Neurofibromin controls cell growth by regulating STAT3 activity in vitro and in vivo

Abstract

Neurofibromatosis 1 (NF1) is an autosomal dominant disorder in which affected individuals are prone to the development of both benign and malignant tumors. Previous studies have shown that the NF1 gene product, neurofibromin, negatively regulates Ras and mammalian target of rapamycin (mTOR) signaling, prompting clinical trials aimed at arresting tumor growth by inhibiting Ras and mTOR hyperactivation. In an effort to identify additional downstream targets of neurofibromin for therapeutic drug design, we employed an unbiased high-throughput chemical library screen (HTS) using NF1-deficient ST88-14 malignant peripheral nerve sheath tumor (MPNST) cells. We identified one previously unrecognized compound, Cucurbitacin-I, for future study, and showed that Cucurbitacin-I inhibits NF1-deficient cell growth by inducing apoptosis. We further show that STAT3, the target of Cucurbitacin-I inhibition, is hyperactivated in NF1-deficient primary astrocytes and neural stem cells as well as mouse glioma and human MPNST cells through Ser-727 phosphorylation, leading to increased cyclin D1 expression. Next, we demonstrated that STAT3 is regulated in NF1-deficient primary mouse glial and human MPNST cells in a TORC1- and Rac1-dependent manner, and that mTOR/Rac1/STAT3 activation controls cyclin D1 levels. Finally, we show that JSI-124 (Cucurbitacin-I) treatment inhibits NF1-deficient ST88-14 tumor growth in vivo. In summary, we used a chemical genetics approach to uncover a novel neurofibromin/mTOR pathway signaling molecule, define its mechanism of action and regulation, and establish STAT3 as a tractable target for future NF1-associated cancer therapy studies.

Introduction

Neurofibromatosis type 1 (NF1) is one of the most common genetic causes for nervous system tumors, and affected patients develop both benign and malignant tumors involving the brain and peripheral nerves (1). Within the central nervous system (CNS), low-grade gliomas of the optic pathway are observed in 15% of children with NF1 (2), while adults with NF1 develop high-grade astrocytomas at a 50-fold increased incidence (3). In addition, benign peripheral nerve sheath tumors (neurofibromas) and malignant peripheral nerve sheath tumors (MPNSTs) are seen in patients with NF1 (4).

With the identification of the NF1 gene in 1990, several investigators found that the NF1 gene product, neurofibromin, functions primarily as a negative regulator of the RAS proto-oncogene, such that NF1 gene inactivation in tumors is associated with increased RAS pathway activation and cell proliferation (5–7). Further studies demonstrated that inhibition of RAS function reduces NF1-deficient cell and tumor growth in vitro and in vivo (8, 9). These exciting observations led to the initiation of clinical trials using anti-RAS biologically-based therapies, including farnesyltransferase inhibitors (FTIs), to treat tumors arising in patients with NF1. Unfortunately, FTI therapy has shown little efficacy in the treatment of NF1-associated plexiform neurofibroma (10).

In light of the limited success of FTI monotherapy in NF1 clinical trials and the fact that RAS activation can mediate cell growth through a multitude of effector proteins, we previously employed an unbiased proteomic method to identify neurofibromin/RAS downstream effector proteins that might serve as improved targets for therapeutic drug design. Using this approach, we found that neurofibromin/RAS growth regulation requires mammalian target of rapamycin (mTOR) function (11). Similar findings were also reported by others (12), prompting preclinical studies with rapamycin demonstrating their efficacy in vivo (13–15).

In an effort to more precisely define the signaling effectors downstream of mTOR that mediate neurofibromin tumor suppression, we have previously shown that mTOR-dependent growth control requires Rac1 activation in NF1-deficient cells (16). Unfortunately, since there are few Rac1-specific pharmaceutical-grade inhibitors suitable for preclinical/clinical study, we next sought to identify additional targets in NF1-deficient cells by means of high-throughput chemical library screening. Using this approach, we identified one previously unrecognized compound (Cucurbitacin-I) and found that Cucurbitacin-I inhibits human NF1-deficient MPNST and mouse Nf1−/− astrocyte growth in vitro. Since Cucurbitacin-I is a potent inhibitor of signal transducer and activator of transcription-3 (STAT3) function, we next demonstrated that STAT3 is hyperactivated in Nf1−/− primary astrocytes, and that STAT3 hyperactivation in NF1-deficient cells results from increased phosphatidylinositol-3 kinase (PI3K)/mTOR/Rac1 pathway signaling. Lastly, we showed that STAT3 inhibition by Cucurbitacin-I blocks human MPNST growth in vivo. Collectively, these findings provide novel insights into neurofibromin growth control, and identify an exciting novel target for future NF1-associated tumor therapeutic drug design.

Materials and Methods

Mice

Six to ten week-old male nu/nu mice were purchased from Taconic Laboratories (Germantown, NY). Nf1^flox/flox mice were intercrossed to generate Nf1^flox/flox pups, while Nf1^flox/flox and Nf1^flox/wt; GFAP-Cre mice were intercrossed to generate mice lacking Nf1 gene expression in GFAP+ (glial) cells (Nf1^GFAPCKO mice), as previously described (17, 18). All mice were used in accordance with established animal studies protocols at the Washington University School of Medicine.

Cell lines

The NF1-deficient ST88-14 malignant peripheral nerve sheath tumor (MPNST) cell line was kindly provided by Dr. Jeff DeClue (NCI, Bethesda, MD). These cells were cultured in Dulbecco’s Modified Eagles’ Medium (DMEM) supplemented with 10% fetal bovine serum and 1% Pen-Strep.

Primary astrocytes

Forebrain astrocyte cell cultures from postnatal day 2 (PN2) Nf1^flox/flox mice were generated as previously described (16, 18, 19). Adenovirus type 5 containing β-galactosidase (Ad5-LacZ) and Cre recombinase (Ad5-Cre) (University of Iowa Gene Transfer Core; Iowa City, IA) were used to produce wild-type (WT) and Nf1-deficient (Nf1−/−) astrocytes, respectively, and neurofibromin loss was confirmed by immunoblot analysis five days after infection [data not shown; (19)]. All experiments were performed on passage 1 or 2 astrocyte cultures.

Pharmacologic inhibitors

Rapamycin (LC Laboratories), LY294002 (Calbiochem), and JSI-124 (Calbiochem) were used. Treatments were for 16–18h unless otherwise indicated. Experiments were performed at least three times with identical results.

High-throughput chemical library screening

The high-throughput compound screen was performed with a cell growth and viability assay in black, clear-bottomed, 96-well culture plates (Corning Costar, Corning, NY) using a Beckman-Coulter Core robotics system, including a FX liquid handler, controlled by the Sagian graphical method development tool (SAMI scheduling software). NF1-deficient ST88-14 cells were seeded in 100 μL/well DMEM supplemented with 10% heat-inactivated fetal bovine serum and 1% glutamine (5,000 cells/well). Plates were maintained in an environmentally-controlled Cytomat incubator until needed for operations, thereby optimizing health and uniform treatment of all plates. Compounds from the NCI 2000 Diversity Set I (in DMSO) were dissolved to 5μM final concentration in DMEM supplemented with 10% heat-inactivated fetal bovine serum and 1% glutamine, and added to cells 24h after cell plating by exchanging the medium. Compound addition was performed with a 96 multichannel head on the FX liquid handler. Compounds were laid out in columns 2–11 of each plate. Control wells for DMSO (0.25% final concentration) and two concentrations of rapamycin (10 nM, 100 nM) were placed in columns 1 and 12. Plates were maintained in the Cytomat incubator for 48h. Cell viability was then determined with resazurin dye (Sigma, St. Louis, MO) (final concentration of 44 μM) after a 3h incubation at 37°C as monitored on a FLUOstar OPTIMA fluorescence reader (BMG Labtech; excitation=544 nm, emission=590 nm).

Statistical analysis and “hit” selection

Cell viability data acquired 48 hours post drug treatment was normalized by dividing treatment fluorescence values by untreated control (DMSO) values. Data were then averaged over three compound replicates. We used two different statistical methods to analyze the data: First, on a plate-by-plate analysis, because the median absolute deviation (MAD) analysis is median-based and is more robust for outliers (20). A hit was defined as values (α) less than -4 MAD from the median (20). A secondary criterion involved the application of a percentile analysis. A hit was defined as less than the 7^th percentile (Supplementary Figure S1). Compounds meeting both criteria were selected for further analysis (Supplementary Table S1). Compounds of interest were selected by review of published structures and literature; those chosen for further characterization are shown in Supplementary Table S2.

Cell proliferation

ST88-14 cells were plated (20,000 cells per well) in 24-well dishes and allowed to adhere for 24 h followed by treatment with Cucurbitacin-I for another 18h. Cells were exposed to [³H]-thymidine (1 μCi/mL) for 4h. 10⁵ astrocytes were allowed to adhere for 24h in 24 well dishes and maintained in serum-free DMEM for 24 h before exposure to [³H]-thymidine (1 μCi/mL). [³H]-thymidine incorporation was determined by scintillation counting (11). All assays were performed in duplicate.

Detection of apoptotic cells

NF1-deficient ST88-14 cells were seeded (50,000 cells/well) in 24 well dishes and treated with Cucurbitacin-I for 18h. The cells were fixed with 4% paraformaldehyde for 1h at room temperature, permeabilized with 0.1% Triton X-100 for 2 min on ice, and the apoptotic cells were detected using a terminal deoxynucleotide transferase-mediated dUTP nick-end labeling (TUNEL) assay kit (Roche, Mannheim, Germany). Nuclei were counterstained with 4,6-diamidino-2-phenylindole (DAPI). Representative photomicrographs were obtained using a fluorescence microscope (Nikon Eclipse TE300 inverted microscope) equipped with a digital camera (Optronics). Apoptotic cells were quantified by counting the number of TUNEL-positive cells as a percentage of the total number of DAPI-stained cells in the field. At least 250 cells were counted.

Immunocytochemistry

WT and Nf1−/− astrocytes (P2) were seeded in 24 well dishes (50,000 cells/well) in astrocyte growth medium for 24h, fixed with 4% paraformaldehyde for 15 minutes at room temperature, permeabilized with 100% methanol, and blocked for 1h in 5% goat serum/0.3% Triton X-100 in PBS at room temperature. To identify phospho-STAT3-expressing cells, astrocytes were incubated with a rabbit anti-phospho-STAT3 (Ser-727) antibody (1:100 dilution, Cell Signaling Technology, Beverly, MA) in 1% BSA/0.3% PBS-Triton X-100 in PBS overnight at 4°C. Alexa Fluor 488-conjugated anti-rabbit IgG secondary antibody (1:200 dilution; Molecular Probes) was used for detection. Nuclei were counterstained with DAPI. Immunolabeling was visualized using the avidin-biotin conjugation method (1:500 dilution; Vectastain ABC Elite; Vector Labs, Burlingame, CA) and 3,3-diaminobenzidine (DAB) tablets (Sigma). Photomicrographs were obtained using a fluorescence Nikon Eclipse TE300 inverted microscope equipped with a digital camera (Optronics).

Retroviral and lentiviral constructs and viral delivery

Mouse-specific small hairpin RNAs (shRNAs) lentiviruses were purchased from Sigma. Raptor (NM_028898; TRCN0000077472) lentivirus was produced and the most effective silencing construct was selected for further study as previously reported (21). Raptor silencing in ST88-14 cells was achieved following three infections with lentivirus. Empty pLKO.1 virus was used as a control.

Murine stem cell virus (MSCV) containing mTOR small interfering RNA (siRNA), constitutively-active Rac1 (Rac1^V12), or dominant negative (DN) Rac1 (Rac1^N17) were generated following 293T cells transfection with τ-helper DNA using Fugene HD (Roche, Mannheim, Germany) (16). 48h later, virus-containing supernatants were filtered through 0.45μM syringe filters. Cells were infected three times and harvested 72h later. MSCV expressing green fluorescent protein (GFP) was included as a control.

Western Blotting

Cells were lysed in standard NP-40 lysis buffer with protease and phosphatase inhibitors for Western blotting as previously described (16). All antibodies were purchased from Cell Signaling Technology (Beverly, MD) and used at a 1:1,000 dilution unless otherwise stated. Primary antibody Rac1 was purchased from Upstate Biotechnology, Temecula, CA. Following horseradish peroxidase–conjugated secondary antibody (Cell Signaling Technology) incubation, detection was accomplished by enhanced chemiluminescence (Amersham Biosciences, Pittsburgh, PA). Densitometry analysis was performed with Gel-Pro Analyzer 4.0 software (Media Cybernetics; Silver Spring, MD) using α-tubulin (Sigma) or non phospho-STAT3, AKT and S6 antibodies for normalization.

Rac1 activation assay

GTP-bound Rac1 was measured using a Rac activation kit (Upstate Biotechnology) according to the manufacturer’s instructions (16). Briefly, ST88-14 cells were lysed, incubated with PAK1-PBD-conjugated agarose beads, washed in lysis buffer, boiled in 2X Laemmli buffer, and bound protein (active Rac1) separated on SDS-PAGE gels for Western blotting. An aliquot of the lysate was saved for Western blotting to ensure equal protein loading.

Immunohistochemical analysis of STAT3 phosphorylation

The human glioma tissue microarray (TMA) slides, containing triplicate cores from 34 patients with sporadic pilocytic astrocytoma, 13 with NF1-associated pilocytic astrocytoma, and 5 cases of normal brain tissue (22), were used in accordance with approved Human Studies Protocols. Immunohistochemistry was performed as previously published (18). Phospho-STAT3 (Ser-727) antibody was used at a 1:100 dilution, and sections were counterstained with hematoxylin. The slides were examined independently by two investigators blinded to both the clinical and pathologic data. Tumors were scored as “positive” if they contained greater than 10% immunoreactive cells or “negative” if they contained fewer than 10% immunopositive cells. A tumor was only scored as “positive” or “negative” if there was concordance between the three cores representing the same tumor on the TMA.

In vivo tumor implantation

One million ST88-14 cells in a total volume of 200 μl were injected subcutaneously (s.c) into the right flank of 6–8 week old male athymic nu/nu mice. Tumors were allowed to grow until they became palpable, and then animals were randomly assigned to two groups. Five mice in each group received daily intraperitoneal (i.p.) injections of vehicle (10% ethanol) or drug (1mg/kg JSI-124 in 10% ethanol) for a total of 5 days. Twice weekly, body weights were recorded and tumor sizes measured by Vernier caliper. Tumor volume (mm³) was calculated using the following equation: (LxW²)/2, where L represents the longest dimension and W the shortest dimension of the tumor (23). Following 5 days of treatment, mice were euthanized and their tumors removed. Apoptotic cell death was detected using a terminal deoxynucleotide transferase-mediated dUTP nick-end labeling (TUNEL) kit in post-fixed tumors sections. Statistical significance was evaluated using the Student’s t test.

Results

Cucurbitacin-I is a novel inhibitor of NF1-deficient cell growth

To identify previously unrecognized downstream targets of neurofibromin relevant to growth regulation, we employed high-throughput chemical library screening (Fig. 1A). This high-throughput screening was based on inhibition of NF1-deficient ST88-14 cell proliferation using DMSO (0.25%) and two concentrations of rapamycin (10 nM, 100 nM) as internal negative and positive controls, respectively. We identified several candidate inhibitors, but only one novel chemical, Cucurbitacin-I. We therefore chose Cucurbitacin-I for more detailed study. To validate the effect of Cucurbitacin-I on NF1-deficient cell growth, we employed [³H]-thymidine and BrdU incorporation: Cucurbitacin-I (10 nM) reduced ST88-14 proliferation, as measured either by [³H]-thymidine incorporation (Fig. 1B) or BrdU incorporation (Supplementary Fig. S2A). Moreover, Cucurbitacin-I also inhibited ST88-14 anchorage-independent cell growth (Supplementary Fig. S2B).

Figure 1. Cucurbitacin-I inhibits NF1-deficient cell growth.

A, Top, ST88-14 cells were treated with the various compounds in all wells of the 96-well plates, except in columns 1 and 12, where ST88-14 cells were treated either with DMSO (0.25%) or with two concentrations of rapamycin (10 nM and 100 nM). Bottom, The chemical structure of Cucurbitacin-I is shown. B, Cucurbitacin-I treatment reduced ST88-14 cell growth by 30% as measured by [³H]-thymidine incorporation (p<0.001). C, Representative TUNEL labeling photomicrographs demonstrate increased numbers of apoptotic ST88-14 cells following Cucurbitacin-I treatment (arrows denote TUNEL-positive cells). The percentages of TUNEL-labeled cells were presented as the mean ± SD of three fields from different separate experiments. Asterisks (*) denote statistically significant differences from vehicle-treated ST88-14 cells. Magnification=20X. Scale bar=100μm. D, Top, STAT3 Ser-727 phosphorylation is reduced by 10 fold in ST88-14 cells following treatment with Cucurbitacin-I. STAT3 Tyr-705 phosphorylation was not detected following 1 min exposure. Bottom, No changes in STAT3 Tyr-705 phosphorylation were observed. Total STAT3 protein expression was used as a protein loading control.

To extend these findings to other Nf1-deficient cell populations, we showed that Cucurbitacin-I (10 nM) inhibited Nf1−/− astrocyte proliferation (Fig. 2B), with minimal effects on wild-type (WT) astrocytes. Similarly, Cucurbitacin-I treatment reduced the proliferation of Nf1-deficient mouse low-grade K1861 glioma cells (Supplementary Fig. S3A).

Figure 2. Cucurbitacin-I inhibits STAT3 activation in Nf1-deficient astrocytes.

A, Top, STAT3 Ser-727 phosphorylation is increased 3-fold in Nf1−/− relative to WT astrocytes. Bottom, Nf1^GFAPCKO mouse brains have increased STAT3 Ser-727 phosphorylation compared to WT (Nf1^flox/flox) mouse brains. STAT3 expression was used as a protein loading control. Right, Increased nuclear STAT3 Ser-727 phosphorylation is observed in Nf1−/− relative to WT astrocytes by immunocytochemistry. Magnification=20X. Scale bar=100μm. B, Cucurbitacin-I treatment reduces Nf1−/− astrocyte proliferation as assessed by [³H]-thymidine incorporation. Asterisks (*) denote statistically significant differences from vehicle-treated Nf1−/− astrocytes. C, STAT3 Ser-727 phosphorylation was decreased following Cucurbitacin-I treatment in Nf1−/− astrocytes (quantitation shown below). D, Representative photomicrographs of human glioma TMA sections stained with STAT3 Ser-727 phospho-antibody. The number of phospho-STAT3 Ser-727-immunopositive specimens in each group is shown in the graph.

To determine whether this reduced cell growth reflected increased cell death, we measured apoptosis in ST88-14 cells treated with 10 nM Cucurbitacin-I using the TUNEL assay. Cucurbitacin-I treatment increased the percentage of TUNEL positive ST88-14 cells by 8-fold compared to vehicle-treated cells (Fig. 1C). Identical results were also observed using K1861 cells (Supplementary Fig. S3B). These findings demonstrate that Cucurbitacin-I decreases NF1-deficient cell growth by inducing apoptosis.

Cucurbitacin-I inhibits STAT3 activation in NF1-deficient cells

One of the mechanisms underlying Cucurbitacin-I function is inhibition of STAT3 function (24, 25). Since neurofibromin has not previously been implicated in STAT3 regulation, we first measured STAT3 activation using phospho-specific antibodies. Cucurbitacin-I treatment of ST88-14 cells (Fig. 1D, top) or K1861 mouse glioma cells (Supplementary Fig. S3C) resulted in suppression of STAT3 activation, as reflected by reduced phosphorylation on Ser-727. In contrast, no change in STAT3 Tyr-705 phosphorylation was observed (Fig. 1D, bottom). Identical results were obtained using a commercial source of Cucurbitacin-I (JSI-124; data not shown).

Second, to directly demonstrate that neurofibromin regulates STAT3 activation, we examined STAT3 phosphorylation in Nf1−/− and WT astrocytes: STAT3 Ser-727 phosphorylation was increased 4.8-fold in Nf1−/− compared to WT astrocytes (Fig. 2A, top). Similar results were observed using brain lysates from Nf1^GFAPCKO mice lacking Nf1 expression in GFAP+ astrocytes: Compared to Nf1^flox/flox (“WT”) mice, Nf1^GFAPCKO mouse brains exhibited 5-fold greater STAT3 (Ser-727) phosphorylation in vivo (Fig. 2A, bottom). No STAT3 Tyr-705 hyperphosphorylation was observed in Nf1−/− astrocytes compared WT astrocytes (data not shown). In addition, primary Nf1−/− neocortical neural stem cells likewise exhibited increased STAT3 Ser-727 phosphorylation relative to their WT counterparts (Supplementary Fig. S4A). Since STAT3 activation (phosphorylation) is associated with its translocation to the nucleus, we found that Nf1−/− astrocytes had greater nuclear STAT3 immunostaining compared to WT astrocytes (Fig. 2A, right).

Third, to show that Cucurbitacin-I inhibits Nf1−/− astrocyte cell proliferation as well as STAT3 activation, we examined the effect of Cucurbitacin-I treatment on [³H]-thymidine incorporation and STAT3 Ser-727 phosphorylation in WT and Nf1−/− astrocytes. We observed a 1.4-fold reduction in cell proliferation (Fig. 2B) and a 3-fold reduction in Ser-727 STAT3 phosphorylation (Fig. 2C) in Nf1-deficient astrocytes following Cucurbitacin-I treatment. The effects of Cucurbitacin-I treatment on Nf1-deficient astrocyte proliferation were similar to those observed following STAT3 shRNAi knockdown (Supplementary Fig. S4B).

Finally, to determine whether STAT3 is also activated in brain tumors from patients with NF1, we examined STAT3 Ser-727 phosphorylation by immunohistochemistry in NF1-associated and sporadic pilocytic astrocytomas (PAs): Whereas 41% of NF1-associated PAs were phospho-STAT3-immunopositive, only 7% of the sporadic PAs were phospho-STAT3-immunoreactive (Fig. 2D). No phospho-STAT3 Ser-727 expression was detected in the normal brain tissue specimens (n = 5). Collectively, these data demonstrate that neurofibromin is a negative regulator of STAT3 activity in vitro and in vivo.

STAT3 activation is regulated by mTOR in NF1-deficient cells

We next sought to determine how neurofibromin regulates STAT3. Our previous studies demonstrated that neurofibromin astrocyte growth control is mediated by AKT/mTOR pathway signaling (11, 16). Here, we showed that pharmacologic inhibition of either PI3-K (LY294002; Fig. 3A) or mTOR (rapamycin; Fig. 3B, left) blocked STAT3 Ser-727 phosphorylation in ST88-14 cells. Second, we demonstrated that rapamycin blocks STAT3 Ser-727 phosphorylation (activation) in both human ST88-14 and mouse K1861 cells (Fig. 3B, right and Supplementary Fig. S3D) as well as Rac1 activation in ST88-14 cells (Fig. 3B, right). Third, we genetically silenced mTOR expression using an mTOR-specific siRNA virus in ST88-14 cells, and found reduced STAT3 Ser-727 phosphorylation following mTOR knockdown (Fig. 3C). Identical results were also observed following rapamycin inhibition (Supplementary Fig. S5A) or mTOR knockdown (Supplementary Fig. S5B) in Nf1−/− astrocytes. Fourth, since rapamycin-sensitive mTOR signaling operates largely through mTOR complex-1 (TORC1), we silenced the major TORC1 protein, raptor, in ST88-14 cells using shRNAi: Raptor knockdown also reduced STAT3 Ser-727 phosphorylation (Fig. 3D). Consistent with TORC1 negative regulation of AKT activity, we also observed increased AKT Ser-473 phosphorylation (activation) in ST88-14 cells following both rapamycin (Fig. 3B, left) and raptor shRNAi knockdown (Fig. 3D).

Figure 3. STAT3 activation is regulated by mTOR in NF1-deficient cells in vitro.

A, The PI3-Kinase inhibitor, LY294002, inhibits ST88-14 STAT3 Ser-727 phosphorylation. Phospho-AKT antibodies were used to demonstrate PI3-Kinase activity. Total AKT and STAT3 served as loading controls. B, Left, Rapamycin inhibited STAT3 Ser-727 phosphorylation and results in increased AKT Ser-473 phosphorylation. Phospho-S6 was used to demonstrate mTOR activity. Total S6, AKT and STAT3 serve as internal loading controls. Right, GTP-bound Rac1 was immunoprecipitated from ST88-14 cells treated with vehicle (ethanol) or rapamycin (10 nM and 100nM) using PAK1-PBD affinity chromatography. Equal protein loading was confirmed by total Rac1immunoblotting. GTP-bound Rac1 activation was decreased in rapamycin-treated cells. STAT3 Ser-727 and S6 phosphorylation were also reduced by rapamycin treatment. C, siRNA-mediated mTOR knockdown inhibited STAT3 Ser-727 phosphorylation. Total mTOR and STAT3 served as loading controls. D, Raptor silencing by shRNAi inhibited ST88-14 STAT3 Ser-727 phosphorylation and resulted in increased AKT Ser-473 phosphorylation. α-tubulin, AKT and STAT3 served as loading controls.

Fifth, based on our studies showing that mTOR regulates Nf1−/− astrocyte growth in a Rac1-dependent manner (16), we examined STAT3 Ser-727 phosphorylation in ST88-14 cells following the introduction of a constitutively-activate Rac1 mutant (Rac1^V12) or a dominant-inhibitory Rac1 mutant (Rac1^N17): Rac1^V12 increased STAT3 Ser-727 phosphorylation (Fig. 4A), whereas Rac1^N17 attenuated STAT3 Ser-727 phosphorylation (Fig. 4B). Similarly, Rac1^N17 expression in Nf1−/− astrocytes reduced STAT3 Ser-727 phosphorylation (Supplementary Fig. S5C, bottom), whereas Rac1^V12 expression in WT astrocytes increased STAT3 Ser-727 phosphorylation (Supplementary Fig. S5C, top). Lastly, as we previously reported (16), active Rac1^V12 expression increased ST88-14 cell growth (Supplementary Fig. S6A), whereas dominant inhibitory Rac1^N17 expression reduced both ST88-14 (Supplementary Fig. S6B) and Nf1−/− astrocyte cell proliferation (Supplementary Fig. S6C). Together, these data demonstrate that mTOR/Rac1 activation regulates STAT3 function in NF1-deficient cells.

Figure 4. Rac1 regulates STAT3 phosphorylation in NF1-deficient cells.

A, Constitutively active Rac1 (Rac1^V12) increases STAT3 Ser-727 phosphorylation and cyclin D1 expression. B, Dominant-inhibitory Rac1 (Rac1^N17) suppresses STAT3 Ser-727 phosphorylation and cyclin D1 expression. α-tubulin and STAT3 serve as loading controls. C, Cyclin D1 expression was reduced following Cucurbitacin-I treatment. α-tubulin was used as a protein loading control.

STAT3 regulates cyclin D1 expression in NF1-deficient cells

To determine whether STAT3 regulates cyclin D1 expression (26), cyclin D1 levels were measured following Cucurbitacin-I treatment. Cyclin D1 expression was reduced in ST88-14 cells (Fig. 4C) and Nf1−/− astrocytes (Supplementary Fig. S5D) after Cucurbitacin-I treatment. Since cyclin D1 is increased following Rac1^V12 expression (Fig. 4A) and decreased following Rac1^N17 expression (Fig. 4B), we conclude that STAT3 regulation of cyclin D1 resulted from Rac1 activation in NF1-deficient cells.

STAT3 inhibition inhibits ST88-14 tumor growth in vivo

To provide in vivo support for STAT3-mediated NF1 growth control, we employed JSI-124 to inhibit STAT3 function in ST88-14 tumor explants. Fifteen days post-implantation, male nu/nu mice with visibly-growing tumors were randomly assigned to daily i.p. injections of either JSI-124 (1mg/kg) or vehicle for 5 days. After 5 days, tumors in the JSI-124-treated group were ~4-fold smaller than those from the vehicle-treated group (p≤0.008) (Fig. 5A–B). Next, to determine whether the decrease in tumor growth reflected increased cell death, we measured apoptosis in frozen tumor sections, and found a 2-fold increase in apoptosis following JSI-124 treatment (p<0.0001; Fig. 5C). Similar results were also obtained in a second independent study (Supplementary Fig. 7A). No effect of JSI-124 treatment on animal body weight (Supplementary Fig. 7B) or apoptosis in the liver (data not shown) was observed. These results demonstrate that STAT3 inhibition results in decreased NF1-deficient cell and tumor growth in vitro and in vivo.

Figure 5. JSI-124 treatment reduced NF1-deficient tumor growth in vivo.

A, Representative tumor sizes (arrows) are shown for each treatment group. B, Tumor volumes were calculated from tumor measurements obtained on the indicated days. Tumor volumes were reduced by ~4-fold (p≤ 0.008) in mice following 5 days of treatment. C, JSI-124 treatment increased apoptosis in vivo by 2-fold relative to vehicle treatment. The percent of TUNEL-positive cells is presented as the mean± SD of ten fields from either JSI-124-treated or vehicle-treated mice. Asterisks (*) denote statistically significant (p<0.0001) differences from vehicle-treated mice.

Discussion

The identification of the NF1 gene and the subsequent discovery that neurofibromin regulates the RAS/mTOR pathway have ushered in a new era of biologically-targeted therapies for NF1-associated tumors. In this regard, clinical studies using RAS inhibitors have already been initiated and those focused on mTOR blockade have recently commenced. Unfortunately, RAS inhibition using farnesyltransferase inhibitors (FTIs) have not shown efficacy for the treatment of NF1-associated peripheral nerve sheath tumors (10). Based on these studies, subsequent translational research has focused on RAS downstream effectors, including Raf/MEK and AKT/mTOR. Several laboratories have recently shown that hyperactivation of the mTOR pathway underlies the growth advantage seen in NF1-deficient glioma and MPNST cells, and that mTOR inhibition, using the macrolide rapamycin, results in decreased optic glioma and MPNST growth in vivo (13–15). However, it is worth noting that rapamycin does not induce apoptosis and does not lead to a durable response following drug cessation (13). In addition, human MPNST cells grown as explants in mice exhibit increased AKT activation following rapamycin treatment as a result of augmented TORC2-mediated AKT phosphorylation (15). While rapamycin analogs are now in clinical trial for human NF1-associated peripheral nerve sheath tumors, this biologically-based therapy may not result in sustained tumor shrinkage.

For these reasons, we sought to discover new targets using an unbiased high-throughput chemical library screening approach. In this screen, we found several promising compounds. A number of our top candidates were known chemotherapeutic agents (mitomycin, daunorubicin, and topotecan), while others were crude inhibitors of major biological processes (bouvardin, tubulosine, and breflate). Only one compound, Cucurbitacin-I, was novel and was therefore selected for further study. Additionally, we chose Cucurbitacin-I, since it is known to inhibit STAT3, a signaling molecule not previously implicated in neurofibromin growth control, and against which inhibitors have been used to treat other cancers (27, 28). Consistent with these other reports, we found that Cucurbitacin-I blocked the proliferation of Nf1−/− astrocytes in vitro and NF1-deficient MPNST cells by suppressing STAT3 activation in vitro and in vivo.

STAT3 is a transcription factor whose activity is regulated by phosphorylation: Tyrosine phosphorylation (residue Tyrosine-705) results in STAT3 dimerization and translocation to the nucleus, whereas transcriptional activation requires phosphorylation on residue Serine-727 (29). In NF1-deficient cells, we observed increased phosphorylation of STAT3 on Ser-727 and not Tyr-705. This is consistent with previous reports demonstrating that RAS activation increases STAT3 Ser-727 phosphorylation (26). Moreover, phosphorylation of STAT3 at Ser-727 is sufficient for STAT3 activation in prostate cancer cells, independent of Tyr-705 phosphorylation (30), and STAT3 Ser-727 phosphorylation is essential for postnatal survival and cell growth in mice, such that mice with a Stat3 allele in which the serine residue has been converted to alanine exhibit increased apoptosis (31).

Since neurofibromin has never been shown to regulate STAT3 activity, we sought to determine whether STAT3 was de-regulated in NF1-deficient cells through previously implicated neurofibromin signaling pathways. We found that STAT3 activity was regulated in NF1-deficient primary cells and tumor cells in an AKT/mTOR-dependent manner. Although unique to neurofibromin/AKT/mTOR pathway regulation, STAT3 hyperactivation is mediated by PI3K/AKT signaling in other cell types (32). Based on our previous studies demonstrating that AKT/mTOR growth regulation in Nf1-deficient astrocytes results from mTOR-mediated Rac1 activation, we show that STAT3 activation is controlled by Rac1 activation downstream of mTOR in Nf1−/− mouse astrocytes and NF1-deficient human MPNST cells. The finding that Rac1 controls STAT3 function in NF1-deficient cells is supported by previous studies demonstrating that Rac1 can bind to and regulate STAT3 activation (33–36). Together, our findings establish a more complete model for neurofibromin growth regulation involving mTOR/Rac1/STAT3 signaling (Fig. 4D).

STAT3 plays a central role in regulating oncogenesis by controlling the transcription of several target genes essential for cell cycle progression, apoptosis, and proliferation (37). In this regard, Ser-727 STAT3 phosphorylation is associated with increased cyclin D1 expression (26). Similarly, we found that cyclin D1 expression was increased in Nf1-deficient astrocytes relative to wild-type astrocytes, and is reduced following either Rac1 or STAT3 inhibition. While cyclin D1 may be partly responsible for promoting cell growth in NF1-deficient cells, other STAT3 target genes, including regulators of apoptosis, are likely involved in facilitating malignant transformation and continued growth following neurofibromin loss. Additional studies will be required to identify and validate these other potential STAT3 transcriptional target genes.

Our finding that neurofibromin regulates STAT3 activation in primary Nf1−/− astrocytes, neural stem cells, and low-grade glioma cells coupled with the observation that STAT3 inhibition blocks Nf1-deficient astrocyte and low-grade glioma cell proliferation has particular relevance to brain tumors in children with NF1. Previously, we have shown that optic glioma growth in Nf1 genetically-engineered mice can be suppressed by treatment with rapamycin; however, this inhibitor is only cytostatic and murine optic glioma growth resumes following the cessation of rapamycin treatment (13). The observation that STAT3 blockade results in apoptosis is exciting, as this anti-tumor effect might result in a durable response similar to what we have observed in Nf1 optic glioma mice treated with conventional genotoxic therapy (temozolomide). Moreover, the potential use of STAT3 inhibitors for anti-glioma therapy is underscored by results from numerous preclinical studies employing high-grade glioma cell lines (25, 38–42). Finally, it is conceivable that STAT3 blockade could be combined with currently employed brain tumor therapies to improve outcome. In this regard, JSI-124 treatment has been shown to sensitize glioma cells to temozolomide, nitrosourea, and cisplatin in a synergistic manner (43). The ability to employ STAT3 inhibitors to reduce the overall dose of alkylating chemotherapy is highly attractive, especially in children whose developing brains are most sensitive to the untoward effects of chemotherapy. Future studies will be required to explore these possibilities and to develop potential strategies to integrate STAT3 adjuvant therapies into our current treatment approach to NF1-associated brain tumor therapy.