Wnt/β-catenin Signaling Contributes to Tumor Malignancy and Is Targetable in Gastrointestinal Stromal Tumor

Gastrointestinal stromal tumor (GIST) is the most common subtype of human sarcoma and typically arises from the stomach or small intestine (1). Most GISTs have an activating mutation in either KIT or PDGFRA (2, 3). Targeting the KIT or PDGFRA oncogenic proteins with the selective tyrosine kinase inhibitor imatinib mesylate has markedly increased survival in advanced GIST (4). Although despite its efficacy, imatinib is rarely curative. The median time to resistance to imatinib is 2 years, most often due to secondary KIT mutations (5).

The molecular basis for aggressive tumor biology in GIST remains poorly defined. For instance, the criteria for assessing the risk of recurrence following the resection of a localized, primary GIST are based on the pathologic factors of tumor mitotic activity, size, and location, while tumor KIT mutation status is not an independent predictor of recurrence, implying that other signaling processes may be involved (6). Multiple regulatory proteins or molecular alterations have been associated with aggressive phenotypes in GIST (7). For example, expression of miR-196a was associated with poor survival and malignant progression in GIST (8) and amplification of cell-cycle genes correlated with increased malignant potential in GIST (9). Deletions of 9p targeting CDKN2A are common in GIST, and loss of p16 has been demonstrated in high-risk GIST (10, 11). In addition, the CINSARC gene signature is linked to expression of p16 and RB1 with increased risk of metastatic disease (12). A subpopulation of GISTs with wild-type KIT and PDGFRA instead has a BRAF mutation (13) or succinate dehydrogenase (SDH) deficiency (14). Very recently, loss of myc-associated protein (MAX) was shown to be an early genomic event of GIST progression (15). A greater understanding of the mechanisms of GIST tumor malignancy is necessary to identify additional therapeutic targets given the limitations of current tyrosine kinase inhibitors once resistance develops.

Wnt pathway components are well known to contribute to tumor progression and metastasis in a variety of cancers (16–18). Wnt signaling includes canonical and noncanonical pathways. Canonical Wnt signaling is initiated by Wnt ligands binding to Frizzled (FZD)/low-density lipoprotein receptor-related protein (LRP) receptor complexes, which inactivate the GSK3β–Axin–APC destruction complex, leading to signaling via the β-catenin/T-cell factor (TCF) pathway (19). The dickkopf-related protein (DKK) family comprises secreted antagonists of Wnt signaling that bind and block the Wnt receptors LRP5/6. Several DKKs have been reported to be downregulated in various cancers, and restoration of their expression can inhibit cell proliferation and tumor growth (17, 20). Wnt/β-catenin signaling plays an important role in tumor initiation and progression through regulation of the cell cycle, stem cell self-renewal, and epithelial-to-mesenchymal transition (19, 21). Aberrant accumulation of nuclear β-catenin is the hallmark of Wnt activation. Stabilized β-catenin associated with its cotranscriptional factor TCF/lymphoid enhancer factor (TCF/LEF) in the nucleus triggers upregulation of proto-oncogenes, such as MYC and cyclin D1. Multiple pathways have been reported to regulate the nuclear translocation and stability of β-catenin (22, 23). The role of Wnt/β-catenin signaling in GIST has not been defined.

In this study, we showed that Wnt/β-catenin components were overexpressed in a subset of human GISTs. Consistently, we found aberrant accumulation of nuclear β-catenin in patient-derived xenografts (PDX) and murine imatinib-resistant models. DKK4 mRNA expression was reduced in most metastases relative to primary, untreated tumors, and overexpression of DKK4 inhibited the Wnt/β-catenin signaling in vitro. Furthermore, COP1 (photomorphogenesis protein 1), a tumor suppressor and E3 ligase, was also involved in the regulation of nuclear β-catenin stability and its transcriptional activity in GIST cells. Finally, we showed that Wnt signaling was required for tumor survival in multiple preclinical GIST models, both in imatinib-naïve and imatinib-resistant GISTs. Therefore, our findings highlight the significant role of Wnt/β-catenin activation in GIST and provide the rationale for targeting Wnt/β-catenin signaling as a potential therapeutic strategy for GIST.

IRB approval was obtained for all human tissue studies. A human GIST tissue paraffin microarray was constructed containing more than 100 archival specimens from patients who underwent resection at Memorial Sloan Kettering Cancer Center (New York, NY) between 1984 and 2004. We also obtained 46 more recent GIST specimens. Mitotic rate (number of mitoses per 50 high-power fields) and mutation status for each specimen were recorded. Single-cell suspensions were generated from fresh tumors after mincing and digestion with 10 mg/mL collagenase type II (Worthington Biochemicals) plus 50 μg/mL DNAse I (Roche Diagnostics) in Ca²⁺ and Mg²⁺-free HBSS (Invitrogen) at 37°C for 30 minutes. The tumor suspensions were filtered through a 100-μm cell strainer and mixed with an equal volume of 100% FBS to quench the collagenase. In some cases, KIT⁺ cells were purified using human CD117 microbeads (Miltenyi Biotec) to a purity that exceeded 90% by flow cytometry.

Human HG130 was derived from an imatinib-resistant liver metastasis that contained KIT exon 13 V654A and KIT exon 11 W557R mutations. Human HG324 was derived from an imatinib-resistant peritoneal GIST metastasis containing a KRAS G12V and SDHB mutation. Bulk tumor cells were freshly isolated from tumors using collagenase as above and injected into the flanks of NSG mice.

The human GIST882 cell line (KIT exon 13 mutant) was kindly provided by Dr. Jonathan Fletcher (Brigham and Woman's Hospital, Dana-Farber Cancer Institute, Boston, MA 02215). Human GIST T1 and murine S2 GIST cells (both KIT exon 11 mutant) have been described previously (24, 25). We established the HG129 cell line from a primary, untreated gastric GIST harboring a KIT exon 11 mutation in 2012 (26). Human HG209 was derived from an imatinib- and sunitinib-resistant peritoneal metastasis and has a KIT exon 11 YIDPTQL 570-576 deletion and a KIT exon 17 point mutation (C. 2446 C>G p. D816H) in 2013. Cell lines were cultured in RPMI1640 supplemented with 10% FBS and 1% penicillin/streptomycin. PKF118-310, XAV939, and G007-LK were purchased from Calbiochem. Imatinib was purchased from LC Laboratories. MG132 and cycloheximide were from Sigma-Aldrich. A Dojindo assay (Cell Counting Kit 8, Dojindo) was carried out at 72 hours to measure the cell viability after incubation of 1 × 10⁴ cells in 96-well plates with PKF118-310, XAV939, G007-LK, imatinib, or media plus 0.5% DMSO.

Total RNA was extracted from human GIST tissues or cells, reverse transcribed, and amplified with PCR TaqMan probes for human DKK4 (Hs00205290_ml), FZD6 (Hs00171574_ml), FZD7 (Hs00275833_sl), LRP6 (Hs00233945_ml), CCND1 (Hs00765553_ml), CTNNB1 (Hs00355049_ml), Axin2 (Hs00610344_ml), TCF4 (Hs00162613_ml), and GAPDH (Applied Biosystems). Quantitative PCR was performed using a ViiA7 real-time PCR system (Applied Biosystems). Data were calculated by the 2^−ΔΔC_t method as described in the manufacturer's instructions and were expressed as fold increase over the indicated controls.

For transient COP1 knockdown or β-catenin knockdown, GIST T1 cells were transfected with 30 nmol/L of On-Target plus SMARTpool siRNA for human COP1 (E-007949-00), On-Target plus SMARTpool siRNA for human β-catenin (L-003482-00), or a nontarget control siRNA (D-001810-10-05; Thermo Scientific) using Lipofectamine RNAiMAX (Invitrogen) for 48 hours. To generate overexpressing cells, GIST T1 were transfected with empty control vector (pCMV6-mock, PS100001), human β-catenin (pCMV6-β-catenin, RC208947), DKK4 (pCMV6-DKK4, RC221217), or COP1 plasmid (pCMV6-COP1, RC210492, all from Origene) using Lipofectamine 2000 (Invitrogen) according to the manufacturer's protocol.

RNA was isolated from human KIT⁺ cells using the RNeasy Plus Mini Kit (Qiagen). Gene expression microarray analysis was performed by the Genomics Core Laboratory of Sloan Kettering Institute using Human Genome U133A 2.0 microarrays (Affymetrix). Microarray data were analyzed using Partek Genomics Suite version 6.5. After log transformation and quantile normalization, ANOVA was performed to compare multiple groups. Statistically significant genes with an FDR <0.05 were selected and analyzed using KEGG Pathway Analysis software (Ingenuity Systems).

Formalin-fixed and paraffin-embedded human tumors, xenograft tumors, and tumors from Kit^V558Δ/+ mice were sectioned at 5-μm thickness. Antigen retrieval was achieved with citrate buffer. IHC was performed using the following antibodies and the corresponding isotype control IgG: anti-human β-catenin IgG (Clone β-catenin 1, DAKO), Ki-67 (Vector Laboratories), and β-catenin (6B3, Cell Signaling Technology).

Western blot of whole protein lysates or nuclear proteins from frozen tumor tissues or cells was performed as described previously (24). Antibodies for cleaved caspase-3 (Asp175; 5A1E), cleaved PARP (Asp214; D64E10), cyclin D1 (92G2), β-catenin (6B3), active β-catenin (D13A1), TCF4 (C48H11), axin 1 (C95H11), tubulin (DM1A), and GAPDH (D16H11) were purchased from Cell Signaling Technology. Other antibodies included lamin B1 (ab16048, Abcam), COP1 (ab56400, Abcam), tankyrase1/2 (H-350, Santa Cruz Biotechnology), and active β-catenin (Clone 8E7, Millipore). ImageJ software was used to measure the relative density for signaling expression.

GIST T1 cells were plated in a 12-well plate (1 × 10⁵ cells/well) and transfected with 0.5 μg of a TCF/LEF reporter plasmid or a negative reporter plasmid (CCS-018L, Qiagen) using Lipofectamine 2000 (Invitrogen) combined with or without an expression vector for β-catenin (pCMV6-β-catenin) or mock control. After 48 hours of transfection, cells were treated with recombinant 150 ng/mL human Wnt3a (R&D Systems) for 4 hours. Cell lysates were analyzed using the Dual-Glo Luciferase Assay System (Promega). Firefly luciferase activity was normalized with an internal control (Renilla luciferase activity) for transfection efficiency. Relative TCF/LEF reporter activity was expressed as fold change over the indicated control.

GIST T1 cells were transiently transfected with control siRNA or COP1 SMARTpool siRNA for 48 hours. Cells were collected after the addition of 200 μg/mL cycloheximide at the indicated time points (0, 1, 2, 4, and 6 hours) with or without 10 μmol/L MG132. Cell lysates from nuclear extracts were used for signaling. Quantification was achieved by ImageJ software. The final β-catenin turnover rate at each time point is the percentage of β-catenin/lamin B1 expressed relative to the t = 0 hour control siRNA.

For PKF118-310 studies in a xenograft model, human GIST T1 cells or HG129 cells (1 × 10⁶) were injected subcutaneously into the flanks of NSG mice. Murine S2 cells were injected into the flanks of C57BL/6J mice (The Jackson Laboratory). After the tumors reached 100 mm³, PKF118-310 (0.8 mg/kg) or vehicle 0.01% DMSO in PBS was injected intratumorally every other day for 12 days. Because PKF118-310 is toxic when administered systemically, we also investigated the in vivo efficacy of G007-LK, another Wnt pathway inhibitor, on tumor growth. The age- and sex-matched Kit^V558Δ/+ mice were treated with imatinib (600 mg/L in drinking water), G007-LK (28.5 mg/kg in MCT solution given by daily intraperitoneal injection; ref. 27), or imatinib plus G007-LK. Control mice received regular drinking water and daily intraperitoneal injection of MCT solution (0.5% methyl cellulose plus 0.2% Tween 80 in sterile water). Mice were sacrificed after 2 weeks of treatment. All mouse studies were approved by the Institutional Animal Care and Use Committee.

Data are expressed as mean ± SEM or median. Differences were detected by unpaired two-tailed Student t tests unless otherwise indicated using Prism 6.0 software (GraphPad Software). P < 0.05 was considered significant.

To identify genes and pathways that might drive GIST tumor malignancy, we compared the transcriptome of freshly isolated KIT⁺ tumor cells from 4 metastatic, resistant tumors to 8 primary, untreated tumors. There were 462 statistically significant genes with an FDR of <0.05 and a fold change >2.0. KEGG pathway analysis revealed differences in cell cycle, DNA replication, and the Wnt signaling pathway (Supplementary Table S1). Specifically, metastatic, resistant tumors (Met/Res) had lower expression of the Wnt inhibitor DKK4 (18.8-fold change) and higher expression of the Frizzled 5 ligand Wnt5A compared to primary, untreated tumors (Prim/UT; 8.6-fold change; Supplementary Table S2). Lack of DKK4 has been shown to promote canonical Wnt signaling and tumor invasion in hepatocellular carcinoma (20). We validated the DKK4 findings in KIT⁺ tumor cells (Supplementary Fig. S1A) and in 46 bulk human GIST tumors by quantitative PCR (Fig. 1A, left). Meanwhile, we found that β-catenin mRNA (CTNNB1) was highly expressed in all GISTs, with a similar magnitude as GAPDH mRNA, and there was no difference between primary, untreated, and metastatic, resistant tumors (Fig. 1A, right). In addition, we found that CTNNB1 was expressed to a greater extent in freshly isolated KIT⁺ tumor cells from 3 patients compared with KIT⁻ (i.e., nontumor) cells (Supplementary Fig. S1B), further supporting the potential importance of β-catenin in GIST tumorigenesis.

To evaluate the potential role of Wnt/β-catenin signaling in GIST biology, we first examined β-catenin staining in a human GIST tissue paraffin microarray comprising more than 100 archival specimens, many of which were from patients who underwent surgery prior to the advent of tyrosine kinase inhibitors. We found high β-catenin staining density in many metastatic tumors, regardless of whether they were untreated or imatinib resistant (Fig. 1B). In addition, we also analyzed the intensity of β-catenin staining in whole tissue sections of 46 additional human GIST specimens (Fig. 1C). There was low or moderate β-catenin staining in many of the primary, untreated tumors. In contrast, a greater percentage of metastatic, imatinib-resistant specimens had high β-catenin staining (Fig. 1D). To determine whether β-catenin was present in the nucleus, we isolated nuclear protein extracts from frozen human GISTs that had high β-catenin staining by IHC. We found that active β-catenin, which was specifically detected by dephosphorylated Ser37 and Thr41 (28), was present in the nuclear extracts of selected primary, untreated, and metastatic, imatinib-resistant tumors (Fig. 1E). Furthermore, we performed quantitative PCR for several Wnt coreceptor and ligand genes in human GIST cell lines as well as freshly isolated KIT⁺ cells from 8 selected human GISTs, including primary, untreated, and metastatic, imatinib-resistant or sensitive tumors. The expression of multiple coreceptor genes of Wnt (FZD6, FZD7, and LRP6) was variable, but the majority of specimens had similar or higher expression than the Caco-2 colorectal adenocarcinoma cell line, which has a high level of Wnt activation (Fig. 1F).

To further assess the role of the Wnt/β-catenin signaling in GIST, we generated PDXs from several patients with metastatic, imatinib-resistant GISTs. Because we have not been able to generate any primary, untreated PDXs thus far, we used GIST T1 xenografts as a comparison because they also expressed highly β-catenin and grow aggressively in vivo. Although β-catenin tended to localize mostly to the membrane and cytoplasm of GIST T1 tumors, it was extensively expressed in the nuclei of the PDX tumors (Fig. 2A). The amount of active nuclear β-catenin was also higher in the imatinib-resistant PDX models compared with GIST T1 xenografts (Fig. 2B). TCF4 protein was also higher in the PDX models. Upregulation of the Wnt target genes cyclin D1 and downregulation of DKK4 were also seen in the PDX models compared with GIST T1 xenografts by quantitative PCR (Fig. 2C), suggesting that activation of the canonical Wnt pathway could contribute to tumor malignancy in the PDX models.

We showed similar findings in the murine imatinib-resistant GIST cell line S2, which we previously derived from a Kit^V558Δ/+ mouse that develops GIST (24). We established an orthotopic liver model of GIST metastasis (Supplementary Fig. S2A). Compared with the original Kit^V558Δ/+ tumor, the S2 liver tumors had lower Kit expression, yet high DOG1 expression (Supplementary Fig. S2B), and similar levels of downstream Kit signaling, such as phospho-STAT3, ERK, and S6 (Supplementary Fig. S2C). There was high nuclear β-catenin staining in murine S2 tumors. In contrast, the Kit^V558Δ/+ mouse tumor had mostly membranous β-catenin staining by IHC (Fig. 2D). By using nuclear and cytoplasmic protein extraction, we detected greater β-catenin nuclear translocation and dephosphorylation in S2 tumors compared with Kit^V558Δ/+ mouse tumors (Fig. 2E). Taken together, these data suggested that there was active Wnt/β-catenin signaling in multiple models of murine and human GIST, as we had observed in the human GIST specimens.

Wnt ligands stimulate canonical Wnt signaling by inactivating the GSK3β–Axin–APC destruction complex, thereby activating the β-catenin/TCF pathway. To test the biological effect of Wnt activation in GIST cells, we stimulated human GIST cell lines and the murine S2 line in vitro with recombinant Wnt3a. We found that active β-catenin was significantly increased (Fig. 3A). The Wnt responsive genes Axin2 and CCND1 were also induced by Wnt3a stimulation (Fig. 3B). To further assess the effect of Wnt3a on β-catenin transcriptional activity, we transfected GIST T1 cells, which have higher TCF4 levels compared with GIST882 cells (Fig. 3A), with a luciferase reporter linked to a TCF/LEF promoter. In addition, we also transfected GIST T1 cells with a mock or a β-catenin expression plasmid. We found that TCF/LEF reporter activity was increased by β-catenin overexpression. The effect was augmented in the presence of Wnt3a (Fig. 3C). As DKK4 mRNA was higher in most primary, untreated GISTs compared with most metastatic, resistant GISTs (Fig. 1A), we overexpressed DKK4 in GIST T1 cells to determine its effect on β-catenin. DKK4 overexpression was confirmed by detection of the vector construct with a FLAG antibody. We found that DKK4 overexpression reduced β-catenin protein levels by 33% (Fig. 3D). Consistent with reduced β-catenin, GSK3β phosphorylated at serine 9, which inhibits formation of the destruction complex (19), was also reduced. Ectopic expression of DKK4 reduced active β-catenin and cyclin D1 levels in the presence of Wnt3a stimulation. Together, these in vitro data demonstrated that Wnt ligands induced β-catenin signaling, while DKK4 antagonized it in GIST cells. Finally, to identify the functional significance of β-catenin expression in GIST cells, we transfected the GIST T1 cells with control siRNA or β-catenin SMARTpool siRNA. Knockdown of β-catenin also decreased cell viability at 72 hours (Fig. 3E).

β-Catenin is unstable due to ubiquitin-mediated proteasomal degradation (19). To address the role of proteasomal degradation of nuclear β-catenin in GIST, we treated GIST T1 cells for 6 hours with the proteasome inhibitor MG132. Upon proteasome inhibition, nuclear dephosphorylated and total β-catenin both stabilized, as expected (Fig. 4A). Incubation of GIST T1 cells with the protein synthesis inhibitor cycloheximide for 6 hours decreased both active and total β-catenin in the nuclei. However, the addition of MG132 restored the accumulation of nuclear active and total β-catenin in GIST T1 cells, indicating that nuclear β-catenin was undergoing proteasomal degradation (Fig. 4B). Several E3 ubiquitin ligases have been shown to promote nuclear β-catenin degradation, such as c-cbl, TRIM33, and COP1 (22, 23, 29). In human GIST specimens, TRIM33 was almost undetectable by Western blot analysis (Supplementary Fig. S3), while COP1 was highly expressed in primary, untreated GISTs, but reduced in many metastatic, resistant GISTs (Fig. 4C). Furthermore, COP1 was highly expressed in GIST T1 cells, especially in the nucleus (Fig. 4D), whereas c-cbl and TRIM33 expression was not detectable. Notably, COP1 functions as an E3 ligase for ETV1 and ETV4 (30, 31). ETV1 is a lineage transcription factor in GIST (32), and we previously explored a critical role of ETV4 in GIST (33). COP1 knockdown with SMARTpool siRNA in GIST T1 cells stabilized active nuclear and total β-catenin levels (Fig. 4D). Consistently, COP1 knockdown significantly increased TCF/LEF reporter activity compared with control siRNA, indicating that COP1 loss facilitated β-catenin–mediated transcription (Fig. 4E). A cycloheximide chase assay further confirmed that COP1 knockdown significantly increased the half-life of nuclear β-catenin (Fig. 4F). To assess whether COP1 binds to β-catenin in GIST T1 cells, we performed COP1 immunoprecipitation (Fig. 4G). There was a direct interaction between endogenous COP1 and β-catenin. Furthermore, MG132 increased the amount of β-catenin that was coimmunoprecipitated with COP1. The binding specificity was confirmed by COP1 knockdown and control IgG immunoprecipitation. Finally, GIST T1 cells were transfected with ectopic COP1. Overexpressing COP1 alone did not decrease the endogenous β-catenin level in GIST T1 cells, suggesting that there was sufficient endogenous β-catenin turnover. However, in the setting of increased β-catenin transcriptional activation induced by ectopic β-catenin, COP1 overexpression inhibited active β-catenin and cyclin D1 (Fig. 4H). Overall, these data suggested that β-catenin stability and transcriptional activity are partially regulated by COP1 in GIST.

To investigate the importance of Wnt signaling in GIST cell survival, we tested the effect of the Wnt inhibitor PKF118-310, a specific inhibitor that disrupts the β-catenin/TCF complex (34). PKF118-310 significantly inhibited the viability of imatinib-sensitive (T1, HG129, and GIST882) and imatinib-resistant (HG209) cells at low IC₅₀ values (Fig. 5A), but did not affect the in vitro viability of a murine hepatocellular carcinoma cell line or B16 melanoma (Supplementary Fig. S4). The mechanism depended on apoptosis as shown by activation of caspase-3 and PARP (Fig. 5B). The Wnt/β-catenin target gene cyclin D1 was markedly reduced, while total β-catenin and TCF4 levels were relatively unaffected. To determine whether Wnt inhibition affected tumor growth in vivo, we treated established flank tumor xenografts of GIST T1 cells and HG129 cells with intratumoral injection of PKF118-310 and found significant tumor regression (Fig. 5C). We confirmed these findings in subcutaneous tumors of the imatinib-resistant murine S2 GIST cell line as PKF118-310 treatment reduced tumor growth (Fig. 5D), tumor β-catenin staining (Fig. 5E), and cyclin D1 expression (Fig. 5F). Thus, targeting β-catenin/TCF was sufficient to inhibit GIST growth in vitro and in vivo in imatinib-sensitive and resistant GIST models.

We next considered whether inhibition of Wnt/β-catenin would enhance the antitumoral effect of imatinib in GIST. Imatinib treatment increased the transcription of CTNNB1 (β-catenin) and Axin2 mRNA in Kit^V558Δ/+ tumors (Supplementary Fig. S5A). Although imatinib reduced active β-catenin protein levels at 1 week of treatment, its expression was restored after an additional week of therapy (Supplementary Fig. S5B), making it an attractive target in addition to KIT. Therefore, we tested whether Wnt inhibition increased the sensitivity of GIST cells to imatinib. Indeed, treatment with PKF118-310 and imatinib further reduced cell viability compared with imatinib alone in the imatinib-sensitive cell lines GIST T1 and HG129 (Fig. 6A, left). We also studied the effect of XAV939 and GK007-LK, which antagonize Wnt signaling by inhibiting tankyrase1/2 (TNKS1/2) and stabilizing axins through the prevention of PARsylation and ubiquitination (27, 35). Similarly, combination therapy with imatinib reduced cell viability in most cases (Fig. 6A, middle and right). We confirmed that tankyrase inhibition increased Axin1 protein levels and found that dual inhibition of Wnt and KIT signaling led to more apoptosis compared with single-drug treatment, as measured by cleaved caspase-3 and PARP (Fig. 6B). As G007-LK can be administered systemically (unlike PKF118-310) and was effective in treating both xenografts and genetically engineered models of colorectal cancer (27), we investigated its efficacy in vivo and synergism with imatinib in genetically engineered Kit^V558Δ/+ mice. G007-LK alone had substantial antitumor activity, which was further increased by combination with imatinib as measured by tumor weight (Fig. 6C). Combination therapy further reduced TNKS1/2, cyclin D1, and phospho-KIT at 2 weeks (Fig. 6D), decreased cellular proliferation as measured by Ki-67 staining, and increased tumor destruction by H&E (Fig. 6E). Taken together, antagonizing Wnt signaling augmented the antitumoral effect of imatinib in vitro and in vivo.

Although aberrant activation of the Wnt/β-catenin pathway has been linked to tumor progression in a variety of cancers (16, 18, 36), its role in GIST is unclear. We identified the Wnt/β-catenin pathway on KEGG pathway analysis comparing mRNA expression of freshly isolated KIT⁺ tumor cells from primary, untreated human GISTs with low mitotic rates to metastatic, resistant tumors with high mitotic rates. We observed that most primary, untreated human GISTs had low staining, while many metastatic tumors (untreated or imatinib resistant) had high levels of β-catenin. Notably, elevated expression of cytoplasmic β-catenin alone has been linked to tumor progression and poor outcome in some human cancers (37). Previously, it was reported that GISTs lacked nuclear β-catenin by IHC, distinguishing it from mesenteric fibromatosis (38), although different techniques may explain the discrepancy with our data. Moreover, in a broad survey of 45 tumors and 23 cell lines, active β-catenin was found to be common in sarcoma and was present in the single human GIST specimen that was tested (39). It remains to be shown whether use of different fixatives or time to tissue fixation affects the quality of β-catenin staining by IHC.

Wnt/β-catenin signaling can be activated in a variety of ways in cancer. In 140 human GISTs analyzed for mutations in over 300 commonly mutated genes in cancer, we have not identified a β-catenin mutation, and only 1 patient had an APC mutation (unpublished data). Instead, we found that the Wnt pathway antagonist DKK4 was substantially lower in metastatic, resistant GISTs compared with primary, untreated GISTs. We also detected other critical components of Wnt/β-catenin signaling in human and murine GISTs, including frizzled receptors, TCF4 and LRP6.

To prove functional Wnt/β-catenin signaling, we performed a number of experiments in human GIST T1 cells. Wnt3a increased TCF/LEF reporter activity. Meanwhile, ectopic expression of DKK4 partially reduced total β-catenin protein levels and inhibitory phospho-GSK3β. In the presence of Wnt3a, DKK4 also reduced active β-catenin and cyclin D1. Epigenetic silencing of Wnt antagonism and aberrant DNA methylation of Wnt pathway genes have been linked to poor prognosis in some cancers (40–42). Several studies have shown that concurrent hypermethylation of multiple Wnt/β-catenin inhibitors leads to gene silencing and contributes to aberrant Wnt pathway activation (43). Further investigation is needed to delineate the mechanism of DKK4 inactivation in GIST.

Although activation of Wnt/β-catenin signaling can occur through a variety of mechanisms, nuclear β-catenin stabilization is essential for its oncogenic function. We showed that proteasome inhibition stabilized nuclear dephosphorylated and total β-catenin and increased TCF/LEF reporter activity, demonstrating that the ubiquitin–proteasome pathway contributes to the regulation of Wnt/β-catenin signaling. Recently, several E3 ligases have been shown to control transcriptionally active nuclear β-catenin (22, 44). Among them, TRIMM33 was almost undetectable in most human GIST specimens and cell lines. COP1 was of particular interest, as it was expressed variably in human GIST specimens and cell lines. Moreover, our unpublished data also showed that in GIST cells, COP1 knockdown with siRNA significantly increased stability of c-jun. c-jun has been reported to stabilize β-catenin via its association with TCF4 and regulate gene transcription stimulated by the canonical Wnt signaling pathway (45). Accordingly, our in vitro data demonstrated that COP1 knockdown increased nuclear β-catenin half-life and stability, and transcriptional activity in a TCF reporter assay. Meanwhile, based on COP1 immunoprecipitation, endogenous COP1 interacted with β-catenin in GIST T1 cells. Similarly, we also detected interaction between endogenous β-catenin and COP1 in some human GIST specimens, including primary, untreated, and metastatic, resistant GISTs (Supplementary Fig. S3). Furthermore, ectopic expression of COP1 reduced active β-catenin and cyclin D1 expression induced by β-catenin overexpression. The coordinated action of COP1 and other E3 ligases, such as β-TrCP, has been reported to negatively regulate β-catenin (29). Although our data demonstrated that COP1 loss contributed to upregulation of β-catenin in GIST, other mechanisms are likely involved and may involve genomic or epigenetic changes.

As the Wnt pathway regulates multiple cellular processes linked to tumor progression, several small-molecule inhibitors or blocking antibodies have been developed for potential therapeutic use (46, 47). Promising data using these agents in different tumor types and animal models have been reported (27, 47). Indeed, we showed that PKF118-310, a selective inhibitor targeting the canonical Wnt/β-catenin signaling cascade, was effective in both imatinib-sensitive and resistant GIST cells in vitro and tumors in vivo. Alternatively, two different tankyrase inhibitors that antagonize Wnt/β-catenin signaling were effective against GIST cell lines in vitro. Strikingly, the tankyrase inhibitor G007-LK had almost as much antitumor efficacy as imatinib in Kit^V558Δ/+ mice. Thus, inhibition of Wnt/β-catenin signaling may be useful in patients with GIST that has become refractory to imatinib or other tyrosine kinase inhibitors. Furthermore, we found that 2 weeks of imatinib therapy increased β-catenin mRNA in Kit^V558Δ/+ tumors and restored active β-catenin levels, suggesting that Wnt/β-catenin signaling is a compensatory mechanism for KIT inhibition. Consistent with these findings, the combination of imatinib and G007-LK was even more effective than either alone in Kit^V558Δ/+ mice.

Although we have demonstrated Wnt/β-catenin activation in human GIST, these findings occurred only in a subset of GIST patients, and more investigation is needed to correlate β-catenin expression with clinicopathologic features. In addition, the efficacy of our in vivo treatments was assessed at short-term intervals, so the toxicity and efficacy of longer treatment durations need to be defined. It is possible that resistance to Wnt inhibition may develop over time. Our data support a clinical trial of Wnt/β-catenin inhibition in GIST. Multiple agents targeting the Wnt/β-catenin pathway are already being tested in patients with cancer (48), which should indicate their safety and effectiveness.

In summary, our findings show that activation of the canonical Wnt pathway and accumulation of nuclear active β-catenin were present in a subset of human GISTs that were treatment naïve, responsive to imatinib, or resistant to imatinib. The mechanism involved reduction of DKK4 and enhanced the nuclear β-catenin stability by COP1 loss. Inhibiting Wnt/β-catenin signaling alone or in combination with imatinib demonstrated antitumor efficacy in multiple cells and preclinical models in GIST. Our data provide the rationale for targeting Wnt/β-catenin signaling as a new strategy in the treatment of GIST.

No potential conflicts of interest were disclosed.

Conception and design: S. Zeng, A.M. Seifert, M.J. Cavnar, R.P. DeMatteo

Development of methodology: S. Zeng, J.Q. Zhang, P. Besmer

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S. Zeng, A.M. Seifert, J.Q. Zhang, T.S. Kim, M.H. Crawley, J.H. Maltbaek, P. Besmer, C.R. Antonescu, R.P. DeMatteo

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S. Zeng, A.M. Seifert, J.Q. Zhang, T.S. Kim, V.P. Balachandran, J.A. Santamaria-Barria, N.A. Cohen, M.J. Beckman, F. Rossi, C.R. Antonescu, R.P. DeMatteo

Writing, review, and/or revision of the manuscript: S. Zeng, A.M. Seifert, J.Q. Zhang, T.S. Kim, V.P. Balachandran, N.A. Cohen, M.J. Beckman, B.D. Medina, F. Rossi, J.K. Loo, J.H. Maltbaek, C.R. Antonescu, R.P. DeMatteo

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): S. Zeng, J.Q. Zhang, J.A. Santamaria-Barria, M.J. Beckman, M.H. Crawley, J.K. Loo, J.H. Maltbaek

Study supervision: S. Zeng, R.P. DeMatteo

We are grateful to the Tissue Procurement Service for assistance in the acquisition of human tumor specimens, Laboratory of Comparative Pathology, Research Animal Resource Center, and Molecular Cytology Core Facility. We thank Russell Holmes for logistical and administrative support.

This work was supported by NIH grants R01 CA102613 and T32 CA09501, Betsy Levine-Brown and Marc Brown, David and Monica Gorin, and the Stephanie and Fred Shuman through the Windmill Lane Foundation (to R.P. DeMatteo); GIST Cancer Research Fund (to R.P. DeMatteo and C.R. Antonescu); F32 CA162721 and the Claude E. Welch Fellowship from the Massachusetts General Hospital (to T.S. Kim); F32 CA186534 (to J.Q. Zhang); and P50 CA140146-01 (to C.R. Antonescu).

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.