Wingless, the Drosophila homolog of the proto-oncogene Wnt-1, can transform mouse mammary epithelial cells

The Wnt gene family constitutes an important group of structurally related genes implicated in regulating a wide variety of developmental processes in both vertebrates and invertebrates (for reviews see Nusse and Varmus, 1992; McMahon, 1992; Moon, 1993). The genes encode secreted proteins which are thought to act as extracellular signaling factors. At least 10 different Wnt genes are so far known in the mouse, 3 or more in Drosophila, and additional members of the family have been described in several other species (Nusse and Varmus, 1992; McMahon, 1992; Moon, 1993; Sidow, 1992; Eisenberg et al., 1992; Russell et al., 1992). The best studied members of the family are the mouse protooncogene Wnt-1 and its Drosophila homolog, the segment polarity gene wingless (Nusse and Varmus, 1982; Rijsewijk et al., 1987a). Much of our present knowledge of Wnt gene function has accrued from analysis of these two genes in disparate biological systems.

Wnt-1 was first identified as a murine oncogene activated in carcinomas induced by the mouse mammary tumor virus (MMTV), although the gene normally functions in the embryonic central nervous system (Nusse and Varmus, 1992; McMahon, 1992). In mid-gestational mouse embryos Wnt-1 is expressed in a spatially restricted pattern within the neural tube (Shackleford and Varmus, 1987; Wilkinson et al., 1987), and the phenotypes of animals homozygous for null alleles demonstrate that Wnt-1 plays an essential role in development of the fetal midbrain and cerebellum (Thomas and Capecchi, 1990; McMahon and Bradley, 1990; Thomas et al., 1991 ; McMahon et al., 1992). Wnt-1 encodes cysteine-rich secreted glycoproteins of 41-44X10³M_r, which in cell culture are found predominantly associated with the extracellular matrix (ECM) or cell surface (Brown et al., 1987; Papkoff et al., 1987; Papkoff, 1989; Bradley and Brown, 1990; Papkoff and Schryver, 1990) and which are believed to function in the manner of growth or differentiation factors (see Nusse and Varmus, 1992, for review).

Mouse mammary tumors induced by MMTV frequently contain proviral DNA insertions at the Wnt-1 locus (Nusse and Varmus, 1982) and these insertions activate expression of wild-type Wnt-1 protein within tissue in which the gene is normally silent (Nusse et al., 1984; van Ooyen et al., 1985). The consequences of this have been duplicated experimentally in transgenic mice designed to express Wnt-1 in their mammary glands: such mice rapidly develop mammary hyperplasia and subsequently carcinomas (Tsukamoto et al., 1988). In addition, the transforming potential of Wnt-1 has been demonstrated in. murine cell culture systems. Expression of Wnt-1 in the mammary epithelial cell line C57MG causes morphological transformation and an apparent loss of contact-inhibition of cell growth (Brown et al., 1986). As a result, cultures grow to higher final cell densities and fail to become quiescent at confluence. Similar effects of Whi-1 have been described in the mammary cell line RAC311 c, which can be made tumorigenic by expression of the gene (Rijsewijk et al., 1987b). In contrast, no manifest phenotypic effects of Wnt-1 have been observed in other cell lines so far tested, including several fibroblast lines, HeLa and MDCK cells (Brown et al., 1986; Jue et al., 1992; Mason et al., 1992). This cell type specificity of Wnt-1 transformation has recently been exploited to demonstrate that fibroblasts expressing Wnt-1, although not themselves phenotypically altered, are able to induce transformation of surrounding C57MG cells in co-cultures (Jue et al., 1992). This effect presumably results from secreted Wnt-1 protein acting in a paracrine manner.

Wnt-1 displays a striking degree of phylogenetic conservation and orthologs have so far been described in human, mouse, reptiles, amphibians, bony and cartilaginous fishes, echinoderms, and Drosophila (Nusse et al., 1984; van Ooyen et al., 1985; Noordermeer et al., 1989; Molven et al., 1991; Busse et al., 1990; Rijsewijk et al., 1987a; Sidow, 1992). The Drosophila ortholog of Wnt-1 is the segment polarity gene wingless, a gene required for correct pattern formation within the embryonic body segments and imaginai discs (Couso et al., 1993; Struhl and Basler, 1993; Babu, 1977; Baker, 1988a; Baker, 1987; Cabrera et al., 1987; Rijsewijk et al., 1987a; Nusslein-Volhard and Wieschaus, 1980). In the embryonic ectoderm, wingless is expressed in narrow stripes of epidermal cells at the posterior boundary of each parasegment (Baker, 1988b; Baker, 1987), and there is both genetic and immunohistochemical evidence that it functions in intercellular signaling to modulate the phenotype and developmental fate of neighboring cells (Wieschaus and Riggleman, 1987; Morata and Lawrence, 1977; Martinez-Arias et al., 1988; DiNardo et al., 1988; Riggleman et al., 1990; Dougan and DiNardo, 1992; van den Heuvel et al., 1989; Gonzalez et al., 1991). Like mouse Wnt-1, Wingless protein is secreted and may be associated with extracellular matrix or cell surfaces in the embryo (van den Heuvel et al., 1989; Gonzalez et al., 1991).

At the amino acid level, Wingless protein and mouse Wnt-1 are 54% identical (Fung et al., 1985; Rijsewijk et al., 1987a). Both are cysteine-rich proteins and all 23 cysteine residues in both proteins are found in equivalent positions (Rijsewijk et al., 1987a). Sequence conservation extends over most of the length of the predicted gene products, with two exceptions. The amino-terminal domains show no significant sequence identity, although both contain secretory signal peptides, and the predicted Wingless protein contains an internal 93 amino acid insertion not present in mouse Wnt-1 (Fung et al., 1985; Rijsewijk et al., 1987a).

Like Wnt-1, many other vertebrate proto-oncogenes display a high degree of evolutionary conservation and homologs of several such genes have been identified in Drosophila (reviewed by Shilo, 1987; Hoffmann, 1989). The ability to perform genetic analysis of these genes in Drosophila offers substantial promise of better understanding of the mechanisms by which their mammalian counterparts operate. In the case of Wnt-1 /wingless, candidate genes in the wingless signaling or response pathway have already emerged from extensive analysis of segment polarity (Ingham, 1991; Peifer and Bejsovec, 1992). However, the relevance of these studies to mammalian Wnt-1 depends in part on the extent to which the sequence homology reflects conservation in the properties of the Wnf-1 and wingless gene products. In this paper we evaluate the action of wingless in cell culture assays commonly used to study Wnt-1 function. We show that the Drosophila gene, like mouse Wnf-1, is capable of causing transformation and mitogene-sis of mouse mammary cells and of converting them to a tumorigenic phenotype. Despite the large phylogenetic distance between Diptera and mammals, our results indicate that the wingless and Wnt-1 gene products share a common mechanism of action and are functionally equivalent in these transformation and oncogenesis assays.

Clonal derivatives of the mouse mammary epithelial cell lines C57MG (Vaidya et al., 1978) and RAC311c (Rijsewijk et al., 1987b) were chosen for their homogeneous flat morphology at confluence and were maintained as previously described for C57MG (Jue et al., 1992).

To construct pMVwg, a 1.9 kb fragment of wingless cDNA (nucleotides 284-2181; Rijsewijk et al., 1987a) was cloned into the MSV-based retroviral vector plasmid pMV7 (Kirschmeier et al., 1988). pMVWnt-1, an equivalent construct containing mouse Wnt-1 cDNA, has been described previously (Jue et al., 1992). Helper-free virus stocks were produced by a two-step procedure in which pMVwg, pMVWnt-1, and pMV7 were first introduced into the ecotropic retroviral packaging cell line E86 (Markowitz et al., 1988b) and stable transfectants were selected in G418 (Brown and Scott, 1987). Virus stocks harvested from pooled populations of resistant colonies were then used to infect the amphotropic packaging cell line AM12 (Markowitz et al., 1988a), again with G418 selection. Virus was harvested both from clonal cultures of infected AM12 cells and from pooled populations. Titers were determined from neo transduction efficiencies upon infection of Rat-2 cells and selection of colonies in G418 (Brown and Scott, 1987). Absence of replication-competent helper virus was verified by assaying the neo-transducing potential of supernatants from infected Rat-2 and C57MG cells.

Focus assays were performed as described previously (Brown et al., 1986). Foci were first visible 1-2 days after the cells reached confluence and were counted and photographed 1-2 days later. Focus-forming units (FFU) were determined from duplicate infections at three different virus dilutions. Cell co-culture assays were performed as described by Jue et al. (1992).

C57MG cells infected with MV7, MVwg or MVWnt-1 were plated at a density of 7.5 x 10⁴ cells per 6 cm dish and grown to confluence over a 5 day period. Using a hematocytometer, total cell numbers in triplicate dishes were determined on three successive days, beginning on the first day that the cultures appeared confluent. Measurements of DNA synthesis were performed on cells plated in parallel with the above. On the day the cultures first reached confluence, methyl-[³H]thymidine (Amersham) was added to the medium at a final concentration of 10 μCi/ml. 24 hours later the cells were washed in PBS and fixed in 3.7% formaldehyde. The monolayers were then washed in 5% TCA, rinsed, and coated with Kodak NTB-2 photographic emulsion. After 2 days exposure, numbers of labelled nuclei per unit area were counted using an inverted microscope.

For indirect immunofluorescence, a polyclonal rat serum raised against a Wingless fusion polypeptide (Gonzalez et al., 1991) was kindly provided by A. Martinez Arias (University of Cambridge, UK). The serum was preadsorbed against methanol-fixed C57MG/MV7 cells. Cells were grown on glass coverslips, fixed for 10 minutes in methanol at -30°C, and incubated with a 1:100 dilution of preadsorbed antiserum for 1 hour at 37°C. Secondary antibody was goat anti-rat IgG conjugated to Texas Red (Jackson Immunoresearch). For staining of ECM-associated Wingless protein, cells were detached by incubating for 5 minutes at 37°C in 1 mM EDTA in PBS and the coverslips were then treated with fixative and immunostained as above. For staining of cell surface-associated Wingless protein, cells were grown on poly-D-lysine-coated coverslips and incubated with wingless antiserum at 1:100 dilution for 12 hours at 4°C prior to fixation and secondary antibody staining as described above.

For immunoblot analysis, conditioned culture medium was precleared at 1000gand proteins sedimented at 100,000 g for 3 hours in the presence of protease inhibitors as described by Jue et al. (1992). Where mentioned, 200 pg/ml heparin (from porcine intestinal mucosa, Sigma) was added to the culture medium for 48 hours. After electrophoresis and transfer to nitrocellulose (Bradley and Brown, 1990), Wingless products were detected using a 1:1500 dilution of rabbit anti-Wingless serum (a gift from R. Nusse, Stanford University) followed by goat anti-rabbit IgG conjugated to alkaline phosphatase (Promega) and were visualized as previously described (Bradley and Brown, 1990).

To express wingless and Whi-1 efficiently in mammalian cells, we used the recombinant retrovirus vectors MVvvg and MVWni-1 derived from the murine sarcoma virus based vector pMV7 (Kirschmeier et al., 1988). These vectors express wingless or VUiZ-1 cDNA from the viral long terminal repeat promoter, and the selectable marker neo from an internal thymidine kinase gene promoter. Helper-free stocks of these viruses, and of the parental vector MV7, were obtained from retroviral packaging cell lines and their neo transducing efficiency was determined as a measure of overall virus titer.

We first tested the transforming potential of wingless in mouse mammary epithelial cells by means of a focus assay in which C57MG cells were infected with MVwg or MVWhf-1 at low multiplicity of infection and allowed to grow to confluence without selection. After they reached confluence, cultures infected with either virus showed distinct foci of morphologically transformed refractile cells within the otherwise flat cuboidal cell monolayer (Fig. 1A-C). The focus-forming efficiency of MVwg virus was approximately five-fold lower than that of MVWnf-1 (Table 1), and the foci induced were generally smaller and took 1-2 days longer to develop than those induced by MVW/tf-1. Nevertheless, the numbers of foci were proportional to the quantities of MV veg virus applied, and the lack of significant focus formation induced by the control virus MV7 confirmed that the transformed foci induced by MVwg were vvbzg/c.v.s’-dcpendent. Staining of nuclear DNA in these cultures showed that the foci constitute regions of increased cell density as well as transformed morphology (data not shown), suggesting that wingless has a mitogenic effect on these cells at confluence as well as a capacity to induce morphological transformation.

We also examined the phenotype of C57MG cells that were selected for infection with the above virus stocks and grown as pooled populations of G418-resistant colonies. 2 days after reaching confluence, cultures infected with the control virus MV7 formed a monolayer of flat cuboidal cells similar to uninfected cultures (Fig. ID). Those infected with MVwg, however, were morphologically transformed and indistinguishable from MVWni-1-infected cultures with respect to their refractile and disordered appearance (Fig. 1E,F). To assess the mitogenic potential of wingless expression in confluent C57MG cells, we first measured the maximal cell densities achieved in the infected cultures: those infected with MVwg grew to 3-to 4-fold higher cell densities than MV7-infected controls and their final densities were equivalent to those of Wnt-1 -transformed cells (Fig. 2A). We next examined DNA synthesis in confluent cultures by measuring [³H]thymidine incorporation into nuclei. Despite their higher cell densities, MVwg-infected cultures were substantially more active in DNA synthesis than control cells, again implying that expression of wingless is mitogenic in confluent cultures of C57MG cells (Fig. 2B).

Taken together, the above results indicate that wingless can induce transformation of C57MG cells with a resulting phenotype equivalent to that obtained by expression of Wnt-l (Brown et al., 1986). Similar morphological changes were also observed when wingless was expressed in RAC3Ilc cells, another mammary line that can be transformed by Wni-1 (Fig. 1G-I; Rijsewijk et al., 1987b). In contrast, no phenotypic effects of wingless were seen in Rat-2 cells (data not shown), consistent with the known cell-type specificity of Whr-1 transformation (Jue et al., 1992).

Unlike C57MG cells, RAC311c cells can be converted to a tumorigenic phenotype by expression of mouse Wnt-\ (Brown et al., 1986; Rijsewijk et al., 1987b). To determine whether Drosophila wingless could substitute for the mouse gene in this oncogenicity assay, pooled populations of infected RAC311 c cells were injected subcutaneously into athymie nu/nu mice. While no tumors were detected at sites injected with MV7-infected control cells, more than 50% of the sites receiving MVwg-infected cells produced tumors within 12 weeks (Table 2). The frequency and latency of these wmgim-induced tumors were not significantly different from those observed with RAC3I Ic cells infected with MVIV///-1 (Table 2).

To confirm that the MVtc.g virus used in these experiments directed synthesis of Wingless protein, and to examine the distribution of the protein, we analyzed MVwg-infected cells by indirect immunofluorescence using an anli-Wingless antiserum (Gonzalez et al., 1991). In methanol-fixed C57MG/MVwg cells we observed a staining pattern consistent with localization of Wingless antigen to the endoplasmic reticulum and Golgi (Fig. 3A). As well as this intracellular staining. in some fields we noticed diffuse extracellular staining beyond the cell boundaries. To examine this further, intact cells were detached from coverslips by treatment with EDTA and the extracellular matrix material remaining on the coverslips was processed for immunostaining. This revealed extensive cloudy staining in the matrix which was specific for cells expressing wingless. If cells were detached when at low density, the patterns of staining often resembled blurred ‘footprints’ of cells, as if corresponding lo regions where cells had previously adhered to the substrate (Fig. 3B). This material did not stain with antibodies to vinculin or a-aetinin (data not shown) and cellular debris was not visible in these regions under phase contrast illumination.

These results therefore indicate that a proportion of the Wingless product secreted from the cells is bound Io ECM material in the immediate vicinity. This finding has been confirmed by immunoblot analysis of ECM fractions (data not shown). In addition, we performed staining of unfixed cell cultures and observed Wingless antigen associated with the cell surface or pericellular matrix (Fig. 3C). Similar patterns of intracellular and extracellular staining were also seen using RAC3I Ic/MVug cells or Ral-2/MVng (data not shown).

To seek evidence of any Wingless protein released into the ceil culture medium, we subjected conditioned medium from C57MG/MVwg cells to centrifugation al 100,000 g and were able to sediment Wingless products under these conditions (Fig. 31)). Their abundance in this pelleted fraction was increased slightly by including heparin in the culture medium (lanes 2 and 3). Immunoblot analysis of these proteins showed that a significant proportion had undergone proteolytic cleavage, although putative full length species of 51-53x10’ M_r were also detected (Tig. 3D).

In Drosophila there is extensive evidence suggesting that wingless acts in intercellular signaling, and we wished to investigate this in the present cell culture system. Since wingless docs not induce transformation of Rat-2 cells, we tested the ability of the gene to function in the paracrine transformation assay described by .lue ct al. (1992). Rat-2 cells infected with either MVu-.g or MV7 were co-cultured with a large excess of uninfected C57MG cells and the mixture allowed to grow to confluence. In control cultures, the Rat-2/MV7 derivatives formed discrete colonies scattered throughout the C57MG cell monolayer (Fig. 4A). In co-cultures containing Wingless-expressing cells, however, we observed rings of morphologically transformed C57MG cells surrounding the majority of Ral-2/MVwg colonies (Fig. 4B). Thus, as a result of expressing Wingless, the Rat-2/MVvvg fibroblasts were able to induce transformation of neighboring mammary cells, some of which were located several cell diameters away from Wingless-expressing cells. These results confirm that the wingless gene can act at a distance Io affect (he phenotype of neighboring cells.

Drosophila homologues of vertebrate proto-oncogenes include genes encoding transcription factors, members of the ras family of GTPases, cytoplasmic and membranebound kinases, transmembrane receptors, and growth factorlike molecules (Shilo. 1987; Hoffmann. 1989; Kaizen et al., 1991; Pulido et al., 1992; van Lohuizen et al., 1991). Although in some cases there is evidence for conservation of transforming potential in chimeric mutant proteins (Holland et al., 1990), normal Drosophila gene products have not so far been shown capable of effecting transformation of mammalian cells. In this report we have used mammary cell transformation assays to evaluate the functional significance of sequence conservation between the Drosophila segment polarity gene wingless and its murine homolog, the proto-oncogene W/Z/-I (Rijsewijk et al., 1987a). We have shown that wingless can induce morphological transformation and mitogenic effects in mouse mammary cell lines with a resulting phenotype comparable to that seen with IV/Z/-I. that the transforming potential of the two genes show similar cell-type specificities, and that RAC3I Ic mammary epithelial cells expressing wingless are tumorigenic in mice. By these criteria, therefore, wingless can act as an oncogene in certain mammalian cell lines.

As well as qualitative similarities in phenotype, the effects of expressing wingless and Wnt-1 in mammary epithelial cells were quantitatively similar in assays of ontogenesis in confluent cell cultures, maximum cell densities, and tumorigenic potential, wingless was approximately 5-fold less efficient than Wnl-1 in focus assays on unsclcctcd C57MG cell monolayers, however. ‘The latter assays, which probably depend on a combination of autocrine and paracrine transformation, are likely to be the most sensitive in revealing differences in transforming potential. We also noted that wingless was somewhat less efficient than Wnl-\ al inducing paracrine transformation in co-culture assays. It remains to be determined whether these differences in activity are intrinsic to the two proteins, or reflect differential efficiencies of their translation or secretion in these cells.

Of the 370 amino acids in mouse Wnt-1 protein, 54% are conserved in the Wingless product. The human WNT-2 gene has also been shown to cause transformation of C57MG cells (Biasband et al., 1992), and when this sequence is taken into account, 122 residues (33% of those in Wnt-1) are common to all three gene products. The conserved amino acids extend over most of the length of the proteins, suggesting that sequences responsible for transformation may not be confined to a single discrete domain. The most striking features of the conserved sequence are the 22 cysteine residues whose relative positions are nearly identical in all three proteins. Assuming that most of these are involved in disulfide bonding, it is likely that a specific tertiary structure is critically important for Wnt protein function. The numerous cysteines are also conserved in most members of the Wnt gene family that have been sequenced to date (Nusse and Varmus, 1992; McMahon, 1992; and refs, therein). The structure they dictate is clearly not sufficient for transformation, however, since mammary cell transformation is not a universal property of all Wnt genes (J. Kitajewski, personal communication).

The protein products of wingless and Wnt-1 show similar secretory properties in the cells used in these experiments. We have previously shown that secreted Wnt-1 protein is associated with the extracellular matrix of cultured cell lines (Bradley and Brown, 1990), and Papkoff and Schryver (1990) have reported Wnt-1 associated with the cell surface or pericellular matrix. Using antibodies to visualize wingless products by immunofluorescence, we have shown here that Wingless protein is present both at the cell surface and in the ECM of C57MG cell cultures. In addition, we detected Wingless product in the conditioned culture medium by centrifuging the medium and analyzing the pellet by immunoblotting. The ease with which this fraction could be sedimented, however, suggests that it is not freely soluble and instead may be present within molecular aggregates, possibly including ECM components not bound to the substratum. One way to explain the distribution of Wingless in these cultures would be to propose that the secreted protein itself is bound to another molecule, such as a proteoglycan, which may exist both as a surface-bound moiety and in a form released into the ECM and culture medium.

The extracellular location of Wingless protein in these mammalian cell cultures is broadly consistent with the distribution of Wingless antigen observed in Drosophila embryos, which can be detected at 1-3 cell diameters distant from cells expressing wingless RNA (van den Heuvel et al., 1989; Gonzalez et al., 1991). More specifically, our data add support to the conclusions of van den Heuvel et al. (1989) that a proportion of Wingless antigen is associated with the ECM in Drosophila. In contrast, a recent study of the distribution of Wingless protein expressed in Xenopus oocytes and embryos found most of the protein to be intracellular and a secreted form was not demonstrated except in the presence of high concentrations of the anionic compound suramin (Chakrabarti et al., 1992). The reasons for this apparent discrepancy are unclear but may be related to the very high level of transient expression achieved in the oocyte system, together with the propensity of Wnt-1 protein to be retained in the endoplasmic reticulum when overexpressed (Papkoff, 1989; Kitajewski et al., 1992).

In view of the secretory nature of Wingless in transformed C57MG and RAC311c cell cultures, the protein presumably acts as an autocrine factor in this system. Although in Drosophila the gene may act in an autocrine manner in certain cells, there is also extensive evidence that the Wingless product acts in the embryo as a paracrine factor (reviewed by Ingham, 1991; Peifer and Bejsovec, 1992). For example, wingless expression in the epidermis of early embryos is necessary for maintenance of engrailed expression in neighboring cells, and Wingless protein can be detected within cells adjacent to those in which it is synthesized (Martinez Arias et aL, 1988; DiNardo et al., 1988; van den Heuvel et al., 1989; Gonzalez et al., 1991; Heemskerk et al., 1991). In what is apparently a combination of autocrine and paracrine effects, wingless causes modulation of Armadillo protein in broad stripes of cells including and surrounding those expressing wingless RNA (Riggleman et al., 1990). In addition, clones of wingless mutant cells behave in a non-autonomous manner in genetic mosaics (Wieschaus and Riggleman, 1987; Morata and Lawrence, 1977). All these data suggest that wingless normally acts in intercellular signaling and the mammary cell transformation assays described here provided an opportunity to test this notion experimentally. Although colonies of Rat-2 fibroblasts expressing Wingless protein are not themselves ostensibly altered, when co-cultured with C57MG mammary epithelial cells they were able to induce transformation of the surrounding mammary cells (Fig. 5). Like mouse Wnt-l, therefore, the wingless gene is able to elicit transformation via a paracrine mechanism (Jue et al., 1992). A paracrine effect of wingless has also recently been demonstrated in co-cultures of Drosophila embryonic cells (Cumberledge and Krasnow, 1993). Collectively these cell culture assays strongly support models of Wingless as an intercellular signaling factor.

In causing transformation of mammary cells, the products of both wingless and Wnt-1 presumably interact with the same cell surface receptors in the target cells. A similar notion is also implied by the recent finding that both genes can induce dorsal mesoderm in Xenopus embryos (Chakrabarti et al., 1992). Specific receptors for Wnt-1 or Wingless have yet to be identified in either mouse or Drosophila, but it seems likely that these too will share sequence homology, at least in their ligand-binding domains. If elements of the signaling pathways activated by these receptors are also conserved in evolution, it is possible that the growth-promoting effects induced by wingless in mouse mammary cells may reflect some of the normal consequences of wingless expression during Drosophila embryogenesis. In this regard it is particularly interesting that wingless appears to have a mitogenic effect on the developing Malpighian tubules of Drosophila, where the gene is normally expressed in the tubule primordia (Skaer and Martinez Arias, 1992). In the absence of wingless function, cells of the tubule anlage fail to proliferate, while over-expression of the gene in wild-type embryos can promote additional cell divisions in the tubules resulting in supernumerary cells (Skaer and Martinez Arias, 1992). In the epidermis, wingless expression is not known to be mitogenic during the establishment of segment polarity, although in the absence of wingless function there is significant cell death accompanying the deletion of specific pattern elements within the body segments (Perrimon and Mahowald, 1987; Klingensmith et al., 1989). Further analysis will be required to determine whether the mitogenic and transforming functions of wingless in mammalian cells depend on the same protein domain(s) required for normal wingless function in Drosophila.

In summary, our results demonstrate that the ability of Whi-1 to cause transformation of mammary epithelial cell lines has been conserved in evolution of the gene from Drosophila to mouse. Despite their sequence divergence, the products of wingless and mouse Wnt-1 show similar biochemical properties and can act via a common mechanism that presumably involves specific cell surface receptors. The conservation of transforming potential demonstrated here emphasizes the validity of studying oncogene homologs in Drosophila, in which genetic analysis may identify additional components of signaling pathways relevant to vertebrate tumorigenesis. Conversely, our data imply that studies of WhM and its effects in cultured cells may be directly relevant to understanding the mechanism of action of Wingless. Finally, the transformation of mouse mammary cell lines provides a convenient cell culture assay for wingless and should facilitate biochemical and functional studies of both wild-type and mutant Wingless proteins to elucidate their roles in intercellular signaling.

We thank Alfonso Martinez Arias, Rocl Nusse, and Enrique Rodriguez-Boulan for generous gifts of reagents, Roger Bradley for helpful advice, and Shall Jue for technical assistance, This work was supported by NIH grants CA47207 and CAI6599, and by funds from the Pew Scholars Program. N. R. R. is supported by the Medical Scientist Training Program of Cornell University Medical College.