EGFR Regulates the Hippo pathway by promoting the tyrosine phosphorylation of MOB1

The Hippo pathway is frequently dysregulated in cancer, leading to the unrestrained activity of its downstream targets, YAP/TAZ, and aberrant tumor growth. However, the precise mechanisms leading to YAP/TAZ activation in most cancers is still poorly understood. Analysis of large tissue collections revealed YAP activation in most head and neck squamous cell carcinoma (HNSCC), but only 29.8% of HNSCC cases present genetic alterations in the FAT1 tumor suppressor gene that may underlie persistent YAP signaling. EGFR is overexpressed in HNSCC and many other cancers, but whether EGFR controls YAP activation is still poorly understood. Here, we discover that EGFR activates YAP/TAZ in HNSCC cells, but independently of its typical signaling targets, including PI3K. Mechanistically, we find that EGFR promotes the phosphorylation of MOB1, a core Hippo pathway component, and the inactivation of LATS1/2 independently of MST1/2. Transcriptomic analysis reveals that erlotinib, a clinical EGFR inhibitor, inactivates YAP/TAZ. Remarkably, loss of LATS1/2, resulting in aberrant YAP/TAZ activity, confers erlotinib resistance on HNSCC and lung cancer cells. Our findings suggest that EGFR-YAP/TAZ signaling plays a growth-promoting role in cancers harboring EGFR alterations, and that inhibition of YAP/TAZ in combination with EGFR might be beneficial to prevent treatment resistance and cancer recurrence.

T he Hippo pathway is a tumor-suppressive signaling route and its downstream targets, Yes-associated protein (YAP) and transcriptional co-activator with PDZ binding motif (TAZ), play a central role in normal tissue growth and organ size 1 . In mammals, the core Hippo kinase pathway consists of mammalian STE20-like kinase 1 and 2 (MST1/2), large tumor suppressor 1 and 2 (LATS1/2), and their adaptor proteins salvador homologue 1 (SAV1) and MOB kinase activator 1A and 1B (MOB1A/B, hereafter MOB1), respectively 2 . MST1/2 phosphorylate the hydrophobic motif of LATS1/2, and subsequently activated LATS1/2 phosphorylate YAP on multiple serine residues (pYAP), leading to its cytoplasmic retention by binding to 14-3-3 and/or degradation through the ubiquitin-proteasome pathway 3 . In the absence of Hippo pathway signaling, LATS1/2 inactivation results in nuclear translocation of hypo-phosphorylated YAP and its interaction with transcription factors including TEA domain family members (TEAD) to enhance the transcription of growth-related genes 4 . YAP/TAZ are aberrantly activated in many types of cancer 5 , including head and neck squamous cell carcinomas (HNSCC), a disease that is diagnosed in around 65,410 new cases each year in the United States alone, resulting in more than 14,620 deaths 6 .
The mechanisms resulting in YAP/TAZ activation in most cancer types are still poorly understood. Specifically for HNSCC, The Cancer Genome Atlas (TCGA) has provided a comprehensive landscape of somatic genomic alterations in this cancer type 7 , which revealed that HNSCC is among the cancers showing the highest incidence of YAP1 gene amplification (6.3% of the cases). In addition, our recent study has uncovered that HNSCCs have frequent alterations of FAT1 (29.8%), which results in YAP activation and its consequent YAP-dependent tumor growth 8 . FAT1 assembles a multimeric Hippo pathway signaling complex, inducing activation of core Hippo kinases by TAO kinases resulting in YAP inactivation 8 . However, it is still possible that other molecular events may control YAP activation in >65% of HNSCC cases that do not exhibit YAP1 or FAT1 genomic alterations, whose elucidation may help reveal new molecular mechanisms controlling the Hippo pathway in cancer.
In this regard, EGFR, one of the ERBB family tyrosine kinases, is amplified and highly overexpressed in HNSCC and lung squamous cell carcinoma, and mutated and activated in many cancer types including lung adenocarcinoma and glioblastoma 7,[9][10][11] . Therefore, EGFR is a widely accepted therapeutic target, either using small molecule tyrosine kinase inhibitors (e.g., erlotinib in lung adenocarcinoma) or blocking antibody (e.g., cetuximab in HNSCC). The link between EGFR activation and the Hippo pathway is still poorly understood 12 , with EGFR failing to reduce the phosphorylation of YAP at S127 and nuclear localization in some cellular systems 13,14 , but inhibiting the Hippo pathway to activate YAP in others [15][16][17][18] .
Here, we show that EGFR activation leads to the phosphorylation of one of the core Hippo pathway components, MOB1 to inhibit LATS1/2 function thus resulting in YAP/TAZ activation in HNSCC cells independent of FAT1 alterations. Remarkably, EGFR-targeting therapies suppress YAP/TAZ, and loss of LATS1/ 2-mediated YAP/TAZ activation confers therapy resistance. These findings contribute to the understanding of the mechanisms by which EGFR-driven signaling networks control YAP/ TAZ activation in normal cells and cancer, and support the therapeutic potential of inhibiting YAP/TAZ function in patients with cancers harboring EGFR alterations to enhance the response to EGFR targeted therapies, and prevent emergence of drug resistance.

Results
EGFR activates YAP/TAZ in HNSCC cells, independently of PI3K. We have recently reported that frequent FAT1 alterations contribute to YAP activation in HNSCC, however many FAT1 wild type HNSCC cases also exhibit nuclear YAP 8 , and as such, the mechanism of YAP activation in HNSCC, and other cancer types, may not yet be fully understood. As a potential upstream activating component, we focused on EGFR, because it is overexpressed or amplified in most HNSCC cases 7 , and the target of the only approved cancer-targeting therapy in this malignancy 19,20 . We first compared EGFR expression and YAP activation among HNSCC cell lines including CAL33 that harbors hemizygous FAT1 K3504X mutation and loss of the remaining allele, and CAL27 cells that have one remaining FAT1 copy 8 . We also used WSU-HN6 cells (herein referred as HN6), which show the highest EGFR expression among our HNSCC cell line panel, but lack FAT1 alterations 21 . Remarkably, HN6 cells showed lower pYAP level and higher expression of YAP-regulated genes CTGF, CYR61, and AMOTL2, and the CTGF and CYR61 protein products ( Fig. 1a and b). We extended this analysis to all cancer types using the Cancer Cell Line Encyclopedia (CCLE) data set (1020 cancer cell lines 22 ). Gene set enrichment analysis (GSEA) revealed that YAP-regulated signatures gene sets (DUPONT: YAP, CORDENONSI_YAP_CONSERVED_SIGNATURE, ZHAO: INDUCED_BY_YAP) including representative YAPregulated genes (e.g., CTGF, CYR61, AMOTL2) were enriched with higher EGFR expression (Fig. 1c, Supplementary Figs. S1, 2a and b). In addition, when HNSCC patients from TCGA were stratified based on mRNA expression of EGFR, CTGF, and CYR61 (all z-score > 0 vs all z-score < 0), the EGFR, CTGF, and CYR61 "high" group (co-overexpression of EGFR and representative YAP-target genes) showed poorer survival with respect to those patients expressing low levels (high group: n = 128, low group: n = 114, Log-rank P = 0.015, Genhan-Breslow-Wilcoxon P = 0.0075). (Supplementary Fig. S3a). This suggests that EGFRactivated YAP/TAZ correlates with poor patient survival.
We next looked to examine whether EGFR can activate YAP/ TAZ, CAL27 cells were treated with EGF. EGF treatment reduced pYAP and increased TAZ levels, as well as CTGF, CYR61, AMOTL2 mRNA and CTGF and CYR61 protein expression, concomitant with the activation of canonical EGFR-downstream pathways including MAPK and PI3K-AKT-mTOR, reflected by increased phosphorylation of ERK1/2, AKT, and S6 ( Fig. 1d and e). Similar results were observed in CAL33 cells, which showed pYAP reduction and increase in YAP/TAZ transcriptional targets (Supplementary Fig. S3b and c). Therefore, EGFR can further activate YAP/TAZ even in the cells harboring FAT1 alterations.
Reconstituted EGFR expression induces hypo-phosphorylation and nuclear translocation of YAP/TAZ, thereby enhancing transcription of YAP/TAZ-regulated genes. To examine the precise mechanism by which EGFR activates YAP/TAZ, we established EGFR-overexpressing HEK293A cells, recapitulating HNSCC and other EGFR overexpressing cancer types. Vector-expressing HEK293A cells showed almost no effect on YAP by EGF treatment, but EGFR-overexpressing HEK293A showed significant pYAP reduction concomitant with CTGF, CYR61, and AMOTL2 mRNA increase, as well as ERK1/2, AKT, and S6 phosphorylation ( Fig. 2a and b). PIK3CA WT and H1047R overexpression slightly increased YAP/TAZ and CTGF/CYR61  Fig. S4a and b), which is consistent with our prior results in HNSCC cells ( Fig. 1f and g). Moreover, knockdown of YAP/TAZ significantly inhibited EGFR-induced CTGF, CYR61, and AMOTL2 expression ( Fig. 2c and d).
Under-phosphorylated and activated YAP/TAZ translocate from cytoplasm into nucleus, where they bind to TEAD transcription factor to act as a co-activator enhancing the transcription of proliferation-related genes 4 . EGFR induced hypo-phosphorylation of YAP and increased YAP-TEAD1 interaction (Fig. 2e), and immunofluorescence staining showed that EGFR activation triggered YAP/TAZ translocation from the cytoplasm into nucleus ( Fig. 2f and supplementary Fig. S4c). In summary, in EGFR expressing cells EGF activation induces hypophosphorylation of YAP and stabilization of TAZ, promote nuclear translocation of YAP/TAZ and their interaction with TEADs, which results in increased transcription of their target genes CTGF, CYR61, and AMOTL2.
EGFR promotes the phosphorylation of MOB1 and LATS1/2 inactivation, independently of MST1/2. Next, we sought to understand the mechanism of how EGFR activates YAP/TAZ. Because LATS1/2 directly phosphorylate YAP/TAZ on serine residues leading to cytoplasmic retention or proteosomal degradation, we examined LATS1/2 activity in the context of EGFR activation. The phosphorylation of the hydrophobic motif of LATS1/2 on threonine 1079 (T1079), which reflects its activity 24 , was reduced by EGFR activation, suggesting that LATS1/2 were inactivated (Fig. 3a). Indeed, in vitro kinase assays showed that LATS1 kinase activity on YAP was suppressed by EGF treatment (Fig. 3b). FBS was used as a positive control, as lysophosphatidic acid (LPA) and sphingosine 1-phosphophate (S1P) in serum inactivate LATS1/2 thereby stimulate YAP through G12/13coupled receptors 13 . In addition, CRISPR/Cas9 engineered LATS1/2 knockout cells abolished pYAP with or without EGF treatment, and increased TAZ level, and the status of YAP/TAZ was not changed further by EGFR activation (Fig. 3c). These data support that LATS1/2 are inactivated by EGFR, thereby promoting YAP activity.
We next attempted to clarify how LATS1/2 activity is suppressed by EGFR activation. Recent studies suggest that Hippo components can be regulated through tyrosine phosphorylation. For example, MST1 can be phosphorylated by FGFR4 and c-Abl, LATS1 by Src, and MOB1 by FAK [25][26][27][28] . Thus, we hypothesized that EGFR stimulation may lead to the phosphorylation of Hippo kinase components to activate YAP/TAZ. To test this hypothesis, we examined tyrosine phosphorylation of MST1, SAV1, LATS1, MOB1, and YAP by EGFR activation. Interestingly, only MOB1 showed tyrosine phosphorylation upon EGFR stimulation in cell in vivo (Fig. 3d). In vitro kinase assays showed that EGFR can directly phosphorylate MOB1 (Fig. 3e), but to explore whether this is also the case in cells in vivo we tested whether MOB1 associates with EGFR by coimmunoprecipitation assays. Although GRB2, an adapter protein acting directly downstream of EGFR, associated tightly with EGFR upon EGF stimulation, MOB1 association with EGFR or GRB2 could not be detected ( Supplementary Fig. S4d). Thus, MOB1 may represent a downstream substrate of EGFR without forming stable protein complexes, which is aligned with the absence of recognizable EGFR-interaction motifs in MOB1, or alternatively, MOB1 may be phosphorylated downstream of EGFR through intermediated receptor or non-receptor tyrosine kinases. Because MOB1 acts as an adaptor protein for LATS1/2, we examined the status of association of MOB1 with LATS1, which was not disrupted by EGFR activation (Fig. 3f). Moreover, since the hydrophobic motif of LATS1/2 is phosphorylated by MST1/2, we examined the activity of MST1/2. Phosphorylation of MST1/2 on threonine 180 and 183 autophosphorylation sites, reflecting MST1/2 activity, did not show differences upon EGFR activation. In addition, phosphorylation of threonine (T)35 of MOB1, a target site of MST1/2, was not affected by EGFR stimulation (Fig. 3g). These data suggest that EGFR activation promotes MOB1 phosphorylation, thus leading to LATS1/2 inactivation but independently of MST1/2.
EGFR inhibition with erlotinib increases pYAP and suppresses transcription of YAP-regulated genes in cancer. HNSCC is characterized by EGFR overexpression and amplification, while non-small cell lung cancer, especially lung adenocarcinoma, harbor frequent activating E746-A750 deletions or L858R mutations in EGFR 29,30 . Given this genetic background, we used increase and suppression of CTGF, CYR61, and AMOTL2 expression ( Fig. 5a and b). To examine the comprehensive transcriptional changes of EGFR inhibition on a global level, we conducted mRNA-sequencing (RNA-seq) of HCC827 cells treated with vehicle or erlotinib, and performed differential gene expression analysis to identify genes whose expression levels were dysregulated in response to erlotinib treatment (Fig. 5c-e and Supplementary Fig. S5a-c). We observed that along with previously reported erlotinib-regulated genes, many genes that are known to be regulated by YAP/TAZ were also suppressed, including CTGF, CYR61, AXL, FGF2, BIRC5, DUSP6, FOSL1, EGR1, HMGA2, AREG, CCND1 (  32 . In addition, YAP/TAZ knockdown in HN6 and HCC827 cells significantly suppressed cell viability, consistent with their growth dependency on YAP/TAZ ( Fig. 5f and g).
Loss of LATS1/2 confers resistance to erlotinib in cancer cells with EGFR alterations. To examine the importance of YAP/TAZ activation under EGFR in HNSCC and lung adenocarcinoma cells, we genome edited the LATS1 and LATS2 genes to activate YAP/TAZ in both cells harboring EGFR alterations. Initially, we took advantage of the CRISPR/Cas9 system to knockout (KO) LATS1 in HN6 and HCC827 cells. LATS1 KO HCC827 cells showed resistance to erlotinib, while LATS1 KO HN6 failed to rescue proliferation, suggesting that LATS1 KO was not sufficient to induce YAP/TAZ activation (Supplementary Fig. S6a and b). Thus, we performed the additional knockdown of LATS2, which partially rescued erlotinib-inhibited CTGF, CYR61, and AMOTL2 expression and conferred resistance to erlotinib (Supplementary Figs. S6c-f). As expected, the basal expression levels in LATS1/2 KO cells were much higher than those of LATS1 KO with siLATS2, suggesting complete knockout of LATS1/2 is required to fully activate YAP/TAZ (Supplementary Fig. S7a and b). We also confirmed that pYAP (S127) was completely dephosphorylated in LATS1/2 KO cells (Fig. 6a). Remarkably, LATS1/2 KO cells completely rescued erlotinib-inhibited CTGF, CYR61, AMOTL2 expression and was sufficient to confer resistance to the growth suppressive effects of erlotinib ( Fig. 6b and c). Both HN6 and HCC827 cells treated with erlotinib resulted in PARP cleavage, a typical molecular event caused by engagement of pro-apoptotic pathways. However, LATS1/2 KO cells showed a reduction in PARP cleavage, supporting that LATS1/2 deficiency confers resistance to erlotinib by promoting cell survival (Supplementary Fig. S7c and d).
Loss of LATS1/2 confers resistance to erlotinib in cancer cells with EGFR alterations in vivo. To further investigate the role of YAP/TAZ activation as a downstream signal of EGFR in HNSCC, we implanted WT and LATS1/2 KO HCC827 cells into NOD-SCID mice, and treated them with erlotinib or vehicle control. While the WT group showed remarkable reduction in tumor volume in response to erlotinib treatment and did not show regrowth after achieving near complete responses, the LATS1/2 KO group exhibited a significant but more limited tumor growth reduction during erlotinib treatment, and rapid regrowth after cessation of erlotinib administration ( Fig. 7a(left), b). In line with the tumor growth curves, the LATS1/2 KO group demonstrated a beneficial response to erlotinib treatment, but a significantly poorer survival compared to WT tumors ( Fig. 7a(right)). Immunohistochemical analysis showed that pEGFR was suppressed in both WT and LATS1/2 groups, and that the percentage of Ki67 positive proliferating cells in erlotinib-treated LATS1/2 group was significantly higher than that of erlotinib-treated WT group (Fig. 7c-e). These results indicate that YAP/TAZ activation may underlie intrinsic as well as acquired resistance to EGFR inhibition in EGFR-altered cancer cells, as judged by reduced tumor growth suppression and rapid tumor relapse.
Although the effector of the Hippo pathway, YAP was initially identified as a substrate of YES and other Src-family kinases, the  role of tyrosine phosphorylation of the core Hippo pathway components has not been studied in detail as compared to the large body of information regarding the regulation of this pathway by serine/threonine protein phosphorylation [38][39][40]   https:// www.phosphosite.org/uniprotAccAction?id=Q9H8S9), showed elevated levels of MOB1 phosphorylated at Y95, supporting our findings that EGFR-MOB1-YAP/TAZ signaling may play an important role in cancers harboring EGFR alterations 41,42 .
Distinct from FAK-induced phosphorylation of MOB1 on Y26 and its dissociation from LATS 28 , however, EGFR activation did not affect the interaction between MOB1 and LATS1/2, while it induced hypo-phosphorylation of the hydrophobic motif (T1079) in LATS1/2 and suppressed their kinase activity. The hydrophobic motif of LATS1/2 can be phosphorylated by activated MST1/2, although EGFR did not change the activity of MST1/2. Similar to MST1/2, mitogen-activated protein kinase kinase kinase kinase (MAP4K) family members, TAOK1 and TAOK3 are also capable of phosphorylating the hydrophobic motif of LATS1/2 [43][44][45][46] . Given that the hydrophobic motif of LATS1 is under-phosphorylated upon EGFR activation and LATS1 phosphorylation remains higher when MOB1 tyrosine phospho acceptor sites are mutated, it is possible that conformational changes triggered by tyrosinephosphorylation interfere with the interaction between LATS1/2 and MST1/2, MAP4Ks, or TAOKs. Further studies, including structural analysis, will be required to clarify the precise role of MOB1 tyrosine phosphorylation in LATS1/2 regulation, including the possibility that this may result in the interaction of the MOB1/ LATS1/2 complex with other upstream components or regulatory molecules. Notably, the distinct regulation of MOB1 tyrosine phosphorylation likely represents a previously unappreciated regulatory signaling node by which multiple receptor and nonreceptor tyrosine kinases may converge with the Hippo pathway to control YAP/TAZ activity Tyrosine kinase inhibitors (TKIs) are well accepted as a molecular targeted therapies for patients with cancers harboring EGFR mutations. EGFR-TKIs including erlotinib and gefitinib (first-generation reversible), afatinib (second-generation irreversible), osimertinib (third-generation irreversible) have been approved for the treatment of lung cancer patients harboring EGFR mutations [47][48][49][50] , and cetuximab has been used for HNSCC and colorectal cancer patients 19,20,51 . While cetuximab is the only FDA-approved cancer-targeting drug for patients with HNSCC, monotherapy response rate is limited (only 10-30%), suggesting the possibility of intrinsic or acquired resistance 29 . Similarly, use of EGFR TKIs in lung cancer show improved response rates (50-80%), but the emergence of intrinsic or acquired resistance, for example, the emergence of EGFR-T790M mutations or the activation of other signaling pathways including MET, AXL, IGF1R, IL-6R, HER2, and HER3 52 , often results in tumor relapse and progressive disease. Of importance, emerging evidence have shown that YAP is overexpressed and contributes to acquired resistance and poor prognosis of cetuximab in HNSCC or EGFR TKIs in lung cancers [52][53][54][55][56] . These prior reports in conjunction with our findings altogether support the theory that a prevalent mechanism of resistance to EGFR-targeted therapies is through the re-activation of YAP/TAZ. Loss of LATS1/2 or other Hippo pathway alterations could confer resistance to erlotinib in HNSCC cells with EGFR overexpression or lung adenocarcinoma cells harboring EGFR mutations. Specifically, our in vivo experiments showing a significant but more limited response to erlotinib in LATS1/2 KO group and rapid tumor regrowth after erlotinib treatment support the idea that YAP/TAZ activation plays an important role in therapy resistance and in tumor recurrence. Therefore, YAP and TAZ may represent mechanistic therapeutic targets in combination with EGFR targeting therapy in order to prevent cancer cells from acquiring resistance and the consequent treatment failure.
Taken together, our study revealed that EGFR promotes MOB1 phosphorylation and suppresses the Hippo pathway, leading to aberrant YAP/TAZ activation in cancers harboring EGFR alterations. These findings support that the EGFR-MOB1-YAP/ TAZ signaling axis may represent a novel therapeutic target for preventing cancer recurrence and progression. Cell culture and transfection. CAL33, CAL27, and HN6 cells were obtained from the NIDCR Oral and Pharyngeal Cancer Branch cell collection 21 . Their identity was confirmed by STR profiling and they were tested free of mycoplasma infection. HEK293 cells were purchased from ATCC (Manassas, VA). CAL33, CAL27, HN6, and HEK293 cells were cultured in DMEM (D-6429, Sigma-Aldrich Inc., St. Louis, MO) supplemented with 10% FBS (Sigma-Aldrich Inc., St. Louis, MO), 1× antibiotic/antimycotic solution (Sigma-Aldrich Inc., MO) and 5 μg/ml plasmocin TM prophylactic (InvivoGen, CA). HCC827 cells were purchased from ATCC and cultured in RPMI 1640 medium, GlutaMAX TM supplement (#61870-036, Thermo Fisher Scientific, CO).
Preparation of recombinant protein. GST-MOB1 was subcloned from HA-MOB1 plasmid into pGEX4T3 vector. pGEX-MOB1 WT was transformed in escherichia coli BL21. E. coli containing pGEX-MOB1 WT were cultured at 37°C for 3 hr, then cultured with 1 mM isopropyl-β-D-thiogalactopyranoside (IPTG) at 25°C overnight. The proteins purification step was performed using magneGST TM protein purification system, following the manufacturer's protocol (Promega, WI).
MOB1 point mutation. MOB1 8YF, 7YF + Y, and 3YF were generated using the QuikChange Lightning Site-Directed Mutagenesis Kit, following the manufacturer's protocol (Agilent Genomics, CA). pcDNA3-HA-MOB1 was used as template for mutagenesis. All mutated sites were validated by sequencing.
Cell viability assay. Cells were plated on 96 well plates. After cells attached on the plate, cells were treated with reagent for 3 days. Aquabluer reagent (#6015, MutliTarget Pharmaceuticals LLC, CO) was applied in the culture medium, incubated for 2 h, then the absorbances were measured by a microplate reader.
Immunofluorescence. Cells were cultured on coverslips coated with Poly-D-lysine hydrobromide (#P7280, Sigma-Aldrich Inc., MO) were rinsed with PBS, fixed with 4 % paraformaldehyde in PBS for 30 min, and permeabilized using 0.5% Triton X-100 with 200 mM glycine for 10 min. Fixed cells were blocked with 3% BSAcontaining PBS for 1 h at room temperature, and incubated with primary antibody overnight at 4°C. Then, incubated with alexa-labeled secondary antibodies (Goat anti-Rabbit IgG Alexa Fluor 488, #A11008, Thermo Fisher Scientific, CO) for 1.5 h at room temperature. Cells were stained with DAPI (#GTX16262, GeneTex, CA) for 10 min at room temperature and mounted. Images were acquired with Zeiss LSM 880 with Airyscan (Carl Zeiss, NY).
RNA sequencing. Samples were sequenced using the Illumina platform. For each sample, paired end sequencing reads were mapped using Bowtie2 version 2.3.4 to GRCh38 reference human genome downloaded from Ensembl. To compute transcript abundance, uniquely mapped reads were quantified using featureCounts version 1.6.3. Counts tables were uploaded to the Galaxy web platform and using the public server at usegalaxy.org, EntrezIDs were converted to gene symbols using the annotatemy IDs tool 59 . Differential gene expression analysis was performed using DESeq2 version 1.18.1, using parametric fit.
Gene set enrichment analysis (GSEA). For the analysis of CCLE data 22 , GSEA (Broad Institute, http://software.broadinstitute.org/gsea/index.jsp) was performed using with 1000 permutations, "Pearson" metric of RNA-seq read counts per gene and a gene set size filter of 15-500. The "C6" gene set database from MSigDB (Cordenonsi: YAP conserved signature") was spiked with "DUPONT: YAP" 60 and "ZHAO: Induced_by_YAP" 61 . For the analysis of the RNAseq data, GSEA was performed the same as above, using a t-test metric for ranking genes.
In vivo mouse experiments. All the animal studies using tumor xenografts studies were carried out according to the UCSD approved protocol (ASP # S15195) in compliance with the IACUC Guide for the Care and Use of Laboratory Mice. Female NOD-scid IL2Rgamma null mice (4-6 weeks of age) were purchased from Charles River Laboratories (Worcester, MA, USA). Cells were transplanted into both flanks (2 million per tumor) of each mouse. When average tumor volume reached a predetermined volume (~150 mm 3 ) the mice were randomized into groups (10 tumors per group). For drug treatment, the mice were treated (oral gavage) with erlotinib (Selleck Chemicals, 50 mg/kg/day) or control diluent (15% Captisol). The mice were euthanized at the indicated time points, when mice succumbed to disease, when tumor growth compromised animal welfare, or when tumor volume reached >200% of initial size at day 1 of treatment. Tumors were isolated for histologic and immunohistochemical evaluation.
Immunohistochemistry. All tissue samples were processed and stained as previously described 62 . The following antibodies were used: pEGFR (catalog number API300AA) was from Biocare Medical (Pacheco, CA, USA). Ki67 (catalog number ab15580) was from Abcam (Cambridge, MA, USA). Samples were scanned with Axioscan Z1 (Zeiss).
Statistics and reproducibility. All data were analyzed using GraphPad Prism version 7 for Windows (GraphPad Software, CA). The data were analyzed by Student's t-test (two-sided) and ANOVA with Tukey-Kramer post hoc test or as indicated in figure legends. All experiments were repeated independently at least three times with similar results, with the exception of the animal studies that were conducted once with a large cohort of mice.