ERBB2-Mediated Transcriptional Up-regulation of the α5β1 Integrin Fibronectin Receptor Promotes Tumor Cell Survival Under Adverse Conditions

Introduction

The ERBB2 proto-oncogene is a member of the EGFR gene family consisting of ERBB1 (EGFR or HER1), ERBB2 (HER2 or NEU), ERBB3 (HER3), and ERBB4 (HER4). The genes encode type I receptor tyrosine kinases (RTK, except for ERBB3, which codes for a receptor harboring an impaired kinase-domain; ref. 1) that are involved in the transmission of proliferative as well as differentiation signals ( 2, 3). ERBB2 plays an important role in the development of human epithelial neoplasias and high levels of ERBB2 expression correlate with poor prognosis in breast and ovarian carcinoma ( 4, 5). Overexpression of the receptor was not only described in human breast and ovarian carcinomas ( 6), but also in other types of cancer, e.g., bladder carcinoma ( 7), non–small cell lung cancer ( 8), in tumors of the gastrointestinal tract ( 9), as well as in other human malignancies ( 10). At least for breast carcinogenesis, it is believed that oncogenic activation of ERBB2 is an early event because it could be found in preinvasive lesions as well as in advanced tumors ( 11, 12). Aberrant ERBB2 activation is implicated in the tumorigenic process through its influence on proliferation, survival of tumor cells, angiogenesis, invasive growth, metastatic behavior, and therapy resistance ( 12– 15). However, as for some other proto-oncogenes, like RAS and RAF, inherent cellular anticarcinogenic mechanisms have been identified preventing aberrant growth of affected cells and leading to a proliferative halt on oncogenic signaling (ref. 16; recent reviews: refs. 17, 18).

Exploiting the tetracycline expression system, we have previously shown that inducible hypermitogenic ERBB2 signaling results in growth arrest of MCF-7 breast carcinoma cells, and that the cyclin-dependent kinase inhibitor, P21, is an important mediator of this cellular response ( 16). Expression screening focused on the regulation of integrin extracellular matrix (ECM) receptors by oncogenic ERBB2 signaling revealed a strong up-regulation of several α-integrin subunits, including α5 integrin, in response to ERBB2 induction. Integrin ligation influences signal transduction cascades that promote cell proliferation, cell migration, cell survival, as well as angiogenesis. Unlike growth factor receptors, they lack intrinsic enzymatic activity but participate in signaling by coclustering with kinases and adaptor proteins. The α5β1 integrin heterodimer constitutes the major cellular receptor for the ECM component fibronectin. Although its effect in tumorigenesis seems to be cell type–specific, there are several studies providing evidence for a pro-tumorigenic role of α5β1 integrin engagement (ref. 19 and references therein). It has recently been shown that adhesion to fibronectin mediated by the heterodimer is responsible for improved survival of different cell lines, including breast carcinoma cells, under adverse conditions ( 20– 23). Furthermore, α5β1 integrin fibronectin receptor ligation is the key mediator of cell survival in an in vitro bone marrow metastasis model in which breast carcinoma cells undergo a transition to a nonproliferative, dormant state in response to fibroblast growth factor 2 (FGF2) treatment ( 24). For patients with breast cancer, the presence of single dormant cells in their bone marrow is indicative of fatal relapse including metastatic disease in bone, and ERBB2 has been implicated in the survival of such disseminated cells ( 25, 26).

In this work, we explore the molecular characteristics and consequences of a hypermitogenic state induced by stringently controlled overexpression of ERBB2 in MCF-7 breast carcinoma cells. In expression analyses, we observe a strong induction of ITGA5 transcripts and protein in response to oncogenic ERBB2 signaling. We characterize this regulatory interaction and investigate the functional effect on tumor cell physiology. Our experiments identify the regulation of cellular adhesion to fibronectin via hypoxia-inducible factor (HIF)-dependent α5β1 integrin induction as a novel mechanism by which oncogenic ERBB2 signaling promotes cellular survival under adverse conditions (serum deprivation, hypoxia, chemotherapeutic drugs) in vitro, thus adding insight to the processes of ERBB2-driven mammary carcinogenesis at the molecular level. This regulatory circuit helps to explain hitherto uncharacterized consequences of oncogenic ERBB2 activation, especially during the early stages of the carcinogenic process, in which strong ERBB2 signals may lead to nondividing, latent tumor cells exhibiting improved survival characteristics.

Materials and Methods

Antibodies and reagents. ERBB2 (sc-284) and actin (sc-1616)-specific antibodies were obtained from Santa Cruz (Heidelberg, Germany), the α5 integrin-specific antibody (610633) used for immunoblotting was purchased from BD PharMingen (Heidelberg, Germany), and the β1 integrin-specific antibody was purchased from BioGenex (San Ramon, CA). The α5 integrin-specific antibody used for flow cytometric detection as well as for function blocking experiments was purchased from Calbiochem (Schwalbach, Germany; integrin α5 fibronectin receptor Ab-1, clone P1D6). Doxycycline was used as the inducer of ERBB2/NeuT in all experiments. It was obtained as the hydrochloride salt from Sigma (Munich, Germany) and stored as a 200 μg/mL aqueous stock solution at −20°C. Human plasma fibronectin was purchased from Calbiochem. RGD (GRGDSPK, Sigma) and RGE (GRGESP, Takara Bio Inc., distributed by Cambrex, Potsdam, Germany) peptides were reconstituted in PBS according to the manufacturer's instructions. Inhibitors PD98059, SB203580, and wortmannin were purchased from Sigma, SP600125 and GÖ6976 were obtained from Calbiochem. They were dissolved as stock solutions in DMSO and added to a final concentration of 10 μmol/L (Wortmannin), 50 μmol/L (PD98059), 20 μmol/L (SB203580), 10 μmol/L (SP600125), and 5 μmol/L (GÖ6976). Solutions of chemotherapeutic drugs [cisplatin (CDDP), 5-fluorouracil (5-FU)] were prepared by the pharmacy of the Mainz University Hospital.

Cell lines, cell culture, and transfections. The MCF-7/pTet-NeuT and enhanced green fluorescent protein (EGFP) cell lines harboring doxycycline-inducible expression constructs respond to doxycycline administration with expression of EGFP alone (MCF-7/pTet-EGFP) or in combination with a mutationally activated ERBB2-isoform isolated from rat glioblastoma (MCF-7/pTet-NeuT). They have been described in detail in ref. ( 16). MCF-7/pTet-ERBB2 cell lines expressing the human ERBB2 protein in response to doxycycline treatment were generated similarly. All MCF-7-derived cell lines were kept in DMEM/10% fetal bovine serum in humidified 5% CO₂ atmosphere at 37°C during routine culture. 184B5 transformed mammary epithelial cells derived from normal mammary tissue obtained from a reduction mammoplasty were obtained from ATCC (CRL-8799) and cultured as described for MCF-7 cells.

The wild-type ERBB2 cDNA was amplified using the following oligonucleotides as primers (ERBB2/Nhe-For, 5′-AAAAAGCTAGCAGCCGCAGTGAGCACCAT-3′; ERBB2/Nhe-Rev, 5′-AAAAAGCTAGCGGTTCACACTGGCACGTC-3′; restriction sites introduced for cloning are underlined). The resulting cDNA was digested and inserted into the NheI site of pINSpBI-NeuT/EGFP following removal of the NeuT cDNA.

Retroviral gene transfer was done using high-titer retroviral stocks generated by transient calcium phosphate transfection of BING amphotropic packaging cells with a pBabe-puro vector containing the NeuT cDNA (pBabe-NeuT) or vector without insert (pBabe) as previously described ( 16). Early-passage 184B5 cells were incubated with titered retroviral supernatants in the presence of 8 μg/mL polybrene for 12 hours. Twenty-four hours after infection, cells were selected for 5 days in the presence of 2 μg/mL puromycin. Individual cell lines emerging after 2 weeks were analyzed for ERBB2 expression by immunoblotting.

Luciferase reporter gene constructs harboring relevant parts of the 5′-flanking region of the ITGA5 gene (as described in ref. 27) were generated inserting a proofreading PCR-amplified DNA fragment of the respective region (the primers used for amplification were ITGA5prom-F, 5′-GTTACTCGAGGTCTTGAACTCCTGGCCTCA-3′ and ITGA5prom-R, 5′-GTTACTCGAGCCGCTCTTCCCTGTCCTG-3′; restriction sites introduced for cloning are underlined) into the XhoI site of pGL2-basic (Promega, Mannheim, Germany). All constructs were verified by sequence analysis.

Transfections for reporter gene analysis were carried out in 24-well tissue culture vessels. One day before transfection, 3 × 10⁴ cells per well were plated for duplicate analyses. The next day, 900 μL of fresh DMEM/10% fetal bovine serum (with or without doxycycline) per well were added, followed by the addition of 1.5 μg pGL2-DNA (pGL2-basic or pGL2-ITGA5), and 0.5 μg of the secreted alkaline phosphatase (SEAP2) control plasmid (BD Clontech, Heidelberg, Germany), which had been complexed with 5 μL FUGENE 6 transfection reagent (Roche, Mannheim, Germany) according to the manufacturer's instructions. To transfect human embryonic kidney (HEK) cells, 1 μg of pGL2-DNA (pGL2-basic or pGL2-ITGA5), 1 μg of pBabe-DNA (pBabe or pBabe-NeuT), and 0.5 μg of SEAP2 control plasmid were treated with FUGENE 6 accordingly. Cotransfection experiments using von Hippel-Lindau (VHL) tumor suppressor protein expression constructs were carried out in analogy: 0.5 μg pGL2-DNA (pGL2-basic or pGL2-ITGA5), 0.5 μg pcDNA3 or pcDNA3-VHL, and 0.1 μg SEAP2 control plasmid complexed with 3 μL of FUGENE 6 transfection reagent were added to NeuT48 cells 16 hours prior to doxycycline induction. Reporter gene assays were done 48 hours after transfection (in the case of VHL cotransfection experiments, reporter gene assays were done 48 hours after the addition of doxycycline, i.e., 64 hours following transfection).

Luciferase assays were done using the firefly luciferase assay kit (Promega). SEAP assays used the Great EscAPe SEAP chemiluminescence assay kit (BD Clontech). Relative light units (RLU) were calculated as the quotient of luciferase and SEAP activities. All transfection experiments were done at least thrice with similar results.

Quantitative real-time reverse transcription-PCR. RNA isolation, reverse transcription, and subsequent quantitative real-time reverse transcription-PCR experiments were done essentially as described ( 16). The primers used for amplification of ITGA5 were ITGA5-F (5′-GGGCCAAGACTTTCTTGCAG-3′) and ITGA5-R (5′-TGAAGAATCCAAGCTTGTAGAGG-3′). Expression of ITGB1 was quantified using the following primers and cycling conditions: ITGB1-F (5′-TACTTGTGAAGCCAGCAACG-3′), ITGB1-R (5′-GGGGTAATTTGTCCCGACTT-3′). Cycling conditions were: for ITGA5 amplification 95°C for 2 seconds, 61°C for 5 seconds, 72°C for 10 seconds, 40 cycles; for ITGB1 amplification 95°C for 2 seconds, 61°C for 5 seconds, 72°C for 13 seconds, 40 cycles; for pyruvate dehydrogenase β (PDH) amplification 95°C for 2 seconds, 61°C for 5 seconds, 72°C for 8 seconds, 40 cycles. Both programs were preceded by an initial incubation for 10 minutes at 95°C. Relative quantification of cDNA concentrations was done using the Data analysis function of the Roche Molecular Biochemicals Light Cycler Software (V. 3.5). Integrin expression levels were calculated in relation to the expression level of PDH as the reference gene [primers for PDH amplification were PDH-F (5′-GGAGTTGAATGTGAGGTGATAAA-3′) and PDH-R (5′-ACGCAGGACCTTCCATGAT-3′)]. For quantification using real-time reverse transcription-PCR, at least two independent experiments with duplicate amplifications were done.

Immunoblotting. Protein extract preparation, immunoblotting, and antibody detection was done as described in ref. ( 16). α5- and β1 integrin-specific antibodies were used as a 1:1,000 dilution.

Flow cytometry. For flow cytometric detection of α5 integrin, subconfluent cells that were grown for 48 hours in the presence or absence of doxycycline in 75 cm² tissue culture flasks were harvested and washed twice with PBS/EDTA. After incubation in PBS/1% bovine serum albumin (BSA) with the primary antibody specific for the α5 integrin protein (1:100 dilution) for 30 minutes at room temperature, the cells were washed thrice with PBS/1% BSA at 4°C. The secondary antibody (phycoerythrin-conjugated goat anti-mouse IgG; obtained from Santa Cruz) was used in a 1:200 dilution in PBS/1% BSA and the resulting cell suspension was kept at room temperature for 20 minutes. Fluorescence-activated cell sorting (FACS) analysis (FACSCalibur, Becton Dickinson, Heidelberg, Germany) was done after three additional washes in PBS/1% BSA at 4°C. The primary antibody was omitted in controls to check for background fluorescence in order to allow assessment of basal cell surface α5 integrin levels.

Adhesion assay and quantification of living cells. Fibronectin-coating of tissue culture vessels was done by adding 50 μL of a fibronectin solution (50 μg/mL) per well of a 96-well tissue culture plate and a 16-hour incubation at 4°C. After removal of excess fibronectin solution, a 10 mg/mL BSA in PBS solution was added, and the plates were kept at 25°C for an additional hour. Subsequently, the wells were washed once with PBS and directly used for adhesion or survival assays. For profiling and adhesion assays, subconfluent cells were cultured in the presence or absence of doxycycline for 48 hours, harvested by incubation in PBS/EDTA for ∼10 minutes and washed once more in PBS/EDTA. Following an additional wash in serum-free DMEM, the cells were resuspended at a final concentration of 2 × 10⁵ cells per mL in DMEM.

Initial profiling experiments determining the expression of α-integrin subunits in NeuT48 and EGFP control cells were done using the α-integrin mediated cell adhesion array (Chemicon, Hofheim, Germany) according to the manufacturer's instructions. Cells (2 × 10⁴; 100 μL) were added to the reconstituted wells and allowed to bind to the wells coated with different antibodies for 2 hours at 37°C. Following aspiration of the medium, wells were washed twice with assay buffer, 100 μL crystal violet containing Cell Stain Solution were added to each well and bound cells were incubated for 5 minutes at room temperature. Subsequently, the staining solution was removed, the cells were washed four times with deionized water, air-dried, and cell-bound dye was extracted by the addition of 100 μL extraction buffer to each well. The amount of solubilized dye was quantified using a MRX II Elisa reader (Dynex Technologies, Inc., Chantilly, VA).

For adhesion assays, 100 μL of the cell suspensions were plated as quadruplicates in a 96-well tissue culture plate previously coated with fibronectin, BSA, or uncoated, and left at 37°C for 1 hour. Subsequently, nonadherent cells were removed by three washings with PBS, and 100 μL DMEM per well were added. To quantify adherent cells, 10 μL of WST-1 reagent (Roche) were added to the wells, and the plate was kept at 37°C for 1 hour, followed by photometric quantification in a MRX II Elisa reader (Dynex). Quantification of plated cells by omitting the washings controlled for differences in the number of respective cell populations.

To quantify the influence of ERBB2-dependent α5β1 integrin induction on cell growth (as the combined effect of proliferation and survival) under adverse conditions, cells grown in the presence or absence of doxycycline for 30 hours were harvested and washed as described for adhesion assays. 5 × 10³ cells (100 μL) per well were plated as triplicates in pretreated 96-well tissue culture vessels (day 0) and kept in DMEM (with or without doxycycline) in a 5% CO₂-humidified atmosphere or hypoxia (5% CO₂, 0.1% O₂, 94.5% N₂). At the indicated time periods, 10 μL of WST-1 reagent (Roche) were added to quantify living cells. Wells of the plating control were quantified directly following plating on day 0. The specificity of α5β1 integrin ligation on cell growth kinetics was shown by adhesion blockade, preincubating the cells either with a function-blocking monoclonal antibody against α5 integrin (clone P1D6), or nonspecific isotype control antibody for 30 minutes at 37°C before plating. Fibronectin specificity was documented by adhesion blockade, preincubating the cells either with an RGD- or with an RGE-containing peptide.

For the analysis of chemoresistance, experimental media containing either specific chemotherapeutic agents dissolved in serum-reduced OptiMEM I (Invitrogen, Karlsruhe, Germany) or control media without drugs were added after cell adherence. Serial dilutions were done to determine and span the effective dose range for each drug. Drugs in use for clinical breast cancer were selected, including antimetabolites (5-FU) and platinum compounds (CDDP). Following incubation for 72 hours at 37°C in a humidified atmosphere containing 5% CO₂, plates were analyzed by adding the WST-1 reagent as described above. The fraction of surviving cells relative to control were plotted against the log of drug concentration, and the IC₅₀ was interpolated from the sigmoidal curve resulting from nonlinear regression applying a modified Hill model ( 28). All assays described were done at least thrice with similar results.

Results

Identification of α-integrins as targets of oncogenic ERBB2 signaling. Specific interactions of tumor cells and surrounding ECM via cellular integrin receptors have been shown to have profound effects on carcinogenesis due to the modulation of cellular variables influencing tumor cell physiology, like proliferation, survival, as well as invasion and metastatic spread. Recent evidence suggests that interactions between integrin ECM receptors and growth factor receptors play a substantial role in carcinogenesis ( 29– 31). Because overexpression of the RTK ERBB2 is causally involved in the development of ∼30% of mammary carcinomas as well as several other epithelial malignancies, we examined whether aberrant ERBB2 signaling results in the deregulation of integrin receptor expression in tumor cells. We therefore analyzed MCF-7 cells conditionally expressing ERBB2 ( 16) for an ERBB2-dependent response of a subset of relevant α-integrin subunits. As shown in Fig. 1 , doxycycline administration led to selective up-regulation of α2, α3, α5, and αV integrin proteins only in NeuT48 cells, suggesting a role for oncogenic ERBB2 signaling in regulating the tumor cell integrin receptor repertoire. Specific adhesion to fibronectin via the α5β1 integrin heterodimer has recently been shown to promote survival of nonproliferative, dormant breast carcinoma cells under conditions of bone marrow metastasis ( 24). Given this apparent analogy to our observations regarding the effect of oncogenic ERBB2 signal transduction in mammary adenocarcinoma cells (i.e., RTK-signaling provoking a proliferative arrest and a concomitant regulation of integrins) we looked more carefully for expression of the α5β1 integrin fibronectin receptor. We therefore analyzed induced and uninduced NeuT48 as well as EGFP10 control cells by flow cytometry for extracellular α5 integrin. As depicted in the histogram plot of Fig. 1B, a strong increase in membrane-resident α5 integrin protein could be detected in NeuT48 cells in response to hypermitogenic ERBB2 signaling: The mean relative fluorescence of doxycycline-induced NeuT48 cells was 127.67, nearly 4-fold higher compared with uninduced NeuT48 or EGFP10 control cells. Taken together, these results clearly show that oncogenic ERBB2 signaling leads to a selective up-regulation of specific α-integrin subunits in the plasma membrane of breast carcinoma cells, including the α5β1 integrin fibronectin receptor.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1.

Oncogenic ERBB2 signaling stimulates cell surface expression of selective α-integrin subunits in mammary adenocarcinoma cells. A, EGFP10 control cells and NeuT48 cells grown for 48 hours in the presence or absence of doxycycline were incubated in wells coated with different α-integrin monoclonal antibodies indicated below the diagram, followed by washing and subsequent staining, dye extraction and photometric determination of bound cells as indicated in Materials and Methods. Columns, OD at 570 nm, which is proportional to the number of bound cells; results of a representative experiment. The negative control well (nc) contains the goat anti-mouse coating antibody. B, EGFP10 control cells (left) and NeuT48 cells (right) grown for 48 hours in the presence (black columns) or absence (open columns) of doxycycline were subjected to FACS analysis following incubation with an α5β1 integrin-specific primary and phycoerythrin-conjugated secondary antibody. Histograms, number of cells displaying the respective fluorescence intensity given on the horizontal axis (which is proportional to the α5β1 integrin expression level); dotted line, distribution of signals detected in control stainings (uninduced cells, primary antibody omitted) to allow assessment of the basal level of cell surface α5 integrin expression.

Identification of ITGA5 and ITGB1 as target genes of oncogenic ERBB2 signaling in mammary adenocarcinoma cells. Given the increase of membrane-resident α5 integrin protein, we next examined whether this up-regulation is the result of an increase in transcripts and/or protein, and whether the heterodimerization partner of α5 integrin, the β1 integrin subunit, is also induced in an ERBB2-dependent manner. As shown in Fig. 2 , we detected a strong increase in transcripts of both genes on induction of oncogenic ERBB2 signaling in several independent MCF-7/pTet-NeuT cell lines ( Fig. 2A). This increase in ITGA5 and ITGB1 RNA abundance resulted in up-regulation of the respective proteins, as shown in the immunoblots of Fig. 2B. In addition, induction of α5 integrin expression in response to ERBB2 signaling was also observed in 184B5 immortalized mammary epithelial cells constitutively overexpressing the oncogenically activated rat NeuT, as well as in MCF-7 cells inducibly expressing wild-type ERBB2 protein ( Fig. 2C). The NeuT48 cells were used to further characterize the α5 integrin response following doxycycline-induction in a time course experiment. As shown in Fig. 2D, a significant elevation of α5 integrin protein expression could be detected 24 hours after induction of oncogenic ERBB2 signaling. Our data thus clearly illustrates that ERBB2 signaling stimulates expression of the ITGA5 and ITGB1 genes encoding the major cellular fibronectin receptor in mammary epithelial cell lines.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2.

Oncogenic ERBB2 signaling up-regulates the α5β1 integrin fibronectin receptor. A, independent MCF-7/NeuT cell lines (NeuT3, NeuT15, and NeuT48) and MCF-7/EGFP control cells (EGFP10) were grown for 48 hours in the absence (open columns) or presence (black columns) of doxycycline. Total RNA was subjected to quantitative real-time reverse transcription-PCR analysis of ITGA5 expression (left) and ITGB1 expression (right). Columns, means of relative integrin mRNA expression levels, results of a representative experiment; bars, SD. B, protein extracts isolated from the cell lines shown in (A) were subjected to SDS-PAGE, blotted and probed for α5 integrin (left) as well as β1 integrin expression (right). Actin-specific staining is shown below as loading control. C, protein extracts isolated from MCF-7 cell lines conditionally expressing human wild-type ERBB2 (left), and 184B5 human transformed mammary epithelial cells constitutively expressing activated NeuT (right) were subjected to SDS-PAGE and Western blotting. Immunodetection of ERBB2/NeuT- (top), α5 integrin- (middle), and actin expression (bottom). MCF-7/ERBB2 cell lines were grown for 48 hours in the presence (+) or absence (−) of doxycycline. 184B5-V, extracts from empty vector-transfected control cells. D, protein extracts of NeuT48 cells cultured in doxycycline-containing DMEM/10% fetal bovine serum for the indicated time periods were blotted and probed for α5 integrin expression. As indicated, NeuT-expression starts around 6 hours following doxycycline induction ( 16). Actin-specific staining is shown as the loading control.

ERBB2 activates the α5 integrin promoter via HIF. To determine whether the strong ERBB2-dependent increase of ITGA5 transcripts is mediated through stimulation of the ITGA5 promoter, we analyzed the effect of oncogenic ERBB2 signaling on an ITGA5 promoter/luciferase reporter construct (pGL2-A5). As shown in Fig. 3A , doxycycline-mediated induction of ERBB2 signaling leads to a massive (>13-fold) increase in luciferase activity in NeuT48 cells, whereas no stimulation was observed when control cells (EGFP10) were transfected with the promoter-construct. Furthermore, we tested whether the influence of NeuT on the ITGA5 promoter is observed in a different cell line (293 HEK cells), and found increasing ITGA5 promoter activity when increasing amounts of NeuT expression plasmid were cotransfected ( Fig. 2B); thus, demonstrating that the regulatory circuit mediating ERBB2-stimulated α5 integrin expression is functional in a different cell type.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3.

ERBB2 signaling activates the ITGA5 promoter using HIF. A, EGFP10 control cells and NeuT48 cells were transfected with the promoterless pGL2-basic vector (pGL2) or with a reporter gene construct carrying the proximal promoter region of the ITGA5 gene (pGL2-ITGA5). Reporter gene analysis after 48 hours of culture in the absence or presence of doxycycline reveals a strong induction of reporter gene activity (expressed as RLU) only in doxycycline-exposed NeuT48 cells transfected with the pGL2-ITGA5 plasmid. Columns, mean values of a representative experiment done in duplicate; bars, SD. B, cotransfection of HEK cells with pGL2-ITGA5 and increasing amounts of the pBabe-NeuT expression plasmid (as indicated below the histogram) yields increasing activity of the ITGA5 promoter. Columns, mean values of a representative experiment done in duplicate; bars, SD. C, NeuT48 cells were cotransfected with pGL2-ITGA5 and either empty (CMV-) or VHL-containing (CMV-VHL) expression plasmids, and the influence of VHL-expression on ERBB2-mediated activation of the ITGA5 promoter was examined in reporter gene analyses. The results show a significant effect of VHL expression on basal (−Dox) as well as ERBB2-stimulated (+Dox) reporter gene activity, suggesting a prominent role of HIF in ERBB2-mediated regulation of α5 integrin expression. Columns, mean values of a representative experiment done in duplicate; bars, SD. D, quantitative real-time reverse transcription-PCR analysis (duplicate readings) of ITGA5 induction after treatment of NeuT48 cells with specific kinase inhibitors in the presence or absence of doxycycline. Inhibitors of PI3K (Wort), ERK1/2 (PD), P38 (SB), c-Jun-NH₂-terminal kinase (SP), and PKC (GÖ), or DMSO (vehicle control) were added to the cells alone or together with doxycycline, and cells were cultured for 36 hours. Relative ITGA5 transcript levels are depicted with mean values in the absence of doxycycline set to 1. Columns, means of a representative experiment; bars, SD.

ITGA5 mRNA induction by hypoxia has previously been shown in several cell lines, including normal and transformed keratinocytes ( 32), colon cancer cells ( 33), as well as cell lines derived from lung and breast cancers ( 34). HIF is a major transcriptional regulator responsible for α5 integrin induction under conditions of low oxygen ( 33). Furthermore, HIF is described to be an important mediator of ERBB2-dependent transcriptional responses, even under normoxic conditions ( 35– 37). We therefore wanted to address the question of whether ERBB2 signaling uses HIF to induce transcription of the ITGA5 gene. The von Hippel-Lindau tumor suppressor (pVHL) is the adaptor protein of an E3 ubiquitin ligase complex which, under normoxic conditions, efficiently targets the α subunits of the HIF for proteasomal degradation. Hence, we used pVHL expression constructs to reduce endogenous HIF activity in ERBB2 overexpressing cells and analyzed the consequences for ITGA5 promoter activity. As shown in Fig. 3C, expression of pVHL leads to a significant decrease of ITGA5 promoter activity in the absence, as well as in the presence, of ERBB2 signaling. Moreover, most of the ITGA5 transcription in response to ERBB2 seems to be controlled by HIF, as can be concluded from the reduction of ITGA5 promoter activity down to the level of uninduced (−Dox) NeuT48 cells when VHL is cotransfected ( Fig. 3C).

In additional experiments, we examined the effect of small molecule inhibitors on ERBB2-dependent ITGA5 transcription to get a first impression about the pathways involved in HIF-mediated induction of α5 integrin following oncogenic ERBB2 signaling. As can be assessed from Fig. 3D, application of GÖ6976 prevented ITGA5 induction, whereas Wortmannin and SB203580 led to a partial decrease of ITGA5 transcripts, indicating a major contribution of protein kinase C (PKC) isoforms to ERBB2-mediated ITGA5 induction.

Taken together, our data show conclusively that ERBB2 signaling uses the transcriptional regulator HIF to activate the promoter of the ITGA5 gene, with PKC, phosphoinositide-3-kinase (PI3K)/Akt, and P38 being involved in this regulatory circuit.

ERBB2 and hypoxia synergistically activate α5 integrin transcription. Given the convergence of growth factor receptor (ERBB2) and hypoxic signal transduction pathways in HIF, we wanted to test whether ERBB2 signal transduction and low O₂ tension act synergistically on ITGA5 transcription. For this purpose, we first looked at whether hypoxic regulation of α5 integrin expression is functional in MCF-7 adenocarcinoma and 184B5 transformed mammary epithelial cells. As shown in Fig. 4A , an increase of ITGA5 transcripts was observed by reducing oxygen tension in both mammary cell lines. In contrast, VHL-deficient renal carcinoma cell lines (CaKi2, 786-O) did not respond to hypoxia by up-regulating the ITGA5 gene (data not shown). The most likely explanation for this observation is that HIF-α is constitutively stabilized in CaKi2 and 786-O cells, therefore α5 integrin expression cannot be up-regulated further by reducing oxygen tension. Luciferase reporter gene assays analyzing the transcriptional hypoxia response of the ITGA5 promoter also showed a hypoxia-dependent up-regulation of reporter gene activity in MCF-7 cells ( Fig. 4B). We then compared the effect of ERBB2 signaling on α5 integrin induction under normoxic and hypoxic conditions. The results of quantitative real-time reverse transcription-PCR experiments depicted in Fig. 4C show that hypoxia (0.1% O₂, 24 hours) in the absence of NeuT signaling, and NeuT induction (36 hours, doxycycline) under normoxic conditions led to a moderate increase of relative ITGA5 transcript abundance (∼6-fold and ∼17-fold, respectively) when compared with the uninduced (−Dox) state under normoxic conditions. However, when both stimuli were applied (36 hours, doxycycline; 0.1% O₂ for 24 hours starting 12 hours after doxycycline administration), a >50-fold elevation of relative ITGA5 transcript levels compared with normoxic/−Dox culture conditions ( Fig. 4C) was observed, demonstrating a synergistic effect of hypermitogenic ERBB2 signal transduction and low oxygen tension on expression of the α5 integrin subunit.

• Download figure
• Open in new tab
• Download powerpoint

Figure 4.

ERBB2 signaling and hypoxia act synergistically on α5 integrin expression. A, real-time quantitative reverse transcription-PCR analysis (duplicate readings) of ITGA5 mRNA expression in different mammary cell lines cultured under normoxic or hypoxic conditions (0.1% O₂) for 24 hours. The relative ITGA5 mRNA abundance under normoxia was set to 1 for each cell line. Columns, means of a representative experiment; bars, SD. B, MCF-7 cells were transfected with either empty (pGL2) or promoter-containing (pGL2-ITGA5) plasmids, left under routine culture conditions for 16 hours, and then kept either under normoxia or hypoxia (0.1% O₂) for an additional 24 hours. Luciferase activities (expressed as RLU) as measured in the resulting cellular extracts are depicted as mean values of duplicate transfections. Columns, means; bars, SD. C, real-time quantitative reverse transcription-PCR analysis (duplicate readings) of ITGA5 mRNA-expression in NeuT48 cells grown in the absence or presence of doxycycline for a total of 36 hours. Where indicated, cells were put into hypoxic atmosphere (0.1% O₂, 24 hours) 12 hours after the addition of doxycycline. The relative ITGA5 mRNA abundance in NeuT48 cells grown under normoxia in the absence of doxycycline was set to 1 to assess transcriptional induction due to the two stimuli (ERBB2-signaling, hypoxia) alone (compare open versus filled column under normoxic conditions, or open columns under normoxic versus hypoxic conditions, respectively) or in combination (compare open column under normoxic versus filled column under hypoxic conditions). Columns, means of a representative experiment; bars, SD.

ERBB2-mediated up-regulation of the α5/β1 integrin fibronectin receptor results in increased fibronectin adhesion, thereby promoting tumor cell survival under adverse conditions and resistance to chemotherapeutic drugs. Adhesion to fibronectin has been shown to confer a selective survival advantage to tumor cells under various “adverse” conditions ( 20, 22– 24, 38, 39). The up-regulation of the α5β1 integrin fibronectin receptor in response to ERBB2 activation prompted us to test whether the cells exhibit an increased adhesiveness towards fibronectin, and whether this has any functional consequences regarding tumor cell pathophysiology. We therefore induced cells to express NeuT for 2 days and quantified their adhesion on uncoated (standard tissue culture plastic), BSA-coated, as well as fibronectin-coated tissue culture vessels. As depicted in Fig. 5A , uninduced NeuT48 cells behave like uninduced and induced control cells in showing no specific change in adhesiveness towards any of the surfaces. However, when NeuT48 cells were induced to express oncogenic ERBB2, the fraction of cells adhering to fibronectin significantly increased by 89% (from 37% to 70% of cells adhering to fibronectin-coated wells; Fig. 5A, right). This increase in fibronectin adhesiveness could be substantially blocked by pretreatment of induced cells with an α5β1 integrin-specific antibody (reducing the ERBB2-mediated increase in adhesiveness by about 60%; data not shown), suggesting that the α5β1 integrin fibronectin receptor is the major cause of increased fibronectin adhesion of mammary adenocarcinoma cells on oncogenic ERBB2 signaling.

• Download figure
• Open in new tab
• Download powerpoint

Figure 5.

ERBB2 signaling induces increased fibronectin adhesiveness of breast carcinoma cells resulting in improved survival under adverse conditions and chemoresistance. A, adhesion of EGFP10 control cells (left) and NeuT48 cells grown for 48 hours in the absence (open columns) or presence (black columns) of doxycycline to different substrates. Percent cells adhering relative to the plating control was calculated for the respective surfaces [untreated tissue culture (standard TC), BSA-treated (BSA), fibronectin-coated (FN)]. Columns, means of quadruplicate readings; bars, SD. B, EGFP10 control cells (left) and NeuT48 cells (middle) were grown in the presence (filled symbols) or absence (open symbols) of doxycycline for 30 hours, plated (day 0, d0) on untreated (▴) or fibronectin-coated wells (▪), and kept under serum-free conditions for the indicated periods of time. The relative number of living cells represents the portion of living cells at a given time point (determined as described in Materials and Methods) expressed as a multiple of the number of cells plated. Columns, means of triplicate readings; bars, SD. C, the same experimental setup as in (B), but the cells were plated (day 0, d0) and kept under hypoxic conditions (0.1% O₂) for the indicated time periods. The relative number of living cells represents the portion of living cells at the respective time point (determined as described in Materials and Methods) expressed as a multiple of the number of cells plated. Columns, means of triplicate readings; bars, SD. D, EGFP10 control cells (left) and NeuT48 cells (right) were grown in the presence (filled symbols) or absence (open symbols) of doxycycline for 30 hours, plated (day 0, d0) on untreated (▴) or fibronectin-coated wells (▪), and exposed to the indicated concentrations of cisplatin (CDDP; top) or 5-FU (bottom) for 72 hours. The fraction of living cells relative to the drug-free control samples was determined for each concentration using the WST-1 reagent as described. Columns, means of triplicate readings; bars, SD. Survival curves were plotted applying nonlinear regression assuming a modified Hill model ( 28); dashed lines, standard tissue culture vessels; solid lines, fibronectin-coated vessels. Resistance factors (RF) are computed by dividing the IC₅₀ of the respective experimental conditions and cell lines.

To test the influence of ERBB2-dependent increase in fibronectin-adhesion on survival, we measured the relative number of living cells as the net effect of cell proliferation and cell death under conditions of serum withdrawal as well as hypoxia when cultured on untreated or fibronectin-coated cell culture vessels. The growth curves depicted in Fig. 5B and C clearly show an ERBB2-dependent prosurvival effect that is dependent on the presence of a fibronectin matrix. As shown in Fig. 5B, the presence of fibronectin leads to a significant increase in the number of oncogenic ERBB2-expressing (i.e., doxycycline-treated) NeuT48 cells over 6 days of culture without serum. In contrast, no such (pronounced) effect could be observed when doxycycline-treated NeuT48 cells were cultured on untreated plastic, or when uninduced NeuT48 or EGFP10 control cells were analyzed. Again, specifically blocking the α5β1 integrin fibronectin receptor lead to a significant decrease of ERBB2-mediated effects, reducing the fraction of living cells cultured without serum for 6 days by 44% compared with the effect of an isotype control antibody (data not shown). In analogy, ERBB2-dependent fibronectin adhesion of MCF-7 breast carcinoma cells was also shown to have a significant effect on survival under conditions of low oxygen. Pretreatment of NeuT48 cells (+Dox) with RGD peptide reduces the fraction of living cells cultured in hypoxia for 2 days by 49% compared to pretreatment with a control peptide (data not shown).

Although even patients with advanced stage breast cancer initially display responses to chemotherapy, metastatic disease almost invariably relapses, becomes resistant to treatment and therefore remains incurable. One of the main factors contributing to the failure of chemotherapy is that the cancer cells are, or become, resistant to drug-induced apoptosis. Members of the human epidermal growth factor receptor family have a major role in the chemoresistance of breast cancer cells ( 40, 41). Although ERBB2 overexpression has been suggested to confer resistance to both hormonal therapy as well as to chemotherapy, and its expression may also predict the response to certain chemotherapy, the underlying molecular mechanisms remain unclear ( 42, 43). On the other hand, for small cell lung cancer it was previously shown that adhesion to ECM within the tumor microenvironment protects tumor cells from cytotoxic drugs commonly used to treat patients with small cell lung cancer ( 39). This mechanism of primary resistance is mediated via signaling through integrins and may be responsible for minimal residual disease and relapse. We therefore hypothesized that ligation of the fibronectin receptor would, at least partially, mediate chemoresistance in the course of hypermitogenic ERBB2 signaling and examined the sensitivity of our MCF-7 lines to agents currently used for breast cancer chemotherapy. Compared with EGFP10 control cells, NeuT48 cells showed a remarkably increased resistance to 5-FU and CDDP ( Fig. 5D). For both drugs, the dose required to kill 50% of ERBB2-expressing cells increased ∼2-fold compared with the uninduced state. This resistance was further enhanced >2-fold, specifically in the ERBB2-positive cells when attached to fibronectin. In contrast, the resistance conferred to EGFP control cells by attachment to fibronectin was minor, whether induced or uninduced. ERBB2-induced up-regulation of α5 integrin may thus be able to confer resistance to apoptosis, a phenomenon referred to as cell adhesion–mediated drug resistance. Although it has not yet been clearly established whether such a mechanism might account for the acquired resistance found in many tumors it provides a good explanation for local recurrence often seen clinically after chemotherapy. In conclusion, our analyses provide direct experimental evidence for an important effect of the ERBB2/fibronectin receptor regulatory circuit on tumor cell survival under physiologically and clinically relevant adverse conditions like lack of growth factors, low oxygen tension, and chemotherapy.

Discussion

Overexpression of the ERBB2 proto-oncogene within solid tumors is associated with poor prognosis ( 4, 44). Breast cancer cell micrometastasis, the dissemination of single, nonproliferating (or dormant) tumor cells even in the earliest phases of mammary carcinogenesis, has a major effect on patient survival ( 45), and ERBB2-positive epithelial tumor cells in bone marrow characterizes a clinically relevant subset of breast cancer micrometastases ( 46). Originally identified as a potent stimulator of mitosis, oncogenic ERBB2 signaling has recently been found responsible for the activation of an initial antiproliferative response in tumor cells ( 16). Thus, it was suggested that an inherent anticarcinogenic program must be overcome in the early stages of carcinogenesis to progress to malignancy (reviewed in refs. 17, 18).

In an attempt to unravel the molecular circuitry of early steps in ERBB2-driven tumorigenesis, we screened MCF-7 mammary adenocarcinoma cells with inducible ERBB2 for the expression of α-integrin subunits of cellular integrin heterodimers because integrins have been shown to significantly contribute to the pathogenesis of malignancies with respect to tumor cell survival and proliferation ( 29– 31). Among the integrins displaying regulation by oncogenic ERBB2 signal transduction in our cells, we selected the constituents of the major cellular fibronectin receptor (α5 and β1 integrin proteins) for further in-depth investigation, mainly because of its known effect on tumor cell survival and proliferation. By specifically adhering to the ECM component fibronectin, the receptor improves the survival of different cell types under adverse conditions (see refs. 20– 24; see refs. 29, 47– 50 for reviews). Thus, adhesion on fibronectin prevents the apoptosis of Chinese hamster ovary cells upon serum withdrawal, and this prosurvival/antiapoptotic effect of fibronectin-adhesion has been attributed to the selective up-regulation of the antiapoptotic BCL-2 protein, involving SHC, FAK, and RAS subsequently activating the PI3K/AKT pathway in these cells ( 20, 51). In the recent work of Korah et al., the interaction of α5β1 integrin and fibronectin maintains survival of different growth-arrested mammary adenocarcinoma cells in an in vitro model system of bone marrow metastasis. Comparable to results obtained in our inducible cell system, the group elegantly shows that selective up-regulation of the α5β1 integrin fibronectin receptor by FGF2 treatment—thereby simulating the conditions of bone marrow metastasis—leads to long-term survival of dormant clones ( 24).

Tumor cell dormancy is a concept used to characterize nonproliferating single tumor cells that have been shed by the primary tumor and extravasated in a distant target organ (refs. 52, 53; see refs. 25, 26, 54– 56 for reviews). Metastatic growth of tumor cells at distant sites is generally considered a late step during tumorigenesis, but current metastasis models hypothesize that dissemination of single tumor cells begins even in the very early phases of malignancy ( 25, 26, 57). Such disseminated micrometastases can be found in a number of organs, like the bone marrow, as solitary cells, where they may survive for a long period of time, eventually leading to metastatic disease even decades later. On the molecular level, dormancy is not well characterized. One prominent feature of dormant cells is the altered spectrum of integrin cell surface receptors involved in survival signaling. Engagement of specific signaling pathways, like the activity ratio of ERK^MAPK to P38^MAPK, seems to be an important determinant of tumor cell fate—i.e., proliferative growth versus dormancy ( 58, 59). Interestingly, the α5β1 integrin fibronectin receptor, in concert with uPAR, is known to modulate the respective mitogen-activated protein kinase (MAPK) activities and therefore represents a critical relay in the tumor cell with respect to proliferation/dormancy ( 58). Given that α5 and β1 integrins, as shown in the present study, are targets of oncogenic ERBB2 signaling, it is intriguing to speculate that this regulatory circuit represents an important mechanism of breast cancer cell dormancy in vivo, that might even be augmented by additional growth factor signals [e.g., FGF2 as documented by Korah et al. ( 24)]. The observation that the ERBB2 conditional cells display a profound activation of both relevant MAPK signal transduction cascades in response to oncogenic ERBB2 signaling, with p38 MAPK being the pathway responsible for transducing the growth arrest signal ( 16), further supports this concept.

Expression profiling experiments have identified a molecular signature associated with bone marrow micrometastasis in human breast cancer ( 60). By using cDNA arrays to comparatively monitor primary tumors of patients with bone marrow–positive and bone marrow-negative breast cancer, the authors showed that differentially expressed genes include those involved in ECM remodeling, adhesion, and signal transduction. In this study, HIF-1α was among those genes most profoundly up-regulated in bone marrow–positive primary tumors. The consistent regulation of several known HIF-1 target genes in the samples analyzed suggests that pathways connected with HIF impinge on bone marrow metastasis ( 60). Whether HIF-dependent regulation of ECM receptors is responsible for the association of HIF-1α overexpression and unfavorable prognosis (shorter overall survival, shorter disease-free survival; refs. 61– 63) in patients with advanced stage breast cancer remains to be determined. Other integrin ECM receptors, like α6β4 and α6β1 integrin heterodimers which are associated with breast cancer progression have been shown to play important roles in cellular responses to hypoxia ( 64, 65). As HIF-1α is known to be stabilized by ERBB2 signaling under normoxic conditions ( 35– 37), we propose that the α5β1 integrin fibronectin receptor is a major effector in this regulatory circuit and may represent the molecular basis for the HIF-1α-dependent aggressiveness observed in ERBB2-overexpressing breast carcinomas. According to this hypothesis, HIF-1α-dependent up-regulation of the α5β1 integrin fibronectin receptor in response to hypoxia (as shown in refs. 33, 34; and in the present study) and/or ERBB2 (this study) might be one decisive factor selecting for living but nondividing, i.e., dormant chemoresistant cells frequently observed in the bone marrow of patients with breast cancer. The functional relevance of the cellular α5β1 integrin fibronectin receptor for breast cancer cell dormancy has recently been described ( 24). Hence, we suggest that the ERBB2-HIF-fibronectin receptor regulatory interaction is involved in increasing the portion of surviving hypoxic cells overexpressing ERBB2 described recently by Dragowska and coworkers ( 66). This notion is supported by our functional studies demonstrating a prosurvival effect of ERBB2 signaling under hypoxia that is dependent on fibronectin adhesion ( Fig. 5C).

Resistance to chemotherapeutic or radiotherapeutic intervention has been associated with ERBB2 overexpression in tumors ( 41, 44, 67– 70). Given the stimulation of fibronectin adhesiveness following oncogenic ERBB2 signaling, we also examined the influence of this interaction on tumor cell chemoresistance and found that fibronectin ligation lead to a 5-fold increase in resistance against drugs used in breast cancer chemotherapy protocols (5-FU and CDDP), further supporting the clinical relevance of the ERBB2/α5 integrin regulatory interaction. One hallmark of tumor cell dormancy is the inherent resistance of nondividing cells against chemotherapeutic agents ( 26, 55, 71). As suggested by the work of Korah et al. ( 24), as well as by our functional in vitro studies, the α5β1 integrin/fibronectin interaction may be a (perhaps essential) marker for dormant cells, and therefore a very interesting target for breast cancer therapy. Given its well accessible cellular location, peptides targeting fibronectin adhesion via the α5β1 integrin heterodimer, or small-molecule inhibitors targeting downstream survival signaling could be used as therapeutic tools—even in combination with strategies directed against the ERBB2 RTK—to prevent signal transduction essential for long-term survival of dormant tumor cells ( 29, 72). Furthermore, the identification of HIF as a central mediator of ERBB2-dependent regulation of tumor cells' fibronectin adhesiveness adds an attractive drug target ( 73) that might be of special importance, given the synergistic activation of fibronectin receptor expression by ERBB2-signaling and low oxygen tension observed in the work presented.

In summary, we identify α5 integrin, as well as its heterodimerization partner, β1 integrin, as targets of oncogenic ERBB2 signaling. Using Tet-inducible epithelial breast adenocarcinoma cells, we show that ERBB2 signaling leads to the accumulation of ITGA5 and ITGB1 transcripts and protein. Functionally, oncogene-induced expression of the α5β1 integrin fibronectin receptor on the surface of MCF-7 cells enhances their adhesiveness to fibronectin, and adhesion on fibronectin leads to chemoresistance and enhanced survival of ERBB2-overexpressing cells under adverse conditions (specifically nutrient deprivation and hypoxia). We discuss that increased fibronectin adhesiveness induced by oncogenic ERBB2 signaling plays an important role in the course of ERBB2-driven carcinogenesis, and that this may be particularly relevant in the subsistence of long-surviving, dormant micrometastatic cells in the early phase following activation of the ERBB2 proto-oncogene. Hence, the novel regulatory interactions described here could lead to additional (complementary) therapeutic options for ERBB2-dependent malignancies.