Introduction

Many strains of rats develop multiple mammary adenocarcinomas when initiated with carcinogens such as methylnitrosourea (MNU). Several strains, however, are completely resistant to mammary tumorigenesis (Wood et al., 2002). The Copenhagen (Cop) rat is the most studied of these strains. In order to characterize the phenotype associated with resistance, we recently examined the mammary glands from both Cop and susceptible Wistar-Furth (WF) rats at various times following MNU treatment (Korkola and Archer, 1999; Korkola et al., 1999). We found that both strains develop the earliest detectable preneoplastic lesions known as intraductal proliferations (IDPs). The majority of IDPs from both strains contain activating mutations in the Ha-ras oncogene that are common in MNU-induced rat mammary adenocarcinomas (Sukumar et al., 1983). By 45 days after MNU treatment, a number of IDPs have progressed to more advanced lesions such as ductal carcinomas in situ and adenocarcinomas in the glands from WF rats, whereas the IDPs in Cop rats fail to progress and instead decline in number such that by 60 days, the glands are essentially free of lesions. These results suggest that initiation of transformation by activation of Ha-ras occurs in both strains, but progression towards malignancy is blocked in Cop rats and preneoplastic lesions are lost. Recently, linkage analysis has shown that at least four loci modify mammary tumorigenesis in the Cop rat, but the genes have yet to be cloned (Shepel et al., 1998).

Wang et al. (1991) showed that tumors can be induced in Cop and WF rats when the Ha-ras oncogene is introduced into mammary epithelial cells in situ using a helper-free, replication-defective retrovirus as vector. The virus preparation was infused directly into each gland via the nipple allowing infection of the ductular epithelium. They showed that the kinetics of tumor development was identical in infected Cop and WF rats. Carcinomas induced in Cop rats, however, were more differentiated and less locally invasive than those in WF rats. Interestingly, the majority of the carcinomas induced in both strains were clonal, demonstrating that while activated Ha-ras initiates neoplastic transformation, progression to carcinomas likely requires additional events. It is possible that proviral insertion within a locus could provide such an event necessary for neoplastic transformation. Wang et al. did not attempt to clone viral integration sites or identify genes involved in tumorigenesis.

In an attempt to identify genes associated with the resistance of the Cop rat to mammary tumorigenesis, we decided to utilize the method of retrovirus-mediated insertional mutagenesis (Ben-David et al., 1990b; Peters, 1990). We chose to use F1 rats from a cross between Cop and susceptible WF rats, since these F1s are known to be resistant to mammary carcinogenesis (Gould et al., 1989) and would be heterozygous at the locus of a resistance or susceptibility gene. We hypothesized that insertional inactivation of a resistance gene or activation of a susceptibility gene in only one allele of the F1 rats may be sufficient to confer susceptibility to v-Ha-ras-induced mammary carcinogenesis. Thus, cloning the site of proviral integration could potentially identify a gene that confers resistance in Cop rats. Here we show that infusion of a replication-defective spleen focus-forming virus harboring an activated v-Ha-ras gene into the main mammary ducts of the F1 rats leads to the formation of adenocarcinomas. Cell lines derived from these tumors are clonal, but do not share a common viral integration site suggesting that a high level of v-Ha-ras expression is able to overcome resistance in the F1 rats. To examine this possibility, we tested the capacity of the cell lines and subclones derived from them for anchorage-independent growth. Some subclones derived from the lines grew well in soft agar while others were unable to grow. These growth characteristics, however, were independent of v-Ha-ras expression levels. The ability of the subclones to grow in soft agar was associated with the size of tumors induced by the cells in nude mice and showed a striking positive association with cyclin D1 expression. This result is consistent with the known ability of exogenous cyclin D1 to stimulate anchorage-independent growth in a breast cancer cell line (Zhou et al., 2000). Interestingly, while resistance to anchorage-independent growth was fully overcome by transfection of cyclin D1 in some clones, in others the effect was partial. Previous studies have demonstrated that stimulation of the Wnt pathway activates -catenin that subsequently upregulates cyclin D1 (Willert and Nusse, 1998). Accordingly, we observed a similar pattern of growth in soft agar as well as upregulation of cyclin D1 following transfection of activated -catenin in these cell lines. The inability of cyclin D1 to fully overcome resistance to anchorage-dependent growth in some cell lines suggests that a resistance gene may block the activity of v-Ha-ras to potentiate growth in soft agar. We relate these findings to the resistance and susceptibility of Cop and WF rats, respectively, to mammary tumorigenesis.

Results

Induction of mammary carcinomas by a retrovirus harboring v-Ha-ras

Infusion of a replication-defective spleen focus-forming virus containing the v-Ha-ras oncogene into the main mammary ducts of F1 (Cop WF) rats produced 51 mammary tumors from 108 successfully infused glands in 15 rats. Tumors first appeared 4–5 weeks following infusion and were harvested over the following 20 weeks when they had grown to about 20 mm in diameter. Histological analysis showed that all of the tumors were typical adenocarcinomas (Russo and Russo, 2000). We randomly selected 17 tumors from seven rats for the production of primary cultures. We obtained cell lines from all of these tumors after about 1 week in culture. Morphological characteristics of the cell lines are illustrated in Figure 1. Most had epithelial morphology (e.g. Figure 1, FE1.3), but some were a mixture of epithelial-like cells with more elongated cells (e.g. Figure 1, FE1.2).

Figure 1.

Epithelial origin of breast cancer cell lines. Morphology of representative tumor-derived cell lines (FE1.2 and FE1.3) was determined by phase-contrast microscopy (magnification: 250). Immunohistochemical analysis was performed in representative subclones from tumor-derived cell lines using antibodies to keratin and vimentin (magnification: 400)

Full figure and legend (397K)

Clonal origin of tumors, cell lines and subclones

To examine the pattern of proviral integration in these cell lines, genomic DNA samples were digested with EcoRI and hybridized with a v-Ha-ras probe. As shown in Figure 2a, all 17 cell lines contained 1–3 integrated proviruses. Similar patterns of proviral integration were also seen in the original tumors used to establish the cell lines (data not shown). The limited number of integrated proviruses suggests that despite the obvious morphological heterogeneity of at least some of the cell lines, they are clonal. To support this notion, we isolated 10 subclones from the cell line FE1.2 and four subclones from the FE1.3 cell line. As described for a number of other subclones derived from chemically induced rat mammary carcinomas (Teodori et al., 2000), the FE1.2 subclones stained positively for vimentin and weakly or negatively for keratin (Figure 1). Some of the subclones were small, round cells with a high nuclear-cytoplasmic ratio (e.g. Figure 1, FE1.2-C4), some were spindle-shaped with a small amount of cytoplasm (e.g. Figure 1, FE1.2-C1), while others were large epithelial-like cells with an abundant cytoskeletal network throughout the cytoplasm (e.g. Figure 1, 2C5). The subclones from FE1.3, like the parental line, had a uniform epithelial-like morphology, but showed a patchy expression of both keratin and vimentin (examples shown in Figure 1). Despite these differences in morphology, Southern blot analysis showed that the subclones isolated from FE1.2 and FE1.3 contain a similar pattern of provirus integration indicating their clonal origin (Figure 2b and c).

Figure 2.

Clonal origin of breast cancer cell lines. (a) Southern blots of DNA isolated from tumor-derived cell lines digested with EcoRI and hybridized with a Ha-ras-specific probe. (b and c) Southern blots of DNA isolated from subclones derived from tumor-derived cell lines FE1.2 (b) and FE1.3 (c) digested with EcoRI and hybridized with a virus-specific probe. In a–c arrows indicate endogenous ras genes

Full figure and legend (171K)

The clonal origin of the carcinomas induced by v-Ha-ras suggested that transformation might require the acquisition of other events, likely through retrovirus-mediated insertional mutagenesis. To examine this possibility, we isolated the single integrated provirus and its junction fragment by screening a genomic library generated from the FA4.4 cell line as described elsewhere (Ben-David et al., 1990a). When the junction fragment was hybridized with genomic DNA prepared from the 17 cell lines, the expected rearrangement because of proviral insertion was only seen in FA4.4 cells. This suggests that the isolated integration site from FA4.4 cells is not a common site for proviral integration. Moreover, analysis of the genomic DNA surrounding this integration site did not reveal the presence of a transcriptional unit. Thus, proviral insertional mutagenesis is unlikely to be involved in the induction of these carcinomas.

Variation in growth in soft agar between cell lines

Although proviral insertion does not appear to cooperate with v-Ha-ras in neoplastic transformation, molecular changes acquired through other mechanisms could complement the activity of ras to lead to tumor formation in the rats. It is possible that a high level of v-Ha-ras expression is able to overcome the suppressor activity of a resistance gene in the F1 rats. We decided to investigate this possibility by first determining the growth properties of our independently derived cell lines. It is well established that expression of v-Ha-ras in epithelial cells can overcome anchorage-dependent growth by inhibiting suspension-induced apoptosis (anoikis) (Rak et al., 1995; Rosen et al., 2000). As shown in Table 1, however, some of the cell lines grew in soft agar while others did not. Interestingly, the ability to grow in soft agar was not associated with the level of v-Ha-ras expression (Table 1).

Table 1 - Three-dimensional growth properties and extent of v-Ha-ras expression of cell lines derived from primary tumors induced in F1 rats by retroviral infection.

Full table

We next examined the ability of the subclones to grow in soft agar. To our surprise, three of the clones derived from the primary line FE1.2 grew extremely well in soft agar, six moderately well and one clone, FE1.2-C5, was unable to grow (Figure 3a). None of the four subclones of FE1.3 (clones C1–C4) was able to grow in soft agar (data not shown). While a somewhat lower level of v-Ha-ras expression was associated with anchorage dependence in FE1.2-C5, high levels of v-Ha-ras were expressed in subclones of FE1.3 that do not grow in soft agar (Figure 3c). We showed that selected FE1.2 and FE1.3 subclones were able to generate tumors in nude mice. The clones that gave rise to the greatest number of clones in soft agar also produced the largest tumors (illustrated in Figure 3b for FE1.2 subclones). The clones that did not grow in soft agar (e.g. FE1.2-C5) gave rise to extremely small tumors.

Figure 3.

Ability of subclones to grow in soft agar correlates with their capacity to generate tumors in vivo. (a) Ability of subclones isolated from tumor-derived cell line FE1.2 to grow in soft agar. (b) Examples of subclone growth in nude mice. Cells (1 10⁶) were injected into the mammary pads of three nude mice. They were observed for tumor development and the size of tumors scored 3 weeks postinoculation (c) Western blot of v-Ha-ras expression in subclones. IEC-18, an intestinal epithelial cell line, is included as a negative control. The expression of ERK is included as a loading control

Full figure and legend (121K)

Anchorage-independent growth and cyclin D1 expression

Since we have previously shown that cyclin D1 is frequently overexpressed in preneoplastic mammary lesions in WF rats but not in Cop lesions (Korkola and Archer, 1999), we examined cyclin D1 expression in representative sample tumors induced in the F1 rats by the retrovirus. All of these tumors overexpressed cyclin D1 but to varying extents (Figure 4a). We next examined the level of cyclin D1 expression in the tumor-derived subclones described above. The ability of the subclones to grow in soft agar showed a striking association with the level of cyclin D1 (Figure 4b). For example, low levels of cyclin D1 were detected in FE1.2-C5 and the four subclones of FE1.3 (FE1.3-C1–C4) that did not grow in soft agar, while significantly higher levels of cyclin D1 were detected in the other subclones that grew well in soft agar. A negligible level of cyclin D1 was seen in the primary epithelial cell line IEC-18 that lacks anchorage- independent growth and was used as a control (Quaroni and Isselbacher, 1981). The high levels of cyclin D1 in nine of the 10 FE1.2 subclones and the low levels in the FE1.3 subclones correspond to the cyclin D1 levels in the tumors from which they were derived (Figure 4a and b).

Figure 4.

Expression of cyclin D1 in tumors and subclones. Western blots of total cellular extracts from (a) a sample of tumors (FA2.1T–FE1.3T) induced in F1 rats by retroviral infusion (subclones FE1.2C1 and FE1.2C5 are included for comparison) and (b) subclones derived from cell lines FE1.2 and FE1.3. Blots were hybridized with antibody to cyclin D1. Expression patterns of -actin are included as loading controls

Full figure and legend (110K)

In order to provide evidence that cyclin D1 is causally involved in the anchorage-independent growth of our subclones, we infected FE1.2-C5 and FE1.3-C3 that expressed low levels of cyclin D1 and did not grow in soft agar, with the retroviral expression vector pMV7plCCND1. This vector contains the entire coding sequence of the human cyclin D1 gene and has been shown to lead to stable overexpression of cyclin D1 in both human and rat cells (Han et al., 1995; Morin et al., 1997). As shown in Figure 5a, retroviral infection led to levels of cyclin D1 in these pooled subclones (FE1.2-C5D1 and FE1.3-C3D1) that were similar to the levels detected in subclone FE1.2-C1 that expressed high levels of cyclin D1 and grew well in soft agar. Pooled cells infected with the vector that lacked the cyclin D1 gene (pMV7pl) served as additional controls (FE1.2-C5V and FE1.3-C3V). It is clear from Figure 5d that expression of cyclin D1 in FE1.2-C5 and FE1.3-C3, enabled them to grow in an anchorage-independent manner, although FE1.2-C5D1 had a higher colony forming efficiency and growth rate than FE1.3-C3D1.

Figure 5.

Effect of overexpression of cyclin D1 and -catenin in breast cancer cell lines on their ability to grow in soft agar. (a) Western blots of cyclin D1 in parental subclones FE1.2-C5 and FE1.3-C3 and pooled cells isolated from these subclones following infection with a retrovirus containing cyclin D1 (FE1.2-C5D1 and FE1.3-C3D1) or the retrovirus lacking cyclin D1 (FE1.2-C5V and FE1.3-C3V). (b) Western blot of cyclin D1 expression in subclones of FE1.2-C5 transfected with an activated form of -catenin (FE1.2-C5Ca1–Ca5) or pooled cells transfected with the vector alone (FE1.2-C5V). (c) Western blots of cyclin D1 expression in the subclone FE1.3-C3 transfected with an activated form of -catenin (FE1.3-C3Ca1–Ca5) or pooled cells transfected vector alone (FE1.3C3V). In (a–c), FE1.2-C1 is a positive control that overexpresses cyclin D1. For all Western blots, a -actin antibody was used as a loading control. Panel d shows colony forming ability in soft agar of cells transfected with cyclin D1 or the activated form of -catenin and their controls as described in panels a–c; bars represent meanss.e.m. for three determinations

Full figure and legend (113K)

Exogenous expression of an activated mutant form of b-catenin in the subclones

Previous studies have indicated that cyclin D1 is downstream of the Wnt pathway and its expression is regulated by -catenin (Willert and Nusse, 1998). In order to determine whether overexpression of cyclin D1 in our subclones is driven by activated -catenin, we transfected subclones FE1.2-C5 and FE1.3-C3 with a dominant-active mutant form of this gene (pCl-neo--cat, S33Y mutant) (Morin et al., 1997). We isolated five individual transfected subclones from each parental subclone (FE1.2-C5Ca1-5 and FE1.3-C3Ca1-5). It is clear from Figure 5b and c that in each case, expression of the activated -catenin led to overexpression of cyclin D1. Pooled cells transfected with the vector only (FE1.2C5V and FE1.3C3V) served as controls. With one exception, the subclones transfected with the activated -catenin grew in soft agar (Figure 5d). As before, the subclones derived from FE1.2-C5 grew more extensively than those derived from FE1.3-C3. The exception, FE1.3-C3Ca1, can be seen to express somewhat lower levels of cyclin D1 than the other transfected subclones and did not grow in soft agar (Figure 5d).

We next examined the cellular localization of -catenin in these subclones. Subclone FE1.2-C1 that grows well in soft agar and overexpresses cyclin D1 (Figure 3a and Figure 4b), clearly shows nuclear staining of -catenin (Figure 6a). FE1.2-C3 that also grows in soft agar and expresses cyclin D1, but not to the same extent as FE12-C1, shows mostly nuclear but some membrane staining (Figure 6b). The two subclones that do not grow in soft agar and express low levels of cyclin D1 (FE1.2-C5 and FE1.3-C3) show mainly membrane localization of -catenin with some cytoplasmic staining (Figure 6c and d). These same subclones transfected with cyclin D1 did not change their localization of -catenin (Figure 6e and f), although FE1.2-C5 showed a change in morphology, with cells becoming more rounded. However, when FE1.2-C5 and FE1.3-C3 were transfected with -catenin (to yield the five individual subclones FE1.2-C5Ca1-5 and FE1.3-C3Ca1-5 described above and shown in Figure 5b–d), there was a similar marked change in the localization of -catenin. FE1.2-C5Ca3, for example, showed intense nuclear staining (Figure 6g) and resembled FE1.2-C1 (Figure 6a) rather than the parental FE1.2-C5 clone (Figure 6c). FE1.3-C3Ca3 showed some nuclear staining together with staining of the membrane and cytoplasm (Figure 6h) resembling FE1.2-C3 (Figure 6b) rather than the parental FE1.3-C3 (Figure 6d). These results indicate that the ability of the cell lines to localize -catenin in the nucleus is critical for cyclin D1 upregulation and anchorage-independent growth.

Figure 6.

Immunohistochemical staining for -catenin in various subclones. (a and b) FE1.2-C1 and FE1.2-C3, respectively – both of these subclones overexpress cyclin D1 and grow in soft agar and are shown as positive controls. (c and d) Parental subclones FE1.2-C5 and E1.3-C3, respectively – both of these subclones weakly express cyclin D1 and do not grow in soft agar. (e and f) FE1.2-C5D1 and FE1.3-C3D1 are pooled cells isolated from the parental subclones following infection with retrovirus containing cyclin D1. (g and h) FE1.2-C5Ca3 and FE1.3-C3Ca3 were prepared by transfecting the parental subclones with the -catenin plasmid. The bars are 5 m

Full figure and legend (448K)

A recent study has shown that c-Jun is directly involved in transcriptional control of cyclin D1 (Wisdom et al., 1999). We have shown a similar level of Jun-terminal kinase (JNK) activity in FE1.2 and FE1.3 cells that expresses high and low levels of cyclin D1, respectively (data not shown). This result excludes this pathway in the regulation of cyclin D1 in our cell lines.

Somatic hybridization of FE1.3 with FE1.2 cells

Our results show that v-Ha-ras is unable to induce cyclin D1 expression and overcome anchorage-dependent growth in some of the breast cancer cell lines. The mechanism for this is unclear, but expression of a dominant acting oncogene or presence of a tumor suppressor gene may be involved. To begin to address this issue, we constructed somatic cell hybrids between FE1.2-C1 cells (high cyclin D1 with ability for anchorage-independent growth) and FE1.3-C3 (low cyclin D1 with no ability for anchorage-independent growth). The hybrid cell line displayed an epithelial-like morphology identical to FE1.3 cells (data not shown). Cyclin D1 was moderately upregulated and the level of v-Ha-ras was unchanged in the hybrid cells compared to the parental cells (Figure 7a). Interestingly, some of the hybrid cells formed small colonies of 20–30 cells in soft agar. These cells, however, developed the morphological characteristics of apoptosis and eventually died (data not shown). Our results indicate that the ability to upregulate cyclin D1 and overcome anchorage-dependent growth may be regulated by the expression of a tumor suppressor gene.

Figure 7.

Somatic hybridization of the breast cancer cell lines and their effect on cyclin D1 expression. Hybrid cells were isolated from fusion of E1.2-C1 and E1.3-C3 cells expressing either puromycin or neomycin resistance genes, respectively. Extract from double drug-resistant cells or the parental cells was used for cyclin D1 expression using Western blot analysis. The level of -actin was used as a loading control

Full figure and legend (107K)

Discussion

To identify genes associated with the resistance of Cop rats to mammary carcinogenesis, we utilized the method of retrovirus-mediated insertional mutagenesis. This method has been successfully applied to identify oncogenes and tumor suppressor genes associated with the induction and progression of a number of types of virally induced cancers (Ben-David et al., 1990b; Peters, 1990). We infused a replication-defective, spleen focus-forming virus harboring a v-Ha-ras gene directly into the main mammary ducts of Cop WF F1 rats for reasons outlined previously. These infections led to the formation of adenocarcinomas in approximately 50% of successfully infused glands.

From the tumors, we were able to generate a number of independently derived cell lines. We also derived a number of subclones from two of our primary cell lines by limiting dilution. Previous studies have demonstrated that cell lines derived from chemically induced rat mammary carcinomas are often heterogeneous with populations of epithelial and myoepithelial cells (Teodori et al., 2000). Furthermore, mammary tumor epithelial cells have the potential to dedifferentiate rapidly into cells with stromal-like features (Rudland et al., 1986). In keeping with these observations, most of our cell lines and subclones had epithelial morphology (Figure 1, FE1.3), but a number of them contained a mixture of polygonal epithelial-like cells with more elongated cells, possibly myoepithelial or fibroblast-like cells (Figure 1, FE1.2). The clonal origin of the cell lines and subclones suggests that they were derived from a primary transformed cell and that there were conversions from epithelial to myoepithelial/fibroblastoid cell types.

The limited number (1–3) of proviral integration sites in the tumors revealed by Southern blotting indicated that they were clonal. Our cell lines and subclones were also clonal, despite some of them having mixed morphological characteristics. The clonal origin of the tumors suggested that retroviral insertional mutagenesis may play a role in their induction. Cloning the junction fragments from one of the cell lines, however, showed there was not a common site for proviral integration. This result suggests that a mechanism in which proviral insertion inactivates a resistance gene in the F1 rats does not seem to be involved in the induction of these tumors. We reasoned that it is likely instead that the high level of v-Ha-ras expression in mammary cells is able to overcome the suppressor activity of a resistance gene in the F1 rats.

To investigate how over expression of v-Ha-ras could overcome resistance in the F1 rats, we determined the growth properties of our cell lines and subclones derived from them. Growth in soft agar requires inhibition of suspension-induced apoptosis (anoikis), a process that plays a central role in carcinogenesis (Frisch and Francis, 1994). Several groups have demonstrated that the expression of activated Ha-ras in epithelial cells of the intestine and other organs induces resistance to anoikis (Frisch and Francis, 1994; Rak et al., 1995; Rosen et al., 2000). Therefore, we examined the capacity of the primary cell lines we derived from the v-Ha-ras-induced tumors for anchorage-independent growth. Some of the cell lines were able to grow in soft agar, but a significant number did not display anchorage-independent growth despite originating from adenocarcinomas. Furthermore, some of the subclones that we showed were clonally related to their corresponding parental cell lines grew extremely well in soft agar, some moderately well, but a significant number did not display anchorage-independent growth. The growth properties were reproduced in nude mice – those cells that grew well in soft agar grew rapidly as tumors, while those that did not grow in soft agar grew extremely slowly.

We next showed that surprisingly, the ability of our cell lines and subclones to grow in soft agar was not associated with the level of v-Ha-ras expression. For example, cell line FE1.3 and its subclones, that were all unable to grow in soft agar, expressed high levels of v-Ha-ras. FE1.2 and nine of the 10 subclones we prepared from this cell line were able to grow, albeit to varying extents, in soft agar and they also expressed v-Ha-ras. The growth of one FE1.2 subclone, FE1.2-C5, however, was anchorage-dependent and it expressed only low levels of ras. These observations suggest that the ability of v-Ha-ras to confer anchorage independence may depend upon the genetic background of the cell lines. Since the lines were derived from heterozygous F1 rats that are resistant to mammary carcinogenesis (Gould et al., 1989), our results raise the possibility that the ability of v-Ha-ras to confer anchorage-independent growth may require the loss of a tumor suppressor gene that controls cell growth both in vivo and in vitro. Presence of such a tumor suppressor gene would also explain our observation that somatic fusion of the anchorage-dependent clone FE1.3-C3 to the anchorage-independent clone FE1.2-C1 did not induce growth in soft agar, although both of these clones express similar levels of v-Ha-ras. It is also possible that the presence of this gene in Cop rats confers resistance to carcinogenesis.

Since the FE1.3 cell line is derived from a mammary carcinoma, it is unclear why these cells failed to grow in soft agar and instead underwent apoptosis. It is possible that in FE1.3 cells growing in vivo, apopotosis is partially circumvented by protective environmental factors secreted from stromal cells surrounding the tumor cells. These factors, however, are not present in soft agar and apoptosis is not blocked. The presence of such factors could increase the survival of cancer cells resulting in the generation of slow growing tumors. This hypothesis is supported by our observation that injection of FE1.3 cells into nude mice induced much smaller tumors than anchorage-independent cell lines that may have lost the tumor suppressor gene.

Cyclin D1 expression has previously been associated with breast cancer development (Zhou et al., 2001) and indeed, a representative sample of the tumors induced using the retrovirus expressed moderate to high levels of cyclin D1. Overexpression of cyclin D1 has also been shown to significantly stimulate anchorage-independent colonization of a human mammary epithelial cell line in soft agar or methylcellulose (Zhou et al., 2000). Interestingly, the ability of our sublones to grow in soft agar showed a striking positive association with the expression of cyclin D1. Furthermore, subclones that were resistant to anchorage-independent growth completely or partially lost this resistance following transfection of cyclin D1. For example, when transfected with cyclin D1, FE1.2-C5 showed robust growth in soft agar, but FE1.3-C3 formed fewer, smaller anchorage-independent colonies. Moreover, somatic cell hybridization that led to moderate increases in the levels of cyclin D1, also led to a temporary formation of small colonies in soft agar compared to the anchorage-dependent parental clone FE1.3. Our results suggest that while a high expression level of cyclin D1 is necessary for anchorage-independent growth in all clones, it is not sufficient to confer full growth capacity in soft agar. We hypothesize that the presence of a tumor suppressor gene in FE1.3 cells suppresses their capacity for full anchorage-independent growth. This tumor suppressor gene may also control the pathway that leads to cyclin D1 upregulation by v-Ha-ras in transformed epithelial cells (Filmus et al., 1994). Furthermore, loss of this tumor suppressor gene may promote the growth of preneoplastic cells in vivo. We have previously demonstrated that a lack of cyclin D1 expression in preneoplastic lesions in Cop rats correlated with the inability of these lesions to progress to cancer (Korkola and Archer, 1999). Our present observations that a lack of cyclin D1 expression in several tumor-derived cell lines plays a role in their inability to grow in soft agar suggest that these cell lines will be useful models of resistance/susceptibility to mammary tumorigenesis in vivo as well as excellent models to study the mechanism of anoikis in breast cancer cells in vitro.

Activation of the Wnt signaling pathway has been shown to lead to accumulation of -catenin in the cytoplasm and its subsequent translocation to the nucleus where it associates with members of the T-cell factor/lymphoid enhancer factor (TCF/LEF) transcription factor family (Miller and Moon, 1996; Willert and Nusse, 1998) to induce transcription and expression of important growth regulatory genes including cyclin D1 (Behrens et al., 1996; Shtutman et al., 1999; Tetsu and McCormick, 1999). Expression of Ha-ras in epidermal keratinocytes has previously been shown to promote cytoplasmic and nuclear accumulation of -catenin (Espada et al., 1999). Recent studies have identified cyclin D1 as a target gene of -catenin in breast cancer (Lin et al., 2000). Furthermore, transgenic mice with the MMTV-LTR driving an activated form of -catenin develop mammary gland hyperplasia and mammary adenocarcinoma (Michaelson and Leder, 2001). It was of importance, therefore, to examine whether -catenin activation is involved in upregulation of cyclin D1 in our cell lines. In our clones that did not grow in soft agar and expressed low levels of cyclin D1, transfection of -catenin led to a marked upregulation of cyclin D1. The cells overexpressing -catenin also acquired the ability to grow to the same extent in soft agar as cells transfected with cyclin D1. Furthermore, the extent of nuclear localization of -catenin in our subclones correlated both with their level of expression of cyclin D1 and ability to grow in soft agar. Interestingly, nuclear localization of -catenin was not seen in an FE1-3 subclone transfected with cyclin D1 that expressed high levels of v-Ha-ras, but grew poorly in soft agar. These results suggest that a tumor suppressor gene in these cells likely inhibits cyclin D1 upregulation through the -catenin pathway

Translocation of -catenin to the nucleus with activation of cyclin D1 transcription may be caused by several mechanisms including mutation of -catenin, deletion of the APC gene or activation of the Wnt pathway (Polakis, 1999). Translocation of -catenin was seen in subclone FE1.2-C1 that grew in an anchorage-independent manner and highly expressed v-Ha-ras. However, this translocation was not seen in the clonally related FE1.2-C5 that neither grew in soft agar nor expressed high levels of v-Ha-ras. This suggests that in FE1.2 subclones, -catenin is activated by v-Ha-ras, and excludes the possibility that an activating mutation is acquired within -catenin in anchorage-independent clones that express high levels of cyclin D1. Since translocation of -catenin was not detected in FE1.3 subclones that express high levels of v-Ha-ras, it is possible that other mutational events upstream of -catenin that are present in FE1.2 cells are necessary to allow its activation by v-Ha-ras. Such events may also play a role in conferring resistance to carcinogenesis in the Cop mammary gland.

In summary, infusion of a retrovirus harboring v-Ha-ras into the main mammary ducts of Cop WF F1 rats led to the formation of mammary adenocarcinomas. We established several cell lines and subclones from these tumors. Despite expressing high levels of v-Ha-ras, some of the cell lines and their subclones were unable to grow in soft agar or form tumors in nude mice. Anchorage-independent growth induced by v-Ha-ras required the expression of cyclin D1 that in turn was regulated by the -catenin pathway. Presence of a tumor suppressor gene in some cells may have blocked their full capacity for growth in soft agar. The resemblance of the growth behavior of our cell lines to the resistance and susceptibility of Cop and WF rats, respectively, to mammary tumorigenesis, may facilitate the identification of breast cancer susceptibility genes as well as the mechanism of anoikis in breast cancer cells in vitro.

Materials and methods

Tumor induction

A replication-defective retrovirus containing v-Ha-ras was generated by cloning v-Ha-ras cDNA into a unique BamHI site in the pSFF retrovirus (Bestwick et al., 1985) The virus was packaged by transfecting the plasmid into GP+E helper-free packaging cells (Markowitz et al., 1988). Helper-free recombinant virus was harvested from these cells and concentrated by ultracentrifugation. To induce tumors in F1 (Cop WF) rats, we followed the procedures of Wang et al. (1991). Briefly, 2 days prior to retroviral infection, the rats (n=15) were given daily s.c. doses of 3 mg/kg perphenazine (Sigma Chemical Co., St Louis, MO, USA). On the day of viral treatment, the animals were given a third dose of perphenazine. After 4 h, the rats were anesthetized and 15 l of a stock of virus solution containing 8 mg/ml polybrene and 2 g/ml indigo carmine was infused directly into the mammary ducts via the nipples. The dye enabled us to ensure that these infusions entered the main ducts and were not subcutaneous. The rats were palpated weekly starting 2 weeks post-infusion. Tumors were harvested when they were >20 mm in diameter or when the animals appeared moribund. A portion of each tumor was fixed in formalin for histological analysis.

Preparation of cell cultures

A small piece of each tumor was digested with 200 U/ml type III collagenase (Worthington Biochemical Co., Lakewood, NJ, USA) for 2 h then introduced into Primeria flasks with -MEM supplemented with 10% FBS, 10 ng/ml EGF, 1 g/ml hydrocortisone and 1 ng/ml 17-estradiol. Subcloning was performed by limiting dilution. Briefly, suspensions of single tumor cell lines were diluted to five cells per ml and seeded into 96-well plates (0.1 ml/well) using the medium described above. After 24 h, wells containing one cell were identified and colonies were allowed to grow for 1 week before being transferred to another 24-well plate. At confluence, clones were transferred to 25 cm² flasks.

Molecular cloning

High molecular weight DNA from a cell line we designated FA4.4 was partially digested with EcoRI and ligated to EcoRI arms of the bacteriophage vector EMBL-4 (Stratagene, La Jolla, CA, USA), as described by the company. Phage plaques were screened with a ³²P-labeled v-Ha-ras probe. After three cycles of phage purification, the inserted fragments were subcloned into plasmid pGEM-7(+) (Promega, Madison, WI, USA).

Tumor DNA and molecular hybridization

High molecular weight DNA was isolated from cell lines and tumor samples, as described (Korkola et al., 1999). DNA was digested with restriction enzymes and electrophoresed on agarose gels. The DNA was acid-depurinated before denaturation and transferred to nitrocellulose. The filters were hybridized with a v-Ha-ras probe, as described elsewhere (Bani et al., 1996).

Infection and transfection experiments

The retroviral expression vector pMV7plCCND1 that contains the entire 1.1 kb coding sequence of human cyclin D1 and the control vector pMV7pl were generously supplied by Dr IB Weinstein, Columbia University (Jiang et al., 1993; Han et al., 1995). Replication-defective viruses were prepared by transfecting the viral vectors into the helper-free packaging cell line GP+E (Markowitz et al., 1988), as described previously (Bani et al., 1996). For viral infection, supernatants from the virus-producing cells were used to infect cells plated at a density of 2 10⁵, and after 48 h incubation G418 resistant cells were pooled and used for experiments. Subclones FE1.2-C5 and FE1.3-C3 were transfected with the plasmid pCl-neo--cat, clone S33Y that contains a dominant-active mutant form of -catenin (generously supplied by Dr B Vogelstein) (Morin et al., 1997). This plamid in which the -catenin gene had been removed was used as the vector control. Transfection was performed using the lipofectamine, according to the manufacturer's protocol (Gibco, Burlington, ON, Canada). Clones expressing -catenin that were resistant to G418 were isolated. Cells transfected with the vector alone were pooled and used as controls.

Growth of cells in soft agar and in nude mice

To monitor growth of cells in soft agar, two layers of agarose were used. The bottom layer contained 0.5% agarose, the top layer 0.3% agarose both in -MEM. Triplicate samples of 5000 cells were seeded on 6 cm plates. FE1.2 clones and subclones were incubated for 5–7 days, while those from FE1.3 were incubated for 14 days. At the end of the incubation period, colonies with >25 cells were counted. Female athymic nude mice (CD1 – Nu/Nu), age 7 weeks, were purchased from Charles River Laboratories (Wilmington, MA, USA) and housed in microisolator cages with sterile bedding. Mice were handled in a unidirectional laminar airflow hood. The tumor cell inoculation procedure has been described in detail elsewhere (Price et al., 1990). Briefly, at 8 weeks of age under ketamine/xylazine anesthesia mice (n=3) received a 1–2 mm incision exposing the right thoracic mammary fat pad. Cells (1 10⁶) were injected in a volume of 50 l and the wound was closed with tissue glue. Tumors were palpated and measured weekly with vernier calipers by a single, blinded observer.

Generation of somatic cell hybrids

To generate somatic cell hybrids, subclones FE1.2-C1 and FE1.3-C3 were infected with MSCV retroviruses carrying genes for puromycin or neomycin, respectively (Bani et al., 1996). Hybrids were prepared by seeding a mixture of the two cell lines (5 10⁵ cells each) in 60-mm dishes 24 h prior to initiation of the cell fusion (Bani et al., 1996). At the time of fusion, cells were washed once with PBS and then treated with 50% polyethylene glycol (PEG) as a cell fusion agent (M_r 4000; Invitrogen, Carlsbad, CA, USA) for 1 min. After removal of PEG by several washes with PBS, fresh medium was added, and the next day cells were harvested and replated at a density of 1 10⁵ cells/100 mm dish in a double selective medium containing 800 g/ml neomycin and 1 g/ml puromycin in -MEM. The resulting doubly resistant cells were pooled and analysed for growth in soft agar. They were also used for Western blot analysis.

Western blotting and immunohistochemistry

Frozen primary tumors were ground to a powder in liquid nitrogen. Western blotting was performed by lysing tumor powder or cells in ice-cold RIPA lysis buffer containing 0.5% NP40, 50 mM Tris-HCl (pH 8.0, 120 mM NaCl, 50 mM NaF, 10 g/ml aprotinin, 100 g/ml leupeptin, and 10 mM phenylmethylsulfonyl fluoride). Lysates were sonicated and clarified by centrifugation. Protein concentrations were determined by the Bradford assay (Bio-Rad, Mississauga, ON, Canada) and 20 g of protein/lane was resolved by SDS/PAGE (10% gels run at 140 V for approx 1 h). After electrophoresis, proteins were transferred onto PVDF membranes (Millipore, Nepean, ON, Canada). Blots were blocked for 30 min at room temperature with 10% fat-free dry milk in PBS-T then reacted with appropriate primary and secondary antibodies and proteins were detected using enhanced chemiluminescence (Amersham, Little Chalfont, Bucks, UK). Primary murine monoclonal antibodies were obtained from the following sources: anti--catenin and anti-Pan-ERK from Transduction Laboratories (Mississauga, ON, Canada) used at dilutions of 1/1000 and 1/3000, respectively; anti-pan-Ras (Ab2) from Oncogene Research Products (San Diego, CA, USA) used at a dilution of 1/100; anti-cyclin D1 from Neomarkers (Fremont, CA, USA) used at a dilution of 1/500; anti--actin from Sigma used at a dilution of 1/50 000. For immunohistochemistry with keratin and vimentin, cells were seeded on sterile cover slips. At 70–80% confluence, they were washed for 3 5 min in PBS, fixed in cold methanol for 10 min, air-dried and stored at -20°C. Following blocking with 1% BSA, cells were incubated overnight at 4°C with antibodies to keratin (Lab Vision, Fremont, CA, USA) or vimentin (Boehringer Mannheim, Germany) at dilutions of 1/50 and 1/3, respectively. After 2 5 min washes in PBS, cells were incubated with secondary antibody (LSAP kit, Dako, Carpinteria, CA, USA) for 30 min followed by 2 5 min washes in PBS and incubation with streptavidin–horseradish peroxidase complex for 30 min. The cells were finally incubated for 5–10 min with the chromagen DAB (Dako) and counterstained with hematoxylin. For immunohistochemistry with -catenin, cells were fixed in 10% paraformaldehyde for 30 min at 4°C and blocked with donkey serum for 30 min at room temperature. Slides were then incubated with the anti--catenin antibody (1/250) for 60 min at room temperature. FITC goat anti-mouse IgG (Jackson Research Laboratories, Bar Harbor, ME, USA) was used as the secondary antibody at a dilution of 1/200. The slides were mounted using GVA solution (Zymed, San Francisco, CA, USA) and photographed using a fluorescence microscope.

References

1. Bani MR, Rak J, Adachi D, Wiltshire R, Trent JM, Kerbel RS and Ben-David Y. (1996). Cancer Res., 56, 3075–3086. | PubMed | ISI | ChemPort |
2. Behrens J, von Kries JP, Kuhl M, Bruhn L, Wedlich D, Grosschedl R and Birchmeier W. (1996). Nature, 382, 638–642. | Article | PubMed | ISI | ChemPort |
3. Ben-David Y, Giddens EB and Bernstein A. (1990a). Proc. Natl. Acad. Sci. USA, 87, 1332–1336. | Article | PubMed | ChemPort |
4. Ben-David Y, Lavigueur A, Cheong GY and Bernstein A. (1990b). N. Biol., 2, 1015–1023.
5. Bestwick RK, Hankins WD and Kabat D. (1985). J. Virol., 56, 660–664.  | PubMed | ChemPort |
6. Espada J, Perez-Moreno M, Braga VM, Rodriguez-Viciana P and Cano A. (1999). J. Cell Biol., 146, 967–980. | Article | PubMed | ISI | ChemPort |
7. Filmus J, Robles AI, Shi W, Wong MJ, Colombo LL and Conti CJ. (1994). Oncogene, 9, 3627–3633. | PubMed | ISI | ChemPort |
8. Frisch SM and Francis H. (1994). J. Cell Biol., 124, 619–626. | Article | PubMed | ISI | ChemPort |
9. Gould MN, Wang B and Moore CJ. (1989). Genes and Signal Transduction in Multistage Carcinogenesis, Colbrun NH (ed). Marcel Dekker: New York, pp. 19–38.
10. Han EK, Sgambato A, Jiang W, Zhang YJ, Santella RM, Doki Y, Cacace AM, Schieren I and Weinstein IB. (1995). Oncogene, 10, 953–961. | PubMed | ISI | ChemPort |
11. Jiang W, Kahn SM, Zhou P, Zhang Y, Cacace AM, Infante A, Doi S, Santella RM and Weinstein IB. (1993). Oncogene, 8, 3447–3457. | PubMed | ISI | ChemPort |
12. Korkola JE and Archer MC. (1999). Carcinogenesis, 20, 221–227.
13. Korkola JE, Wood GA and Archer MC. (1999). Breast Cancer Res., 1, 88–94.
14. Lin SY, Xia W, Wang JC, Kwong KY, Spohn B, Wen Y, Pestell RG and Hung MC. (2000). Proc. Natl. Acad. Sci. USA, 97, 4262–4266. | Article | PubMed | ChemPort |
15. Markowitz D, Goff S and Bank A. (1988). J. Virol., 62, 1120–1124. | PubMed | ISI | ChemPort |
16. Michaelson JS and Leder P. (2001). Oncogene, 20, 5093–5099. | Article | PubMed | ISI | ChemPort |
17. Miller JR and Moon RT. (1996). Genes Dev., 10, 2527–2539. | Article | PubMed | ISI | ChemPort |
18. Morin PJ, Sparks AB, Korinek V, Barker N, Clevers H, Vogelstein B and Kinzler KW. (1997). Science, 275, 1787–1790. | Article | PubMed | ISI | ChemPort |
19. Peters G. (1990). Cell Growth Differ., 1, 503–510.  | PubMed | ISI | ChemPort |
20. Polakis P. (1999). Curr. Opin. Genet. Dev., 9, 15–21. | Article | PubMed | ISI | ChemPort |
21. Price JE, Polyzos A, Zhang RD and Daniels LM. (1990). Cancer Res., 50, 717–721. | PubMed | ISI | ChemPort |
22. Quaroni A and Isselbacher KJ. (1981). J. Natl. Cancer Inst., 67, 1353–1362.
23. Rak J, Mitsuhashi Y, Erdos V, Huang S-N and Kerbel RS. (1995). J. Cell Biol., 131, 1587–1598. | Article | PubMed | ISI | ChemPort |
24. Rosen K, Rak J, Leung T, Dean NM, Kerbel RS and Filmus J. (2000). J. Cell Biol., 149, 447–456. | Article | PubMed | ISI | ChemPort |
25. Rudland PS, Paterson FC, Monaghan P, Davies AC and Warburton MJ. (1986). Dev. Biol., 113, 388–405.
26. Russo J and Russo IH. (2000). J. Mammary Gland. Biol. Neoplasia, 5, 187–200. | Article | PubMed | ISI | ChemPort |
27. Shepel LA, Lan H, Haag JD, Brasic GM, Gheen ME, Simon JS, Hoff P, Newton MA and Gould MN. (1998). Genetics, 149, 289–299. | PubMed | ISI | ChemPort |
28. Shtutman M, Zhurinsky J, Simcha I, Albanese C, D'Amico M, Pestell R and Ben-Ze'ev A. (1999). Proc. Natl. Acad. Sci. USA, 96, 5522–5527. | Article | PubMed | ChemPort |
29. Sukumar S, Notario V, Martin-Zanca D and Barbacid M. (1983). Nature, 306, 658–661. | Article | PubMed | ISI | ChemPort |
30. Teodori L, Tagliaferri F, Stipa F, Valente MG, Coletti D, Manganelli A, Guglielmi M, D'Angelo LS, Schafer H and Gohde W. (2000). In Vitro Cell Dev. Biol. Anim., 36, 153–162.
31. Tetsu O and McCormick F. (1999). Nature, 398, 422–426. | Article | PubMed | ISI | ChemPort |
32. Wang B, Kennan WS, Yasukawa-Barnes J, Lindstrom MJ and Gould MN. (1991). Cancer Res., 51, 5298–5303.
33. Willert K and Nusse R. (1998). Curr. Opin. Genet. Dev., 8, 95–102. | Article | PubMed | ISI | ChemPort |
34. Wood GA, Korkola JE and Archer MC. (2002). Carcinogenesis, 23, 1–9.
35. Wisdom R, Johnson RS and Moore C. (1999). EMBO J., 18, 188–189. | Article | PubMed | ISI | ChemPort |
36. Zhou Q, Hopp T, Fuqua SAW and Steeg PS. (2001). Cancer Lett., 162, 3–17. | Article | PubMed | ISI | ChemPort |
37. Zhou Q, Wulfkuhle J, Ouatas T, Fukushima P, Stetler-Stevenson M, Miller FR and Steeg PS. (2000). Breast Cancer Res. Treat., 59, 27–39.

Acknowledgements

We thank Dr B Vogelstein for the -catenin plasmid and Dr IB Weinstein for the cyclin D1 plasmid. We also thank Drs J Filmus and Kirill Rosen for their constructive comments on the manuscript and Robin Duncan for help with the experiments with nude mice. This work was supported by a grant from the Canadian Breast Cancer Research Initiative.