Altered mammary epithelial development, pattern formation and involution in transgenic mice expressing the EphB4 receptor tyrosine kinase

The mammary gland is an excellent example of the inductive principle maintained in an adult organ. Tissue composition and the state of differentiation change drastically according to functional requirement and are dependent on mesenchymal-parenchymal interactions, extracellular matrix and local paracrine and endocrine interactive stimuli (Schmeichel et al., 1998). Furthermore, members of at least two gene families specifying developmental pathways during embryonic development (Wnt and Hox) are expressed in a developmental stage-specific manner in the mammary gland (Gavin and McMahon, 1992; Friedmann et al., 1994).

The mammary gland consists of two main components, the ectodermal parenchyma and the mesodermal stroma. The parenchyma, which is composed of secretory and ductal epithelial cells, contractile myoepithelial cells and pluripotent stem cells, develops and functions within the stroma, which consists of fibroblasts and adipose cells (Smith and Cepko, 2001). Unlike other organs, the mammary gland develops mainly in the juvenile and adult organism. With the onset of ovarian function at puberty, the rudimentary epithelial anlagen are induced to proliferate and to invade the surrounding fatty tissue, giving rise to a primitive epithelial ductal tree characteristic of the virgin gland. During pregnancy, the mammary epithelium differentiates and expands drastically until the entire gland is filled with secretory epithelium producing milk to nourish the young. After weaning, the mammary epithelium regresses by massive apoptotic cell death (Richert et al., 2000). Although the mediators of the complex interplay involved in mammary gland development and function are not fully characterized, protein tyrosine kinases, either as receptors or intracellular signal transducers, have been implicated (Fox and Harris, 1997; Hynes et al., 1997).

The murine EphB4 receptor protein tyrosine kinase was originally isolated from the mature mouse mammary gland, and tightly controlled expression was observed during mammary gland development and experimental carcinogenesis (Andres et al., 1994). The Eph family of receptor protein tyrosine kinases (RPTKs), which has 14 characterized members, represents the largest family of RPTKs to date (Pasquale, 1997). The ligands of the Eph family RPTKs, the protein ephrins, are also membrane associated either by a glycosyl-phosphatidylinositol tail (ephrin A family) or are bona fide transmembrane proteins (ephrin B family) (Pandey et al., 1995). The cytoplasmic domains of both the receptors, as well as the ligands (ephrin-B family), become phosphorylated on conserved tyrosine residues following interaction, suggesting that signalling cascades can ensue not only from the receptors but also from the ligands. This may provoke bi-directional signalling and mutual cell-cell communication (Holland et al., 1996; Brückner et al., 1997). This contention is supported by the demonstrations that both Eph receptors and ephrin ligands interact with PDZ-domain-containing proteins (Hock et al., 1998; Torres et al., 1998; Brückner et al., 1999; Lin et al., 1999), proteins implicated in the formation of submembranous scaffolds for the assembly of macromolecular signalling complexes (Garner et al., 2000). Possible mechanism(s) for modulating both receptor and ligand activities exist. Both molecule types associate with phosphatases (Dodelet and Pasquale, 2000) and recently an extracellular metalloprotease, Kuzbanian, has been found to associate with the ephrin-A2 ligand. This protease is activated after ligand activation and cleaves the extracellular moiety of the ligand molecule, thereby terminating ligand signalling and the physical association between cells (Hattori et al., 2000).

The observation that Eph receptors and their ligands exhibit reciprocal expression patterns during embryonic development has led to the suggestion that these molecules play a role in the development and patterning of a variety of tissues during embryogenesis. Indeed, Eph family members are involved in gastrulation, cell migration from the neural crest, segmentation of the early embryo and formation of the somites (Holder and Klein, 1999). The best-studied system so far is the development of the nervous system, where the Eph family and its ligands have a pivotal role in axon guidance, fasciculation and, together with NMDA receptors, in synaptogenesis (Klein, 2001). Recently, it has been shown that the Eph family, in particular EphB4 and its ligand ephrin-B2, are intimately involved in the development of the vascular system during embryogenesis (Wang et al., 1998; Gerety et al., 1999). In contrast to embryonic development, little is known about the function of the Eph family in post-natal and adult life, although some members, including EphB4, are expressed in adult organs such as the kidney, lung and mammary gland (Andres et al., 1994).

We have previously investigated the expression of the EphB4 receptor and the ephrin-B2 ligand proteins during mammary gland development. Expression of both was parenchyma specific, developmentally regulated and estrogen dependent, implicating this receptor-ligand pair in the hormone-dependent morphogenesis of the mammary gland (Nikolova et al., 1998). Expression of the ephrin-B2 ligand was confined to the epithelial cells, whereas the EphB4 receptor was expressed in both the myoepithelial and epithelial cells. Interestingly, the epithelial expression of EphB4 was only observed during proliferative phases of mammary gland development, such as puberty and the follicular phase of the cycle (Nikolova et al., 1998). We have now established transgenic mice exhibiting overexpression of the EphB4 receptor in the mammary epithelium to investigate the effects of untimely epithelial expression of this receptor on glandular growth and differentiation. We demonstrate that unscheduled expression of EphB4 interferes with the architecture of the mammary epithelial tree, alters the response of the epithelial cells to proliferative and apoptotic signals and contributes to the invasive phenotype of mouse mammary tumours.

An expression cassette comprising the 1.2 kb MMTV-LTR promoter and a 0.3 kb fragment encoding the SV40 late polyadenylation site was used for construction of the transgene. This expression cassette confers dexamethasone-inducible expression on linked oncogenes in cell culture (Jaggi et al., 1986). A 0.2 kb fragment encoding the SV40 splice donor/acceptor sequence and a 3.5 kb fragment encompassing the murine EphB4 cDNA depleted of its own polyadenylation site were inserted adjacent to the MMTV-LTR promoter (Andres et al., 1994). Transgenic mice were established commercially by pronuclear injection using fertilized eggs from a F2 C57Bl6×DBA/2 hybrid cross (Animal Facility of the Centre for Molecular Biology, University of Heidelberg, Germany). Mouse tail DNA was prepared according to Andres et al. (Andres et al., 1991), and transgenic mice were identified by PCR using primers recognizing sequences in the LTR promoter and the EphB4 cDNA. The PCR reactions were performed using the PCR Core Kit according to the manufacturers instructions (Roche Diagnostics, Rotkreuz, Switzerland). Two transgenic lines were obtained with stable integration and high expression of the transgene in the mammary gland. Animals of these lines were bred to homozygosity. Homozygosity was ascertained by Southern blotting (Andres et al., 1991) and confirmed by crossing suspected homozygous animals repeatedly with control animals and determining the pattern of transgene inheritance. In order to establish double transgenic mice, EphB4 transgenic animals were first crossed over six generations with inbred C57Bl6 mice (Charles River Wiga, Sulzfeld, Germany) in order to reduce the complexity of the genetic background. MMTV-LTR-neuT transgenic mice of the FeBV strain (Muller et al., 1988) were obtained commercially (Charles River Wiga, Sulzfeld, Germany). Tumour development was followed in the F1 generation of EphB4/C57Bl6 and neuT/FeBV crossings. EphB4 negative littermates served as controls in all experiments.

The fourth inguinal mammary glands were used routinely for histological examination. For whole-mount staining, the mammary glands were spread on coated slides and fixed for four hours in Carnoy’s solution (ethanol:chloroform:glacial acid, 6:3:1). Tissues were washed for 15 minutes in 70% ethanol, rehydrated and stained overnight in carmine alum (2 g/L carmine, 5 g/L potassium sulfate). The next day, tissues were washed in 70% ethanol, dehydrated, cleared with xylene and mounted in Eukitt. For paraffin embedding, the contralateral fourth inguinal mammary glands were fixed for 24 hours in 4% formaldehyde, dehydrated and embedded in paraffin. 4 μm sections were either stained with hematoxilin and eosin or were subjected to immunohistochemical detection of the EphB4 protein as described previously (Nikolova et al., 1998). Sections were examined using a Leica DMRD microscope and images were recorded digitally using a DC200 camera and the Leica LMS programme (Leica, Glattbrug, Switzerland).

Animals were injected with 200 μg per g body weight of bromodeoxyuridine (BdUr, Sigma, Buchs, Switzerland) in PBS three hours before sacrificing. Incorporation of BdUr into DNA was determined immunohistochemically on formaldehyde fixed sections using anti-BdUr-specific antibodies (Roche Diagnostics) and peroxidase-labelled AB Complex (Dako, Glostrup, Denmark). Apoptotic cell death was analysed by the TUNEL assay on formaldehyde fixed sections using the In Situ Cell Death Detection Kit with TMR-red-labelled dUTP, according to the manufacturers instructions (Roche Diagnostics, Rotkreuz, Switzerland).

For RNA analyses, the third mammary gland was snap frozen in liquid nitrogen and stored at –70°C until required. RNA preparation and northern blot analyses were done as described in Andres et al. (Andres et al., 1994). Equal loading of the gels was verified by ethidium bromide staining. For RT-PCR, the RNA was treated with 10 U of DNase for 60 minutes at room temperature and re-isolated by phenol extraction and ethanol precipitation. RT-PCR was performed using the Titan One Tube RT-PCR Kit according to the manufacturers instructions (Roche Diagnostics, Rotkreuz, Switzerland). The primers used to detect transgene-derived transcripts were directed to sequences corresponding to the 3′ untranslated region of the EphB4 cDNA and the SV40 polyadenylation signal of the transgene construct yielding a transgene-specific fragment of 480 bp. The correctness of the PCR product was verified by cloning and sequencing. The absence of contaminating DNA was controlled by conventional PCR. For protein analyses, the snap-frozen contra-lateral third mammary glands were macerated in SDS-PAGE sample buffer, boiled and subjected to immunoprecipitation and/or western blot analyses. Protein concentration was determined visually by amido black staining of 5 μl aliquots of extracts spotted onto nitrocellulose and approximately equal loading confirmed by Coomassie staining of western blot filters. The EphB4 antibodies (Nikolova et al., 1998) and the clone 4G10 phospho-tyrosine antibodies were utilised using conventional immunoprecipitation and western blotting methodology (Küng et al., 1997).

In order to investigate the role of the EphB4 receptor in the biology of the mammary gland, we have established transgenic mice bearing the murine EphB4 receptor cDNA under the control of the MMTV-LTR promoter. This promoter induces expression of linked transgenes in the mammary gland predominantly during late pregnancy and lactation (Hennighausen, 2000). Two transgenic lines, line 3 and 9, were obtained and bred to homozygosity. Both lines transmit the transgene to the progeny in a Mendelian manner typical of somatic integration and exhibit high transgene expression in the lactating mammary glands. In the first instance, we analysed transgene expression throughout mammary gland development. Northern blot analysis revealed strong transgenic EphB4 expression, which was induced at puberty and increased during pregnancy (Fig. 1A). Transgene expression culminated at early lactation and declined after day three of post-lactational involution. In contrast, expression of the endogenous EphB4 gene is downregulated during pregnancy, is essentially absent during lactation and re-appears after day three of post-lactational involution (Andres et al., 1994). As the transcripts of the transgene and the endogenous gene were indistinguishable in size, the transgenic origin of EphB4 overexpression observed at the RNA level was verified by RT-PCR analysis taking advantage of the SV40 polyadenylation signal sequence specific for the transgene (Fig. 1B). The discrepancy between northern blot and RT-PCR results for transgene expression in mature virgins most probably reflects differential transgene expression during the estrous cycle (Andres et al., 1995). RT-PCR analysis of the identical RNAs using an ephrin-B2 specific primer pair showed no striking difference in endogenous ephrin-B2 ligand expression between transgenic and control animals (Fig. 1B). As previously observed, ephrin-B2 expression was less tightly controlled during the life cycle of the mammary gland (Nikolova et al., 1998).

Western blot analyses revealing that the EphB4 receptor protein levels in mammary glands of transgenic animals paralleled those observed at the RNA level (Fig. 1C). Breeding the mice to homozygosity for the EphB4 transgene resulted in a marked increase in transgene expression (Fig. 1C). Immunoprecipitation of protein extracts from lactating mammary glands with a phospho-tyrosine-specific antibody followed by western blot analysis with anti-EphB4 antibody revealed that a portion of the transgenic EphB4 receptor molecules was phosphorylated at tyrosine residues (Fig. 1D). Immunohistochemical staining confirmed, as expected, the absence of EphB4 protein in control lactating animals (Fig. 1E). In contrast, in lactating transgenic mammary glands, the EphB4 protein was readily detected and localized to the parenchymal cells often concentration at the cell membrane. These results confirm that the transgene protein is made by the expected cell type and exhibits the correct subcellular localization. Interestingly, the immunohistochemical staining revealed patchy expression of the transgene in the mammary tissue: the epithelial cells showed large variations in the amount of EphB4 protein in the lobular units and even within the same alveolus (Fig. 1F). In summary, animals homozygous for the EphB4 transgene exhibit considerable overexpression of the EphB4 receptor, predominantly during pregnancy, lactation and early involution. All data presented in this paper were derived using homozygote animals, as the phenotypic consequences observed in heterozygote animals were very similar but had lower penetrance.

Whole-mount staining of the entire mammary epithelial tree at different stages of development served to assess the effect(s) of unscheduled EphB4 expression on the gross architecture of the mammary gland. In control animals, at the onset of ovarian function at 3.5 weeks of age, the rudimentary mammary epithelial anlagen are induced to proliferate and invade the fatty tissue by twisting and branching, giving rise to the epithelial tree characteristic of the mature mammary gland (Fig. 2A,C). At 10 weeks of age, the entire fat pad was populated by epithelium and the pubertal growth phase was concluded, as evidenced by the reshaping of terminal end buds to ductal structures (Fig. 2E,G,I,K). In contrast, the mammary anlagen of EphB4 transgenic animals exhibited severe growth retardation. At 5.5 weeks of age, the epithelial network was rudimentary, with reduced side-branching activity and an absence of second and third order branching (Fig. 2B,D). At 10 weeks of age, the epithelial network was expanded and the extent of the epithelialization of the gland and the ductal morphology were similar to that found in 5.5 week-old control females (Fig. 2F,H). In mature transgenic females at 13.5 weeks, the epithelial network was clearly developed; however, epithelial-free areas of adipose tissue and a large number of growing terminal end buds were still visible, indicating that the growth phase had not yet been concluded (Fig. 2J,L).

In order to investigate the proliferative activity of mammary epithelial cells during pubertal development, animals were injected with BdUr prior to sacrifice, and DNA synthesizing cells detected immunohistochemically. As expected, high proliferative activity was seen in the mammary epithelium of five-week-old control mice (Fig. 3A). In contrast, very low cell proliferation was observed in age-matched transgenic females (Fig. 3B). At 10 weeks of age, cell proliferation had essentially ceased in control animals, whereas age-matched transgenic animals exhibited substantial proliferative activity (Fig. 3C,D). In mature 20-week-old females, almost no difference was evident in the extent of epithelialization and proliferation between control and transgenic mammary glands, indicating that in the transgenic animals mammary epithelial growth had caught up (data not shown).

The main growth phase of the mammary gland occurs during pregnancy when the epithelial cells proliferate profusely, differentiate and at lactation fill the entire gland with secretory epithelium. Whole-mount staining of mammary glands during pregnancy revealed that in transgenic animals, epithelial growth was also retarded during pregnancy (Fig. 4A,B). Furthermore, perturbances in the lobular architecture were also evident: less lobular units were formed in the transgenic animals. The single lobules, however, contained more but smaller alveolar units than in control animals, giving the transgenic lobules a collapsed appearance (Fig. 4C,D). The different growth kinetics were also reflected by the extent of cell proliferation. At day 13 of pregnancy, considerably less epithelial proliferation was observed in the transgenic mammary glands compared with control females (data not shown). In contrast, at day 18 of pregnancy, when cell proliferation normally ceases and is replaced by epithelial differentiation, the mammary glands of transgenic females exhibited the highest proliferative activity (Fig. 4E,F). In the normal mammary gland, apoptotic cell death is essentially limited to the involution process and the anestrous phase of the cycle. Interestingly, we observed significant apoptotic cell death in transgenic mammary glands between day 13 and 18 of pregnancy, with most around day 16 (Fig. 5B,D). In contrast, apoptotic cell death was rare in mammary glands of control animals at any stage of pregnancy. (Fig. 5A,C).

In the lactating mammary gland 10 days after parturition, both the control and transgenic glands exhibited similar extents of epithelialization. In transgenic, lactating mammary glands, however, the alveolar architecture was highly irregular and the epithelial cells appeared to have partly lost cell-cell contact with their neighbours. Cells characterized by areas of strong basophilic staining were often observed exfoliating into the lumen. Furthermore, the luminal space was frequently reduced in volume and the secreted material in the alveolar lumen had a much denser appearance than in the lactating mammary glands of control animals. (Fig. 6A,B). This phenotype was not seen homogenously throughout the gland; areas of near normal morphology could also be detected. Despite the irregular architecture of the mammary glands at lactation, transgenic females were able to nourish their young normally, indicating that sufficient functionality of the mammary gland was achieved. Indeed, no striking differences in the amount and quality of milk protein expression could be detected in transgenic and control animals (data not shown).

After weaning of the young, the mammary gland involutes and most of the secretory epithelium undergoes apoptotic cell death (Strange et al., 1992). We have compared the involution process in transgenic and control animals in situ using the TUNEL assay. In control animals, massive apoptosis was initiated at day one of involution, culminated at days two to three and after day six, cell death was essentially replaced by tissue remodelling (Fig. 7A,C,E,G) (Strange et al., 1992). In contrast, in transgenic animals, apoptosis was first detected at day three of involution and was still prominent at day six of involution (Fig. 7B,D,F,H).

Interestingly, we frequently detected BdUr incorporation in nuclei of parenchymal cells of transgenic mammary glands at day two of weaning (Fig. 8B). This is in contrast to control animals, where only occasional BdUr incorporation was seen during early involution (Fig. 8A). These observations indicate that in the transgenic animals the involution process is delayed and that the epithelial cells respond differently to the apoptotic signals induced by weaning, many replicating their DNA. Clusterin (SGP-2) is a secreted protein originally described as an early marker of cell death in the prostate and the mammary gland (Montpetit et al., 1986; Strange et al., 1992). Recent observations, however, indicate that clusterin is a cytoprotective protein expressed in cells destined to survive apoptosis (Wilson and Easterbrook-Smith, 2000). Northern blot analyses revealed that transgenic animals expressed significantly higher levels of clusterin RNA during the first two days of involution and that expression subsided to control levels at day three of involution, paralleling the appearance of TUNEL positive cells in the transgenic animals (Fig. 8C,D).

The delay in the involution process was also evident from histological examination of involuting mammary glands. Mammary glands of control animals at day two of involution were characterized by dilated alveoli with many dead cells shed into the lumen (Fig. 9A). Stage-matched transgenic mammary glands also exhibited engorgement of the alveoli; however, dead cells within the lumen were only occasionally observed (Fig. 9B). In transgenic animals, histological evidence of apoptosis was first observed three days after weaning (Fig. 9C,D), and after six days, the extent of involution was almost similar to control mice (Fig. 9E,F). Whole-mount staining of regressed mammary glands five weeks after the last lactating period revealed that involution had occurred in both transgenic and control animals. In transgenic animals, however, the involution process was not as extensive and a larger number of alveolar buds remained in the tissue (Fig. 9G,H).

Overexpression of EphB4 has been described in a subset of human and mouse mammary tumour cells (Dodelet and Pasquale, 2000). This, and the observation that transgenic animals exhibited less extensive epithelial regression after weaning, prompted us to investigate the influence of EphB4 expression on mammary tumour formation. The results are summarized in Fig. 10. After an observation time of up to one year, none of the MMTV-EphB4 transgenic females developed mammary tumours. In order to investigate if EphB4 can influence tumour formation, we have crossed the EphB4 transgenic animals with transgenic animals bearing the neuT oncogene under the control of the MMTV LTR promoter. Females bearing the neuT oncogene in a C57Bl6 genetic background developed mammary gland tumours after a latency time of 6.5 months. Although tumours were allowed to grow to a considerable size, tumour growth was always local and metastasis formation was never observed (Fig. 10A). Interestingly, the latency time of mammary tumour appearance in EphB4/neuT double transgenic animals was reduced by about 50% to 3.4 months. Moreover, tumour metastasis to the lung was observed in five out of six double transgenic animals (Fig. 10A). Both the primary mammary carcinomas and the lung metastases of double-transgenic animals were histologically very similar to the solid mammary carcinomas developing in the single transgenic NeuT animals, suggesting that EphB4 expression did not alter the tumour type but its invasive behaviour (Fig. 10B-D). These results indicate that, although EphB4 expression itself does not correlate with an enhanced incidence of tumour development, EphB4 can cooperate with the neuT oncogene and results in earlier tumorigenesis and a more aggressive, invasive tumour phenotype.

We have established transgenic mice overexpressing the EphB4 receptor PTK in the mammary parenchyma. To our knowledge, these MMTV-EphB4 transgenic mice represent the first in vivo model to investigate the effects of deregulated overexpression of the EphB4 receptor in adult animals. Our results have revealed the involvement of this receptor in the regulation of tissue architecture and cellular growth response in the adult mammary gland. It is, however, unclear which mechanism the unscheduled overexpression of the transgene protein uses to exert its effects. We have demonstrated phosphorylation of transgenic EphB4 receptors molecules on tyrosine residues, suggesting that the observed effects may not only be due to overexpression but also to activation. The observation that the greatest effects were seen at developmental stages when endogenous ephrin-B2 is also expressed suggests that transgenic EphB4-ephrinB2 receptor-ligand interactions may at least in part be responsible for the observed phenotype. This may be an oversimplification of the situation given the probable in vivo promiscuity of both Eph receptors and ephrin ligands (Dodelet and Pasquale, 2000). However, a ligand-independent mechanism is also conceivable, relying on activation as a consequence of overexpression. Indeed, Zisch et al. (Zisch et al., 1997) have observed Eph receptor activation solely as a consequence of overexpression.

The EphB4 transgenic mice exhibited developmentally regulated transgene expression consistent with the properties of the MMTV-LTR promoter – the highest expression being detected in lactating mammary glands (Hennighausen, 2000; Dickson et al., 2000). The observed patchy expression pattern of the transgene appears to be a characteristic of the normal lactating mammary gland, as the same phenomenon is also true for milk proteins. It is thought that this phenomenon may be the consequence of local tissue regeneration or local adaptation to variable milk consumption by the young (Bchini et al., 1991; Wilde, 1999) and reflects the remarkable plasticity of this organ. Alternatively, the patchy EphB4 expression could also reflect an attempt of the organ to compensate for possible detrimental effects by silencing transgene expression (Clark, 1998).

Transgenic mammary glands were characterized by disturbed development of the epithelial tree. Beginning at puberty, transgenic epithelial ducts exhibited less branching activity and developed less alveolar buds. This phenomenon was even more evident during pregnancy-induced morphogenesis of the mammary gland. Although not as extensive, this phenotypic consequence is very reminiscent of the defects observed in the mammary epithelium of progesterone receptor (PR) knockout mice (Brisken et al., 1998). These PR knockout animals have confirmed the observations made in vitro that progesterone is responsible for the development of ductal side branches during pubertal and pregnancy-induced development. The local regulators of the frequency of and spacing of the side-branches as well as of bifurcation of the alveolar buds are still unknown. The phenotype observed in the EphB4 transgenic females suggests that this receptor could serve as a negative local control element in these processes.

The low branching activity of the epithelial tree in the EphB4 transgenic mice resulted in a smaller number of individual lobules during pregnancy. The single lobules of the transgenic animals, however, contained more, but smaller, alveoli than control animals. Furthermore, the histological appearance of the single alveoli at lactation of the transgenic animals revealed an irregular, fragile morphology, suggestive of perturbances in cell-cell and cell-matrix interactions. E-and P-cadherin, members of the cell adhesion molecule gene family, are important factors regulating mammary alveolar growth and function. Recent data have described a close relationship between E-cadherin and the Eph receptor family in terms of expression, localization and activation (Zantek et al., 1999; Orsulic and Kemler, 2000). Moreover, ectopic expression of EphA4 in early Xenopus embryos has been shown to disrupt the cadherin-mediated cell adhesion during gastrulation (Winning et al., 1996; Jones et al., 1998). In addition to the cross-talk with the cadherin cell adhesion molecules, the Eph family has also been shown to mediate cell attachment or detachment by regulating integrin function (Huynh-Do et al., 1999; Davy and Parker, 2000). Thus, it is likely that the observed phenotypic consequences of EphB4 overexpression on the mammary epithelial morphology may, at least in part, be due to a disturbed cell-cell and cell-matrix attachment.

The most striking effect observed in the EphB4 transgenic mice concerned the altered proliferative and apoptotic response of the mammary epithelial cells. The analysis of BdUr incorporation revealed that a considerable portion of the transgenic mammary epithelial cells undergo DNA synthesis at day two of involution, a time point at which in control glands apoptosis is initiated and little if any BdUr incorporation was seen. The subsequent, approximately one day delayed, induction of epithelial regression in transgenic glands makes it unlikely that the BdUr-positive cells finished the cell cycle and underwent mitosis. Conceivably, these cells are arrested at the G2/M checkpoint of the cell cycle and subsequently undergo apoptosis. Normally, cells predetermined to proliferate or undergo cell death enter the cell cycle, traverse early G1 phase, whereupon their exact fate is fulfilled (King and Cidlowski, 1998). Recent evidence suggests that, apart from being involved in the control of cell migration, the Eph family is also involved in the regulation of cell death and survival. EphA4 is transiently expressed in motoneurons, which are predestined to undergo cell death during the development of the spinal cord (Ohta et al., 1996). Moreover, several members of the EphA family are responsible for inhibiting cell migration and induction of cell death in ventral spinal cord neurons (Yue et al., 1999). In contrast, overexpression of the ectodomain of the EphB2 receptor in the subventricular zone of the adult brain inhibited cell migration but stimulated proliferation of neuroblasts (Conover et al., 2000).

Our observation of DNA synthesis in the involuting mammary epithelium may reflect an altered response of EphB4 overexpressing cells to death/survival signals in early G1. Furthermore, the high proportion of apoptotic cell death observed in transgenic animals during the proliferative phase at pregnancy may similarly reflect an altered cellular response to proliferative signals, possibly correlated to a disturbed functioning of the E-cadherins. In mammary epithelial cells, E-cadherin can regulate cell survival by activating the retinoblastoma (Rb) gene, and via Rb it can initiate a growth signal conflict in an epithelial cell population induced to undergo apoptosis (Day et al., 1999). It is interesting that in the same time frame that DNA synthesis was observed in the absence of apoptosis, clusterin, a gene involved in protecting cells from apoptosis by unknown mechanisms (Wilson and Easterbrook, 2000), was highly induced in the transgenic mammary glands. This suggests that epithelial survival is, at least transiently, favoured in the transgenic mammary glands predestined for involution.

Although EphB4 overexpression apparently interferes with the growth response of the mammary epithelial cells during involution, tumour formation was not observed in the EphB4 transgenic females. In contrast, in neuT/EphB4 double transgenic females, tumour formation was accelerated and tumour growth was more aggressive than in the single transgenic neuT animals. Similar observations have been made in transgenic animals overexpressing the bcl-2 gene in the mammary epithelium where a delay in post-lactational involution was observed. Similar to the results reported here, bcl-2 overexpression alone did not lead to tumorigenesis; instead, it favoured tumour formation in c-myc/bcl2 double transgenic mice (Jäger et al., 1997). Interestingly, we have observed a considerable increase in the tumour latency time in the single transgenic neuT females crossed into the C57Bl6 genetic background over that in NeuT transgenic females of the pure FeBV strain. This is in agreement with the observations made with other oncogene-bearing mice or with spontaneous tumour formation that the high tumour resistance of the C57Bl6 strain is genetically determined (Macleod and Jacks, 1999). In the EphB4/neuT transgenic females, tumours not only appeared with reduced latency but also metastasised to the lung. Several reports have positively correlated overexpression of Eph family members with carcinogenesis, although direct evidence for the transforming potential is still missing (Dodelet and Pasquale, 2000). Our results indicate that EphB4 overexpression by itself is not tumorigenic but instead favours an invasive phenotype on transformed cells. This notion is further supported by our previous observation that in experimental mouse mammary tumours induced by the Ha-ras oncogene only the anaplastic, invasive tumour cells expressed high levels of EphB4, whereas the majority of the tumour mass was EphB4 negative (Nikolova et al., 1998). Thus, disturbance in the expression of molecules involved in the control of pattern formation may not directly induce transformation but may be instrumental in the acquisition of the malignant, invasive phenotype, possibly by modulating integrin and cadherin functions. It remains to be elucidated if epithelial overexpression of EphB4 can also provoke increased angiogenic potential of mammary cells and thereby facilitate tumour invasiveness.

The authors wish to thank R. R. Friis and R. Strange for stimulating discussions and Barbara Krieger for photographic artwork. This work has been supported by the Swiss National Science Foundation (31-54315.98), the Novartis Research Foundation, the Foundation for Cinical-Experimental Cancer Research, the Swiss Life Insurance Foundation, the Stiftung zur Krebsbekämpfung, and the ‘Stiftung zur Förderung der Wissenschaftlichen Forschung an der Universität Bern’. R.J. was supported by the Bernese Cancer League.