LZTR1 Mutation Mediates Oncogenesis through Stabilization of EGFR and AXL

Leucine zipper‐like transcription regulator 1 (LZTR1) is a Kelch-BTB-BACK domain-containing protein that functions as substrate adaptor of a CRL complex, CRL3^LZTR1 (1, 2). To date, a full characterization of the function of the LZTR1 protein is still lacking, but interest in LZTR1 function continues to increase due to the growing number of cancer types that are reported to harbor genetic alterations of the LZTR1 gene. After our initial discovery of LZTR1 mutations and deletions in glioblastoma multiforme (GBM; ref. 1), genetic alterations of LZTR1 have been identified in numerous cancers including hepatocellular, esophagogastric, colorectal, and breast carcinoma, lung adenocarcinoma, clonal hematopoiesis disorders, and a variety of human tumor predisposition syndromes (1, 3–16). For example, germline monoallelic mutations of LZTR1 were reported in patients with schwannomatosis, a genetic disorder characterized by highly invalidating tumors of the peripheral nervous system (PNS; refs. 4, 17). Mutations of LZTR1 involve both Kelch and BTB-BACK domains, substrate and CUL3 interacting domains, respectively, albeit without clustering in hot spots. In vitro experiments and data from human genetic studies indicated that LZTR1 mutations are loss-of-function events and pointed to a potential tumor suppressor function for this gene (1, 3–5). However, the role of LZTR1 as a tumor suppressor gene in the whole organism remained to be charted.

Similarly, the identity of the protein substrates targeted by LZTR1-mediated ubiquitylation and the oncogenic functions implemented by the deregulation of specific LZTR1 substrate(s) in LZTR1-mutant disorders remained puzzling. Although initial reports suggested that RAS proteins interact with LZTR1 and undergo direct ubiquitylation by CRL3^LZTR1 complexes (18, 19), later studies failed to confirm these interactions (20). Instead, the RAS family member RIT1 was identified as a substrate of LZTR1 through RIT1-specific protein interaction assay, and it was suggested that dysregulation of RIT1 contributes to the activation of MAP kinase signaling in Noonan syndrome caused by pathogenic mutations of LZTR1 (20). However, this study left open the question of the broad cellular scope of LZTR1 substrates and especially their role in sporadic tumors and cancer predisposition syndromes such as schwannomatosis sustained by LZTR1 inactivation.

Here, we used multiple unbiased proteomics screens to capture substrates of LZTR1-mediated ubiquitylation/degradation and uncovered EGFR and AXL, two oncogenic receptor tyrosine kinases (RTK), as novel protein substrates of LZTR1 activity. Interaction with and ubiquitylation by LZTR1 of both RTKs was consequent to ligand-induced activation, leading to lysosomal mediated degradation and signaling downregulation. We modeled LZTR1 tumor suppressor function in a mouse model sustained by genetic inactivation of Lztr1 and Cdkn2a in the nervous system. Compound mice developed tumors resembling neurofibroma, malignant PNS tumors, and schwannomas, all characterized by elevated expression of EGFR and AXL. Finally, we report that LZTR1 germline and somatic mutations identified in patients with cancer fail to balance EGFR and AXL cellular levels. Consequently, LZTR1 inactivation induces vulnerability to the concurrent inhibition of the two RTKs, thus providing new treatment opportunities to patients affected by cancer harboring mutation of LZTR1.

To identify the protein substrates targeted by LZTR1-dependent ubiquitin-mediated degradation, we performed three orthogonal and unbiased screens (Fig. 1A). First, to uncover ubiquitylation targets of LZTR1, quantitative ubiquitin diglycine (diGly) proteomics, a peptide antibody-based affinity approach that enriches for and identifies endogenously ubiquitylated proteins, was performed in Lztr1^fl/fl;Rosa-CreER mouse astrocytes, the cellular model most frequently used for induction of high-grade glioma in mice (21, 22). In this cellular system, the endogenous Lztr1 gene was deleted by 4-hydroxytamoxifen (4-OHT)–induced recombination and LZTR1 reexpressed by the tetracycline-inducible system (Fig. 1A). We found that 1,076 proteins had higher ubiquitylation in LZTR1-reconstituted cells compared with Lztr1-deleted astrocytes (Fig. 1B; Supplementary Table S1). Second, Tandem Mass Tag (TMT)–based quantitative proteomics was used to identify proteins elevated by genetic deletion of the Lztr1 gene in mouse embryonic fibroblasts (MEF; Fig. 1A). We detected a set of 33 proteins that accumulated at higher levels in constitutive Lztr1 knockout MEFs (Fig. 1C; Supplementary Table S2). Third, we probed the LZTR1 cellular interactome by immunoprecipitation/mass spectrometry of U87 human glioma cells stably expressing LZTR1 wild-type or the LZTR1^R810W mutant that we initially reported as somatic mutation in human GBM (Fig. 1A). The LZTR1^R810W protein harbors a loss-of-function mutation in the BTB-BACK domain that impairs CUL3 recognition, thus stabilizing the protein without affecting the substrate-interacting Kelch domain (1). The analysis uncovered 184 proteins specifically interacting with LZTR1 wild-type and LZTR1^R810W mutant with spectral counts > 2-fold in LZTR1^R810W compared with LZTR1 wild-type (Fig. 1D; Supplementary Table S3), a criterion that we applied to identify candidate substrates of LZTR1, which more stably interact with the ineffective mutant. The RTK epidermal growth factor receptor (EGFR) was the only hit common to all three screens (Fig. 1B–E). Another oncogenic RTK, AXL (23), also scored as a positive hit in two of the three screens (Fig. 1B, D, and E). RIT1, a RAS-related small GTPase recently reported as a substrate of LZTR1 (20), also emerged as a positive hit from two of the three screens (Fig. 1B, C, and E). Using independent TMT-based quantitative proteomics experiments in HeLa cells, we confirmed that CRISPR/Cas9-mediated knockout of LZTR1 consistently increased EGFR and AXL proteins compared with control cells (Fig. 1F).

Given the large spectrum of cancer types harboring genetic alterations of LZTR1, we sought to define the activity of LZTR1 in cells originating from different tissues and the relationship with the other described LZTR1 substrates. Genetic deletion of the Lztr1 gene in mouse astrocytes and CRISPR/Cas9-mediated knockout of LZTR1 in immortalized Schwann cells SW10, U87 malignant glioma, and HeLa cells consistently induced accumulation of EGFR and AXL in the absence of change of mRNA (Fig. 2A–D; Supplementary Fig. S1A and S1B). Accumulation of RIT1 was also readily detected in LZTR1-deleted cells (Fig. 2A–D). However, deletion of LZTR1 did not change the steady-state levels of any of the RAS isoforms, KRAS, HRAS, and NRAS (Supplementary Fig. S1C). Similarly the levels of ERBB2, ERBB4, and FGFR3 were unchanged, thus highlighting the specificity of LZTR1 activity toward the RTKs EGFR and AXL (Supplementary Fig. S1D). We then tested the consequences of ectopically increasing LZTR1 expression on endogenous and exogenous EGFR and AXL in several cell types. Expression of LZTR1 in SW10 cells caused a noticeable decrease of endogenous EGFR and AXL proteins but only minimal change of RIT1 (Fig. 2E) and led to destabilization of exogenously expressed EGFR and AXL in a dose-dependent manner (Fig. 2F and G). RAS isoforms were not affected by LZTR1 expression (Supplementary Fig. S1E). The effect of LZTR1 on EGFR and AXL protein stability was confirmed by the markedly prolonged half-life under cycloheximide (CHX) treatment in cells with knockout of LZTR1 (Fig. 2H–K) and accelerated protein turnover by increased LZTR1 (Fig. 2L–O). Finally, immunoprecipitation from lysates of cells expressing LZTR1-FLAG and EGFR-GFP or AXL-HA demonstrated that LZTR1 physically interacts with EGFR (Fig. 2P and Q) and AXL (Fig. 2R and S). Glutathione S-transferase (GST) pulldown assays with different GST-EGFR protein domains and LZTR1-FLAG from HEK293T cell lysates revealed that the intracellular domain of EGFR interacts with LZTR1 (Supplementary Fig. S2A and S2B). Reciprocal experiments with purified GST-LZTR1 deletion polypeptides identified the Kelch domain of LZTR1 as the EGFR-binding region (Supplementary Fig. S2C and S2D).

Following ligand-mediated activation and endocytosis, RTKs are targeted to the lysosome for degradation (24–26). LZTR1-mediated destabilization of EGFR and AXL was relieved by the lysosomal inhibitor bafilomycin A1 (BFLM A1) but not the proteasome inhibitor bortezomib (BRTZ; Fig. 3A and B). In contrast, RIT1 reduction by LZTR1 was rescued by BRTZ but not BFLM A1, a finding consistent with the engagement of different proteolytic machinery for these substrates of LZTR1 (Supplementary Fig. S3A). Confirming published findings (20), RIT1 physically interacted with LZTR1 in the absence of association with EGFR and AXL, thus suggesting independent regulation of substrates by LZTR1 (Supplementary Fig. S3B). Consistent with the CULLIN-mediated mechanism of degradation by the CRL3^LZTR1 complex, the NEDD8-activating enzyme (NAE) inhibitor MLN4924, which blocks neddylation of CULLIN and inactivates CRLs (27), rescued LZTR1-induced destruction of EGFR, AXL and RIT1 (Fig. 3C and D; Supplementary Fig. S3C). Next, we examined the ubiquitylation status of EGFR and AXL in the presence of LZTR1. Expression of LZTR1 increased the formation of high-molecular-weight polyubiquitylated EGFR, and this effect was abolished when cells were treated with MLN4924 (Fig. 3E). LZTR1 also increased the polyubiquitylated species of AXL, which accumulated at much higher levels when cells were treated with the lysosome inhibitor chloroquine (CQ), whereas the proteasome inhibitor MG132 was ineffective (Fig. 3F). When coexpressed with ubiquitin wild-type or ubiquitin variants in which all lysine residues had been changed to arginine except for lysine-48 (K48) or lysine-63 (K63), LZTR1 induced EGFR and AXL ubiquitylation by ubiquitin wild-type and predominantly K63 but minimally K48, indicating that LZTR1 promotes preferentially K63-mediated ubiquitylation linkage, which is the key signal for efficient targeting of RTKs to the lysosomal degradation pathway (Fig. 3G and H; ref. 28).

Missense point mutations of LZTR1 that have been reported in sporadic tumors and the germline of patients with schwannomatosis target the Kelch domain, which is required for substrate recognition (1, 4, 8, 29, 30), but also the BTB-BACK domain, which interacts with CUL3 (Supplementary Fig. S3D). To interrogate the cancer-specific significance of the inactivation of EGFR and AXL by LZTR1, we analyzed a panel of mutations of LZTR1 and found that most mutants compromised the degradation capacity of LZTR1 toward EGFR and AXL (Fig. 3I and J, for mutants in the Kelch domain; Fig. 3K and L for mutants in the BTBBACK region). Consistent with impaired degradation of EGFR, most Kelch domain mutated proteins failed to interact with EGFR in coimmunoprecipitation assays (Supplementary Fig. S3E). RIT1 destabilization was also impaired by all LZTR1 mutants tested except the R170Q mutation, which exhibited an activity comparable to LZTR1 wild-type but was ineffective toward AXL and EGFR (Supplementary Fig. S3F). Therefore, LZTR1 mutations that are causally implicated in cancer and schwannomatosis generally compromise the ability of this ubiquitin ligase to regulate EGFR and AXL protein balance.

RTK signaling begins with the ligand–receptor interaction resulting in receptor dimerization, autophosphorylation of the C-terminal tail, and recruitment of transducers of signaling. As negative feedback regulation, ligand binding to RTKs also triggers internalization and degradation of the activated receptor with consequent signal termination (31–33). We found that EGF and GAS6 stimulation promoted the interaction of LZTR1 with the active, phosphorylated form of EGFR and AXL, respectively, when tested by coimmunoprecipitation (Fig. 4A and B). EGF-induced stimulation of LZTR1–EGFR complex formation was impaired for the kinase-defective mutant EGFR^K721R, further supporting the conclusion that EGFR phosphorylation and activation promotes binding to LZTR1 (Fig. 4C). To directly determine whether the interaction with LZTR1 controls EGFR and AXL degradation initiated by EGF and GAS6, respectively, we treated cells expressing EGFR or AXL with EGF or GAS6 for different time points in the presence or absence of LZTR1. LZTR1 markedly increased the degradation of EGFR and AXL at the time points analyzed (Fig. 4D and E). LZTR1 also enhanced the EGF-induced ubiquitylation of EGFR (Supplementary Fig. S3G). Conversely, constitutive Lztr1 deletion in MEFs impaired EGF-induced quenching of EGFR protein and signaling, maintaining active phospho-EGFR and preventing the loss of phospho-ERK (Fig. 4F and G). Similarly, CRISPR/Cas9-mediated knockout of LZTR1 in HeLa cells stabilized EGFR and phospho-EGFR and resulted in persistent activation of MAPK when the signaling mechanisms were interrogated in the context of increasing doses of EGF (Fig. 4H). Knockout of LZTR1 caused comparable stabilization of EGFR and increased signaling when the receptor was activated by epiregulin, an EGFR low-affinity ligand (34), indicating that the negative regulation of LZTR1 upon EGFR is independent of the particular ligand activating the receptor (Fig. 4I). Similarly, knockout of LZTR1 attenuated GAS6-mediated AXL degradation and prolonged STAT3 signaling (Fig. 4J).

Receptor ubiquitylation is a sorting signal that targets activated RTKs at the cell surface to the lysosome by trapping them within clathrin-coated pits and multivesicular endosomes (35). Ubiquitylation of RTKs is not necessary for internalization but is required for lysosome targeting and degradation (36, 37). To determine the role of LZTR1 for EGFR and AXL endocytosis and trafficking to the lysosome, we first examined the subcellular localization of LZTR1 in relationship with EGFR and AXL upon ligand stimulation. When expressed as EGFR–GFP fusion, the GFP signal colocalized with LZTR1-FLAG at the cell periphery 5 minutes after the addition of EGF at 4°C (Supplementary Fig. S4A). Endogenous AXL and LZTR1 colocalized at the cellular membrane after the addition of GAS6 to the culture (Supplementary Fig. S4B). Next, we analyzed the localization of ligand-occupied EGFR and AXL in LZTR1 wild-type and knockout cells. For the EGFR-EGF assay, we used EGF conjugated with Alexa-647 to track ligand–receptor interaction. In control cells analyzed 5 minutes after cell stimulation with EGF-Alexa-647, the EGF–EGFR ligand–receptor complex colocalized with the early endosome marker EEA1, and this colocalization decreased 45 minutes after EGF stimulation (Fig. 5A and B, red bars). As expected, colocalization of EGF-Alexa-647 with the lysosomal marker LAMP1 was minimal 5 minutes after EGF stimulation and increased after 45 minutes (Fig. 5C and D, red bars). In cells harboring knockout of LZTR1, we initially detected a reduction of the EGF-EEA1 early endosome colocalization 5 minutes after EGF addition, but the early endosome compartmentalization of the EGF–EGFR complex remained stable 45 minutes after EGF stimulation, thus persisting at levels slightly higher than those of control cells (Fig. 5A and B, yellow bars). Loss of LZTR1 impaired lysosomal targeting of EGFR, as shown by the markedly reduced pool of EGFR colocalizing with the lysosomal marker LAMP1 45 minutes after EGF stimulation compared with control cells (Fig. 5C and D, yellow bars). For the AXL-GAS6 assay, we tracked AXL trafficking after GAS6 addition by immunofluorescence of the endogenous AXL protein. As for EGFR, colocalization of AXL with EEA1 was reduced in LZTR1 knockout cells 15 minutes after stimulation with GAS6 compared with wild-type cells (Fig. 5E and F, yellow bars). A comparable statistically significant reduction in colocalization of AXL with LAMP1 was obtained in LZTR1 knockout cells relative to wild-type when the analysis was performed 60 minutes after stimulation with GAS6 (Fig. 5G and H, yellow bars). These results indicate that in the absence of LZTR1, EGFR and AXL are not efficiently sorted to the lysosome but remain in the early endosomal compartment and are likely recycled to the plasma membrane with defective termination of signaling.

In humans, the LZTR1 gene is targeted by loss-of-function somatic mutations in numerous cancers and the germline of patients with schwannomatosis (1, 3–16). To ask whether genetic deletion of Lztr1 in the mouse delivers a model of tumorigenesis and whether loss of LZTR1 activity in the whole organism is associated with accumulation of EGFR and AXL, we generated constitutive and conditional Lztr1 knockout mice (38). Constitutive Lztr1^−/− mice died at mid-gestation with severe and progressive apoptotic cell death of liver cells (Supplementary Fig. S5A). Embryos also exhibited abnormal expansion of the ventricular/subventricular zone (VZ/SVZ) of the telencephalon and a reduced cortical plate (CP), suggestive of a relative block of neurogenesis and neuronal differentiation (Supplementary Fig. S5B and S5C). At this embryonic stage, dorsal root ganglia (DRG) can be clearly identified (39, 40). Although expression of AXL and EGFR is tenuous in these structures, Lztr1-deleted DRG had higher levels of AXL and EGFR than wild-type DRG (Supplementary Fig. S5D). To explore the consequences of targeted deletion of Lztr1 in the nervous system, we crossed Lztr1^fl/fl mice with the GFAP-Cre deleter strain. GFAP, a marker of neural stem cells and astrocytes in the central nervous system (CNS), is expressed by immature Schwann cells and neural crest–derived skin precursors that have been proposed as a cell of origin of peripheral nerve sheet tumors in humans and mice (41–43). Lztr1^fl/fl;GFAP-Cre mice were viable, exhibited minor neurologic symptoms such as abnormality of the hind limb reflex, and were tumor-free. We analyzed EGFR and AXL protein by immunofluorescence and found that they accumulated in Lztr1^fl/fl;GFAP-Cre mouse brain at levels higher than the wild-type littermates (Supplementary Fig. S6A and S6B). Although EGFR expression appeared widespread, expression of AXL accumulated in the area surrounding the lateral ventricles and the rostral migratory stream (Supplementary Fig. S6B). The absence of tumor formation in the nervous system of the Lztr1 conditional knockout mouse was in line with the absence of tumor phenotypes in mice harboring conditional deletion of individual tumor suppressor genes (44–46). To ask whether LZTR1 loss cooperates with inactivation of other tumor suppressor genes for tumor development in the nervous system, we generated Lztr1^fl/fl;Cdkn2A^fl/fl;GFAP-Cre compound mice. Loss-of-function germline mutations of CDKN2A occur in the germline of patients predisposed to tumors in the PNS. Somatic deletions of CDKN2A are frequently associated with progression to malignancy of PNS tumors and are the most frequent genetic alterations in malignant peripheral nervous system tumors (MPNST; refs. 47–52). Lztr1^fl/fl;Cdkn2A^fl/fl;GFAP-Cre mice exhibited significantly reduced survival compared with Lztr1^+/+;Cdkn2A^fl/fl;GFAP-Cre and control mice (Fig. 6A). The primary reason for lethality was tumor development in the PNS and hematopoietic lineage (Fig. 6B). Conversely, we did not observe tumors in the brain of Lztr1^fl/fl;Cdkn2A^fl/fl;GFAP-Cre mice. Approximately 50% of Lztr1^fl/fl;Cdkn2A^fl/fl;GFAP-Cre mice developed PNS tumors in the skin, soft tissues, and DRG that showed positivity for neural cell markers (Fig. 6C; Supplementary Fig. S6C). Approximately 30% of the PNS tumors exhibited histologic and immunophenotypic features compatible with schwannomas, including a tendency to form pseudopalisading structures, positivity for the Schwann cell markers S100β and SOX10, and diffuse positivity for calretinin, a neural protein recently included in the panel of markers differentiating schwannoma from neurofibroma in humans (Fig. 6C; Supplementary Fig. S6C; refs. 53, 54). The morphology and immunophenotyping of tumors not classified as schwannoma-like in Lztr1^fl/fl;Cdkn2A^fl/fl;GFAP-Cre mice confirmed their origin from cells from the PNS and broadly indicated transition toward more malignant and anaplastic lesions, similar to MPNSTs. EGFR and AXL accumulated at high levels in schwannoma-like and MPNST-like tumors (Fig. 6C; Supplementary Fig. S6D). Cells isolated from schwannoma-like tumors that developed in Lztr1^fl/fl;Cdkn2A^fl/fl;GFAP-Cre+ mice (E0954) were expanded ex vivo and retained high expression of the Schwann cell markers S100β and calretinin (Fig. 6D). They also exhibited elevated EGFR and AXL, which were downregulated following lentivirus-mediated reconstitution of LZTR1 (Fig. 6E), leading to reduced cell proliferation (Fig. 6F). In addition to tumors of the PNS, 36% of the Lztr1^fl/fl;Cdkn2A^fl/fl;GFAP-Cre mice developed splenomegaly and enlarged liver associated with myeloproliferative disorders spanning from extramedullary hematopoiesis to overt myelogenous leukemia (Supplementary Fig. S6E). The spectrum of myeloid tumors observed in Lztr1^fl/fl;Cdkn2A^fl/fl;GFAP-Cre mice is consistent with the hematopoietic expression of the Cre driver by the GFAP promoter (55) and the sensitivity of the hematopoietic lineage to LZTR1 mutations for tumor development in humans (3, 56).

The above findings suggest that EGFR and AXL may aberrantly accumulate in human tumors initiated by inactivation of LZTR1. To test this hypothesis, we compared EGFR and AXL protein levels in schwannomas from individuals carrying germline mutations of LZTR1 and affected by schwannomatosis, schwannomas from individuals carrying germline mutations of SMARCB1, and sporadic/nonsyndromic schwannoma. EGFR and AXL proteins were analyzed by IHC and quantified with the digital H-SCORE (D-H-SCORE), an accurate and reproducible method for quantitative immunostaining (57). EGFR and AXL accumulated in schwannomas from LZTR1 germline mutant schwannomatosis patients at levels significantly higher than those detected in SMARCB1-mutant or sporadic schwannomas (Fig. 6G and H). However, expression of RIT1, KRAS, and HRAS did not vary significantly between sporadic or LZTR1-mutant germline schwannoma (Supplementary Fig. S7A and S7B). To validate the finding that RTK signaling is enhanced in LZTR1 germ­line mutant schwannoma compared with LZTR1 wild-type schwannomatosis tumors, we analyzed the transcriptomic profiles from RNA-sequencing (RNA-seq) data of schwannomatosis-schwannoma from patients with germline mutations of LZTR1 (n = 10) and schwannomatosis-schwannoma from patients without LZTR1 germline mutations (n = 14; ref. 58). By performing differential gene-expression analysis and the robust Gene Set Test–Mann–Whitney–Wilcoxon (MWW) test statistics (59), we found that EGF-dependent RTK activation and RTK activities are significantly higher in tumors from patients with LZTR1 mutation compared with tumors harboring wild-type LZTR1 (Supplementary Fig. S7C and S7D). The independent analysis of single-sample enrichment scores for EGFR and RTK gene signatures confirmed that response to EGF and RTK activity and regulation are the biological pathways significantly increased in tumors from patients with LZTR1 germline mutations (Supplementary Fig. S7E and S7F). There was no significant difference in the occurrence of chromosome 22q loss of heterozygosity between the two groups (Fisher exact test P value = 1), suggesting that RTK pathway activation is unlikely to be linked to this genetic event (Supplementary Fig. S7C).

Finally, we asked whether the accumulation of RTKs in cells harboring loss of LZTR1 affected the proliferation potential and induced vulnerability to inhibition of EGFR and AXL kinase activity. For these experiments, we considered two EGFR inhibitors, afatinib and osimertinib, and the AXL-specific inhibitor bemcentinib (60). Afatinib is an irreversible pan-ERBB inhibitor that targets wild-type EGFR (61). Osimertinib binds to the EGFR kinase irreversibly by targeting the cysteine-797 residue in the ATP binding site via covalent bond formation (62). Although the inhibitory activity of osimertinib against EGFR^T790M mutation is well established in lung cancer, the inhibition of the wild-type EGFR kinase is still considerable (63–66). To evaluate target engagement by the two EGFR inhibitors in cells isolated from a schwannoma-like mouse tumor (E0954), we analyzed the phospohrylation of EGFR after treatment with different concentrations of osimertinib and afatinib. Both inhibitors decreased the phosphorylation of Tyr-1068 of EGFR, although afatinib showed a more pronounced block than osimetinib at the lowest concentrations (Supplementary Fig. S8A and S8B). When analyzed at a single dose for the effect on clonogenicity of schwannoma-like cells from Lztr1^fl/fl;Cdkn2A^fl/fl;GFAP-Cre⁺ mice (E0954), both afatinib and osimertinib as well as the AXL inhibitor bemcentinib were ineffective. However, treatment with each EGFR inhibitor in combination with bemcentinib severely reduced clonogenicity (Supplementary Fig. S8C–S8F). To validate this result, we determined the drug combination effects at multiple doses using the Bliss score. The E0954 schwannoma-like cells exhibited high sensitivity to combination treatment with afatinib or osimertinib and bemcentinib (Fig. 7A and B; Bliss synergistic score: 13.69 and 16.5, respectively). In contrast, the combination of osimertinib plus bemcentinib lacked a synergistic effect in cancer cells derived from a mouse model of FGFR3–TACC3 gene fusion (refs. 59, 67; Fig. 7C; Bliss score: −0.59), thus supporting the specificity of therapeutic synergy. The molecular synergy of osimertinib and bemcentinib was evident in the inhibition of the downstream signaling, with complete elimination of AKT phosphorylation and significant reduction of ERK phosphorylation by the drug combination compared with treatment with the individual drugs (Fig. 7D). We also sought to confirm whether EGFR and AXL inhibitors have synergistic effects in immortalized astrocytes isolated from Lztr1^fl/fl;Rosa-CreER mice either left untreated or treated with 4-OHT to delete the Lztr1 gene. Lztr1 deletion by 4-OHT treatment markedly increased colony formation (Supplementary Fig. S8G). Recombined and control cells were treated with the EGFR inhibitor osimertinib or afatinib and the AXL inhibitor bemcentinib individually and in combination. Coinhibition of EGFR and AXL with osimertinib or afatinib and bemcentinib even at nanomolar concentration enhanced the killing of cells lacking active LZTR1 (4-OHT-treated) without affecting the viability of wild-type astrocytes (ethanol-treated; Supplementary Fig. S8H and S8I). Finally, we examined the antitumor effects of EGFR and AXL inhibitors in vivo. In these experiments, we tested afatinib and osimertinib alone or in combination with bemcentinib. Mice were injected subcutaneously with schwannoma-like cells from Lztr1^fl/fl;Cdkn2A^fl/fl;GFAP-Cre+ mice (E0954) and treated with (i) bemcentinib at 50 mg/kg body weight, osimertinib at 5 mg/kg body weight, bemcentinib at 50 mg/kg body weight plus osimertinib at 5 mg/kg body weight, or DMSO as control; (ii) afatinib at 10 mg/kg body weight, bemcentinib at 50 mg/kg body weight, bemcentinib at 50 mg/kg body weight plus afatinib at 10 mg/kg body weight, or DMSO as control (5 days in/one day off for both treatment regimens). Survival was evaluated by the log-rank test. For both therapy regimens, single-drug treatments were indistinguishable from control (Fig. 7E–G). Combination treatment of afatinib and bemcentinib prolonged survival compared with single-drug and vehicle treatment (Fig. 7E; P = 0.0092). However, this treatment regimen was considerably toxic in the mouse and required significant dose reduction. The combination of osimertinib plus bemcentinib was more effective in reducing tumor growth and improving mouse survival compared with single-drug treatment and controls (Fig. 7F and G; P < 0.0001). Taken together, in vitro and in vivo experiments validated the combination of EGFR and AXL inhibitors for the treatment of LZTR1-mutant cancer.

Our study shows that LZTR1 is recruited to ligand-activated EGFR and AXL, which are directly ubiquitylated by CRL3^LZTR1 complexes and consequently targeted to the lysosome for proteolysis. Thus, LZTR1 controls the timing of termination of the signal generated by the two RTKs. In the absence of LZTR1, EGFR and AXL remain in the endosome compartment and are likely recycled to the plasma membrane. Remarkably, most LZTR1 mutations of the Kelch and BTB-BACK domains found in different sporadic cancer types and the germline of patients with schwannomatosis have lost the ability to destabilize EGFR and AXL. The effect of Kelch domain mutants is associated with loss of the interaction between the LZTR1 mutants and the RTK substrates, whereas the mutants in the BTB-BACK domain specifically impair the formation of a complex with CUL3 (68). Consistent with these findings, EGFR and AXL specifically accumulate in tumors from LZTR1-mutant schwannomatosis, resulting in deregulated downstream signaling. We also found that schwannoma-like tumor cells lacking the Lztr1 gene exhibited vulnerability to combinatorial inhibition of EGFR and AXL. This finding was confirmed in multiple cell systems in vitro and in vivo and was similarly reproduced when using different small-molecule inhibitors of EGFR.

The analysis of the Lztr1-deficient mouse model showed that constitutive homozygous deletion of Lztr1 is embryonically lethal at mid-gestation because of severe liver degeneration. Knockout embryos also exhibited expansion of the germinal areas of the telencephalon with reduction of the differentiated CP. Thus, the phenotypes induced by Lztr1 deficiency in the mouse differ from the pathologic traits that have been reported in mice harboring gain-of-function mutation of the LZTR1 substrate RIT1, which recapitulate the dysmorphic alterations in Noonan syndrome (20). Interestingly, conditional knockout of Lztr1 and Cdkn2a under the GFAP promoter, which is widely active in progenitor and glial cells in the CNS and PNS, did not trigger tumorigenesis in the brain but induced PNS tumors including schwannoma-like tumors and MPNST and generated sensitivity to developing myeloid neoplasms. These results demonstrated for the first time the tumor suppression function of LZTR1 in the whole organism in vivo and unmasked the striking sensistivity of cells of the PNS to neoplastic transformation by LZTR1 inactivation. Deregulation of EGFR and AXL signaling in LZTR1-mutant cells points to a broad level of complexity of the control of cell growth and survival because RTKs are also coupled to non–RAS-dependent pathways (69, 70). EGFR and AXL are among the most studied tumor-specific RTKs undergoing genetic alterations that generate oncogenic signals driving multiple cancer types. However, besides copy-number gain or activating mutations, posttranslational accumulation of the EGFR and AXL proteins is frequently observed in the absence of gene amplification (71–73). Furthermore, ligand-induced activation of these RTKs is a potent oncogenic mechanism in diverse cancer types (74–76). We suggest that the frequent loss-of-function alterations of the LZTR1 gene across multiple tumor types and in tumor predisposition syndromes increase RTK protein stability and promote tumor initiation and progression by activating distinct cancer hallmarks such as epithelial-to-mesenchymal transition and invasion (77, 78).

By simultaneously targeting EGFR and AXL for degradation, LZTR1 not only restrains EGFR but also blocks the parallel and redundant cancer signaling mediated by AXL. EGFR and AXL can be directly targeted by small-molecule inhibitors. The therapeutic effect of single treatment with the EGFR inhibitors afatinib and osimertinib or the AXL inhibitor bemcentininb was minimal. However, when used in combination, EGFR and AXL inhibitors exhibited specific antitumor effect against Lztr1-deficient immortalized astrocytes, PNS tumor cells, and tumor models in vivo, indicating that when treated with a single RTK inhibitor, the persistent activity of EGFR or AXL is sufficient to sustain survival and growth of LZTR1-mutant cells.

Overall, our data can be safely interpreted to indicate that the combinatorial inhibition of EGFR and AXL may represent a therapeutic opportunity for LZTR1-mutant tumors. The occurrence of loss-of-function alterations of LZTR1 in glioblastoma and the lack of effective therapies for patients with this tumor type underscore the relevance of follow-up studies with EGFR and AXL inhibitors targeting LZTR1-mutant glioblastoma.

HeLa (ATCC, CCL-2, RRID:CVCL_0030), HEK293T (ATCC, CRL-11268, RRID:CVCL_1926), SW10 (ATCC, CRL-2766, RRID:CVCL_6458), and U87-MG (HTB-14, RRID:CVCL_0022) cell lines were acquired through ATCC. Cell lines were cultured in DMEM supplemented with 10% fetal bovine serum (Sigma). Cells were routinely tested for Mycoplasma contamination using Mycoplasma Plus PCR Primer Set (Agilent) and were found to be negative. MEFs were isolated from embryonic day 13.5 (E13.5)-old LZTR1^+/+ and LZTR1^−/− embryos and astrocytes LZTR1^fl/fl-Rosa-CreER were isolated from postnatal day 5 pups according to published protocols (79, 80). Deletion of Lztr1 in astrocytes was achieved by treating cells with 350 nmol/L 4-hydroxytamoxifen for 4 days. EGF and epiregulin stimulation was performed by adding the ligand at the concentrations indicated in the individual figures after 16 hours of starvation from serum. Cells were transfected with Lipofectamine 2000 (Invitrogen) or calcium phosphate. Cells were transduced using lentiviral particles in a medium containing 4 μg/mL of polybrene (Sigma). Evaluation of protein half-life was performed by treating cells with 50 μg/mL of CHX for the indicated times and analyzed by Western blot. The half-life of EGFR and AXL was quantified by densitometry using ImageJ processing software (NIH, RRID:SCR_003070). Densitometry values were analyzed by Prism 6.0 (RRID:SCR_002798).

cDNAs for LZTR1 and AXL were amplified by PCR and cloned into the pCDNA3.1 or pLOC vectors in frame with FLAG or HA tag at the C-terminus. Expression of LZTR1 using a doxycycline-inducible system was achieved by cloning FLAG-tagged LZTR1 cDNA into pINDUCER-TR3G. EGFR-GFP was a gift from Alexander Sorkin (University of Pittsburgh School of Medicine; Addgene plasmid # 32751, RRID:Addgene_32751). GST-LZTR1 full-length and deletion mutants (LZTR1-1-425, LZTR1-426-840, LZTR1-426-640, LZTR1-641-840) and GST-EGFR deletion mutants (EGFR-1-620, EGFR-643-953, EGFR-643-1186, EGFR-953-1186) were cloned into pGEX-4T-3 and resulting plasmids verified by Sanger sequencing. EGFR kinase activity defective mutant, K721R and LZTR1 Kelch domain mutants were generated by site-directed mutagenesis using the QuickChange Site-Directed mutagenesis kit (Agilent) and verified by Sanger sequencing.

For CRISPR/Cas9-mediated LZTR1, we used a lentiviral vector (LV01, Sigma). sgRNA sequences are:

• control nontargeting: CGCGATAGCGCGAATATATT

• hLZTR1-1: CACCCACGAACTCGTCGCA (HSPD0000047303);

• hLZTR1-2: GACTTCGACCATAGCTGCT (HSPD0000047304);

• mLZTR1-1: CATGGAAGAGCCTCCCGCT (MMPD0000065735);

• mLZTR1-2: GACAACAACATTCGCAGTG (MMPD0000065733).

Lentiviral particles were produced by cotransfecting lentiviral vectors with pCMV-ΔR8.1 and pMD2.G plasmids into HEK293T cells as previously described (81). After lentiviral infection, cells were selected with puromycin, seeded in a 96-well plate at a density of 0.5 cells per well, and independent clones were isolated, expanded, and analyzed by Western blot for expression of the targeted genes.

For the global ubiquitylation enrichment study, immortalized LZTR1^fl/fl-Rosa-CreER astrocytes were treated with 4-OHT as described above to induce deletion of LZTR1. Cells were then infected with lentivirus for doxycycline-inducible reconstitution of LZTR1 and treated with doxycycline or vehicle to express LZTR1 in the absence or presence of MG132. Cells were lysed/homogenized by bead-beating in 9 mol/L urea and 100 mmol/L ammonium bicarbonate, supplemented with protease inhibitors. Lysates were cleared by centrifugation at 21,000 × g for 30 minutes at 4°C, and protein concentration was measured by BCA. Total protein (5 mg) from each sample was reduced with 10 mmol/L DTT for 30 minutes at 56°C in an air thermostat, cooled down to room temperature, and alkylated with 11 mmol/L CAA at room temperature for 30 minutes, and alkylation was then quenched by the addition of an additional 5 mmol/L DTT. Samples were diluted 6-fold with 50 mmol/L ammonium bicarbonate and digested overnight with trypsin (1:50) at 37°C. The next day, the digestion was stopped by the addition of 0.25% TFA (final v/v) and centrifuged at 10,000 rpm for 10 minutes at room temperature to pellet precipitated lipids, and the cleared supernatant was collected. Supernatants were desalted on a SepPak C18 cartridge 500 mg and dried using vacuum centrifugation. Desalted dried peptides (5 mg) were resuspended in 1.4 mL of ice-cold IAP buffer (50 mmol/L MOPS (pH 7.2), 10 mmol/L sodium phosphate, and 50 mmol/L NaCl) and centrifuged at maximum speed for 5 minutes at 4°C to remove any insoluble material. Supernatants (pH ∼7.5) were incubated with the washed PTMScan Ubiquitin Remnant Motif (K-ε-GG) Kit #5562 (Cell Signaling Technology) antibody beads for 2 hours at 4°C. After centrifugation at 2,000 × g for 1 minute, beads were washed three times with ice-cold IAP buffer and three with ice-cold HPLC water. The ubiquitinated peptides were eluted twice with 0.15% TFA, desalted using SDB-RP StageTip, and dried via vacuum centrifugation. Desalted peptides were injected in an EASY-Spray PepMap RSLC C18 50 cm × 75 cm ID column (Thermo Scientific) connected to an Orbitrap Fusion Tribrid (Thermo Scientific). Peptide elution and separation were achieved at a nonlinear flow rate of 250 nL/minute using a gradient of 5% to 30% of buffer B (0.1% (v/v) formic acid, 100% acetonitrile) for 110 minutes with the temperature of the column maintained at 50°C during the entire experiment. The Thermo Scientific Orbitrap Fusion Tribrid mass spectrometer was used for peptide tandem mass spectroscopy (MS/MS). Survey scans of peptide precursors were performed from 375 to 1,500 m/z at 120K full-width at half maximum (FWHM) resolution (200 m/z) with a 2 × 10⁵ ion count target and a maximum injection time of 50 ms. The instrument was set to run in top speed mode with 3-second cycles for the survey and the MS/MS scans. After a survey scan, MS/MS was performed on the most abundant precursors, i.e., those exhibiting a charge state from 2 to 6 of greater than 5 × 10³ intensity, by isolating them in the quadrupole at 1.6 Th. We used collision-induced dissociation (CID) with 35% collision energy and detected the resulting fragments with the rapid scan rate in the ion trap. The automatic gain control target for MS/MS was set to 1 × 10⁴ and the maximum injection time was limited to 35 ms. The dynamic exclusion was set to 15 s with a 10 ppm mass tolerance around the precursor and its isotopes. Monoisotopic precursor selection was enabled. Raw mass spectrometric data were analyzed using the MaxQuant environment v.1.6.1.0 (82) and Andromeda for database searches (83) at default settings with few modifications. The default was used for first search tolerance and main search tolerance (20 ppm and 6 ppm, respectively). MaxQuant was set up to search with the reference mouse proteome database downloaded from UniProt. MaxQuant performed the search trypsin digestion with up to 2 missed cleavages. Peptide, site, and protein false discovery rates (FDR) were all set to 1%. The following modifications were used for protein identification and quantification: oxidation of methionine (M), acetylation of the protein N-terminus, DiGly (K), and deamination for asparagine or glutamine (NQ). Results obtained from MaxQuant were filtered by localization probability of ubiquitination site > 0.90; intensity of diGly peptide > median; summed score for the individual peptides > 15th percentile; posterior error probability < 0.05; intensity fold change between samples with the expression of LZTR1 and control > 2.5.

To identify the proteins that accumulated following LZTR1 loss, immortalized LZTR1 wild-type and LZTR1^−/− MEFs or HeLa cells were collected in cold PBS and centrifuged at 4,000 RPM for 5 minutes and processed as described (84). Briefly, frozen cell pellets were lysed by bead-beating in 8 M urea and 200 mmol/L EPPS (pH 8.5), supplemented with protease inhibitors. Samples were reduced with 5 mmol/L TCEP and alkylated with 10 mmol/L iodoacetamide (IAA) that was quenched with 10 mmol/L DTT. A total of 200 μg of protein was chloroform−methanol precipitated. Protein was reconstituted in 200 mmol/L EPPS (pH 8.5) and digested by Lys-C overnight and trypsin for 6 hours, both at a 1:50 protease-to-peptide ratio. Digested peptides were quantified using a Nanodrop and 100 μg from each sample was labeled with 800 μg TMT reagent using 10-plex TMT kit1. TMT labels were checked, 0.5 μg of each sample was pooled, desalted, and analyzed by short SPS-MS3 method, and using normalization factor, samples were bulk mixed at 1:1 across all channels. 900 μg of the bulk mixed sample was used for total proteome analysis. Mixed TMT-labeled samples were vacuum centrifuged and desalted with C18 Sep-Pak (200 mg) solid-phase extraction column. The desalted sample was fractionated using BPRP chromatography. Peptides were subjected to a 50-minute linear gradient from 5% to 42% acetonitrile in 10 mmol/L ammonium bicarbonate pH 8 at a flow rate of 0.6 mL/minutes over Water X-bridge C18 column (3.5 μm particles, 4.6 mm ID, and 250 mm in length). The peptide mixture was fractionated into a total of 96 fractions, which were consolidated into 36 fractions. Fractions were subsequently acidified with 1% formic acid, and vacuum centrifuged to near dryness and desalted via SDB-RP StageTip. Fractions were dissolved in 10 μL of 3% acetonitrile/0.1% formic acid injected using SPS-MS3. The UltiMate 3000 UHPLC system (Thermo Scientific) and EASY-Spray PepMap RSLC C18 50 cm × 75 μmol/L ID column (Thermo Scientific) coupled with Orbitrap Fusion (Thermo Scientific) were used to separate fractioned peptides with a 5% to 30% acetonitrile gradient in 0.1% formic acid over 45 minutes at a flow rate of 250 nL/minute. After each gradient, the column was washed with 90% buffer B for 10 minutes and reequilibrated with 98% buffer A (0.1% formic acid, 100% HPLC-grade water) for 40 minutes. The full MS spectra were acquired in the Orbitrap Fusion Tribrid Mass Spectrometer (Thermo Fisher Scientific) at a resolution of 120,000. The 10 most intense MS1 ions were selected for MS2 analysis. The isolation width was set at 0.7 Da and isolated precursors were fragmented by CID at normalized collision energy (NCE) of 35% and analyzed in the ion trap using “turbo” scan speed. Following the acquisition of each MS2 spectrum, a synchronous precursor selection (SPS) MS3 scan was collected on the top 10 most intense ions in the MS2 spectrum. SPS-MS3 precursors were fragmented by higher energy collision-induced dissociation at an NCE of 65% and analyzed using the Orbitrap. Raw mass spectrometric data were analyzed using Proteome Discoverer 2.3 to perform database search and TMT reporter ion quantification. TMT tags on lysine residues and peptide N termini (+229.163 Da) and the carbamidomethylation of cysteine residues (+57.021 Da) were set as static modifications, whereas the oxidation of methionine residues (+15.995 Da), and deamidation (+0.984) on asparagine and glutamine were set as a variable modification. Data were searched against a UniProt Mouse or human database with peptide-spectrum match (PSM) and protein-level FDR at 1% FDR. The signal-to-noise (S/N) measurements of each protein were normalized so that the sum of the signal for all proteins in each channel was equivalent to account for equal protein loading. Protein identification and quantification were analyzed using the R software environment. Proteins were selected according to >1.5-fold change between LZTR1 knockout and wild-type samples.

LZTR1 complexes were purified from the U87 cells transduced with lentivirus expressing LZTR1-FLAG, LZTR1-FLAG-R810W, or the empty vector. Cellular lysates were prepared in 50 mmol/L Tris-HCl, pH 7.5, 150 mmol/L NaCl, 1 mmol/L EDTA, 1% NP40, 0.5% sodium deoxycholate, 0.1% sodium dodecyl sulfate, 1.5 mmol/L Na3VO4, 50 mmol/L sodium fluoride, 10 mmol/L sodium pyrophosphate, 10 mmol/L β-glycerolphosphate, and EDTA-free protease inhibitor cocktail (Roche). Lysates were immunoprecipitated with FLAG affinity matrix (Sigma-Aldrich, cat. # F2426, RRID:AB_2616449) and eluted with FLAG peptide. Eluates were separated on 4% to 12% gradient SDS-PAGE and stained with SimplyBlue (Thermo Fisher Scientific). Protein gel slices were excised and in-gel digestion was performed as previously described (85), with minor modifications. Gel slices were washed with 1:1 acetonitrile and 100 mmol/L ammonium bicarbonate for 30 minutes and then dehydrated with 100% acetonitrile for 10 minutes until shrunk. The excess acetonitrile was then removed, and the slices were dried in a speed vacuum at room temperature for 10 minutes. Gel slices were reduced with 5 mmol/L DTT for 30 minutes at 56°C in an air thermostat, cooled down to room temperature, and alkylated with 11 mmol/L IAA for 30 minutes with no light. Gel slices were then washed with 100 mmol/L of ammonium bicarbonate and 100% acetonitrile for 10 minutes each. Excess acetonitrile was removed and dried in a speed vacuum for 10 minutes at room temperature, and the gel slices were rehydrated in a solution of 25 ng/μL trypsin in 50 mmol/L ammonium bicarbonate for 30 minutes on ice and digested overnight at 37°C in an air thermostat. Digested peptides were collected and further extracted from gel slices in an extraction buffer (1:2 ratio by volume of 5% formic acid: acetonitrile) at high speed, shaken in an air thermostat. The supernatants from both extractions were combined and dried in a speed vacuum. Peptides were dissolved in 3% acetonitrile/0.1% formic acid. Desalted peptides were injected in an EASY-Spray PepMap RSLC C18 50 cm × 75 cm ID column (Thermo Scientific) connected to an Orbitrap Fusion Tribrid (Thermo Scientific) as described in the “Global ubiquitylation enrichment” section above. The following modifications were used for protein identification and quantification: oxidation of methionine (M), acetylation of the protein N-terminus, and deamination for asparagine or glutamine (NQ). Results obtained from MaxQuant were assembled in Scaffold for data visualization. A specificity score of proteins interacting with LZTR1 proteins was computed for each polypeptide as described (86). Briefly, we compared the number of peptides identified from our mass spectrometry analysis to those reported in the CRAPome database that includes a list of potential contaminants from affinity purification-mass spectrometry experiments (www.crapome.org). The specificity score is computed as [(#peptide*#xcorr)/(AveSC*MaxSC*# of Expt.)], where #peptide, identified peptide count; #xcorr, the cross-correlation score for all candidate peptides queried from the database; AveSC, averaged spectral counts from CRAPome; MaxSC, maximal spectral counts from CRAPome; and # of Expt., the total found number of experiments from CRAPome. Finally, proteins were filtered by PSM = 0 in the empty vector (EV) control; PSM > 0 in the cells expressing wild-type (WT) and R810W-mutant (RW) LZTR1; RW-LZTR1 vs. WT-LZTR1 PSM fold change > 2.

PGEX 4T-3-EGFR and PGEX 4T-3-LZTR1 full-length or deletion mutant plasmids were introduced into E. coli strain BL21 (DE3). Protein expression was induced with 200 μmol/L isopropyl-β-D-1-thiogalactopyranoside (IPTG) for 4 to 5 hours at 30°C. Bacteria were harvested and resuspended in lysis buffer containing 1× phosphate-buffered saline, 0.5% Triton X-100, 1× protease inhibitors cocktail (Roche), 1 mmol/L PMSF, and 200 μg/mL lysozyme, and incubated on ice for 30 minutes. After centrifugation at 17,000 × g for 15 minutes at 4°C, supernatants were incubated with Glutathione S Sepharose beads (Cytiva) for 1 hour at 4°C, washed 3 times in lysis buffer and stored in PBS.

Cells were lysed in RIPA buffer (50 mmol/L Tris-HCl, pH 7.5, 150 mmol/L NaCl, 1 mmol/L EDTA, 1% NP40, 0.5% sodium dexoycholate, 0.1% sodium dodecyl sulfate, 1.5 mmol/L Na₃VO₄, 50 mmol/L sodium fluoride, 10 mmol/L sodium pyrophosphate, 10 mmol/L β-glycerolphosphate and EDTA-free protease inhibitor cocktail; Roche) or NP40 lysis buffer (50 mmol/L Tris-HCl, pH 7.5, 150 mmol/L NaCl, 1 mmol/L EDTA, 1% NP40, 1.5 mmol/L Na₃VO₄, 50 mmol/L sodium fluoride, 10 mmol/L sodium pyrophosphate, 10 mmol/L β-glycerolphosphate and EDTA-free protease inhibitor cocktail; Roche). Lysates were centrifuged 14,000 rpm for 15 minutes at 4°C and supernatant was collected. For immunoprecipitation, cell lysates were incubated with the antibody for GFP (Santa Cruz Biotechnology, cat. #sc-9996, RRID:AB_627695) or EGFR (Santa Cruz Biotechnology, cat. #sc-373746, RRID:AB_10920395) and protein G/A beads (Santa Cruz Biotechnology, cat. #sc-2003, RRID:AB_10201400) or FLAG M2 affinity gel (Sigma-Aldrich, cat. #F2426, RRID:AB_2616449) and HA affinity gel (Sigma-Aldrich, cat. #E6779, RRID:AB_10109562) at 4°C overnight. Beads were washed with lysis buffer four times and eluted in 2× SDS sample buffer. Immunoprecipitates or lysates were separated by SDS-PAGE and transferred to polyvinylidene difluoride or nitrocellulose membranes. Membranes were blocked in TBS-T (0.1% Tween20) with 5% nonfat milk or BSA and probed with primary antibodies.

Antibodies and working concentrations are: GFP (1:1000; Santa Cruz Biotechnology, cat. #sc-9996, RRID:AB_627695); LZTR1 (1:500; Santa Cruz Biotechnology, cat. #sc-390166, RRID:AB_2910196); EGFR (1:500; Santa Cruz Biotechnology, cat. #sc-373746, RRID:AB_10920395); AXL (1:1000; Santa Cruz Biotechnology, cat. #sc-166269, RRID:AB_2243305); c-MYC (1:500; Santa Cruz Biotechnology, cat. #sc-40, RRID:AB_627268); GST (1:1,000; Santa Cruz Biotechnology, cat. #sc-138, RRID:AB_627677); ERK (1:2,000; Cell Signaling Technology, cat. #9102, RRID:AB_330744); phospho-ERK (1:2,000; Cell Signaling Technology, cat. #4370, RRID:AB_2315112); phospho-EGFR (1:1,000; Cell Signaling Technology, cat. #3777, RRID:AB_2096270); AXL (1:2,000; Cell Signaling Technology, cat. #8661, RRID:AB_11217435); phospho-AXL (1:500; Cell Signaling Technology, cat. #5724, RRID:AB_10544794); DYKDDDDK Tag (1:1,000; Cell Signaling Technology, cat. #14793, RRID:AB_2572291); β-actin (1:10,000; Sigma-Aldrich, cat. #A5441, RRID:AB_476744); mono- and polyubiquitinilated-HRP (1:1,000; Enzo Life Sciences, cat. #BML-PW0150-0025, RRID:AB_2051892); EGFR (1:2000; Abcam, cat. #ab52894, RRID:AB_869579); RIT1 (1:500; Abcam, cat. #ab53720, RRID:AB_882379); HA (1:1,000; Roche, cat. #12158167001, RRID:AB_390915); AXL (1:1,000; R&D Systems, cat. #AF854, RRID:AB_355663). Horseradish peroxidase-conjugated secondary antibodies were purchased from Invitrogen and ECL (Amersham) or Super Signal West Femto (Thermo Scientific) was used for detection.

HEK293T and HeLa cells were transfected with a plasmid expressing EGFR-GFP, HA-Ub, and LZTR1-FLAG as indicated in figures and treated with MLN4924 (1 μmol/L), CQ (100 μmol/L), MG132 (20 μmol/L), or EGF (20 ng/mL or 100 ng/mL) and ubiquitylation assay was performed under denaturing condition. Cells were lysed in 1% SDS and boiled at 100°C for 10 minutes. Lysates were diluted with tris buffered saline containing 1% NP40 and centrifuged 14,000 rpm for 15 minutes at 4°C. Immunoprecipitation was performed using 500 μg to 1 mg of cellular lysates using HA affinity gel (Sigma-Aldrich, cat. #E6779, RRID:AB_10109562), AXL (Santa Cruz Biotechnology, cat. #sc-166269, RRID:AB_2243305), or EGFR (Santa Cruz Biotechnology, cat. #sc-373746, RRID:AB_10920395) antibodies at 4°C overnight followed by protein G/A beads for 90 minutes at 4°C. Beads were washed with lysis buffer four times and eluted in 2 × SDS sample buffer. Protein samples were separated by SDS-PAGE and transferred to the PVDF membrane and analyzed by Western blot.

Cells were fixed with 4% paraformaldehyde in PBS for 10 minutes, washed with cold PBS three times, and permeabilized and blocked with 0.02% saponin or 0.5% Triton X-100 and 5% BSA in PBS for 1 hour. Cells were incubated with antibodies; EEA1 antibody (1:200; BD Biosciences, cat. #610457, RRID:AB_397830); LAMP1 antibody (1:500; Abcam, cat. #ab24170, RRID:AB_775978); LAMP1 (1:700; Santa Cruz Biotechnology, cat. #sc-20011, RRID:AB_626853); AXL (1:1,000; Cell Signaling Technology, cat. #8661, RRID:AB_11217435); AXL (1:400; Santa Cruz Biotechnology, cat. #sc-166269, RRID:AB_2243305); FLAG (1:1,000; Sigma-Aldrich, cat. #F1804, RRID:AB_262044) overnight at 4°C or 1 hour at room temperature. Cells were washed with 0.02% Saponin in PBS for three times and incubated with fluorescence-conjugated secondary antibodies; goat anti-rabbit IgG (H+L) cross-adsorbed secondary antibody, Cyanine3 (Thermo Fisher Scientific, cat. #A10520, RRID:AB_253402); Cy3 AffiniPure donkey anti-goat IgG (H + L; Jackson ImmunoResearch Labs, cat. #705-165-147, RRID:AB_2307351); goat anti-mouse IgG (H+L), superclonal recombinant secondary antibody, Alexa-Fluor 555 (Thermo Fisher Scientific, cat. #A28180, RRID:AB_2536164); goat anti-rabbit IgG (H+L) highly cross-adsorbed secondary antibody, Alexa-Fluor 647 (Thermo Fisher Scientific, cat. #A-21245, RRID:AB_2535813); and goat anti-mouse IgG (H+L) cross-adsorbed secondary antibody, Alexa-Fluor 647 (Thermo Fisher Scientific, cat. #A-21235, RRID:AB_2535804) for 30 minutes at room temperature. After three washes, DNA was counterstained with DAPI (Sigma). Fluorescence microscopy was performed on a Nikon A1R MP microscope using a 100×, 1.45 Plan Apo Lambda lens or Olympus IX70 microscope with 40× objective. Image analysis was performed using ImageJ (RRID:SCR_003070). At least 30 cells per sample (4–10 cells per field) were randomly selected and colocalization between EGF or AXL and EEA1 or LAMP1 was quantified using the ImageJ plug-in JACoP.

Human and mouse tissue preparation and immunostaining were performed as previously described (86, 87). Human schwannoma samples analyzed by IHC had been stored in the Onconeurotek tumorbank (certified NF S96 900) and received the authorization for analysis from the ethical committee (CPP Ile de France VI, ref A39II), and French Ministry for research (AC 2013-1962). Briefly, tumor sections were deparaffinized in xylene and rehydrated in a graded series of ethyl alcohol. Antigen retrieval was performed in citrate solution pH = 6.0 using a decloaking chamber. After peroxidase blocking in 3% H₂O₂ for 15 minutes, slides were blocked for 1 hour in 10% goat serum, 0.25% Triton X-100, 1× PBS and then incubated at 4°C overnight with antibodies; EGFR antibody (1:2,500; Abcam, cat. #ab52894, RRID:AB_869579); AXL antibody (1:250; Cell Signaling Technology, cat. #8661, RRID:AB_11217435); RIT1 (1:500; Abcam, cat. #ab53720, RRID:AB_882379); KRAS (1:800; Proteintech, cat. #12063–1-AP, RRID:AB_878040); HRAS (1:150; Novus Biologicals, cat #NBP2–42864); S100B (1:400; Abcam, cat. #ab52642, RRID:AB_882426); calretinin (1:400; Abcam, ab244299); SOX10 (1:500; Abcam, ab227680); AXL (1:2,000; R&D Systems, cat. #AF854, RRID:AB_355663). Sections were then incubated with biotinylated anti-rabbit or anti-mouse antibody followed by streptavidin–peroxidase. Reaction for schwannoma tissue sections was developed by 3,3′-diaminobenzidine (DAB) and counterstained with hematoxylin. Images were acquired under 10× magnification using an Olympus 1 × 70 microscope equipped with a digital camera section (6–25 images/section). Quantification of digital H-SCORE was obtained using ImageJ (NIH) with specific built-in color deconvolution plug-in for evaluation of hematoxylin and DAB staining (57). For mouse tumor immunofluorescence, TSA-Cy3 was used (Akoya Biosciences) and nuclei were counterstained with DAPI (Sigma). Semiquantitative IHC analysis of PNS tumors was performed by scoring cells with positive signal for each marker according to the estimated percentage of positive cells: +++: positive cells ≥50% (diffuse); ++: positive cells between 25% and 50% (diffuse); +: positive cells between 10% and 25% (diffuse); +/−: positive cells between 10% and 25% (only in some fields with other fields negative); −: negative.

We obtained heterozygous Lztr1^{tm1a(EUCOMM)Wtsi} mice through the European Community Mouse Mutagenesis consortium (38, 88). Lztr1^tm1a heterozygous mice are phenotypically normal. The Lztr1-deficient mice (LZTR1^{tm1a(EUCOMM)Wtsi}) carry a knockout-first allele, in which a cassette including LacZ and neo genes were inserted at position 17,518,027 of Chr 16 (intron 5–6) of the Lztr1 gene. The cassette includes an FRT site followed by a lacZ sequence and a loxP site. This first loxP site is followed by neomycin under the control of the human beta-actin promoter, SV40 polyA, a second FRT site, and a second loxP site. A third loxP site is inserted downstream of the targeted exon 7 at position 17,518,811. LoxP sites thus flank exon 7. We generated the “conditional” (floxed) allele by intercrossing them with ROSA26^Sortm1(FLP1) mice. Lztr1^fl/flmice were then intercrossed with Cdkn2a^fl/fl and GFAP-Cre mice to generate Lztr1^fl/fl;Cdkn2a^fl/fl;GFAP-Cre⁺ mice or with Rosa-cre-ER mice to obtain Lztr1^fl/fl;Rosa-cre-ER mice from which astrocytes were isolated.

For in vivo treatment experiments, 1 × 10⁶ tumor cells from Lztr1^fl/fl;Cdkn2a^fl/fl;GFAP-Cre⁺ mice were injected subcutaneously in the right flank of female and male nu/nu mice in 150 μL volume of cell cultured media with 50% Matrigel. Mice carrying 250 to 300 mm³ subcutaneous tumors (approximately 16 days from injection) were treated with (i) bemcentinib at 50 mg/kg body weight, osimertinib at 5 mg/kg body weight, bemcentinib at 50 mg/kg body weight plus osimertinib at 5 mg/kg body weight, or DMSO as control; (ii) afatinib at 10 mg/kg bodyweight, bemcentinib at 50 mg/kg body weight, bemcentinib at 50 mg/kg body weight plus afatinib at 10 mg/kg body weight, or DMSO as control (5 days on/one day off for both treatment regimens). Mice were weighed daily. Tumor diameters were measured every 2 days with a caliper, and tumor volumes were estimated using the formula: width² × length/2 = V (mm³). Mice were euthanized when the tumor size reached 2,000 mm³ or if the skin was ulcerated, according to Institutional Animal Care and Use Committee (IACUC) recommendation.

All experiments using mouse models were approved by the IACUC at Columbia University and performed according to IACUC guidelines and regulations.

Astrocytes were treated with 350 nmol/L 4-hydroxytamoxifen or ethanol for 4 days. Cells (1,000) were plated in 6-well plates and cultured in DMEM/F12 medium supplemented with N-2, B-27, EGF, and FGF2 for 2 weeks. Cells were stained with crystal violet and colonies were counted.

Five hundred cells were seeded in triplicate wells of 6-well plates. Twenty-four hours after seeding, cells were treated with osimertinib, afatinib, and bemcentinib for 72 hours as indicated in Supplementary Fig. S8C and S8E. Cells were washed to remove the drugs and cultured in a growth culture medium until colonies were detected by light microscopy. Cells were stained with crystal violet and colonies were counted.

Astrocytes were treated with 350 nmol/L 4-hydroxytamoxifen or ethanol for 4 days and then cultured in DMEM/F12 medium supplemented with N-2, B-27, EGF, and FGF2 for 4 additional days. Cells isolated from tumors occurring in Lztr1^fl/fl;Cdkn2a^fl/fl;GFAP-Cre⁺ mice were cultured in the medium as described above. FGFR3–TACC3 gene fusion–harboring cells were similarly cultured. Cells were plated at 4,000 cells/well in 130 μL of the medium in opaque white 96-well plates, and after 24 hours, cells were treated in six replicates with 3-fold serial dilutions of osimertinib or afatinib and bemcentinib drug combinations as indicated in Fig. 7; Supplementary Fig. S8H and S8I for 96 hours. Viability was determined using CellTiter-Glo assay reagent (Promega, G7570) and GloMax-Multi+ Microplate Multimode Reader (Promega). Experiments were repeated three times with similar results. The Bliss score was obtained using SynergyFinder (https://synergyfinder.fimm.fi; ref. 89).

Gene-expression RNA-seq data of 24 Schwannomatosis-related schwannoma and their genetic annotations were retrieved from a published study (58). RNA-seq raw data were mapped to human reference (hg19) using STAR (90), and featureCounts (91) was used for transcript level quantification. Differential gene-expression analysis was performed to compare LZTR1 germline mutant (n = 10) and WT (n = 14) samples, using EDAseq R package (92). Functional enrichment was analyzed by the MWW–Gene Set Test (59) to identify significantly overrepresented pathways in the LZTR1-mutant subset compared with the WT, and to compute the normalized enrichment score (NES) in each single sample (Gene Set Test–MWW NES > 0.58 and FDR < 0.05).

Results in graphs are expressed as means ± SD as indicated in figure legends, for the indicated number of observations. Statistical significance was determined by the Student t test (two-tailed, unequal variance) using GraphPad Prism 8.0 software package (GraphPad Inc. RRID:SCR_002798) or statistical functions in Excel. P < 0.05 is considered significant and is indicated in figure legends.

The data generated in this study are available within the article and its supplementary data files.

F. Bielle reports grants from AbbVie, nonfinancial support from Bristol Myers Squibb, and other support from Bristol Myers Squibb outside the submitted work. M. Eoli reports grants from the Italian Ministry of Health (RRC) during the conduct of the study. M. Sanson reports personal fees from Genenta outside the submitted work. A. Iavarone reports a planned patent application on this work pending. No disclosures were reported by the other authors.

A. Ko: Conceptualization, data curation, funding acquisition, investigation, methodology, writing–original draft. M. Hasanain: Methodology, investigation. Y. Oh: Methodology, investigation. F. D'Angelo: Data curation. D. Sommer: Methodology. B. Frangaj: Methodology. S. Tran: Resources. F. Bielle: Resources. B. Pollo: Resources. R. Paterra: Resources. K. Mokhtari: Resources. R.K. Soni: Methodology. M. Peyre: Resources. M. Eoli: Resources. L. Papi: Resources. M. Kalamarides: Resources. M. Sanson: Resources. A. Iavarone: Conceptualization, supervision, funding acquisition, writing–original draft. A. Lasorella: Conceptualization, data curation, supervision, funding acquisition, investigation, writing–original draft.

We thank the Proteomics and Macromolecular Crystallography Shared Resource, Herbert Irving Comprehensive Cancer Center for the proteomics experiments. This work was supported by NIH R01CA101644, U54CA193313, R01CA131126, and R01CA239721 to A. Lasorella; U54CA193313, R01CA179044, R01CA190891, and a Synodos for Schwannomatosis Consortium program (CTF-2018-18-005B) by the Children's Tumor Foundation to A. Iavarone; a fellowship from the Children's Tumor Foundation to A. Ko (CTF-2020-01-007); and INCa-DGOS-INSERUM_12560 (SiRIC CURAMUDS) and the Ligue Nationale contre le Cancer (LNCC; Equipe labellisée) to M. Sanson. A. Ko, A. Iavarone, and A. Lasorella are inventors of a patent application based on this work.

The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.