Kinase-Inactive Glycogen Synthase Kinase 3β Promotes Wnt Signaling and Mammary Tumorigenesis

Glycogen synthase kinase 3β (GSK3β) is a serine/threonine kinase that was originally found to have a pivotal role in glycogen metabolism and insulin-mediated signaling but is now recognized to function in multiple biological pathways. More than 40 proteins have been reported to be phosphorylated by GSK3β, including over a dozen transcription factors (1). Recently, attention has focused on the developmental role of GSK3β. During fly development, the GSK3 homologue, zeste-white 3, is a negative regulator of wingless (wg) signaling, the agonist responsible for normal wing development. The vertebrate homologues of wg, the Wnts, are responsible for embryonic patterning beginning with the establishment of the embryonic axes (2). In Xenopus development, the dorsoventral axis is established by dorsal accumulation of β-catenin, a critical coactivator in the Wnt signaling pathway (3). Positive elements of the Wnt canonical pathway (e.g., β-catenin) produce ectopic axes when injected ventrally, whereas inhibitors or negative regulators of the pathway antagonize dorsalization. Rat, human, and Xenopus GSK3β block dorsalization, whereas inactive mutants of GSK3β act as dominant negatives of the normal enzyme function, inducing axis duplication when injected ventrally in the embryo (4–6).

Biochemical studies have elucidated the role of GSK3β in the canonical Wnt signaling pathway. In the absence of Wnt signals, free cytoplasmic β-catenin is incorporated into a cytoplasmic complex that includes Axin, GSK3β, and adenomatous polyposis coli (APC). This enables casein kinase I to phosphorylate β-catenin, creating a consensus site on β-catenin for phosphorylation by GSK3β. The phosphorylated β-catenin is then targeted for ubiquitin-mediated proteasomal degradation (7). This process is opposed by casein kinase 2 (CK2), which phosphorylates β-catenin in the armadillo repeat region, stabilizing it and promoting Wnt signaling and dorsal axis formation (8–10). Wnt signaling via Dishevelled (Dsh) inactivates GSK3β and prevents it from phosphorylating β-catenin, reducing its affinity for axin and APC and stabilizing it in the cytoplasm. As β-catenin accumulates, it translocates into the nucleus, where it binds to T-cell factor (TCF) and lymphoid-enhancing factor (LEF) transcription factors and dramatically increases their transcriptional activity. Genes up-regulated by TCF/LEF include embryologic genes, such as siamois and engrailed (11), and adult proto-oncogenes, such as c-myc and cyclin D1 (12–14).

In the mammary gland, canonical Wnt signaling seems to play a role in both development and cancer. Wnt 6 and Wnt 10b are expressed on the surface of the ectoderm in mammary placodes and buds beginning on embryonic day 11.25 and are essential for initiation of mammary morphogenesis (15, 16). LEF-1 is expressed in the mammary gland beginning at embryonic day 11.5, and in LEF-1 deficient mice, the gland fails to develop (17). ΔN89-β-catenin, a mutant of β-catenin that lacks the NH₂-terminal GSK3β phosphorylation sites and is thereby stabilized, promotes precocious alveolar development during puberty (18). Negative regulators of Wnt signaling block mammary gland development: ectopically expressed Dickkopf, a Wnt pathway inhibitor, blocks early mammary gland development (19); other negative regulators of β-catenin inhibit alveolar development in pregnancy (20).

Although a role for the Wnt pathway is well recognized in colon cancer, where, for example, mutations of the APC and β-catenin genes are found in both sporadic and inherited cancers, little is known about the Wnt pathway in human mammary tumorigenesis. Overexpression of several Wnts has been reported in breast cancer (21–24) and amplification of the Dsh downstream messenger has been seen in 50% of primary breast tumors (25). Up-regulation of β-catenin mRNA levels has been detected by microarray analysis in human breast cancer (26) and elevation of β-catenin protein expression has been reported in 60% of human breast cancer tissues (27). Detection of β-catenin by immunohistochemistry has been associated with poor outcome (28, 29). These events have been modeled in mice, as mammary gland tumors develop in transgenic mice overexpressing genes in the Wnt signaling pathway, including Wnt 1 (30), Wnt 10b (31), ΔN89-β-catenin (18), and cyclin D1 (32). In contrast, in transgenic mice that overexpress Axin, the expression of cyclin D1 is attenuated and increased apoptosis occurs in the mammary epithelia (33). Overexpression of the regulator CK2α also promotes mammary tumorigenesis (34). In this regard, GSK3β has not been studied. In this article, we use a novel mouse model to explore the effect of Wnt pathway deregulation on the development of breast cancer, with emphasis on GSK3β as a pivotal kinase regulator in this pathway. Because inactive mutants of GSK3β can activate canonical Wnt signaling, we hypothesized that overexpression of kinase-inactive GSK3β (KI-GSK3β) in mammary glands would work as a dominant negative, antagonizing the endogenous activity of GSK3β. Consistent with this, we find that kinase-inactive murine GSK3β (KI-mGSK3β) stabilizes β-catenin expression and catalyzes its localization to the nucleus and up-regulates the downstream target gene, cyclin D1, in vitro. In vivo, transgenic mice overexpressing the mutant form of GSK3β under the control of the mouse mammary tumor virus (MMTV)-long terminal repeat (LTR) develop mammary tumors. Analyses of the tumors show overexpression of β-catenin as well as cyclin D1. Thus, antagonism of GSK3β activity is oncogenic in the mammary epithelium; mutation or pharmacologic down-regulation of GSK3β could promote mammary tumors.

Cloning of mouse glycogen synthase kinase 3β and kinase-inactive murine glycogen synthase kinase 3β. Based on the sequence of human GSK3β, we synthesized a pair of primers and amplified a portion of the mGSK3β cDNA. A randomly primed murine spleen Lambda ZAP II cDNA library (Stratagene, La Jolla, CA) was screened with the radiolabeled murine PCR product and a single clone was isolated that contained a full-length GSK3β open reading frame (ORF) based on bidirectional sequencing (DNA/Protein Core Facility, Boston University School of Medicine, Boston, MA). Double-stranded site-directed mutagenesis was done on the mGSK3β sequence to introduce a mutation into the ATP-binding site, changing KKV at amino acids 85 to 87 to MII with the expectation of creating an inactive enzyme, mGSK3β-KI. This created an additional BclII site, so PCR was used to amplify the fragment containing the mutation and distinguish it by BclII digestion, and it was also confirmed by sequencing. For in vitro experiments, the KI-mGSK3β cDNA was subcloned into pcDNA3.0 (Invitrogen, Carlsbad, CA). Using PCR, a hemagglutinin (HA) tag was added 5′ to the KI-mGSK3β coding sequence in this vector. The sense oligonucleotide primer contained HA tag and the KI-mGSK3β sequence (5′-GGGGTACCACCACCATGGCCTACCCATACGACGTACCAGACTACGCATCGGGGCGACCGAGAACCACC-3′) and an antisense oligonucleotide was from the end of the ORF (5′-GGTCTAGAGCTCCTGGGGGCTGTTCAGG-3′). Parallel experiments to those with the HA-KI-mGSK3β were carried out with a myc-tagged kinase-inactive mutant of rat GSK3β (rGSK3β), myc-rGSK3β(K85R).

Cell culture. C57MG normal mouse mammary epithelial cells were grown in DMEM supplemented with 10% fetal bovine serum, 4 mmol/L glutamine, 50 units/mL penicillin, and 50 mg/mL streptomycin (Cellgro, Mediatech, Inc., Herndon, VA) in a 5% CO₂ incubator at 37°C. Transfections were done using LipofectAMINE 2000 (Invitrogen) or the Amaxa nucleofection system according to the manufacturer's instructions. The total amount of transfected DNA was kept constant by adding plasmid vector DNA when necessary. For small interfering RNA (siRNA) experiments, cells were transfected with SMARTpool mGSK3β siRNA or siCONTROL (Dharmacon, Chicago, IL). Alternatively, cells were treated with the GSK3 inhibitors SB216763 (Sigma, St. Louis, MO) or TDZD-8 (Calbiochem, San Diego, CA).

Western blot analyses. Protein extracts were prepared by homogenizing frozen tumors or mammary gland specimens in lysis buffer as described (34). Primary antibodies were the following monoclonal antibodies: anti-β-catenin (BD, Lexington, KY), anti-β-actin (Sigma), anti–cyclin D1 (Calbiochem), anti-α-tubulin (Sigma), anti-HA.11 (Covance/Babco, Richmond, CA), anti-c-myc (Roche, Indianapolis, IN), and anti-SP1 (BD). For quantitative analysis of each band, integrated pixel density minus background density was determined using a Fluor-S MultiImager and analysis was done using Quantity One software (Bio-Rad, Hercules, CA).

For nuclear and cytoplasmic separations, cells were washed, harvested with ice-cold PBS, and centrifuged at 960 × g for 5 minutes at 4°C. The pellet was suspended in 2 volumes of ice-cold, low-salt buffer [10 mmol/L HEPES (pH 7.9), 1.5 mmol/L MgCl₂, 10 mmol/L KCl, 0.05% NP40, 0.5 mmol/L DTT] supplemented with protease inhibitor cocktail (Sigma) and incubated on ice for 30 minutes. Following 15 minutes of centrifugation at 10,600 × g, the supernatants were frozen as cytoplasmic extracts. Nuclei were extracted with 2 volumes of ice-cold, high-salt buffer [20 mmol/L HEPES (pH 7.9), 1.5 mmol/L MgCl₂, 0.42 mol/L NaCl, 0.2 mmol/L EDTA, 0.5 mmol/L DTT, 0.5 mmol/L phenylmethylsulfonyl fluoride, 25% (v/v) glycerol] and incubated at 4°C for 40 minutes. Nuclear extracts were cleared by centrifugation at 20,800 × g for 15 minutes.

Immunofluorescence microscopy. C57MG cells were transfected with HA-KI-GSK3β using the Amaxa nucleofection system and plated on glass coverslips. Twenty-four hours later, the transfected cells were transferred into 1% fetal bovine serum in DMEM, starved overnight, and then stained. Cells were washed thrice with cold PBS and then fixed and permeabilized with 4% paraformaldehyde and 0.5% Triton X-100 for 10 minutes, blocked with 3% bovine serum albumin for 30 minutes, and subsequently incubated at 4°C with primary anti-β-catenin antibody and secondary FITC-conjugated anti-mouse IgG (Sigma) for 60 minutes each. For nuclear staining, we used Hoechst dye by adding a 1:100 dilution of 100 μg/mL stock to the medium of the cells 20 minutes before the beginning of the experiment. Pictures were taken in a fluorescence microscope (Nikon, Japan) fitted with a digital camera (Diagnostic Instruments, Sterling Heights, MI). The software used was Spot Advance (Diagnostic Instruments).

β-catenin protein stability. C57MG cells (0.5 × 10⁶) were transiently transfected with different amounts of the HA-KI-GSK3β construct. After 24 hours, the cells were starved in 1% fetal bovine serum and 16 hours later were treated with 50 μg/mL cycloheximide to block de novo protein synthesis. Samples were taken at the beginning of the experiments and at 2-hour intervals. At each time point, the cells were washed in cold PBS and pelleted, proteins were extracted, and Western blotting was done for β-catenin.

Quantitative real-time PCR and semiquantitative reverse transcription-PCR. Reactions (25 μL) were prepared by mixing 12.5 μL of Taqman Universal PCR Master Mix (Applied Biosystems, Foster City, CA), 5 ng of the relevant cDNA, and 1.25 μL of an Assay-on-Demand gene expression product for cyclin D1 (CCND1) or β-glucuronidase (GUSB) as an endogenous control. Quantitative real-time PCR (qPCR) was done in an ABI Prism 7000 Sequence Detection System (Applied Biosystems). The initial step was for 10 minutes at 95°C and then 40 cycles of denaturation at 95°C for 15 seconds and annealing/extending at 60°C for 1 minute. Background signal was eliminated and Ct values were determined using the SDS version 1.1 analysis software (Applied Biosystems). Standard curves for CCND1 and GUSB were done to confirm that the sample Ct values were within the range. For reverse transcription-PCR (RT-PCR), total RNA (1 μg) was reverse transcribed using the ProSTAR First-Strand RT-PCR kit (Stratagene). PCR with specific primers for cyclin D1 (sense 5′-CCCTCCGTATCTTACTTCAA-3′ and antisense 5′-GATGGTCTGCTTGTTCTCAT-3′) was done in a thermal cycler (MJ Research, Watertown, MA) by denaturing at 95°C for 3 minutes and then 30 cycles of denaturing at 95°C for 30 seconds, annealing at 53°C for 30 seconds, and extending at 72°C for 30 seconds.

Transgenic animals. The KI-mGSK3β cDNA was subcloned into a vector in which the MMTV-LTR directs expression to the mammary epithelium, with ras 5′ untranslated sequences provided upstream of the cDNA and a SV40 intron and polyadenylation signal downstream (35). Plasmid sequences were removed by restriction digestion at the SalI and SpeI sites, and the excised transgene construct was gel purified and microinjected into pronuclei of fertilized one-cell zygotes of FVB/N mice in the Transgenic Core Laboratory at Boston University School of Medicine. Three independent transgenic lines were obtained, and female transgenic mice were continuously bred to induce transgene expression through activation of the hormone-dependent MMTV-LTR. Mice were monitored weekly for the appearance of tumors. Mice (n = 117) were sacrificed and tissues were collected for histopathological analysis, cell culture, and RNA and/or protein analyses. To assess expression levels of the transgene in the mouse organs, tissues were collected from 6- to 8-week-old females that were pregnant and from females that developed mammary tumors.

Expression analyses. For expression analysis, total RNA was extracted from mouse tissues. After DNase treatment (Roche), RNA was reextracted and ethanol precipitated. RNA (5-10 μg) was then reverse transcribed using the ProSTAR First-Strand RT-PCR kit. PCR was done with sense GSK3β (CGAGACACACCTGCACTCTT) and SV40 primers described above, for 35 cycles, to detect the transgene. Both spliced and unspliced transgene mRNA could be amplified (614 and 547 bp), as there is a splice donor and acceptor in the amplified region of the SV40 poly(A) tail. The quality of first-strand synthesis was verified with HPRT amplification (36).

Histology. On necropsy, tumors and organs were removed and immediately fixed in Optimal Fix (American Histology Reagent Co., Inc., Lodi, CA). The tissues were processed, embedded in paraffin, and sectioned at 7 μm. The sections were mounted on glass slides and stained with H&E using routine laboratory procedures in the Transgenic Core Pathology Laboratory at the University of California-Davis (Davis, CA). Immunohistochemistry for cytokeratins, smooth muscle actin, hair keratin, and estrogen and progesterone receptors were done as described previously (37). Images were captured with ×10, ×20, and ×40 objectives using a Carl Zeiss (Thornwood, NY) Axiocam camera and processed using Adobe Photoshop (Adobe Systems, Inc., San Jose, CA) software. Sections were compared with other specimens in the extensive mouse mammary tumor database. ³

Expression of myc-rGSK3β(K85R) up-regulates β-catenin in mammary epithelial cells. As a first step to determine whether dysregulation of GSK3β could play a role in mammary carcinogenesis, we manipulated its levels in cells in vitro and assessed the expression of β-catenin as an indicator of Wnt signaling. Initial experiments were carried out with a previously described rat construct (4). C57MG cells, a nonmalignant cell line derived from murine breast epithelial cells, were transiently transfected by nucleofection with increasing amounts of myc-tagged kinase-inactive mutant form of rGSK3β, myc-rGSK3β(K85R), construct. pEGFP-C1 was cotransfected; the transfection efficiency using this method was 75% (data not shown). After 48 hours, protein extracts were analyzed for β-catenin expression. β-catenin protein levels were up-regulated with increasing expression of the myc-rGSK3β(K85R) (Fig. 1A), consistent with increasing activation of the Wnt pathway.

myc-rGSK3β(K85R) stabilizes endogenous β-catenin protein. In the absence of Wnt signals, β-catenin is incorporated into a cytoplasmic complex, including GSK3β, which targets it for degradation. To validate that the changes in β-catenin expression are due to ability of the myc-rGSK3β(K85R) to stabilize β-catenin protein, we studied the half-life of β-catenin. C57MG cells were transiently transfected with increasing amounts of the myc-rGSK3β(K85R) construct. Forty-eight hours after transfection, the cells were treated with cycloheximide to block de novo synthesis of β-catenin, and periodically, cells were harvested and proteins were subjected to immunoblotting to assess β-catenin stability. The half-life of β-catenin was increased, being 1.5, 5.4, 26, and 11 hours in the cell lines expressing 0, 2, 4, or 6 μg of the myc-rGSK3β(K85R), respectively (Fig. 1B).

β-catenin is localized to the nucleus in mammary cells expressing myc-rGSK3β(K85R). When Wnt signaling is activated, free β-catenin is translocated into the nucleus to stimulate transcription of proto-oncogenes, such as cyclin D1 and c-myc. To investigate if myc-rGSK3β(K85R) can influence the translocation of β-catenin from the cytoplasm to the nucleus, cytoplasmic and nuclear extracts were prepared from the transfected C57MG cells. In the control cells transfected with the empty vector alone, the levels of β-catenin protein in the cytoplasm were higher than in the nucleus. In contrast, in the presence of increasing amounts of myc-rGSK3β(K85R), the levels of β-catenin in the nucleus were significantly higher than in the cytoplasm (Fig. 2A). Quantification showed a ratio of nuclear-to-cytoplasmic expression of 0.3, 1.9, 3.9, 3.5, 9.8, and 14.8 in cells transfected with 0, 1, 2, 4, or 6 μg of the myc-rGSK3β(K85R), respectively (Fig. 2A).

A murine kinase-inactive form of glycogen synthase kinase 3β. We cloned the full-length ORF of mGSK3β from a phage library; its sequence was identical to the sequence in Genbank (gi:7025914). To create an inactive enzyme that might function as a dominant negative for signaling in the Wnt pathway, we mutated the ATP-binding site, similar to what was done for the human GSK3β (5). In vitro translated wild-type (WT) and mutant GSK3β constructs were subjected to a kinase assay using a GSK3β substrate peptide derived from cyclic AMP–responsive element-binding protein (38). The activity of the mutant was <20% of that of the WT enzyme, consistent with its design as a kinase-inactive mutant (data not shown). Moreover, to further confirm the ability of the KI-mGSK3β construct to act as a dominant negative, we tested its ability to produce axis duplication in Xenopus embryos as has been reported for the mutated inactive human GSK3β (5). RNA for KI-mGSK3β was transcribed in vitro and injected ventroequatorially into Xenopus embryos, and these embryos developed ectopic axes consistent with activation of the Wnt pathway (data not shown).

Wnt pathway activation in cells in vitro by HA-KI-mGSK3β. We confirmed that HA-KI-mGSK3β acts in the same fashion as myc-rGSK3β(K85R). Steady-state elevation of β-catenin expression was detected in C57MG cells transiently transfected with increasing expression of HA-KI-mGSK3β (Fig. 3A). The half-life of β-catenin in the transfected cells was as long as 26.6 hours (4 μg plasmid) compared with only 2.2 hours in untransfected cells (Fig. 3B). Immunofluorescence was used to confirm nuclear translocation of the up-regulated β-catenin. C57MG cells transfected with 4 μg of the HA-KI-mGSK3β construct or empty vector were fixed and stained with a primary monoclonal antibody against β-catenin and a secondary antibody conjugated with FITC along with Hoechst dye to identify the nuclei. In the control cells (Fig. 2B, a and d), most of the β-catenin is located in the cytoplasm and in the plasma membrane. In contrast, the cells transfected with the HA-KI-mGSK3β plasmid exhibited strong nuclear staining (Fig. 2B, b and e). As a control for nonspecific fluorescence, the FITC secondary antibody was used alone on the same transiently transfected cells (Fig. 2B, c and f). Thus, the mutant HA-KI-mGSK3β acts similarly to the myc-rGSK3β(K85R) in promoting β-catenin stabilization and nuclear translocation, consistent with canonical Wnt pathway activation.

Inhibition of murine glycogen synthase kinase 3β using small interfering RNA or pharmacologic inhibitors up-regulates β-catenin expression. We compared our in vitro results with the HA-KI-mGSK3β construct to other methods of inhibiting GSK3β. C57MG cells were transiently transfected by nucleofection with SMARTpool siRNA, a pool of four specific siRNAs for mGSK3β, or siCONTROL. GSK3β proteins levels were reduced significantly in cells transfected with the siRNAs for mGSK3β compared with siRNA control or untransfected cells (Fig. 4A). There was an inverse correlation between the expression of GSK3β and β-catenin, which was up-regulated in cells transfected with GSK3β siRNA. Alternatively, C57MG cells were treated with the GSK3 inhibitors SB216763 or TDZD-8 at 20 and 5 μmol/L, respectively. As expected, the kinase inhibitors did not alter the levels of GSK3β but up-regulated β-catenin expression, consistent with inhibition of GSK activity (Fig. 4B and C). Thus, in vitro, the mutant HA-KI-mGSK3β construct acts similarly to specific GSK3β siRNA and pharmacologic inhibitors.

Cyclin D1 is up-regulated in mammary cell lines transfected with HA-KI-mGSK3β. When β-catenin translocates into the nucleus during canonical Wnt signaling, it binds the factors of the TCF/LEF family and dramatically increases their transcriptional activity, stimulating the expression of proto-oncogenes, such as cyclin D1 (13, 14). In a preliminary experiment, we used the TOPFLASH/FOPFLASH TCF/LEF luciferase reporter system (39) and showed that cotransfection of the reporter along with HA-KI-mGSK3β resulted in a 10-fold increase in luciferase activity compared with controls (data not shown). To show this for an endogenous biologically relevant gene, we measured levels of cyclin D1 in transiently transfected C57MG cells. Semiquantitative PCR with specific primers for cyclin D1 suggested that there was an increased amount of cyclin D1 mRNA in cells expressing the HA-KI-mGSK3β (data not shown). These results were confirmed using qPCR, where an increase in cyclin D1 mRNA of up to ∼7-fold was seen (Fig. 3C). Consistent with these results, the levels of cyclin D1 protein were higher in cells expressing HA-KI-mGSK3β compared with vector-transfected cells (Fig. 3A, bottom). Similar results were obtained with the myc-rGSK3β(K85R) construct (Fig. 1C), suggesting that kinase-inactive forms of GSK3β are able to promote complete and functional Wnt signaling, including up-regulation of target genes.

Transgenic expression of kinase-inactive murine glycogen synthase kinase 3β in the mouse mammary gland. The KI-mGSK3β construct was subcloned into a MMTV-LTR vector (40), designed for hormone-dependent transgene expression in the adult mammary gland, and microinjected into fertilized FVB/N mouse oocytes. Three founders were identified by Southern blotting; their offspring had similar expression and phenotypes, so the data were pooled for all three lines. To verify the expression of the transgene mRNA, a transgene-specific RT-PCR assay was employed. During pregnancy, which activates transgene expression, the mammary glands and other epithelial tissues, including kidney, small intestine, salivary gland, and spleen, expressed transgene-specific transcripts (data not shown), a pattern of expression has been seen with other MMTV transgenes (35). To determine whether the transgene protein is functional in vivo as a Wnt pathway activator, we assayed for expression of GSK3β and β-catenin protein in the mammary gland. Although the untagged kinase-inactive protein could not be distinguished from the WT kinase by immunoblot, we found an increase in total GSK3β protein in the mammary gland of the transgenic compared with WT mice along with a significant increase in β-catenin protein, consistent with Wnt pathway activation by the transgene (Fig. 5A).

Mammary tumors in MMTV-KI-mGSK3β transgenic mice. MMTV-KI-mGSK3β mice develop and breed normally. To promote transgene expression from the hormone-dependent MMTV-LTR, female mice were continuously bred and pups were removed after 7 days of lactation. A cohort of 117 transgenic female mice derived from the three independent transgenic lines was observed for 2 years. Sixty-two percent of the mice developed mammary tumors at a median age of 22 months, with no significant difference in incidence among the lines. Although they occur long after pregnancy, the tumors continue to express the transgene (Fig. 5B). The pooled mammary tumor incidence is illustrated in a Kaplan-Meier plot (Fig. 5C). The tumors and other mammary glands and organs were harvested for histologic and molecular analyses.

Detailed histopathological analyses were done on 54 of the female transgenic mice that had evidence of mammary tumors (Table 1; Fig. 6). Most of the tumors were adenocarcinomas (Fig. 6A), including variants such as papillary carcinomas (Fig. 6B), and the tumors were frequently associated with invasive growth (Fig. 6C). The most common histologic subtypes were pilar (squamous) tumors (n = 13), papillary tumors (n = 12) typically with micropapillary components, glandular tumors (n = 8), and myoepithelial tumors (n = 8 spindle cell tumors and n = 1 adenomyoepithelioma). The tumors tended to be stroma rich, to contain inflammatory infiltrates, and to keratinize. Immunohistochemistry was negative for estrogen and progesterone receptors, except in the spindle cell tumors, which had perinuclear estrogen receptor staining and were also positive for smooth muscle actin (data not shown). Immunofluorescence for cytokeratin 1 (Fig. 6E), cytokeratin 5, cytokeratin 6, and hair keratin (not shown) confirmed transdifferentiation into epidermal and pilar structures. Other mice had hyperplastic and dysplastic mammary lesions without tumors. Twenty-one mice had other malignancies, including lymphomas (n = 8), leukemias (n = 6), bronchoalveolar lung tumors (n = 5), and hepatoma (n = 2). The incidence of nonmammary neoplasms was very similar to that reported for WT mice of the same FVB/N strain (41) and thus most likely does not result from an effect of the transgene.

Transgenic expression of kinase-inactive murine glycogen synthase kinase 3β up-regulates β-catenin expression and cyclin D1 in vivo. To confirm that tumorigenesis in the MMTV-KI-mGSK3β was occurring in association with activation of Wnt signaling, we assayed the expression levels of β-catenin in mammary glands and tumor tissues from the transgenic mice. β-catenin protein was up-regulated in the tumor samples in six of seven transgenic mice (Fig. 7A, quantification in Fig. 7B). In these tumors, cyclin D1 was also up-regulated (Fig. 7A).

Using qPCR, we compared expression of β-catenin and cyclin D1 in the mammary glands of WT female FVB/N mice and MMTV-KI-mGSK3β transgenics. We found that the presence of the transgene resulted in a detectable increase in cyclin D1 mRNA (Fig. 7C, black columns) but not β-catenin mRNA (Fig. 7C, gray columns) in the premalignant mammary gland as well as in malignant mammary gland.

We have engineered a murine KI-GSK3β by altering three residues (85-87) in the ATP-binding site of mGSK3β, similar to what was done for the human GSK3β (5). We compared it to a previously described GSK3β(K85R) construct that has been shown to activate Wnt signaling in rat-1 fibroblasts and PC12 cells (42). The construct had minimal kinase activity against a peptide substrate, consistent with its design as a kinase-inactive mutant. To confirm its ability to act as a dominant negative and stimulate Wnt signaling in the mammary epithelium, we tested the construct both in vitro and in vivo. In vitro, we used C57MG cells, a nonmalignant cell line derived from murine breast epithelial cells. These cells are responsive to Wnt signaling; expression of Wnt in C57MG cells causes morphologic transformation and apparent loss of contact inhibition of cell growth (43). We studied the expression, half life, and subcellular localization of β-catenin and other target proteins and genes in C57MG cells transfected with increasing amounts of HA-KI-mGSK3β. We found that KI-mGSK3β stabilizes β-catenin expression and promotes its translocalization to the nucleus. We also detected morphologic differences in C57MG cells transfected with HA-KI-mGSK3β; the cells were less spread out and there was a reduction in membrane staining of β-catenin probably because of diminished interaction between β-catenin and E-cadherin in those cells. The previously reported myc-rGSK3β(K85R) was used in the same assays and gave similar results (4). We compared these findings to other known methods for inhibiting GSK3β, such as siRNA and pharmacologic inhibitors, and obtained similar results. The key role of GSK3β as a negative regulator of Wnt signaling has also been shown using siRNA in mouse P19 cells (44).

In vitro, the HA-KI-mGSK3β promoted Wnt-dependent transcription based on its ability to activate a TCF/LEF-dependent reporter and to up-regulate cyclin D1 mRNA and protein. Cyclin D1 overexpression has been found in 50% of patients with breast cancer, but only 15% to 20% of these cases show gene amplification of cyclin D1 (45–47). Other mechanisms, such as up-regulation of gene transcription and translation, also play roles in overexpression of cyclin D1 in breast cancer; in fact, GSK3β has been shown to directly regulate cyclin D1 turnover in vitro (48). β-catenin transactivation is correlated significantly with cyclin D1 overexpression in both 8 breast cancer cell lines and 123 primary breast cancer specimens (27).

In vivo, KI-mGSK3β caused dorsal axis duplication in Xenopus embryos as shown previously for inactive Xenopus, rat, and human GSK3β (4–6). Based on these promising results, we then established transgenic mice overexpressing KI-mGSK3β under the control of the MMTV-LTR. Other investigators have developed transgenic models using the WT enzyme to study its role in phosphorylation of tau in the brain (49, 50), regulation of glucose metabolism in muscle (51), and its role in cardiac development (52, 53).

Using the MMTV promoter, KI-mGSK3β was expressed in the adult mammary gland in response to steroid hormones and in some other epithelial tissues and T lymphocytes of pregnant mice. This pattern of expression fits with what has been seen with other MMTV transgenes (35). However, in spite of its expression in several epithelial tissues, it causes primarily mammary tumors in FVB/N mice. More than 60% of the female KI-mGSK3β mice developed mammary tumors at a median age of 22 months. The histology and pattern of cytokeratin expression of these tumors has been compared with other murine mammary tumors and is similar to that of tumors due to other mutations in the Wnt signaling pathway, including mice with Wnt 1, Wnt 10b, and β-catenin transgenes and APC gene mutations (37, 54). Characteristic for these Wnt pathway-induced mammary tumors are ductular architecture, well-developed stroma, myoepithelial, acinar, or glandular differentiation, and squamous metaplasia (37). As in other mouse models with Wnt pathway activation, some KI-GSK3β transgenic tumors showed transdifferentiation into epidermal and pilar structures accompanied by typical cytokeratin and hair keratin expression (54). This histologic determination is supported by our molecular results, as expression of KI-mGSK3β leads to up-regulation of β-catenin and cyclin D1.

Although tumors developing in the KI-mGSK3β mice could theoretically result from activation of pathways other than the canonical Wnt pathway, the demonstration of up-regulation of β-catenin and the transcriptional up-regulation of the Wnt target cyclin D1 in the transgenic tumors is a strong evidence that Wnt pathway activation is a major effect of the transgene, consistent with its ability to mediate this in cells in vitro. Moreover, constitutive expression of Akt, in another signaling pathway that is inhibited by GSK3β, causes delayed mammary involution but not mammary tumors when it is expressed in the mammary gland using the MMTV promoter (55, 56). This supports our contention that KI-mGSK3β is acting through the Wnt pathway.

Thus, our experiments show that the KI-GSK3β can promote mammary tumorigenesis. Other mutant Wnt genes, now well accepted to be important in human tumorigenesis, were first identified in animal models. Wnt-1 itself was cloned as a common insertion site for MMTV in murine mammary tumors (57); mutation of APC in intestinal polyps and cancers was first found as an ethylnitrosourea-induced mutation in APC^min mice (58). The current study suggests that GSK3β has the capability to be a tumor suppressor, and mutations could be sought in human specimens from breast and other cancers. In addition, as inhibitors of GSK3β enter clinical trials for treatment of diabetes, consideration should be given to the possibility that such drugs might up-regulate Wnt signaling and promote mammary or other tumors.

Grant support: National Institute of Environmental Health Sciences grant P01 ES11624 (D.C. Seldin); U.S. Army Medical Research and Materiel Command predoctoral award W8IXWH-04-1-0375 and Norman G. Levinsky Memorial Fellowship program (M. Farago); Department of Medicine, Boston University School of Medicine pilot research grant and American Cancer Society institutional training grant (I. Dominguez); and Deutsche Forschungsgemeinschaft grant BA 1433/4-1 (A. Rosner).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

We thank Sandip Patel for technical assistance with the development of the KI-mGSK3β transgenic mice, Jessica Murray for assistance in statistical analyses, Patrick Hogan for animal care, Diane Song for providing plasmids and for help with the reporter assay, J.E. Walls for assistance with the immunohistochemistry, Dr. K. Miyoshi for carrying out the cytokeratin immunofluorescence, Drs. G. Shyamala and T-T. Sun for kindly providing the antibodies, and the members of the laboratory for helpful discussions and review of the article.