N-cadherin/FGFR promotes metastasis through epithelial-to-mesenchymal transition and stem/progenitor cell-like properties

Abstract

N-cadherin and HER2/neu were found to be co-expressed in invasive breast carcinomas. To test the contribution of N-cadherin and HER2 in mammary tumor metastasis, we targeted N-cadherin expression in the mammary epithelium of the MMTV-Neu mouse. In the context of ErbB2/Neu, N-cadherin stimulated carcinoma cell invasion, proliferation and metastasis. N-cadherin caused fibroblast growth factor receptor (FGFR) upmodulation, resulting in epithelial-to-mesenchymal transition (EMT) and stem/progenitor like properties, involving Snail and Slug upregulation, mammosphere formation and aldehyde dehydrogenase activity. N-cadherin potentiation of the FGFR stimulated extracellular signal regulated kinase (ERK) and protein kinase B (AKT) phosphorylation resulting in differential effects on metastasis. Although ERK inhibition suppressed cyclin D1 expression, cell proliferation and stem/progenitor cell properties, it did not affect invasion or EMT. Conversely, AKT inhibition suppressed invasion through Akt 2 attenuation, and EMT through Snail inhibition, but had no effect on cyclin D1 expression, cell proliferation or mammosphere formation. These findings suggest N-cadherin/FGFR has a pivotal role in promoting metastasis through differential regulation of ERK and AKT, and underscore the potential for targeting the FGFR in advanced ErbB2-amplified breast tumors.

INTRODUCTION

Epithelial-to-mesenchymal transition (EMT) is a key process underlying tumor cell dissemination from the primary tumor to distant organs. During EMT, carcinoma cells upregulate EMT transcription factors and matrix metalloproteases to maintain a mesenchymal/invasive phenotype.^1,2 Oncogenes that promote neoplastic transformation are aided by pro-invasive molecules to lead the metastatic process. The HER2/neu gene, also known as ErbB2, is amplified in 25% of human breast cancers, and is associated with poor prognosis, but is not, however, the main driver of metastasis.^3,4 We have previously shown that N-cadherin or CDH2, a cell–cell adhesion molecule with pro-invasive activity,^5,6 is upregulated in 52% of moderately-to-poorly invasive duct carcinomas (IDCs).^7,8 Examination of these tumors suggested that N-cadherin is expressed in HER2-amplified tumors, implying that these molecules might converge in driving malignancy.

To study the potential association of HER2 and N-cadherin in clinical breast specimens, N-cadherin expression was introduced in the mammary epithelium of MMTV-Neu mice expressing an activated form of ErbB2/Neu (NeuNT), harboring a point mutation in the transmembrane domain.⁹ The MMTV-NeuNT mouse is commonly used to study signal transduction by ErbB2/Neu in mammary cancer.⁹ Mice expressing NeuNT and N-cad under the control of the MMTV-LTR promoter (Neu-N-cad) developed enhanced pulmonary metastasis relative to control NeuNT (Neu) mice, independently of primary tumor growth. Primary cell lines, derived from Neu-N-cad mammary tumors, exhibited dramatic increases in invasiveness, EMT, fibroblast growth factor receptor (FGFR) expression and EMT transcription factors. Moreover, Neu-N-cad tumor cell lines exhibited stem/progenitor cell-like properties, involving mammosphere formation and aldehyde dehydrogenase (ALDH) activity, which were inhibited by FGFR tyrosine kinase inhibition. Evaluation of the signal transduction pathways downstream of N-cad and FGFR showed differential activation of the AKT and ERK pathways, leading to metastasis. MEK1 inhibition by PD0325901 suppressed carcinoma cell proliferation, cyclin D1 expression, mammophere formation and Aldefluor activity, but had no effect on invasion or EMT. Conversely, AKT inhibition by MK2206, suppressed tumor cell invasion, through Akt 2 inhibition, and Snail expression, but had no effect on cell proliferation or mammosphere formation. These findings demonstrate that N-cadherin upregulation in ErbB2/Neu amplified tumors promotes metastasis by potentiating FGFR activity, leading to differential signaling by AKT and ERK, resulting in EMT, invasion, proliferation and stem/progenitor cell-like properties. These findings underscore the potential for targeting the FGFR in HER2-positive breast cancers.

RESULTS

N-cadherin and HER2 are co-expressed and co-localized in IDCs Although HER2 or N-cadherin (N-cad) was each shown to be expressed in invasive breast cancers,^3,7 little is known about their co-expression in human breast tumors. We examined 99 breast cancers, which were moderately or poorly differentiated IDCs. These included 32 HER2-positive and 67 HER2-negative breast cancers, for which N-cad expression was determined by immuno-histochemistry using a human N-cadherin-specific antibody.⁷ N-cad was highly expressed in 17 of the 32 HER2-positive cases (53%) and in 36 out the 67 HER2-negative (54%) cases, suggesting that N-cad and HER2 can be co-expressed in tumors, without, however, evidence of any significant association (P =0.96) between N-cad and HER2 status. Given these results, we can estimate a 95% confidence interval of 36–70% for the true population proportion of co-expression of HER2 positive and N-cad upregulation among these invasive breast cancers. The expression of N-cad in HER2-positive and HER2-negative tumors is illustrated in Figure 1a and Supplementary Figure 1. Co-localization of HER2 and N-cad was observed in a majority of the HER2/N-cad co-expressing tumors (Figure 1a and Supplementary Figure 1). Regardless of HER2 status, we also found that N-cad was expressed in 35% of estrogen receptor (ER)-positive and 55% of estrogen receptor (ER)-negative breast cancers; however, the difference between the two groups was not statistically significant (P =0.07 by χ²). These data suggest that N-cad and HER2 can coincide in invasive breast cancers and that N-cad might be expressed in both luminal and basal breast cancer types.

Figure 1.

N-cadherin is co-expressed and co-localized with HER2 in invasive breast cancers; N-cadherin expression in the MMTV-NeuNT mouse model promotes tumor metastasis. (a) Invasive duct carcinomas cancers (N =35) were co-immunostained for HER2 (fluorescin isothiocyanate (FITC)) and N-cad (tetramethyl rhodamine isothiocyanate (TRITC)). N-cad/HER2 co-expression in tumors, which were HER2 +/N-cad + (top panels), HER2 −/Ncad + (middle panels) or HER2 +/N-cad − (bottom panels), is shown in representative images. Merged channels (right panels) show co-localization of HER2 and N-cad in these tumors. (b) Primary tumor onset (top panel) was measured in 25 Neu and 25 Neu-N-cad mice and was determined to be of 25–27 weeks after birth. Tumor mass (bottom panel) was determined as the aggregate tumor mass in 25 mice at 60 days post onset. (c) Lung metastasis was tested in 10 sections of 15 mice and the number of foci per lung section are shown as mean±s.e.m. Statistical analysis comparing Neu and Neu-N-cad mice was done using the Mann–Whitney U-test; P<0.05. (d) Mammary tumor extracts from eight individual Neu and Neu-N-cad mice were immunoblotted for N-cad, ErbB2 and β-actin. (e) Sections from Neu-N-cad (1eft panels) and Neu (middle and right panels) mammary tumors were immunostained for N-cad (top) and E-cad (bottom). N-cad upregulation in Neu tumors (right top panel), which were also positive for E-cad (right, bottom panel), is shown. (f) Neu (top) and Neu-N-cad (bottom) lung metastases from 10 mice were co-immunostained with N-cad (FITC) and smooth muscle actin (TRITC). (g) Neu-N-cad metastases were stained for ErbB2 (top) and N-cad (bottom).

N-cadherin enhances mammary tumor metastasis in the MMTV-NeuNT mouse model

To test whether N-cad influences malignant progression in HER2-positive breast cancers, we crossed the MMTV-Neu mouse expressing an activated form of Neu (NeuNT) to the MMTV-N-cadherin mouse.^9,10 The MMTV-NeuNT mouse develops mammary tumors by 24 weeks of age (6 months) and low levels of pulmonary metastasis by 32 weeks of age (8 months), whereas the MMTV-N-cad mouse is non-tumorigenic.^6,11 Compared with MMTV-NeuNT (Neu) mice, MMTV-NeuNT-N-cad (Neu-N-cad) mice did not exhibit any change in tumor onset or mass, suggesting that N-cad may not affect tumor initiation or growth at the primary site (Figure 1b). Lung metastasis was, however, substantially increased in Neu-N-cad mice as compared with Neu mice. The occurrence of metastasis was higher in Neu-N-cad mice in which 80% of mice generated lung foci as compared with 10% of Neu mice. Metastatic growth was determined by both the number of metastatic foci (Figure 1c) and metastatic index, which integrates the number and size of foci¹² (Supplementary Figure 2A). Both measurements revealed a sixfold increase in metastasis in Neu-N-cad mice over Neu mice. This effect was not observed earlier than 60 days post onset (8 months of age), and was not evaluated at later time points because of large tumor burden.

We controlled for ErbB2/Neu and N-cad expression in mammary tumors from eight individual mice from each genotype, at 60 days post onset, by immunoblotting. As expected, ErbB2 was expressed in most tumors, whereas N-cad was only detected in Neu-N-cad tumors (Figure 1d). Immunostaining of tumors for both E-cad and N-cad revealed junctional localization of both proteins (Figure 1e). Interestingly, Neu tumors exhibited spontaneous N-cad expression in a small percentage (1.59±0.37) of cells in the tumor, mostly in areas near vessels (Figure 1e and Supplementary Figure 5B). Despite occasional N-cad expression in primary tumors, Neu metastases were negative for N-cad (Figure 1f, top panel), implying potential mesenchymal-to-epithelial transition (MET) at distant sites. Indeed, staining of pulmonary metastases for E-cad and pan-cytokeratin showed expression of both markers in Neu foci, a pattern that was also observed in Neu-N-cad mets (Supplementary Figure 3A). Consistent with N-cad effects on invasion, Neu mets formed preferentially intravascular emboli compared with a more extravasatory pattern of Neu-N-cad mets (Figure 1f). This was illustrated by co-staining of N-cad in carcinoma cells and smooth muscle actin in vessels (Figure 1f). Furthermore, ErbB2 and N-cad expression were maintained in Neu-N-cad mets, indicating sustained expression in disseminated cells (Figure 1g).

N-cadherin enhances tumor cell invasiveness in vitro

To investigate the mechanism of metastasis promotion by N-cad, we derived primary epithelial cell lines from Neu and Neu-N-cad mammary tumors (designated as Neu 1, Neu-N-cad 1, 2, 3). Given the challenge in generating NeuNT-based primary tumor cell lines, because of apoptosis in vitro, we included another Neu cell line (Neu 2), which was derived from the wild-type MMTV-Neu mouse, to confirm findings in NeuNT (Neu 1) cells. As in tumors in vivo, Neu-N-cad mammary tumor cell lines expressed N-cad and E-cad at cell–cell contacts (Figure 2a), whereas Neu cell lines expressed low levels of N-cad (Figure 2a). In contrast, E-cad was found in both cell types (Figure 2a). The relative metastatic potential of these cell lines was confirmed in vivo upon tail vein into syngenic FVB/N female mice. Neu-N-cad cell lines generated multiple foci (mean±s.e.m. =53±13) in the lungs, whereas Neu cell lines were unable to colonize the lungs (Figure 2b). Similar effects were obtained upon orthotopic injection; Neu-N-cad cells formed foci (mean±s.e.m. =17±3.5) in the lungs as compared with none by Neu tumors. Consistent with these data, Neu-N-cad cell lines displayed Matrigel invasion in vitro, which was increased by threefold, relative to Neu control cell lines (Figure 2c). In support of the latter, Neu-N-cad cell lines exhibited a mesenchymal morphology (Figure 2c), and upregulated MMP-2 and MMP1 mRNA (Figure 2d), relative to the epitheloid appearance to Neu cells (Figure 2c). By contrast, histological examination of the primary tumors did not reveal any differences in morphology between Neu and Neu-N-cad tumors (Figure 2f). However, staining of the tumors for pan-cytokeratin or vimentin, indicated diminished cytokeratin and augmented vimentin staining in random areas of Neu-N-cad tumors relative to Neu controls (Supplementary Figure 3B).

Figure 2.

N-cadherin enhances invasiveness and ERK activation in Neu-N-cad tumor cell lines and metastases. (a) Neu and Neu-N-cad mammary tumor cell lines were immunostained for E-cad (tetramethyl rhodamine isothiocyanate TRITC) or N-cad (fluorescein isothiocyanate FITC), both showing membrane localization; lysates from Neu 1, 2 and Neu-N-cad 1, 2, 3 cell lines were immunoblotted for N-cad, E-cad or β-actin. (b) Two Neu and Neu-N-cad cell lines were injected into the tail vein of FVB female mice (N =5), and 6 weeks later, lung sections were scored for metastatic foci. Neu-N-cad cells (bottom) colonized the lungs with numerous foci (53±15; P<0.05) compared with Neu cells, which did not generate foci (top). (c) Neu and Neu-N-cad cells were subjected to Matrigel invasion for 24 h and the migrated cells were photographed (top panels), and their morphology compared with cells grown in culture dishes (bottom panels). The number of invaded cells was scored and the average invasion of Neu and Neu-N-cad cell lines was determined in 5–10 experiments. Data are shown as mean ±s.e.m. Statistical analyses were performed using the Mann–Whitney U-test; P<0.05. (d) Quantitative real-time PCR analysis of MMP1, MMP2, MMP9 in Neu versus Neu-N-cad cell lines was performed on RNA extracted from each of the cell lines; MMP mRNA levels were standardized to GADPH mRNA levels. Data are shown as fold increase in mRNA levels in Neu-N-cad 1, 2. 3 cells relative to Neu 1 cells. Results are shown as mean±s.e.m.; P<0.05, t-test. (e) Neu 1, Neu 2 or Neu-N-cad 1, 2, 3 cell lines were serum-starved and stimulated with or without 50 ng/ml FGF-2 for 10 min. Cell lysates were immunoblotted with anti-phosphorylated or total ERK. (f) Sixty days post onset, mammary tumors from three Neu and Neu-N-cad mice were extracted and immunoblotted with an anti-p-ERK or ERK antibody. Sections from three Neu and Neu-N-cad tumors were immunostained using an anti-p-ERK antibody. Representative images are shown. (g) Lung tissues from nine Neu or Neu-N-cad mice (from 60 days post onset) were stained for p-ERK by immunohistochemistry using DAB detection. Two representative Neu (left panels) and Neu-N-cad (right panels) lungs showing p-ERK staining in metastases are displayed. Phospho-ERK immunoreactivity was determined in mets from both genotypes by Image J and shown as mean±s.e.m.; P<0.05, t-test. GAPDH, glyceraldehyde 3-phosphate dehydrogenase.

N-cadherin enhances fibroblast growth factor (FGF)-2-dependent ERK phosphorylation in ErbB2/Neu-driven tumor cells and metastases

We have previously shown that N-cad promotes invasiveness by enhancing FGF-2-stimulated ERK phosphorylation in the MCF-7 cell line and the MMTV-PyMT model.^5,6,13 These results were recapitulated in the NeuNT-N-cad model, showing dramatic increases in FGF-2-stimulated ERK phosphorylation (p-ERK), in Neu-N-cad cell lines, relative to low p-ERK levels in Neu cell lines (Figure 2e). To assess ERK activation in metastasis in vivo, we examined p-ERK levels in Neu and Neu-N-cad metastases from 60 days post onset (8 months), a time of peaking metastasis in Neu-N-cad mice. Immunoblotting of tumor extracts from 10 individual mice from each genotype, of which 3 individual tumors are shown, with anti-p-ERK antibody, revealed unchanged p-ERK levels between Neu and Neu-N-cad tumors (Figure 2f). These data were also confirmed by anti-p-ERK immunostaining of primary tumors (Figure 2f). In contrast, immunohistochemistry of lung tissues, also from 60 days post onset, indicated a 2.4-fold increase in p-ERK staining levels in Neu-N-cad relative to Neu mets (Figure 2g), suggesting ERK activity is enhanced in metastasis.

As ErbB2 is a potent activator of the PI3K/Akt pathway,⁴ we tested for changes in AKT phosphorylation; the latter was enhanced in Neu-N-cad cells, but was not as dependent on FGF-2 as ERK phosphorylation (Supplementary Figure 2B). By contrast, p38 or JNK phosphorylation and AKT, p38 or JNK levels were unchanged by FGF-2 (Supplementary Figure 2B).

N-cadherin upregulates FGFR expression and phosphorylation in tumor cells

FGF activation of ERK was shown to be potentiated by N-cad in the MCF-7 and PyMT mammary tumor models.^6,13 We therefore examined whether increases in p-ERK in Neu-N-cad cells were due to FGFR upmodulation. Immunoblotting of Neu and Neu-N-cad cell lysates using a pan-FGFR monoclonal antibody recognizing FGFR1-4,¹⁴ showed that Neu-N-cad cell lines expressed markedly increased levels of pan-FGFR relative to Neu cell lines (Figure 3a). The latter was matched by increased FGFR activity, as shown by FGFR tyrosine phosphorylation (Figure 3b).

Figure 3.

N-cadherin causes FGFR upregulation and phosphorylation leading to invasion. (a) Neu 2, Neu-N-cad 1, 2, cell lines as well as Neu 1 and Neu-N-cad 3 cell lines were serum starved and stimulated with or without 50 ng/ml FGF-2 for 10 min. Cell lysates were immunoblotted with anti-pan-FGFR or anti-α-tubulin. (b) Two Neu and Neu-cad cell lysates were immunoblotted with anti-phosphorylated FGFR or anti-α-tubulin. (c) RNA from Neu and Neu-N-cad cell lines was subjected to quantitative real-time PCR analysis using Taqman primers for FGFR1 and FGFR2 or for (d) FGFR2 isoform splice variants IIIb and IIIc relative to GADPH. The relative fold increase in mRNA levels in Neu-N-cad cell 1, 2, 3 cell lines relative to Neu cells are shown. Results are shown as mean±s.e.m.; P<0.05, t-test. (e) The effect of FGFR inhibition on Matrigel invasion of Neu and Neu-N-cad cells was determined upon treatment of cells with dimethyl sulfoxide (DMSO) or 0.5 μM of the FGFR inhibitor, PD173074. Data are shown as mean±s.e.m.; P<0.05, t-test. (f) Neu and Neu-N-cad cell lines were treated with DMSO or 0.5 μM PD173074 for 16 h, fixed and stained with an anti-ZO-1 antibody and counterstained with 4′,6-diamidino-2-phenylindole. GAPDH, glyceraldehyde 3-phosphate dehydrogenase.

As FGFR1 and FGFR2 were both found to co-immunoprecipitate with N-cad in cancer cells,⁵ we examined their relative expression by quantitative real-time PCR. FGFR1 mRNA was increased by 2.5- and 1.6-fold in Neu-N-cad relative to Neu cell lines (Figure 3c). Although FGFR2 mRNA levels were unchanged, there was a striking change in FGFR2 mRNA isoform expression. Namely, Neu-N-cad cells exhibited a 7- to 20-fold decrease in the epithelial FGFR2 IIIb mRNA isoform, and a 5-fold increase in the mesenchymal FGFR2 IIIc mRNA relative to Neu cell lines (Figure 3d). These data implied that N-cad promotes invasion through FGFR activation. Indeed, treatment of cells with 0.5 μM of the FGFR tyrosine kinase inhibitor, PD173074, inhibited Matrigel invasion of Neu-N-cad cells, and to a lesser extent, also of Neu cells (Figure 3e). The latter is likely due to N-cad upregulation, which activates the FGFR in Neu cells. This idea was further substantiated by that N-cad small interfering RNA (siRNA) knockdown in Neu cells desensitized the cells to FGFR inhibition. This was shown by the inability of PD173074 to inhibit Neu invasion in the absence of N-cad (Supplementary Figure 4A). Moreover, FGFR inhibition caused a change from a mesenchymal to an epitheloid morphology in Neu and Neu-N-cad cells, as shown by increased ZO-1 recruitment to cell–cell contacts (Figure 3f). We conclude that N-cad sensitizes tumor cells to FGFR inhibition, which reverses the invasive phenotype.

N-cadherin stimulates Snail and Slug upregulation in a FGFR-dependent manner

We thus tested whether the N-cad/FGFR pro-invasive effects were associated with EMT. We tested for changes in the expression of the EMT transcription factors Twist 1, Snail, Slug, Zeb1 and Zeb 2 in Neu-N-cad versus Neu cells, using quantitative real-time PCR analysis. Snail mRNA was increased by 33-, 20- and 11-fold in Neu-N-cad cell lines (Figure 4a). Similarly, Slug mRNA was increased by 11-, 5- and 4-fold, whereas Twist 1 mRNA was unchanged relative to Neu cell lines (Figure 4b). Moreover, Neu-N-cad cell lines exhibited a three- to fourfold in Zeb1, and a three- to ninefold increase in Zeb 2 mRNA, which also coincided with increases in vimentin mRNA (Figure 4c). Confirming the effects of N-cad on EMT, Neu-N-cad cell lines expressed higher levels of fibronectin and vimentin, which were matched by decreases in pan-cytokeratin levels when compared with Neu cell lines (Figure 4d).

Figure 4.

N-cadherin stimulates EMT transcription factors in a FGFR-dependent manner. (a–c) RNA from Neu1 and Neu-N-cad 1, 2, 3 cell lines was tested by quantitative real-time PCR for expression of (a) Snail, (b) Twist1 and Slug, (c) Zeb1, Zeb 2, and vimentin. RNA from each of the cell lines was analyzed relative to GADPH mRNA. Data are shown as fold increase in mRNA levels in Neu-N-cad cell lines relative to Neu 1 cells. Results are shown as mean±s.e.m.; P<0.05, t-test. (d) Neu1, Neu 2 and Neu-N-cad 1, 2 and 3 cell lines were immunblotted with an antibody to N-cad, fibronectin, vimentin, pan-cytokeratin or β-actin. (e) Neu-N-cad 1 and 3 cell lines were treated with dimethyl sulfoxide or 0.5 μM PD173074 for 16 h and stimulated with 50 ng/ml FGF-2 for 10 min. Cell lysates were immunoblotted with an antibody to p-FRS2α, p-ERK, Snail, Slug, Zeb-1 or β-actin. (f) Neu 1 and Neu 2 cell lines were treated as in e and immunoblotted for p-ERK, Slug or α-tubulin. GAPDH, glyceraldehyde 3-phosphate dehydrogenase.

To determine whether the FGFR regulates EMT, Neu-N-cad cells were treated with the FGFR tyrosine kinase inhibitor, PD173074,¹⁵ which inhibits FRS2 phosphorylation (Figure 4e), known to couple FGFR to ERK activation.¹⁶ Indeed, 0.5 μM PD173074 suppressed Snail and Slug expression in Neu-N-cad cell lines, but had no effect on Zeb1 (Figure 4e) or Zeb 2 levels (not shown). Slug expression, which was low in Neu cell lines, was also reduced by PD173074 (Figure 4f), whereas Snail expression was undetected in both Neu cell lines (not shown).

N-cadherin/FGFR activation of ERK is critical for cell proliferation but not invasion

We previously showed that N-cad expression in MCF-7 and PyMT mammary tumor cells promoted invasion through ERK phosphorylation leading to MMP-9 gene expression.^5,6 However, treatment of Neu or Neu-N-cad cells with 0.5 μM of the MEK1 inhibitor, PD0325901, which suppressed p-ERK (Figure 5b), did not affect Matrigel invasion as compared with FGFR inhibition by PD173074, which greatly reduced invasiveness of Neu-N-cad cells (Figure 5a). In addition, PD0325901 had no significant effect on Snail or Slug expression (Figure 5b). These data implied that ERK activation by FGFR controls the proliferation of Neu-N-cad cells, but not their invasion or EMT. Indeed, FGFR inhibition by PD173074, which inhibits ERK phosphorylation, suppressed cyclin D1 expression (Figure 5c) and 5-bromo-2′-deoxyuridine (BrdU) uptake in Neu-N-cad cell lines (Supplementary Figure 2C). These effects were reproduced by ERK inhibition by PD0325901, which also suppressed cyclin D1 expression (Figure 5d) and BrdU levels (Supplementary Figure 2C).

Figure 5.

N-cadherin stimulates cyclin D1 expression via FGFR-ERK and not FGFR-AKT activation. (a) Neu and Neu-N-cad cells were treated with dimethyl sulfoxide (DMSO), 0.5 μM PD173074 (FGFR inhibitor) or 0.5 μM PD0325901 (MEK1 inhibitor) for 16 h and were tested for Matrigel invasion; the relative invasion of drug-treated versus untreated cells in triplicate experiments cells is shown as mean±s.e.m.; differences between DMSO and drug-treated cells were significant; P<0.05, t-test. (b) Neu-N-cad 1 cells were treated with DMSO or PD0325901 and cell lysates were immunoblotted with an antibody to p-ERK, Snail, Slug or total ERK. (c, d) Cell lysates from Neu-N-cad 1 and 3 cell lines treated with DMSO, 0.5 μM PD173074 (c) or 0.5 μM PD0325901 (d) for 16 h were immunoblotted with an anti-p-ERK, cyclin D1 or tubulin antibody. (e) Neu-N-cad 1 and 3 cell lines were treated with DMSO or 0.5 μM PD173074, and were immunoblotted with an anti-p-Akt or Akt antibody. (f) Neu-N-cad 1 and 3 cell lines were treated with DMSO or 0.5 μM MK2206 (AKT inhibitor), and cell lysates were immunoblotted an anti-p-Akt, cylin D1 or α-tubulin antibody. (g) Neu1 and Neu-N-cad 1 cell lines were treated with DMSO or 0.5 μM MK2206 for 16 h and tested for Matrigel invasion. Results are shown as mean±s.e.m.; P<0.05, t-test. (h) Cell lysates from Neu-N-cad 1 in (f) were western blotted for p-AKT, Snail, Slug, p-GSK3 (tyr 216; activated GSK3) and α-tubulin.

In contrast, FGFR-AKT inhibition did not affect cell proliferation. Although PD173074 reduced AKT phosphorylation in Neu-N-cad cell lines (Figure 5e), direct inhibition of AKT by MK2206, did not significantly reduce cyclin D1 (Figure 5f) or BrdU levels (Supplementary Figure 2D) in both cell types. By contrast, MK2206 reduced Matrigel invasion of Neu-N-cad but not of Neu cells (Figure 5g), and suppressed Snail, but not Slug, expression (Figure 5h), likely due to GSK3 activation as a result of Akt inhibition, resulting in Snail ubiquitination and degradation.¹⁷ As the Akt family is comprised of three isoforms involving Akt 1, Akt 2 and Akt 3, which exert differential effects on cell migration,^18–20 we investigated which of these isoforms promote invasive migration of Neu-N-cad cells. We have previously shown that mammary N-cad expression in the PyMT mouse attenuates Akt 3 expression, and that Akt 3 overexpression in PyMT-N-cad cells suppresses invasion.²¹ Several reports have demonstrated anti-motile effects of Akt1 and pro-motile effects of Akt 2 in breast cancer cells.^22–25 We therefore examined the expression pattern of Akt isoforms in Neu versus Neu-N-cad cells by western blotting using Akt isoform-specific antibodies. The expression of Akt1 or Akt 2 was found to be similar in Neu and Neu-N-cad cells (Supplementary Figure 4B). Akt 3 expression, although variable, was not significant different between Neu-N-cad and Neu cell lines, except for a strong reduction in Neu-N-cad 1 cells (Supplementary Figure 4B). To further examine the contribution of the Akt isoforms to invasive migration, we knocked down each of the isoforms in Neu-N-cad cells using siRNA or small hairpin RNA (shRNA). We found that Akt1 siRNA knockdown did not significantly affect Neu-N-cad cell invasiveness relative to control cells (Supplementary Figure 4C). However, Akt 2 siRNA, which caused ~50% depletion of Akt 2 levels in both Neu-N-cad 1 and 3 cell lines, reduced Matrigel invasion also by ~50% (Supplementary Figure 4C). Moreover, knockdown of Akt 3 by two independent shRNAs,²¹ significantly increased the invasiveness of Neu-N-cad cell lines by 1.7-fold (Supplementary Figure 4E) as was the case for PyMT-N-cad cells.²¹ These data are consistent with the known pro-motile activity of Akt 2 and the counter-motile effect of Akt 3 observed in breast cancer cells.^18,21,23,25 We thus conclude that Akt 2 is likely the isoform-promoting invasiveness of Neu-N-cad cells, where it may overcome inhibition by Akt3.

N-cadherin induces tumorspheres and ALDH activity in an FGFR-ERK-dependent manner

As EMT is thought to generate tumor cells with stem-like properties,²⁶ we tested whether Neu-N-cad cells exhibited features of tumor-initiating cells such as mammosphere formation.^27,28 The ability of Neu and Neu-N-cad cells to form mammospheres was tested in media supplemented with growth factors including FGF-2. The latter was shown to be essential for the maintenance of cancer stem cells in the tumorspheres.²⁹ Interestingly, 70% of Neu-N-cad cells formed mammospheres as compared with 1.5% of Neu cells, which were in small clusters (Figure 6a).

Figure 6.

N-cadherin promotes mammosphere formation and increases ALDH1 expression, which are both suppressed by FGFR or MEK1 inhibitors. (a) Neu 1, 2 and Neu-N-cad 1, 3 single-cell suspensions were plated on a 60-mm ultra-low attachment tissue culture dish at 1 × 10⁵/ml in Dulbecco’s modified Eagle’s medium/F12 containing 10 ng/ml FGF-2, 4 μg/ml heparin, 20 ng/ml epidermal growth factor 5 μg/ml insulin and 5 μg/ml hydrocortisone for 5 days and photographed by bright-field microscopy. The number of mammospheres per microscopic field over five fields was determined. Data are shown as the mean number of mammospheres per 100 cells performed in triplicates; P<0.05, t-test. (b) Neu-N-cad 1 or Neu-N-cad 3 single-cell suspensions were plated under mammosphere growth conditions with dimethyl sulfoxide (DMSO), 0.5 μM of PD173074, PD0325901 or MK2206; mammospheres were photographed and counted. (c) Secondary mammosphere formation from primary mammosphere cultures that were treated with DMSO, PD173074 or PD0325901 was tested. First generation mammospheres were dissociated in trypsin and single cells were replated at 5 × 10³/ml for 7 days, in the absence of inhibitors. Results are mean of mammospheres±s.e.m. from triplicate experiments; P<0.05, t-test. (d) Neu 1, Neu-N-cad 1 and 3 cell lines were assessed for ALDH activity using the Aldefluor assay. Neu-N-cad 1 (bottom panels) and Neu-N-cad 3 (Supplementary Figure 5A) showed threefold increase in the Aldelfuor-positive cell population compared with Neu cells (top panels). Left panels: Negative controls include treating cells with DEAB, an irreversible inhibitor of ALDH. (e) Neu-N-cad 1 cells were treated with DMSO, 1 μM PD173074 or PD0325901 for 24 h and assessed for ALDH activity; left panels: negative control of cells treated with the ALDH inhibitor (DEAB, diethylamino benzaldehyde; SSC, side scatter).

To evaluate the role of FGFR-ERK activation in tumorsphere formation, Neu-N-cad cells were treated with 0.5–1.0 μM PD173074 (FGFR inhibitor) or PD0325901 (MEK1 inhibitor; Figure 6b). This resulted in mammosphere inhibition by either drug (Figure 6b). By contrast, AKT inhibition by MK2206 did not significantly reduce sphere formation (Figure 6b). Moreover, replating cells from mammosphere cultures, which were treated with either inhibitor, in the absence of further drug treatment, failed to generate secondary mammospheres (Figure 6c). These data suggested that the FGFR-ERK axis drives the proliferation of Neu-N-cad cells, and hence, their tumorsphere-forming capability. Concordant with this idea, PD173074 and PD0325901, which inhibited cyclin D1 expression (Figure 5d), also reduced BrdU uptake into Neu-N-cad cells (Supplementary Figure 2C), whereas AKT inhibition by MK2206 did not affect cell proliferation (Supplementary Figure 2D). These data implied that the FGFR-ERK axis controls the proliferation of Neu-N-cad cells.

Another marker of human or mouse mammary cancer stem/progenitor cells, which was shown to be more associated with luminal progenitor cells than with the mammary stem cells, is ALDH.^27,30–32 We thus tested whether N-cad caused changes in ALDH1 activity. Interestingly, Neu-N-cad cell lines exhibited a threefold increase in ALDH1 activity, relative to Neu cells, as measured by the Aldefluor assay (Figure 6d). Furthermore, FGFR or MEK1 inhibition by PD173074 or PD0325901 reduced the fraction of ALDH1-positive Neu-N-cad 1 and 3 cell lines from 30 to 15% and 17%, respectively (Figure 6e and Supplementary Figure 5A). These data indicate that the FGFR-ERK axis regulates the proliferation of Neu-N-cad cells, which behave like luminal progenitor cells. Consistent with these data, immunostaining of Neu and Neu-N-cad primary tumors for ALDH1 and N-cad revealed co-localization of these markers in Neu tumors, an effect that was exacerbated in Neu-N-cad tumors (Supplementary Figure 5B).

Finally, to test for differential tumorigenic potential, we assayed for relative ability of Neu and Neu-N-cad cells to form mammary tumors in host mice. Implantation of tumor cells into the mammary fat pads of syngeneic or athymic nude mice showed that Neu-N-cad cells formed large tumors by 6-weeks post inoculation (Supplementary Figure 5C), whereas Neu cells were unable to grow tumors. This might be due to a lower survival capability, and hence, apoptosis of Neu cells in vivo. Consistent with these data, others have shown that mammary implantation of ALDH-positive versus -negative tumor cells from the same tumor led to tumor formation by ALDH-positive, and not by ALDH-negative cells, demonstrating a link between ALDH activity and tumor-initiating potential.³³ The cumulative evidence suggests that the N-cad/FGFR axis enhances metastasis because of differential effects of AKT and ERK activation, leading to invasion, EMT, proliferation and stem/progenitor properties.

DISCUSSION

It is clear that HER2/neu has a critical role in breast cancer progression; however, the upregulation of other receptor tyrosine kinases, or constitutive downstream signaling, have limited the efficacy of HER2-targeted therapies.³⁴ Hence, the elucidation of pathways that drive metastasis of HER2-amplified tumors will undoubtedly yield novel insights into mechanisms of metastasis and/or HER2 drug resistance. We show that N-cadherin can be co-expressed with HER2 in a large number of moderately-to-poorly IDCs, suggesting that these molecules may act in concert to drive malignant progression. Interestingly, N-cad expression was evenly distributed among HER2-positive and HER2-negative breast cancers, implying no association between N-cad and HER2 status. However, these data suggest that among these HER2-positive breast cancers, N-cad expression is not uncommon. Regardless of HER2 status, we found that N-cad was expressed in 35% of ER-positive and 55% of ER-negative breast cancers, although the difference between the two groups was not statistically significant (P = 0.07 by χ²). These observations implied that N-cad might be expressed in luminal and basal breast cancers. Hence, expression of N-cad in the MMTV-NeuNT tumor model may reflect changes occurring in ER-positive tumors as this model expresses a luminal gene profile.³¹ We found that N-cadherin overexpression in MMTV-NeuNT tumors enhanced metastasis likely by promoting invasion, EMT and stem/progenitor cell-like properties. Interestingly, MMTV-NeuNT tumors displayed spontaneous upregulation of N-cadherin in a small subset of carcinoma cells, implying ErbB2 might promote EMT via N-cad upregulation.^17,35 Consistent with our findings in the MCF-7-N-cad and MMTV-PyMT-N-cad breast tumor models,^5,6,13 N-cad was found to potentiate FGFR signaling in MMTV-NeuNT-N-cad tumors. N-cad/FGFR induced striking EMT involving Snail and Slug upregulation, as well as decreased epithelial and increased mesenchymal gene expression. Moreover, N-cad expression was associated with downregulation of the epithelial FGFR2 IIIb splice variant, and upregulation of the mesenchymal FGFR2 IIIc isoform mRNA,^36,37 underscoring findings that splicing events underlie EMT.³⁸ Although EMT facilitates metastasis, the reverse process of MET is also necessary for metastatic progression.³⁹ The EMT–MET interconversion is likely to mediate the reciprocal switching between motile and stationary states or between invasion and proliferation during metastasis.^39,40 Our data indicate that Neu and Neu-N-cad metastases express a highly epithelial phenotype, as shown by robust E-cadherin and pan-cytokeratin expression.

Elucidation of N-cad/FGFR downstream signaling showed that FGF-2 stimulated robust ERK phosphorylation in Neu-N-cad cells compared with Neu control cells. We further showed that ERK phosphorylation controls the proliferation, but not the invasion, of Neu-N-cad cells. These findings are in contrast to those in MCF-7-N-cad and PyMT-N-cad cells, where ERK activation was important for invasion,^5,6 suggesting the oncogenic context, or the target cell of malignation, may account for these differences. We speculate that the FGFR-ERK axis drives the proliferation of mammary Neu-N-cad cells, which were derived from the primary tumors. Interestingly, Neu-N-cad cell lines displayed metastatic ability when injected into the mammary fat pads of syngeneic mice, suggesting a prior selection of these cells in culture, likely due to a pro-survival effect of N-cad.⁴¹ This is consistent with our in vivo data, showing that p-ERK staining was enhanced in Neu-N-cad metastases but not in primary tumors. In variance from the tumors, Neu-N-cad cell lines exhibited higher p-ERK and BrdU levels than Neu cell lines in vitro, suggesting that Neu-N-cad cell lines might have originated from a ‘metastatic pool’ rather than from the indolent bulk of the tumor.

In contrast to ERK, AKT phosphorylation was shown to control invasion and EMT, but had no effect on proliferation or tumor-sphere formation. The effects of AKT on EMT may be linked to AKT phosphorylation/inactivation of GSK3, which causes Snail stabilization, whereas the effect of AKT on invasion is consistent with the differential effects of the AKT isoforms on invasive migration. Although AKT1 was shown to inhibit cell migration, AKT2 was found to promote this process in breast cancer cells.^18–25 We have recently shown that N-cadherin expression in the MMTV-PyMT model promotes cell migration through AKT3 attenuation, and that AKT3 overexpression in these cells reduced cell migration, suggesting a similar effect to AKT1.^18,21 N-cad expression in the MMTV-NeuNT-N-cad model did not significantly affect the expression of any of the AKT isoforms. However, knockdown of AKT2 in Neu-N-cad cells inhibited invasion, whereas AKT3 shRNA knockdown enhanced invasion. These data suggested that AKT2 is likely the isoform contributing to Neu-N-cad cellular invasiveness. Moreover, these findings confirmed those in PyMT-N-cad cells, where AKT3 was found to suppress invasive migration.

Consistent with EMT and metastasis, Neu-N-cad cells formed extensive tumorspheres as compared with Neu cells. We speculate that this difference might be because of enhanced cell proliferation in Neu-N-cad cells, due to FGFR-ERK activation leading to cyclin D1 expression and BrdU uptake. Moreover, Neu-N-cad cell lines displayed hallmarks of stem/progenitor cells including EMT and Aldefluor activity.^26,27,30 ALDH1 is a well-recognized marker of human mammary stem/progenitor cells, which was also found to reside in mouse mammary luminal progenitor cells.^{30–33,42,43} It was, however, found to be more associated with luminal progenitor cells than with mammary stem cells. These data are concordant with that MMTV-Neu tumors express a lumimal gene profile.³¹

From a therapeutic point of view, our findings demonstrate that as HER2-amplified tumors progress to metastasis, they may upregulate N-cad, FGFR or both to enhance a stem/progenitor cell potential that contributes to metastasis. This may explain the resistance of HER2-positive tumors to Trastuzumab or Lapatinib, possibly due to upregulation of non-HER receptors including the FGFR.³⁴ Indeed, overexpression of FGFR was observed in the HER2-positive MDA-MB-453 cell line, which is resistant to trastuzumab, and where HER2 and FGFR were shown to cooperate in Cyclin D1 expression.^44,45 Moreover, a recent report showed that pan-HER inhibition synergizes with FGFR inhibition in suppressing malignancy in a xenograft model of HER2/FGFR-positive breast cancer cells.⁴⁶ Regardless of HER2, others have shown that triple negative breast cancer cell lines, which most express N-cadherin, are sensitive to FGFR inhibition by PD173074.⁴⁷ Thus, a therapy consisting of FGFR inhibitors, alone or with anti-HER2 agents, might overcome the metastasis of HER2-amplified tumors or triple negative breast cancers.

MATERIALS AND METHODS

Human archived breast tumors

Formalin-fixed and paraffin-embedded moderately or poorly differentiated, IDCs (N =99) were retrieved from the Department of Pathology at the Mount Sinai School of Medicine. Theses tumors represent a cross-section of all tumors from breast cancer patients who presented at Mount Sinai Hospital. These specimens were IRB exempt as these were anonymous cases that were not traceable to the patient’s information. The status of ER, PR (progesterone receptor), HER2 and grade of these IDCs are known. To determine co-expression and colocalization, 35 IDCs were co-stained with monoclonal anti-HER2 (clone CB11; Invitrogen, Carlsbad, CA, USA) and polyclonal anti N-cadherin (Santa Cruz, Santa Cruz, CA, USA), followed by detection with secondary Alexafluor-conjugated antibodies. Images were captured using a Zeiss Axioskop 2 mot plus upright microscope (Zeiss, Thornwood, NY, USA) fitted either with the Axiocam MRm (monochrome) or MRc (color) CCD cameras using AxioVision Software (Zeiss, Munich, Germany) using × 20 or × 40 magnifications.

Animals

Animals were housed and maintained by the Animal Studies Institute at Albert Einstein College of Medicine. Animal protocols used for this study were reviewed and approved by the AECOM Institute for Animal Studies. MMTV-NeuNT in the FVB/N background (NG.TK), expressing a point mutation in the transmembrane domain,⁹ were obtained from Charles River (Hopkinton, MA, USA). FVB/N mice expressing mouse N-cadherin in the mammary epithelium were as described.¹¹ Matings were performed with MMTV-Neu male hemizygous mice and MMTV-Ncad (+/−) or MMTV-NeuNT homozygous male mice were interbred with female MMTV-Ncad (+/+), (+/−) or (−/−) mice. MMTV-NeuNT-N-cad mice were identified by tail PCR using primer sets for N-cadherin: 5′-TGGAGAGACTTCTGAAAC AGC-3′ and 5′-CCATTCATCAGTTCCATAGGTTG-3′ and for Neu 5′-CGGAACCC ACATCAGGCC-3′ and 5′-TTTCCTGCAGCAGCCTACGC-3′. Transgenic animals were identified by PCR of a distinct product of 446 bp for N-cadherin and 550 bp for Neu.

Tumor onset and growth

Tumor onset was determined by mammary gland palpation in 20–30 Neu and Neu-N-cad mice. Tumor growth was measured at 30 and 60 days post onset. At each time point, mice were killed and mammary tumors were excised and weighed in 20–30 mice per group. Data were displayed as mean tumor onset time or mass±s.e.m. Statistical analyses were performed comparing individual time points using the Mann–Whitney U-test with significant differences P<0.05.

Lung metastasis

Neu and Neu-N-cad mice were killed at 30 and 60 days post tumor onset. Lungs were inflated by tracheal cannulation of 10% formalin and fixed for 24–48 h. Formalin-fixed lungs were paraffin embedded and sectioned on a tissue microtome (Leica Microsystems, Leica, Mars, PA, USA) at 5 μm. Lungs were serial sectioned through the tissue and sets of five serial sections, at 50-μm intervals, were captured and mounted on slides and the number of foci was determined on 10 non-duplicating sections per mouse. Data were displayed as mean number of metastases±s.e.m. Statistical analyses comparing Neu and Neu-N-cad mice used the Mann–Whitney U-test; P<0.05. In addition, metastasis index was evaluated by a stereological method using paraffin-embedded whole lungs. Lungs were systematically sectioned through the entire lung with one 5 μm section taken in every 0.5 μm of lung thickness. All sections were stained with hematoxylin and eosin and images of whole lungs were taken using a Zeiss SV11 microscope with a Retiga 1300 digital camera (Zeiss, Göttingen, Germany) and analyzed using Image J. The metastasis index represents the total volume of metastases normalized to total lung volume.

TaqMan real-time PCR

Total RNA was extracted from each cell line using RNeasy Mini Kit (Qiagen, Valencia, CA, USA). Real-time reverse transcription-PCR was carried out using total RNA samples in triplicates, TaqMan RNT-to-Ct 1-Step Kit (Applied Biosystems, Carlsbad, CA, USA), and gene-specific TaqMan probes (Applied Biosystems). Glyceraldehyde 3-phosphate dehydrogenase probe was used as endogenous control. The following are Gene Expression Assay IDs (Applied Biosystems) for each gene; glyceraldehyde 3-phosphate dehydrogenase (Mm99999915_g1), Snail (Mm00441533_g1), Slug (Mm00441531_m1), Twist1 (Mm00442036_m1), Twist2 (Mm00492147_m1), Zeb1 (Mm00495564_m1), Zeb2 (Mm00497193_m1), Vimentin (Mm00449208_m1), ERα (Mm00433149_m1), FGFR1 (Mm00438930_m1), FGFR2 (Mm03053754_s1), FGFR2 IIIb (Mm01275521_m1), FGFR2 IIIc (Mm01269938_m1), FGFR3 (Mm03053890_s1) and FGFR4 (Mm00433314_m1). Comparative CT (ΔΔCT) real-time analysis was performed using StepOne Software in StepOnePlus Real-Time PCR system (Applied Biosystems). Relative quantification of mRNA expression was depicted using GraphPad Prism software (La Jolla, CA, USA) and error bars indicate minimum and maximum values of 95% confidence level.

SiRNA, shRNA, transfection and transduction

Control siRNA or siRNA for mouse N-cadherin, Akt 1 and Akt2 were obtained from Santa-Cruz and transfected into cells per the manufacturer’s protocol. Two independent mouse Akt3 shRNAmirs in the pGIPZ lentiviral vector were purchased from Open Biosystems (Huntsville, AL, USA). A pGIPZ non-silencing shRNAmir vector along with Tat, Rev, Gag/Pol and VSV-G packaging vectors were included. Cells were transfected with appropriate retroviral vector using Lipofectamine LTX reagent (Invitrogen).

Antibodies

Anti-human HER2 (clone CB11) was from Invitrogen. Mouse monoclonal anti-N-cadherin or E-cadherin were from Invitrogen. Rabbit polyclonal antiE-cadherin or N-cadherin antibody was from Cell Signaling Inc., (Danvers, MA, USA) Antibodies to ErbB2, p-ErbB2, p-FGFR, p-FRS2, p-ERK and ERK, p-AKT, AKT, Akt1, Akt2, Akt3, p-p38 p38 and p-GSK3 (anti p-tyrosine 216) were from Cell Signaling Inc. Antibodies to Slug, Snail, Twist1, cyclin D1 were from Cell Signaling and antibodies to Zeb1 or Zeb 2, smooth muscle actin, actin or tubulin antibodies were from Sigma (St Louis, MO, USA). Anti-pan FGFR recognizing all four isoforms was obtained from Selective Genetics Inc. (San-Diego, CA, USA). Anti-vimentin, pan-cytokeratin, fibronectin were from Sigma. Anti-ALDH1 was obtained from Invitrogen. Anti-BrdU, Alexafluor-conjugated secondary antibodies or horseradish peroxidase-conjugated antibodies were from Invitrogen. Enhanced chemilumescent (ECL) reagents were from Pierce (Thermo-Scientific, Rockford, IL, USA).

Immunoblotting

Cells or tissues were extracted in solubilization buffer (50 mM Tris-HCl pH 7.5), 150 mM NaCl, 0.5 mM MgCl₂, 0.2 mM MEGTA, 1% Triton X-100) including protease and phosphatase inhibitors. Thirty micrograms protein were loaded on 7.5% SDS–polyacrylamide gels and transferred to Immobilon membranes. Blots were probed overnight at 4 °C with indicated antibodies and developed by ECL.

Signaling pathway inhibitors

The FGFR1/2 inhibitor PD173074 and the MEK1 inhibitor PD0325901 were obtained from Pfizer Inc., (New York, NY, USA). The AKT inhibitor MK2206 was from Tocris Inc., (Bristol, UK). Cells were plated at 1 × 10⁶ in/2 ml in growth media/10% fetal bovine serum (FBS) in 35 mm Petri dishes and allowed to adhere for 18 h, were then treated for 16 h in media containing 0.5% FBS with either dimethyl sulfoxide (control) with 0.5–1 μM PD173074, PD0325901, MK2206, diluted in dimethyl sulfoxide.

Immunofluorescence

Cells were plated on collagen-coated coverslips for 24 h, fixed in 3.7% paraformaldehyde for 15 min at room temperature, permeabilized for 2 min in 0.1% Triton X-100, washed in phosphate-buffered saline (PBS) and blocked in 2% bovine serum albumin (BSA)/PBS for 1 h, incubated with 2 μg/ml of primary antibody in PBS/0.5% BSA, washed in PBS, followed by 1 h incubation with 1:5000 Alexfluor-conjugated antibodies (Invitrogen), washed and counterstained with 4′,6-diamidino-2-phenylindole. Immuno-fluorescence images were captured using a Zeiss Axioskop 2 mot plus upright microscope fitted either with the Axiocam MRm (monochrome) or MRc (color) CCD cameras using AxioVision Software.

BrdU staining

Neu-N-cad cells were grown on coverslips in 6-well plates at a density of 1 × 10⁵/ml in Dulbecco’s modified Eagle’s medium/10% FBS. One hour before ending the experiment, cells were pulsed with 10 μM BrdU (Sigma), washed with PBS, fixed in 70% ethanol for 30 min and washed with PBS. DNA was denatured by 3 N HCl for 20 min, followed by PBS wash and incubation in 0.1 M sodium tetraborate for 5 min. Cells were then blocked in PBS/3% BSA and incubated with anti-BrdU-FITC antibody (Invitrogen) for 30 min followed by washes in PBS. Coverslips were mounted in Prolong gold antifade reagent with 4′,6-diamidino-2-phenylindole (Invitrogen). BrdU-positive cells were counted in 500 cells over five fields. Results are mean −/+ s.e.m. and statistical significance was at P<0.05 using the paired t-test.

Immunohistochemistry

Formalin-fixed/paraffin-embedded tissues were sectioned at 5 μm thickness, deparafinized in xylene and rehydrated in a series of decreasing ethanol/H₂0 solutions. Antigens were retrieved by steam heat for 20 min in a 0.01 M pH 6.0 trisodium citrate buffer, followed by 1 h blocking in PBS/5% serum/2%BSA. For chromogenic detection, endogenous peroxidase activity was blocked by incubating tissue with 2% H₂O₂ for 10 min. Tissues were washed in PBS and incubated with primary antibodies diluted in blocking buffer overnight at 4 °C, washed in PBS and incubated with horseradish peroxidase or fluorescently labeled antibodies for 1 h at room temperature. DAB (3,3′-Diaminobenzidine) chromogen was added according to manufacturer’s instruction (DAKO, carpinteria, CA, USA), and tissues were counterstained with Harris Hematoxylin (Fisher Scientific, Pittsburg, PA, USA).

Mammary tumor cultures

Primary mammary tumors were excised and washed in PBS. Tumor pieces were incubated at 37 °C for 3 h in Medium 199 with 2 mg/ml collagenase, 100 U/ml hyaluronidase, 100 U/ml penicillin/streptomycin, 2 mg/ml of BSA. Digested material was pelleted and cells were plated at 1 × 10⁷ cells per 10 cm dish with DME/10%FBS/10 μg/l insulin and 10 ng/ml epidermal growth factor. Epithelial cells were grown in culture till they reached crisis and resulting cell pools were used in the analysis.

Invasion assays

Invasion assays using a Boyden chamber were done as described.⁵ Twenty-four-well Transwell 8-μm filters were coated with 10 μg growth factor-reduced Matrigel (Becton Dickinson, Mississauga, ON, Canada). Fibroblast conditioned medium was used as a chemoattractant. Cell suspensions were plated at 1 × 10⁵ cells/0.5 ml DME, 0.1% BSA per well into the upper compartment of the Boyden chamber for 18 h. Cells on top the filters were removed, and cells penetrating the filters were stained with 0.5% crystal violet or 4′,6-diamidino-2-phenylindole to stain nuclei. Filters were counted and invaded cells were expressed as the average number of migrated cells bound per microscopic field over four fields per assay in triplicate experiments. Statistical significance was determined by one-tailed t-test; P<0.05.

Mammosphere formation

Single-cell suspensions of Neu or Neu-N-cad cells were plated on a 60-mm ultra-low attachment tissue culture dish (Corning, Lowell, MA, USA) at a density of 1 × 10⁵/ml in Dulbecco’s modified Eagle’s medium/F12 containing 10 ng/ml FGF-2, 4 μg/ml heparin, 20 ng/ml epidermal growth factor, 5 μg/ml insulin, 5 μg/ml hydrocortisone. Inhibition of mammospheres was tested with PD173074, PD0325901, MK2206 or dimethyl sulfoxide, which were added at 0.5 μM or 1 μM in the medium and replenished every day for 5 days. For secondary sphere formation, primary spheres were dissociated in 1:1 trypsin/Dulbecco’s modified Eagle’s medium solution at 37 °C, and mechanically by passing through a 23-G needle. Single cells were replated at 5 × 10³/ml, and incubated in 37 °C, 5% CO₂ for 7 days. At the end of the treatment, cells were transferred to a 35-mm MatTek dish and mammospheres as well as total cells (including mammospheres, single cells and clusters) per microscopic field over five fields were counted. Mammospheres were expressed as the average number of mammospheres per 100 cells