MMTV-Wnt1 (a gift from Dr. Harold Varmus) and MMTV-ILK [21] mice were backcrossed to the FVB strain. MMTV-Wnt1 males were then bred with MMTV-ILK female to generate the double-transgenic MMTV-Wnt/ILK mice. Mice were tail clipped and genotyped by PCR by using forward 5'-CCACACAGGCATAGAGTGTCTGC -3' and reverse 5'-GGACTTGCTTCTTATAGCC-3' primers for Wnt1 transgene [25] and forward 5'-CATGTATGCACCTGCCTG-3' and reverse 5'- TATGTCACACCACAGAAG 3' primers for the ILK transgene [21]. Nulliparous females were monitored biweekly for the appearance of a small palpable tumor nodule in the mammary fat pads. Ages were recorded at the time of tumor discovery and tumors were measured every 2 to 3 days with a calliper by using the modified ellipsoid formula (LxW²)/2 [26]. All animal procedures were done in accordance with protocols approved by the Institution Animal Care Committee at the Jack Bell Research Centre and University of British Columbia in Vancouver.

For whole-mount staining, the fourth mammary glands from transgenic and control mice were harvested and fixed in Carnoy's fixative (60% ethanol, 30% chloroform, 10% acetic acid) for 2 to 4 h, rehydrated and stained in carmine alum (Stem Cell Technologies, Vancouver, BC, Canada) overnight. After dehydration, samples were spread on a microscope slide, cleared in xylene, and images were acquired by using a dissecting microscope. For histologic analysis, mammary and tumor tissue samples were fixed in 10% neutral buffer formalin overnight at room temperature and were subsequently rinsed and kept in 70% ethanol until embedded in paraffin and sectioned at 5-μm thickness. Sections were either stained with hematoxylin and eosin (H&E) or processed for immunohistochemical analysis and counterstained by following standard procedures. Sections were incubated with rabbit anti FOXA1 (Abcam ab55178, Cambridge, MA), monoclonal β-catenin (BD-610153), polyclonal CD24 (BD-557436) (BD Transduction Laboratories, Mississauga, ON, USA), monoclonal PCNA (sc-56), polyclonal Ki-67 (Neomakers RM9106-S0) and polyclonal ER-α sc-542 (Santa Cruz Biotechnology Inc., Santa Cruz, CA) antibodies. Mouse and rabbit IgG antibodies were used as controls (Santa Cruz Biotechnology). All antibodies were used at 1:100 dilutions. Antibody binding was detected by using the Dako Cytomation Envision System (Dako Cytomation, Carpinteria, CA) and Vector NovaRED substrate kit (Vector Laboratories Inc., Burlingame, CA), according to the manufacturer's protocol.

Tumors were finely chopped with a razor blade and then digested in DMEM containing 10% collagenase/hyaluronidase mixture (Stem Cell Technologies) for 1.5 h at 37°C. Antibodies used against mouse antigens were purchased from BD Pharmingen, unless otherwise specified, and included CD24-APC (BD-101813); CD-61-PE (BD-555735); CD29-FITC (Biolegend-12343); biotinylated CD31 (BD-553371), CD45 (Biolegend-103103), Ter119 (Biolegend-116203); and streptavidin-PE-Texas red (BD-551487). Antibody staining and cell sorting was performed as previously described [8]. For labeling of intracellular epitopes, cells were cultured on poly-L-lysine-coated coverslips, and were fixed in 4% paraformaldehyde and permeabilized in 0.2% Triton-X 100 for 4 min at room temperature before blocking in 2% BSA in PBS at room temperature for 30 min. Antibodies used were cytokeratin 14 (Covance, PRB155P) and cytokeratin 18 (Progen Biotechnik, GP CK18.2).

Mammary gland and tumor samples were flash frozen in liquid nitrogen and lysed in buffer containing 1% Triton X-100, 50 mM Hepes (pH = 7.5), 150 mM NaCl, 10% glycerol, and 1 mM and 2 mM EDTA supplemented with the appropriate protease and phosphatase inhibitor [27]. Protein concentrations were determined by using the BCA Protein Assay (Pierce, Rockford, IL, USA) according to the manufacturer's recommendations. Western blots were performed by using the following antibodies: monoclonal β-catenin (BD Transduction Laboratories 610153), monoclonal cyclin D1/D2 (Upstate Biotech 05-362, Lake Placid, NY), polyclonal Ruvb-like 1 (Protein Tech. Group 10210-2-AP, Chicago, IL), and monoclonal β-actin (Sigma A5441, St. Louis, MO) as loading control. All antibodies, unless otherwise indicated, were used at 1:1,000 dilution.

Gene-expression profiling was performed at the Centre for Translational and Applied Genetics (CTAG, Vancouver, BC, Canada) by using Affymetrix GeneChip Mouse Exon 1.0 ST Arrays according to the manufacturer's instructions. In brief, total RNA was extracted from the snap-frozen tumor tissue by using the GeneChip Whole Transcript (WT) Sense Target (ST) Labeling Assay (Affymetrix), as per manufacturer's recommendations, before being hybridized to a GeneChip Mouse Exon 1.0 ST Array (Affymetrix). RNA, 1 μg, was used as starting material, and the rRNA was reduced by using the RiboMinus Transcriptome Isolation Kit (Invitrogen), as per Affymetrix recommendations. Each chip was run through the GeneChip Fluidics Station 450 for washing and staining before scanning the chip on the GeneChip Scanner 3000 7G (Affymetrix). The corresponding microarray data were deposited into the Gene Expression Omnibus, with accession number GSE19300.

Quantitative real-time PCR (Q-RT-PCR) was conducted in 384-well plates on an Applied Biosystems (Foster City, CA, USA) Q-RT-PCR instrument by using Roche Universal Probe Library (UPL) (Roche Applied Science, Laval, Quebec, Canada) according to the manufacturer's instructions. In brief, 1 μg of total RNA was used in a 40-μl reaction to make cDNA. Subsequently, 10 μl of the Q-RT-PCR mixture containing 100 nM UPL probe, 200 nM of each primer, and TaqMan PCR master mix (Applied Biosystems) was loaded into each well. After preliminary 95°C incubation, the samples were read for 40 cycles (95°C, 30 sec; 60°C, 30 sec; 72°C, 60 sec). The values for mRNA expression were normalized by using β-actin or GAPDH or both as the housekeeping genes. All Q-RT-PCR primers were designed by using the Roche Applied Science online assay design center for UPL system, and were purchased from Invitrogen (Burlington, ON, Canada). All assays were set up for a relative quantitation method in which mean Ct values from MMTV-Wnt1 samples were used as calibrators for data analysis for MMTV-Wnt/ILK samples. The relative fold change (FC) of gene expression between MMTV-Wnt1 and MMTV-Wnt/ILK samples was calculated by using the standard 2^-ΔΔct method.

All the statistical analysis was performed as specified in this section, unless otherwise indicated in the figure legends. The percentage tumor-free data were plotted by using Kaplan-Meier curves, and differences in tumor free rates were assessed with the Cox model. Tumor-volume data were analyzed by using linear regression. Affymetrix GeneChip expression data were processed by using ArrayAssist software (ref ArrayAssist v1.x). Probe-set summarization and background correction of expression values was performed by using the RMA algorithm (ArrayAssist ExonRMA), and chip-to-chip variation was corrected by using quantile normalization. Quantitative RT-PCR data (cycle time or Ct data) were obtained by using RQ Manager Software (Applied Biosystems). Ct data were analyzed by using linear mixed effects, allowing for the biologic sample replicates and Q-RT-PCR plate-to-plate variation, as well as adjustment for to the reference loading control β-actin or GAPDH or both. Results were exponentiated to obtain natural scale fold-change estimates (that is, 2^-ΔΔCt) and confidence intervals. The null-effect reference line at 1.0 is shown in the fold-change plots, and statistically significant fold changes are indicated by whisker 95% confidence intervals that do not cross the reference line. Benjamini-Hochberg (BH) adjustment of P values to correct for multiple comparisons was also performed on Q-RT-PCR data, and only genes with P value of < 0.05 were considered significantly differentially expressed.

To begin investigating the cooperation between ILK and Wnt signaling pathways in mammary tumor formation we needed to determine how overexpression of ILK would affect Wnt1-induced mammary tumor development. Double-transgenic MMTV-Wnt/ILK mice were created by crossing the previously available MMTV-Wnt1 male and the MMTV-ILK female (see Materials and Methods) (Figure 1a). In addition to genotyping to confirm the presence of both transgenes (that is, MMTV-Wnt1 and MMTV-ILK) (Additional file 1, Figure S1), quantitative RT-PCR confirmed overexpression of human ILK in tumors of the double-transgenic mice (Figure 1b), whereas the level of Wnt1 overexpression remained fairly consistent in both transgenic models (data not shown). After genotyping, double-transgenic MMTV-Wnt/ILK female mice and their littermate controls (MMTV-Wnt1, MMTV-ILK, and FVB) were monitored and palpated weekly for the onset of tumors. The percentage of animals in each cohort remaining free of palpable mammary tumors was plotted at weekly intervals as a function of age for virgin females (Figure 2a). Mammary tumors appeared earlier in bi-transgenic animals overexpressing both Wnt and Ilk oncogenes. About half of the bi-transgenic mice examined (n = 23) developed palpable mammary tumors by the age of 19 weeks (T_1/2 = 19), whereas fewer than 30% of the Wnt1 transgenic and none of the ILK or FVB control mice had tumors. Furthermore, by age 50 weeks, all of the bi-transgenic mice exhibited tumors, at which time about 25% of the Wnt1 transgenic mice (n = 31) still were tumor free (Figure 2a). In addition, the tumors that arose in both Wnt1 and Wnt/ILK transgenic mice were lobuloalveolar adenocarcinomas, as was previously reported for the Wnt1 transgenic mice.

To determine the growth rate after tumor onset, the sizes of the mammary tumors were periodically measured. We observed a significant difference in the growth rate of the mammary tumors between MMTV-Wnt1 and MMTV-Wnt/ILK transgenic mice. In general, tumors arising in MMTV-Wnt/ILK mice grew at an accelerated rate, forming large tumors of 2,000 mm³ or more within 14 days, whereas tumors in MMTV-Wnt1 typically did not grow to this size in 50 days (Figure 2b). Therefore, based on the calliper measurement of individual tumors from each group, tumors from MMTV-Wnt/ILK mice showed a tumor-doubling time of 4 days, whereas the doubling time of the tumors in the MMTV-Wnt1 group was 3 times longer, averaging 12 days (Figure 2c).

To determine whether the observed accelerated tumor incidence starts with a higher proliferation during mammary gland development, we examined the pre-neoplastic mammary glands in all transgenic models. The mammary gland architecture was therefore analyzed with whole-mount preparation. Whereas the mammary glands taken from a wild-type FVB female mouse revealed a branching ductal structure typical of a virgin female, reflected by the presence of an ordered ductal network with side branching, the pre-neoplastic mammary glands from MMTV-Wnt/ILK transgenic mice exhibited increased ductal branching with more secondary and ternary side branches, as well as a denser alveolar development at an earlier age (8 weeks old) compared with either MMTV-Wnt1 mice or controls (10 to 12 weeks old) (Figure 3a). The areas of dense alveoli are indicative of mammary gland hyperplasia (black arrows in Figure 3a). In addition, the presence of hyperplastic mammary tissue was confirmed by hematoxylin and eosin (H&E) staining of transgenic mammary gland sections (Figure 3b, a-f). The mammary glands of MMTV-Wnt/ILK mice showed a marked enhancement in epithelial hyperplasia as well as multilayered and disorganized ductal epithelium, compared with the mammary glands of the MMTV-Wnt1 mice.

Immunohistochemical staining also was used to examine the pre-neoplastic mammary glands for both PCNA (proliferating cell nuclear antigen)- and Ki67-positive cells, two factors known to represent proliferation. Whereas PCNA is maximally detected in the S phase, Ki67 is preferentially expressed during later stages of the cell cycle [28], providing different staining patterns, but both indicative of the proliferation status of the tissue. As previously shown [9], we also confirmed a greater proliferation status in the mammary gland cells in MMTV-Wnt1 transgenic mice compared with non-transgenic FVB controls (Figure 3b, g-j). However, the mammary glands of MMTV-Wnt/ILK mice showed an even greater increase in both PCNA (Figure 3b, g-j) and Ki67 (Figure 3c) staining over those observed in the glands of MMTV-Wnt1 transgenic mice (Figure 3b, i-l, and 3c). Altogether, these results suggest a further enhancement in the proportion of the proliferating cells as the result of the mutual input from both ILK and Wnt signaling pathways in the double-transgenic mice. The pre-neoplastic mammary glands of the double-transgenic mice on average showed about a twofold increase in proliferation activity based on Ki67 staining compared with the glands of the MMTV-Wnt1 model (Figure 3d).

To investigate the potential role of some of the known downstream targets of both Wnt and ILK signaling pathways in the enhanced development of tumors in Wnt/ILK double-transgenic mice, we performed a series of Western blots on protein extracts isolated from 20 randomly chosen tumors. In addition, we included protein extracts from two normal mammary glands harvested from FVB mice as a control (Figure 4a). The Western-blot data were quantified by densitometry analysis (Figure 4b). All values were normalized to β-actin for the corresponding sample on each gel. We observed a clear difference in the expression of some of the downstream molecular effectors of the Wnt signaling pathway, such as cyclin D1, β-catenin, and suggestive evidence of a difference for RuvB-like 1. Western-blot analysis of β-catenin showed greater than twofold increase in β-catenin expression in Wnt/ILK double-transgenic tumors compared with Wnt1-induced mammary tumors (Figure 4a,b). More interesting, β-catenin immunohistochemical staining revealed a marked increase in nuclear accumulation of β-catenin in MMTV-Wnt/ILK tumors compared with a more-cytoplasmic distribution of β-catenin observed in Wnt1 tumors (Figure 4c). It is known that an increase in Wnt-mediated β-catenin is attributable to increased stability through the inhibition of the ubiquitin-mediated degradation [29,30]. In MMTV-Wnt/ILK tumors, further activation of ILK, which has also been shown to regulate β-catenin stabilization and nuclear localization (reviewed in [15]), further contributed to this increase. Quantification of the proportion of the cells staining positive for nuclear β-catenin to the total number of the cells in five randomly chosen fields of view shows a 3.5-fold increase in nuclear positive β-catenin staining in MMTV-Wnt/ILK tumors (Figure 4d). In addition, RuvB-like 1, a TATA box-binding protein essential for cell growth [31], which participates in β-catenin-mediated transactivation by binding to nuclear β-catenin [32], also showed about a 1.8-fold increase in MMTV-Wnt/ILK tumors. Interestingly, RuvB-like 1 was also recently found to interact with ILK in a proteomics analysis for the identification of the components of the ILK interactome [33].

Previous studies have shown an increased abundance of cyclin D1 in Wnt1 mammary tumors [34,35]. Although cyclin D1 levels were increased in Wnt1 tumors compared with normal mammary epithelium, we found an additional fourfold increase in the cyclin D1 level in Wnt/ILK double-transgenic mammary tumors. We also observed an increase in GSK-3 Ser-9 phosphorylation in MMTV-Wnt/ILK tumors. Both cyclin D1 and phosphorylation of GSK-3β are regulated by ILK as well as by Wnt1 [20]. Other signaling molecules, such as P-ERK and fibronectin, also showed a marginal increase in the MMTV-Wnt/ILK tumor samples (Figure 4b). However, this upregulation was not statistically significant with the sample size tested.

Some of the other downstream targets of Wnt and ILK pathways, such as c-myc, Lef-1, MMP-2, MMP-7 (data not shown), ^ser473P-AKT (Additional file 1, Figure S2), did not show a marked difference in expression in the mammary tumors between the two transgenic groups. Overall, these data suggest activation of collaborative oncogenic signaling pathways contributing to the more-aggressive growth phenotype of the MMTV-Wnt/ILK tumors.

To identify other possible signaling molecules and events involved in increased tumor incidence in MMTV-Wnt/ILK transgenic tumors, we performed global gene-expression analysis on RNA isolated from four independent tumors from each MMTV-Wnt1 and MMTV-Wnt/ILK transgenic model. Gene-expression levels were determined by using the Affymatrix GeneChip Mouse Exon 1.0 ST. The exploratory expression data obtained from the Affymetrix scanner were processed by using ArrayAssist software (ArrayAssist v1.x) controlling for false-negative and false-positive results. Probe-set summary and background correction of expression values were performed by using the RMA algorithm (ArrayAssist ExonRMA), and chip-to-chip variation was corrected by using quantile normalization. From the gene-chip data, we identified 200 known or predicted genes showing a significant fold change (data not shown). Volcano or scatterplot was used to highlight the most differentially expressed transcripts as the function of significance (determined based on P value) over differential expression (determined as fold change) (Additional file 1, Figure S3). Subsequently, all exon-level and gene-level probe sets showing an absolute fold change > 2.5 and an unadjusted P value < 0.05 were reviewed (Table S1). Categorization of these differentially expressed genes according to their molecular-function groups, as shown in Additional file 2, suggests a change in the program of gene expression, and particularly a transcriptionally active signaling network in Wnt/ILK mammary tumors (Additional file 2). Included among these are Sirt1, Hey2, Ncoa4, and Foxa1, all of which have been implicated in the progression of luminal breast cancer and ER-α-dependent signaling, as well as ER-α (that is, Esr 1) (Additional file 2). Ncoa4 (nuclear receptor co-activator 4) and FOXA1 (forkhead-box A1), both have been shown to act as transcriptional activators of estrogen (ER)-α [36,37]. Sirt1 transcription factor has been shown to regulate Hey2, a downstream effecter of Notch1 signaling [38,39], one of the more prominent activated pathways in the luminal-restricted mammary progenitor cells [40,41]. This functional classification also facilitated the selection of the gene candidates for further validation (Additional file 2). This microarray analysis with the limited number of chips used was an initial approach in obtaining an exploratory list of potential gene targets. We, therefore, followed up by selecting 26 of these genes for validation by quantitative reverse transcriptase PCR (Q-RT-PCR).

Transcript levels for each test gene were normalized to β-actin or GAPDH transcript levels or both in the same extract. Logarithmic results were exponentiated to obtain linear-scale fold-change estimates (that is, 2^-ΔΔCt) and confidence intervals. The results of the Q-RT-PCR assays demonstrated consistent differential expression for 17 of the 26 genes identified and tested from the Affymetrix analysis (Figure 5).

Expression profiling, by microarray DNA analysis, revealed a marked upregulation in FOXA1 in mammary carcinomas of the Wnt/ILK double-transgenic mice (Table S1). Upregulation of FOXA1 also was confirmed by Q-RT-PCR in a larger tumor sample set (n = 8) as well as at the protein level by immunohistochemical staining (Figure 6). Quantitative RT-PCR showed an average increase of fourfold in FOXA1 expression (Figure 5). The fact that FOXA1 controls the downstream transcription of ER-regulated genes [37,42,43] led us to examine further the expression of ER-α, which was also identified in the Affymetrix analysis. ER-α also showed evidence of fold change in our exon-level analysis on eight different exon probe sets, with an average fold change of 2.56 and a P value of 0.031.

At the protein level, although the tumors from both transgenic models were ER-α positive by immunohistochemistry, a higher level of staining was evident in sections from the double-transgenic tumors (Figure 6a). In addition, Q-RT-PCR revealed a 2.4-fold increase in ER-α expression at the mRNA level in MMTV-Wnt/ILK tumors (Figure 6b). Because the expression of ER-α is considered one of the hallmarks of luminal-type breast tumor [42], and FOXA1 is considered a marker of luminal cell differentiation during normal mammary gland development [44], we sought to examine the potential difference in the level of mammary gland luminal epithelial fraction in the two transgenic models.

Recently, CD24 was used as a single marker for isolation of pure populations of the luminal and basal compartments of the mouse mammary epithelium. Cells with a high expression of CD24 were identified to represent the mouse mammary luminal epithelial cells [41,45,46]. Other studies showed high expression of CD24 on luminal-restricted colony-forming cells [41,45,46]. In our model, initial immunohistochemistry revealed a greater presence of CD24-positive cells in both mammary tumors and pre-neoplastic mammary glands of MMTV-Wnt/ILK mice compared with the same tissue type in MMTV-Wnt mice (data not shown). This suggests a possible role for ILK in the background of MMTV-Wnt1 in expanding the luminal cell lineage with markedly enriched tumor-forming capacity.

However, because analysis of CD24 level by IHC is not an optimal way for such a comparison to be made, we further examined different epithelial subpopulations in tumors from both transgenic models with FACS analysis. We specifically looked at the expression of the cell-surface markers CD24 and CD61 within the Lin^-CD29^lo fraction (Figure 7a). Within the CD29^loCD24^hi fraction, also known as the luminal subpopulation [47], we determined the level of CD61⁺, which is the enriched fraction for the luminal progenitor cells [8,48] (Figure 7a). Single-cell suspensions from mammary glands of FVB mice were used as a control for staining with these cell-surface markers (Figure 7a). As suspected, single-cell suspensions prepared from MMTV-Wnt/ILK tumors (n = 3), depleted of hematopoietic and endothelial cells [8] (referred to as "Lin^-"), showed a more than two fold expansion in the percentage of CD61⁺ luminal progenitor cells within the total luminal fraction, CD29^loCD24^hi population (Figure 7a,b). The purity of the cell populations was evaluated by determining the expression of the luminal lineage marker keratin 18 (K18) and the myoepithelial marker keratin 14 (K14) (Figure 7c). As expected, K18 expression was highest in the CD29^loCD24^hi luminal fraction, whereas expression of K14 was restricted to CD29^hiCD24^med myoepithelial fraction in both transgenic models (Figure 7c). These data further suggest the expansion of the luminal progenitor lineage as the result of the mutual input of ILK and Wnt1 signaling.