MCF10A cells were obtained from the American Type Culture Collection and LNCaP cells were obtained from European Collection of Cell Cultures. MCF10A was cultured in DMEM-F12 medium (Invitrogen, Paisley, UK) as previously described (18). LNCaP cells were cultured in RPMI (Invitrogen) containing 10% foetal calf serum (FCS) (Sigma-Aldrich, Dorset, UK). All cell lines were maintained at 37°C and 5% CO₂ in the tissue culture incubator and were routinely tested using MycoAlert^® Mycoplasma detection assay (Lonza, Walkersville, MD, USA).

CSMD1 shRNA pRS vectors (HuSH™, OriGene, Rockville, MD, USA) were transformed into NEB 5-α competent E. coli (New England BioLabs, Ipswich, MA, USA) and grown in L broth containing 50 µg/ml ampilicillin (Sigma-Aldrich). Sequences of CSMD1 shRNA constructs and their cognate CSMD1 mRNA regions are shown in Table I. For each cell line, 1×10⁶ cells/ml were transfected with either shCSMD1 or shcontrol constructs using Lipofectamine™ 2000 (Invitrogen). Cells were then grown in media that contained the appropriate dose (0.5 µg/ml, MCF10A; 0.01 µg/ml, LNCaP) of puromycin (Sigma-Aldrich) for 2 weeks. Single cell colonies were selected by serial dilutions in 96-well plates.

RNA was extracted from cells using TRIzol (Invitrogen). First strand cDNA synthesis was then performed using ThermoScript (Invitrogen) and random hexamers (Promega, Southampton, UK) following the manufacturer's instructions. To test the specificity of shRNA constructs, CSMD3 was amplified. PCR reactions were performed using HotStarTaq (Qiagen, West Sussex, UK) following the manufacturer's instructions. The final extension step was for 10 min at 68°C. Annealing temperature of the CSMD3 primers was 52°C and 50°C for the housekeeping gene RPLP0. PCR products were analysed in 1.5% agarose gel containing ethidium bromide. For primer sequences of CSMD3 and the housekeeping gene RPLP0, see Table II.

CSMD1 and RPLP0 expression levels were investigated in all shCSMD1 and shcontrol clones using the relative standard curve method. TaqMan reactions, using sensiMix dT master mix (Quantace, London, UK), were performed in duplicates and run on an ABI7500 (Applied Biosystems, Warrington, UK). For sequences of primers and probes, see Table III.

IHC was performed to confirm the knockdown of CSMD1 protein expression in shRNA clones. Cells were pelleted, formalin fixed and paraffin-embedded and CSMD1 was stained using chicken anti-CSMD1 antibody at a 1:3000 dilution, as previously described (14). Negative controls, in which the pre-immune serum was applied, and positive controls of normal breast tissue were included in each batch of IHC.

Cell viability was evaluated using the MTT (3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide) assay. Cells (1×10⁴ cells/ml) were incubated with MTT solution (1 mg/ml) (Sigma-Aldrich) for 3–4 h at 37°C. Crystals were dissolved in propan-1-ol and their optical densities (OD) were quantified at 570 nm (Opsys MR Plate reader; Dynex Technologies, West Sussex, UK).

Confluent monolayers were incubated in mitomycin C (10 µg/ml) (ICN Biomedicals, Costa Mesa, CA, USA) for 2 h at 37°C to inhibit cell division. Wounds were introduced into the monolayer by scratching with a P200 micropipette tip. Wound closure was followed for 96 h and phase contrast images were captured using an Olympus digital still camera attached to an Olympus inverted microscope with a 4X objective lens. The wound area was imaged using ImageJ and expressed as a percentage of the wound area covered at each time point relative to the surface area of the wound at time zero.

Invasion assays were carried out as previously described (20). The upper sides of 12-well format transwell inserts (BD Biosciences, Franklin Lakes, NJ, USA) were coated with Matrigel at a 1:10 dilution in serum-free medium (SFM). The undersides of the inserts were coated with fibronectin (10 µg/ml in PBS). Medium containing FCS (1 ml) was added to the base of each well. Cells at 5×10⁴ cells/ml in SFM were placed on the top of inserts, for 16 h. Invasive cells that penetrated through to the underside of the inserts were identified by fixation in 4% paraformaldehyde and staining with crystal violet. Cells were then destained using 2% SDS and the OD was measured using the Opsys MR Plate reader at 570 nm. The percentage of invasion was calculated independently using the following equation: Cell invasion (%) = OD of cells in the bottom/OD of cells in the bottom + OD of cells in the top.

Plates (96-well) were coated with fibronectin (10 µg/ml in PBS) or Matrigel (20 µg/ml in SFM). Non-specific adhesion sites were blocked by pre-incubation in 0.5% bovine serum albumin in DMEM or RPMI. Cells at 4×10⁵ cells/ml were pipetted onto the coated plates at 37°C and allowed to attach for 30 min or 2 h. Following this, unattached cells were aspirated and the remaining attached cells were fixed in 4% paraformaldehyde and stained with crystal violet. Cells were then destained using 2% SDS and OD was measured using the Opsys MR Plate reader at 570 nm.

3D Matrigel cultures of MCF10A cells were performed as previously described (18). Briefly, 8-well glass slides (Nunc™; Nalge Nunc International Corp., Rochester, NY, USA) were coated with growth factor reduced Matrigel (BD Biosciences). shcontrol or shCSMD1 MCF10A cells (5000 cells/ml in 2% Matrigel assay medium) were seeded onto the slides and incubated at 37°C for up to 26 days. Cells were fed every 3–4 days with 2% Matrigel assay medium.

Fluorescent staining of acinar structures was performed as previously described (18). Phalloidin (Invitrogen), was used as a cytoskeletal marker at a 1:50 dilution. Rabbit polyclonal active caspase-3 antibody (Abcam, Cambridge, UK) was used as an apoptotic marker at a 1:100 dilution. Fluorescence staining was imaged using a Nikon confocal microscope (Nikon Eclipse TE2000-E; Nikon UK, Kingston Upon Thames, UK). Images of acinar structures were captured by serial confocal sectioning and viewed using confocal software. Phase contrast images were captured using an Olympus digital still camera attached to an Olympus inverted microscope (Olympus CKX41). All images were converted to TIFF format and figures were assembled and annotated using Adobe Photoshop 7.0^®.

Initially the distribution of the data from all the functional assays except wound healing assay was tested using Shapiro-Wilk or Kolmogorov-Smirnov tests. When the data were normally distributed, a t-test was used to compare between means. A Mann-Whitney U test (non-parametric) was used when the data were not normally distributed. All tests were 2-sided and performed using SPSS software version 15. A p-value ≤0.05 was considered to be statistically significant.

To help determine the function of CSMD1, shRNA gene silencing methodology was used to create stable cell lines with suppressed CSMD1 expression. Two different cell lines were analysed: MCF10A (normal breast) and LNCaP (prostate). Single cell colonies were selected from all cell lines and the mRNA expression level of CSMD1 was investigated using qRT-PCR. Colonies that showed the best level of mRNA knockdown were chosen for subsequent experiments. Two separate shCSMD1 cell lines were generated from the MCF10A cells (termed clone 1 and 2) and a single cell line was generated from the LNCaP cells. MCF10A shCSMD1 clones 1 and 2 showed a 66% (SD±8.9%) and 66.9% (SD±12%) reduction in CSMD1 mRNA expression, respectively, when compared to shcontrol cells. The LNCaP cell line showed a 62% (SD±12%) CSMD1 mRNA knockdown, relative to shcontrols (Fig. 1A).

To confirm these levels of knockdown at the protein level, CSMD1 protein expression was investigated using IHC since at the time a CSMD1 antibody suitable for westerns was not available. MCF10A shCSMD1 colonies showed lower CSMD1 protein expression compared to the shcontrol. However, unexpectedly, the LNCaP shCSMD1 cells did not show a reduced level of protein expression compared to shcontrols cells in contrast to the high reduction in CSMD1 mRNA expression observed by qRT-PCR (Fig. 1B).

To test the specificity of the shCSMD1 constructs the expression of CSMD3 mRNA was investigated using RT-PCR. Since normally none of the tested cell lines express CSMD2 its expression was not tested. The shCSMD1 constructs did not affect the expression pattern of CSMD3 in either of the cell lines examined. Moreover, the shcontrol plasmids had no effect on the expression pattern of CSMD3 (Fig. 1C).

To investigate the morphology of the shCSMD1 cells, the F-actin stain Phalloidin was used to examine the cells. Staining revealed that in all lines examined, loss of CSMD1 expression resulted in misshapen cells that tended to grow individually. In contrast, shcontrol cells mainly grew as colonies. In MCF10A, many shCSMD1 cells possessed lamellipodia-like protrusions suggestive of a motile phenotype (Fig. 2, arrow). LNCaP shCSMD1 cells displayed very long filopodia-like protrusions (Fig. 2, arrows). In all cases these features were not observed in the appropriate shcontrol cell lines.

Having shown that CSMD1 silencing affected cell morphology, using MCF10A cells as a model we focused on understanding its functional effects by investigating a number of important cellular processes. Since minimal knockdown at the protein level was observed in the LNCap cell line further functional work was not performed in these cells. Cell proliferation was assayed using the MTT protocol. Silencing CSMD1 expression did not significantly affect cell proliferation until 72 h post-plating when increases of 34% (SD±4%, p=0.003) and 56% (SD±3.3%, p=0.001) were observed in MCF10A shCSMD1 clones 1 and 2 (Fig. 3).

Since the morphology of the MCF10A shCSMD1 cells suggested that they were more motile than their shcontrol counterparts (Fig. 2), we next investigated the role of CSMD1 in cell migration. Wound healing assays were performed after inhibiting MCF10A proliferation by treating cells with mitomycin C prior to the experiment. In the shcontrol cells 14.15% (SD±3.6%) and 31.55% (SD±4.3%) of the wound were covered during the first 16 h and 24 h after wounding respectively. In the shCSMD1 clones 1 and 2, 37.02% (SD±8.5%, p=0.07) and 56.15% (SD±7.1%, p=0.018) of the wound were covered during the first 16 h. Moreover, after 24 h, cells in shCSMD1 clones 1 and 2 covered 69.9% (SD±6.6%, p=0.02) and 88.2% (SD±0.9%, p=0.003) of the wound (Fig. 4A and B), confirming that suppression of CSMD1 expression promotes a more motile phenotype in MCF10A cells.

Since loss of CSMD1 promoted cell migration, we next studied its role in cell invasion. At 16 h after plating, suppression of CSMD1 expression resulted in 33% (SD±3%, p<0.001) and 26% (SD±6%, p<0.001) increases in cell invasion of MCF10A shCSMD1 clones 1 and 2, respectively, compared to shcontrol (Fig. 4C).

Shared homology with Cub domain containing protein 1 (CDCP1), which is involved in cell-matrix interactions (21,22), suggested that CSMD1 could play a role in cell adhesion. To examine this possibility adhesion assays using Matrigel and fibronectin were performed 30 min and 2 h after cell plating. Loss of CSMD1 expression caused ~34% (SD±8.5%, p=0.001) and ~17% (SD±15%, p=0.06) decreases in the adhesion of MCF10A shCSMD1 clones 1 and 2 to fibronectin after 30 min, compared to shcontrol. After 2 h, this difference had increased to ~36% (SD±10%, p=0.01) and ~40% (SD±15%, p=0.004) respectively (Fig. 5A). Following CSMD1 silencing, adhesion of MCF10A shCSMD1 clones 1 and 2 to Matrigel decreased by 20% (SD±13.5%, p=0.02) and 16% (SD±8.5%, p=0.05) after 30 min compared to shcontrol. Strikingly, 2 h after plating these differences had increased to 44% (SD±7.6%, p=0.0006) and ~42% (SD±9.6%, p=0.001) (Fig. 5B).

We have previously observed that CSMD1 is highly expressed in well-differentiated areas of breast cancer samples, suggesting that CSMD1 might play a role in mammary duct formation (14). To examine this further, an MCF10A 3D Matrigel model was established and the effects of CSMD1 knockdown on the morphogenesis of mammary acini were investigated. Loss of CSMD1 expression resulted in a larger number of acini and larger areas of flat cells (Fig. 6A, arrows). Acini in 3 fields from 3–5 independent experiments were counted and averaged every 2–4 days. There was a statistically significant increase in the average number of shCSMD1 acini compared to shcontrol at all time points (p<0.05) (Fig. 6B).

Phalloidin staining and confocal microscopy revealed that shCSMD1 acini were irregular in shape and heterogeneous in size (Fig. 7A). The diameter of each acinus was measured and the mean diameter of acini from shcontrol cells determined (85 µm). This was then used as a cut-off to distinguish between large and small acini. Noteworthy, the proportion of large acini in MCF10A shCSMD1 clones 1 and 2 were 33% (SD±21%, p=0.1) and 43% (SD±15%, p=0.03) higher than in the shcontrol cell line (Fig. 7B). Furthermore, confocal imaging revealed that in shCSMD1 clones 1 and 2, respectively, 55% (SD±9.8%, p=0.045) and 90% (SD±8%, p=0.008) of acini failed to form a lumen compared to shcontrol cells (Fig. 7C).

In this model system lumen formation depends on apoptosis, raising the possibility that a failure to generate acini with a lumen might be due to differences in developmentally regulated programmed cell death in cells where CSMD1 expression is inhibited. To test this, acini at day 12 (before the completion of lumen formation in shcontrol cells) and day 26 were stained with antibodies specific for active caspase-3. At day 26, lumens of shcontrol acini showed very weak staining compared to shCSMD1 lumens (Fig. 7A, arrows). However, at day 12, shcontrol acini showed strong staining of active caspase-3 (Fig. 7D). These observations confirm that during normal acinar development in this model system a lumen is generated by temporally coordinated apoptotic processes. Apoptosis is then severely suppressed or absent after the completion of lumen in shcontrol acini. In contrast, because shCSMD1 acini failed to form a lumen, apoptotic processes remain active until day 26.