mTOR kinase inhibition disrupts neuregulin 1-ERBB3 autocrine signaling and sensitizes NF2-deficient meningioma cellular models to IGF1R inhibition

Meningiomas (MNs), arising from the arachnoid/meningeal layer, are nonresponsive to chemotherapies, with ∼50% showing loss of the Neurofibromatosis 2 (NF2) tumor suppressor gene. Previously, we established NF2 loss activates mechanistic target of rapamycin complex 1 (mTORC1) and mechanistic target of rapamycin complex 2 (mTORC2) signaling, leading to clinical trials for NF2 and MN. Recently our omics studies identified activated ephrin (EPH) receptor and Src family kinases upon NF2 loss. Here, we report increased expression of several ligands in NF2-null human arachnoidal cells (ACs) and the MN cell line Ben-Men-1, particularly neuregulin-1/heregulin (NRG1), and confirm increased NRG1 secretion and activation of V-ERB-B avian erythroblastic leukemia viral oncogene homolog 3 (ERBB3) receptor kinase. Conditioned-medium from NF2-null ACs or exogenous NRG1 stimulated ERBB3, EPHA2, and mTORC1/2 signaling, suggesting pathway crosstalk. NF2-null cells treated with an ERBB3-neutralizing antibody partially downregulated mTOR pathway activation but showed no effect on viability. mTORC1/2 inhibitor treatment decreased NRG1 expression and downregulated ERBB3 while re-activating pAkt T308, suggesting a mechanism independent of NRG1–ERBB3 but likely involving activation of another upstream receptor kinase. Transcriptomics after mTORC1/2 inhibition confirmed decreased ERBB3/ERBB4 while revealing increased expression of insulin-like growth factor receptor 1 (IGF1R). Drug treatment co-targeting mTORC1/2 and IGF1R/insulin receptor attenuated pAkt T308 and showed synergistic effects on viability. Our findings indicate potential autocrine signaling where NF2 loss leads to secretion/activation of NRG1-ERBB3 signaling. mTORC1/2 inhibition downregulates NRG1-ERBB3, while upregulating pAkt T308 through an adaptive response involving IGF1R/insulin receptor and co-targeting these pathways may prove effective for treatment of NF2-deficient MN.

Meningiomas (MNs), arising from the arachnoid/meningeal layer, are nonresponsive to chemotherapies, with 50% showing loss of the Neurofibromatosis 2 (NF2) tumor suppressor gene. Previously, we established NF2 loss activates mechanistic target of rapamycin complex 1 (mTORC1) and mechanistic target of rapamycin complex 2 (mTORC2) signaling, leading to clinical trials for NF2 and MN. Recently our omics studies identified activated ephrin (EPH) receptor and Src family kinases upon NF2 loss. Here, we report increased expression of several ligands in NF2-null human arachnoidal cells (ACs) and the MN cell line Ben-Men-1, particularly neuregulin-1/heregulin (NRG1), and confirm increased NRG1 secretion and activation of V-ERB-B avian erythroblastic leukemia viral oncogene homolog 3 (ERBB3) receptor kinase. Conditioned-medium from NF2-null ACs or exogenous NRG1 stimulated ERBB3, EPHA2, and mTORC1/2 signaling, suggesting pathway crosstalk. NF2-null cells treated with an ERBB3-neutralizing antibody partially downregulated mTOR pathway activation but showed no effect on viability. mTORC1/2 inhibitor treatment decreased NRG1 expression and downregulated ERBB3 while re-activating pAkt T308, suggesting a mechanism independent of NRG1-ERBB3 but likely involving activation of another upstream receptor kinase. Transcriptomics after mTORC1/2 inhibition confirmed decreased ERBB3/ERBB4 while revealing increased expression of insulin-like growth factor receptor 1 (IGF1R). Drug treatment co-targeting mTORC1/2 and IGF1R/insulin receptor attenuated pAkt T308 and showed synergistic effects on viability. Our findings indicate potential autocrine signaling where NF2 loss leads to secretion/activation of NRG1-ERBB3 signaling. mTORC1/2 inhibition downregulates NRG1-ERBB3, while upregulating pAkt T308 through an adaptive response involving IGF1R/insulin receptor and co-targeting these pathways may prove effective for treatment of NF2-deficient MN.
The disease neurofibromatosis 2 (NF2) is characterized by multiple nervous system tumors, including bilateral vestibular schwannomas and intracranial meningiomas (MNs), and is caused by mutations in the NF2 gene (1)(2)(3). Separately, sporadic MNs are the most common primary intracranial tumors in adults, with 50% having biallelic, somatic inactivation of NF2. Benign MNs (World Health Organization grade I) are most common; however, they often cause severe neurologic morbidity and mortality because of compression of adjacent brain or spinal cord. Atypical (World Health Organization grade II) or anaplastic (World Health Organization grade III) MNs display more aggressive clinical behavior with rapid growth and increased recurrence rates (4,5). With nonresponse to chemotherapies, the current standard of care for MN is maximal surgical resection, whereas radiation is reserved for recurrent or aggressive tumors. MNs that progress despite surgery and radiation show high morbidity and mortality (1,6). Therefore, effective noninvasive therapies are much needed for both NF2associated and sporadic MNs.
To identify other relevant drug targets for NF2, we recently performed large-scale kinome and transcriptome analyses of our AC and MN cell models, which showed increased activation and expression of several EPH receptor family tyrosine kinases, Src family kinase members and c-KIT, which are all targets of dasatinib (21). In follow-up studies, we confirmed the expression/activation of EPH receptor family tyrosine kinases as well as downstream Src family kinases in NF2-null MN and schwannoma models and reported that combining a dual mTORC1/mTORC2 inhibitor, AZD2014 or INK128 (also called TAK-228), with dasatinib is synergistic in these models (22,23). Here, extending analyses of transcriptomic data from our previous studies, we have identified several ligands, including NRG1, which encodes neuregulin-1. Although mTOR activation upon NF2 loss is well established, the upstream regulation of this activation and downstream consequences of mTOR activation remain unclear. Therefore, we set out to examine the role of neuregulin-1/heregulin (NRG1) in the context of NF2 loss. Our findings reveal that NF2 loss leads to NRG1 secretion that in turn activates the V-ERB-B avian erythroblastic leukemia viral oncogene homolog 3 (ERBB3) receptor kinase in an autocrine fashion. In addition, secreted and exogenous NRG1 activates downstream mTOR signaling and EPHA2. Interestingly, mTORC1/mTORC2 inhibition in our MN cellular models disrupts the NRG1-ERBB3 signaling while increasing pAkt T308, but not pAkt S473, because of an adaptive response that likely involves upregulation of insulinlike growth factor receptor 1 (IGF1R) expression and activation. Our results further show that combined inhibition of mTOR and IGF1R signaling is synergistic in MN cells and provide a compelling basis for in vivo testing in animal models of NF2.

NF2-deficient AC and MN cells secrete NRG1 ligand that activates the ERBB3 receptor and downstream signaling
We next examined whether NF2-deficient AC cells and Ben-Men-1 cells (Fig. 1B) secrete NRG1 ligand. Media from NF2(+) and NF2(−) AC-CRISPR cells were collected at 48 h following a change to serum-free medium and then concentrated by centrifugal filtration. Immunoblotting revealed NRG1 secretion in 48 h-concentrated medium (48 h-CM) from NF2(−) AC cells and Ben-Men-1, whereas 48 h-CM from NF2(+) ACs or 0 h-CM from NF2(−) ACs showed no signal for secreted NRG1 (Fig. 1C). As a confirmation, we also performed qPCR analysis in Ben-Men-1 cells where NF2 had been reintroduced by lentiviral transduction and observed that reexpression of NF2 (NF2-CSCW2) demonstrated a significant decrease in NRG1 expression compared with empty vector (V-CSCW2) (Fig. S1).
Next, in conditioned medium experiments, we incubated NF2(+) ACs with medium from NF2(+) or NF2(−) ACs. In cells treated with NF2(−) 48 h-CM, we observed increased phosphorylation of ERBB3 Y1197 (pERBB3) receptor compared with 48 h-CM from NF2(+) ACs. We also found increased phosphorylation of the receptor EPHA2 S897 (pEPHA2) as well as activated mTOR signaling, as shown by increase in mTORC1 pathway readout pS6 S240/44 and mTORC2-SGK1 pathway readout pNDRG1 T346 (Fig. 1D). We confirmed these results using exogenous NRG1 to stimulate NF2-null ACs and observed not only activation of pERBB3 but also pEPHA2, pS6K T389 (mTORC1 readout), and pAkt S473 (another mTORC2 readout) (Fig. 1E). These data suggest that loss of NF2 in AC or MN cells leads to secretion of NRG1 ligand and activation of ERBB3 receptor in an autocrine fashion. In addition, upregulation of NRG1-ERBB3 signaling can cross-talk to EPHA2 receptor as well as activate downstream mTORC1/2 signaling.
Ben-Men-1 cells with MM-121 inhibited the basally activated and NRG1-stimulated pERBB3 Y1197, both at 2 h and 24 h time points (Fig. 2B). Further, similar to lapatinib, NRG1stimulated cells treated with MM-121 also showed attenuated pEPHA2 S897, pAkt S473 (mTORC2 readout), and pS6K T389 (mTORC1 readout) at 2 h and 24 h treatment, as well as decreased pS6 S240/44 (mTORC1) in NF2-null ACs at 24 h and Ben-Men-1 at 2 h and 24 h (Fig. 2C). Moreover, when examining basally activated downstream signaling, we found that unlike lapatinib, treatment with MM-121 was capable of inhibiting pEPHA2 and mTORC1 signaling (pS6K and pS6) at 24 h in NF2-null ACs. It should be noted that, consistent with our previous reports, activated pAkt S473 was below detectable level in serum-deprived NF2-null ACs. In Ben-Men-1 cells, MM-121 treatment also downregulated basal pAkt (mTORC2), pS6K, and pS6 (mTORC1) at 2 h and 24 h, but had no effect on pEPHA2 (Fig. 2D). Here, our results confirm the signaling cross-talk seen in our conditioned media and NRG1-stimulation experiments and suggest that activated NRG1-ERBB3 signaling may be partially responsible for activated EPHA2 and mTORC1/2 pathways in NF2-null AC and Ben-Men-1 cells. We also show that direct inhibition of NRG1 ligand binding to ERBB3 is more effective than lapatinib to downregulate the basally activated signaling pathways.
ERBB3 receptor by preventing NRG1 ligand binding can lead to downregulation of ERBB3 as well as EPHA2 and mTORC1/ 2 signaling. However, our viability assays indicate that specific inhibition of NRG1-ERBB3 signaling in NF2-null AC and Ben-Men-1 cells is not sufficient to affect cellular growth.
increased pAkt T308, similar to INK128 alone (Fig. 3C). Transcriptome analyses of our NF2-null ACs after treatment with INK128 showed a decrease in NRG1, ERBB3, and ERBB4, suggesting that NRG1-ERBB pathway is downregulated at the transcription level by INK128 treatment (Table 2).
Interestingly, RNA-seq analyses also showed a significant increase in IRS2 and IGF1R in NF2-null ACs treated with INK128 ( Table 2). Treatment of two independent NF2-null AC-CRISPR clones and Ben-Men-1 cells with INK128 at 2 h and 24 h revealed increased expression of the IGF1R receptor by immunoblotting, confirming our RNA-seq results (Fig. 4A). Further, using an antibody that recognizes both pIGF1R Y1135/36 and pIR Y1150/51, we also found increased levels of pIGF1R/pIR as well as increased pAkt T308, but no activation of pAkt S473 after 24 h in NF2-null ACs and Ben-Men-1 cells (Fig. 4B). Treatment with INK128 at both 24 h and 48 h showed similar inhibition of pAkt S473 in our cells (data not shown). RTKs, including IGF1R/IR, are clearly established to activate downstream phosphoinositide 3-kinase (PI3K)/ PDK1-dependent recruitment and phosphorylation of Akt T308 at the plasma membrane (32). We examined whether INK128 treatment might induce a similar signaling mechanism in NF2-null cells. Co-treatment of our NF2-null AC and Ben-Men-1 cells with INK128 combined with IGF1R inhibitor BMS-754807 downregulated both pIGF1R and pAkt T308 (Fig. 4, B-C), whereas co-treatment of INK128 with lapatinib or erlotinib had no effect on pAkt T308 levels (Fig. 4C). In addition, inhibition of both mTORC1/2 and IGF1R also downregulated pFOXO1/3a, a functional downstream phospho-target of Akt, compared with INK128 alone. (Fig. 4D). Together, these data suggest that mTOR kinase inhibition disrupts and downregulates NRG1-ERBB3 signaling, while inducing an adaptive response of PDK1-dependent pAkt T308 activation that involves upregulation of IGF1R in NF2deficient AC and MN cells.

Co-treatment with INK128 and BMS-754807 reveals synergistic effects
We examined the effects of mTOR kinase inhibition and IGF1R inhibition, singly and combined, on cell viability in NF2-null cells including NF2-null ACs, two immortalized MN lines Ben-Men-1 and MN1-LF, and two primary MN lines MN646C and MN658 (Fig. S3). For single drug testing, in agreement with our previous reports using mTORC1/2 inhibitors (15,22), cell viability assays in NF2-null cell lines demonstrated a reduction upon INK128 treatment. Moreover, INK128 treatment showed superior efficacy in all cell lines tested compared with BMS-754807 with IC50s in the nanomolar range for INK128 treatment compared with micromolar IC50 range for BMS-754807 (Fig. 5A, Table S1). Therefore, these data suggest a relative insensitivity to IGF1R inhibition alone in these cell lines.
Based on our RNA-seq and immunoblotting data showing upregulated IGF1R expression and activation in response to dual mTORC1/2 inhibition, we next performed co-treatment assays. Dose-response testing in Ben-Men-1 cells for BMS-754807 in the presence of 400 nM INK128 (BMS + 400 nM INK128) revealed a left-shifted DRC, with a decreased IC50 of 3.2 μM (MR, 77.7%) compared with 4.4 μM (MR, 75.7%) for BMS-754807 alone (Fig. 5B). We further extended the cotreatment dose-response testing in MN1-LF cells. DRCs for BMS-754807 co-treated with increasing doses of INK128 again revealed left-shifted DRCs suggesting increased sensitivity to IGF1R inhibition in a dose-dependent manner compared with BMS-754807 alone (Fig. 5C, Table S1).
Expanding on these results, employing 10 × 10 dose matrix testing, we next asked whether co-treatment with INK128 and BMS-754807 has a synergistic effect. For drug synergy analysis, we utilized the web-based SynergyFinder application (33), which generates a synergy score by comparing drug treatment data to four different synergy reference models: (i) Highest Single Agent, (ii) Loewe Additivity, (iii) Bliss Independence, and (iv) Zero Interaction Potency. Co-treatment of Ben-Men-1 and MN1-LF cells as well as two human primary NF2-null MN lines, MN646C and MN658, with INK128 and BMS-754807 showed true synergistic effects, defined as a score >5.0 for all four reference models (34) (Fig. 6, Table 3). Taken together, cell viability assays strongly support our immunoblotting results and suggest a potential mechanism for drug synergy where mTOR kinase inhibition leads to IGF1R-PDK1-pAkt T308 pathway activation that in turn sensitizes NF2-null cells to IGF1R inhibition.

Discussion
The ERBB family of RTKs (also known as HER family) include four distinct receptors, ERBB1 (EGFR/HER1), ERBB2 (neu/HER2), ERBB3 (HER3), and ERBB4 (HER4). They are often overexpressed, mutated, or amplified in human cancers, thus making them important therapeutic targets (35). Many extracellular ligands can bind ERBB receptors, initiating a Adaptive signaling upon mTOR inhibition in NF2-deficiency cascade of biochemical and signaling events. Specifically, NRG1 and NRG2 can bind ERBB3 and ERBB4 receptors. Elevated NRG1 expression and activated ERBB3 are seen in many types of human cancers supporting the rationale to target the NRG1-ERBB3 axis (36)(37)(38)(39). More importantly, the NRG1 ligand and its target the ERBB family receptors have been implicated in schwannoma tumorigenesis (40)(41)(42) leading to preclinical testing of EGFR/ERBB2 kinase inhibitor lapatinib in a vestibular schwannoma model (27). A subsequent Phase II clinical trial of lapatinib for patients with NF2 and progressive vestibular schwannoma showed tumor regression and improvement of hearing in 4 of 17 patients treated (43). Based on these reports, we focused our study on elevated NRG1 expression, detected in our MN models by transcriptome analyses, to define the role of the NRG1-ERBB3 axis in MN with NF2 loss. In addition to confirming the secretion of NRG1 and activation of downstream signaling by NRG1, our results here show that a multi-ERBB inhibitor lapatinib, but not EGFRspecific inhibitor erlotinib inhibits the NRG1-stimulated activation of ERBB3 (pERBB3), EPHA2 (pEPHA2), and mTOR (pAkt, pS6K) (Fig. 2), suggesting that NRG1-induced activation is not EGFR-dependent. Lapatinib however was unable to block basal activation of EPHA2 and mTOR signaling in NF2null cells.
Interestingly, using a systems biology approach, a human monoclonal antibody seribantumab (MM-121), that inhibits NRG1-stimulated ERBB3 signaling with low nanomolar IC 50 values compared with lapatinib and other ERBB inhibitors, was identified to be more effective in blocking ligand-induced activation of ERBB3 signaling network, which led to a Phase II clinical trial of human cancers (29,30,44). Therefore, we set out to test the ability of MM-121 to block NRG1-induced signaling and effectively inhibit cell proliferation in NF2-null MN cells. Our results show that MM-121 not only blocks NRG1-stimulated but also basal ERBB3 activation and . ImageJ/Fiji quantitation shows phosphorylated relative to total protein, with control lanes normalized to 1. PDK1, phosphoinositide-dependent kinase-1; mTOR, mechanistic/mammalian target of rapamycin; mTORC1, mechanistic target of rapamycin complex 1.
Adaptive signaling upon mTOR inhibition in NF2-deficiency downstream signaling upon NF2 loss; however, it is not effective in inhibiting proliferation of MN cells, suggesting that specific NRG1-ERBB3 signaling may not be a major player for MN growth. Our findings in fact show that NRG1 expression is regulated by mTORC1 as seen by a significant decrease in NRG1 expression after either mTORC1 inhibition by rapamycin or mTOR kinase inhibition (Fig. 3A), supporting the possibility that NRG1 expression and autocrine signaling could be partly because of mTORC1/mTORC2 activation upon NF2 loss (Fig. 7A).
A previous study demonstrated that mTOR kinase inhibition of breast cancer cells for 24 h led to inhibition of pAkt S473 but activation of pAkt T308, along with downstream effectors of Akt, because of activation of several RTKs, including ERBB family members as well as IGF1R and IR. Treating breast cancer cell lines and in vivo xenograft models with a combination of a mTOR kinase inhibitor and a multi-ERBB inhibitor lapatinib abolished Akt signaling and resulted in cell death and tumor regression (31). Our results in MN cells similarly show that mTOR kinase inhibition persistently inhibits pAkt S473 while activating pAkt T308.
Conversely, unlike breast cancer cells, mTOR inhibition downregulates NRG1 expression as well as pERBB3 activation, and treatment with MM-121 was unable to block activation of pAkt T308 induced by mTOR kinase inhibition (Fig. 3C), which was consistent with our RNA-seq results where we noted a decrease in ERBB3 and ERBB4 expression. Interestingly, upon mTOR kinase inhibition, we observed a significant increase in the expression of receptor genes IGF1R, INSR, and IGF2R, as well as IRS2, encoding insulin receptor substrate 2. It is interesting to note that insulin receptor substrate (IRS) proteins are recruited to active IGF/insulin receptors where they are phosphorylated, triggering PI3K-PDK1-Akt signaling (45). We confirmed that increased expression and activation of IGF1R/IR in our NF2-deficient cells is responsible for phosphorylation and activation of Akt at T308, a direct target site of PDK1. A recent report on melanoma revealed that long-term mTOR kinase inhibition leads to re-activation of both pAkt S473 and pAkt T308 and that induction of IGF1R/IRdependent PI3K activation and Akt phosphorylation is mediated by an integrin/focal adhesion kinase/IGFR-dependent process (46). In our MN models, the activation of pAkt Adaptive signaling upon mTOR inhibition in NF2-deficiency T308 by mTOR kinase inhibition, similar to melanoma cells, is mediated by IGF1R/IR; however, unlike melanoma, pAkt S473 remains fully inhibited by mTOR kinase inhibition in MN cells. Taken together, our results support the existence of cell-/context-dependent mechanisms for adaptive signaling observed upon mTOR kinase inhibition.
Treatment of NF2-associated MN and schwannomas with rapamycin analogs was found to be cytostatic (18)(19)(20). Further, mTORC1 inhibition with rapamycin and its analogs is known to relieve the negative feedback inhibition on IRS-1 (47,48) and Grb10 (49,50), as well as other negative regulation of mTORC2 independent of IRS-1 and Grb10 (51,52), thus activating the PI3K-mTORC2-Akt prosurvival pathway. Therefore, mTOR kinase inhibitors were developed to overcome the limitations of rapamycin by effectively inhibiting both mTORC1 and mTORC2 (53). Our previous studies revealed mTORC2-SGK1 activation in addition to mTORC1 activation in NF2 tumors and demonstrated mTOR kinase inhibitors to be more effective than rapamycin (15,17), which led to ongoing clinical trials with mTOR kinase inhibitor AZD2014 for MN. Our continued studies here show that mTOR activation in NF2-deficient cells leads to increased expression and secretion of NRG1 ligand in an autocrine fashion (Fig. 7A) and that mTOR kinase inhibition disrupts the NRG1-ERBB3 signaling. Nevertheless, inhibiting mTOR kinase induces an adaptive response involving IGFR/IR expression and activation in NF2-null MN cells (Fig. 7B), leading to activation of prosurvival pathways, including Akt T308 and  Table S1. Synergy scoring scale is shown with red representing synergism and green representing antagonism (A-D). Bliss, Bliss independence; HSA, highest single agent; Loewe, Loewe additivity; ZIP, zero interaction potency.
Forkhead box protein O (FOXO) phosphorylation, which needs to be taken into account when considering combination treatment approaches. Importantly, our findings show that combining an IGFR inhibitor with an mTOR kinase inhibitor has synergistic effects, thus setting the stage for further in vivo studies and potential translation to the clinic for more effective treatment of NF2-associated tumors.

Cell lines and reagents
Human cell lines included an NF2-null immortalized MN line Ben-Men-1, NF2 AC-CRISPR, and two independent human primary MN for which cell line establishment and growth conditions have been previously described (15,54). All primary cultures were collected following Massachusetts General Hospital Human Subjects protocols for tumor acquisition after informed consent. In addition, we generated an immortalized human MN line, MN1-LF from an independent NF2-null human primary MN cell line derived from a surgical resection. Immortalization methods and confirmation was carried out on a fee-for-service basis by Alstem, Inc. Briefly, 1 × 10 5 cells were transduced with lentivirus encoding SV40 large T antigen and a puromycin resistance gene. Cells were infected at a multiplicity of infection of 2. Following selection with 1.5 μg/ml puromycin for 3 days, cells underwent an additional three passages. PCR amplification of SV40 was performed to confirm immortalization followed by expansion and freezing of cells.
Re-expression of NF2 in Ben-Men-1 cells Re-introduction of NF2 was performed using the full length cDNA coding sequence of NF2 (isoform 1, NCBI accession #NM_000268) cloned into the lentiviral vector CSCW2 followed by packaging as previously reported (16). For reexpression, Ben-Men-1 cells were transduced with either NF2-CSCW2 or V-CSCW2 at a multiplicity of infection of 50 along with 8 μg/ml polybrene by spin-infection as previously described (15). Cells were then expanded and harvested for RNA/cDNA synthesis for quantitative RT-PCR, and protein lysates were generated for immunoblotting.

Transcriptome/RNA-seq and analysis
Baseline transcriptome analyses between NF2(−) versus NF2(+) ACs and Ben-Men-1 were previously described (21). For transcriptome analyses, RNA was isolated from all MNrelevant lines using TRIzol reagent (15596026, ThermoFisher Scientific) according to the manufacturer's instructions. Pelleted cells were resuspended TRIzol reagent using microdounce homogenization then extracted with chloroform, followed by isopropanol precipitation of RNA from the aqueous phase and an 80% ethanol wash. RNA pellets were solubilized in 30 to 50 μl of RNase-free water (Ambion, AM9937). RNA quality was assessed using the Agilent Bioanalyzer Tapestation 2200 (Agilent Technologies).
Adaptive signaling upon mTOR inhibition in NF2-deficiency next day, cells were rinsed twice with PBS and once in serumfree DMEM (SF-DMEM), followed by addition of 10 ml SF-DMEM per plate. Cells were then incubated for 48 h. As a control, 10 ml of SF-DMEM was also added to cells and immediately collected (0 h). Next, 0 h-or 48 h-incubated medium was collected, briefly spun to remove debris, and then applied to an Amicon Ultra-15 3K filtration unit (EMD Millipore) followed by centrifugation at 3000g for 45 to 50 min. When total retentate volume was 700 μl, retentate was then transferred to an Amicon Ultra-0.5 ml 3K filtration unit (EMD Millipore), in two additions, and spun at 14,000 rpm until retentate reached a final volume of 200 μl. The resulting 0 hor 48 h-concentrated media (0 h-CM or 48 h-CM) was collected, protease inhibitor cocktail (P8340, Sigma) was added (1 × final concentration), followed by immunoblotting of 40 μl for each sample.
For conditioned medium experiments, isogenic NF2expressing and NF2-null AC-CRISPR cell lines were seeded in full growth medium. The next day, cells were rinsed twice with PBS followed by addition of SF-DMEM to each cell line, which served as the source of conditioned medium. Cells were then incubated for 48 h, and then conditioned medium was harvested. For treatment, NF2-expressing AC-CRISPR cells were seeded in full growth medium and incubated until 70% confluency. The growth medium was then removed, and cells were rinsed twice in PBS followed by 6 h treatment using respective conditioned medium harvested above. Treated cells were then lysed, followed by immunoblotting.

Cell viability and combination drug screening
Drug screening assays for all cell lines were carried out in a 384-well format using the CellTiter-Glo cell viability kit (Promega) according to the manufacturer's instructions. Relative luminescence units were measured using the EnVision 2103 Multilabel Reader (Perkin Elmer). For single drug DRCs, cells were seeded 24 h before drug treatment at 400 cells/well. Treatments using MM-121, INK128, and BMS-754807 were performed in full growth conditions for 72 h, and treatment dosage is outlined in the figure legends. Percent viability was calculated (relative luminescence units of drug treated versus vehicle treated), and DRCs and drug concentrations inhibiting cell growth by 50% (IC50) were calculated using GraphPad Prism 8 by nonlinear regression (curve fit) analysis with ± standard error of the mean determined (at least three replicates/cell line). For drug synergy testing, INK128 and BMS-754807 were arrayed in a standard 10 × 10 dose matrix (three replicates/cell line). Cells were seeded as above and treated for 72 h, and percent viability was calculated for each treatment dosage point. Synergy scores were generated using the web-based SynergyFinder application (https:// synergyfinder.fimm.fi/) (33).

Data availability
All data described in this study are contained within the manuscript with the exception of the entire dataset for transcriptome changes in INK128-treated cells (described in Table 2), which is available upon request.
Acknowledgments-We would like to thank Anna D. Ware and Breelyn Karno for technical help.
Author contributions-R. L. B. contributed to design of experiments, acquisition of data, analysis and interpretation of data, and drafting and revising the article; S. E. contributed to analysis and interpretation of data, and drafting and revising the article; L. W. contributed to acquisition of data; J. T. J., S. R. P. and J. F. G. contributed to interpretation of data, and drafting and revising the article; V. R. contributed to conception of study, design of experiments, interpretation of data, drafting and revising the article, and final approval of version to be published.
Funding and additional information-This study was supported by Children's Tumor Foundation-Synodos for NF2, NF Northeast, U.S. Army Medical Research Neurofibromatosis Program W81XWH1810547, National Institutes of Health R01 NS113854, and the S. Sydney DeYoung Foundation. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.
Conflict of interest-The authors declare that they have no conflicts of interest with the contents of this article.