The following primary antibodies were used: c-Myc (9E10, to detect Myc-tagged protein), DVL2 and DVL3 (Santa Cruz Biotechnology, Inc., Santa Cruz, CA, USA); DVL1 (R&D Systems, Abingdon, UK); active β-catenin (anti-ABC; Upstate, Billerica, VA, USA); α-tubulin (DM1A) (Neomarkers, Fremont, CA, USA); cyclin D1 (SP4) (Cell MARQUE, Rocklin, CA, USA) for immunohistochemistry (IHC); cyclin D1 (Chemicon, Billerica, MA, USA) for western blotting; bromodeoxyuridine (BrdU) (Roche, Basel Switzerland); p21^Cip1 (OP64-100UG) (Oncogene Research Products, Cambridge, MA, USA); ERK and P-ERK (Thr202/Tyr204) (Cell Signaling Technology, Danvers, MA, USA); total β-catenin and CD31 (BD Pharmingen, Franklin Lakes, CA, USA); and active β₁-integrin (Clone HUTS-4, MAB2079Z; Chemicon).

As secondary antibodies we used anti-rabbit and anti-mouse antibodies (GE Healthcare, Little Chalfont, UK; LI-COR Bioscience, Lincoln, NE, USA), anti-rat antibody (GE Healthcare) or anti-goat antibody (DAKO A/S, Glostrup, Denmark) coupled to horseradish peroxidase (HRP) or IRDye 800CW.

For IHC we used Biotin-SP-conjugated affinipure donkey anti-rabbit, anti-mouse, anti-rat IgG (Jackson ImmunoResearch, West Grove, PA, USA) and goat anti-rat ALEXA 568 (Molecular Probes, Eugene, OR, USA). Recombinant Wnt3a was purchased from R&D Systems. Y27632 was purchased from Sigma-Aldrich (St. Louis, MO, USA). The cDNA encoding Myc/His-tagged human sFRP1 in pCDNA was provided by Jeffrey Rubin (NCI, Bethesda, MD, USA) and was recloned into the pBabePuro retroviral vector. Conditioned media (CM) from Wnt1-producing cells, from sFRP1-producing cells, and purified sFRP1 were prepared as previously described 7.

MDA-MB-231/sFRP1-P1 cells and control-P1 cells were seeded on 12-well plates and were transfected with a mixture of Super TOPFlash plasmid and pRL-CMV (Promega, Madison, WI, USA) to assay TCF promoter activity using Fugene6 (Roche) according to the manufacturer's instructions. Luciferase activities were measured 48 hours later using the Dual-Luciferase Reporter Assay System (Promega) and Mithras LB940 (Berthold Technologies, Bad Wildbad, Germany) according to the manufacturer's instructions. The fold activation was normalized against renilla luciferase. Nine wells were used per condition and the average and standard error are shown in the graph.

The human breast cancer cell line MDA-MB-231 (ATCC, Manassas, VA, USA) was cultivated in DMEM, 10% heat-inactivated FCS (Amimed, Allschwil, Switzerland) supplemented with penicillin and streptomycin (Sigma-Aldrich). All transfections were performed using FuGENE 6 Transfection Reagent (Roche) following the manufacturer's guidelines. MDA-MB-231 cells were stably transfected with pCDNA3.1(+) (Invitrogen, Carlsbad, CA, USA) encoding Myc/His-tagged human sFRP1 or empty pCDNA3.1(+) as control. After selection with 1 mg/ml G-418, three clones of MDA-MB-231/sFRP1 and three control clones were isolated. Equal cell numbers of these clones were pooled before some experiments (MDA-MB-231/sFRP1-P1 and MDA-MB-231/control-P1). A second pool of sFRP1-expressing MDA-MB-231 cells (MDA-MB-231/sFRP1-P2) and control cells (MDA-MB-231/control-P2), each representing > 100 clones, was generated by infecting the cells with pBabePuro encoding Myc/His-tagged human sFRP1 or empty pBabePuro followed by selection with 2 μg/ml Puromycin (Sigma-Aldrich).

Cell proliferation was measured either by counting cell numbers with a Vi-Cell XR cell viability analyzer (Beckman Coulter, Fullerton, CA, USA) on selected days after seeding 200,000 cells on six-well plates or using the YOPRO cell viability assay (Invitrogen) 3 days after seeding 1,000 cells on a 96-well plate, according to the manufacturer's instructions. Anoikis was measured by seeding cells in 1% FCS-containing medium on polyHema-coated plates to prevent adhesion. Cells were harvested 24 hours later, stained with propidium iodide. The cell cycle distribution was analyzed with a FACScalibur (Becton Dickinson, San Jose, CA, USA) using the Cellquest software. A representative cell cycle distribution of three MDA-MB-231/sFRP1 clones and three control clones is shown. Unless otherwise noted, P values were calculated using Student's t test.

Cells were lysed in 1% Nonidet P-40, 50 mM Tris pH 7.5, 120 mM NaCl, 5 mM ethylenediamine tetraacetic acid, 1 mM ethylene glycol tetra-acetic acid (EGTA), 2 mM sodium vanadate, 20 mM β-glycerophosphate, 10 μg/ml aprotinin, 10 μg/ml leupeptin, 0.5 mM phenylmethanesulphonylfluoride (PMSF), 50 mM NaF, 1 mM dithiothreitol for 5 minutes on ice before collecting lysates. Debris was removed by centrifugation at 4°C and the protein concentration was determined using the Bradford reagent (BioRad, Hercules, CA, USA).

For western blotting, protein loading buffer was added to 30 to 50 μg total protein and the samples were denatured for 10 minutes at 95°C prior to separation on SDS-polyacrylamide gels and blotting by semi-dry transfer for 90 minutes on PVDF membranes (Millipore, Billerica, MA, USA). Membranes were blocked using 10% horse serum in Tris-buffered saline- Tween buffer for 1 hour (0.2 M NaCl, 25 mM Tris, pH 7.5, 0.5 ml/l Tween-20), except for p21^Cip detection where PBS- Tween buffer was used instead of TBS-Tween buffer for blocking. Blots were incubated with primary antibodies at room temperature for 1 hour or at 4°C overnight, followed by 30-minute incubation with secondary antibodies. These antibodies were anti-rabbit-HRP, anti-mouse-HRP (1:5000) or anti-goat-HRP (1:5000) for detection of luminescence, which was carried out using ECL (GE Healthcare) according to the manufacturer's instructions and using X-OMAT LS films (Kodak, New York, NY, USA). Secondary antibodies IRDye 800CW goat anti-rabbit-IgG or anti-mouse-IgG (1:10,000; LI-COR Biosciences) were detected with the LI-COR Odyssey system according to the manufacturer's instructions (LI-COR Biosciences). Quantification of protein expression was carried out using Odyssey 2.1 (LI-COR Biosciences).

Cells were seeded on six-well plates and grown to confluency. Monolayers were scratched, and in the indicated experiments the media were changed to Wnt1 CM or control CM. When using recombinant Wnt3a and Y27632, 90 minutes before the scratch was made the media were changed to DMEM 10% FBS containing 100 μg/ml Wnt3a (R&D Systems) and/or 5 μM Y27632 (Sigma-Aldrich). Pictures of randomly-chosen nine wound edges per condition were taken at time 0 and at the indicated time points using Nikon DIAPHOT (Nikon, Tokyo, Japan). The recovered area was calculated using ImageQuant TL (GE Healthcare). In some experiments, purified sFRP1 was added to the CM 7.

MDA-MB-231/sFRP1-P1 and control-P1 cells (0.3 × 10⁶) were incubated for 30 minutes on ice with an antibody recognizing active β₁-integrin (final concentration 20 ng/μl) in HEPES/NaCl buffer. This was followed by incubation for 30 minutes on ice with fluorescein isothiocyanate-conjugated (FITC) donkey anti-mouse IgG secondary antibody (Jackson Laboratories) (diluted 1:250) in Flow PBS (1 × PBS, 2% horse serum, 0.1% sodium azide). Fluorescence was measured using a FACSCalibur machine (Becton Dickinson) and the percentage of gated cells stained with active β₁-integrin was calculated.

Female Balb/c nude mice 7 to 10 weeks old were obtained from Charles River Laboratories (L'Arbresle, France) and were maintained in accordance with the Swiss guidelines for animal safety. Mammary tumors were established in mice (5 to 8 mice per group) by injecting 0.5 to 1.0 × 10⁶ control or sFRP1-expressing MDA-MB-231 cell lines in 100 to 150 μl PBS into the fourth right-side mammary fat pad. The tumor size was measured two or three times per week using a gage, and the volume was calculated considering the tumor as an oval according to the formula:

Statistical analyses were performed with two-way repeated-measures analysis of variance (repeated measures ANOVA (RM ANOVA)).

To assay tumor cell proliferation, 100 μg/g body weight of BrdU (Cell Proliferation Kit II; Roche) was intraperitoneally injected into tumor-bearing mice that were sacrificed 2 hours later. Tumors were excised and washed with PBS before fixation in 4% paraformaldehyde (PFA) at 4°C for 24 hours, and BrdU detection was performed as previously described 13. To examine experimental metastasis, 1.0 × 10⁶ MDA-MB-231/sFRP1-P2 cells or control-P2 cells were injected into the tail vein of female Balb/c nude mice (5 to 6 mice per group). Fifty-three days after the injection, the mice were sacrificed, lungs were dissected and the total number of surface lung metastases was determined. For western analyses, excised tumors were snap frozen and pulverized in liquid nitrogen and lysed in SDS buffer (100 mM Tris- HCl pH 7.6, 2% SDS, 10 mM dithiothreitol, 2 mM sodium vanadate, 0.5 mM ethylenediamine tetraacetic acid) by incubation at 95°C for 10 minutes.

To detect functional vessels in tumors, 100 μl of 2 μg/μl solution of fluorescein-labeled Lycopersicon esculentum lectin (Vector Labs, Burlingame, CA, USA) was injected into tail veins of tumor-bearing mice 14, and the mice were sacrificed 5 minutes later. Tumors were excised, fixed in 4% paraformaldehyde in PBS for 48 hours at 4°C, followed by an overnight incubation in 30% sucrose in PBS at 4°C, and then embedded in tissue-Tec O.C.T. Compound 4583 (Sakura, Tokyo, Japan). Frozen sections (9 μm) were subjected to IHC analysis to detect tumor-associated vessels using rat anti-mouse CD31 (diluted 1:100; BD Pharmingen) and goat anti-rat ALEXA 568 (diluted 1:200; Molecular Probes). Staining was performed using Discovery XT (Ventana Medical Systems, Inc., Tucson, AZ, USA). Pictures were taken with a Z1 microscope (Carl Zeiss, Jena, Germany) and were analyzed with IMARIS (Bitplane, Zurich, Switzerland) to calculate the co-localized area. For detection of cyclin D1, frozen tumor sections (9 μm) were subjected to IHC using SP4 (diluted 1:100) and Biotin-SP-conjugated affinipure donkey anti-rabbit IgG (diluted 1:100). Staining was carried out using Discovery XT with sCC1 pretreatment. Pictures were taken with an Eclipse E600 (Nikon) and were analyzed with IMARIS (Bitplane) to calculate the signal intensity.

Cultured cells were collected when plates were 70 to 80% confluent and total RNA was extracted using the RNeasy Mini kit (Qiagen, Venlo, The Netherlands). To extract total RNA from tumors, dissected tumors were put into RNAlater (Qiagen) overnight at 4°C, followed by RNA extraction using TRIzol reagent (Invitrogen) and washing using the RNAeasy Mini kit according to the manufacturer's instructions. RNA from mammary tumors (six MDA-MB-231/sFRP1-P1 tumors and five control tumors) and cultured cells (three MDA-MB-231/sFRP1 clones and three control clones) were individually amplified and labeled using the Ambion MesageAMP III RNA Amplification Kit (Applied Biosystems, Austin, TX, USA). Biotinylated, fragmented cRNA was hybridized to Affymetrix U133 plus 2.0 human GeneChips™ (Affymetrix, Santa Clara, CA, USA).

Expression values were estimated using the GC-RMA implementation found in Genedata's Refiner 4.5 software (Genedata AG, Basel, Switzerland). Quantile normalization and median scaling were performed in order to standardize array signal distributions to facilitate the comparison between in vitro cultured cells and in vivo tumor samples. Probesets showing statistically different expression profiles (one-way analysis of variance with P < 0.01; Benjamini and Hochberg Q values determined to minimize the false discovery rate) and specific pairwise fold changes were clustered by rank correlation with R > 0.8 for the first criterion and R > 0.885 for the second criterion using the Profile Distance Search function of Genedata's Analyst 4.5 tool (Genedata AG, Basel, Switzerland). All of the microarray data are stored in Gene Expression Omnibus [GEO:GSE13806].

For the quantitative real-time PCR, each sample cDNA was made from 2.5 μg RNA using Ready-To-Go™ You-Prime First-Strand Beads (GE Healthcare). Quantitative real-time PCR was performed with ABI Prism 7000 (Applied Biosystems, Austin, TX, USA) using ABsolute SYBR Green ROX Mix (THERMO Scientific, Waltham, MA, USA) following the manufacturer's guidelines. The primer sequences used for quantitative real-time PCR are as follows: human c-Myc forward, 5'-CCTACCCTCTCAACGACAG-3'; human c-Myc reverse, 5'-CTTGTTCCTCCTCAGAGTCG-3'; human β-actin forward, 5'-TGTCCACCTTCCAGCAGATGT-3'; and human β-actin reverse, 5'-CGCAACTAAGTCATAGTCCGCC-3'.

Interference with autocrine WNT signaling via sFRP1 has been shown to block in vitro proliferation of human breast cancer cell lines 67. In the following experiments we examined the effects of blocking WNT signaling using the basal-like breast cancer model MDA-MB-231 15. Vectors encoding Myc-tagged sFRP1 and the empty control were transfected into MDA-MB-231 cells, which express no sFRP1 mRNA (data not shown), and stable clones were selected in G418-containing medium. Three MDA-MB-231/sFRP1 clones expressing moderate to strong levels of the Myc-tagged sFRP1, as well as control clones, were selected for further analyses (Figure 1a).

WNT pathway activity was examined in the cells using various markers. As a consequence of WNT binding to FZD, cytoplasmic scaffolding proteins of the Dishevelled family (DVL1, DVL2 and DVL3) become phosphorylated on serine and threonine residues. DVL phosphorylation, which is the most proximal signaling event downstream of FZD activation, can be monitored by a decrease in the electrophoretic mobility of p-DVL 1617. DVL1, DVL2 and DVL3, total β-catenin and stabilized active β-catenin were examined by western analysis in the individual clones. In the sFRP1-expressing cells, there was a decrease in the level of the three p-DVLs, compared with the vector control cells. Furthermore, the level of total β-catenin and active β-catenin was reduced in the two clones expressing the highest level of sFRP1 (Figure 1a).

Next we examined transcriptional activity of the canonical WNT signaling pathway following transient expression of the TOPFlash TCF reporter plasmid in the cells. TOPFlash luciferase reporter activity showed a significant 2.7-fold decrease in the sFRP1-expressing cells compared with controls (Figure 1b), confirming that ectopic expression of sFRP1 reduces canonical pathway activity in MDA-MB-231 cells. The phosphorylation status of ERK, another signaling protein that is active in MDA-MB-231 cells, was not altered in the sFRP1-expressing cells (Figure 1a), suggesting that the effects of sFRP1 are specific for the WNT pathway.

For further studies, the three sFRP1-expressing clones were pooled (P1) and a second pool of sFRP1-expressing MDA-MB-231 cells consisting of > 100 clones was generated (P2). Corresponding control pools, control-P1 and control-P2, were also generated. Quantification of a western analysis shows that MDA-MB-231/sFRP1-P1 has 2.8-fold higher levels of sFRP1 than does MDA-MB-231/sFRP1-P2 (Figure 1c).

Next we examined effects of sFRP1 on tumor cell proliferation. MDA-MB-231 parental cells were inhibited by treatment with sFRP1 CM (see Additional data file 1). Furthermore, ectopic expression of sFRP1 in MDA-MB-231 cells also caused a decrease in proliferation (Figure 1d, e). One day after plating there was no significant difference between the control cells and the sFRP1-expressing cells. After 3 days, however, we observed a significant difference in the sFRP1-P1 cells and sFRP1-P2 cells compared with control cultures. This effect appears to be dependent on sFRP1 expression levels since, in comparison with controls, there is a 31% and a 16% reduction in proliferation of MDA-MB-231/sFRP1-P1 (Figure 1d) and MDA-MB-231/sFRP1-P2 (Figure 1e) cells at day 3, respectively.

To test the in vivo effects of sFRP1 expression, control and sFRP1-expressing MDA-MB-231 cells were injected into the mammary fat pads of female nude mice and tumor growth was monitored. There was a significant reduction in tumor outgrowth in mice injected with MDA-MB-231/sFRP1-P1 cells compared with control cells (P < 0.01, two-way RM ANOVA) (Figure 2a). Furthermore, the time to detection of the first tumors was much shorter after injection of MDA-MB-231/control-P1 cells, compared with MDA-MB-231/sFRP1-P1 cells (23 days vs. 35 days, respectively) (Figure 2b). Moreover, three mice injected with the MDA-MB-231/sFRP1-P1 cells remained tumor-free at day 45, when the experiment was terminated. In contrast, all of the mice injected with MDA-MB-231/control-P1 cells had tumors (Figure 2b). Western analysis carried out on tumor lysates revealed that sFRP1 was present in tumors arising from MDA-MB-231/sFRP1-P1 cells and WNT signaling was downregulated, since there was a decrease in the amount of p-DVL3 in comparison with control tumors (Figure 2c).

MDA-MB-231/sFRP1-P1 cells were tested in two additional experiments that yielded similar results (see Additional data file 2). While there was some variation in the time of tumor onset in the individual experiments, the time to appearance of the first tumor was consistently longer following injection of the MDA-MB-231/sFRP1 cells, in comparison with control cells (Figure 2b; see Additional data file 2). The data pertaining to the number of tumor-free mice were pooled for the three experiments, showing that 68% of the mice injected with MDA-MB-231/sFRP1-P1 cells were tumor-free at day 39, while only 20% of the control animals were free of tumors (Figure 2d). The data on tumor outgrowth kinetics were pooled for two experiments (Figure 2e), yielding a highly significant difference in the outgrowth ability of sFRP1-expressing cells (P < 0.0001, two-way RM ANOVA).

MDA-MB-231/sFRP1-P2 cells, which express 2.8-fold less sFRP1 than do the MDA-MB-231/sFRP1-P1 cells (Figure 1c), also grew more slowly than control cells following injection in the mammary gland (Figure 2f). Although the effect on tumor growth did not reach significance using two-way RM ANOVA and all animals had tumors at the end of the experiment (Figure 2g), tumor onset was delayed significantly in the cohort injected with MDA-MB-231/sFRP1-P2 cells (P = 0.026, log-rank test) (Figure 2g). The in vivo results together with the data on in vitro proliferation (Figure 1d, e) suggest that higher levels of sFRP1 cause a stronger blockade of WNT pathway activity, leading to a more pronounced effect on the time to tumor onset and tumor outgrowth.

Acquisition of migratory ability by cancer cells is an important characteristic contributing to metastatic tumor cell spread 18. Accordingly, we also examined the effect of WNT signaling on the migratory ability of MDA-MB-231 cells, using a wound healing assay. Confluent monolayers of MDA-MB-231/sFRP1-P1 cultures and control-P1 cultures were scratched and the medium was changed to Wnt1 CM or control CM. In response to Wnt1 CM, the control MDA-MB-231 cells migrated significantly more rapidly into the wounded area compared with cultures treated with control CM (Figure 3a, gray bars). In contrast, Wnt1 treatment of MDA-MB-231/sFRP1-P1 cells did not significantly stimulate migration (Figure 3a, black bars), reflecting the ability of sFRP1 to block Wnt1-mediated FZD activation.

We also performed a migration assay using the parental MDA-MB-231 cells treated with purified recombinant Wnt3a to confirm the results obtained with Wnt1 CM (recombinant Wnt1 is not commercially available). Wnt3a stimulated the canonical pathway as shown by the increased level of active β-catenin (Figure 3b), and increased the migratory ability of the cells in a wound closure assay (Figure 3c). In summary, in MDA-MB-231 cells, activation of the WNT pathway has a positive effect on cell migration; while sFRP1 lowers the proliferative and migratory ability of the MDA-MB-231 cells.

Next we tested the effect of WNT pathway blockade on the in vivo metastatic ability of MDA-MB-231 cells. Populations of MDA-MB-231/sFRP1 cells and control cells were injected into nude mice through the tail vein 19; 53 days later the mice were sacrificed and the lungs were analyzed for metastatic foci. There was a dramatic difference in the number of metastatic foci arising from sFRP1-expressing cells compared with control cells. A typical lung from an animal injected with control MDA-MB-231 cells and with sFRP1-expressing cells is shown (Figure 4a). A quantitative analysis of the lungs revealed that there is a significant decrease in the number of metastases arising in mice injected with sFRP1-expressing cells in comparison with control MDA-MB-231 cells (Figure 4b). In conclusion, sFRP1-mediated blockade of WNT signaling impairs the in vitro migratory ability and the in vivo metastatic ability of MDA-MB-231 tumor cells.

Our next goal was to uncover the mechanism underlying the ability of sFRP1 to decrease the mammary tumor-forming potential of MDA-MB-231 cells. This could be tumor cell intrinsic, resulting from downregulation of WNT signaling, and/or extrinsic via secreted sFRP1 effects on tumor-associated cells; both possibilities were tested. In vivo tumor cell proliferation was evaluated by examining BrdU incorporation in control tumors and sFRP1-expressing tumors. Incorporated BrdU was detected with a specific antiserum (Figure 5a) and staining was quantified. There was a 70% reduction in BrdU staining in tumors arising from MDA-MB-231/sFRP1-P1 cells compared with control tumors (Figure 5b). Apoptosis, as measured by western blotting for cleaved caspase-3, was low in the MDA-MB-231 control tumor lysates and was not increased in the MDA-MB-231/sFRP1-P1 tumor lysates (data not shown). Moreover, expression of sFRP1 in MDA-MB-231 cells did not render them sensitive to anoikis-induced cell death (Figure 5c), which rules out the possibility that slower tumor outgrowth reflects lower numbers of viable cells at the time of inoculation. In summary, the results suggest that sFRP1 downregulation of WNT signaling has a strong effect on tumor cell proliferation, but not on survival.

sFRP1 has been reported to block in vivo neovascularization 20. We therefore considered the possibility that the density or the functionality of the tumor-associated vessels might be impaired in sFRP1-expressing tumors. Vasculature was visualized by tail vein injection of tumor-bearing mice with FITC-labeled L. esculentum lectin 14 5 minutes before the animals were sacrificed, which allows only functional vessels to be perfused. Tumor sections were prepared and the associated endothelial cells were stained for CD31, while functional vessels were visualized via the FITC signal. There was no significant difference in the total vessel area or the ratio of FITC-positive/CD31-positive vessels in sFRP1-expressing tumors compared with control tumors (Figure 5d, e), suggesting that sFRP1 does not influence the number or the functionality of tumor-associated blood vessels. In summary, these results suggest that sFRP1-mediated blockade of WNT pathway activity in tumor cells is an important factor contributing to the slower outgrowth of the MDA-MB-231/sFRP1 tumors in mammary glands.

To identify target genes that are controlled by sFRP1 expression and might influence proliferation of the MDA-MB-231 cells, we undertook a genome-wide transcriptome analysis using microarrays. RNA isolated from individual tumors arising after injection of MDA-MB-231/sFRP1-P1 cells and control-P1 cells, as well as RNA from in vitro cultured MDA-MB-231/sFRP1-P1 cells and control-P1 cells, was analyzed.

Considering data generated from six sFRP1-expressing tumors and five control tumors, there were 1,753 probesets (1,246 genes) whose signals changed more than 1.5-fold (P < 0.01, Student's t test) in the tumors arising from MDA-MB-231/sFRP1-P1 cells compared with tumors arising from control-P1 cells. The same analysis performed on in vitro cultured samples revealed 428 probesets (332 genes) that had a 1.5-fold difference (P < 0.01, Student's t test). Only 69 probesets (54 genes) overlapped between the two analyses, clearly showing the substantial difference in gene expression between in vivo tumors and in vitro cultured cells.

All of the microarray data were examined for characterized WNT pathway target genes. We identified 16 genes where at least one of the probesets showed a tendency for suppression in MDA-MB-231/sFRP1-expressing cells and tumors (see Additional data files 3 and 4).

In vitro proliferation of MDA-MB-231/sFRP1-P1 cells was decreased by 30% compared with control cultures (Figure 1), while the in vivo effects of WNT pathway blockade appeared to be much stronger. For example, in comparison with control MDA-MB-231 cells, outgrowth of MDA-MB-231/sFRP1-P1-generated tumors was significantly slower; and tumor-free mice remained in this cohort in each experiment (Figure 2). We therefore decided to identify transcripts that were only altered in vivo in MDA-MB-231/sFRP1-expressing tumors. We screened the previously mentioned 1,753 probesets that were originally identified by a comparison of the MDA-MB-231/sFRP1 tumors with control tumors (see above) using the Profile Distance Search function of Genedata's Analyst 4.5 tool. In this analysis we looked for genes whose expression was significantly altered only in tumors arising from MDA-MB-231/sFRP1-P1 cells compared with tumors arising from control-P1 cells, with cultured MDA-MB-231/sFRP1-P1 cells or with cultured control-P1 cells. This resulted in 135 probesets (106 genes) that were downregulated (Figure 6a) and 84 probesets (62 genes) that were upregulated (Figure 6b) only in the sFRP1-positive tumors (see Additional data files 5 and 6 for lists of genes and their fold change).

The microarray analysis showed that the signal from a probeset for CCND1 (encoding cyclin D1) was downregulated and the probeset for CDKN1A (gene encoding p21^Cip1) was upregulated in vivo, in tumors resulting from MDA-MB-231/sFRP1-P1 cell injection (Figure 6c). The CCND1 promoter has a consensus lymphoid enhancer binding factor-1 binding site 21, and in some cancer models its expression is controlled by β-catenin/TCF activation 22. Our results suggest that CCND1 might also be a direct β-catenin/TCF target in MDA-MB-231 cells.

Both of these genes were analyzed further based on their known roles in cell cycle regulation and proliferation. Cyclin D1 was examined by IHC in tumor sections using a specific antiserum. Quantification of the staining showed a 30% decrease in cyclin D1 in sFRP1-expressing tumors compared with control tumors (Figure 7a). Western analysis for cyclin D1, carried out on lysates prepared from MDA-MB-231/sFRP1-P1 cultures and control cultures, revealed no significant difference in expression (Figure 7b). Western analysis revealed that p21^Cip1 was present in tumors resulting from injection of MDA-MB-231/sFRP1-P1 cells, while none was detectable in control tumors (Figure 7c). Considering the in vitro cultured cells, neither MDA-MB-231/sFRP1-P1 cells nor control-P1 cells had detectable levels of p21^Cip1 protein (Figure 7d). As predicted from the array data (Figure 6c), therefore, the decrease in cyclin D1 and the increase in p21^Cip1 are observed in vivo in the sFRP1-expressing tumors.

Considering that c-Myc is a WNT pathway target 2324 that regulates cyclin D1 22 and p21^Cip1 expression 25, we also examined c-Myc. There were no significant changes in c-Myc RNA levels in sFRP1-expressing tumors; however, c-Myc protein levels were lower in all of the sFRP1-expressing tumors compared with the control tumors (see Additional data file 7). In summary, the in silico analysis carried out on RNA of tumors arising from the MDA-MB-231/sFRP1 cells revealed altered levels of target genes (CCND1 and CDKN1A) that probably contribute to the anti-proliferative effects of sFRP1 expression. Moreover, our results also show the strong influence that the in vivo tumor environment has, not only on gene expression, but also on c-Myc protein. Decreased c-Myc levels might also contribute to the in vivo activity of sFRP1.