Targeted Inhibition of the Dual Specificity Phosphatases DUSP1 and DUSP6 Suppress MPNST Growth via JNK

Malignant peripheral nerve sheath tumors (MPNSTs) are devastating chemotherapy- and radiation-resistant soft tissue sarcomas. Rare in the general population, half of MPNSTs arise in patients with the autosomal dominant genetic disorder called neurofibromatosis type 1 (NF1). MPNSTs in patients with NF1 frequently occur in the context of preexisting benign plexiform neurofibroma, and are the leading cause of death in adults with NF1 (1–3). NF1 loss results in activation of RAS proteins that stimulates the MAPK family including ERK1 and ERK2, the JNKs 1, 2, and 3, and the p38 family of kinases (α, β, δ, and γ) resulting in tumor cell growth. However, from gene expression microarray data, we and others have observed that key negative regulators of MAPK signaling are also increased in NF1-deleted tumors (4, 5). Dual specificity phosphatase (DUSP) expression is transcriptionally upregulated by increased ERK activity, providing a negative feedback loop to suppress RAS signaling (6). Within the DUSP family of proteins, the MAP kinase phosphatases (MKP) dephosphorylate tyrosine and serine/threonine residues on ERK, JNK, and p38 (7). DUSPs are subdivided on the basis of cellular localization and substrate specificity. There are over 10 members of this family including DUSP1/MKP-1, which is a nuclear MKP, and DUSP6/MKP-3, which is cytosolic (8). DUSPs substrate specificities depend on the cellular context (8).

Studies have described MAPK negative regulators, including DUSP6, as tumor suppressors. Thus, decreased expression of DUSP genes and proteins can occur in human cancer, including in tumors from pancreas and lung (9, 10). In contrast, DUSPs are upregulated in some leukemia types including pre-B acute lymphoblastic leukemia (ALL) and DUSP1/6 inhibition has antitumor effects in gastric, breast, and leukemia cancer models (11–14). In these cases, it is believed that DUSPs help tumors adapt to excessively high levels of growth factor signals. DUSP inhibition can acutely hyperactivate MAPK signaling, triggering activation of effectors such as p53, ATM, and CHK1/2, which arrest cell growth and/or induce cell death. If NF1-deleted cells with RAS activation depend on DUSPs for growth and survival, then inhibiting DUSP could be effective in NF1 and in the many sporadic tumor types now known to show NF1 loss (15).

Despite advances in understanding MPNST biology, there are no effective available MPNST treatments (2). Some potential targets are growth factor receptors, and the PI3K-AKT-mTOR and Wnt/β-catenin signaling pathways (16, 17). Targeting MEK activity provides growth suppression in certain MPNST models, but effects, when present, are transient (5, 18). Aurora kinase inhibition is cytostatic (19) and with combination therapy may show additive effects (20).

Targeting DUSP is an active area of preclinical development. (E)-2-benzylidene-3-(cyclohexylamino)-2,3-dihydro-1H-inden-1-one (BCI) is a specific allosteric inhibitor of DUSP1/6 (12, 13, 21). Recently, in mouse models of leukemia and gastric cancer, BCI was effective in killing tumor cells in vivo (12–14, 21). We hypothesized that inhibiting DUSPs might be selectively toxic to MPNST cells. In this study, we conducted molecular analyses to better understand the role of DUSP1 and DUSP6 in MPNSTs, using a combination of MPNST cell lines, in vivo cell line xenografts and, a novel patient-derived xenograft (PDX) MPNST model. The results from our studies demonstrate that DUSP1/6 protein expression are important for growth and survival in NF1-deleted MPNSTs and suggest that targeting DUSPs may be a therapeutic option for treating MPNSTs.

Cell lines included immortalized human Schwann cells (iHSC; ref. 22) with creation and characterization of NF1-deficient HSC1λ cell lines (Williams and colleagues, unpublished). Briefly an early, common, exon of NF1 (exon 10) was targeted to generate indel mutations using CRISPR/Cas9. Those found to be harboring homozygous frameshift mutations resulting in an early termination codon were considered NF1^−/−. Cell lines showing homozygous wild-type (WT) sequences were considered isogenic WT controls. MPNST cell lines, ST8814, S462.TY, STS26T, and 90-8, were obtained as described previously (23, 24). All MPNST cell lines were derived from patients with NF1, except STS26T. Cell lines were maintained in DMEM and 10% FBS and were confirmed negative for Mycoplasma prior to use. Verification of cell lines was performed by short tandem repeat DNA profiling (Genetica DNA Laboratories). Inhibitors used were BCI (Axon Medchem), PD0325901 MEK inhibitor (gift from Kevin M. Shannon, University of California, San Francisco, CA), and JNK inhibitors, JNK-IN-8 (Selleckchem) and SP600125 (LC Laboratories). Drugs were prepared as 10 mmol/L stocks in DMSO and were diluted into media for cell culture treatments. Cell viability was quantified 72 hours after treatment by CellTiter 96 AQueous One Solution Cell Proliferation Assay (MTS; Promega) as per manufacturer's instructions. Combination index data analysis was done using Calcusyn software (Version 1). Combination indices (CI) less than 0.9 indicate synergy, 0.9 to 1.1 indicate additivity, and greater than 1.1 indicate antagonism.

Cells were serum starved for 16 hours in DMEM media (Life Technologies). MPNST cells were pretreated with inhibitor for 15 to 60 minutes then stimulated with DMEM and 10% FBS (Gemini Bio Products) for 15 to 60 minutes. The iHSCs were stimulated with DMEM-F12 with N2 supplement (Life Technologies) and 10 ng/mL β-Heregulin (β-HRG). Signaling experiments were terminated with a cold PBS wash, and cell lysates were generated for Western blotting. Cell death was analyzed with the In Situ Cell Death Detection Kit (terminal deoxynucleotidyl transferase–mediated dUTP nick end labeling, TUNEL) and TMR red (Roche), by fluorescence microscopy.

Tissues and cells were lysed in a RIPA buffer supplemented with protease and phosphatase inhibitors (Roche). Protein concentrations were determined with the BCA Protein Assay Reagent (Thermo Fisher Scientific). Antibodies used were Dusp1 (Millipore, 07-535), DUSP6 (Abcam, 76310), p38 (Abcam, 7925), p-p38 (Life Technologies, 44684), p53 (Santa Cruz Biotechnology, sc-47698). Antibodies from Cell Signaling Technology were p-DUSP-1 (2857), ERK, p-ERK (4695, 4370), JNK, p-JNK (9252, 4668), c-jun, p-c-jun (2315, 2361), p-p38 (9211), PARP (9542), cyclin D1 (2922), p-ATM (13050), p-RB (retinoblastoma; 9301), and B-actin (5125). Blots were analyzed by densitometry using ImageJ software.

For lentiviral small hairpin RNA (shRNA) infection, MPNST cells at 70% confluence were infected with lentiviral particles containing shRNAs targeting DUSP1 (TRCN0000002516, TRCN0000002517) or DUSP6 (TRCN0000002436, TRCN0000002437) at a multiplicity of infection (MOI) of 5. These were purchased from the Lenti-shRNA Library Core [Cincinnati Children's Hospital Medical Center (CCHMC), Cincinnati, OH; originally from the Sigma Mission Consortium shRNA library]. pLKO.1-puro control was used as the nontargeting (NT) control. Cell lines were infected overnight in the presence of 10 μg/mL polybrene. Stable cell populations were selected with puromycin (2 μg/mL) 48 hours later.

Total RNA was isolated from cells using the RNeasy Kit (Qiagen). The cDNA was made using the High-Capacity cDNA Reverse Transcription Kit (Life Technologies). Real-time PCR was performed on cDNA using SYBR Green (Applied Biosystems). Relative quantification was performed using the ΔΔC_t method, n = 6.

Formalin-fixed, paraffin-embedded tissue blocks were sectioned at 5 μm. Tissue sections were deparaffinized in xylene and rehydrated in ethanol, then water. Sections were stained with hematoxylin and eosin (H&E) or Masson trichrome, in which blue stain indicates collagen. For IHC, sections were subjected to heat-induced antigen retrieval in citrate buffer. Sections were incubated with primary antibody overnight at 4°C and biotinylated secondary antibodies were detected with horseradish peroxidase–conjugated Streptavidin (Elite ABC, Vector Laboratories) treated with DAB substrate (Vector Laboratories). Sections were counterstained with hematoxylin. The primary antibodies used for IHC include Anti-Human Nucleoli Antibody (Abcam, ab190710) and Cleaved Caspase 3 (CC3; Cell signaling Technology No. 9661).

Copy-number alteration (CNA) data on 51 primary MPNSTs (NF1-associated and sporadic) was from a published GSE3388 dataset [Agilent Human Genome CGH Microarray Kit (4 × 44K); ref. 25]. A circular binary segmentation (CBS) algorithm was applied to the log₂ ratios of intensity values from tumor and normal to reduce local noise effects. CBS calculates a likelihood-ratio statistic for each array probe by permutation to locate change points (26). After initial segmentation, smoothing, and postsegmentation normalization processes, CGHcall algorithm (27) was applied to assign each segment an aberration label. We chose the 5-state (amplification, gain, normal, loss, and double loss) prediction design and the probability of each state was calculated, then one state with the highest probability was chosen as a final CNA calling label for each segment.

We reprocessed the published data [Gene Expression Omnibus (GEO) accession #:GSE14038, Affymetrix GeneChip HU133 Plus 2.0] using Bioconductor's affy package (28). Differentially expressed genes between 2 groups were predicted using an adjusted t test after multiple testing (Benjamini–Hochberg) and Bioconductor's limma package (29). ANOVA test was performed using R (v3.3.1).

All animal studies were performed according to the guidelines of the Institutional Animal Care and Use Committee of CCHMC. Mouse neurofibroma was obtained from Nf1 fl/fl;Dhh-Cre mice as described previously (30). For the cell line–based xenograft model, S462.TY cells (1.25 × 10⁶) in 30% Matrigel (BD Biosciences) were injected subcutaneously into the flank of 5- to 6-week-old female athymic nude mice. For the PDX MPNST model, biopsied tumor tissue was implanted subcutaneously directly into NOD.Cg-Prkdc^scid Il2rg^tm1Wjl/SzJ (NSG) mice. Tumors used for the study were grown in a donor mouse to a volume of approximately 2,500 mm³. The tumor was harvested, cut into 1.5-mm³ pieces, and inserted subcutaneously into the flank of each mouse. Mice bearing F3–F4 grafts were used in this study. Drug dosing was started when tumors reached 150 to 200 mm³. We measured tumors and weighed mice every 3 days. Tumor volume was calculated as follows: L × W2 (π/6), where L is the longest diameter and W is the width. The length of treatment was determined by the time tumors in each vehicle-treated group reached a maximum allowable tumor volume of 2,500 mm³ (21 days for the PDX and 27 days for the S462.TY model).

Analyses were performed in GraphPad Prism software version 7.0. Data were mean ± SD or SEM as indicated and were analyzed using ANOVA with Tukey post hoc test or a Student t test, 2-sided. P values <0.05 were considered significant.

We used published gene expression profiles to analyze the differential expression of DUSP genes (5). DUSP1 and DUSP6 mRNAs were consistently elevated in neurofibroma and MPNSTs as compared with human Schwann cells, which is consistent with their transcriptional upregulation by mutant NF1 (Fig. 1A). In a genetically engineered mouse model of neurofibroma, we examined DUSP protein levels by immunoblot analysis. DUSP1 was 4× higher in neurofibroma compared with Nf1^+/+ (WT) sciatic nerves (Fig. 1B). Anti–p-DUSP1 detects a phosphorylation site known to stabilize the protein (31); p-DUSP1 expression was elevated 7× (Fig. 1B). DUSP6 protein was 13× higher in neurofibroma compared with Nf1^+/+ sciatic nerves (Fig. 1B). Compared with WT nerve, neurofibromas consistently showed elevated p-ERK, p-c-jun and decreased p-JNK, with variable levels of p-p38 (Supplementary Fig. S1A). In iHSCs lacking NF1, DUSP1 was 62% higher, p-DUSP1 2× and DUSP6 3× higher than isogenic WT iHSC (Fig. 1C), consistent with the idea that in Schwann cells, NF1 deletion can drive increases in DUSP protein expression. DUSP protein was also elevated in human MPNST cell lines (n = 4), showing an average 30% increase in DUSP1 and an average 3× higher expression of p-DUSP1 compared with iHSC (Fig. 1C). DUSP6 protein was 4 to 15× higher in MPNSTs compared with iHSC, with the highest expression in the NF1-deleted MPNST cell line, ST8814 (Fig. 1C). Compared with iHSC, MPNST cell lines have increased p-ERK with variable levels of p-JNK and p-p38 (Supplementary Fig. S1B).

DUSP1 and DUSP6 are on human chromosomes 5q35.1 and 12q21.33, respectively. Previous studies of MPNSTs have identified chromosomal gains in these regions (25, 32). We reasoned that during the transformation process, some MPNSTs may genetically alter DUSPs to adapt to increased growth factor signaling. Amplifications of HGF, MET, and PDGFRA genes are present in some MPNSTs, suggesting a role in NF1 tumor progression (33). We analyzed data from human plexiform neurofibroma and primary MPNSTs (NF1-associated and sporadic) to evaluate the genomic status of DUSP1 and DUSP6 (25). We found no copy-number changes in plexiform neurofibroma (Fig. 1D). In NF1-associated MPNSTs, 31.25% of the tumors had copy-number gains in DUSP1 or DUSP6, with 12% having gains in both (Fig. 1D). In sporadic MPNSTs, 34% had gains in DUSP1 or DUSP6 with 5.7% having amplifications of DUSP6 (Fig. 1D). Single-copy losses of DUSP1 or DUSP6 were detected in 25% of NF1-associated MPNSTs with DUSP1 or DUSP6 loss at 8.6% and 5.7% in sporadic MPNSTs (Fig. 1D). Therefore, copy-number changes are not likely to be a major contributor to DUSP1/6 expression in most MPNSTs.

We used shRNA to knock down DUSP1 or DUSP6 in the MPNST cell line ST8814. Using 2 specific shRNAs and NT controls, we validated DUSP1/6 knockdown of mRNA and a 55% to 75% decrease in protein (Fig. 2A–D). When DUSP1 was knocked down, levels of DUSP6 did not change and vice versa (Fig. 2C and D). Proliferation was decreased by 28% to 43% on DUSP1 knockdown and 40% to 51% on DUSP6 knockdown (Fig. 2E and F). DUSP1 knockdown cells treated with DUSP6 shRNA showed a further 30% decrease in proliferation (Fig. 2E). DUSP6 knockdown cells treated with DUSP1 shRNA showed a further 20% decrease in proliferation (Fig. 2F). Thus, directly decreasing DUSP1/6 levels, alone or in combination, decreases proliferation of MPNST cells.

We next tested the consequences of knockdown of DUSP1 or DUSP6 on MAPK signaling in the MPNST cell line, ST8814. In DUSP1 knockdown, 15 to 60 minutes after serum stimulation, p-ERK and p-JNK were 1.75× higher compared with NT controls (Fig. 3A). The phosphorylated JNK substrate p-c-jun was 2× higher, and p-p38 was unchanged (Fig. 3A). In contrast, in DUSP6 knockdown ST8814 cells, p-ERK was only modestly increased (1.3×; Fig. 3B). DUSP6 knockdown resulted in 2.58× higher p-JNK, and its substrate p-c-jun was 2× higher; p-p38 was unchanged compared with NT controls (Fig. 3B). DUSP1 knockdown cells treated with DUSP6 shRNA showed a 26% increase in p-ERK and larger increases in p-JNK and p-c-jun (71% and 56%, respectively) compared with DUSP1 knockdown alone (Fig. 3C). Total protein levels of these markers were unchanged except for c-jun in DUSP6 knockdown, which showed a modest increase (Supplementary Fig. S2A–S2C). We confirmed these results in the S462.TY MPNST cell line (Supplementary Fig. S2D and S2E). DUSP knockdown results in decreased proliferation and increased levels of p-ERK, p-JNK, and p-c-jun (Supplementary Fig. S2D and S2E). These data suggest that in MPNST cells DUSP1/6 play major roles in dephosphorylation of MAPK.

We assessed the effect of DUSP1/6 downregulation on cell-cycle regulation and cell growth markers. Interestingly, upon DUSP1/6 silencing, cyclin D1 levels were reduced despite the high levels of MAPK signaling (Fig. 3D). Although sustained activation of ERK is required for cell-cycle progression, hyperactivation of ERK or JNK signaling can lead to stabilization of p53 that can induce cell-cycle arrest or cell death (34, 35). Indeed, after knockdown of DUSP1, DUSP6, or both, p53 and p-ATM were increased, suggesting that this pathway is active in MPNST cells (Fig. 3D). In addition, knockdown of DUSP1 or DUSP6 induced a 64% increase in PARP cleavage and increased TUNEL⁺ cells compared with NT controls, indicating the induction of apoptotic cell death (Fig. 3D and E). Collectively, our results identify key pathways by which DUSPs are likely to regulate proliferation and cell death in MPNST cells.

To verify these results, we tested the effect of the DUSP1/6 inhibitor, BCI, on the growth of the NF1-deleted MPNST cell lines, ST8814 and S462.TY, and on NF1^+/+ cell line, iHSC-1λ, and the sporadic MPNST cell line, STS26T. We found that 2 to 4 μmol/L BCI significantly inhibited MPNST cell growth 3 days after treatment (Fig. 4A). These concentrations were similarly effective in leukemia and gastric cancer cell lines (13, 14). The iHSC-1λ and STS26T MPNST cell lines were less sensitive to BCI (Fig. 4A). We evaluated MAPK pathway activation by Western blot analysis. Cells were serum-starved overnight, incubated with 2 μmol/L BCI, and then stimulated with serum-containing media. After 1 hour, p-ERK, p-JNK, p-c-jun, and total c-jun were elevated in the BCI-treated MPNST cell lines, ST8814 and S462.TY, but did not change in iHSC-1λ (Fig. 4B; Supplementary Fig. S3A). There were no detectable differences in p-p38 (Fig. 4B). In contrast, in STS26T cells only p-p38 was increased and cyclin D1 decreased (Supplementary Fig. S3B). Within 24 hours, BCI decreased total PARP and increased cleaved PARP and CC3, indicative of apoptotic cell death in NF1-deficient ST8814 and S462.TY cells (Fig. 4C). Also, there were 1.5 to 2× increases in TP53 and p-ATM in these BCI-treated cells correlating with reduced phospho-RB levels and reduced cell proliferation (Fig. 4C).

Given that blocking DUSPs, genetically or pharmacologically, increased p-JNK, we evaluated the possibility that JNK is a critical mediator of decreasing cell proliferation and/or increasing cell death, by testing whether JNK inhibition would alter BCI effects in MPNSTs. In ST8814 MPNST cells treated with the irreversible JNK inhibitor JNK-IN-8 at 0.6 to 2.5 μmol/L, cell proliferation was modestly increased (Supplementary Fig. S3C). These concentrations inhibit JNK activity as indicated by reduced p-c-jun (Supplementary Fig. S3D; ref. 36). Targeting DUSP and JNK in combination reduced the effectiveness of BCI with combination indexes 1.2-3.9 indicating antagonism (Fig. 4D). Similar effects were seen with the JNK inhibitor, SP600125 (Supplementary Fig. S3E and S3F). These data suggest that blocking DUSP phosphatase activity using BCI increases p-JNK–causing effects on MPNST cells. Given that DUSP6 knockdown more effectively hyperactivates MAPKs in MPNST ST8814 cells (Fig. 3B), our data suggest that BCI largely inhibits DUSP6 in NF1-deleted MPNSTs.

In MPNST cells, DUSP inhibition modestly increased phosphorylation of the MEK substrate ERK (Fig. 4B). We tested whether DUSP inhibition alters the effects of MEK inhibition in MPNST cell lines (Supplementary Fig. S4A and S4B). At 0.5 to 2 μmol/L, BCI with 1.25 or 2.5 μmol/L PD0325901, MEK inhibitory concentrations (5), there were no additive effects in ST8814 (CI, 0.9; Supplementary Fig. S4A and S4B) or STS26T (data not shown). In S462.TY, there was evidence of modest synergy (CI, 0.29) over a narrow concentration range (Supplementary Fig. S4A and S4C). The effect of PD0325901 in combination with genetic knockdown of Dusp1/6 had no additive or antagonistic effect on cell proliferation (Supplementary Fig. S4D). Thus, DUSP effects are mediated by JNK and combinations with MEKi may be modestly effective in some MPNST cell lines.

Because NF1 deletion in iHSC increased DUSP1 and DUSP6 protein levels (Fig. 1C), we used these cell lines to determine whether BCI would inhibit proliferation in a NF1-dependent manner in Schwann cells. There was a therapeutic window in which NF1-deleted iHSCs were more susceptible compared with WT cells to the antiproliferative effects of BCI (0.25–1 μmol/L; Fig. 5A). The largest difference was at 1 μmol/L (Fig. 5A). DUSP inhibition had lesser effects on iHSC NF1 WT cells, in which the baseline activity of MAPK (p-ERK) was barely detectable (Fig. 5B). Consistent with direct inhibition of DUSPs with shRNA and with BCI in MPNSTs, 1 μmol/L BCI enhanced p-ERK, p-JNK, and p-c-jun signaling in NF1-deleted iHSC after 1 hour, (Fig. 5B; Supplementary Fig. S5A). Constitutive MAPK activity in the iHSC NF1^−/− cells was strongly increased by treatment with BCI and correlated with increased levels of TP53 and p-ATM and reduced pRB and cyclin D1 (Fig. 5C). Evidence of cell death was variable, with evidence of cleaved PARP but not CC3, suggesting reduced proliferation as the predominate effect in these cells (Fig. 5C). As in MPNST cells, targeting DUSP and JNK in combination diminished the effectiveness of BCI in NF1^−/− iHSC (CI > 1.3; Fig. 5D; Supplementary Fig. S5B). At these concentrations of JNK-IN-8, p-c-jun was reduced (Supplementary Fig. S5C). Thus, in NF1^−/− Schwann cells and in MPNST cells with NF1 deletion, DUSPs contributes to cell proliferation and survival. These data suggest that DUSPs act as oncogenes in this context.

We tested whether BCI might be a useful therapeutic in vivo using an established cell line–based xenograft model, S462.TY MPNST cells (TY xenograft) engrafted subcutaneously into athymic nude mice. Tumor-bearing mice were injected with BCI 10 mg/kg, i.p., once daily, 5×/week for 27 days. No differences were detectable in tumor volume (Fig. 6A). In spite of the similar tumor volumes, excision of BCI-treated tumors revealed edema associated with encapsulated necrotic tissue that was absent in vehicle-treated mice; areas of necrosis ranged from 8% to 23% of section area (Fig. 6B and C).

We next used a novel PDX MPNST model which was developed by directly implanting patient tumor tissue into mice. The patient tumor had mutations in NF1 and in other genes common in human MPNSTs (CDKN2A/B loss and SUZ12 mutation), Rictor was also amplified. The resulting xenograft MPNSTs maintained characteristics of the original tumor histologically (Supplementary Fig. S6A) and genetically with NF1 and SUZ12 mutations. Grafted tumors remained positive for human nucleoli antigen (Supplementary Fig. S6B). BCI treatment at both doses significantly inhibited the growth of the PDX MPNSTs (Fig. 6D). In tissue sections, both the TY xenograft and the PDX MPNST model showed areas of cell apoptosis (CC3⁺ cells) and necrotic cell death on histology (TY, 8%–23%; PDX, 2%–17% vs. VEH 1%–2%; Fig. 6E and F). Both models also had areas of fibrosis/collagen deposits as indicated by trichrome staining (Supplementary Fig. S6C).

In the S462.TY xenograft, samples were collected on treatment day 27, 2 hours following the last BCI treatment, and analyzed for MAPK signaling and proliferation/cell death markers. BCI-treated tumors showed 60% to 90% increases in p-ERK, p-JNK, and p-c-jun, 60% lower levels of cyclin D1, and variable levels of TP53 versus controls (Fig. 6G; Supplementary Fig. S6D and S6E). In the PDX model, after 21 days of treatment, there was a trend toward MAPK activation (Supplementary Fig. S6F). Given that signaling data may be confounded by areas of necrosis and collagen/fibrosis, in the PDX model, signaling was analyzed after 12 days of BCI. These BCI-treated tumors had a significant 2.5 to 3.5× increased p-ERK, p-JNK, and p-c-jun, 40% lower levels of cyclin D1, and a 46% increase in TP53 (Fig. 6H; Supplementary Fig. S6G and S6H). Upon discontinuing treatment in the PDX model, tumors showed regrowth and areas of proliferation among areas of necrosis and collagen deposits (Supplementary Fig. S7A). This suggests that BCI affects the MAPK pathway in vivo to transiently inhibit cell proliferation and induce cell death in MPNSTs.

Interestingly, in a separate study using the PDX model, we found that BCI (10 mg/kg, i.p., 5×/week) was as effective as the MEKi PD0325901 (1.5 mg/kg, 5×/week) in inhibiting tumor growth (Supplementary Fig. S7B). We also tested PD0325901 (1.5 mg/kg, 5×/week for 27 days), versus a combination of BCI and PD0325901 (10 mg/kg, 5×/week for 27 days and 1.5 mg/kg, 5×/week for 27 days, respectively) in the TY xenograft model. As described previously, tumor volumes of MEK inhibitor-treated tumors were significantly lower compared with vehicle (Supplementary Fig. S7C). However, there was no significant tumor volume difference between the combination of BCI and PD0325901 versus PD0325901 only (Supplementary Fig. S7C). From histologic analysis, BCI-treated tumors showed more necrotic cell death compared with MEK alone or treatment with the MEKi and DUSPi combination (Supplementary Fig. S7D). Areas of necrosis were also observed in the PDX xenograft model after BCI but not PD0325901 treatment (Supplementary Fig. S7D). Collectively, our data using 2 different in vivo xenograft MPNST models suggests that targeting DUSP1/6 with a single agent, BCI, suppresses MPNST growth and induces cell death.

Novel targets and treatment options for NF1-mediated MPNSTs are urgently needed. DUSPs regulate MAPK signaling by affecting diverse cellular processes including cell proliferation, cell survival, and cell differentiation (37). We posited that understanding how DUSP family proteins affect NF1-mutant MPNSTs would suggest new approaches for MPNST therapy. Our analysis of DUSP1 and DUSP6 shows that both DUSP proteins are overexpressed in MPNSTs compared with normal tissues under basal conditions, in cells and in growing tumors. DUSP1 and DUSP6 may have partially redundant functions in MPNSTs, and it is possible that other DUSPs, including DUSP22, also play roles. DUSP1/6 are upregulated in NF1-mutant neurofibroma, iHSC, and in MPNST cells. In these cell types, DUSPs promote cell proliferation in vitro, as reducing levels of either or both, genetically or with the small-molecule inhibitor BCI, decreases cell proliferation. Each DUSP1 and DUSP6 also protect NF1-mutant MPNST cells from cell death, as silencing DUSP1 or DUSP6 in MPNST cells increased cell death in vitro. In 2 in vivo model systems, BCI, which specifically targets both DUSPs (12, 21) caused necrotic and apoptotic cell death.

In most normal cells, DUSPs are transcriptionally upregulated in response to MAPK signaling, and function as negative feedback regulators to suppress RAS/MAPK signaling. DUSPs are often upregulated in cancers harboring activating mutations in RAS or BRAF (6). The mechanisms underlying DUSP upregulation vary. Patients with pre-B ALL showed reduced CpG methylation of the DUSP6 promoter as compared with normal pre-B cells (13). Gene amplification or increased transcription due to deletion of NF1 and increased MAPK signaling may also play roles. In agreement with data from Courtois-Cox and colleagues (4), we find transcriptional upregulation of DUSPs in NF1 mutant cells. In addition, in 30% of MPNSTs, the DUSP1 or DUSP6 locus, or both, are amplified.

Genetic knockdown of DUSP1 decreased tumorigenesis in orthotopic pancreatic tumors and in non–small cell lung cancer (38, 39). DUSP6 prosurvival effects were also described in chemoresistance screens, in glioblastoma cells, and gastric cancer (14, 40, 41). The utility of the DUSP1/6 inhibitor, BCI, or its analogs, as therapeutics remains to be tested. BCI is a small molecule identified in a zebrafish chemical screen that prevents catalytic phosphatase activity upon substrate binding, contributing to its specificity (21). An allosteric binding site for BCI exists within the phosphatase domains of both DUSP6 and DUSP1, but not in the related DUSP5; activity toward other DUSPs has not been excluded to date (12, 21). BCI demonstrated single-agent efficacy at doses 5 to 40 mg/kg, i.p daily (or twice daily) for up to 1 month in mouse models of gastric cancer and leukemia. These regimens effectively inhibited DUSP1/6 without noted toxicity, as did our doses within this range in MPNST xenografts (12–14). Inhibition of the MEK/ERK pathway combined with DUSP inhibition in vivo did not alter tumor volume or cell death when compared with single agents. Further studies may identify improved dosing regimens and schedules that maximize efficacy.

Previous work showed that inhibiting DUSPs can result in increased phosphorylation of ERK, JNK, and/or p38 MAPKs (13). A predominant role for p-p38 was demonstrated in breast cancer and in therapy-resistant chronic myeloid leukemia (11, 12). In gastric cancer models, DUSP inhibition increased p-ERK, inducing cell death (14). In non–small cell lung cancer, phosphorylated JNK and p38 affected tumor progression upon DUSP1 silencing (39). Indeed, recent DUSP6 knockout studies failed to show expected effects on p-ERK; instead, p-JNK increased (42). These studies highlight the tumor-specific MAPK substrate effects of DUSPs. In MPNST cells, DUSP knockdown or DUSP inhibition with BCI caused increased p-ERK, p-JNK, and phosphorylation of the JNK substrate c-Jun. The finding that blocking JNK with either of 2 unrelated antagonists rescues the effects of BCI suggests that in the NF1-deficient immortalized Schwann cells and MPNST cells we evaluated, DUSP1 and DUSP6 act largely on the JNK arm of the MAPK pathway, supporting the idea that JNK activation plays a critical role in BCI's antitumor effect in MPNSTs.

The duration of MAPK activation is controlled by cell expression of negative regulators of MAPK signaling, including GAPs, such as NF1, and phosphatases such as DUSPs (4). MAPK is known to inhibit growth and promote cell death through altering phosphorylation of RB and ATM, or through TP53 (43). Given that TP53, RB, and ATM are central regulators of the cell cycle, the genetic status of these tumor suppressors may influence sensitivity to DUSP inhibitors. The incidence of mutation of TP53 is 40% in MPNSTs (44). In MPNSTs with WT TP53, DUSP inhibition with acute reactivation of JNK signaling appears to stabilize TP53, enabling the apoptotic pathway as occurs in other settings (34, 35). In ovarian cancer cells, JNK promotes PARP phosphorylation leading to PARP degradation (45). Total and cleaved PARP play important roles in modulation of apoptosis and necrosis. Inhibition of DUSPs and PARP may induce DNA damage and prevent repair, potentially sensitizing radio-resistant MPNST cells to radiation therapy and cell death.

The S462.TY cell line model has a high tumorigenic potential in mice and so is widely used to evaluate therapies for MPNSTs. However, a single model cannot accurately represent the complexity and diversity of MPNST tumors. Models in which patient-derived tissues, including MPNSTs, are directly grafted into immunocompromised mice are increasingly used to test therapeutics (46). We describe a PDX model that contains common cancer-associated mutations found in patient MPNSTs: NF1, CDKN2A/B, SUZ12 (47–49), and retains patient MPNST histologic features. BCI treatment in this model slowed tumor growth and also promoted cell death. With BCI treatment, the differences in tumor volume between the models may have been due to edema, with a delay in clearance of the dying cells and collagen deposits that contribute to tumor volume. Such effects can confound analysis of drug effects on tumor volume, as shown after other drug treatments in sarcoma and xenografted tumors (50). Importantly, in both models, single-agent BCI treatment was sufficient to cause cell death in vivo. For comparison, MEK inhibitor treatment transiently lowers tumor volumes, with no evidence of increased cell death (5).

In NF1-mediated transformation to MPNSTs, tumor cells appear to maintain MAPK activity within a narrow range to enable survival. DUSP inhibition, with increased MAPK signaling through JNK (and possibly other substrates), slows proliferation and can promote cell death. Our data support the development of agents targeting both DUSP1 and DUSP6 for MPNSTs.

D.A. Largaespada is an employee of Surrogen, Inc. and B-MoGen Biotechnologies, reports receiving commercial research grants from Genentech, and holds ownership interest (including patents) in NeoClone Biotechnology, ImmuSoft (previously Discovery Genomics, Inc), B-MoGen Biotechnologies, and Recombinetics, Inc. No potential conflicts of interest were disclosed by the other authors.

Conception and design: A. Ramkissoon, K.B. Williams, D.A. Largaespada, N. Ratner

Development of methodology: A. Ramkissoon, D. Milewski, K.B. Williams, R.L. Williams, K. Choi, N. Ratner

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A. Ramkissoon, K.E. Chaney, D. Milewski, A.A. Miller, T.V. Kalin, J.G. Pressey, S. Szabo, M. Azam, N. Ratner

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A. Ramkissoon, K. Choi, A.A. Miller, S. Szabo, M. Azam, N. Ratner

Writing, review, and/or revision of the manuscript: A. Ramkissoon, D. Milewski, K. Choi, J.G. Pressey, N. Ratner

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): R.L. Williams

Study supervision: A. Ramkissoon, M. Azam, N. Ratner

The authors thank Dr. Meenu Kesarwani for providing reagents and critical review of the manuscript, Dr. Tilat Rizvi for advice on IHC, Jonathan Fletcher for tumor dissection assistance, and Kierra Ware for immunoblot assistance. This work was supported by R01 NS086219 (to D.A. Largaespada and N. Ratner), a grant from the Children's Tumor Foundation-Synodos (to D.A. Largaespada), and CancerFree KIDS grants (to A. Ramkissoon). A.A. Miller was supported by a summer fellowship (NIH T35DK060444).

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.