The Wnt/Wg signal transduction pathway is an evolutionarily conserved pathway that plays an important role in the developmental program of many organisms (for review see 1234). Genetic epistasis tests in Drosophila, in combination with biochemical experiments performed in Xenopus and in tissue culture cell lines, have contributed much information to the molecular mechanisms underlying Wnt signaling. In recent years, the discovery of the genes and their protein products that are involved in Wnt signaling has led to the development of a complex signaling network that includes various scaffolding proteins, kinases, and transcription factors. For a detailed description of the proteins involved, please see The Wnt Genes Webpage 5. With respect to the canonical Wnt/β-catenin pathway, in the absence of Wnt/Wg signaling, cells maintain low cytoplasmic levels of the oncoprotein β-catenin or Armadillo (Arm), its Drosophila homolog. β-catenin is a multifunctional molecule that acts at the plasma membrane in adherens junctions, can be found in the cytoplasm, and is also detected in the nucleus as part of a transcription factor complex. Upon phosphorylations at its amino terminus by an isoform of casein kinase 1 (CK1) and by glycogen synthase kinase-3β (GSK-3β), β-catenin is targeted for ubiquitin-mediated degradation by the "destruction complex" consisting of GSK-3β, the tumor suppressor protein adenomatous polyposis coli (APC), axin, and a member of the SCF ubiquitin ligase complex, β-TrCP/Slimb 67. In order to initiate signaling, Wnt must bind to its receptor Frizzled and its co-receptor LRP, leading to β-catenin stabilization. β-catenin levels now rise in the cytoplasm, reaching a critical amount that enables β-catenin to translocate to the nucleus where it interacts with Lef/TCF 89 and activates target gene expression.

Besides playing a role in differentiation and development, the Wnt/Wg signaling pathway has also been implicated in tumorigenesis 7. The Wnt-1 oncogene was originally discovered as int-1, a gene whose activation upon insertion of the mouse mammary tumor virus results in the formation of mouse mammary tumors 1011. Although Wnt-1 is not expressed in the normal mouse mammary gland, expression of Wnt-1 in transgenic mice results in the formation of mammary tumors. In addition, overexpression of Wnt-1 in mouse mammary epithelial cell lines such as C57MG or RAC311 results in their transformation 12. Furthermore, in Rat1 fibroblasts, Wnt-1 overexpression induces serum-independent cellular growth in a manner that correlates with an increase in cytoplasmic β-catenin 13. These results suggest that perhaps Wnt-1-mediated transformation occurs through an increase in cytoplasmic β-catenin levels because transformation can take place in the absence of Wnt-1 signaling upon overexpression of a stable form of β-catenin. Other cell lines, such as NIH3T3 cells, have also been employed to study the response of endogenous β-catenin to Wnt-1 signaling and have arrived at similar conclusions: Wnt-1 signaling allows for accumulation of β-catenin to levels that enable it to form a complex with Lef/TCF proteins to activate transcription 14.

Expression of some Wnts has been correlated with the development of human cancer. Recently, increased expression of Wnt-1 mRNA was observed in cell lines derived from a human gastric cancer (OKAJIMA), pancreatic cancer (BxPC-3), and in 50% of primary gastric cancers 15. The Wnt 2 gene is up-regulated at the mRNA level in gastric and esophageal carcinomas as well as in colorectal tumors at various stages 16. Wnt-5a mRNA is expressed at very low levels in breast cell lines and normal breast tissue, but benign proliferations and invasive cancers exhibit a ten-fold and four-fold higher level of Wnt-5a mRNA, respectively 17. Wnt-5a is also up-regulated at the RNA level in lung, breast, prostate carcinomas and melanomas 18. Wnt10B mRNA is elevated in primary breast carcinomas and in several noncancerous and cancerous breast cell lines 19. Finally, Wnt-13 mRNA is expressed in several adult tissues and in three human cancer cell lines: HeLa cells (cervical cancer) and in MKN28 and MKN74 (gastric cancer), potentially implicating this gene in tumorigenesis 20. Thus, the up-regulation of Wnts is correlated with the development of multiple tumor types, providing additional evidence that Wnt proteins and the pathway(s) they regulate play an important role in the control of cell growth and differentiation.

Mutations in several components of the Wnt signaling pathway, such as APC and β-catenin, have also been linked to tumorigenesis. APC is a tumor suppressor gene that is mutated in up to 80% of human colon carcinomas 21. Mutations of APC result in the development of a form of inherited colon carcinoma called Familial Adenomatous Polyposis (FAP). Individuals with FAP develop multiple colonic polyps throughout their life, predisposing them to colon cancer. Elevated levels of β-catenin can also contribute to tumorigenesis. High levels of β-catenin are associated with several human cancers, including colon carcinomas, melanomas, pilomatricomas and hepatocellular carcinomas, either due to a nonfunctioning APC protein or to mutations that eliminate the phosphorylation sites within β-catenin 622232425.

The diverse roles of Wnts in both development and tumorigenesis have fostered the search for Wnt responsive genes in these processes. Wg signaling in Drosophila is known to transcriptionally activate the expression of engrailed and Ultrabithorax through the Armadillo/dTCF complex 2627. Besides these Drosophila homeobox genes, Wnt signaling through β-catenin in Xenopus results in the transcriptional induction of two additional homeobox genes, siamois 28 and twin 29. Other responsive genes of Wnt/Wg that are transcriptionally activated by β-catenin/Lef-TCF include the oncogenes cyclin D1 3031, c-myc 32, and WISP-1 3334; c-jun and fra-1 35, as well as the Xenopus fibronectin gene 36, connexin43 37, matrilysin 38, BTEB2 39, Wrch-1 40, and membrane-type matrix metalloproteinase-1 (MT1-MMP) 41.

The canonical Wnt signaling pathway relies upon the activation of responsive gene expression through the β-catenin/Lef-TCF complex in response to the binding of Wnt to its receptor Frizzled. Evidence exists, however, to suggest that Wnts can signal through a β-catenin/Lef-TCF-independent mechanism to activate downstream gene expression. These "non-canonical" pathways rely on either (1) the phosphatidylinositol (PI) pathway to activate protein kinase C (PKC) and raise levels of intracellular calcium (Ca²⁺) in order to regulate responsive gene expression (the Wnt/Ca²⁺ pathway) 42 or (2) Wnt signaling through Dsh and the JNK (the Wnt/PCP pathway) 4344. For example, Xenopus Wnt-5a, a Wnt that does not induce ectopic axis formation nor stabilizes β-catenin, employs the Wnt/Ca²⁺ pathway to exert its effects 45. Thus, it is apparent that not all Wnts initiate the transcriptional activation of their responsive genes through the β-catenin/Lef-TCF complex, suggesting that β-catenin stabilization is not the sole end result of all Wnt signaling and that other pathways can be stimulated upon binding of Wnt to the Frizzled receptor.

The current list of Wnt/Wg responsive genes contains primarily those genes whose expression is activated through the canonical pathway. In an attempt to identify additional downstream responsive genes of the Wnt signaling pathway that are relevant to transformation and potentially tumorigenesis, we performed a suppression subtractive hybridization (SSH) screen between a Wnt-1-expressing mouse mammary epithelial cell line (C57MG/Wnt-1) and the parental cell line (C57MG). SSH offers many advantages when compared to conventional methods, such as differential display (DD) and representational difference analysis (RDA), employed to identify mRNA and/or cDNA differences between two populations. Differential display is more suitable to different cell types rather than one cell type subjected to two different conditions, as differences in PCR amplification can make control mRNAs appear differentially expressed between two populations, and results are often not reproducible upon Northern blot analysis 46. Representational difference analysis 47 does not rely on suppression PCR to prevent the amplification of undesirable sequences. SSH, on the other hand, permits the rapid isolation of both rare and abundant messages from the tester population in two hybridization steps followed by two PCR reactions. It has been widely used over the past five years to identify syntenin and other TNF-inducible genes in endothelial cells 48, to contrast gene expression profiles between non-tumorigenic rat embryo fibroblasts and H-ras transformed cells 49, to identify differences between the bacterial genomes Escherichia coli and Salmonella typhimuriu 50, and to generate a testis-specific cDNA library, where rare sequences were enriched over 1000-fold in one round of subtractive hybridization 51.

Using the technique of SSH, we previously reported the identification and characterization of a mouse homolog of BTEB2 39, Wrch-1 40, and WISP-1 3334. This report details the remaining results of the subtractive hybridization screen; in particular, the identification of 59 additional genes confirmed to be up-regulated in response to Wnt signaling. These results reveal the gene expression changes that occur in a tissue culture model system in response to Wnt signaling and may help explain the resulting transformed phenotype.

cDNA subtractive hybridization has identified several putative Wnt-1 responsive genes at the mRNA level in a mammary epithelial cell line that overexpresses Wnt-1 (C57MG/Wnt-1). The up-regulation of these candidate genes was confirmed by Northern blot analysis and differential screening of material isolated from three independent sources: 1) C57MG cells infected in an independent retrovirus infection assay, 2) co-culture of C57MG cells with a cell line that overexpresses Wnt-1 (QT6Wnt-1), and 3) C57MG cells that overexpress a stable form of β-catenin. In addition, some of these putative Wnt-1 responsive genes were confirmed by semi-quantitative RT-PCR analysis using RNA obtained from C57MG/Wnt-1 cells and the parental C57MG cells. Independent retroviral infection of C57MG cells with a Wnt-1 or empty retroviral vector confirmed that, of the 67 genes subjected to Northern blot analysis, QRT-PCR and/or differential screening, approximately 93% (62) were up-regulated ≥1.5-fold by Wnt-1 signaling.

Signaling pathways give rise to a cascade of gene expression, with the immediate response genes being activated first, usually in the absence of any new protein synthesis 646566. The function of such immediate response genes ranges from regulators of the cell cycle to acting as transcription factors 6768. The initial activation of transcription factors then permits the activation of a second wave of gene expression, the early or middle response genes 6469. The late response genes are activated last, and may or may not be transcriptionally regulated by the early response genes 7071. In the case of Wnt-1 transformation of C57MG cells, we hypothesized that these late response genes may function to maintain the transformed state of these cells throughout their lifetime in tissue culture. Therefore, the screen was performed using C57MG cells that had been stably expressing Wnt-1 for approximately three weeks to ensure the identification of all three categories of response genes. In addition, total RNA was extracted when cells were approximately 95% confluent to ensure the effects of Wnt-1-induced transformation and to enable us to determine how gene expression is regulated in response to the mitogenic effects of Wnt signaling. It is evident to us, however, that by using this time point of three weeks, we may potentially miss identifying some immediate (and perhaps early) response genes that are only transiently activated in response to Wnt signaling, as such genes may not remain induced over such an extended period of time. In addition, as is true of all non-saturation screens, some known targets of Wnt signaling were not identified, such as cyclin D1. We recognize that this is a limitation of the screen methodology but are still encouraged by the number of Wnt responsive genes that we have identified.

The sequences having database homology and that were confirmed to be up-regulated in response to Wnt signaling in one (or more) of the independent assays could be categorized based upon their putative function as follows: cell cycle exit/differentiation, transcription factors, calcium-related, ECM/cytoskeleton, p53-related, GTP-binding proteins, ubiquitin-related, and other. One category of interest was that of the transcription factors, as these might include 1) genes activated, along with β-catenin, to mediate responsive gene expression and/or 2) novel β-catenin responsive genes. We believe that the identification of a battery of transcription factors up-regulated in response to Wnt signaling may aid us in developing a model in which the activation of these transcription factors, along with the β-catenin/Lef-TCF complex, mediates the up-regulation of other genes by Wnt signaling, including those identified in our screen. Another category of interest consisted of several genes involved in remodeling the extracellular matrix (ECM) and the cytoskeleton. The identification of these genes as potential Wnt-1 responsive genes is interesting in light of the ability of Wnt-1 to radically alter the morphology of C57MG cells, a process that involves changing the cytoskeleton of these cells and their surrounding ECM. Finally, connexin43 was identified as a putative responsive gene of Wnt-1 signaling in our assay. The identification of this gene as a responsive gene of the Wnt-1 signaling pathway is expected given previous evidence that it is a veritable Wnt-1 target 63 and lends credence to the ability of the screen to identify true responsive genes of the Wnt-1 signaling pathway.

Several of the genes confirmed by independent retroviral infection of C57MG cells, along with the putative transcription factors, were chosen for analysis in the co-culture assay. The co-culture assay was performed in two phases, an early co-culture of one to four hours, and a later co-culture of six to forty-eight hours. Several of the Wnt-1 responsive genes highly up-regulated by retroviral infection were not induced in the co-culture assay. At first this result seemed surprising, but it can probably be explained by the conditions under which these genes were identified: the screen, as well as the independent retroviral infection, was performed on cells that had been passaged in tissue culture for three weeks after Wnt-1 retroviral infection.

Concomitant with the co-culture assay, differential screening was employed in order to confirm the screen results. Many of the genes confirmed by independent retroviral infection through Northern blot analysis were also found to be induced in the differential screening assay using C57MG cells infected with a Wnt-1 retrovirus as the starting material. Although the fold induction was not identical for both assays, this is to be expected given that the methodology for each is different. Furthermore, differential screening relies upon the ability of RNA to be converted to labeled cDNA and for each gene to be equally represented, in excess, on the nitrocellulose membrane, in order to hybridize with the labeled cDNA. Given these caveats, it is not surprising to see differences between the fold activation for each gene as determined by Northern blot analysis and differential screening. The conversion of RNA to labeled cDNA assumes that all transcripts will be converted and labeled with the same efficiency by the reverse transcriptase enzyme. It is highly possible that some transcripts will be more difficult to convert to cDNA given the nature of their sequence. This would give rise to an unequal representation of particular transcripts within the cDNA mixture and could explain the differences observed in fold activation by the differential screening method. These inherent limitations to the differential screening technique must be taken into account and hence the results should be confirmed by Northern blot analysis and/or QRT-PCR to determine if in fact the gene is regulated by Wnt-1 signaling.

The differential screening assay was also performed with a C57MG cell line that stably overexpresses a non-degradable β-catenin mutant, and the results were compared to those found for a differential screening assay conducted with a twelve hour QT6Wnt-1+ C57MG co-culture. Several of the genes transcriptionally up-regulated by the Wnt-1 co-culture were also found to be up-regulated by overexpressing β-catenin. This is to be expected, given that the canonical Wnt-1 signaling pathway relies upon the β-catenin/Lef-TCF transcription factor complex to mediate the activation of downstream responsive genes.

These results provide us with a molecular signature of the changes that occur in a mammary epithelial cell line (C57MG) upon transformation by Wnt-1. Although it is not possible to distinguish between genes induced directly and indirectly by Wnt signaling using the methods described herein (because we looked three weeks after retroviral infection and also did not analyze the promoters of the identified genes for putative TCF binding sites), the techniques utilized here still have provided us with a useful tool to rapidly assess changes induced during the transformation phenotype. With these results in hand, we now have a snapshot of the gene expression changes that take place after three weeks of Wnt-1 overexpression, and we can use this information to understand how the protein products of these up-regulated genes help establish and/or maintain the transformation phenotype. We surmise that many changes take place in the cell upon reception of Wnt, including, on a molecular level, the activation of various downstream signaling pathways mediated by other transcription factors, calcium, phosphatases, and GTP binding proteins. These in turn go on to activate other genes that lead to the morphological changes apparent upon transformation; for example, the activation of various ECM and cytoskeletal proteins to remodel the cell architecture. It is possible that such a model can be used to investigate the roles of Wnts, and how they transduce their signals, in other systems.

LT carried out all of the molecular studies and experiments described herein, with the exception of the actual screen process, sequencing, and QRT-PCR experiments that were performed by DP. LT, in conjunction with Dr. Arnold J. Levine, conceived of the study and participated in its design and coordination. All authors read and approved the final manuscript.