieved. Adhesion assays were performed as previously described 27 . Cells were plated at a density of 1 × 10⁵ cells/well into 24-well dishes previously coated with 500 ng/cm² fibronectin or 200 ng/cm² vitronectin (Sigma), and blocked with 1% BSA.

429T and 429T-ALK3^QD cells were pre-treated with 2 μg/mL Dox for 2 days, or left untreated, while grown in monolayer culture. One hundred thousand cells/well were then seeded in quadruplicate into a 24-well Ultra-Low Attachment cluster plate (Corning) and the same culture conditions (i.e. +/- Dox) were maintained during spheroid formation. Images were captured at the 2-day timepoint using an Olympus IX70 inverted microscope at 100 × magnification and Image Pro 6.2 software.

Statistical analyses were performed using the nonparametric Mann-Whitney test with 95% confidence intervals (GraphPad Prism 4). Values of significance are indicated in the Legends to Figures.

Previous work from our laboratory demonstrated that primary cultures of normal human OSE cells and EOC cells possess an intact BMP4 signalling pathway, but more importantly, that primary EOC cells exhibit striking changes in morphology, adhesion, motility and invasion in response to BMP4 stimulation 2 23 . Additionally, EOC cells exhibit heightened responses in gene expression following treatment with exogenous BMP4 in comparison to normal OSE cells; for example, ID1 gene expression is increased ~10- to 15-fold in EOC cells, compared to 2- to 3-fold in normal OSE 2 23 . To further our understanding of the potential differential responses to BMP4 signalling in these cell types, we quantified BMP receptor expression levels in normal OSE and EOC cells. BMP4 signalling is mediated primarily by the ALK3, ALK6, and BMPR2 receptors. In Northern blot analyses using RNA isolated from actively growing primary cultures of OSE and EOC cells, and the EOC cell lines CaOV3, SkOV3, and OVCA429, we readily detected expression of the type I BMP receptor ALK3 and the type II receptor BMPR2 mRNA (Fig. 1). After normalization to GAPDH expression, no significant differences in the mRNA level of ALK3 or BMPR2 were detectable when normal OSE is compared with primary EOC samples. Expression of ALK6 mRNA was largely undetectable in all primary cell samples. Additionally, we have determined that BMPR2 protein expression as well as the BMP-specific R-Smads, Smad1 and Smad5, remain unchanged between primary OSE and EOC cells (T Shepherd & M Nachtigal, unpublished observations). By comparison, immortalized EOC cell lines expressed higher levels of ALK3, ALK6, and BMPR2 than the primary cell samples. These data are consistent with quantitative RT-PCR results (data not shown) suggesting that ALK3 and BMPR2 are likely the chief receptors used by BMP4 to confer activated signalling in primary cultures of human OSE and EOC cells.

Our goal was to examine the consequence of BMP signalling activity on the metastatic potential of ovarian cancer cells. We engineered OVCA429 cells that possess Dox-inducible expression of the mutated BMP type I receptor ALK3^QD. The ALK3^QD receptor harbours a Q-to-D point mutation at amino acid 233 in the GS domain thereby replacing the necessary activation by BMPR2-mediated phosphorylation in response to BMP ligand binding 28 . We chose to use OVCA429 cells since they have a BMP receptor expression profile similar to primary OSE and EOC cells (Fig. 1), and exhibit robust BMP4 signalling activity in response to BMP4 treatment (Fig. 2A). We generated five independent clones of 429T-ALK3^QD cells, and two of these exhibited low basal expression in untreated cells with readily-detectable ALK3^QD expression in response to treatment with 2 μg/mL Dox for 24 h (Fig. 2B). As a functional readout to confirm intact ALK3^QD receptor signalling, we observed Dox-induction of ID1 and ID3 mRNA expression in 429T-ALK3^QD cells as compared with 429T controls (Fig. 2C). High basal levels of ID1 expression in untreated cells is due to the presence of serum in the growth media required to sustain rtTA expression and Dox-inducible activity in Tet-On cells.

OVCA429 cells consistently form ascites and intraperitoneal tumour lesions when injected directly into the peritoneal cavity of nude mice 24 thereby providing a useful in vivo model to study ovarian cancer metastasis. Female CD-1 nu/nu mice were injected i.p. with 5 × 10⁵ cells of 429T-ALK3^QD or 429T controls. Mice were subsequently divided into two groups (n = 15 per group), a group was either fed a normal chow diet or Dox-containing chow to induce ALK3^QD expression. After an experimental endpoint of 90-100 d, tumours formed on several surfaces of the peritoneal cavity and measurable ascites developed in mice in all four treatment groups (Table 1). Notably, the proportion of Dox-treated mice injected with 429T-ALK3^QD cells that developed measurable ascites was reduced when compared to control mice (28.6% versus 40-46.7%). In addition, we observed a decrease in the mean number of macroscopic tumour nodules (≥1 mm) that formed within the peritoneal cavity due to ALK3^QD expression (Fig 3A). Together, ALK3^QD expression was associated with reduced intraperitoneal tumour and ascites formation compared with controls; however, this difference was not statistically significant. We noted that the number of animals developing tumours in the untreated (minus Dox) 429T-ALK3^QD group had a lower mean number of tumours (6.2 ± 2.0) than the 429T control groups. This raises the possibility that ALK3^QD is expressed at very low levels in the absence of Dox, undetectable by expression analysis, while still producing a modest functional response. Expression of ALK3^QD was confirmed in tumours that developed in Dox-treated 429T-ALK3^QD-injected mice (Fig. 3B). Macroscopic tumours adherent to the peritoneal wall were observed in tumour-bearing mice, and histologic analysis demonstrated foci of penetration into the underlying smooth muscle (Fig. 3C), but this invasion was not evident in tumours isolated from Dox-treated 429T-ALK3^QD mice. The majority of tumour nodules were detected in the adipose tissue of the omentum juxtaposed to the pancreas, and the serosa of the spleen, intestine and liver (Table 1). Interestingly, only one Dox-treated 429T-ALK3^QD female mouse harboured experimental metastasis to the omentum, the most common site for tumours in all other groups.

We next performed a series of experiments to investigate the underlying basis for reduced in vivo metastatic potential of ovarian cancer cells due to constitutively-active BMP signalling. A number of genes involved in cancer metastasis are putative and/or known targets for active BMP signalling in ovarian cancer cells 23 . A well-established and distinguishing molecular signature of epithelia-mesenchymal transition and increased malignancy is the upregulated expression of Snail and Slug transcription factors (encoded by SNAI1 and SNAI2, respectively) and the subsequent downregulation of their target gene CDH1 encoding E-cadherin 29 30 . By performing quantitative RT-PCR we determined that Dox-treatment of 429T-ALK3^QD cells led to the 4-fold upregulation of SNAI1 (Snail) and 2.5-fold upregulation of SNAI2 (Slug) when compared with 429T control cells (Fig. 4). In a reciprocal fashion, ALK3^QD signalling reduced CDH1 expression by more than 40% versus control cells. As a positive control for activated BMP signalling, we detected a substantial increase in both ID1 (~8-fold) and ID3 (~9-fold) mRNA expression in ALK3^QD-expressing cells (Fig. 4).

Cell adhesion is likely a key checkpoint regulating the efficient spread of ovarian cancer cells during metastasis. Potential molecular targets of ALK3^QD signalling mediating the observed reduction in ovarian cancer cell adhesion are genes encoding integrins and/or extracellular matrix (ECM) components. We performed gene expression analyses for several specific integrins and ECM components known to be expressed by OSE and EOC cells 31 32 . Dox-treated 429T-ALK3^QD cells possessed a 64% decrease in ITGB1 (β₁-integrin) and 43% decrease in ITGB3 (β₃-integrin) mRNA expression versus 429T controls (Fig. 4). In addition, we observed a reciprocal elevated expression of ECM genes FN1 (fibronectin) and VTN (vitronectin) mRNA in ALK3^QD-expressing cells by 49% and 69%, respectively.

Similar to previous observations when primary EOCs and established EOC cell lines were treated with exogenous BMP4 ligand 2 23 , induction of ALK3^QD did not affect the overall growth rate of OVCA429 cells (data not shown). We observed, however, that when Dox-treated cells reached confluence, both 429T-A44 and 429T-A54 cells had an altered morphology as compared with Dox-treated 429T cells or untreated cells (Fig. 5A). In the untreated state, OVCA429 cells appear as a cobblestone epithelial monolayer. Induction of ALK3^QD expression in 429T-A44 and 429T-A54 cells produced a more mesenchymal-like cellular phenotype than uninduced cells, with clusters of refractile, spindle-shaped cells.

To address whether this altered phenotype in response to ALK3^QD expression affected the motility of the cells, standard in vitro wound-healing assays were performed. There was no difference in motility rates among Dox-treated and untreated 429T, 429T-A44 and 429T-A54 cells (data not shown). When cells were observed during the progress of cell migration (i.e. at 24 hours after wounding), Dox-treated 429T-A44 and 429T-A54 cells possessed a more spindle-shaped morphology as they migrated into the manufactured wound space with cells possessing long, cytoplasmic projections extending in the direction of cell movement (Fig. 5B). Interestingly, 429T and 429T-ALK3^QD cells are poorly invasive through matrices composed of either rat tail collagen or gelatin, regardless of Dox-treatment (data not shown). This lack of invasiveness is identical to the parental OVCA429 cell line (Mujoomdar & Nachtigal, unpublished observations). Taken together, these data provide further validation that ALK3^QD expression in OVCA429 cells results in morphological changes paralleling the observed alterations in EMT marker expression.

A critical step controlling ovarian cancer metastasis is the ability of EOC cells to adhere to surfaces at secondary sites 1 . The ALK3^QD-mediated decreases in β₁- and β₃-integrin mRNA expression in 429T-ALK3^QD ovarian cancer cells would predict a decrease in cell adhesion. Timed cell-detachment assays demonstrated that Dox-treated 429T-ALK3^QD cells have increased sensitivity to trypsinization when compared to control cells (Fig. 6A). To directly investigate the adhesive properties of these cells to specific substrates, we performed adhesion assays using tissue culture dishes precoated with the extracellular matrix (ECM) components fibronectin (FN) and vitronectin (VTN), which are produced by EOC cells 32 . As compared to 429T control cells, 429T-ALK3^QD cell lines had a reduced ability to attach to FN- and VTN-coated wells when induced to express ALK3^QD (Fig. 6B&C).

Multicellular aggregates of EOC cells, termed spheroids, are present in the ascites of patients and are hypothesized to play an essential role in ovarian cancer metastasis 33 . Cell-cell cohesion in spheroids is mediated primarily by cadherins junctions and integrin interactions 34 . Thus, we sought to determine whether the reduced E-cadherin and β1-/β3-integrin expression due to ALK3^QD signalling affects the ability of OVCA429 cells to form spheroids. After culturing cells for 2 d on Ultra-Low Attachment cluster plates, Dox-treated 429T-ALK3^QD spheroids were smaller in size and less dense-appearing than untreated cells and 429T control cells (Figure 7).