Activation of the A2b1 Integrin Prevents C-Erbb2-Induced Scattering and Apoptosis of Human Mammary Epithelial Cells in Collagen

Home , Integrin

Activation of the a2b1 integrin prevents c-erbB2-induced scattering and apoptosis of human mammary epithelial cells in collagen

Dan BaeckstroÈ m*,1, Pei-Juan Lu2 and Joyce Taylor-Papadimitriou2

1Department of Medical Biochemistry, University of GoÈteborg, Box 440, SE-405 30 GoÈteborg, Sweden; 2Breast Cancer Biology Laboratory, Imperial Cancer Research Fund, 3rd Floor, Thomas Guy House, Guy's Hospital, London SE1 9RT, UK

Constitutive overexpression of c-erbB2 in the mammary Where the signals may generate negative as well as epithelial cell line MTSV1-7 has been shown to result in positive eects on cell growth, the clones selected from epithelial-mesenchymal conversion, anchorage-indepen- transfections of constitutively expressed genes will not dent growth and loss of organized morphogenesis in re¯ect any major growth inhibitory eects of the collagen. To elucidate the events leading to this drastic transfected genes. A more satisfactory approach to change, MTSV1-7 cells and its subclone HB2 (which examining the multiple eects which an expressed gene shows a more strictly epithelial phenotype) were may have on cell behaviour is therefore to use a transfected with the hybrid trk-neu receptor consisting construct where either expression or function of the of the extracellular domain of the trkA nerve growth gene can be regulated after isolation of the transfected factor (NGF) receptor and the transmembrane and clones. We have therefore taken this approach to cytoplasmic domains of c-erbB2 (neu). In cells expressing studying the eects of signalling from the c-erbB2 this construct, c-erbB2 homodimerization can be mi- homodimer in human mammary epithelial cells. micked by addition of NGF. In trk-neu transfectants of The c-erbB2 receptor is overexpressed in a propor- HB2 cells, modest expression led to increased cell tion of breast cancers and over-expression is associated proliferation upon NGF treatment. When clones with with a poor prognosis (De Potter and Schelfhout, higher expression levels were grown in collagen, NGF 1995). c-erbB2 is a tyrosine kinase receptor of the instead induced cell scattering, diminished viability and epidermal growth factor (EGF) receptor family which dramatically increased apoptosis. Interestingly, both the is normally involved in the transduction of signals dissociation of colonies and loss of cell viability could be induced by growth factors such as EGF, heregulin and completely reversed by treatment of the cells with others. This occurs through heterodimerization with antibodies that activate the adhesive capacity of the other members of the EGF receptor family, (Alroy and a2b1 integrin. Long-term NGF treatment of high- Yarden, 1997). Upon ligand binding, c-erbB2 hetero- expressing transfectants generated ®broblastic clones dimerizes with other members of the EGF receptor displaying a reduced expression of integrin a2 and E- family, leading to tyrosine phosphorylation of the cadherin, and extensive apoptosis in collagen. These receptor subunits' cytoplasmic domains. Homodimer- results, which indicate that strong c-erbB2 signalling ization of c-erbB2, which also leads to tyrosine may lead to downregulation and/or inactivation of the phosphorylation, is not induced by ligands but is a2b1 integrin, promoting apoptosis in collagen, provide thought to occur in carcinoma cells overexpressing c- one possible explanation to the increased apoptosis erbB2. frequently seen in early tumour development. Oncogene To study the in¯uence of c-erbB2 overexpression in (2000) 19, 4592 ± 4603. luminal mammary epithelial cells (the cells from which breast cancers develop), the non-malignant mammary Keywords: epithelial-mesenchymal conversion; E-cad- epithelial cell line MTSV1-7 originally derived from herin; heregulin; morphogenesis; anchorage-indepen- cells cultured from human milk, has been used dent growth (D'Souza et al., 1993; D'Souza and Taylor-Papadimi- triou, 1994). In these experiments, transfected clones showing constitutive expression and activation of the c- Introduction erbB2 receptor showed varying degrees of inhibition of expression of epithelial speci®c markers, with very high The experimental approach of transfecting oncogenes expression of c-erbB2 inducing complete epithelial- or proto-oncogenes into cells is widely used to mesenchymal conversion, an event found to be induced investigate the eects of over-expression on cell by other oncogenes (Oft et al., 1996). Concomitant phenotype. While similar signal transduction pathways with the downregulation of expression of epithelial involving many oncogenes are found in most cell types, speci®c markers such as E-cadherin, the a2 integrin it is becoming clear the signals generated may induce subunit was also downregulated in the c-erbB2 widely diering responses, from mitosis to apoptosis or transfectants. While this integrin is widely expressed dierentiation. The responses can be in¯uenced by in mesenchymal as well as epithelial cells, its function several factors, among which the cell phenotype and in the major collagen binding integrin a2b1 in the strength of the signal generated are important. mammary epithelial cells gives it an important role in the interaction of the cells with the extracellular matrix. Furthermore, expression of the a2 subunit is reduced in some primary breast cancers (Berdichevsky et al., *Correspondence: D BaeckstroÈ m Received 20 March 2000; revised 20 June 2000; accepted 7 July 1994b; Zutter et al., 1993). This event, aecting cell ± 2000 matrix interactions, may play a role in the initiation of c-erbB2 homodimerization in mammary epithelial cells D BaeckstroÈm et al 4593 metastasis by allowing the cancer cells to detach, baseline phosphorylation also occurred, presumably invade and proliferate. However, for this to happen, as a result of spontaneous homodimerization. The the cells also have to acquire the ability to proliferate appearance of more than one band in this region is without the attachment to the extracellular matrix, presumably a result of dierences in mobility induced which is necessary for the survival of normal epithelial by dierent degrees of phosphorylation. cells (Frisch and Francis, 1994). The order of events is therefore crucial in determining the survival of the cells NGF-induced changes in morphology, proliferation and proceeding toward outright malignancy. apoptosis of trk-neu transfectant clones To examine the sequence of events occurring when c- erbB2 is overexpressed in human mammary epithelial In the short term, all trk-neu transfectants maintained cells, and to be able to follow eects which may an epithelial morphology when grown on tissue culture negatively aect cell growth, we have used the `trk-neu' plastic, even after treatment with NGF for 1 week hybrid receptor (Sachs et al., 1996). In this construct (after several passages in the presence of NGF, the extracellular domain of the trkA nerve growth however, changes in the morphology of the high- factor (NGF) receptor is fused to the transmembrane expressing HB2/tnz34 and MTSV1-7/tnp20 clones were and cytoplasmic domains of c-erbB2, thus allowing seen, as will be discussed below). When grown for 1 NGF to act as a ligand, inducing homodimerization of week in collagen, all trk-neu-expressing clones ex- the c-erbB2 cytoplasmic domain in transfected cells. hibited a morphological response to NGF (Figure 2a). We have transfected the trk-neu hybrid into MTSV1-7 In contrast to the low-expressing clone HB2/tnp26, cells and also into HB2 cells (a subclone of MTSV1-7 which displayed increased colony size without other cells which display a uniformly spherical colony morphological changes upon NGF treatment, the morphology in collagen). Studies with these transfec- other, more high-expressing clones showed varying tants have shown that signalling from the c-erbB2 degrees of NGF-induced cell scattering (quantitated in homodimer can confer a growth advantage or impair Figure 2b). In the MTSV1-7/tnp20 clone, cell scattering cell proliferation, depending on the level of expression was evident also in untreated cells, but signi®cantly of the receptor. Thus short term treatment with NGF increased upon NGF treatment. of cells in collagen resulted in growth stimulation in The morphological observations suggested that in transfectants expressing low levels of receptor, while the low-expressing HB2/tnp26 transfectant, prolifera- the growth of high expressing clones was inhibited. tion in collagen was stimulated by NGF-induced Signi®cantly, the growth inhibition could be reversed receptor homodimerization, whereas in high-expressing by activation of the function of the a2b1 integrin, clones, the net eect of NGF was inhibitory. This was although the levels of integrin expressed were not con®rmed by MTS proliferation assay of cells growth aected. Longer term treatment of the high expressing in collagen (Figure 3). A comparison with Figure 1a cells with NGF however did lead to a marked shows that in the HB2 transfectants, proliferation upon reduction in the level of expression of the a2 integrin NGF treatment was inversely correlated to the level of subunit, concomitant with a conversion to a mesench- trk-neu expression. On plastic, the changes in pro- ymal phenotype, loss of expression of E-cadherin and liferation induced by NGF were similar, but generally impairment of growth potential which was not less pronounced (data not shown). reversible with integrin activating antibodies. In the clones HB2/tnp38 and HB2/tnz34, cell scattering in collagen was accompanied by the appearance of apparently nonviable cells surrounded by cell fragments (Figure 4a). This observation, Results together with the reduced proliferation seen in these clones (Figure 3) raised the possibility that the Generation of trk-neu transfectant clones1 morphogenetic events induced by trk-neu homodimer- Transfectants were generated by co-transfection of the ization was accompanied by an apoptotic response. In pBATtrk-neu and pBABEpuro plasmids (tnp clones) situ ¯uorescent TUNEL assay of cells grown in or by transfection with the pcDNA3.1zeo/trk-neu collagen was therefore performed on these cell lines, plasmid (tnz clones) using calcium phosphate precipita- showing a dramatic increase in the frequency of stained tion. Mixtures of puromycin-resistant clones obtained nuclei in NGF-treated cells (Figure 4b and quantitated from transfection with the pBABEpuro plasmid (HB2/ in Figure 7b). Apoptosis, which was also determined puro mix and MTSV1-7/puro mix) were used as by assaying fragmented morphology of propidium control transfectants. In the selected transfectant clones iodide-stained nuclei (data not shown), was strongly MTSV1-7/tnp20, HB2/tnp26, and HB2/tnp38 and associated with the NGF-induced scattered morphol- HB2/tnz34, the trk-neu hybrid could be detected at ogy whereas untreated cells growing in coherent the expected 160-kDa position in Western blot using colonies were rarely stained. It should be noted that the 21N antiserum against an epitope in the cytoplas- the quantitation of TUNEL-stained cells had to be mic domain of c-erbB2 (Figure 1a). It is evident from performed after only 4 days of culture since the this blot that the trk-neu hybrid was expressed at clustered growth of untreated cells prevented accurate widely dierent levels in the transfectant clones. As shown in Figure 1b, brief (15-min) NGF treatment of the transfectants led to increased tyrosine phosphorylation in of a band coinciding in size with the transfected 1The properties of the clone HB2/tnp26 have been described in hybrid receptor, whereas control transfectants showed detail elsewhere (BaeckstroÈ m et al., 2000) and data regarding this no response to NGF. In addition to the inducible clone will only be displayed here in comparison to the other, more phosphorylation of trk-neu, a signi®cant level of high-expressing clones.

Oncogene c-erbB2 homodimerization in mammary epithelial cells D BaeckstroÈm et al 4594

Figure 1 Expression and NGF-inducible tyrosine phosphorylation of the trk-neu hybrid receptor in trk-neu transfectants of HB2 and MTSV1-7 cells as measured by Western blot of total cell lysates separated on 6% SDS ± PAGE gels. (a) Measurement of trk- neu expression as detected by the 21N rabbit antiserum directed against the cytoplasmic domain of c-erbB2. Transfectants display a band of varying intensity around 160 kDa, the expected size of trk-neu. (b) Phosphotyrosine blot showing the eect of NGF treatment on the tyrosine phosphorylation of the trk-neu hybrid receptor in transfected clones. Cells were serum-starved for 8 ± 16 h before NGF stimulation for 15 min. Lysates of untreated and NGF-treated cells were subjected to Western blot and probed with an antibody to phosphotyrosine

counting of nuclei in larger colonies. Judging from the NGF in HB2/tnz34 cells was not accompanied by any morphological observations, the apoptotic response major downregulation of integrin a2 (Figure 5b); would be expected to become more pronounced when similar results were obtained with integrin b1 and E- the cells were grown for a longer time. cadherin (not shown). These results indicate that short- term c-erbB2 signalling via activation of the trk-neu hybrid cannot signi®cantly reduce the expression of Expression of integrins and E-cadherin in trk-neu these adhesion molecules on the protein level. In transfectants addition to the short-term experiments, measurements The a2b1 integrin is the major collagen binding integrin were also made on transfectants kept for 10 passages in in HB2 and MTSV1-7 cells (Berdichevsky et al., NGF, also without detectable changes in the overall 1994a), and the response of HB2 cells in collagen to levels of integrins a2 or b1 (data not shown). morphogenetic stimuli has been shown to be modulated by antibodies that modify the function of the a b 2 1 Reversal of NGF-induced scattering and apoptosis by integrin (Alford et al., 1998; Berdichevsky et al., activation of the a b integrin 1994a). Constitutive overexpression of c-erbB2 in 2 1 MTSV1-7 cells has been shown to be associated with The morphological changes in NGF-treated trk-neu scattered colony morphology and downregulation of transfectants in collagen strongly indicated that the the expression of integrin a2 as well as E-cadherin induced c-erbB2 homodimer signalling aected cell- (D'Souza et al., 1993; D'Souza and Taylor-Papadimi- collagen interactions. On the other hand, no signi®cant triou, 1994). Furthermore, promoter activity analyses changes in the levels of collagen-binding integrins could have indicated that c-erbB2 overexpression can down- be detected (the a1 integrin is not expressed by HB2 cells regulate the transcription of integrin a2 in a Ras- (Berdichevsky et al., 1994a)). These observations raised dependent manner (Ye et al., 1996). It was therefore of the possibility that the function of the integrins, rather interest to elucidate the eect of induced c-erbB2 than their abundance, was being modulated by the c- homodimer signalling in the trk-neu transfectants on erbB2 signalling. In order to test this hypothesis, HB2/ integrin and E-cadherin expression. In Western blots of tnp38 and HB2/tnz34 cells were treated (in the presence trk-neu transfectants grown for 1 week on plastic or or absence of NGF) with antibodies that modulate the collagen in the presence or absence of NGF, no major activity of the a2b1 integrin: TS2/16, which is a well- dierence in expression of the integrins a2, a3 or b1 or documented activator of the b1 subunit (van de Wiel-van of E-cadherin could be detected, as exempli®ed in Kemenade et al., 1992); HAS4 (Tenchini et al., 1993), Figure 5a (a3 not shown) with the clone HB2/tnz34 which binds the a2 subunit and has been shown to which displays the most dramatic morphological possess some features of an integrin-activating antibody reaction to NGF treatment. Moreover, by indirect (such as decreasing cell motility on collagen and immuno¯uorescent staining of cells grown in collagen, inhibiting HGF-induced morphogenesis (Alford et al., it could be shown that the cell scattering induced by 1998)); and P5D2 and P1E6, which block the ligand-

Oncogene c-erbB2 homodimerization in mammary epithelial cells D BaeckstroÈm et al 4595

Figure 2 Morphology of trk-neu transfectants when grown in three-dimensional culture in polymerized collagen I. Single-cell suspensions were mixed with a neutralized solution of collagen I, allowed to set and then kept with regular growth medium with or without 50 ng/ml NGF for 1 week. (a) Micrographs showing bright-®eld images of representative colonies with enlarged details. (b) Quantitation of cell scattering in HB2 transfectants based on evaluation of at least 100 colonies from three dierent samples. Colonies were classi®ed as scattered if approximately 25% or more of the cells in one colony appeared as single detached cells

binding capacity of integrins b1 and a2, respectively. These results suggest that the internal signals generated Both the activating antibodies were able to inhibit NGF- by phosphorylation of the c-erbB2 receptor aect the induced cell scattering in collagen (Figure 6); moreover, conformation of the a2b1 integrin and thus the the apoptotic response and the decrease in proliferation functional interaction of the cells with the collagen seen in NGF-treated transfectants were both com- matrix, an interaction which is necessary for cell pletely reversed in the presence of these antibodies survival. (Figure 7). Since the integrin-activating antibodies stimulated proliferation also in the NGF-untreated c-erbB2 homodimer signalling and heregulin treatment transfectants, one may assume that the background elicit distinct responses in HB2/tnz34 cells phosphorylation observed (Figure 1B) confers some degree of integrin inactivation also in the absence of In a previous paper (BaeckstroÈ m et al., 2000) we have NGF. In contrast, the blocking antibodies P1E6 and described how the growth factor heregulin induces P5D2 did not aect the NGF-induced scattering, striking morphogenesis and stimulates proliferation in (Figure 6); instead, it exacerbated the proliferative HB2 cells grown in collagen, and that these responses decrease (Figure 7a [only P5D2 shown]) and apoptosis are dependent on c-erbB2 as well as c-erbB3 (most (not shown) displayed by the NGF-treated transfectants. likely as a result of heregulin-induced c-erbB2/c-erbB3

Oncogene c-erbB2 homodimerization in mammary epithelial cells D BaeckstroÈm et al 4596 heterodimerization). It was therefore of interest to increase in proliferation and a spiky colony morphol- analyse the eect of heregulin on trk-neu transfectants ogy neither of which is as pronounced as in heregulin in which the NGF-induced c-erbB2 homodimer signal- treated parental HB2 cells; however, they are in sharp ling led to a proliferative decrease. As shown in Figure contrast to the scattering and loss of viability seen 8, heregulin treatment of HB2/tnz34 cells elicits an upon NGF treatment. This result strengthens the conclusion that c-erbB2 homodimerization and heregulin-induced c-erbB2/c-erbB3 heterodimerization can give rise to widely distinct biological eects.

Phenotypic changes in trk-neu transfectants following prolonged NGF treatment As related above, when the HB2/tnp26 and HB2/tnp38 transfectants were grown for 10 passages or more in the presence of NGF, no changes in cell morphology could be detected. However, in the case of the high- expressing HB2/tnz34 and the MTSV1-7/tnp20 transfectants, treatment of the cells with 50 ng/ml NGF for several passages led to the emergence of an increasing frequency of ®broblastic cells interspersed with the dominating epithelial phenotype as shown for MTSV1- 7/tnp20 cells in Figure 9a. Immuno¯uorescent staining for intermediate ®lament components of this cell mixture revealed that the ®broblastic cells had acquired expression of the mesenchymal marker vimentin, whereas the epithelial marker keratin 18 was variably downregulated (Figure 9b,c). In order to further analyse the properties of the cells in later passages, ®broblastic and epithelial sub-clones were isolated from MTSV1-7/tnp20 cells that had been treated with NGF for 10 passages and the expression Figure 3 Eect of NGF treatment on proliferation of trk-neu of integrins a2 and b1, E-cadherin and the trk-neu transfectants in collagen. Cells were plated in collagen gels and hybrid were determined by Western blot analysis. As grown for 1 week in the presence or absence of NGF before exempli®ed in Figure 10, the ®broblastic subclones addition of MTS reagent. The fold increase in proliferation was were characterized by strong reductions in the levels of calculated by dividing the absorbance values for NGF-treated cells with those for untreated cells. The graph shows mean and integrin a2 and E-cadherin. Furthermore, the overall s.d. values for three separate experiments performed in triplicate expression level of the trk-neu construct was higher in

Figure 4 Cell scattering in collagen of NGF-treated trk-neu transfectants is associated with apoptosis. (a) Light micrograph of NGF-treated HB2/tnz34 cells in collagen showing scattered cells surrounded by cell fragments. (b) In situ ¯uorescent TUNEL assay of HB2/tnz34 cells. PI, propidium iodide staining showing all nuclei; FITC, TUNEL staining with FITC-labelled dUTP, showing apoptotic nuclei. The upper panels (7NGF) show a coherent colony typical of cells grown without NGF treatment; the lower panels (+NGF) show a colony of NGF-treated cells showing characteristic scattering and abundant apoptotic cells

Oncogene c-erbB2 homodimerization in mammary epithelial cells D BaeckstroÈm et al 4597

Figure 5 In¯uence of NGF treatment on integrin and E-cadherin expression in the trk-neu transfectant clone HB2/tnz34. (a) Western blot analysis for the expression of a2 or b1 integrins and E-cadherin following treatment with 50 ng/ml NGF for 1 week. Controls and NGF-treated cells were grown for 7 days on plastic before lysis, separation of proteins on 6% SDS ± PAGE and transfer to PVDF membranes. All samples were reduced except for in the b1 integrin blot. (b) Expression of integrin a2 in HB2/ tnz34 cells grown in collagen for 1 week in the absence or presence of 50 ng/ml NGF as measured by immuno¯uorescence microscopy. The micrographs show representative colonies stained in the collagen gel with the antibody HAS4 directed against the a2 integrin subunit. In order to preferentially stain extracellular integrins, cells were not permeabilized in these samples the ®broblastic than in the epithelial subclones or in erbB2 overexpression in mammary epithelial cells the progenitor MTSV1-7/tnp20 cells, indicating that (D'Souza et al., 1993; Sheeld, 1998). When plated the ®broblastic cells had originated from the most in soft agar, none of the trk-neu transfectants highly-expressing subpopulations within these transfec- (including the ®broblastic subclones from MTSV1-7/ tants. tnp20 cells) showed any growth with the exception of The ®broblastic cells were also characterized by a HB2/tnz34 (Table 1), where colony formation was low growth rate which was further reduced upon NGF doubled upon NGF treatment. Since HB2/tnz34 is the treatment (Figure 11a). This eect was especially most high-expressing of the trk-neu transfectant clones pronounced in collagen, where a very high frequency described here, these results suggest a critical threshold of apoptotic cells was observed (Figure 11b). In of c-erbB2 signalling below which anchorage-indepen- contrast to the behaviour of NGF-treated HB2/tnp38 dent growth cannot occur. or HB2/tnz34 cells (Figure 6), the impaired growth and frequent apoptosis of these cells could not be ameliorated by treatment with integrin-activating Discussion antibodies (data not shown). Epithelial subclones, although morphologically indistinguishable from the Overexpression of the type I tyrosine kinase receptor c- progenitor MTSV1-7/tnp20 cells both on plastic and in erbB2 has been implicated in the aetiology of mammary collagen (data not shown), displayed a minor reduction carcinoma and agents that interfere with its function are in growth rate suggesting that a more subtle being evaluated for their therapeutic potential in the phenotypic change had occurred in these cells as a treatment of breast cancer (Shak, 1999). The immediate result of prolonged NGF treatment. signalling eect of c-erbB2 overexpression is generally believed to be ligand-independent homodimerization leading to continuous activation of the dimerization Anchorage-independent growth of trk-neu transfectants partners' tyrosine kinase domains. In the present Anchorage-independent growth assayed by colony investigation, we have attempted to analyse the relation- formation in soft agar is closely linked to a malignant ship between this signalling event and conversion of cell phenotype and has been reported to result from c- phenotype on dierent time scales using an inducible

Oncogene c-erbB2 homodimerization in mammary epithelial cells D BaeckstroÈm et al 4598

Figure 7 Proliferation (a) and apoptosis (b) in collagen of high- expressing HB2/trk-neu transfectants in response to NGF and antibodies to the integrin subunits a2 and b1. Single-cell suspensions of the cells were incubated with the antibodies indicated before embedding in collagen and culture in the absence or presence of NGF. In the proliferation assay (a), HB2/tnp38 cells were grown for 7 days before addition of the MTS reagent. Fold changes in HB2 cell proliferation were calculated relative to antibody-untreated HB2 cells and changes in HB2/tnp38 cell proliferation were calculated relative to antibody- and NGF- untreated HB2/tnp38 cells. In the apoptosis assay (b), HB2/tnz34 cells were grown for 4 days before ®xation and staining for apoptotic nuclei using the TUNEL method. The antibodies are: HAS4 against the a2 subunit which is a putative activator of Figure 6 Morphology of HB2/tnz34 cells grown in collagen with ligand binding; P1E6, an inactivating a2 antibody; TS2/16, an and without NGF treatment in the presence or absence of activating b1 antibody and P5D2, an inactivating b1 antibody monoclonal antibodies against the a2 or b1 integrin subunits. (a) Micrographs showing bright-®eld images of representative colonies. (b) Quantitation of cell scattering based on evaluation of the NGF ligand, do indeed result in a stimulation of of at least 100 colonies from three dierent samples. Colonies cell growth. However, where the intensity of the signals were classi®ed as scattered if approximately 25% or more of the is increased, in the HB2 clones expressing higher levels cells in one colony appeared as single detached cells. The of the receptor, short term (4 ± 7 days) treatment with antibodies are: HAS4 against the a2 subunit which is a putative activator of ligand binding; TS2/16, an activating b1 antibody; NGF induces colony disruption and growth inhibition. and P5D2, an inactivating b1 antibody. The a2-inactivating This eect is seen when cells are grown in a collagen antibody P1E6 had a similar eect as P5D2 (not shown) matrix, where the cell ± matrix interaction and survival of these cells depend on the expression and function of the a2b1 integrin (Alford et al., 1998; Berdichevsky et system, the `trk-neu' hybrid receptor. In trk-neu al., 1994a). Although the level of the a2b1 integrin was transfected cells, homodimerization of the c-erbB2 not reduced by short term treatment with NGF, the cytoplasmic domain can be initiated and interrupted by function of the integrin appeared to be impaired. This the addition and removal of NGF, respectively. The can be deduced from the observation that the negative development of human mammary epithelial cell lines eects of NGF on colony morphology and cell expressing the trk-neu receptor has also allowed the proliferation in collagen of the high expressing clones analysis of the eect of increasing the signals from the c- could be reversed by treatment with antibodies which erbB2 homodimer on the behavioural properties of these activate the a2b1 integrin. Furthermore, the morphol- cells without prejudicing the eects towards stimulation ogy of scattered NGF-treated transfectants closely of cell growth. resembled that of parental HB2 cells treated with The data obtained with the trk-neu transfectants sub-lethal doses of blocking antibodies to the a2 or b1 indicate that modest levels of expression of the integrin subunits (Alford et al., 1998; Berdichevsky et receptor, and therefore of signals generated on addition al., 1994a). These ®ndings strongly indicate that

Oncogene c-erbB2 homodimerization in mammary epithelial cells D BaeckstroÈm et al 4599

Figure 8 Morphogenetic (a) and proliferative (b) responses of HB2/tnz34 cells to recombinant human heregulin b1177 ± 244.(a) representative colony morphologies of HB2/tnz34 cells grown for 1 week in collagen without growth factors, in the presence of 10 ng/ml heregulin b1177 ± 244, or 50 ng/ml NGF. (b) proliferation of HB2 and HB2/tnz34 cells in collagen in response to varying concentrations of heregulin b1177 ± 244 in the presence or absence of 50 ng/ml NGF. Cells were grown for 1 week in collagen and then quantitated using the MTS reagent. Fold changes in HB2 cell proliferation were calculated relative to HRG-untreated HB2 cells and changes in HB2/tnz34 cell proliferation were calculated relative to HRG- and NGF-untreated HB2/tnz34 cells signalling from the c-erbB2 homodimer can negatively (Figure 10) and tyrosine phosphorylation (not shown) regulate the function of the a2b1 integrin, leading to of the trk-neu hybrid was signi®cantly higher than in decreased viability. One possible mechanism for this the epithelial transfectants from which they originated. regulation is suggested by the results of Lo et al. (1999) It thus appears that in the MTSV1-7/tnp20 cells that who found enhanced interaction between focal adhe- had the highest degree of trk-neu expression, prolonged sion kinase and paxillin in NIH3T3 cells upon c-erbB2- NGF-induced signalling exceeded a critical threshold transfection. An antagonism between sustained c-erbB2 level where gain of ®broblastic phenotype and loss of signalling and integrin-mediated matrix adhesion is epithelial markers appeared more or less simulta- also suggested by the results of Tagliabue et al. (1996) neously. who observed laminin-dependent complex formation Epithelial-mesenchymal conversion has earlier been between c-erbB2 and the a6b4 integrin, leading to observed as a result of constitutive overexpression of surface downregulation of c-erbB2 and growth inhibi- full-length c-erbB2 in MTSV1-7 cells (D'Souza et al., tion. However, as the present results show, the 1993; D'Souza and Taylor-Papadimitriou, 1994). The outcome of c-erbB2 signalling is highly dose-dependent resulting cell line, ce1, also displayed decreased levels and also likely to be strongly in¯uenced by the cell of E-cadherin and integrin a2. In addition, the ce1 cells type. In this context, it is interesting to note that the had acquired anchorage-independent growth and in¯uence of EGF on integrin expression and function tumorigenicity as assayed by growth in soft agar and has been reported to be either positive or negative tumour formation in mice, respectively. Other studies depending on the degree of receptor expression have also reported these changes in c-erbB2 over- (Genersch et al., 1996; Krensel and Lichtner, 1999). expressing mammary epithelial cells (Baasner et al., When NGF treatment of high-expressing trk-neu 1996; Sheeld, 1998). However, none of the ®bro- transfectants was extended to a period of several blastic clones derived from the MTSV1-7/tnp20 cells weeks, subpopulations appeared which displayed a grew in soft agar, regardless of NGF treatment. The ®broblastic morphology, vimentin expression and loss same behaviour was also found in the other trk-neu of E-cadherin and integrin a2. In ®broblastic clones transfectants with the exception of HB2/tnz34 cells isolated from MTSV1-7/tnp20 cells, the expression where a modest, but NGF-enhanced colony frequency

Oncogene c-erbB2 homodimerization in mammary epithelial cells D BaeckstroÈm et al 4600

Figure 10 Analysis of the expression of integrin a2, E-cadherin and the trk-neu hybrid (as detected by the 21N antiserum against the c-erbB2 cytoplasmic domain) by Western blot in MTSV1-7/ tnp20 cells and representative epithelial and ®broblastic subclones obtained after prolonged NGF treatment. The control transfectant MTSV1-7/puro mix is included for comparison. E, epithelial phenotype; F, ®broblastic. Lysates were separated on a 6% SDS ± PAGE gel before transfer to PVDF membranes

transfectants described here (as measured in Western blots using the 21N antiserum, data not shown). The signalling pathway leading from c-erbB2 signalling to anchorage independence is likely be dependent on both c-src and PI(3)-kinase, since inhibition of either has been shown to block growth in soft agar of c-erbB2- transfected mammary epithelial cells (Ignatoski et al., 2000; Sheeld, 1998). The response to c-erbB2 homodimer signalling described here does not appear to be a general eect of activation EGF receptor family members, as treatment of HB2/tnz34 cells with the growth factor Figure 9 Morphology and intermediate ®lament expression of heregulin, which requires the participation of c-erbB3 MTSV1-7/tnp20 transfectants after prolonged NGF treatment. The phase-contrast image (a) shows the mixed morphology and/or c-erbB4 in functional receptor heterodimers, resulting from NGF treatment of this transfectant: groups of resulted in compact colony morphology and prolif- coherently growing epithelial cells (left) typical of the original erative stimulation in collagen, a phenotype widely MTSV1-7/tnp20 transfectant clone were interspersed with scat- dierent from that seen upon NGF treatment. Since all tered cells of ®broblastic morphology (right). In the immuno- ¯uorescence micrographs (b, c), the cells have been probed with available data strongly indicate that the c-erbB2/c- antibodies against keratin 18 (b) or vimentin (c) followed by erbB3 heterodimer is the major heregulin receptor in FITC-labelled secondary antibody (green) and nuclear counter- epithelial cells, including HB2 cells (BaeckstroÈ m et al., staining with propidium iodide (red). Dense clusters of epithelial 2000), one may conclude that the signalling molecules cells were strongly stained with the keratin 18 antibody, whereas recruited by the c-erbB2/c-erbB3 heterodimer are the scattered ®broblastic cells mostly displayed weak or no staining (b). The reverse is true for the vimentin staining (c). All dierent from those recruited by the c-erbB2 homo- scale bars=100 mm dimer (Khwaja et al., 1997) One possible mediator of such signalling which could be recruited dierently could be PI(3)-kinase, which is known to promote cell was observed. One may thus conclude from the present survival under conditions of matrix deprivation data that although the conversion to a ®broblastic (Khwaja et al., 1997) and which is much more phenotype is closely linked to downregulation of eciently recruited by c-erbB3 than by c-erbB2. Other integrin a2 and E-cadherin, it is not necessarily linked recent studies have showed dierences in the in¯uence to anchorage independence. Nowak et al. (1997) have on adhesive and proliferative properties between proposed that only very high levels of c-erbB2 dierent EGF receptor dimers. Also using arti®cial signalling will confer a fully malignant phenotype, dimerization of receptor hybrids, Muthuswamy et al. and this hypothesis could explain the discrepancy in (1999) showed a strikingly higher focus-forming ability phenotype between ce1 cells and the trk-neu transfec- in rat ®broblasts upon c-erbB2 homodimerization tants, since the expression of c-erbB2 in ce1 cells is compared to heterodimerization between c-erbB2 and higher than trk-neu expression in any of the the EGF receptor. Adelsman et al. (1999) have

Oncogene c-erbB2 homodimerization in mammary epithelial cells D BaeckstroÈm et al 4601 Table 1 Colony formation frequency in soft agar of trk-neu transfectants with and without treatment with 50 ng/ml NGF. Data show mean values and standard deviations of triplicate samples from a representative experiment (repeated twice) 7NGF +NGF Cell line % colonies s.d. % colonies s.d.

HB2 50.5 ± 50.5 ± HB2/tnp26 50.5 ± 50.5 ± HB2/tnp38 50.5 ± 50.5 ± HB2/tnz34 8.6 1.7 18.1 3.1 MTSV1-7/tnp20 50.5 ± 50.5 ± subclone 5 (fibroblastic) 50.5 ± 50.5 ± subclone 6 (epithelial) 50.5 ± 50.5 ± subclone 9 (epithelial) 50.5 ± 50.5 ± subclone 12 (fibroblastic) 50.5 ± 50.5 ± T47D (positive control) 63.7 2.9 N/D N/D

negatively aects cell-collagen interaction in mammary epithelial cells both by aecting integrin function and expression. These anti-adhesive eects would be advantageous to carcinoma cells by promoting tissue disruption and invasive behaviour. However, c-erbB2- induced adhesion downregulation does not appear to be automatically linked to protection against the apoptotic response that follows loss of functional matrix contact in epithelial cells; rather, c-erbB2 signalling must exceed a critical threshold level for such protection to occur. This conclusion would oer one explanation to the often reported phenomenon that tumour development is frequently accompanied by increased apoptosis (Naik et al., 1996; Soini et al., 1998). A cell embarking on a tumorigenic pathway by overexpressing c-erbB2 would thus have to pass a bottleneck in viability until either the level of c-erbB2 signalling becomes high enough or some other phenotypic or genotypic change occurs to prevent anoikis; during this interval, an increase in the generation of apoptotic cells would be expected.

Materials and methods

Growth factors and antibodies NGF from mouse submaxillary gland was purchased from Promega. Recombinant human heregulin b1177 ± 244, a frag- Figure 11 Proliferation and apoptosis of MTSV1-7/tnp20 cells ment (Mr=7884) containing the receptor-binding EGF-like and epithelial and ®broblastic subclones in the absence or domain was a gift generously provided by Dr MX presence of 50 ng/ml NGF. (a) Proliferation assays of cells grown Sliwkowski, Genentech, South San Francisco, CA, USA. on plastic or in collagen for 1 week. Cell densities in triplicate samples were quantitated by addition of the MTS reagent and The monoclonal antibodies (mAbs) TS2/16 and P5D2 which measuring absorbance at 450 nm after blank subtraction. (b) activate and suppress the binding capacity of integrin b1, Quantitation of apoptosis in cells grown in collagen for 4 days respectively, were prepared from hybridoma cells available using TUNEL assay with FITC-labelled dUTP. The increase in from ATCC. The mAb HAS4 against integrin a2 was a gift apoptosis upon NGF treatment in subclone 5 is signi®cant generously provided by Dr Fiona Watt, Imperial Cancer (P50.02, Student's t-test) Research Fund, London UK. The rabbit antisera AB1936 and AB1920 against integrins a2 and a3, respectively, as well as the function-blocking integrin a2 mAb P1E6 (MAB1950) observed a marked dierence between EGF and were purchased from Chemicon. The 21N rabbit antiserum heregulin-b in their capacity to stimulate b1 integrin against a region in the cytoplasmic domain of c-erbB2 was function in mammary carcinoma cells. Since c-erbB2 obtained from NeoMarkers. The anti-vimentin mAb Vim3B4 has been shown to be a preferred heterodimerization was purchased from DAKO. For the detection of keratin 18, partner for the other EGF receptor family members the mAb Da7 (Lauerova et al., 1988) was used. Phosphotyr- (Graus-Porta et al., 1997), it is evident that dierences osine was detected using the mAb 4G10 from Upstate in the heterodimerization partners' substrate domains Biomedical. The mAb HECD-1 against E-cadherin was purchased from Takara Biomedicals. and their accessibility to docking proteins strongly modulate the biological eects of c-erbB2 activation. Taking together the phenotypical responses of Expression vectors MTSV1-7 and HB2 cells to inducible c-erbB2 homo- The pBATtrk-neu vector was a gift generously supplied by dimerization, one can conclude that c-erbB2 signalling Drs Martin Sachs and Walter Birchmeier of the Max Planck-

Oncogene c-erbB2 homodimerization in mammary epithelial cells D BaeckstroÈm et al 4602 Institut, Berlin, Germany. The pcDNA3.1zeo/trk-neu plasmid volumes were scaled up to allow covering of the plug. After was generated by excision of the trk-neu cDNA using SalI TUNEL reaction and propidium iodide (PI) counterstaining, and cloning into the XhoI site of pcDNA3.1zeo+ (Invitro- plugs were immersed in 10 mM tris-HCl, pH 8.0, 50% gen). glycerol for 1 h and then put inside a ring of silicone rubber on a glass slide and overlayed with a cover glass. Light Cell culture The cell lines MTSV1-7 (Bartek et al., 1991) pressure was applied to the cover glass to reduce the sample and HB2 (Berdichevsky et al., 1994a), derived from human thickness and the silicone rubber was allowed to set before mammary epithelial cells obtained from mothers' milk, were viewing. The frequency of apoptotic cells was evaluated as cultured in DMEM supplemented with 10% foetal calf serum the ratio of TUNEL-stained nuclei (FITC ¯uorescence) to all (FCS), 5 mg/ml hydrocortisone and 10 mg/ml insulin. Trans- nuclei (PI ¯uorescence) in three random ®elds/plug in fectants carrying puromycin or zeocin resistance genes were triplicate, giving a minimum nucleus count of 300/data point. grown in the presence of 0.5 mg/ml puromycin or 5 mg/ml zeocin, respectively. Immunofluorescence microscopy Cell culture in collagen was performed by suspending single cells in a neutralized solution of bovine dermal For analysis of cells grown on plastic, cells were plated in 35- collagen I (Vitrogen 100, Collagen Biomaterials) to a ®nal mm tissue culture dishes at low density (2000 cells/dish) and cell concentration of 104 cells/ml and dispensing 100 mlof allowed to grow for 1 week. The cells were then ®xed in cold suspension/well in 96-well plates. Cells were allowed to set at (7208C) methanol for 5 min. After three washes with PBS, 378C for 90 ± 120 min during which time the plates were the dishes were incubated with 20% FCS in PBS for 1 h inverted every 15 min to ensure an even distribution of cells followed by an incubation with monoclonal antibodies within the collagen gel. The wells were then fed 200 ml regular against vimentin or keratin 18 for 1 h. After three washes growth medium and refed after 24 h with 100 ml medium with PBS, ¯uorescein-conjugated rabbit anti-mouse immu- containing growth factors, antibodies etc. where applicable. noglobulin (DAKO, dilution 1/50 in PBS with 5% FCS) was added. After 1 h, the dishes were washed once with PBS containing 5 mg/ml propidium iodide and twice with PBS. Transfections Before viewing under a ¯uorescence microscope, 100 mlof MTSV1-7 and HB2 cells were transfected with trk-neu- 10 mM tris-HCl, pH 8.0, 50% glycerol was added and a containing vectors by means of calcium phosphate-DNA 32 mm circular coverglass was placed over the cells. coprecipitation using standard techniques (Sambrook et al., Cells grown in collagen in 96-well plates were ®xed, washed 1989). Resistant clones were isolated by ring cloning and and mounted as described above for the TUNEL apoptosis tested for expression of the c-erbB2 cytoplasmic domain in assay. In some analyses, permeabilization with Triton X-100 immuno¯uorescent staining, NGF-induced morphogenesis in was omitted. collagen and tyrosine phosphorylation. Western blot Proliferation assays Lysates were prepared from cells grown to 80 ± 90% Cells were plated from a single cell suspension into 96-well con¯uence in 6-cm dishes on plastic or a layer of 500 ml plates (for plating of cells in collagen, see `Cell culture'). polymerized collagen I. Cells were washed three times with After having grown for the required time period, 20 ml/well of PBS before lysis in 50 mM tris-HCl, pH 7.5, 0.15 M NaCl, MTS reagent (CellTiter96 AQueous One Solution Prolifera- 1% Triton X-100, 5 mM EDTA, 0.5 mM phenylmethyl tion Assay, Promega) was added to each well. Absorbance at sulphonyl ¯uoride, 100 KIE/ml aprotinin, 50 mg/ml leupeptin. 450 nm was read after 2 ± 5 h and the net absorbance was In phosphotyrosine analyses, cells were kept with serum-free calculated from triplicate data points after subtraction of medium 8 ± 16 h before stimulation and washed and lysed in blank absorbance (medium only or collagen+medium only). the presence of 1 mM Na3VO4. The linearity of the assay at the cell densities used was The protein concentrations of the lysates were determined veri®ed in separate experiments. using the BCA protein assay (Pierce) and equal amounts of protein for each sample were loaded onto 6% SDS ± PAGE gels. For the analysis of E-cadherin and integrin a2, the Soft agar assay samples were reduced, whereas unreduced samples were used Cells were plated at 350 cells/well in 24-well plates in 0.3% for the analysis of integrin b1. The separated proteins were agar on top of a 0.5% agar base. Both agar layers contained transferred to PVDF membranes (Hybond-P, Amersham DMEM with supplements as described above. Two days after Pharmacia) by electroblotting at 3 mA/cm2 for 2 h in 25 mM plating, 1 ml medium (with or without 50 ng/ml NGF) was tris, 0.192 M glycine, 20% methanol. The membranes were added to each well. After 2 weeks in culture, colony blocked in PBS with 5% nonfat dry milk, 0.1% Tween-20. formation was evaluated by counting the frequency of Membranes were then incubated with primary antibody colonies containing more than 10 cells. The breast cancer diluted in blocking solution. Membranes were washed in cell line T47D was used as a positive control. PBS, 0.1% Tween-20 and bound antibodies were detected using horseradish peroxidase-conjugated secondary antibodies in combination with the Enhanced Chemiluminescence In-gel apoptosis assays (ECL) system (Amersham Pharmacia). Apoptotic nuclei of cells grown in collagen were stained by a modi®ed version of the terminal dUTP nick-end labelling (TUNEL) protocol (Apoptosis Detection System, Fluores- cein, Promega). Brie¯y, cells were grown in collagen in 96- well plates as described above, the medium was replaced with Acknowledgements cold 4% paraformaldehyde in PBS and the ®xation was The authors thank Dr Martin Sachs and Prof Walter allowed to proceed for 2 h on ice before removing the cell- Birchmeier for the gift of the pBAT/trk-neu plasmid and containing collagen plugs from the plates using a 23-gauge Dr Fiona Watt for supplying the HAS4 antibody. The needle (®xation before removing the plugs ensured excellent P5D2 hybridoma developed by EA Wayner was obtained preservation of colony morphology). The plugs were then from the Developmental Studies Hybridoma Bank devel- washed and stained according to the manufacturer's descrip- oped under the auspices of the NICHD and maintained by tion but all incubation times were doubled and reagent The University of Iowa. This work was supported by Assar

Oncogene c-erbB2 homodimerization in mammary epithelial cells D BaeckstroÈm et al 4603 Gabrielsson's Fund, the Swedish Cancer Fund (grant no. European Commission (fellowship grants no. 975098), the Imperial Cancer Research Fund and the ERBCHBGCT 94-0556 and ERB4001GT965003).

References

Adelsman MA, McCarthy JB and Shimizu Y. (1999). Mol. Lauerova L, Kovarik J, Bartek J, Rejthar A and Vojtesek B. Biol. Cell., 10, 2861 ± 2878. (1988). Hybridoma, 7, 495 ± 504. Alford D, BaeckstroÈ m D, Geyp M, Pitha P and Taylor- Lo SS, Lo SH, Wang SC and Hung MC. (1999). Mol. Papadimitriou J. (1998). J. Cell. Sci., 111, 531 ± 532. Carcinog., 25, 150 ± 154. Alroy I and Yarden Y. (1997). FEBS Lett., 410, 83 ± 86. Muthuswamy SK, Gilman M and Brugge JS. (1999). Mol. Baasner S, Vonmelchner H, Klenner T, Hilgard P and Cell. Biol., 19, 6845 ± 6857. Beckers T. (1996). Oncogene, 13, 901 ± 911. Naik P, Karrim J and Hanahan D. (1996). Genes. Dev., 10, BaeckstroÈ m D, Alford D and Taylor-Papadimitriou J. 2505 ± 2516. (2000). Int. J. Oncol., 16, 1081 ± 1090. Nowak F, Jacquemin-Sablon A and Pierre J. (1997). Exp. Bartek J, Bartkova J, Kyprianou N, Lalani EN, Staskova Z, Cell Res., 231, 251 ± 259. Shearer M, Chang S and Taylor-Papadimitriou J. (1991). OftM,PeliJ,RudazC,SchwarzH,BeugHandReichmann Proc. Natl. Acad. Sci. USA, 88, 3520 ± 3524. E. (1996). Genes. Dev., 10, 2462 ± 2477. Berdichevsky F, Alford D, D'Souza B and Taylor-Papadi- Sachs M, Weidner KM, Brinkmann V, Walther I, Obermeier mitriou J. (1994a). J. Cell. Sci., 107, 3557 ± 3568. A, Ullrich A and Birchmeier W. (1996). J. Cell. Biol., 133, Berdichevsky F, Wetzels R, Shearer M, Martignone S, 1095 ± 1107. Raemakers FCS and Taylor-Papadimitriou J. (1994b). Sambrook J, Fritsch EF and Maniatis T. Molecular Cloning. Mol. Cell. Dierentiation, 2, 255 ± 274. A Laboratory Manual, 2nd edn. Cold Spring Harbor D'Souza B, Berdichevsky F, Kyprianou N and Taylor- Laboratory Press, Cold Spring Harbor. Papadimitriou J. (1993). Oncogene, 8, 1797 ± 1806. Shak S. (1999). Semin Oncol., 26, 71 ± 77. D'Souza B and Taylor-Papadimitriou J. (1994). Proc. Natl. Sheeld LG. (1998). Biochem. Biophys. Res. Commun., 250, Acad. Sci. USA, 91, 7202 ± 7206. 27 ± 31. De Potter CR and Schelfhout A-M. (1995). Virchows Arch., Soini Y, Paakko P and Lehto VP. (1998). Am.J.Pathol.,153, 426, 107 ± 115. 1041 ± 1053. Frisch SM and Francis H. (1994). J. Cell. Biol., 124, 619 ± Tagliabue E, Ardini E, Pellegrini R, Campiglio M, Bufalino 626. R, Jeschke M, Colnaghi MI and Menard S. (1996). Br. J. Genersch E, Schuppan D and Lichtner RB. (1996). J. Mol. Cancer, 74, 1427 ± 1433. Med., 74, 609 ± 616. Tenchini ML, Adams JC, Gilberty C, Steel J, Hudson DL, Graus-Porta D, Beerli RR, Daly JM and Hynes NE. (1997). Malcovati M and Watt FM. (1993). Cell. Adhes. EMBO J., 16, 1647 ± 1655. Commun., 1, 55 ± 66. Ignatoski KM, Maehama T, Markwart SM, Dixon JE, van de Wiel-van Kemenade E, van Kooyk Y, de Boer AJ, Livant DL and Ethier SP. (2000). Br.J.Cancer,82, 666 ± Huijbens RJ, Weder P, van de Kasteele W, Melief CJ and 674. Figdor CG. (1992). J. Cell. Biol., 117, 461 ± 470. Khwaja A, Rodriguez-Viciana P, WennstroÈ mS,WarnePH Ye JP, Xu RH, Taylor-Papadimitriou J and Pitha PM. and Downward J. (1997). EMBO J., 16, 2783 ± 2793. (1996). Mol. Cell. Biol., 16, 6178 ± 6189. Krensel K and Lichtner RB. (1999). Int. J. Cancer, 80, 546 ± Zutter MM, Mazouijan G and Santoro SA. (1993). Am. J. 552. Pathol., 142, 1439 ± 1448.

Oncogene