Canonical Wnt/β-catenin Signaling Drives Human Schwann Cell Transformation, Progression, and Tumor Maintenance

Malignant peripheral nerve sheath tumors (MPNSTs) are soft tissue sarcomas that are believed to originate in the Schwann cell or Schwann cell precursors (1). These tumors can occur in the context of Neurofibromatosis Type 1 syndrome (NF1), which occurs in 1 in 3,000 live births, but can also occur spontaneously in the general population (2, 3). Because of the incomplete understanding of the genes and pathways driving MPNST development and progression, the current treatment for patients is surgical resection of the tumor, if possible, followed by nonspecific, high-dose chemotherapy (4, 5). These therapies often prove ineffective, and subsequently, patients with MPNSTs suffer very poor 5-year survival rates of less than 25% (3, 4). This shows the urgent need for a more complete understanding of the genetic events that drive these tumors, to develop novel targeted therapies to treat these patients.

It is known that biallelic loss of the Neurofibromin 1 gene (NF1) in Schwann cells is the pathologic cause of the benign neurofibromas seen in NF1 patients, but secondary genetic changes, many of which remain unknown, must occur for these benign tumors to transform into MPNSTs (1, 5–8). Ten percent of NF1-associated neurofibromas will undergo malignant transformation, the leading cause of death in adult NF1 patients (4, 5). MPNSTs can also form spontaneously, in the absence of NF1 loss, and the genes responsible for spontaneous MPNST formation are also largely unknown (3, 4). Loss of PTEN expression and overexpression of EGF receptor (EGFR) are 2 changes often seen in both spontaneous MPNSTs and NF1-associated MPNSTs, but there are likely many other important genetic changes and signaling pathways yet to be identified (5, 8–14).

Canonical Wnt/β-catenin signaling has been shown to play a role in many types of cancer, including colorectal, lung, breast, ovarian, prostate, liver, and brain tumors (15). However, this pathway has not been directly implicated in neurofibromas or MPNSTs. In other cell types, Wnt signaling can be activated in cancer through a variety of mechanisms, including activating mutations in β-catenin (CTNNB1), overexpression of Wnt ligand genes, inactivating mutations in AXIN1, GSK3B, and APC (all encoding members of the β-catenin destruction complex), and promoter hypermethylation of negative regulators of Wnt signaling (16). Another mechanism of Wnt pathway activation recently shown in colorectal cancer is overexpression of R-spondins due to gene fusions (17). R-spondins are secreted ligands that can potentiate Wnt signaling in the presence of Wnt ligands (18). Wnt signaling can also be activated via crosstalk with other signaling pathways, including the phosphoinositide 3-kinase (PI3K)/AKT/mTOR pathway, where loss of PTEN can activate PKB/AKT, causing the phosphorylation and inactivation of GSK3B, which results in stabilization of β-catenin protein (19). Growth factor signaling pathways can also activate Wnt signaling, such as in the case of stimulation of the EGF receptor (EGFR), which results in the activation of β-catenin/T-cell factor (TCF)/lymphoid enhancer factor (LEF)–dependent transcription of genes such as CyclinD1 (CCND1), C-Myc (MYC), and Survivin (BIRC5) (20, 21). Notably, some human MPNSTs have been shown to have loss of PTEN expression and/or express high levels of active EGFR (9, 11, 12, 14). β-catenin–dependent transcription can promote progression through the cell cycle, stem cell self-renewal, and epithelial to mesenchymal transition, all of which play a role in tumor initiation and progression (15, 16, 22). Because of the importance of this pathway in driving tumorigenesis in many types of cancer, the development of small-molecule inhibitors that target Wnt signaling is rapidly underway and has the potential for profound clinical benefits for patients with Wnt-driven tumors (16, 23, 24).

We have implicated canonical Wnt signaling in the development and progression of peripheral nerve tumors by using a murine Sleeping Beauty (SB) forward genetic screen (25). In addition, we show that several well-established murine models of neurofibromas and MPNST development also exhibit activation of the Wnt signaling pathway. Activation of this pathway has been confirmed in human patient samples by gene expression microarray analysis and tissue microarray (TMA) studies. Activation of Wnt signaling occurs in human tumors through multiple mechanisms, including downregulation of β-catenin destruction complex members and overexpression of R-spondin 2 (RSPO2). We show that activating canonical Wnt signaling is sufficient to induce transformed properties in immortalized human Schwann cells in vitro. We also show that downregulating canonical Wnt signaling reduces the oncogenic and tumorigenic phenotypes observed in human NF1-associated and sporadic MPNST cell lines as measured by cell viability, colony formation, and xenograft tumor growth. Furthermore, we found that small-molecule inhibitors of Wnt signaling inhibit MPNST cell viability with little effect on normal human Schwann cells. These inhibitors show marked synergistic effects when combined with the mTOR inhibitor RAD-001, which has previously been shown to be moderately effective in preclinical MPNST models (26–28). These results suggest that Wnt/β-catenin signaling is a novel drug target for patients with MPNSTs.

To identify genes and pathways that drive Schwann cell tumor development and progression, a forward genetic screen using the SB transposon system was conducted in mice, similar to many that have been previously published (29). The full details of this screen will be published separately (Rahrmann and colleagues, submitted for publication). This screen identified several members of the canonical Wnt/β-catenin signaling pathway in the development of benign neurofibromas and MPNSTs (Supplementary Table S1). This screen was conducted using a CNPase promoter driving Cre expression that, in combination with a Cre/Lox-regulated Rosa26-SB11^LSL transgene, allows for T2/Onc mutagenic transposon mobilization in Schwann cells and their precursors, the cell of origin for peripheral nerve sheath tumors (30). A common transposon insertion site (CIS)-associated gene list was generated independently for benign neurofibromas and MPNSTs, as diagnosed by histopathologic examination (Rahrmann and colleagues, submitted for publication). We required that integrations map uniquely to the murine genome with a sequence length that would preclude random mapping. Furthermore, we required that each integration be present at more than 1 of 10,000 of the total sequences present to exclude potential non-driver mutations that are found in minor subclones or due to artifacts (31). When the entire CIS gene list for MPNSTs was analyzed using Ingenuity Pathway Analysis (Ingenuity Systems, www.ingenuity.com), we found enrichment for genes that are present in the Wnt/β-catenin pathway (P = 3.93E-4). Furthermore, when immunohistochemical (IHC) analysis was conducted on murine tumors from this SB screen at various stages of progression, the level and nuclear localization of β-catenin increased with tumor progression (Fig. 1A). In addition to tumors from mice induced with SB, 4 other well-established mouse models of MPNST showed nuclear β-catenin expression (Fig. 1B; refs. 13, 14, 32–34). These mouse models represent both NF1-associated and sporadic MPNST development by a variety of genetic mechanisms. Interestingly, tumors from the rapidly forming, highly aggressive models, Dhh-Cre; Nf1^fl/fl; Pten^fl/fl and Dhh-Cre; Pten^fl/fl; CNP-EGFR, stained very strongly for nuclear β-catenin, whereas the slower-forming NPcis and PLP-Cre; Nf1^fl/fl tumors had less total β-catenin, but still high nuclear staining. Thus, the intensity of nuclear β-catenin staining may correlate with and indeed drive tumor aggressiveness.

To determine the role of Wnt/β-catenin signaling in human Schwann cell tumors, we assayed the expression and localization of β-catenin and expression of known Wnt target genes in human Schwann cells and MPNST cell lines. Activated β-catenin is found in the nucleus where it can be transcriptionally active, whereas inactive β-catenin is cytoplasmically localized or membrane bound (22). In a TERT/CDK4^R24C immortalized human Schwann cell line (iHSC2λ; Margaret Wallace, manuscript in preparation), β-catenin was expressed, but localized primarily in the cytoplasm (Fig. 1C). In contrast, both sporadic and NF1-associated MPNST cell lines (STS-26T and ST8814, respectively) expressed a similar level of β-catenin, but with predominantly nuclear localization. Western blot analysis showed that in all 4 MPNST cell lines analyzed, the level of MYC protein was increased, and in 2 of 4 MPNST cell lines, CCND1 protein was increased compared with the immortalized human Schwann cell line (Fig. 1D). These data suggest that known β-catenin outputs are increased in MPNST cell lines compared with immortalized human Schwann cells.

To characterize whether alterations in regulators of Wnt signaling underlie pathway activation in MPNST cell lines, we assayed Wnt regulators by gene expression microarray in purified Schwann cells taken from human peripheral nerve, neurofibromas, and MPNSTs, as well as solid tumors at various stages of disease (ref. 35; Fig. 2). Several members of the β-catenin destruction complex, including APC and GSK3B, showed downregulation with tumor progression in a subset of cases. The Wnt ligand genes WNT2, WNT5A and WNT5B showed an increase in many or most samples. These data suggest that multiple mechanisms activate canonical Wnt signaling in human Schwann cell tumors.

To further investigate activation of Wnt signaling in human tumors, and to account for regulation of Wnt signaling by post-translational mechanisms such as protein degradation or localization, we constructed a human TMA composed of 30 benign dermal neurofibromas (dNF), 32 plexiform neurofibromas (pNF), and 31 MPNSTs. The fraction of tumors positive for β-catenin increased with tumor progression (55.9% dNF, 86.2% pNF, 96.9% MPNSTs, P = 0.0001 Fisher's exact test; Fig. 3A). In addition, the percentage of tumors with nuclear β-catenin was higher in the MPNSTs than in the neurofibromas (Fig. 3B and C). As a transcriptional output of Wnt signaling, we also quantified the percentage of CCND1-positive tumors and found that it also increased in the higher-grade tumors (67.6% dNF, 64.7% pNF, 94.1% MPNST, P = 0.0051 Fisher's exact test; Fig. 3D). Furthermore, the intensity of CCND1 staining increased with tumor progression (P = 0.0356 Fisher's exact test; Fig. 3E and F). As an additional transcriptional output of Wnt signaling, we assessed C-MYC staining intensity and found that it too showed a trend toward increased expression with tumor progression (P = 0.1603 Fisher's exact test; Supplementary Fig. S1A and S1B). This suggests that the acquisition of nuclear β-catenin in these tumors is activating known β-catenin targets, and correlates with tumor progression. To confirm this, we assessed β-catenin and CCND1 staining on a pNF that transformed into an MPNST. Although there was only a single patient represented on our TMA with a paired pNF and MPNST, we found that β-catenin level and nuclear localization increased in the MPNST. Furthermore, while the pNF was negative for CCND1, the MPNST was positive for CCND1 (Fig. 3G).

We next functionally validated a subset of the β-catenin–regulatory genes identified in our SB screen that were also implicated in human tumors (Supplementary Table S1 and Fig. 2). We hypothesized that if Wnt/β-catenin signaling played a role in the development of peripheral nervous system tumors, then activation of this pathway in immortalized human Schwann cells may be sufficient to drive a more transformed phenotype in these cells. To activate Wnt signaling, we expressed an activated form of CTNNB1 (CTNNB1^S33Y) in immortalized human Schwann cells (iHSC1λ and iHSC2λ). Cells overexpressing CTNNB1^S33Y showed increased levels of total β-catenin protein, but, interestingly, only the iHSC1λ cells showed increased activated β-catenin (Fig. 4A). Expression of CTNNB1^S33Y resulted in increased expression of MYC, CCND1, AXIN2, LEF1, and BIRC5 in both iHSC1λ and iHSC2λ cell lines, showing activation of genes that have been shown to be β-catenin targets (Supplementary Fig. S2A and S2B). CTNNB1^S33Y expression also resulted in increased cell viability (Fig. 4B), but did not change soft agar colony formation (Fig. 4C). However, we did observe that cells overexpressing CTNNB1^S33Y stopped growing as a monolayer, as seen in nontransformed cells, and began growing as 3-dimensional colonies (Fig. 4D).

In addition to activating Wnt signaling by overexpressing CTNNB1^S33Y, we used short hairpin RNAs (shRNA) to knock down AXIN1 and GSK3B, both encoding members of the β-catenin destruction complex identified in our SB screen (Supplementary Table S1) and known tumor suppressor genes in other types of cancers (15). These genes were knocked down in both immortalized human Schwann cell lines (iHSC1λ and iHSC2λ) using shRNAs and validated by quantitative PCR (qPCR) analysis (Fig. 4E and Supplementary Fig. S2I) and Western blot analysis (Supplementary Fig. S2C and S2F). We show that knockdown of either AXIN1 or GSK3B activated genes that have been shown to be to be targets of β-catenin, including CCND1, MYC, AXIN2, LEF1, and BIRC5 (Supplementary Fig. S2D, S2E, S2G, and S2H). Knockdown of AXIN1 or GSK3B was sufficient to induce oncogenic properties in immortalized human Schwann cells as shown in vitro by a significant increase in cell viability (Fig. 4F and Supplementary Fig. S2J) and anchorage-independent growth (Fig. 4G and Supplementary Fig. S2K). When the immortalized human Schwann cells that expressed either AXIN1 or GSK3B shRNAs were injected into immunodeficient mice, they were unable to induce tumor formation (data not shown), suggesting that while activation of Wnt signaling is sufficient to induce some oncogenic properties in Schwann cells in vitro, it is not sufficient to induce tumorigenic properties in vivo.

We next sought to determine whether the inhibition of Wnt/β-catenin signaling in MPNST cell lines was sufficient to reduce cellular viability and anchorage-independent growth. To reduce Wnt signaling, we knocked down CTNNB1 and TNKS in 2 MPNST cell lines, S462-TY (NF1-associated MPNST cell line; ref. 36) and STS-26T (sporadic MPNST cell line; ref. 37) using shRNA vectors. We chose to knockdown CTNNB1, as it plays a direct role in the transcription of Wnt-dependent genes by binding TCF/LEF in the nucleus and acting as a transcriptional activator (22, 23). TNKS was also knocked down because it was identified in our forward genetic screen, and inhibition of TNKS is known to stabilize AXIN1, leading to the degradation of β-catenin protein (38). TNKS is also the target of many small-molecule inhibitors of the Wnt pathway (38). Knockdown of CTNNB1 and TNKS by shRNA was confirmed by qPCR (Fig. 5A and D). Reduction in either CTNNB1 or TNKS was sufficient to decrease expression of β-catenin transcriptional targets, as shown by a reduction in the expression of MYC, CCND1, AXIN2, LEF1, and BIRC5 (Supplementary Fig. S3A–S3D). Knockdown of either CTNNB1 or TNKS reduced cell viability (Fig. 5B and E) and anchorage-independent growth (Fig. 5C and F) in both S462-TY and STS-26T cells.

Overexpression of GSK3B, a member of the β-catenin destruction complex, was also sufficient to reduce the oncogenic properties of both the NF1-associated and sporadic MPNST cell lines. GSK3B overexpression was confirmed by Western blot analysis and qPCR and resulted in reduced β-catenin total protein levels (Fig. 5G and Supplementary Fig. S3E and S3F). The reduction in known Wnt signaling outputs was also shown by a decrease in MYC, CCND1, AXIN2, LEF1, and BIRC5 transcript levels (Supplementary Fig. S3E and S3F). GSK3B overexpression resulted in a decrease in cell viability (Fig. 5H) and a significant decrease in soft agar colony formation in both cell lines (Fig. 5I). These results suggest that these MPNST cell lines depend on Wnt signaling for rapid proliferation and robust colony-forming abilities, and reduction in Wnt signaling can reduce these oncogenic properties.

To investigate the role of Wnt/β-catenin signaling in vivo, the S462-TY and STS-26T cell lines expressing either CTNNB1 shRNA or GSK3B cDNA were injected into immunodeficient mice, which were monitored for the rate of tumor formation and growth. Reduction in CTNNB1 by shRNA delayed the rate of tumor growth in the NF1-asociated cell line (Supplementary Fig. S4A and S4B). Tumor sections derived from cells expressing the CTNNB1 shRNA showed fewer Ki67-positive cells compared with tumors derived from nonsilencing (NS) shRNA–expressing cells (Supplementary Fig. S4C). Overexpression of GSK3B resulted in reduction in tumor growth in the spontaneous MPNST cell line (Supplementary Fig. S4D and S4E). Tumor sections from cells overexpressing GSK3B also showed fewer Ki67-positive cells compared with tumors derived from NS shRNA–expressing cells (Supplementary Fig. S4F). Despite a significant delay in tumor onset, tumor volume at 40 days after injection was similar in control and experimental groups. Evaluation of gene expression in these tumors indicated that knockdown of CTNNB1 and overexpression of GSK3B was not maintained over the course of the experiment (data not shown).

R-spondins are secreted ligands that potentiate Wnt signaling when Wnt ligand is available (18). Microarray analysis showed that RSPO2 was highly expressed in 2 of 6 MPNST cell lines and 2 of 13 primary MPNSTs evaluated compared with purified normal human Schwann cells or normal nerve (Fig. 6A). Furthermore, it has been shown that one mechanism for RSPO2 overexpression is a deletion-mediated gene fusion between exon 1 of EIF3E and exon 2 of RSPO2, resulting in the overexpression of native RSPO2 protein (17). To determine whether similar gene fusions occur in Schwann cell tumors, primers were designed to amplify any fusion transcripts between EIF3E and RSPO2 in cDNA libraries made from 5 human MPNST cell lines and 2 immortalized human Schwann cell lines. We identified a fusion transcript expressed by the S462 MPNST cell line that was not present in any of the other MPNST or immortalized human Schwann cell lines (Fig. 6B). Expression of this fusion correlated with a 1,295-fold increase in RSPO2 expression in the S462 cell line (Fig. 6C). The fusion transcript that was identified included exon 1 of EIF3E fused to a 122 base pair region of chromosome 8 located between EIF3E and RSPO2, fused to exon 2 of RSPO2 (Fig. 6D). To determine whether S462 cells require RSPO2, we used an shRNA construct to knock down its expression. Depletion of RSPO2 significantly reduced expression of Wnt target genes and cell viability (Fig. 6E and F). These data suggest that overexpression of RSPO2 may drive a subset of human MPNSTs, and further support the model that Wnt signaling is activated in Schwann cells by multiple mechanisms.

Our functional data suggested that targeting Wnt/β-catenin signaling with small-molecule inhibitors might be effective at reducing MPNST cell viability and tumor forming properties. To test this hypothesis, we exposed a panel of 5 MPNST cell lines (S462, S462-TY, ST8814, T265, and STS-26T) and 2 immortalized human Schwann cell lines (iHSC1λ and iHSC2λ) to 2 compounds that inhibit Wnt signaling, XAV-939 and IWR-1. These compounds both function to stabilize AXIN1 by inhibiting TNKS, resulting in an increase in β-catenin phosphorylation and subsequent degradation (38). We found that normal Schwann cells are highly resistant to proliferation inhibition and cell death by these 2 compounds as shown by high IC₅₀ values, whereas the MPNST cell lines were quite sensitive with relatively low IC₅₀s of 0.047 to 2.22 μmol/L for XAV-939 and 0.055 to 186.87 μmol/L for IWR-1 (Fig. 7A; CalcuSyn Version 2.1, BioSoft).

We next conducted a screen to identify targeted therapies that synergized with Wnt signaling inhibitors to induce apoptosis in MPNST cell lines. Several drugs were screened, including inhibitors of PI3K (PI-103), mTOR (RAD-001 and rapamycin), and MEK (PD-901 and AZD-6244). Of these drugs, the most significant synergism was seen with RAD-001. It has been shown that RAD-001 is modestly effective in several preclinical models of peripheral nerve tumors, but the effects of this drug are largely cytostatic (26, 27). In most cases, treatment of the S462, T265, and STS-26T MPNST cell lines with RAD-001, IWR-1, or XAV-939 as single agents at their IC₅₀ concentrations resulted in less than 50% terminal deoxynucleotidyl transferase–mediated dUTP nick end labeling (TUNEL)-positive cells, showing that although 50% of proliferation was inhibited, only a minority of cells underwent apoptosis (Fig. 7B and C). In contrast, when the MPNST cell lines were treated with their IC₂₅ dose of RAD-001 in combination with either IWR-1 or XAV-939 also at their IC₂₅ dose, the percentage of TUNEL-positive cells often went above 50%, and in the S462 cell line, even approached 100%, showing that co-targeting the Wnt and mTOR pathways is synergistic in inducing apoptosis.

Analysis of MPNST tumor genomes has shown a wide variety of genetic changes ranging from point mutations to entire chromosome gains and losses (39). Determining which of these changes contribute to tumorigenesis and which changes are simply “passenger” mutations is crucial in identifying targets for therapeutic intervention (1, 39, 40). Prior studies have focused on NF1 and p53/Rβ-regulated pathways in MPNSTs; yet, there remains a need to broaden our understanding of the genetic alterations that contribute to Schwann cell tumorigenesis. The SB transposon system is an unbiased, powerful tool to identify oncogenes and tumor suppressor genes (29). In an SB screen conducted to find genetic drivers of MPNSTs and neurofibromas, we uncovered many alterations that affect canonical Wnt/β-catenin signaling, suggesting this pathway is a likely driver of Schwann cell tumor development, progression, and maintenance (Supplementary Table S1) (25). We showed that murine tumors from both the SB screen and from 4 other established MPNST models showed nuclear-localized β-catenin, suggesting that they have activated canonical Wnt signaling (Fig. 1). In addition, we identified activation of Wnt signaling in a majority of human MPNST cell lines and primary human tumors investigated (Figs. 1–3).

In this study, we showed that multiple genetic alterations correlate with activation of the canonical Wnt/β-catenin pathway, and that these alterations can induce oncogenic properties in human Schwann cells and are required for tumor maintenance in MPNST cells. A subset of human Schwann cell tumors shows downregulation of β-catenin destruction complex components, such as APC and GSK3B (Fig. 2). Microarray analysis also showed overexpression of several Wnt ligands and RSPO2, a secreted ligand that can potentiate Wnt signaling, in a subset of human tumors (Fig. 6). Interestingly, we found that while CCND1 protein levels correlate with tumor progression, CCND1 mRNA expression is not as convincing (Figs. 2 and 3). It should be noted that the microarray data suggest that in whole tumors, CCND1 may be decreased, but in purified cells, it seems to be unchanged or slightly upregulated. This may be due to the effects of contaminating cells within the whole tumor, such as fibroblasts, macrophages, and endothelial cells. In addition, we believe that the discrepancy seen between the TMA and the microarray gene expression of CCND1 may be due to the fact that the CCND1 transcript is highly regulated at the translational level. Because of the long 5′ untranslated region and secondary structure of CCND1 mRNA, it is often poorly translated, although many factors, including PI3K/mTOR signaling, can increase the translation of this transcript (41). On the basis of the mRNA expression data, MYC may be a more consistent marker for activated Wnt signaling, whereas CCND1 may be a better marker of Wnt signaling activation by IHC.

We tested the functional importance of Wnt pathway activation in Schwann cell tumorigenesis through gain- and loss-of-function studies in vitro. Using gene overexpression and knockdown in immortalized human Schwann cells and MPNST cell lines, we showed that aberrant expression of Wnt regulators resulted in phenotypic and functional changes in these cells. Downregulation of GSK3B and AXIN1 and overexpression of CTNNB1 were sufficient to induce transformed properties in immortalized human Schwann cells, and downregulation of CTNNB1 and TNKS, as well as overexpression of GSK3B, reduced the oncogenic properties of MPNST cell lines and delayed tumor growth in vivo (Figs. 4 and 5 and Supplementary Figs. S2 and S4).

Recently, a novel mechanism for activating Wnt signaling was shown in colorectal cancer in which gene fusions occurred between EIF3E and RSPO2 due to a deletion on chromosome 8, resulting in the overexpression of native RSPO2 protein (17). We identified a similar fusion transcript in a human MPNST cell line with elevated RSPO2 expression and showed that knockdown of RSPO2 was sufficient to reduce Wnt signaling outputs and cellular viability in this cell line (Fig. 6). Further work is needed to determine the frequency of R-spondin fusion transcript expression in primary human Schwann cell tumors. It will also be necessary to identify whether expression of R-spondin fusion transcripts in MPNSTs is mediated by chromosome 8 deletions similar to those seen in colorectal cancer, or if there is another mechanism by which this occurs, such as transcription-mediated gene fusion. On the basis of our results, a drug that could block the function of RSPO2 may be therapeutically beneficial for patients with MPNSTs overexpressing RSPO2.

In addition to the mechanisms described above, there are likely other mechanisms by which this pathway is activated in human tumors. For example, activation of PKB/AKT, a common phenomenon in MPNSTs, leads to the phosphorylation and inactivation of GSK3b, resulting in stabilization of the β-catenin protein (19, 42). Loss of PTEN expression has been shown in human MPNSTs, and this could activate Wnt signaling through activation of AKT (12, 13). It has also been shown that in cells expressing EGFR, EGF stimulation can result in the activation of β-catenin/TCF/LEF-dependent transcription, and EGFR expression has been implicated in MPNST development as well (9, 11, 20). Recently, Mo and colleagues (43) showed a novel mechanism of NF1-associated MPNST progression in which autocrine activation of CXCR4 by CXCL12 mediates tumor progression through PI3K and β-catenin. Although no insertions were identified in Cxcr4 or Cxcl12 in our SB screen, gene expression data from human tumors showed statistically significant differential regulation of CXCR4 in MPNST tumor-to-nerve comparison (fc = 5.3×, P = 0.016), but not in MPNST cell-to-NHSC comparison (fc = 0.85×, P = 0.16). In contrast, CXCL12 showed a statistically differential expression pattern in MPNST cell-to-NHSC comparison (fc = 2.7×, P = 0.02), but not in MPNST tumor-to-nerve comparison (fc = 4.7×, P = 0.15). This study by Mo and colleagues (43) complements our work in showing yet another mechanism by which human Schwann cell-derived tumors activate Wnt signaling, leading to tumor progression. Additional work will be required to determine the contribution of each of these pathways to Wnt activation in human MPNSTs.

It is unclear from this study whether Wnt pathway activation is important for tumor initiation or for tumor progression, but it seems likely that it may play a role in both processes. It is clear that a subset of benign human neurofibromas exhibit activated Wnt signaling, which may play a role in the initiation of these benign tumors (Fig. 3 and Supplementary Fig. S1). A larger percentage of high-grade tumors (MPNSTs) have activation of this signaling pathway, making it plausible that this pathway also plays a role in progression from a benign tumor to malignancy. It is also likely that this pathway may cooperate with other pathways that play a role in the development or progression of MPNSTs. For example, Wnt signaling activation is an initiating event for colorectal cancer, leading to hyperplasia, but other genetic changes must occur for a tumor to form (44). Future work will need to focus on identifying cooperating mutations or signaling pathways that function in concert with Wnt signaling in MPNST formation. Indeed, loss of NF1 expression is a good candidate for cooperating alteration, and future experiments in which NF1 is lost and the Wnt pathway is activated in normal Schwann cells will be critical to our understanding of cooperating genes and pathways.

Although the Wnt pathway is known to play a critical role in oncogenesis, the development of clinically beneficial targeted therapies has lagged. The discovery of small-molecule inhibitors of the Wnt/β-catenin pathway is being aggressively pursued, as these therapies have the potential to benefit patients suffering from a wide variety of cancers (16, 23, 24).Wnt/β-catenin signaling is critical for many normal developmental and cellular processes; thus, inhibition of this pathway may have many unwanted and negative side effects (16, 45). To this end, inhibitors targeting specific Wnt pathway molecules (such as RSPO2) and functions in tumorigenesis may be the key to identifying a successful targeted therapy (16, 45). We show that inhibiting Wnt signaling in vitro is an effective cytostatic therapy, but when combined with RAD-001, it creates a potent cytotoxic therapy (Fig. 7). Further work will need to be done to investigate the use of Wnt pathway inhibitors using in vivo models of MPNST, as these may be a valuable new class of targeted therapies for patients with MPNSTs.

Formalin-fixed, paraffin-embedded tissues were sectioned at 5 μm, mounted, and heat-fixed onto glass slides to be used for IHC analyses. Briefly, the glass section slides were dewaxed and rehydrated through a gradual decrease in ethanol concentration. The antigen epitopes on the tissue sections were then unmasked using a commercially available unmasking solution (Vector Laboratories) according to the manufacturer's instructions. The tissue section slides were then treated with 3% hydrogen peroxide to remove endogenous peroxidases. Blocking was conducted at room temperature in normal goat serum (5% serum in PBS) in a humidified chamber for one hour. Sections were then incubated overnight at 4°C in a humidified chamber with primary antibody (β-catenin, 1:100, Cell Signaling). After primary incubation, sections were washed thoroughly in PBS before incubating with goat anti-rabbit horseradish peroxidase (HRP)–conjugated secondary antibody (Santa Cruz Biotechnology). After 3 washes with PBS, the sections were treated with freshly prepared 3,3′-diaminobenzidine substrate (Vector Laboratories) and allowed to develop before stopping the reaction in water once adequate signal was obtained. Finally, sections were then lightly counter-stained with hematoxylin, dehydrated through gradual increase in ethanol concentration, cleared in Citrosol (Fisher Scientific), and mounted in Permount (Fisher Scientific).

Cultured immortalized Schwann cells (iHSC1λ and iHSC2λ) were both derived from a patient's normal sciatic nerve, are NF1 wild-type, and were immortalized by hTERT and CDK4^R24C to allow in vitro studies (Dr. Margaret Wallace, manuscript in preparation). Immortalized human Schwann cell lines and MPNST cell lines [S462 (46), S462-TY (36), ST8814 (47), T265 (48), and STS-26T (37)] were maintained in Dulbecco's Modified Eagle Medium (DMEM) supplemented with 10% FBS and penicillin/streptomycin (Cellgro) and cultured on tissue culture-treated plates under standard conditions of 37°C and 5% CO2. No authentication of cell lines was done by the authors.

For immunofluorescence assays, cells were grown to 80% confluency on 8 chambered slides (Lab-TekII). Cells were fixed in 10% formalin and washed with PBS with 0.1% Tween-20. Cells were incubated in β-catenin primary antibody (1:100, Cell Signaling) at 4°C overnight, followed by 1-hour room temperature incubation in anti-rabbit AlexaFluor 488 secondary antibody (Invitrogen). TUNEL staining was conducted using the In Situ Cell Death Detection Kit, POD (Roche). Slides were mounted using Prolong Gold Antifade Reagent with 4′, 6-diamidino-2-phenylindole (DAPI; Invitrogen) and images using a Zeiss Axiovert 25 inverted microscope. For analysis of cell death, total cells and TUNEL-positive cells were counted and averaged over 3 independent frames with 50 to 100 cells per frame.

One million cells were lysed using an NP-40 buffer (50 mmol/L Tris-HCl, pH 7.6, 150 mmol/L NaCl, 1% NP-40, 5 mmol/L NaF, 1 mmol/L EDTA) containing a protease inhibitor (Roche) and phosphatase inhibitors (Sigma). Whole-cell lysates were cleared by centrifugation. Protein samples were prepared in an SDS solution with reducing agent (Invitrogen) and run on 10% Bis-Tris pre-made gels (NuPage, Invitrogen). Gels were transferred onto polyvinylidene difluoride membranes using the iBlot system (Invitrogen) and activated in 100% methanol. Membranes were blocked in filtered 5% bovine serum albumin (BSA) for 2 hours at room temperature, followed by a 4°C overnight incubation in primary antibody. The primary antibodies used in this study were activated β-catenin (1:1,000), total β-catenin (1:100), GSK3B (1:1,000), AXIN1 (1:1,000), and glyceraldehyde-3-phosphate dehydrogenase (GAPDH, 1:2,000; Cell Signaling). Following primary antibody incubation, membranes were thoroughly washed in TBS with 0.1% Tween-20 (TBST) and incubated in goat anti-rabbit IgG-HRP–conjugated secondary antibody (Santa Cruz, 1:4,000 in 0.5% BSA, 1 hour at room temperature). Blots were thoroughly washed in TBST and developed using the SuperSignal WestPico Chemiluminescence Detection Kit (Thermo Scientific). Densitometry quantification was done using ImageJ software and normalized to GAPDH (49).

We used the published data (GEO accession GSE14038, Affymetrix GeneChip HU133 Plus 2.0; ref. 39) for gene microarray analysis. This dataset used custom CDF (custom GeneChip library file) based on RefSeq target definitions (Hs133P RefSeq Version 8) for accurate interpretation of GeneChip data (50). The dataset in Fig. 2 includes 86 microarrays on purified human Schwann cells and primary tissues taken from a normal sciatic nerve, dNFs, pNFs, and MPNST cell lines. The heatmap was generated using CRAN's pheatmap package (http://cran.r-project.org/web/packages/pheatmap/index.html), and Refseq IDs were replaced with corresponding HUGO Gene Nomenclature Committee official gene symbols. Statistical comparisons were done using R/Bioconductor's Limma package (http://www.bioconductor.org) and GeneSpring GXv7.3.1 (Agilent Technologies). Differentially expressed genes were defined as genes with expression levels at least 3-fold higher or lower in target groups (MPNST) compared with NHSC after applying Benjamini and Hochberg false discovery rate correction (P ≤ 0.05).

Representative areas of disease were identified on hematoxylin and eosin-stained sections for 30 neurofibromas, 32 plexiform neurofibromas, and 31 MPNSTs. TMA blocks consisting of duplicate 1.0 mm core samples were constructed with a manual tissue arrayer (MTA-1, Beecher Inc) and limited to 64 cores per recipient block. IHC for CCND1 (SP4) monoclonal antibody (Neomarkers, Thermo Scientific, 1:50) and β-catenin (6B3) monoclonal antibody (Cell Signaling; 1:200) was conducted using an automated IHC staining platform (Nemesis 7200, Biocare) following standard IHC protocols (51). Digital images of IHC-stained TMA slides were obtained as previously described by Rizzardi and colleagues (52). P values were determined using a 3 × 2 (staining positivity) or 3 × 3 (staining intensity or localization) Fisher's exact test to determine differences in staining between the 3 tumor types.

GSK3B and Luciferase cDNA (Invitrogen) were each cloned into a vector containing a CAGGS promoter to drive cDNA expression followed by an IRES-GFP. These plasmid vectors were transfected into cells using the NEON transfection system, according to the manufacturer's protocol (Invitrogen). Three independent shRNAs targeting AXIN1, GSK3B, CTNNB1, TNKS, RSPO2, and a non-silencing control shRNA were purchased from OpenBiosystems. Lentiviral particles containing the shRNAs were produced in 293T cells using the Trans-Lentiviral Packaging Kit (Thermo Scientific). A dsRed control and CTNNB1^S33Y were cloned into a vector containing a CAGGS promoter to drive cDNA expression followed by an IRES-GFP and also transduced into cells using lentiviral particles. Viral supernatant was collected after 24 hours of viral production, cleared, and applied directly to the cells for 6 hours. Following viral transduction, cells underwent selection in 4 μg/mL puromycin (Invitrogen). shRNA expression was validated by GFP expression, Western blot analysis, and qPCR. Data are shown for the one shRNA construct that gave the greatest knockdown as determined by Western blot analysis and/or qPCR.

RNA was extracted from 1 million cells using the High Pure RNA Isolation Kit (Roche). RNA was analyzed by nanodrop (Thermo Scientific) and by agarose gel electrophoresis for quantification and quality control. One μg of RNA was used to synthesize cDNA using the Transcriptor First Strand Synthesis (Roche) with both random hexamer and oligo dT primers. qPCR reactions were conducted using LightCycler 480 SYBR I Green (Roche) and run on an Eppendorf Mastercycler ep gradient S. Primer sequences can be found in Supplementary Table S2 (21, 53, 54). Data were analyzed using RealPlex software, calibrated to ACTB levels and normalized to either overexpression control cells (expressing CAGGS > Luciferase-IRES-GFP or dsRed) or non-silencing shRNA–expressing cells and averaged over 3 experimental replicates.

Cellular viability assays were set up in a 96-well format with 100 cells plated per well in DMEM full media containing 4 μg/mL puromycin (Invitrogen). Readings were taken every 24 hours over 6 days by the MTS assay (Promega). Absorbance was read at 490 nm to determine viability and 650 nm to account for cellular debris on a BioTek Synergy Mx automated plate reader.

Six-well plates were prepared with bottom agar composed of 3.2% SeaPlaque Agar (Lonza) in DMEM full media and allowed to solidify before 10,000 cells in top agar (0.8% SeaPlaque Agar in DMEM full media) were plated and allowed to solidify. DMEM full media with 4 μg/mL puromycin (Invitrogen) was plated over the cells and cells were incubated under standard conditions (5% CO2, 37°C) for 14 days. Top media were removed and cells were fixed in 10% formalin (Fisher Scientific) containing 0.005% crystal violet (Sigma) for 1 hour at room temperature. Formalin was removed and colonies were imaged on a Leica S8 AP0 microscope. Twelve images per cell line were taken and automated colony counts were done using ImageJ software (49). Results shown are a representative example of at least 3 independent experiments.

A total of 1.5 million cells in serum-free DMEM with 33% Matrigel Basement Membrane (BD Biosciences) were injected into the back flanks of athymic Foxn1^nu/nu mice (Charles River Laboratories). Each mouse was injected on the right flank with cells expressing a specific shRNA or a cDNA and on the left flank with control cell lines expressing a non-silencing shRNA or a Luciferase cDNA with at least 4 mice per experiment. Tumor volume was measured biweekly using calipers, and mice were sacrificed when the tumor reached approximately 10% of the total body weight or 40 days after injection. Tumors were harvested for IHC and Western blot analysis. All animal work was conducted according to the University of Minnesota's approved animal welfare protocol.

RAD-001, IWR-1, and XAV-939 were solubilized in dimethyl sulfoxide and then subsequently diluted in sterile PBS. A total of 1,200 cells per well of a 96-well plate were treated with varying concentration of drug in quadruplicate and assayed for cell viability using the MTS assay (Promega) to determine the IC₅₀ values. All data analysis was done using CalcuSyn software (CalcuSyn Version 2.1, BioSoft).

D.A. Largaespada has received a commercial research grant from Genentech; has ownership interest (including patents) in NeoClone Biotechnology, Inc. and Discovery Genomics, Inc.; and is a consultant/advisory board member of NeoClone Biotechnology, Inc. and Discovery Genomics, Inc. No potential conflicts of interest were disclosed by the other authors.

Conception and design: A.L. Watson, E.P. Rahrmann, N. Ratner, D.A. Largaespada

Development of methodology: A.L. Watson, C.B. Conboy, B.R. Wahl, A.E. Rizzardi, M.H. Collins, N. Ratner, D.A. Largaespada

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A.L. Watson, E.P. Rahrmann, B.S. Moriarity, C.B. Conboy, A.D. Greeley, A.L. Halfond, L.K. Anderson, B.R. Wahl, V.W. Keng, A.E. Rizzardi, C.L. Forster, M.H. Collins, M.R. Wallace, S.C. Schmechel, D.A. Largaespada

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A.L. Watson, K. Choi, C.B. Conboy, A.D. Greeley, L.K. Anderson, A.E. Rizzardi, M.H. Collins, A. Sarver, D.A. Largaespada

Writing, review, and/or revision of the manuscript: A.L. Watson, E.P. Rahrmann, B.S. Moriarity, K. Choi, C.B. Conboy, A.D. Greeley, A.L. Halfond, L.K. Anderson, B.R. Wahl, V.W. Keng, A.E. Rizzardi, M.H. Collins, A. Sarver, M.R. Wallace, S.C. Schmechel, N. Ratner, D.A. Largaespada

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A.L. Watson, E.P. Rahrmann, A.D. Greeley, A.E. Rizzardi, C.L. Forster

Study supervision: A.L. Watson, D.A. Largaespada

This work was supported by the NIH (P50 NS057531), The Children's Tumor Foundation, The Zachary NF Fund, and The Jacqueline Dunlap NF Fund. A.L.Watson is funded by the 2011 Children's Tumor Foundation Young Investigators Award (2011-01-018). K. Choi is funded by the National Cancer Institute Training Grant 5T32CA059268-15. C.B. Conboy is funded by NIH F30 CA171547 and NIH MSTP grant T32 GM008244. These studies used BioNet histology and digital imaging core facilities, which are supported by NIH grants P30-CA77598 (D. Yee), P50-CA101955 (D. Buchsbaum), and KL2-RR033182 (B. Blazar), and by the University of Minnesota Academic Health Center.