Proteomic screening identifies a YAP-driven signaling network linked to tumor cell proliferation in human schwannomas

Abstract

Background

Inactivation of the NF2 gene predisposes to neurofibromatosis type II and the development of schwannomas. In vitro studies have shown that loss of NF2 leads to the induction of mitogenic signaling mediated by receptor tyrosine kinases (RTKs), MAP kinase, AKT, or Hippo pathways. The goal of our study was to evaluate the expression and activity of these signaling pathways in human schwannomas in order to identify new potential therapeutic targets.

Methods

Large sets of human schwannomas, totaling 68 tumors, were analyzed using complementary proteomic approaches. RTK arrays identified the most frequently activated RTKs. The correlation between the expression and activity of signaling pathways and proliferation of tumor cells using Ki67 marker was investigated by reverse-phase protein array (RRPA). Finally, immunohistochemistry was used to evaluate the expression pattern of signaling effectors in the tumors.

Results

We showed that Her2, Her3, PDGFRß, Axl, and Tie2 are frequently activated in the tumors. Furthermore, RRPA demonstrated that Ki67 levels are linked to YAP, p-Her3, and PDGFRß expression levels. In addition, Her2, Her3, and PDGFRß are transcriptional targets of Yes-associated protein (YAP) in schwannoma cells in culture. Finally, we observed that the expression of these signaling effectors is very variable between tumors.

Conclusions

Tumor cell proliferation in human schwannomas is linked to a signaling network controlled by the Hippo effector YAP. Her2, Her3, PDGFRß, Axl, and Tie2, as well as YAP, represent potentially valuable therapeutic targets. However, the variability of their expression between tumors may result in strong differences in the response to targeted therapy.

Biallelic inactivation of the NF2 tumor suppressor gene leads to the development of intracranial tumors such as schwannomas, meningiomas, and ependymomas. These lesions can be sporadic or develop in the context of an inherited familial disease called neurofibromatosis type 2. Patients who are born heterozygous for NF2 are predisposed to the development of multiple tumors upon subsequent loss of the second allele.¹

To this day, surgery and radiotherapy remain the most frequent options for treatment. The use of chemotherapy has long been hindered by the lack of clearly identified therapeutic targets. Since the discovery of the NF2 gene in 1993,² a major effort has been made to understand how merlin, the NF2 gene product, regulates cell proliferation and tumor growth. Several proteins playing key roles in these processes have emerged as candidate therapeutic targets. The inactivation of NF2 was shown to trigger the activation of Ras and Rac and the stimulation of downstream mitogenic p42/44 MAPK and PAK1/2 pathways.^3,4 It has also been demonstrated that loss of merlin expression induces the activity and surface expression of various receptor tyrosine kinases (RTKs). In the latter case, it is due to an increase of their transport to the plasma membrane,^5–7 leading to stimulation of promitogenic signaling pathways. Recently, a new function for merlin in the nucleus was discovered.⁸ In this compartment, the binding of merlin to D-Caf1 abolishes its ubiquitin ligase activity and leads to the inhibition of proliferation. Finally, the regulation of the Hippo signaling pathway by merlin has emerged in recent years as a major mechanism controlling cell proliferation and survival in various organisms and tissues.^9,10 Indeed, in several models, loss of merlin expression leads to nuclear accumulation of Yes-associated protein (YAP), which is the major effector of the Hippo pathway.^11,12 There, it stimulates transcription of a wide set of promitogenic and anti-apoptotic target genes.¹³ Hence, the growth advantage conferred by the loss of NF2 appears to be the consequence of an accumulation of distinct signaling dysfunctions.

The relevance of these mechanisms of growth control by merlin has also been tested on tumor cell cultures derived from human biopsies. Receptors such as IGF1R, Axl, or PDGFRß and signaling pathways such as MAPK, Rac/PAK, JNK, or β-catenin were shown to be activated in response to merlin loss of expression.^14–16 However, the tumor cell cultures always represent a simplified model of their tumor of origin. In this context, evaluation of the expression and activity of signaling pathways in surgical biopsies represents a necessary complement to the previous approaches. Immunohistological evaluation of the expression and activity of signaling proteins in tumor sections has provided important information on the biology of NF2-related tumors.^17–19 Nevertheless, new proteomic approaches such as reverse phase protein arrays (RPPA) allow more precise means for quantifying signaling protein expression and activity on large sets of biological samples.²⁰ Therefore, combining different types of proteomic analysis for the study of tumors constitutes a powerful approach for identifying the signaling events that promote their growth.

Using a set of more than 40 human schwannomas, we have analyzed the status of several major signaling pathways that were previously shown to be regulated by merlin. First, using receptor tyrosine kinase (RTK) arrays and Western blotting, we have identified which RTKs are the most frequently activated in schwannomas and could therefore represent a potential therapeutic target. Then, RPPA was used on our tumor set to measure the expression and activity of 35 proteins comprising the major mitogenic signaling pathways. Our results showed that expression levels of YAP and several of its target genes, notably several receptors identified through RTK arrays, are linked to proliferation. However, immunohistochemical analysis revealed a strong variability of protein expression from tumor to tumor, suggesting that the response to targeted chemotherapy would likely be highly variable. Altogether, our work identifies a signaling network associated with YAP protein and linked to schwannoma cell proliferation that could represent a set of valuable therapeutic targets.

Statement of translational relevance

Patients affected by the familial syndrome Neurofibromatosis type 2 develop multiple tumors when the NF2 suppressor gene is inactivated. Schwannomas are the most frequent ones and the number of chemotherapeutic options is yet very limited. Hence, the identification of potential therapeutic targets is a major objective of the research effort on this tumor type. In this study, we have combined several proteomic approaches to investigate the expression and activity of mitogenic signaling pathways regulated by NF2. Our work shows that proliferation of tumor cells is correlated to the activation of a signaling network linked to the Hippo effector Yap and several of its target genes. Hence, our study provides evidences that Yap as well as its target genes PDGFRß, Her3, Her2 and Axl represent potentially new therapeutic targets for the treatment of schwannomas.

Material and Methods

Tumors and Patients

Sixty-eight patients (33 women and 35 men) who underwent surgery for a schwannoma in Beaujon Hospital (Clichy, France) were retrospectively studied (Supplementary Table 2). Fourteen patients had a NF2 disease, while the other patients had developed sporadic tumors. For all patients, the diagnosis was established with routine formalin-fixed, paraffin-embedded material (hematoxylin and eosin sections). Frozen material was available for 33 tumors and stored at −80°C. The study was conducted after approval by the Comité Consultatif des Personnes Participant à une Recherche, and samples were obtained from participants who provided informed consent.

Immunoblotting

All extracts from HEI193 cells or human schwannomas were prepared with a RPPA extraction buffer (see Reverse Phase Protein Arrays section). Using a dounce homogenizer, tumor biopsies were lysed until complete solubilization. Following clarification of the extracts by centrifugation (20 000 g for 10 min at 4°C), protein concentration was measured by Bradford assay (Biorad). For Western blotting, proteins were separated by SDS-PAGE and transferred to nitrocellulose. Membranes were incubated with primary antibodies overnight at 4°C in phosphate-buffered saline, Tween 0.1%, and fetal bovine serum 10%. Antibodies directed against phospho-Akt₄₇₃, phospho-MAPK_202/204, phospho-GSK3ß₉, Axl, PDGFRß, EGFR, YAP, and Stat3 are described in the Immunohistochemistry section. We also used antibodies to mtor (2972; Cell Signaling Technology CST; Ozyme), phospho-mtor₂₄₈₁ (2971; CST), JNK (9252; CST), phospho-JNK_183/185 (9251; CST), AKT1 (9272; CST), MAPK (9102; CST), GSK3ß (9332; CST), p38 SAPK_180/182 (9212; CST), phospho-STAT3₇₀₅ (9131; CST), IGF1R (3027; CST), c-Met (3127; CST), Her2 (2165; CST), MST1 (3681; CST), phospho-MST1/2_183/180 (3681; CST), Lats1 (3477; CST), phospho-Lats1₉₀₉ (9157; CST), phospho-YAP₁₂₇ (4911; CST), phospho-Taz₈₉ (sc-17610; Santa Cruz/Tebu-Bio), Tie2 (C-20; Santa Cruz), Her3 (C-17; Santa Cruz), MST2 (pab0892-P; Covalab), Lats2 (pab0891-P; Covalab), Taz (pab0893-P; Covalab), survivin (614701; Biolegend), and Actin (A3853; Sigma Aldrich).

Immunofluorescence

HEI193 cells were infected with a lentivirus expressing GFP-YAP (Genecopoeia EX-Z0483-Lv122; Tebu-Bio) and selected with 5 µg/mL of puromycine. Cells were fixed on glass coverslips for 20 minutes in 4% paraformaldehyde. Coverslips were mounted on a glass slide with Citifluor mounting media containing 4’,6-diamidino-2-phenylindole (Citifluor Ldt). Pictures were acquired using a Leica DM 6000B epifluorescence microscope and a 63X oil immersion objective.

Luciferase Activity Assay

HEI GFP-YAP cells were seeded in a 25 mm dish and transfected 24 hours later under the following conditions: 1 µg of 8_GTIIC-Luc vector (kind gift from S. Piccolo laboratory) or p∂51-Luc, 0.5 µg of pRL-TK vector (Promega) was added to 100 µL NaCl 150 mM and 4 µL of Lipofectamine 2000 (Life Technologies). Luciferase activity was quantified 48 hours later using the Dual-Luciferase Reporter Assay System (Promega) according to the manufacturer's recommendations. Experiments were performed in triplicate.

Immunohistochemistry

Tissue microarrays (TMAs) were constructed from representative blocks from 14 NF2 partients and 31 sporadic patients. The construction of these TMAs was performed using a tissue arrayer (Beecher). In order to obtain a representative sampling of the tumors, each specimen was represented by three 1 mm cores, taken randomly in the tissue microarray block. In total, 2 blocks of TMA were constructed (including 45 tumors).

Sections of TMA blocks (3 µm) were immunolabeled (streptavidin-peroxidase protocol; immunostainer BenchMark Ventana) with antibodies to: ErbB2 (Ab-17; LabVision), ErbB3 (SGP1; LabVision) EGFR (2232; CST), PDGFR-ß (3162; CST), phospho-MAPK_202/204 (9101; CST), Stat3 (9132; CST), phospho-Akt₄₇₃ (D9E; 4060; CST), phospho-GSK3ß₉ (9331; CST), YAP (H-125; Santa Cruz/Tebu-Bio), Axl (C-20; Santa Cruz), and ß-catenin (clone 14; 610153; BD Biosciences). All antibodies are reactive in paraffin-embedded sections. Immunostaining of paraffin sections was done after dewaxing and rehydrating slides. Antigen retrieval was conducted by pretreatment at 95°C. Endogenous peroxidase was blocked with 0.5% hydrogen peroxide in water for 30 minutes. Substitution of the primary antibody with phosphate-buffered saline was used as a negative control.

For staining evaluation, 3 sections of each tumor were evaluated independently by 2 investigators (AC and DL). The percentage of positive cells was evaluated for each protein. The pattern of expression (cytoplasmic, membranous, and nuclear) was noted. All cores were evaluated separately; the mean score calculated was used to represent the whole immunoreactivity in tumors. Tumors were considered negative for a marker if no stained cells were detected in the 3 sections.

Receptor Tyrosine Kinase Arrays

Tumors were extracted in the provided extraction buffer using a dounce homogenizer. Following clarification of the extracts by centrifugation (20 000 g for 10 min at 4°C), protein concentration was measured by Bradford assay (Biorad). Proteome Profiler RTK Arrays (R&D) were incubated O/N with 1 mg of tumor extract and developed following the guidelines from the manufacturer.

Sample Preparation

Tissue samples were disrupted in Laemmli buffer (50 mM Tris [pH = 6.8], 2% SDS, 5% glycerol, 2 mM DTT, 2.5 mM EDTA, 2.5 mM EGTA, 1x HALT phosphatase inhibitor (78420; Perbio), protease inhibitor cocktail complete MINI EDTA-free (1836170; Roche; 1 tablet/10 mL), 2 mM Na3VO4, and 10 mM NaF) using a TissueLyser (Qiagen), and two 5 mm stainless beads per sample. Extracts were then boiled for 10 minutes at 100°C, passed through a fine needle to reduce viscosity and centrifuged for 10 minutes at 15 000 rpm. The supernatant was harvested and stored at −80°C. Protein concentration was determined (ref 23252; Pierce BCA reducing agent compatible kit). Samples were deposited onto nitrocellulose-covered slides (Schott Nexterion NC-C) using a dedicated arrayer (2470 Arrayer; Aushon Biosystems). Four serial dilutions, ranging from 500 to 62.5 µg/mL, and 3 technical replicates per dilution were deposited for each sample. Arrays were revealed with specific antibodies (see Supplementary Table 3 for a complete list of antibodies references) or without primary antibody (negative control), using Autostainer Plus (Dako). Briefly, slides were incubated with avidin, biotin, and peroxidase-blocking reagents (Dako) before saturation with TBS (Tris-buffered saline) containing 0.1% Tween-20 and 5% bovine serum albumin (BSA) (TBST-BSA). Slides were then probed overnight at 4°C with primary antibodies diluted in TBST-BSA. After washes with TBST, arrays were probed with horseradish peroxidase-coupled secondary antibodies (Jackson ImmunoResearch Laboratories) diluted in TBST-BSA for 1 hour at room temperature (RT). To amplify the signal, slides were incubated with Bio-Rad Amplification Reagent for 15 minutes at RT. The arrays were washed with TBST, probed with Alexa647-Streptavidin (Molecular Probes) diluted in TBST-BSA for 1 hour at RT, and washed again in TBST. For staining of total protein, arrays were incubated 15 minutes in 7% acetic acid and 10% methanol, rinsed twice in water, incubated 10 minutes in Sypro Ruby (Invitrogen), and rinsed again. The processed slides were dried by centrifugation and scanned using a GenePix 4000B microarray scanner (Molecular Devices). Spot intensity was determined with MicroVigene software (VigeneTech Inc).

Data Processing

Intensities are measured on a serial dilution for each sample in order to evaluate the dynamic range of detection for each antibody. Data were quantified by computing a relative protein expression level for each sample, followed by a normalization step that corrected for nonbiological bias using negative control slides and Syro Ruby slides. Quantification and normalization were simultaneously performed by NormaCurve.⁴⁹ An additional normalization step removed variability due to array effect and was performed by median loading. All primary antibodies used in RPPA had been previously tested by Western blotting to assess their specificity for the protein of interest.

RNA Extraction and Quantitative Real Time Polymerase Chain Reaction

Total RNA from HEI193 control and stably expressing GFP-YAP were extracted using RNeasy Mini kit (Qiagen) according the manufacturer's instructions. One microgram of extracted RNA was reverse transcribed with random primers following the manufacturer's protocol (Invitrogen; SuperScript III). For the Q-RT-PCR, 10 ng of cDNA was added in a reaction mix containing 0.5 µM of each primer in the LightCycler SybR green master mix (Roche) with RNase-free water adjusted to a total volume of 25 μl. The PCR program included a denaturation step at 95°C for 5 minutes followed by 45 cycles of 95°C for 10 seconds, 60°C for 15 seconds, and 72°C for 5 seconds. All reactions were performed in sextuplicate in 96-well plates. The quantification was done using the ΔΔCt method by comparing the expression of the target genes in GFP-YAP cells versus normal cells relative to the endogenous reference, the 18S ribosomal RNA (Eukaryotic 18S rRNA Endogenous Control; Applied Biosystems/Invitrogen). The following primer pairs were designed, validated, and used: survivin: (F) CAGTGTTTCTTCTGCTTCAAGG and (R) CTTATTGTTGGTTTCCTTTGCAT; erbb2: (F) CTTTGCTGTCCTGTTCACCA and (R) TCATCATCTTCACATTGAGTAGGC; erbb3: (F) CTGATCACCGGCCTCAAT and (R) GGAAGACATTGAGCTTCTCTGG; PDGFR: (F) CATCTGCAAAACCACCATTG and (R) GAGACGTTGATGGATGACACC.

Statistical Analysis

For tissue microarrays and RPPA results analysis, the Spearman' rank correlation coefficient was used to assess the independent significance of the various markers. The critical level of statistical significance was set at a P value of <.05.

Results

A Limited Set of Receptor Tyrosine Kinases Are Frequently Activated in Human Schwannomas

From Drosophila eye to mammalian Schwann cells, merlin appears to be a major regulator of RTK homeostasis. In these tissues, loss of merlin expression leads to increased RTK-mediated signaling, either due to accumulation of RTKs at the plasma membrane or loss of proper downregulation of receptor activity.^6,7 The misregulation of RTK activity is believed to be an important step in schwannoma development and growth. Hence, deciphering the landscape of activated RTKs in this type of tumor should help to identify possible therapeutic targets for which inhibitors already exist. We used RTK array technology on a set of 29 human sporadic schwannomas to generate the activity profile of 42 different RTKs (out of an estimated total of about 60 RTKs). We considered as activated, receptors that gave a signal that was above the signal of any of the negative controls. Three receptors were found activated in more than 50% of the tumors Her3 (86%), Axl (55%), and Ron (86%). Four more receptors were found activated in 25%–50% of the samples: EGFR (48%), Her2 (31%), ROR2 (45%), and Tie2 (34%). Activated PDGFRß, Her4, and VEGFR3 were detected in 21%, 21%, and 17%, respectively. Finally, five receptors were found at least once in our tumor set: Tyro3 (3%), FGFR2a (7%), c-Ret (7%), EphB2 (3%), and MuSK (3%) (Fig. 1A and B and Supplementary Table 1). To confirm expression of the detected active receptors, we performed Western blots on extracts from a subset of the schwannomas used for RTK arrays. The expression of EGFR, Her2, Her3, PDGFRß, Tie2, and Axl were confirmed (Fig. 1C). Interestingly, the expression levels varied widely from one tumor to another. We could not confirm the expression of Ron and ROR2 using 2 different antibodies to each receptor, although these antibodies were able to detect the proteins overexpressed in cell culture (not shown). We also examined the expression of several other RTKs that were not found to be activated by RTK arrays. IGFR1 was expressed in schwannoma, and we could detect low levels of c-Met (Supplementary Fig. 1A). This result may be due to the lack of sensitivity of arrays for these receptors or to their weak activation in the tumors. Nevertheless, this suggests that additional RTKs may be active in the tumors but not detected with RTK array. Because Western blot and RTK arrays are performed using extracts from tumors that may contain a fraction of nontumor tissue that also expresses RTKs, they are susceptible to generating false-positive signals. For example, PDGFRß signal could come from endothelial cells of blood vessels and not from tumor cells per se. We used immunohistochemistry (IHC) to evaluate the tissue expression pattern of the most frequently activated RTKs on a tissue microarray regrouping of 40 schwannoma biopsies (partially overlapping with the set used for RTK arrays and Western blots; see Supplementary Table 2). Her2, Her3, PDGFRß, and Axl were detected in tumor cells, although at variable levels (Fig. 2). Surprisingly, EGFR was not found in any of the samples, although it was easily seen in a control placental tissue (Fig. 2). This result suggests that EGFR signal obtained by Western blot and RTK arrays may have come from contaminating tissues. In conclusion, our analysis demonstrates that Her2, Her3, Axl, Tie2, and PDGFRß, but not EGFR, are frequently activated in the tumors, therefore defining the profile of RTKs that could be targeted for the treatment of schwannomas.

Fig. 1.

Only a limited set of receptor tyrosine kinases (RTKs) are frequently activated in human schwannomas. (A) Four examples of RTK arrays membrane incubated with extracts from human vestibular schwannomas. RTKs were considered activated when the signal was more intense than the control immunoglobulin spots. Strong signals at each corner correspond to a phosphotyrosine positive control. (B) Quantification of the percentage of tumors expressing specific activated RTKs. Fifteen out of 42 RTKs were detected in their activated form. The active RTKs that were not detected are also indicated. (C) Expression of EGFR, Her2, Her3, Axl, PDGFRß, and Tie2 were confirmed by Western blot in a series of 12 human schwannoma extracts. Extracts from HEI193 human schwannoma cell line (SC) were used for comparison. Actin was used as a loading control.

Fig. 2.

Immunohistochemical evaluation of receptor tyrosine kinase expression in human schwannomas. The expression of Her2, Her3, Axl, PDGFRß, and EGFR was evaluated by immunohistochemistry on schwannoma sections. The percentage of positive tumors is indicated on the right. No EGFR was detected in schwannomas, although the antibody gave a clear signal in the control placenta section (insert).

Expression and Activity of Major Mitogenic Signaling Pathways Are Heterogeneous in Human Schwannomas

Loss of NF2 leads to activation of several major mitogenic signaling pathways.^4,9,21–23 We used Western blotting to evaluate the expression of P42/44 MAPK, P38 SAPK, P46/55 JNK, mTor, Akt, STAT3, GSK3ß, and YAP in our human schwannoma biopsies (Fig. 3A). All tumors tested were positive for the various proteins. All of them also expressed several other components of the Hippo signaling pathway such as the cotranscription factor Taz and the kinases MST1/2 (Supplementary Fig. 1B). In contrast, Lats1/2 expression was rarely observed in schwannomas (Supplementary Fig. 1B). Specific expression in the tumor cells was assessed by IHC on schwannoma sections. All proteins were found to be expressed in more than 90% of the tumors. Phosphorylated forms of P42/44, P38, Akt, m-Tor, and Stat3 were also observed in most tumors by Western blot (Fig. 3A) or by IHC (Fig. 3B). Activated YAP was found in the nucleus of schwannoma cells in all tumors (Fig. 3B). However, despite being detected in 98% of the tumors, ß-catenin was only found in the cytoplasm, suggesting that the canonical Wnt pathway is not activated in schwannomas (Fig. 3B and Supplementary Fig. 2A). This was in clear contrast to the nuclear signal that we observed in a pancreatic tumor that was used as a positive control (Supplementary Fig. 2B). Remarkably, although most tumors expressed signaling proteins and their phosphorylated forms, we noticed great variability in the percentage of positive cells in each tumor (Fig. 4A). As presented in Fig. 4B, this percentage can fluctuate from less to 25% to more than 75% depending on the receptor tested. For example, Her3 was expressed in more than 75% of the cells in 42.5% of the tumors. Axl presented a more homogeneous distribution with 79% of the tumors presenting positive staining in more than 75% of the cells. In contrast, only 14% of tumors showed more than 75% of positive labeling for PDGFRß. We also observed that nearly two-thirds of the tumors (69%) expressed Her2 in more than 50% of the cells; however, the active phosphorylated form of Her2 was detected in more than 50% of cells in only 34% of the tumors (not shown). This result confirmed the differences we observed between Western blot and RTK arrays results for the expression of total and activated receptors. They suggest that the amount of receptor expressed will not necessarily reflect its activity in the tumor. Finally, for all the proteins tested by IHC, we couldn't find any significant differences in the expression patterns between sporadic and familial schwannomas.

Fig. 3.

Major mitogenic signaling pathways are active in human schwannomas. (A) The expression and phosphorylation of major signaling pathways known to be regulated by merlin were evaluated by Western blot in a series of 12 human schwannomas. HEI193 extracts were used for comparison. Actin was used as a loading control. (B) Immunohistochemistry was used to evaluate the expression of phosphorylated Akt₄₇₃, GSK3ß₉, and MAPK_202/204 as well as YAP, STAT3, and ß-catenin in human schwannoma sections. The enlarged insert for ß-catenin clearly shows negative nuclear staining. The percentage of tumor with a positive detection of the markers is indicated.

Fig. 4.

Schwannomas are very heterogeneous for the percentage of cells expressing signaling effectors. (A) The heterogeneity of the expression of Her2, Her3, PDGFRß, phospho-MAPK_202/204, STAT3, and YAP was evaluated by immunohistochemistry. The top panel provides an example of a low percentage of positive cells, whereas the lower panel represents tumors with a high percentage of labeled cells. (B) The graph represents the percentage of tumors that express Her2, Her3, PDGFRß, or Axl in <25%, between 25% and 50%, between 50% and 75%, and in more than 75% of cells. (C) The same experiment was performed for the activated forms of MAPK, Akt, STAT3, and YAP. The phosphorylated forms of MAPK and Akt were independently evaluated in the cytoplasm and in the nucleus.

Most tumors expressed phospho-P42/44 MAPK_202/204, nuclear Stat3, and nuclear YAP in <50% of cells. Only phospho-AKT₄₇₃ was detected in more than 50% of cells in a large majority of the tumors. Altogether, a surprisingly large fraction of tumors expressed detectable levels of signaling effectors in <50% of the tumor cells (Fig. 4C). This heterogeneity was observed both in Antoni A and B areas as shown for YAP staining in Supplementary Fig. 2C. We next determined if the percentage of positive cells for the various markers was linked or if it fluctuated independently. Using the Spearman correlation test, we found no significant correlation between the percentages of cells expressing nuclear YAP protein (Table 1) with any other marker tested. However, we showed that the percentage of cells expressing Her3, Her2, and PDGFRß were correlated (Table 1). We also observed that phospho-P42/44 MAPK_202/204 and cytoplasmic phospho-Akt₄₇₃ were linked to activated (nuclear) STAT3. In both cases, these correlations suggest that the expression of growth factor receptors and the activation of signaling pathways might be coregulated in the tumors. Indeed, in the case of nuclear phospho-P42/44 MAPK_202/204 and STAT3, double labeling on the same sections showed coexpression at the cellular level (see Supplementary Fig. 2D).

Table 1.

Correlation between the percentages of positive cells in human schwannomas

Her2
Her3	0.036–0.33
PDGFRß		0.006–0.42
Axl
p-MAPK cyto.
p-MAPKnucl.					2.84E-27–0.98
STAT3 nucl.			0.044–0.32	0.033– 0.35	0.017–0.38	0.023–0.36
YAP nucl.
p-AKT473 cyto.	0.05–0.33						0.002–0.49
P-AKT473 nucl.									0.038–0.35
	Her2	Her3	PDGFRß	Axl	p-MAPK cyto.	p-MAPK nucl.	STAT3 nucl.	YAP nucl.	p-AKT473 cyto.	p-AKT473 nucl.

Immunohistochemistry was used to evaluate the percentage of positive cells in each schwannoma for the indicated signaling effectors. The P value and correlation coefficients are indicated when they are significant (P <.05). The results suggest that several receptors and signaling pathways are coregulated in the tumors.

In conclusion, we observed that the expression levels for most of the signaling effectors we tested are very heterogeneous in schwannomas. Hence, signaling effectors in which expression levels or activity are correlated to clinical parameters such as tumor cell proliferation would likely represent the most relevant targets for therapy.

Proliferation of Tumor Cells Is Correlated with a Signaling Network Linked to YAP

RPPA is a powerful technique that allows precise quantification of the expression levels and the activity (using phospho-specific antibodies) of proteins in tumor extracts.²⁰ As an approach to identifying pathways that could be linked to the proliferation of tumor cells, we applied this technique to study our schwannoma samples. We used a set of 35 antibodies (Supplementary Table 3) to identify correlations between the expression of RTKs, effectors of major signaling pathways, their phosphorylated forms, and the levels of Ki67 (a key indicator of tumor cell proliferation). The analysis was performed on 30 human sporadic schwannomas. We showed that the expression of Ki67 correlated with the level of only 5 markers (Fig. 5A). Hence, Ki67 correlated positively with phospho-Her3₁₂₈₉, PDGFRß, phospho-PDGFRß₁₀₂₁, and YAP and negatively to phospho-Stat3. In addition, the analysis indicated that YAP levels are correlated with PDGFRß and phospho-PDGFRß₁₀₂₁. Interestingly, phospho-Her3₁₂₈₉ expression was also linked to phospho-P38_180/182, phospho-Her2₈₇₇, and phospho-Met_1234/1235. YAP activity is notably inhibited by phosphorylation on serine 127 upon activation of the Hippo signaling pathway, which prevents its nuclear localization. However, Ki67 was not linked to phosphorylated YAP₁₂₇ (P = .7) but to the expression level of YAP. Using IHC, we observed that the majority of tumors with strong cytoplasmic YAP staining also displayed elevated levels of nuclear YAP. Similarly, low cytoplasmic YAP staining was usually associated with low levels of nuclear YAP (Fig. 5B). When we used the Spearman correlation test on cytoplasmic and nuclear YAP intensity, we observed a significant correlation between them (P = .00009, ρ = 0.38). These results indicate that the activity of YAP in tumors, which is mediated by the nuclear pool, is primarily controlled by its expression level in the cells.

Fig. 5.

Reverse phase protein array analysis of human schwannomas reveals signaling network correlated with proliferation marker ki67. (A) The expression of YAP, PDGFRß, phospho-PDGFRß₁₀₂₁, phospho-Her3₁₂₈₉, and Phospho-STAT3₇₀₅ are correlated with the expression of Ki67 in human schwannoma samples. The P value and correlation coefficient are indicated for each correlation. For these 5 markers, the other significant correlations are indicated. (B) YAP intensity in the nucleus and the cytoplasm were scored by immunohistochemistry. High nuclear staining was primarily associated with high cytoplasmic staining and vice versa. The percentage of tumors represented in each case is provided. A Spearman correlation test indicates that the intensities in the nucleus and the cytoplasm are linked (P = .0009), suggesting that the activity of YAP is a function of its expression level.

PDGFRß ligand PDGF, Axl and its ligand Gas6, and Her2 were described as transcriptional targets of YAP in different cell types.^24–26 In order to test the role of YAP in the activation of receptors and signaling pathways linked to the Ki67 in schwannomas, we overexpressed GFP-YAP in the human schwannoma cell line HEI193²⁷ (Fig. 6A). By Western blot, we found that YAP stimulated PDGFRß and Her2 expression. However, Axl levels remained essentially unaffected. More surprisingly, Her3 levels were strongly increased in GFP-YAP expressing cells, although this receptor was never listed as a transcriptional target of YAP (Fig. 6B). Using quantitative PCR, we showed that YAP overexpression stimulated the transcription of PDGFRß and Her2. Survivin, which is a well-characterized transcriptional target of YAP, was used as a control and showed a robust response to GFP-YAP (Fig. 6C). Once again, Axl showed no induction upon YAP overexpression. Very interestingly, Her3 transcription was the most strongly stimulated by YAP. Furthermore, when HEI cells expressing GFP-YAP were treated with verteporfin, (a drug that blocks YAP transcriptional activity by dissociating it from the Tead DNA-binding proteins), we observed a decrease of PDGFRß and Her3 by Western blot and quantitative PCR (6D and 6E). Hence, it appears that the expression of PDGFRß, Her3, and Her2 is under the control of YAP in schwannomas, the first two depending on YAP/Tead interaction. Altogether, our results suggest that the proliferation of tumor cells in human schwannomas is under the control of a limited signaling network linked to YAP.

Fig. 6.

The most frequently activated receptor tyrosine kinases in schwannomas are transcriptional targets of YAP. (A) A HEI193 human schwannoma cell line was created that stably overexpresses GFP-YAP contructs. The fusion protein accumulates in the nucleus (left). Western blot confirms the overexpression of GFP-YAP in the cells. (B) Western blot shows that the expression of Her2, Her3, PDGFRß, and survivin is markedly induced upon GFP-YAP expression. Axl levels appear essentially unaffected. (C) Quantitative Real Time-PCR confirms that the induction of Her2, Her3, and PDGFRß by GFP-YAP is of transcriptional orgin. Survivin, a known target gene of YAP, was used as a positive control. On the contrary, Axl is not induced by YAP. The average induction folds are indicated. The graph is representative of 6 experiments. (D) HEI193 cells overexpressing GFP-YAP treated with the YAP/Tead dissociating compound verteporfin show a strong decrease in Tead-mediated transcriptional activity measured by luciferase reporter assay. (E) The protein levels of Her3, PDGFRß, and survivin decrease in response to treatment with verteporfin. (F) Similarly, quantitative Real Time-PCR indicates that verteporfin treatment of HEI193 GFP-YAP cells reduces mRNA levels of survivin, Her3, and PDGFRß. Her2 expression is unaffected. (* = P value <.05; N/S = not significant)

Discussion

In the present study, we have combined Western blotting, IHC, RTK arrays, and RPPA to identify signaling modules that are linked to tumor cell proliferation in human schwannomas and could represent potential therapeutic targets.

Our screening of activated RTK is the first to provide a precise landscape of RTK activation in a large series of human schwannomas. We showed that only 6 RTKs (out of 42 studied) are found to be activated in more than 20% of the tumors (Her2, Her3 and Her4, Axl, PDGFRß, and Tie2). It has been known for a long time that ErbB receptors are important for Schwann cell biology. This family of receptors was also shown to be expressed in schwannomas and to support schwannoma cell proliferation in culture.^28,29 In addition, Neuregulin1, the ligand of Her3, was shown to be expressed in human schwannomas and to promote tumor cell proliferation via an autocrine loop.^18,30 Remarkably, Her3 is phosphorylated in a large majority of the tumors. Her3 does not possess intrinsic tyrosine kinase activity and cannot homodimerize.³¹ It is phosphorylated on tyrosines primarily through dimerization with other ErbB family members. Therefore, its frequent phosphorylation may reflect dimerization with other activated Her members. This hypothesis is supported by RPPA data indicating that the levels of phosphorylated Her3₁₂₈₉ and phosphorylated Her2₈₇₇ are correlated. Nevertheless, most of the tumors tested showed an activation of Her3 in the absence of activated Her2 or Her4, as assessed by RTK arrays (Supplementary Table 1). It is possible that Her3 dimerized with other receptors in the tumors. It has been shown that Her3 can associate with c-Met.³² We could detect c-Met by Western blot in only a subset of tumors (see Supplementary Fig. 1). RPPA analysis showed a positive correlation between phospho-Her3₁₂₈₉ and phospho-Met_1234/1235. However, the activated receptor was not detected by RTK array, suggesting that it might be expressed at low levels and only be weakly activated in schwannomas. c-Met expression has been reported in schwannomas in several studies,^33,34 and its expression even at a low level could have a significant role in schwannoma development. EGFR is another candidate for dimerization with Her3. The question of EGFR expression in schwannomas has been the object of conflicting reports.^28,35 Our IHC clearly demonstrated that schwannomas do not express detectable levels of EGFR. This absence of EGF receptor expression likely explains the failure to inhibit schwannoma growth of patients treated with the EGFR inhibitor Erlotinib.³⁶ Hence, our results confirm that this receptor is unlikely to be an efficient therapeutic target in human schwannomas. In the case of Her2, our results may help to explain the outcome of a recent clinical trial using lapatinib,³⁷ an inhibitor of both EGFR and Her2. The impact of this treatment on schwannoma growth is likely to result from the sole inhibition of Her2, given the absence of EGFR expression that we observed. Upon lapatinib administration, 23.5% of the patients showed significant tumor reduction.³⁷ This number is close to the percentage of tumors in which we detected an activated form of Her2 by RTK arrays. It is also similar to the percentage of tumors that express Her2 in more than 75% of cells. Whether the clinical response to lapatinib is linked to these parameters remains speculative at this point but deserves further investigation. In conclusion, our results point to Her3 as a promising therapeutic target for the treatment of schwannomas. Furthermore, dual Her2/Her3 inhibitors may show a greater efficacy for a subset of patients.

The inhibition of angiogenesis remains the unique chemotherapeutic strategy for treating schwannomas to this day. Targeting VEGF has demonstrated some efficacy, although response to treatment is not seen for all patients.³⁸ Very interestingly, our analysis of RTK expression shows that, in addition to VEGFR, 3 other receptors involved in angiogenesis regulation are frequently expressed in schwannomas (ie, Axl, Tie2 and PDGFRß). Axl is a member of the TAM (tyro3-Axl-Mer) family of receptors. It is involved in many biological processes such as survival, proliferation or migration, and its expression is altered in many different types of cancers.³⁹ In Schwann cells, stimulation by its ligand, Gas6, promotes proliferation and survival through a FAK/Src/NFkappaB pathway.^17,40 Axl was also shown to promote angiogenesis. One mechanism proposed is that Axl stimulates the expression of angiopoïetin-2 (Ang2), which is a ligand of Tie2 receptor.⁴¹ Tie2 has 2 ligands: Ang1 and Ang2. Ang1 stimulates Tie2 and leads to blood vessel stabilization, whereas Ang2 acts antagonistically and promotes vessel sprouting. Ang1 and Ang2 display both antitumorigenic and protumorigenic effects. Nevertheless, concomitant inhibition of Ang2 or Tie2 and VEGF efficiently induces apoptosis of endothelial cells and appears to be a promising strategy for the inhibition of angiogenesis. Indeed, several clinical trials based on this principle are currently in progress.⁴²

Studies performed on cultured cells have increased the understanding of how merlin controls proliferation at the molecular level. Defining which signaling pathways are correlated with proliferation in tumors represents an important step for identifying potential therapeutic targets. Therefore, we decided to use RPPA to evaluate protein expression in tumor biopsies. RPPA has been used to study signaling in various types of cancers,⁴³ but this is the first time this strategy has been applied to the study of human schwannomas. We observed that a small set of signaling effectors correlates positively with the proliferation marker Ki67 levels. These include YAP, phospho-Stat3₇₀₅, PDGFRß, phospho-PDGFß₁₀₂₁, and phospho-Her3₁₂₈₉. The inverse correlation of phospho-Stat3₇₀₅ to Ki67 levels is unexpected, given that activation of Stat3 appears essentially pro-proliferative and anti-apoptotic in cancer cells and schwannoma cells in culture.^21,44 However, several reports showing that the activation of Stat3 is linked to a better prognosis in various cancers,⁴⁵ indicate that its role in tumorigenesis depends on the tumor type. YAP is the main effector of the Hippo signaling pathway. Upon loss of NF2, YAP is dephosphorylated, allowing its entry into the nucleus where it activates the transcription of pro-proliferative and anti-apoptotic targets. Remarkably, Ki67 correlates with the amount of YAP but not with the level of phospho-YAP_S127. It is possible that variations of YAP levels in the tumors are from transcriptional origin. However, changes in YAP levels could also be due to modulation of its degradation. Indeed, CK1 and LKB1, as well as Lats, have been shown to play a major role in YAP degradation.^46,47

A remarkable aspect of the present study is the fact that 3 of the most frequently activated RTKs in our tumor samples are potential targets of YAP. First, a recent work showed that PDGF is transcriptionally upregulated by YAP.²⁴ A second study identified PDGFRß promoter sequences by ChIP-onChIP, using antibodies to YAP and TEAD suggesting that PDGFRß is a transcriptional target of YAP.²⁶ We previously observed that the PDGFRß receptor is transcriptionally upregulated in Schwann cells and schwannomas upon NF2 inactivation.⁴⁸ In this report, we confirm that overexpression of YAP in the HEI193 human schwannoma cell line induces PDGFRß mRNA and protein levels. In addition, the overexpression of YAP strongly induces Her2 and Her3 levels in Hei193 cells. The fact that YAP is a cotranscription factor, together with the decrease in mRNA levels in response to YAP inhibitor verteporfin, supports the idea that Her3 and PDGFRß could be transcriptional targets of YAP/Tead in schwannomas. Indeed, promoter sequence analysis shows several Tead-binding sequences for PDGFRß and Her3 (data not shown). The absence of Her2 response to verteporfin suggests that Her2 expression requires a DNA binding partner distinct from Tead. Due to the lack of antibodies specific enough, we could not use RPPA to evaluate whether Axl expression was linked to Ki67 or to other markers. However, Axl receptor and its ligand Gas6 were previously shown to be bona fide target genes of YAP.²⁵ Taken together, our observations strongly suggest that the proliferation of tumor cells in human schwannomas is, at least in part, under the control of a signaling module where the Hippo effector YAP stimulates the expression of several target genes, notably Her3, PDGFRß, PDGF, and probably Her2 and Axl. These receptors, once activated by their ligands (some of which are known targets of YAP) would stimulate mitogenic downstream signaling pathways that contribute to tumor growth (see Fig. 7 for a model).

Fig. 7.

Model of signaling pathway regulation in human schwannomas. When expressed in Schwann cells, merlin inhibits growth factor receptor delivery at the plasma membrane, as we previously demonstrated (1). Merlin also blocks YAP nuclear accumulation (2) leading to the inhibition of proliferation (3). In schwannomas, where merlin expression is lost (4), YAP accumulates in the nucleus (5) and stimulates the expression of PDGFRß, Her2, Her3, and possibly Axl as well as some of their ligands such as PDGF and Gas6 (6). In the absence of merlin, these receptors accumulate at the cell surface (7) where they stimulate mitogenic signaling pathway (8) proliferation and survival (9). In parallel, Tie2 expression may stimulate the development of new blood vessels (10) contributing to tumor development.

Schwannomas are usually considered to be a histologically homogenous type of tumor. However, their response to anti-VEGF or anti-Her2 therapies proved to be variable. One parameter that may account for this result is the heterogeneity in the expression level of the target proteins. This is likely to be an important parameter for evaluating a candidate therapeutic target. Our IHC study of human schwannomas has shown that these tumors are remarkably heterogeneous in the expression of most signaling effectors as tested. Although they were detected in almost all tumors, the percentage of positive cells was frequently less than 50% and some even less than 25%. We can only speculate about the origins of this observed variability, but our results suggest that a therapeutic strategy focusing on a single target is unlikely to be efficacious for the majority of patients. Indeed, this prediction has been confirmed by the results of a clinical trial targeting Her2.³⁷ One option for overcoming this limitation could be to target several pathways simultaneously. In this case, inhibiting targets that are not coregulated in the tumors might prove to be the most efficient strategy.

Altogether, our work shows that combining proteomic strategies to measure protein expression and activity in NF2- deficient tumors is a powerful approach for identifying and evaluating candidate therapeutic targets. It also provides important clues that can be used to better understand the mechanisms that promote tumor growth. This type of analysis could also be used to study other NF2-deficient tumors such as meningiomas or ependymomas. Their comparison with schwannomas could tell if the forces that sustain tumor growth are common. Similarly, comparing sporadic to inherited tumors, or NF2-deficient to NF2-expressing tumors would certainly provide important informations on the mechanisms of tumor development and the role of Merlin in this process. Finally, a proteomic study comparing schwannomas and meningiomas would certainly lead to the identification of therapeutic targets common to both types of tumors. This would be of great relevance for NF2 patients who frequently develop several tumors of both types simultaneously.

Supplementary Material

Supplementary material is available online at Neuro-Oncology (http://neuro-oncology.oxfordjournals.org/).

Funding

This work was supported by Institut National du Cancer (INCA), Association pour la recherche sur le Cancer (ARC), Ligue contre le Cancer (LCC to A.B.), Association Neurofibromatose et Recklinghausen (ANR).