1. Cancer Biology
  2. Genetics and Genomics
Download icon

Genome-wide CRISPR screens of oral squamous cell carcinoma reveal fitness genes in the Hippo pathway

  1. Annie Wai Yeeng Chai
  2. Pei San Yee
  3. Stacey Price
  4. Shi Mun Yee
  5. Hui Mei Lee
  6. Vivian KH Tiong
  7. Emanuel Gonçalves
  8. Fiona M Behan
  9. Jessica Bateson
  10. James Gilbert
  11. Aik Choon Tan
  12. Ultan McDermott
  13. Mathew J Garnett
  14. Sok Ching Cheong  Is a corresponding author
  1. Head and Neck Cancer Research Team, Cancer Research Malaysia, Head and Neck Cancer Research Team, Malaysia
  2. Wellcome Sanger Institute, Wellcome Genome Campus, United Kingdom
  3. Department of Biostatistics and Bioinformatics, Moffitt Cancer Center, United States
  4. Oncology R&D AstraZeneca, CRUK Cambridge Institute, United Kingdom
  5. Department of Oral & Maxillofacial Clinical Sciences, Faculty of Dentistry, University of Malaya, Malaysia
Research Article
  • Cited 0
  • Views 1,347
  • Annotations
Cite this article as: eLife 2020;9:e57761 doi: 10.7554/eLife.57761

Abstract

New therapeutic targets for oral squamous cell carcinoma (OSCC) are urgently needed. We conducted genome-wide CRISPR-Cas9 screens in 21 OSCC cell lines, primarily derived from Asians, to identify genetic vulnerabilities that can be explored as therapeutic targets. We identify known and novel fitness genes and demonstrate that many previously identified OSCC-related cancer genes are non-essential and could have limited therapeutic value, while other fitness genes warrant further investigation for their potential as therapeutic targets. We validate a distinctive dependency on YAP1 and WWTR1 of the Hippo pathway, where the lost-of-fitness effect of one paralog can be compensated only in a subset of lines. We also discover that OSCCs with WWTR1 dependency signature are significantly associated with biomarkers of favorable response toward immunotherapy. In summary, we have delineated the genetic vulnerabilities of OSCC, enabling the prioritization of therapeutic targets for further exploration, including the targeting of YAP1 and WWTR1.

eLife digest

Many types of cancer now have 'targeted treatments', which specifically home in on genes cancer cells rely on for survival. But there are very few of these treatments available for the most common type of mouth cancer, oral squamous cell carcinoma, which around 350,000 people are diagnosed with each year.

Designing targeted treatments relies on detailed knowledge of the genetic makeup of the cancer cells. But, little is known about which genes drive oral squamous cell carcinoma, especially among patients living in Asia, which is where over half of yearly cases are diagnosed. One way to resolve this is to use gene editing technology to find the genes that the cancer cells need to survive.

Now, Chai et al. have used a gene editing tool known as CRISPR to examine 21 cell lines from patients diagnosed with oral squamous cell carcinoma. Most of these lines were from Asian patients, some of whom had a history of chewing betel quid which increases the risk of mouth cancer. By individually inactivating genes in these cell lines one by one, Chai et al. were able to identify 918 genes linked to the survival of the cancer cells. Some of these genes have already been associated with the spread of other types of cancer, whereas others are completely unique to oral squamous cell carcinoma. The screen also discovered that some cell lines could not survive without genes involved in a signalling pathway called Hippo, which is known to contribute to the progression of many other types of cancer.

Uncovering the genes associated with oral squamous cell carcinoma opens the way for the development of new targeted treatments. Targeted therapies already exist for some of the genes identified in this study, and it may be possible to repurpose them as a treatment for this widespread mouth cancer. But, given that different cell lines relied on different genes to survive, the next step will be to identify which genes to inactivate in each patient.

Introduction

Head and neck squamous cell carcinoma (HNSCC) is a heterogeneous tumor arising from the mucosal surfaces lining the upper aerodigestive tract. The commonest subtype, oral squamous cell carcinoma (OSCC) is especially prevalent among Asian countries (Bray et al., 2018). OSCC has been associated with distinct risk habits such as betel quid chewing, tobacco smoking and alcohol consumption (Shield et al., 2017). The 5-year survival rate for OSCC is about 50% (Kumar et al., 2016) and surgery remains the mainstay of treatment. Cetuximab, an inhibitor of the epidermal growth factor receptor (EGFR), is used in combination with platinum-based chemotherapy for the treatment of advanced OSCC (Vermorken et al., 2008). However, the improvement in survival remains marginal (Vermorken et al., 2008). More recently, immune checkpoint inhibitors have been approved for the treatment of advanced and metastatic OSCC (Cohen et al., 2019). Although an improvement in patients’ outcome is anticipated with the advancement of immunotherapy, clinical trial outcomes showed an average objective response rate of only 13–36% (Bauml et al., 2017; Burtness et al., 2019; Ferris et al., 2016), and the factors determining response towards checkpoint inhibitors are still largely unknown. This underscores the need to identify further therapeutic targets for OSCC.

Genomic sequencing technology has enabled the delineation of the comprehensive mutational and transcriptomic landscape of cancers, including OSCC (The Cancer Genome Atlas Network, 2015; Pickering et al., 2013). However, the functional significance of most of these genetic alterations remains unclear and little is known about their value as therapeutic targets for OSCC. Identifying the genetic dependencies of OSCC will, therefore, be critical for the development of novel therapies. Genome-scale functional genetic screens allow the high-throughput identification of genes that govern cell survival (Gerhards and Rottenberg, 2018). Previously, such genes were identified using RNA interference (RNAi) technology (McDonald et al., 2017; Tsherniak et al., 2017). More recently, essential genes have been identified through the use of CRISPR-Cas9 technology due to its high specificity and efficiency compared to RNAi (Gerhards and Rottenberg, 2018). Several studies using genome-wide CRISPR-Cas9 screen have already shown promising outcome in identifying novel cancer-specific vulnerabilities that are useful drug targets (Steinhart et al., 2017; Wang et al., 2017), as well as improving the understanding of drug mechanism of action (Barazas et al., 2018; Hou et al., 2017).

The Cancer Dependency Map project (a consortium effort by the Wellcome Sanger Institute and the Broad Institute) have conducted CRISPR-Cas9 screen on a large number of cell lines including some OSCC models (Behan et al., 2019; Meyers et al., 2017). However, there is a lack of representation of Asians OSCC, such as those associated with betel quid chewing habit, one of the major risk factors of OSCC in many Asian countries (Shield et al., 2017). Further, the comparison of genomics data across different populations has revealed distinctive features in the different populations (Chai et al., 2020).

To identify genetic vulnerabilities in OSCC, we performed genome-wide CRISPR-Cas9 screens on 21 highly annotated OSCC cell lines, most of which are unique models derived from Asian patients (Fadlullah et al., 2016). Our study contributes to approximately one-third of the OSCC functional genetic screens currently available globally, expanding the representation of this heterogeneous disease (Behan et al., 2019; Meyers et al., 2017). In addition to finding known genetic vulnerabilities, we also uncover novel candidate genes essential for OSCC survival that can facilitate the development of new targeted therapies for OSCC. We validated the essentiality of Yes-associated protein 1 (YAP1) and WW domain-containing transcription regulator protein 1 (WWTR1) and revealed mutually exclusive dependency and compensable functions of these paralogs in different subsets of OSCC models. We identified OSCC tumors with a gene expression signature similar to cell lines with validated dependencies. Among which, OSCC resembling the WWTR1-dependent cell lines showed significant enrichment of immune-related pathways and are associated with biomarkers of response towards checkpoint inhibitors. In summary, our study demonstrated the robustness of genome-wide CRISPR-Cas9 screen in identifying genetic vulnerabilities in diverse OSCC models, offering new molecular insights into this disease.

Results

Genome-wide CRISPR screens in 21 OSCC cell lines

In order to identify genetic vulnerabilities of OSCC, particularly those that are more relevant to tumors of Asian origin, we conducted genome-wide CRISPR-Cas9 knockout screens (Figure 1A and Figure 1—figure supplement 1A). We screened a unique set of 14 well-characterized OSCC cell lines termed the ORL-series [ORL-48, –115, −136, –150, −153, –156, −166, –174, −188, –195, −204, –207, -214, –215]. These were established from the tumors of Malaysian OSCC patients (Fadlullah et al., 2016) and are comprehensively annotated with whole-exome sequencing (WES) and RNA sequencing data. In addition, we screened a further seven OSCC cell lines [BICR10, Ho-1-u-1, HSC-2, HSC-4, PE/CA-PJ15, SAS and SCC-9] sourced from commercial cell line repositories. Demographic details of the patients from whom the 21 OSCC cell lines were derived are shown in Figure 1A. The presence of mutations and copy number alterations in the top five significantly mutated genes are indicated. Overall, we find that our selection of cell lines represents the diversity of mutated driver genes observed in OSCC.

Figure 1 with 1 supplement see all
Genome-wide CRISPR-Cas9 screen on 21 OSCC cell lines.

(A) Demographic details and genomic profile of the patients of which the 21 OSCC cell lines were derived from, with bar charts depicting the number of essential genes identified by MAGeCK. The presence of mutations/copy number alterations in the top five mutated genes in OSCC is shown. Numbers in first column indicated frequency of mutations (%) among OSCC tumors from TCGA (The Cancer Genome Atlas Network, 2015) while second column indicated frequency of mutation (%) among 21 OSCC cell lines. Bar charts in the lower panel shows the number of significant fitness genes (those with MAGeCK FDR less than or equal to 5%), with the orange bars representing the number of non-core fitness genes. Abbreviations: Ind – Indian; Chi – Chinese, Mal – Malay; Jap – Japanese, Cau – Caucasian; F – Female; M – Male; G – Gingiva; T – Tongue; BM – Buccal Mucosa; FOM – Floor of Mouth; LN – derived from lymph node metastasis. (B) Pie charts showing the proportion of fitness genes among the 18,010 genes screened. 918 non-core fitness genes were shortlisted after filtering out the core fitness genes. (C) Bar chart depicting the number of non-core fitness genes that are found in 1 to 21 dependent cell lines.

Identification of core and context-specific fitness genes

Fitness genes were identified after an unsupervised computational correction with CRISPRcleanR (Behan et al., 2019; Iorio et al., 2018), followed by mean-variance modeling and systematic ranking of significantly depleted genes using MAGeCK (Li et al., 2014; Figure 1—figure supplement 1B). At a false discovery rate (FDR) of 5%, the number of significantly depleted genes ranged from 525 genes in ORL-156 to 1399 genes in ORL-215 (Figure 1A and Supplementary file 1).

As we aimed to identify genetic vulnerabilities of OSCC that can be safely targeted therapeutically, we filtered the significantly depleted genes to exclude previously defined core fitness genes (Behan et al., 2019; Hart et al., 2014; Hart et al., 2017; Meyers et al., 2017), that were found to be essential across the many cell lines from different lineages, and are likely toxic to the cells when targeted. In general, more than 80% of all the MAGeCK hits were found to be core fitness genes and were filtered out (Figure 1B and Supplementary file 1). Overall, among the 18,010 genes screened, 2539 (14%) were found to be significant MAGeCK hits in at least one cell line and following the removal of core fitness genes (Supplementary file 2), 918 context-specific fitness genes were shortlisted for further prioritization (Figure 1B). About 40% (366 genes) were uniquely essential in a single cell line, while the remaining 60% (552 genes) were essential in at least two cell lines, hence are recurrent context-specific essentialities (Figure 1C).

Capturing of known HNSCC cancer genes and pathways

Pathway enrichment analyses on the 918 context-specific fitness genes was conducted on each individual cell line (Figure 2—figure supplement 1A). Consistent with the highly heterogeneous nature of OSCC, diverse pathways were enriched across these cell lines. Pathways enriched among the 918 genes revealed several cancer-related pathways, potentially comprising important cancer-specific targets (Figure 2—figure supplement 1B and Supplementary file 3A-B). Whilst pathways such as ubiquitin-mediated proteolysis and cellular senescence are common across all cell lines, pathways such as NF-kappa B and MAPK signaling pathways are selectively enriched only in subsets of cell lines.

Next, we sought to determine which components of the common oncogenic pathways altered in HNSCC (The Cancer Genome Atlas Network, 2015) were required for cancer cell fitness and annotated the genes with the frequency of dependency (Figure 2A). Notably, most of the fitness genes are either existing drug targets or deemed clinically actionable. For example, drugs targeting PIK3CA and CDK6 are already in clinical trials for HNSCC treatment [NCT01816984, NCT02537223, NCT03356223, NCT03356587]. We also examined the dependency profile of 44 cancer genes with driver mutations known to be associated with HNSCC (Bailey et al., 2018; Martincorena et al., 2018; Figure 2B) and found that more than half of these cancer genes were dispensable for OSCC survival. Oncogene addiction has been the promising source of finding the Achilles heel for successful molecular targeted therapy (Weinstein and Joe, 2008). Based on the WES data of the 21 OSCC cell lines, we found 43 genes with driver mutations in at least one cell line and plotted the CRISPR score (measure of sgRNA depletion in the CRISPR screen) to examine if there is any differential dependency associated with the mutations (Figure 2C and Supplementary file 4). Dependencies on mutated PIK3CA were observed in four OSCC cell lines with a hotspot mutation in E545K (BICR10), Q546R (ORL-150) and H1047R (HSC-2), and to a lesser extent, in ORL-115 with H1047L mutation. Intriguingly, HSC-4, which harbors the same E545K mutation as BICR10, did not show any dependency on the mutated PIK3CA, this is consistent with findings from Project Score (Behan et al., 2019). A splice site driver mutation in PTEN co-occurred in HSC-4 and may have counteracted the oncogene addiction effect on the mutated PIK3CA, as suggested previously in breast cancer (Lazaridis et al., 2019). Dependency on NFE2L2 was observed in Ho-1-u-1 and BICR10 however the mechanism of activating this oxidative pathway differed between these two cell lines. Oncogene addiction is observed for Ho-1-u-1 with a hotspot mutation (E82D) in NFE2L2, that has been shown to enhance its transcriptional activity and promoting cell proliferation (Shibata et al., 2008). On the other hand, BICR10 harbors an inactivating mutation on KEAP1 (R320Q) a negative regulator of NFE2L2. The R320Q mutation has been reported to stabilize NRF2 (encoded by NFE2L2) and enhances cell fitness as reported previously in lung cancer (Hast et al., 2014). Finally, the only cell line that shows dependency on HRAS is ORL-214 which carries a mutation in HRAS (G12C). Another gene, encoding ZFP36L1 with truncating mutation at S324 in ORL-48 also showed preferential dependency, suggesting that the effect of this mutation should be studied further.

Figure 2 with 1 supplement see all
Identification and analysis of targetable genes and pathways in OSCC.

(A) Common oncogenic pathways altered among HNSCC samples of TCGA were annotated with frequency of dependency. Non-core fitness genes are indicated in red and the percentage of OSCC cell lines that were dependent on the genes are shown. The Drug Gene Interaction database (DGIdb) (http://www.dgidb.org/) was used to determine if the gene is clinically actionable while the availability of drugs were determined using Open Targets Platform (https://www.targetvalidation.org/). (B) Heatmap of gene essentiality of the 44 HNSCC cancer genes in the 21 OSCC lines. These are consensus cancer genes for HNSCC curated from Bailey et al., 2018 and Martincorena et al., 2018. (C) CRISPR scores of genes with driver mutations in at least one of the 21 OSCC cell lines. Cell lines labeled in green with mutation are examples of those showing oncogene addiction on mutated genes, for example on PIK3CA (ORL-150, BICR10, and HSC-2) and NFE2L2 (Ho-1-u-1). (D) Pathway enrichment analysis for fitness genes that are differentially enriched among the seven betel-quid associated OSCC. (E) Distribution of the 918 fitness genes based on their small molecule inhibitors tractability assessment. Tractability is defined as detailed in Behan et al., 2019 where: tractability group one included targets with approved drugs (Bucket 1) or drugs in clinical/pre-clinical development (Bucket 2, 3); tractability group two included targets with evidence supporting tractability albeit no drugs are available yet; while the least tractable group three included targets that lacks evidence informing tractability.

Identification of unique dependencies in betel-quid-associated OSCC

Studies on the genomic landscape of Asian and Caucasian OSCC have revealed distinct molecular differences, suggesting that some population-specific risk habits might have contributed to these differences (The Cancer Genome Atlas Network, 2015; Chai et al., 2020; Hsieh et al., 2001; India Project Team of the International Cancer Genome Consortium, 2013; Zanaruddin et al., 2013). Betel quid chewing is frequently associated with OSCC in Asia (Guha et al., 2014; India Project Team of the International Cancer Genome Consortium, 2013; Shah et al., 2012). In this study, several Asian-derived OSCC models that were associated with betel-quid chewing were included, and we had the opportunity to determine if there are differences in genetic dependencies between these OSCC (ORL-115, ORL-136, ORL-174, ORL-195, ORL-204, ORL-207, ORL-214) (n = 7), with those that are not associated with betel quid chewing (n = 14) (Figure 2D). Of the 110 fitness genes uniquely seen in betel-quid-associated OSCC, the NF-kB signaling pathway stands out as one of the significantly enriched pathway. The fitness genes from this pathway that are unique to the betel quid-associated OSCC include NFKB2, TNFAIP3, CSNK2A1, and TRIM25. When cross checking with the DepMap and Project Score data, three out of four (75%) of these genes (NFKB2, TNFAIP3, and TRIM25) were not found among the screened OSCC models, which were mostly derived from Caucasians, or from Asians not known to chew betel quid. Interestingly, our findings are coherent with previous studies from Taiwan and India, where betel quid chewing is common, which have demonstrated that extract from the areca nut of betel quid can directly activate the NF-kB signaling pathway, favoring OSCC cells survival (Chiang et al., 2008; Islam et al., 2019; Lin et al., 2005).

Tractability of the identified fitness genes

Given that many of the reported HNSCC-related cancer genes do not appear to be fitness genes, we sought to determine which of the 918 genes could potentially be tractable using previously defined frameworks (Behan et al., 2019; Brown et al., 2018; Figure 2E and Supplementary file 5). From the 918 genes, 45 genes fall into the tractability group 1, some examples of genes in this group include EGFR, PIK3CA, CDK4, and CDK6, where anticancer drugs targeting these genes are already approved or clinical trials for the treatment of HNSCC are on-going, demonstrating the robustness of our results. When classified based on protein function using PANTHER (Mi et al., 2013), most of those in tractable group 1, are transferase (kinases) (32%) and oxidoreductase (30%), including genes like CDK4, CDK6 and PIK3CA; and several genes in the family of NADH:ubiquinone oxidoreductase such as NDUFB9 and NDUFC2 (Figure 2—figure supplement 1C). Interestingly, emerging oncology and non-oncology drugs such as the HDAC inhibitors (Yoon and Eom, 2016) and miglustat, an approved drug for Gaucher’s disease (Barth et al., 2013), were amongst potential drug repurposing candidates that target fitness genes in tractability group 1 (targeting HDAC2 and UGCG respectively). The only transcription factor that falls within tractability group one is the ESR2, with several antagonists available owing to its well-studied ligands and structure (Ho, 2004; Figure 2—figure supplement 1C). A further 210 genes (23%) belong to tractability group 2, which harbors novel targets that have evidence supporting their tractability. Albeit no drugs are currently in clinical trials, these may hold potential for future drug development. For example, several companies are developing drugs that could target YAP1, SLC2A1 and PTPN11 which are in tractability group 2. However, about 70% of the 918 genes belong to tractability group 3 (least tractable or lacking evidence) where significant efforts in understanding their structure and function would be necessary to evaluate their tractability. Consistent with previous reports (Behan et al., 2019), the least tractable group comprises mainly of nucleic acid-binding proteins and transcription factors, such as the Kruppel-like factors (KLF) gene families (KLF4 and KLF5) and zinc finger proteins (ZNF148 and ZNF236) (Figure 2—figure supplement 1C).

Fitness genes in copy number amplified regions

HNSCC belongs to the ‘C class’ tumor, where the landscsape of genomic alterations is dominated by copy number alterations including recurrent chromosomal gains and losses (Ciriello et al., 2013). The most frequently reported copy number gains occurr in chromosomes 3q, 5 p, 7 p, 8q, and 11q (The Cancer Genome Atlas Network, 2015; Salahshourifar et al., 2014). To identify putative oncogenes that are essential for OSCC within these amplified regions, we evaluated the number of candidate fitness genes before and after CRISPRcleanR correction for copy number bias (Figure 3—figure supplement 1A). After correction, no enrichment/bias was found in the frequently amplified chromosome, demonstrating effective correction of copy number bias (Figure 3—figure supplement 1B). KEGG pathway analysis of the 152 genes from the five amplified regions showed enrichment of several oncogenic signaling pathways such as the small cell lung cancer, Hippo signaling pathway and ErbB signaling pathway (Figure 3—figure supplement 2A). We focused our analysis on the Hippo signaling pathway (Figure 3—figure supplement 2B), which has recently been implicated for major oncogenesis role in squamous cell carcinoma, including OSCC (Campbell et al., 2018; Ge et al., 2011; Hiemer et al., 2015; Wang et al., 2018). Furthermore, YAP1 or WWTR1 amplifications occur in approximately 19% of HNSCC (YAP1–5.5%, WWTR1–14.3%), which puts it among the top five cancers with the highest amplification of these genes amongst 33 cancer types (Wang et al., 2018).

Differential dependency pattern on YAP1 and WWTR1

YAP1 and WWTR1 (also known as TAZ) are transcription co-activators, which are the major effectors of the Hippo pathway (Guo and Teng, 2015; Wang et al., 2018). YAP1 and WWTR1 are paralogs with ~46–60% similarity in their amino acid sequence (Guo and Teng, 2015). They were shown to have both overlapping and distinct roles in different contexts (Guo and Teng, 2015; Plouffe et al., 2018). In OSCC, overexpression of YAP1 and WWTR1 has been shown to increase proliferation, survival and migration, mainly via interaction with the transcriptional enhanced associate domain (TEAD) transcription factors (Hiemer et al., 2015).

Interestingly, across the 21 OSCC cell lines, there is a subset of lines that show significant dependency on only one of the paralogs, while another subset of lines does not exhibit significant dependency on either YAP1 or WWTR1 (Figure 3A). With the exception of two cell lines, ORL-153 and ORL-215, YAP1-dependent lines and WWTR1-dependent lines are mutually exclusive. There were seven lines that were highly dependent on YAP1 (ORL-48, ORL-136, ORL-156, ORL-204, ORL-207, SAS, SCC-9), four highly dependent on WWTR1 (ORL-174, ORL-188, ORL-214, PE/CA-PJ15), while the other eight were not dependent on either (Figure 3A). Notably, WWTR1 gene locus (3q25) is at close proximity to the locus of PIK3CA, SOX2 and TP63 at 3q26-28, whereby their focal amplification is frequently reported in HNSCC (The Cancer Genome Atlas Network, 2015; Figure 3—figure supplement 2C). The majority of the WWTR1-dependent cell lines have copy number amplification for these genes on 3q25-28 (Figure 3A), but only WWTR1 CRISPR scores are significantly different between those with and without copy number amplification (p<0.001) (Figure 3—figure supplement 2D). Only two of nine cell lines with PIK3CA copy number amplification are dependent on PIK3CA itself. This suggests that copy number amplification of WWTR1 may constitute to a functional oncogenic role of WWTR1 in OSCC, instead of being a passenger gene that is co-amplified with the canonical HNSCC oncogene, PIK3CA. Notably, we also observed an enrichment of PIK3CA mutations (p=0.0003) among cell lines that are neither dependent on YAP1 or WWTR1, whereby five out of six such lines have PIK3CA hotspot mutations (BICR10, HSC-2, HSC-4, ORL-115, and ORL-150) (Figure 3A).

Figure 3 with 4 supplements see all
YAP1 and WWTR1 of the Hippo pathway are selectively essential in a distinct subset of OSCC cell lines.

(A) Essentiality profile (depicted with CRISPR scores heatmap) of YAP1, WWTR1, PIK3CA, TP63 and SOX2 across 21 OSCC cell lines derived from the CRISPR/Cas9 screen. Dependency on these genes were depicted as grey box in the bottom panel, according to the MAGeCK definition of significant depletion at FDR ≤ 0.05. No cell lines were dependent on TP63 or SOX2. The degree of essentiality differs across the lines. Some subsets of the cell lines are only dependent on either YAP1 or WWTR1, while neither gene appears to be essential in another subset of cell lines. PIK3CA, TP63 and SOX2 are genes implicated in HNSCC carcinogenesis that are often co-amplified with WWTR1, located on chromosome 3q25-28. All WWTR1-dependent cell lines had copy number amplification on these genes while all PIK3CA mutated cell lines are not dependent on either YAP1 or WWTR1. (B) Western blot images showing the protein level of YAP1 and WWTR1 on day 4 upon transducing the Cas-9 expressing cell lines with lentivirus carrying gene-specific sgRNA. Two sgRNAs were used per target gene. (C) Co-competition assay was used to validate the essentiality of YAP1 and WWTR1 on the selected cell lines. The growth of the BFP-positive transduced population was compared to the non-transduced population throughout the 18 days assay. The percentage of BFP-positive cells obtained at different time points were normalized to the day 4 readings for respective sgRNA (except ORL-204 which had time points normalized to the day 6 readings for respective sgRNA). PLK1 is a core essential gene included as a positive control. Negative controls include CHAT which is a non-essential gene across the panel of cell lines, and NT serves as a non-targeting control. Data are shown as mean ± SD (n = 2 biological repeats). (D) qPCR results show suppression of downstream targets of YAP1 and WWTR1 only when the respective fitness gene is being knocked-out. Down-regulation of CTGF and CYR61 gene expression was observed when YAP1 is knocked-out in the YAP1-dependent cell lines (ORL-48 and ORL-204). In the WWTR1-dependent cell lines (ORL-214, PE/CA-PJ15), CTGF and CYR61 expression is only suppressed when WWTR1 is knocked-out. Data are shown as mean ± SD (n = 2 independent experiments with technical triplicates).

Intriguingly, mutually exclusive copy number gains of chromosome 3q and 11q22 (where YAP1 is mapped to) have been reported in squamous cell carcinoma (Campbell et al., 2018) and consistently, YAP1 and WWTR1 amplification were also found to be mutually exclusive in HNSCC (Wang et al., 2018). In our study, despite having cell lines dependent on YAP1, none of the 21 OSCC cell lines shows copy number amplification of YAP1 or neighbouring genes on the chromosome 11q22. This suggests that other non-genomic mechanisms could likely be in place to activate YAP1.

To investigate if expression of YAP1 and WWTR1 is associated with the dependency, we examined baseline mRNA and protein expression of YAP1 and WWTR1 in representative OSCC lines (Figure 3—figure supplement 2E–F). Among the YAP1-dependent lines (ORL-48, ORL-204), overexpression of YAP1 mRNA and protein levels were observed. Further, these lines have low protein expression of WWTR1. The mRNA and protein expression of WWTR1 are relatively higher than YAP1 among the WWTR1-dependent cell lines (ORL-214 and PE/CA-PJ15), and those that were not affected when either YAP1 or WWTR1 is knocked-out (‘non-dependent’). To determine if the association between dependency and gene expression is exclusive to OSCC, we computed the differential dependency score of YAP1 and WWTR1 and their gene expression for 273 cancer cell lines from Project Score (Behan et al., 2019; Figure 3—figure supplement 3A). Generally, cell lines that are dependent on YAP1 have higher YAP1 expression, and similarly, WWTR1-dependent lines have higher expression of WWTR1 compared to YAP1. WWTR1 gene expression showed significant negative correlation with its dependency in these 273 cancer cell lines (Pearson R = −0.570, p-value=3e-25) and showed non-significant negative correlation among our 21 OSCC lines screened (Pearson R = −0.354, p-value=0.116) (Figure 3—figure supplement 3B–C). In other words, the higher dependency on WWTR1 gene is associated with higher WWTR1 gene expression. Interestingly, this observation was also seen in other cancer types, including non-small cell lung carcinoma, squamous cell lung carcinoma, glioblastoma, breast carcinoma, and lung adenocarcinoma (Figure 3—figure supplement 3D), based on the Project Score data. Among these cancers, non-small cell lung carcinoma showed the highest percentage of WWTR1-dependency (30%), with the strongest correlation (Pearson’s R = −0.934, p-value=0.0021).

To validate the differential dependency for YAP1 and WWTR1, we performed single-gene knockout using two sgRNAs per gene and investigated the growth inhibition effect of gene knockout using co-competition assay, as previously reported (Behan et al., 2019). Two cell lines each from the YAP1-dependent (ORL-48 and ORL-204), WWTR1-dependent (ORL-214 and PE/CA-PJ15) and YAP1/WWTR1 non- dependent (BICR10, HSC-2) groups were used. The efficacy of protein knockout using individual sgRNAs was assayed with western blotting (Figure 3B).

The results obtained from the co-competition assay corroborated our CRISPR screen data (Figure 3C). More than half of the transduced cell population was depleted following YAP1-knockout in ORL-48 and ORL-204 cells, but not when WWTR1 was knocked-out. Likewise, the growth inhibition in ORL-214 and PE/CA-PJ15 was only seen following WWTR1-knockout, but not upon YAP1-knockout. On the other hand, the fraction of transduced cells in BICR10 and HSC-2 did not show any prominent changes upon knockout of either YAP1 or WWTR1. These experiments were validated using clonogenicity assays (Figure 3—figure supplement 4A). Together, these results support the differential dependency pattern on YAP1 and WWTR1. Besides, another YAP1-dependent line, SAS has recently been reported to harbor a fusion protein of YAP1 and MAML2 (Picco et al., 2019). Our co-competition assay also confirms the dependency on YAP1 in these cells and the differential depletion of sgRNA targeting early and late exons of YAP1 suggested that this oncogenic fusion protein provided a survival advantage (Figure 3—figure supplement 4B–E). In OSCC, CTGF, and CYR61 are two canonical transcriptional targets of YAP1 and WWTR1 (Hiemer et al., 2015). Consistent with their known pro-survival properties, more substantial reduction of CTGF and CYR61 gene expression were seen only when the respective upstream fitness genes were knocked-out, as measured by qPCR (Figure 3D and Figure 3—figure supplement 4F).

Since YAP1/WWTR1 were both known to regulate proliferation and apoptosis, we next investigated whether the depletion of YAP1 and WWTR1 affects cell proliferation or apoptosis in the selected OSCC cells. Consistently, YAP1 depletion in ORL-48 resulted in significant reduction in viable cells, to a level comparable to the depletion of PLK1 (Figure 4A), and this was reflected in the increase in apoptotic cells (Figure 4B). Further, WWTR1 depletion in ORL-214 showed significantly lower percentage of viable cells when compared with the control, as well as significant increase in apoptotic cells (Figure 4A–B). Overall, these results confirm the dependency on either YAP1 or WWTR1 for survival, whereby depletion of the respective fitness genes resulted in increased apoptosis. By contrast, depletion of either YAP1 or WWTR1 did not affect the survival of HSC-2, as confirmed by the apoptotic assay (Figure 4A–B).

YAP1/WWTR1 knockout impairs proliferation and induces apoptosis in dependent cells.

(A) Knocking out YAP1 or WWTR1 significantly inhibited cell proliferation in the respective dependent cell lines. Cell viability of HSC-2 (a YAP1- and WWTR1-non-dependent cell line) was not affected by depletion of either YAP1 or WWTR1. Experiments are repeated twice in technical triplicates. (n = 3 technical repeats for two biological repeats). Data shown as mean ± SD. *p<0.01; **p<0.001. (B) Apoptosis assay revealed that YAP1 depletion in YAP1-dependent ORL-48 resulted in significant increase in apoptotic cells while WWTR1 depletion in WWTR1-dependent ORL-214 resulted in an increase in apoptotic cells. No significant changes in apoptotic cells were observed in HSC-2 when either YAP1 or WWTR1 was depleted. Experiments are repeated twice in technical triplicates. Data of one experiment are shown as mean ± SD (n = 3 technical repeats). *p<0.01; **p<0.001.

YAP1 and WWTR1 shows compensatory roles in OSCC cell lines

The lack of dependency on either YAP1 or WWTR1 in the non-dependent lines is intriguing, given the importance of these genes in most OSCC cancer cells. As YAP1 and WWTR1 share high structural homology and have common downstream targets, we hypothesized that YAP1 and WWTR1 could provide compensatory functions to maintain the survival of BICR10 and HSC-2 cell lines when either one of the genes was knocked-out.

To confirm our hypothesis, we knocked-out both YAP1 and WWTR1 simultaneously by co-transducing the cell lines with lentivirus carrying blue fluorescence protein (BFP)-tagged YAP1 sgRNA and mCherry-tagged WWTR1 sgRNA (Figure 5A). The co-competition assays showed that the population of BICR10 and HSC-2 with the double knockout of both YAP1 and WWTR1 depleted drastically (Figure 5B) compared to when each gene was knocked-out individually (Figure 3C). This suggests that in this YAP1/WWTR1 compensable subset of cell lines, the paralogs can compensate for the function of one another to activate the downstream mechanisms required to maintain cell fitness. This was substantiated by quantitative-PCR (qPCR) of downstream targets CTGF and CYR61 where substantial down-regulation in double knockout cells were observed compared to when each gene is knocked-out individually (Figure 5C).

Simultaneous depletion of YAP1 and WWTR1 on BICR10 and HSC-2 inhibited cell growth, suggesting compensatory role of YAP1 and WWTR1.

(A) Western blot showing the protein level of YAP1 and WWTR1 in BICR10 and HSC-2 Cas9 cell lines upon different combinations of YAP1 and WWTR1 sgRNA co-transduction. sgRNAs that targeted YAP1 were tagged with BFP marker and WWTR1 sgRNAs were tagged with mCherry marker respectively, they were simultaneously transduced into the Cas9-expressing cell lines to achieve double knockout of YAP1 and WWTR1. (B) YAP1 and WWTR1 double knockout cells show inhibited growth in co-competitive assay. Simultaneous knockout of YAP1 and WWTR1 in Cas9-expressing BICR10 and HSC-2 resulted in the depletion of sgRNAs-transduced populations, which was not observed upon the depletion of either gene alone. Results shown were normalized to day 4. Data are shown as mean ± SD (n = 2 independent experiments with technical triplicates). (C) qPCR results revealed that strong suppression of CTGF and CYR61 gene expression in BICR10 and HSC-2 is only seen when YAP1 and WWTR1 are simultaneously knocked-out. qPCR was performed in technical triplicates. Data are shown as mean ± SD (n = 3 technical repeats).

YAP1/WWTR1-dependency associated gene signatures in OSCC

Next, we sought to determine whether the differential dependency on YAP1 and WWTR1 is also relevant in OSCC tumors. We first derived the gene expression signatures representing the three groups using differentially expressed gene (DEG) analysis based on the OSCC cell lines with validated dependency (Figure 6—figure supplement 1A–B and Supplementary file 6). Using the ‘YAP1 dependency signature score’, ‘WWTR1 dependency signature score’ and ‘Compensable signature score’ [see materials and methods], OSCC cell lines were clustered into three broad groups based on their dependency on YAP1, or WWTR1 (Figure 6A). Using the same algorithm, we then computed the dependency signature score for each of the 315 OSCC tumors from the TCGA HNSCC cohort. From the heatmap and clustering analysis based on their dependency signatures, the three groups were also observed among the OSCC tumors (Figure 6B). To define representative ‘core’ samples, we found 41 OSCC tumors (13%) with high YAP1 dependency signature score (>0.5); 30 OSCC (9.5%) with high WWTR1 dependency signature score (>0.5) and 34 OSCC (11%) with high Compensable signature score (>0.5). Using these core OSCC samples and cell lines with validated dependency, we then used GSEA to identify hallmark pathways that are enriched in each of these groups (Figure 6C and Supplementary file 7). YAP1-dependent cell lines and tumors showed enrichment in hallmarks related to cell cycle, such as the E2F targets, G2M checkpoint, MYC targets, and DNA repair pathways (Figure 6C). This is consistent with previous reports that have demonstrated that the transcription factors E2F and MYC are critical downstream regulators of YAP/TEAD-mediated activation of cell cycle genes (Kapoor et al., 2014; Pattschull et al., 2019). While OSCC with high WWTR1 dependency signature score showed high expression of genes in several hallmarks related to immunity, such as the interferon responses, inflammatory responses and the complement pathway. This association is aligned with the recent findings that WWTR1 may play a role in immunity by upregulating PD-L1 expression (Janse van Rensburg et al., 2018). On the other hand, cell lines and tumors with high Compensable signature score showed enrichment in several metabolism-related hallmarks, such as fatty acid metabolism and xenobiotic metabolism. Notably, all three YAP1/WWTR1 compensable cell lines (BICR10, HSC-2 and HSC-4) harbor PIK3CA mutation, and that alterations in the PI3K signaling pathway have been linked to multiple metabolic dysregulations in cancer (Hao et al., 2016).

Figure 6 with 3 supplements see all
Identification of YAP1/WWTR1 dependency associated gene signatures in OSCC tumors samples.

(A) Heatmap with hierarchical clustering using the computed signature scores showed clustering of the OSCC cell lines into three groups, based on their validated dependency. (B) Heatmap with hierarchical clustering of OSCC tumors using the computed signature scores. The three clusters found from cell lines were also present among the OSCC tumors. For labeling convenience, OSCC with high YAP1 dependency signature score is referred as ‘YAP1-dependent’; ‘WWTR1-dependent’ – high WWTR1 dependency signature score; ‘Compensable’ – high compensable signature score. (C) GSEA of the OSCC cell lines and those OSCC tumors with similar gene signatures showed overlapping hallmarks enrichment, while distinct hallmarks were associated with each of the three groups. Only common hallmarks (between cell lines and tumors) with positive normalized enrichment score (NES) of >0.5 were shown. Full GSEA results can be found in Supplementary file 7. (D) OSCC with high WWTR1 dependency signature score showed significantly lower stromal infiltration, but higher immune infiltration compared to the other two groups. (E) By comparing the immune expression signatures from Thorsson et al., 2018, OSCCs with high WWTR1 dependency signature score are associated with significantly higher IFN-gamma response and cytolytic CD8 T-cells and lower TGF-beta response. (F) mRNA expression of PD-L1 (CD274) were significantly elevated among OSCC tumors with high WWTR1 dependency signature score. (G) OSCC with high WWTR1 dependency signature score showed significantly higher enrichment score for the 18-gene T-cell inflamed GEP, which is a clinically validated biomarker of response towards checkpoint blockade. For panels D-G, unpaired Welch’s t-test with Welch’s correction was used due to unequal sample size (n = 43, 31, 30, respectively).

OSCC with WWTR1 dependency signature and immune biomarkers

Among the plethora of diverse functions for YAP1 and WWTR1 evidence for their critical roles in mediating immune response have recently emerged (Geng et al., 2017; Janse van Rensburg et al., 2018; Pan et al., 2019). Specifically, WWTR1 but not YAP1 was shown to be essential for TH17 cell differentiation (Geng et al., 2017), and constitutively active WWTR1 (TAZ-S89A) was shown to induce PD-L1 expression to a much greater extent than YAP1 (YAP1-S127A) (Janse van Rensburg et al., 2018). Consistently, the results from the enrichment analysis of WWTR1-dependent OSCC were dominated by immune-related hallmarks. To investigate the association of YAP1/WWTR1 dependency with the immunity of OSCC, we mined the previously defined immune landscape of core samples from TCGA for comparison. Using the stromal and immune signature defined previously (Yoshihara et al., 2013), ssGSEA enrichment scores revealed that OSCC with high WWTR1 dependency signature score showed significantly lower stromal signature, but higher immune signature, when compared with OSCC of high YAP1 dependency or Compensable signature scores (Figure 6D). This suggests an enrichment of immune cell infiltration among OSCC with high WWTR1 dependency signature score. Next, we also assessed the other immune-related scores from previous work (Thorsson et al., 2018) and confirmed that these OSCC were associated with significantly enriched interferon-gamma (IFNγ) response signatures and reduced transforming growth factor-beta (TGFβ) response (Figure 6E and Figure 6—figure supplement 2A–B). Consistent with that, they also possess significantly higher cytolytic T cells (CD8) score than the other two groups (Figure 6E). This interesting observation led us to postulate that OSCC cancers with a gene expression signature associated with WWTR1 dependency might be more vulnerable to checkpoint blockade. To extend this hypothesis, we examined the level of several predictive biomarkers for immune checkpoint inhibitor response that had been tested and validated in the clinical setting (Ayers et al., 2017; Cristescu et al., 2018). Interestingly, the PD-L1 mRNA expression and the 18-gene T-cell inflamed gene expression profile (GEP) enrichment scores of the OSCC with high WWTR1 dependency signature score were significantly higher than the other two groups (Figure 6F–G). They were also associated with higher tumor mutational burden (TMB) albeit not statistically significant when compared with the other two groups (Figure 6—figure supplement 2C). To ensure that the association with immune signatures seen is specific to WWTR1 but not caused by other co-amplified genes in 3q25-28 (such as PIK3CA, TP63, and SOX2), we examined the changes in gene expression level of PD-L1 upon depletion of WWTR1 and other co-amplified genes. qPCR results revealed strongest suppression of PD-L1 gene expression upon WWTR1 knockout, but remain largely unchanged when the other genes were knockout (PIK3CA, TP63, and SOX2) (Figure 6—figure supplement 3A). We also examined a microarray dataset from Hiemer et al., 2015 and showed that a significant correlation of WWTR1 gene expression with PD-L1 expression was observed, but no correlation in gene expression was seen between PIK3CA, TP63 or SOX2 and PD-L1 (Figure 6—figure supplement 3B). Together, these results suggested that OSCC resembling the gene signature with WWTR1-dependent cell lines may be associated with better respond to immunotherapy.

Discussion

New therapeutic targets are urgently needed for the development of OSCC treatment. However, genomics studies have shown that oncogenic mutations in OSCC are largely limited to PIK3CA and HRAS and even in these, mutations are only found in a small subset of patients. Therefore, determining the oncogenic pathways and specific therapeutic targets have not been straight forward for this disease. We performed genome-wide CRISPR-Cas9 screens in a unique collection of OSCC cell lines derived from patients with diverse risk habits to identify genetic vulnerabilities that can serve as a basis for further therapeutic development for OSCC. Adding to existing genomics datasets in OSCC, this approach identifies fitness genes required for the survival of cancer cells where targeting these will result in the killing of these cells. The majority of the cell lines used in this study were from Asian patients where OSCC is more prevalent, and where causative factors include betel quid chewing in addition to smoking that is more often found in Caucasian patients (Cheong et al., 2017; Kumar et al., 2016). The molecular drivers of these patients are under-characterized as genomics data on Asian patients remain limited (India Project Team of the International Cancer Genome Consortium, 2013; Su et al., 2017).

Here, we identified 918 fitness genes in OSCC. Pathway enrichment analysis revealed that these genes were highly associated with diverse cancers pathways, confirming the robustness of our screen and pipeline in identifying targetable genetic dependencies. These included known candidate genes that are already being investigated in clinical trials for OSCC or other cancers such as CDK6, PIK3CA, and FGFR1, as well as novel genes that are yet to be explored as therapeutic targets, including those within the oxidative stress pathway (KEAP1, NFE2L2). Notably, we showed that about 5% (45/918) of these genes are highly tractable with approved drugs, or have drugs that are in late-stage of clinical testing, demonstrating that these screens could help to prioritize drugs that could be repurposed for OSCC treatment. We compared our screens with previous genome-wide RNAi screens and found that candidate genes such as those relating to the cell cycle (CKAP5, KPNB1, RAN, TPX2, and KIF11) were also identified in our dataset (Martens-de Kemp et al., 2013). However, due to stringent filtering of core-fitness genes (Behan et al., 2019; Hart et al., 2014; Hart et al., 2017; Meyers et al., 2017), these were no longer within the list of OSCC non-core fitness genes.

Notably, the unique inclusion of several Asian OSCC models known to be associated with betel-quid chewing in our screen enabled us to identify that the NF-kB signaling pathway is among one of the most significantly enriched dependencies among betel-quid associated OSCC compared to cancers not associated with this risk habit. This finding is in line with the past studies reporting direct activation of the NF-kB signaling, upon treatment of OSCC cell lines with the extract of areca nut which is the main component in the betel quid (Chiang et al., 2008; Lin et al., 2005). By contrast, members of this signaling pathway (TRIM25, NFKB2, TNFAIP3) did not seem to be fitness genes in the other cell lines not associated with betel quid chewing, nor those from the Project Score or DepMap (Behan et al., 2019; Meyers et al., 2017). Members of the NF-kB signaling pathways have been proposed as therapeutic targets for inflammatory diseases and cancers, with various types of inhibitors being developed. In the event of the development of NF-kB inhibitosr with promising clinical utililty in the future, it would be of interest to investigate if increase efficacy of this inhibitor would be seen among the betel-quid-associated OSCC. Appreciating this differences in dependencies in Asian OSCC that remains undiscovered in existing large genetic screens could have significant implications, especially when employing precision medicine in the different populations. We acknowledge, however, that further in-depth investigations including a larger sample size and further representative models, are needed in order to confirm our findings and to inform the basis of developing potential targeted therapy against the unique vulnerabilities among the Asian OSCC.

Driver mutations are often expected to be robust biomarkers in precision medicine. However, upon analysing the dependency profiles on cancer genes that are commonly mutated in HNSCC and those with driver mutations among our 21 OSCC cell lines, we show that with the exception of some genes with driver mutations leading to oncogene addiction (PIK3CA, HRAS and NFE2L2), most other driver mutations did not confer preferential gene dependency and their value as a drug target remains unclear.

Given the propensity of copy number alterations in driving OSCC, a ‘C’ class tumor (The Cancer Genome Atlas Network, 2015; Ciriello et al., 2013), we investigated the commonly amplified genomic regions to look for functionally important candidate genes. One of the pathways that was significantly enriched was the Hippo pathway. Within this pathway, we focused our analyses on YAP1 and WWTR1, two paralogs that show differential dependency pattern in our 21 OSCC cell lines. Copy number amplifications of 11q22 and 3q25 (where YAP1 and WWTR1 are located, respectively) are common events reported in OSCC, often in mutual exclusive manner (Campbell et al., 2018; Wang et al., 2018). These two genes are the major effectors negatively regulated by the Hippo pathway that is increasingly reported to play multiple roles in carcinogenesis, as reviewed comprehensively in recent years (Dey et al., 2020; Santos-de-Frutos et al., 2019). However, the majority of the studies focused on either one of the paralogs or had assumed similar functions between paralogs (Santos-de-Frutos et al., 2019; Zanconato et al., 2016). Emerging evidence demonstrate that YAP1 and WWTR1 have distinct roles where they partner with different transcription factors, drive different downstream effectors and also modulate the tumor microenvironment distinctively (Callus et al., 2019; Janse van Rensburg et al., 2018; Kaan et al., 2017; Plouffe et al., 2018). Our study revealed the intricate dominance of dependency on either one of the paralogs, despite the other not being deleted or loss. More intriguingly, the subset of lines thought not to be dependent on either YAP1 or WWTR1, were actually lines where the remaining paralog is able to compensate for the lost paralog, therefore, enabling the cells to continue surviving.

In the context of OSCC/HNSCC, studies demonstrating the distinct roles and regulatory mechanism of YAP1 and WWTR1 are emerging. Analysis of genome-wide transcriptional changes upon knockdown of YAP1 or WWTR1 in OSCC showed that YAP1 had a more prominent role in transcriptional regulation (Hiemer et al., 2015). Using a tongue orthotopic mouse model with the deletion of MOB1A/B, Omori et al., 2020 provided strong evidence that YAP1 acted as a strong driver in OSCC tumor initiation and progression, whereby WWTR1 did not seems to play an equivalent role (Omori et al., 2020). The differences between YAP1 and WWTR1 can also be further exemplified in terms of their interaction with upstream/downstream pathways, that would involve other frequently co-amplified genes such as PIK3CA, TP63, and SOX2. The co-occurrence of amplifications in these genes that are part of the extended signaling network of the Hippo pathway underscores the critical role of the Hippo pathway in driving the OSCC tumorigenesis. Overexpression of PIK3CA was shown to be correlated with YAP1 activation and associated with poor clinical outcome (García-Escudero et al., 2018). Further, the activation of the PI3K through mitogenic signaling inhibits the Hippo pathway leading to YAP1 activation and cell growth (Fan et al., 2013). On the other hand, WWTR1 was shown to act upstream of SOX2, facilitating stemnesses in HNSCC (Li et al., 2019). Intriguingly, while WWTR1 knockdown led to reduction of SOX2 mRNA and protein expression, this was not seen when YAP1 was knockdown in the HNSCC cells (Cal27 and Fadu) (Li et al., 2019; Huang et al., 2017). In support of that, an inverse relationship between the expression levels of YAP1 and ΔNp63 were reported in lung SCC tumor samples (Huang et al., 2017). However, this was not seen in HNSCC tumor samples and cell lines (Ge et al., 2011), suggesting a cancer/context-specific regulatory mechanism might be in place. Consistent with the oncogenic roles of YAP1 reported in OSCC (Hiemer et al., 2015; Omori et al., 2020), another study showed that p63, together with the co-expressing chromotin remodeling factor, ACTL6A, can drive YAP1 activation, suppressing differentiation and promoting cell proliferation in HNSCC (Saladi et al., 2017). Similar observations between WWTR1 and TP63 have not been reported thus far. Hence, understanding the context in which the OSCC lines can be either YAP1-dependent/WWTR1-dependent or having compensable YAP1/WWTR1 is important, as YAP1 and WWTR1-dependency appear to be associated with the enrichment of distinct pathways. Current inhibitors of YAP1 such as verteporfin and CA3 that blocks its interactions with the TEAD transcription factors also targets WWTR1 (Song et al., 2018; Zhang et al., 2015); therefore, this data underscores the need to develop more specific inhibitors to prevent the targeting of many different downstream pathways. Intriguingly, we also observed enrichment of PIK3CA mutant (p=0.0003) among OSCC lines that are compensable for YAP1 or WWTR1. As recent studies have provided evidence that YAP1 and WWTR1 could mediate mutant PIK3CA-induced tumorigenesis (Zhao et al., 2018) andother studies also suggested crosstalk between these Hippo pathway effectors with the PI3K-Akt pathway (García-Escudero et al., 2018), confirmatory and mechanistic studies will be needed to delineate why YAP1 and WWTR1 function can be compensated in these PIK3CA-mutated cell lines, while distinct dependencies on either paralog are observed in PIK3CA wild-type lines. The functional loss of mutated FAT1 has also been reported to be associated with YAP1 activation in head and neck cancer (Martin et al., 2018), however, no enrichment of FAT1 mutation was seen among the YAP1-dependent nor WWTR1-dependent models in this study.

We also provided tissue-relevant insights of our findings by including analysis of the OSCC tumors from TCGA. Since YAP1 and WWTR1 are transcription co-factors that could regulate a plethora of gene transcriptions, we devised an analysis workflow utilizing the DEGs among the three subsets of OSCC models. We have identified 105 OSCC tumors that show highly similar gene expression signature as the cell lines, which are predicted to share the same dependency pattern. Similarly, comparison of the OSCC tumors based on their gene signatures revealed significant differences in terms of their enriched gene sets and immune signatures. As checkpoint blockade is approved for the treatment of recurrent and metastatic OSCC, understanding how YAP1 and WWTR1 influence the immune microenvironment could provide clues on the combination therapies that could increase the subset of patients responding checkpoint inhibitors. In particular, OSCC with high WWTR1 dependency signature score are significantly associated with various biomarkers that were predicted to show good response toward checkpoint inhibitors. This finding is consistent with the recent discovery that constitutively active WWTR1 induces PD-L1 expression, to a greater extent than YAP1 (Janse van Rensburg et al., 2018), and that tumors with YAP1 amplifications have low T-cell infiltration (Saloura et al., 2019). These observations have clinical implication as anti-PD1 is an approved therapy for HNSCC and therefore, further investigation and validation will be needed to confirm this observation and its clinical impact. While many companies are developing novel inhibitors targeting YAP1-TEAD transcriptional activity, which should be effective against all other OSCC, combination with checkpoint inhibitor could be considered for those OSCC with WWTR1 dependency signatures. We acknowledge, however, that whilst the dependency observed in cell lines could be recapitulated in OSCC, OSCC tissues would be much more heterogenous and could harbor specific genetic abrogations that could be the dominant driver of tumorigenesis. Therefore, further validation of the association between WWTR1-dependency signature and response to checkpoint inhibitors should be validated particularly in the context of clinical trials involving checkpoint inhibitors.

The roles of the Hippo signaling pathway and its effectors YAP1 and WWTR1 in cancer immunity remains unclear. The inactivation of the Hippo pathway through the loss of LATS1/2 was reported to cause the induction of anti-tumor immune response and inhibition of HNSCC tumor growth, via the hyperactivation of YAP1/WWTR1 (Barth et al., 2013; Moroishi et al., 2016). This demonstrates that components of the Hippo signaling pathway could also modulate the host tumor microenvironment in addition to what we have demonstrsated in cancer cells. The design and models used in our study have not been set up to examine this where the inherent limitation of using cell lines do not consider the components of the tumor microenvironment in the in vitro screening. Nonetheless, our findings provided a novel insight linking the intricate dependency on YAP1 and WWTR1 with differential state of the immune microenvironment in OSCC which warrants further investigation with the use of immune-competent mouse models, before further clinical evaluation can be made.

In summary, our study is the first large-scale CRISPR-Cas9 screen and focused analysis conducted on large panels of Asian derived OSCC cell lines and provided a cancer-specific overview of the fitness genes landscape, affording opportunities for further therapeutic targets development. The ability to scrutinize the functional genomics of these fitness genes/pathways to a greater detail was also exemplified in this study.

Materials and methods

Cell lines

Request a detailed protocol

Fourteen OSCC cell lines (referred to as the ORL- series) were derived spontaneously from surgically resected OSCC tissue specimens in Cancer Research Malaysia. Briefly, tissues were collected in α-MEM containing 20% (v/v) FBS, 200 iu/l penicillin, 200 μg/ml streptomycin and 0.1 μg/ml of fungizone. Subsequently, tissues were washed in absolute ethanol for 20–30 s and then washed twice with phosphate-buffered saline (PBS) under sterile conditions. Tissues were minced, washed twice in culture media and re-suspended in α-MEM containing 20% (v/v) FBS, 200 iu/l penicillin, 200 μg/ml streptomycin, 0.4 ng/ml EGF, 2 μg/ml hydrocortisone and 2 mM L-glutamine, and seeded into tissue culture dishes. Cultures were continuously maintained for more than 100 population doublings (Fadlullah et al., 2016). HSC-2, HSC-4, and SCC-9 were isolated from squamous cell carcinoma of various oral regions, by which the surgically excised tumors were minced and disaggregated to single cells. Epithelial cells proliferated from the explants were then sub-cultured continuously (Momose et al., 1989; Rheinwald and Beckett, 1981). No special immortalization methods were detailed for BICR10, PE/CA-PJ15, Ho-1-U-1 and SAS (Edington et al., 1995, Berndt et al., 1997, Miyauchi et al., 1985, Takahashi et al., 1989) nor for all other cell lines used in this study. All these OSCC cells lines were maintained in Dulbecco’s Modified Eagle’s Medium(DMEM)/Nutrient Mixture F-12 medium (Gibco) supplemented with 10% (v/v) heat inactivated fetal bovine serum (Gibco), and 100 IU Penicillin/Streptomycin (Gibco). All lines were incubated in a humidified atmosphere of 5% CO2 at 37°C. The lines were authenticated by STR profiling using Promega PowerPlex16HS Assay (Promega, Wisconsin, United States), with the data giving more than 80% match to the respective donor or reference as deposited in the databases of cell line resources (such as ATCC, DSMZ, JCRB). Cell lines were routinely tested for the presence of mycoplasma with MycoAlert mycoplasma detection kit (Lonza, Basel, Switzerland). Only mycoplasma-free cell lines were used in all experimentation.

Plasmid transfection and virus transduction

Request a detailed protocol

HEK293 cells were transfected using jetPRIME transfection reagent (Polyplus Transfection) according to the manufacturer’s instructions. Briefly, transfection complex consisting of jetPRIME buffer, jetPRIME reagent, vector of interest, pMD2.G and psPAX2 were prepared and mixed with Opti-MEM (Gibco). Next, the transfection complex medium was added to HEK293 cells with 90% confluency. After overnight incubation at 37°C, 5% CO2, transfection complex medium was replaced with fresh DMEM high glucose (Gibco) complete medium. Medium containing virus was collected at 48- and 72 hr post-transfection and filtered using PVDF 0.45 μm syringe filter.

To perform virus transduction, selected cell lines were transduced with lentivirus containing the vector of interest, in the presence of 8 μg/ml polybrene. After overnight incubation, medium containing lentivirus was replaced with fresh DMEM/F12 complete medium. Cells were incubated for 48 hr and harvested to evaluate the transduction efficiency via flow cytometry analysis with BD LSR Fortessa X-20 cell analyser (BD Biosciences). Gating strategy for flow cytometry analysis of transduced cells carrying fluorescence marker is exemplified in Supplementary file 8.

Generation of Cas9-expressing cell lines

Request a detailed protocol

Selected cell lines were transduced with lentivirus containing the pKLV2-EF1α-Cas9Bsd-W (Addgene plasmid # 68343, gift from Kosuke Yusa). Cells stably expressing Cas9 enzyme was established via blasticidin selection 3 days post-transduction. Cas9 enzyme cutting efficiency was routinely checked via lentivirus transduction of the reporter plasmid pKLV2-U6gRNA5(gGFP)-PGKmCherry2AGFP-W (Addgene plasmid # 67982, gift from Kosuke Yusa). The efficiency of Cas9 cutting activity was accessed using a reporter plasmid and analyzed using flow cytometry analysis. The screening will only be conducted on those cell lines with >80% Cas9 cutting activity, as indicated by the efficiency of GFP knockout in cell lines transduced with the reporter plasmid.

Genome-wide CRISPR-Cas9 knockout screening

Request a detailed protocol

The Human Improved Genome-wide Knockout CRISPR Library v1 (Addgene plasmid #67989, gift from Kosuke Yusa) containing 90,709 gRNAs targeting a total number of 18,010 protein-coding genes was used in the genome-wide CRISPR-Cas9 screening (Tzelepis et al., 2016). For each of the OSCC cell lines, a total of 60 million Cas9-expressing cells were transduced with the CRISPR library lentivirus at a multiplicity of infection (MOI) of 0.3. Polybrene at 8 μg/ml was added to increase transduction efficiency. All screenings were performed in triplicates. Library representation was evaluated by the percentage of BFP-expressing cells, determined using flow cytometry on day 4 post-transduction. The library representation was minimally kept at 100X coverage of the library, equivalent to 30 million cells total, or 10 million cells expressing BFP before proceeding to puromycin selection (2.0 μg/ml) for 3–4 days, to select for successfully transduced cells. Following complete selection with puromycin, a minimum of 75 million cells were maintained throughout the 18 days screen. BFP expression was monitored to ensure selection was adequate. On day 18 post-transduction, 60 million cells were pelleted down for genomic DNA extraction.

Genomic DNA extraction of post-CRISPR screened cells

Request a detailed protocol

Genomic DNA was extracted from 60 million post-CRISPR screened cells using QIAGEN blood and cell culture DNA Maxi kit (Qiagen), according to manufacturer’s instruction. Extracted DNA was quantified using Qubit 2.0 fluorometer (Thermo Fisher Scientific).

Library preparation of genomic DNA

Request a detailed protocol

To prepare and generate Illumina libraries for deep sequencing, amplification of sgRNA was performed using Q5 Hot Start High-Fidelity 2 × Master Mix and forward/reverse primers pair (gLibrary-HiSeq_50bp-SE-U1 F and R) in 50 μl reaction, as previously described4. For the CRISPR library v1 plasmid, 10 independent PCR reactions were set up using 2.0 μg of the plasmid. While for the CRISPR-screen cell lines, 2.0 μg of genomic DNA harvested from day-18 post-transduction was used in each of the 36 independent PCR reactions. The PCR conditions were as follows: 98°C for 30 s, 26–28 cycles of 98°C for 10 s, 61°C for 15 s and 72°C for 20 s, and the final extension, 72°C for 2 min. The PCR products were analyzed on 2% agarose gel and additional PCR cycles were added if necessary.

About 5 μl of PCR products were pooled from all 36 reactions and QIAquick PCR purification kit (QIAGEN) was used to purify the amplified gRNA. Concentration of purified PCR products were quantified with Qubit dsDNA broad-range (BR) assay kit (Thermo Fisher Scientific), using the Qubit 2.0 fluorometer (Thermo Fisher Scientific). PCR enrichment was then carried out using 200 pg of purified PCR products with 2x KAPA HotStart ReadyMix. 1 μl of forward P5 fusion primer (PE 1.0 p5 Top_PE_C) and 1 μl of different reverse primers (indexed iPCRTags) were used. The PCR conditions were as follows: 98°C for 30 s, 10–12 cycles of 98°C for 10 s, 66°C for 15 s and 72°C for 20 s, and the final extension, 72°C for 5 min. Finally, SPRISelect beads (Beckman Coulter) were used to purify the PCR products at a PCR-product-to-bead ratio of 1:0.8. Purified libraries were dissolved in 30 μl nuclease-free water and quantified using Agilent High Sensitivity DNA kit (Agilent Technologies) on Agilent 2100 Bioanalyzer (Agilent Technologies).

Deep sequencing of post-CRISPR screen libraries

Request a detailed protocol

Purified libraries of the triplicate screens tagged with different iPCR tags were pooled and sequenced at about 300x coverage on Illumina HiSeq 2500 with 19 bp single-end (SE) deep sequencing at the Wellcome Sanger Institute (WSI). About 30–40 million reads were obtained for each of the three replicates. Sequences of the Read one sequencing primer (U6-Illumina-seq2) and index sequencing primer can be found in Supplementary file 9.

CRISPR data processing

Request a detailed protocol

sgRNA raw counts in each triplicate of the CRISPR screen were generated using the in-house script developed at WSI, as previously described (Behan et al., 2019; Iorio et al., 2018). The CRISPRcleanR tool was downloaded as an R package (https://github.com/francescojm/CRISPRcleanR) and used for pre-processing of the sgRNA raw counts (Iorio et al., 2018). This tool allows for an unsupervised correction for copy number amplification bias and other gene-independent responses when subjected to CRISPR-Cas9 targeting, hence reducing false-positive call for essential genes (Iorio et al., 2018). Briefly, the ccr.NormfoldChanges function was used to compute the median-ratio normalization of raw counts and log2 fold-changes for all sgRNAs, averaging from triplicates. The built-in KY_library_v1.0 was used for library annotation and sgRNAs with read counts less than 30 in the plasmid were excluded. Then, the ccr.logFCs2chromPos and ccr.GWclean function of the CRISPRcleanR were used to perform genome mapping and sorting of the sgRNAs, followed by the correction of the gene-independent responses to compute corrected log-fold changes. As part of the quality assessment, Pearson correlation test was used to compare sgRNA raw counts between replicates of the same cell line, while precision-recall was assessed as previously described, using sets of known essential and non-essential genes (Behan et al., 2019; Supplementary file 10).

The corrected gene-level log-fold changes were quantile normalized and corrected for batch effect using ComBat (Leek et al., 2012), we refer this CRISPRcleanR corrected, quantile normalized and ComBat corrected log-fold changes as the ‘CRISPR score’, with a negative value indicating the extent of depletion of sgRNA counts targeting the gene when compared with the initial plasmid library.

The function ccr.correctCounts used inverse-transformation method to generate CRISPRcleanR-corrected counts, which is used as input files for MAGeCK analysis (Iorio et al., 2018; Li et al., 2014). For each of the 21 OSCC cell lines, MAGeCK analysis was performed using default parameters, except that normalization is set to ‘none’, as the input corrected counts had already been normalized using CRISPRcleanR. A false discovery rate cut-off of 5% (FDR ≤ 0.05) was applied to identify the significantly depleted genes in each cell line, defined here as MAGeCK hits.

To remove potential false-positive hits, RNA-seq expression data of each of the 21 cell lines were utilized to filter out MAGeCK hits with negligibly low or no reported reads (Fragments Per Kilobase of transcript per Million mapped reads (FPKM) <0.5).

Next, in order to identify and prioritize genes that can be safely targeted for the treatment of OSCC, we curated a list of core fitness essential genes from four different sources and used it to further filter the list of MAGeCK hits from each cell line. The first two sources were the core essential genes (CEG) list, published by Hart et al., 2014 and the subsequent updated CEG2 list, published in 2017 (Hart et al., 2017). The third source is the ‘common-essential genes’, downloaded from Broad’s Institute Cancer Dependency Map database (18Q3 release) (Meyers et al., 2017). We also utilized the list of pan-cancer core-fitness genes compiled by Project Score of Cancer Dependency Map at WSI (Behan et al., 2019). The full list of all genes from these four sources was tabulated in Supplementary file 2.

WES and identification of driver mutation

Request a detailed protocol

All 21 OSCC cell lines were subjected to WES at the WSI using HiSeq2500. WES data were processed using an established pipeline as previously described to identify driver mutations (Iorio et al., 2016). A total of 43 genes with driver mutation in at least one cell line were identified (Supplementary file 5).

Pathway enrichment analysis

Request a detailed protocol

KEGG pathway enrichment analysis was performed using the over-representation analysis function at the ConsensusPathDB (http://ConsensusPathDB.org) (Kamburov et al., 2013). A threshold of minimum two genes overlapping with the gene set of a given pathway and p-value cut-off of 0.05 were applied. Enriched pathways were ranked by q-value.

Differential fitness genes analysis for betel-quid-associated OSCCs

Request a detailed protocol

OSCC models with known betel quid chewing as the only risk habit were included in this analysis – (n = 7; ORL-115, ORL-136, ORL-174, ORL-195, ORL-204, ORL-207 and ORL-214), to compare with the other OSCCs not associated with betel quid chewing (n = 14; ORL-48, ORL-150, ORL-153, ORL-156, ORL-166, ORL-188, ORL-215, BICR10, Ho-1-u-1, HSC-2, HSC-4, SAS, SCC-9, PE/CA-PJ15). Venn diagram was used to depict the number of unique and overlapping fitness genes that are found between those OSCCs with or without association with betel quid chewing.

Functional classification and tractability assessment

Request a detailed protocol

To assess the tractability of the 918 non-core fitness genes, we utilized the genome-wide target tractability assessment pipeline as previously described (Behan et al., 2019; Brown et al., 2018). Based on assessment for small molecules tractability, essential genes were assigned into tractability bucket 1 to 10, with decreasing tractability. Next, essential genes in each tractability group were further classified into protein classes using the PANTHER database online tool (http://www.pantherdb.org/) (Mi et al., 2013).

Validation sgRNA design and cloning

Request a detailed protocol

To validate the results obtained from the screen, individual targeted genetic knockouts were generated using CRISPR/Cas9 and plasmid expressing sgRNAs targeting the gene of interest. Two sgRNA sequences were used for each target gene, one was selected from the CRISPR library v1 while another sequence was designed using the Genetic Perturbation Platform (GPP) web portal (https://portals.broadinstitute.org/gpp/public/analysis-tools/sgrna-design). List of sgRNA used and their sequences can be found in Supplementary file 11. pKLV2-U6gRNA5(BbsI)-PGKpuro2ABFP-W and pKLV2-U6gRNA5(BbsI) PGKpuro2AmCherry-W were linearized using BbsI enzyme (NEB R0539S) and the concentration was adjusted to 20 ng/μL. Target oligos were phosphorylated and annealed using T4 PNK (NEB M0201). The thermocycler condition used are as follows: 37°C for 30 min, 95°C for 5 min, followed by a ramp down to 25°C at 0.1°C /s. The annealed oligos were next diluted twice for prior to ligation: 1st dilution = 139 μL EB buffer + 2 μL of 10 μM double-stranded oligos; 2nd dilution = 57 μL EB buffer + 3 μL of 1st dilution. Following that, overnight ligation at 4°C was carried out using T4 ligase (NEB M0202S) and 10X ligase buffer (NEB M0202S). The ligation products were then transformed into DH5α chemically- generated competent cells and plated onto Luria Broth (LB) agar plates containing 100 μg/ml Ampicillin. The plasmids were then extracted using QIAprep Spin Miniprep Kit (QIAGEN) and the sgRNA sequences were verified by Sanger sequencing prior to use.

Co-competition assay

Request a detailed protocol

The relative growth rate of sgRNA-transduced and non-transduced cells was compared using co-competiton assay, as described previously (Behan et al., 2019; Tzelepis et al., 2016). Briefly, in order to achieve single gene-specific knockout, the Cas9-expressing cell lines were transduced at 30–70% transduction efficiency, with lentivirus carrying gene-specific sgRNA in pKLV2-U6gRNA5(BbsI)-PGKpuro2ABFP-W. Using flow cytometry, the percentage of BFP-positive sgRNA-transduced cells was measured between day 4 and day 18 post-transduction. The results obtained from days 8, 11, 15 to 18 were normalized to the percentage of BFP-positive transduced cells on day 4 or 6 to investigate the relative growth changes of the transduced population following gene depletion. For each target gene (YAP1 and WWTR1), two different sgRNA were used, one from the Kosuke Yusa’s CRISPR Library v1 (‘Y1K’ – sgRNA targeting YAP1; ‘W1K – sgRNA targeting WWTR1’) and another independently designed sgRNA using Broad’s sgRNA-designer tool (‘Y2B’ – sgRNA targeting YAP1; ‘W2B’ – sgRNA targeting WWTR1). sgRNA targeting a core fitness gene, Polo-like kinase 1 (PLK1) was included as a positive control, choline acetyltransferase (CHAT)-targeting sgRNA was used as a non-fitness gene negative control and a non-targeting (NT) sgRNA w also included.

To achieve double gene knockout, sgRNAs targeting two different genes were cloned into either one of the plasmids tagged with BFP or mCherry (pKLV2-U6gRNA5(BbsI)-PGKpuro2ABFP-W or pKLV2-U6gRNA5(BbsI)-PGKpuro2AmCherry-W). Thereafter, the changes in the BFP- and mCherry double-positive cell population was measured as mentioned above.

Lysates preparation and western blotting

Request a detailed protocol

To determine the baseline protein expression level, OSCC parental cell lines were seeded in 100 mm3 dish and cultured until they reached 70–80% confluency. To assess the differential expression of the protein of interest after gene knockdown, Cas9-expressing cells were transduced with the target sgRNA at above 90% transduction efficiency. Next day, transduced cells were selected using 2 μg/ml of Puromycin. Day four post-transduction, percentage of BFP-expressing cells were determined using flow cytometry and total cell lysates (TCL) were extracted with RIPA buffer (50 mM Tris pH8, 1% (v/v) NP-40, 0.5% (w/v) sodium deoxycholate, 0.1% (w/v) SDS, 150 mM NaCl) supplemented with Halt Protease and Phosphatase Inhibitor (PI) Cocktail (Pierce Biotechnology) on ice. TCL was collected by centrifugation and quantified using the BCA method (Thermo Fisher Scientific). About 20 µg of the TCL was resolved on SDS-PAGE gel and proteins were transferred onto PVDF membranes (Millipore). Membranes were blocked with 5% (w/v) milk in TBST (0.1% (v/v) Tween 20) and probed with primary antibodies (1:1000 dilution in 1% (w/v) BSA) overnight at 4°C. Horseradish peroxidase (HRP)-conjugated secondary antibodies (1:10,000 dilution in 5% milk) were probed for one hour at room temperature. For signal development, WesternBright Quantum HRP substrate (Advansta Inc) was used and visualized using the FluorChem HD2 imaging systems (Alpha Innotech). To normalize for loading, the blots were re-probed with an anti-tubulin monoclonal antibody (1:1000 dilution in 1% BSA) and processed as described above. List of primary and secondary antibodies used is found in Supplementary file 12. Uncropped western blot images can be found in Supplementary file 13.

Total RNA extraction and qPCR

Request a detailed protocol

Cas9-expressing cells were transduced with selected sgRNA to achieve single gene or double gene knockout, as described above. On day 4 post-transduction, total RNA was extracted using TRIzol Reagent (Thermo Fisher Scientific). Total RNA (1 μg) was used for reverse transcription to complementary DNA (cDNA) using high-capacity cDNA reverse transcription kit (Applied Biosystems). Real-time quantitative PCR was performed using 1 μl of 5x diluted cDNA with PowerUp SYBR Green Master Mix and corresponding primers in 7500 Real-Time PCR System (Applied Biosystem). All reactions were performed in technical triplicates and repeated twice. Cycling conditions used are as follows: 50°C for 2 min, 95°C for 2 min, 40 cycles of 95°C for 15 s, and 60°C for 1 min. A default melt curve stage was included to allow inspection of primer specificity. Ribosomal protein L13 (RPL13) was used as an endogenous reference control for normalization. Sequences of all primers used can be found in Supplementary file 9.

Clonogenic assay

Request a detailed protocol

On 4- or 6 days post-transduction of sgRNA-containing lentiviruses, 2000 cells were seeded into six-well plate. After a week, cells were fixed using ice-cold methanol followed by staining with crystal violet solution.

Cell viability assay

Request a detailed protocol

The 3-(4,5-dimethylthiazol-2-yl)−2,5-diphenyltetrazolium bromide (MTT) assay was used to access the effect of target gene knockdown/knockout on cell viability. Briefly, 4- or 6 days post-transduction with sgRNA, 2000 cells were seeded in triplicates in 96-well plate. 72 hr later, 50 μl MTT was added to each well and incubated for 4 hr at 37°C. After removing the media, 150 μl dimethyl sulfoxide was added to dissolve the formazan crystal and optical density was measured using Synergy H1M microplate reader (BioTek Instruments, USA) at 570 nm.

Apoptosis assay

Request a detailed protocol

On 4- or 6 days post-transduction of sgRNA-containing lentiviruses, 30,000 cells were seeded into 24-well plate and harvested after 72 hr. Cells were pelleted down and washed with PBS. For detection of apoptotic cells, the cell pellet was resuspended in 1x Annexin V buffer containing 2.5 μl of Annexin V solution (BD Biosciences) and 2.5 μl of propidium iodide and incubated for 15 min in the dark. The proportion of apoptotic cells were analyzed using the LSR Fortessa X-20 cell analyser (BD Biosciences) and FlowJo (version 10.5.3, BD Biosciences), considering all single- and double-stained cells as apoptotic cells. Gating strategy for detection of apoptotic cells using flow cytometry can be found in Supplementary file 8.

DEGs signature based on dependency on YAP1 or WWTR1

Request a detailed protocol

Three representative cell lines with validated dependency on YAP1 or WWTR1 were used to derive gene expression signatures. ‘YAP1-dependent’ – ORL-48, ORL-204, SAS; ‘WWTR1-dependent’ – ORL-214, PE/CA-PJ15, ORL-174; ‘Compensable’ – BICR10, HSC-2 and HSC-4. DEGs for each group were computed using the limma package (Bioconductor) on the iRAP-processed, ComBat corrected FPKM matrix for these nine cell lines. Non-overlapping DEGs with significant p-value threshold <0.01 and log fold change >2 were retained. The final list of DEGs is found in Supplementary file 6. Gene expression data of the HNSCC cohort in TCGA in the form of RSEM was downloaded from cbioportal (https://www.cbioportal.org/) (Gao et al., 2013). Gene expressions of all DEGs were then extracted for the 315 OSCC samples.

For all DEGs, Z-score was computed and a ‘dependency signature score’/ ‘compensable signature score’ was generated for each cell line/tumor sample, taking the difference between the average of all Z-score of upregulated DEGs and that of downregulated DEGs. For example, ‘YAP1-dependency signature score’ = (average of all Z-score of upregulated DEGs among YAP1-dependent cell lines) - (average of all Z-score of downregulated DEGs among YAP1-dependent cell lines).

Subsequently, cell lines and OSCC tumors were analyzed using hierarchical clustering and visualize with heatmap (generated using Morpheus, Broad Institute: https://software.broadinstitute.org/morpheus) using the ‘YAP1 dependency signature score’, ‘WWTR1 dependency signature score’ and ‘Compensable signature score’.

Correlation analysis and gene set enrichment analysis

Request a detailed protocol

Core OSCC samples with >0.5 dependency signature score in one of the three groups were identified. There was a total of 43 OSCC samples with high YAP1 dependency signature score; 31 with high WWTR1 dependency signature score and 30 with high Compensable signature score. Clinical and genomic data of these core OSCC samples were accessed from cBioPortal (Gao et al., 2013). Gene set enrichment analyses (GSEA) for the three groups of core OSCC samples were performed using the Broad Institute’s Molecular Signatures Database (MSigDB) hallmark gene sets as reference database (Liberzon et al., 2015). Signatures reflective of the immune landscape of these core OSCC samples were extracted from the supplementary table 1 of Thorsson et al., 2018. Single-sample GSEA (ssGSEA) was performed using the GenePatterns web-tool (https://www.genepattern.org/) (Reich et al., 2006) with the 18-genes T-cell inflamed Gene Expression Profile (GEP) gene set, which was found to be predictive biomarkers for response to pembrolizumab in HNSCC clinical trial (Ayers et al., 2017; Cristescu et al., 2018).

Statistical analysis

Request a detailed protocol

All statistical significance analyses were performed using unpaired parametric two-tailed t-test in GraphPad Prism (version 8, GraphPad Software Inc) unless otherwise stated. Unpaired t-test with Welch’s correction (Welch’s t-test) was used for all analyses in Figure 6 and Figure 6—figure supplement 2 due to unequal sample size. For estimation of the Pearson correlation, the cor.test function in Rstudio (version 1.2.1335, Rstudio Inc) was used.

Code availability

Request a detailed protocol

No unreported or custom code was used in this study. Open source softwares were used for data analysis and codes are available upon request.

Appendix 1

Key Resources Table

Appendix 1—key resources table
Reagent type (species) or resourceDesignationSource or referenceIdentifiersAdditional information
Recombinant DNA reagentHuman Improved Genome-wide Knockout CRISPR Library v1
(pooled library)
Addgene#67989Containing 90,709 gRNAs, selectable with puromycin
Recombinant DNA reagentpKLV2-EF1α-Cas9Bsd-W (plasmid)Addgene#68343To express Cas9 constitutively, selectable with blasticidin
Recombinant DNA reagentpKLV2-U6gRNA5 (gGFP)-PGKmCherry2AGFP-W
(plasmid)
Addgene#67982Reporter plasmid
Recombinant DNA reagentpKLV2-
U6gRNA5(BbsI)-PGKpuro2ABFP-W (plasmid)
Addgene#67974For cloning of single sgRNAs, tag with blue fluorescence protein (BFP)
Recombinant DNA reagentpKLV2-U6gRNA5(BbsI)-PGKpuro2AmCherry-W (plasmid)Addgene#67977For cloning of single sgRNAs, tag with mcherry
Recombinant DNA reagentpMD2.G (plasmid)Addgene#12259For lentivirus generation
Recombinant DNA reagentpsPAX2 (plasmid)Addgene#12260For lentivirus generation
Chemical compound, drugjetPRIME reagent and bufferPolyplus transfection114–75For transfection
Cell line (Homo sapiens)Cells derived from oral squamous cell carcinomaCancer Research Malaysia Fadlullah et al., 2016, PMID:27050151ORL-48Maintained in DMEM/F12 supplemented with 10% FBS, 1% penicillin/streptomycin
Cell line (Homo sapiens)Cells derived from oral squamous cell carcinomaCancer Research Malaysia Fadlullah et al., 2016, PMID:27050151ORL-115Maintained in DMEM/F12 supplemented with 10% FBS, 1% penicillin/streptomycin
Cell line (Homo sapiens)Cells derived from oral squamous cell carcinomaCancer Research Malaysia Fadlullah et al., 2016, PMID:27050151ORL-136Maintained in DMEM/F12 supplemented with 10% FBS, 1% penicillin/streptomycin
Cell line (Homo sapiens)Cells derived from oral squamous cell carcinomaCancer Research Malaysia Fadlullah et al., 2016, PMID:27050151ORL-150Maintained in DMEM/F12 supplemented with 10% FBS, 1% penicillin/streptomycin
Cell line (Homo sapiens)Cells derived from oral squamous cell carcinomaCancer Research Malaysia Fadlullah et al., 2016, PMID:27050151ORL-153Maintained in DMEM/F12 supplemented with 10% FBS, 1% penicillin/streptomycin
Cell line (Homo sapiens)Cells derived from oral squamous cell carcinomaCancer Research Malaysia Fadlullah et al., 2016, PMID:27050151ORL-156Maintained in DMEM/F12 supplemented with 10% FBS, 1% penicillin/streptomycin
Cell line (Homo sapiens)Cells derived from oral squamous cell carcinomaCancer Research Malaysia Fadlullah et al., 2016, PMID:27050151ORL-166Maintained in DMEM/F12 supplemented with 10% FBS, 1% penicillin/streptomycin
Cell line (Homo sapiens)Cells derived from oral squamous cell carcinomaCancer Research Malaysia Fadlullah et al., 2016, PMID:27050151ORL-174Maintained in DMEM/F12 supplemented with 10% FBS, 1% penicillin/streptomycin
Cell line (Homo sapiens)Cells derived from oral squamous cell carcinomaCancer Research Malaysia Fadlullah et al., 2016, PMID:27050151ORL-188Maintained in DMEM/F12 supplemented with 10% FBS, 1% penicillin/streptomycin
Cell line (Homo sapiens)Cells derived from oral squamous cell carcinomaCancer Research Malaysia Fadlullah et al., 2016, PMID:27050151ORL-195Maintained in DMEM/F12 supplemented with 10% FBS, 1% penicillin/streptomycin
Cell line (Homo sapiens)Cells derived from oral squamous cell carcinomaCancer Research Malaysia Fadlullah et al., 2016, PMID:27050151ORL-204Maintained in DMEM/F12 supplemented with 10% FBS, 1% penicillin/streptomycin
Cell line (Homo sapiens)Cells derived from oral squamous cell carcinomaCancer Research Malaysia Fadlullah et al., 2016, PMID:27050151ORL-207Maintained in DMEM/F12 supplemented with 10% FBS, 1% penicillin/streptomycin
Cell line (Homo sapiens)Cells derived from oral squamous cell carcinomaCancer Research Malaysia Fadlullah et al., 2016, PMID:27050151ORL-214Maintained in DMEM/F12 supplemented with 10% FBS, 1% penicillin/streptomycin
Cell line (Homo sapiens)Cells derived from oral squamous cell carcinomaCancer Research Malaysia Fadlullah et al., 2016, PMID:27050151ORL-215Maintained in DMEM/F12 supplemented with 10% FBS, 1% penicillin/streptomycin
Cell line (Homo sapiens)Cells derived from oral squamous cell carcinomaWellcome Sanger InstituteBICR10Maintained in DMEM/F12 supplemented with 10% FBS, 1% penicillin/streptomycin
Cell line (Homo sapiens)Cells derived from oral squamous cell carcinomaWellcome Sanger InstituteHO-1-u-1Maintained in DMEM/F12 supplemented with 10% FBS, 1% penicillin/streptomycin
Cell line (Homo sapiens)Cells derived from oral squamous cell carcinomaWellcome Sanger InstituteHSC-2Maintained in DMEM/F12 supplemented with 10% FBS, 1% penicillin/streptomycin
Cell line (Homo sapiens)Cells derived from oral squamous cell carcinomaWellcome Sanger InstituteHSC-4Maintained in DMEM/F12 supplemented with 10% FBS, 1% penicillin/streptomycin
Cell line (Homo sapiens)Cells derived from oral squamous cell carcinomaWellcome Sanger InstitutePE/CA-PJ15Maintained in DMEM/F12 supplemented with 10% FBS, 1% penicillin/streptomycin
Cell line (Homo sapiens)Cells derived from oral squamous cell carcinomaWellcome Sanger InstituteSASMaintained in DMEM/F12 supplemented with 10% FBS, 1% penicillin/streptomycin
Cell line (Homo sapiens)Cells derived from oral squamous cell carcinomaWellcome Sanger InstituteSCC-9Maintained in DMEM/F12 supplemented with 10% FBS, 1% penicillin/streptomycin
Cell line (Homo sapiens)Human embryonic kidney cellsWellcome Sanger InstituteHEK293Maintained in DMEM (high glucose) supplemented with 10% FBS, 1% penicillin/streptomycin
AntibodyAnti-human YAP (D8H1X), Rabbit monoclonalCell Signaling Technology#14074WB - (1:1000)
AntibodyAnti-human TAZ (D361D), Rabbit monoclonalCell Signaling Technology#70148WB - (1:1000)
AntibodyAnti-human actin (clone C4), Mouse monoclonalMerck MiliporeMAB1501WB – (1:5000)
AntibodyGoat anti-rabbit IgG-HRPSouthern BiotecSB4010-05Secondary antibody; WB – (1:10000)
AntibodyGoat anti-mouse IgG-HRPSouthern BiotecSB1010-05Secondary antibody; WB – (1:10000)
Commercial assay or kitQ5 Hot Start High-Fidelity 2X Master Mix
New England BiolabsM0494S
Commercial assay or kitQIAquick PCR purification kitQiagen28104
Commercial assay or kitQubit dsDNA broad-range (BR) assay kitThermo Fisher ScientificQ32850
Commercial assay or kitKAPA HotStart ReadyMixRocheKK2601
Commercial assay or kitAgilent High Sensitivity DNA kitAgilent technologies5067–4626
Commercial assay or kitQIAprep Spin Miniprep KitQiagen27104
Commercial assay or kitcDNA reverse transcription kitApplied Biosystems, Thermo Fisher Scientific4368813
Commercial assay or kitPowerUp SYBR Green Master MixApplied Biosystems, Thermo Fisher ScientificA25742
Commercial assay or kitFITC Annexin V Apoptosis Detection Kit IBD Bioscience556547
Chemical compound, drugBbsI enzymeNew England BiolabsR0539S
Chemical compound, drugT4 Polynucleotide kinase enzymeNew England BiolabsM0201
Chemical compound, drugT4 ligase and 10x ligase bufferNew England BiolabsM0202S
Chemical compound, drugWesternBright Quantum HRP substrateAdvanstaK-12042-D20Substrate for Western Blot
Chemical compound, drugTRIzol ReagentThermo Fisher Scientific15596026
Software, algorithmCRISPRcleanR (version 0.5)Iorio et al., 2018 PMID:30103702
Software, algorithmMAGeCK (version 0.5.7)Li et al., 2014 PMID:25476604
Software, algorithmRstudio (version 1.2.1335)Rstudio Inc
Software, algorithmGraphPad Prism (version 8)GraphPad Software Inc

References

  1. 1
  2. 2
  3. 3
  4. 4
  5. 5
  6. 6
  7. 7
    Oral squamous cell carcinoma invasion is associated with a laminin-5 matrix re-organization but independent of basement membrane and hemidesmosome formation clues from an in vitro invasion model
    1. A Berndt
    2. P Hyckel
    3. A Könneker
    4. D Katenkamp
    5. H Kosmehl
    (1997)
    Invasion & Metastasis 17:251–259.
  8. 8
  9. 9
  10. 10
    Pembrolizumab alone or with chemotherapy versus cetuximab with chemotherapy for recurrent or metastatic squamous cell carcinoma of the head and neck (KEYNOTE-048): a randomised, open-label, phase 3 study
    1. B Burtness
    2. KJ Harrington
    3. R Greil
    4. D Soulières
    5. M Tahara
    6. G de Castro
    7. A Psyrri
    8. N Basté
    9. P Neupane
    10. Åse Bratland
    11. T Fuereder
    12. BGM Hughes
    13. R Mesía
    14. N Ngamphaiboon
    15. T Rordorf
    16. WZ Wan Ishak
    17. R-L Hong
    18. R González Mendoza
    19. A Roy
    20. Y Zhang
    21. B Gumuscu
    22. JD Cheng
    23. F Jin
    24. D Rischin
    25. G Lerzo
    26. M Tatangelo
    27. M Varela
    28. JJ Zarba
    29. M Boyer
    30. H Gan
    31. B Gao
    32. B Hughes
    33. G Mallesara
    34. D Rischin
    35. A Taylor
    36. M Burian
    37. T Fuereder
    38. R Greil
    39. CH Barrios
    40. DO de Castro Junior
    41. G Castro
    42. FA Franke
    43. G Girotto
    44. IPF Lima
    45. UR Nicolau
    46. GDJ Pinto
    47. L Santos
    48. A-P Victorino
    49. N Chua
    50. F Couture
    51. R Gregg
    52. A Hansen
    53. J Hilton
    54. J McCarthy
    55. D Soulieres
    56. R Ascui
    57. P Gonzalez
    58. L Villanueva
    59. M Torregroza
    60. A Zambrano
    61. P Holeckova
    62. Z Kral
    63. B Melichar
    64. J Prausova
    65. M Vosmik
    66. M Andersen
    67. N Gyldenkerne
    68. H Jurgens
    69. K Putnik
    70. P Reinikainen
    71. V Gruenwald
    72. S Laban
    73. G Aravantinos
    74. I Boukovinas
    75. V Georgoulias
    76. A Psyrri
    77. D Kwong
    78. Y Al-Farhat
    79. T Csoszi
    80. J Erfan
    81. G Horvai
    82. L Landherr
    83. E Remenar
    84. A Ruzsa
    85. J Szota
    86. S Billan
    87. I Gluck
    88. O Gutfeld
    89. A Popovtzer
    90. M Benasso
    91. S Bui
    92. V Ferrari
    93. L Licitra
    94. F Nole
    95. T Fujii
    96. Y Fujimoto
    97. N Hanai
    98. H Hara
    99. K Matsumoto
    100. K Mitsugi
    101. N Monden
    102. M Nakayama
    103. K Okami
    104. N Oridate
    105. K Shiga
    106. Y Shimizu
    107. M Sugasawa
    108. M Tahara
    109. M Takahashi
    110. S Takahashi
    111. K Tanaka
    112. T Ueda
    113. H Yamaguchi
    114. T Yamazaki
    115. R Yasumatsu
    116. T Yokota
    117. T Yoshizaki
    118. I Kudaba
    119. Z Stara
    120. WZ Wan Ishak
    121. SK Cheah
    122. J Aguilar Ponce
    123. R Gonzalez Mendoza
    124. C Hernandez Hernandez
    125. F Medina Soto
    126. J Buter
    127. A Hoeben
    128. S Oosting
    129. K Suijkerbuijk
    130. A Bratland
    131. M Brydoey
    132. R Alvarez
    133. L Mas
    134. P Caguioa
    135. J Querol
    136. EE Regala
    137. MB Tamayo
    138. EM Villegas
    139. A Kawecki
    140. A Karpenko
    141. A Klochikhin
    142. A Smolin
    143. O Zarubenkov
    144. BC Goh
    145. G Cohen
    146. J du Toit
    147. C Jordaan
    148. G Landers
    149. P Ruff
    150. W Szpak
    151. N Tabane
    152. I Brana
    153. L Iglesias Docampo
    154. J Lavernia
    155. R Mesia
    156. E Abel
    157. V Muratidu
    158. N Nielsen
    159. V Cristina
    160. T Rordorf
    161. S Rothschild
    162. R-L Hong
    163. H-M Wang
    164. M-H Yang
    165. S-P Yeh
    166. C-J Yen
    167. N Ngamphaiboon
    168. N Soparattanapaisarn
    169. V Sriuranpong
    170. S Aksoy
    171. I Cicin
    172. M Ekenel
    173. H Harputluoglu
    174. O Ozyilkan
    175. K Harrington
    176. S Agarwala
    177. H Ali
    178. R Alter
    179. D Anderson
    180. J Bruce
    181. B Burtness
    182. N Campbell
    183. M Conde
    184. J Deeken
    185. W Edenfield
    186. L Feldman
    187. E Gaughan
    188. B Goueli
    189. B Halmos
    190. U Hegde
    191. B Hunis
    192. R Jotte
    193. A Karnad
    194. S Khan
    195. N Laudi
    196. D Laux
    197. D Martincic
    198. S McCune
    199. D McGaughey
    200. K Misiukiewicz
    201. D Mulford
    202. E Nadler
    203. P Neupane
    204. J Nunnink
    205. J Ohr
    206. M O'Malley
    207. B Patson
    208. D Paul
    209. E Popa
    210. S Powell
    211. R Redman
    212. V Rella
    213. C Rocha Lima
    214. A Sivapiragasam
    215. Y Su
    216. A Sukari
    217. S Wong
    218. E Yilmaz
    219. J Yorio
    (2019)
    The Lancet 394:1915–1928.
    https://doi.org/10.1016/S0140-6736(19)32591-7
  11. 11
  12. 12
    Genomic, pathway network, and immunologic features distinguishing squamous carcinomas
    1. JD Campbell
    2. C Yau
    3. R Bowlby
    4. Y Liu
    5. K Brennan
    6. H Fan
    7. AM Taylor
    8. C Wang
    9. V Walter
    10. R Akbani
    11. LA Byers
    12. CJ Creighton
    13. C Coarfa
    14. J Shih
    15. AD Cherniack
    16. O Gevaert
    17. M Prunello
    18. H Shen
    19. P Anur
    20. J Chen
    21. H Cheng
    22. DN Hayes
    23. S Bullman
    24. CS Pedamallu
    25. AI Ojesina
    26. S Sadeghi
    27. KL Mungall
    28. AG Robertson
    29. C Benz
    30. A Schultz
    31. RS Kanchi
    32. CM Gay
    33. A Hegde
    34. L Diao
    35. J Wang
    36. W Ma
    37. P Sumazin
    38. H-S Chiu
    39. T-W Chen
    40. P Gunaratne
    41. L Donehower
    42. JS Rader
    43. R Zuna
    44. H Al-Ahmadie
    45. AJ Lazar
    46. ER Flores
    47. KY Tsai
    48. JH Zhou
    49. AK Rustgi
    50. E Drill
    51. R Shen
    52. CK Wong
    53. JM Stuart
    54. PW Laird
    55. KA Hoadley
    56. JN Weinstein
    57. M Peto
    58. CR Pickering
    59. Z Chen
    60. C Van Waes
    61. SJ Caesar-Johnson
    62. JA Demchok
    63. I Felau
    64. M Kasapi
    65. ML Ferguson
    66. CM Hutter
    67. HJ Sofia
    68. R Tarnuzzer
    69. Z Wang
    70. L Yang
    71. JC Zenklusen
    72. JJ Zhang
    73. S Chudamani
    74. J Liu
    75. L Lolla
    76. R Naresh
    77. T Pihl
    78. Q Sun
    79. Y Wan
    80. Y Wu
    81. J Cho
    82. T DeFreitas
    83. S Frazer
    84. N Gehlenborg
    85. G Getz
    86. DI Heiman
    87. J Kim
    88. MS Lawrence
    89. P Lin
    90. S Meier
    91. MS Noble
    92. G Saksena
    93. D Voet
    94. H Zhang
    95. B Bernard
    96. N Chambwe
    97. V Dhankani
    98. T Knijnenburg
    99. R Kramer
    100. K Leinonen
    101. Y Liu
    102. M Miller
    103. S Reynolds
    104. I Shmulevich
    105. V Thorsson
    106. W Zhang
    107. R Akbani
    108. BM Broom
    109. AM Hegde
    110. Z Ju
    111. RS Kanchi
    112. A Korkut
    113. J Li
    114. H Liang
    115. S Ling
    116. W Liu
    117. Y Lu
    118. GB Mills
    119. K-S Ng
    120. A Rao
    121. M Ryan
    122. J Wang
    123. JN Weinstein
    124. J Zhang
    125. A Abeshouse
    126. J Armenia
    127. D Chakravarty
    128. WK Chatila
    129. I de Bruijn
    130. J Gao
    131. BE Gross
    132. ZJ Heins
    133. R Kundra
    134. K La
    135. M Ladanyi
    136. A Luna
    137. MG Nissan
    138. A Ochoa
    139. SM Phillips
    140. E Reznik
    141. F Sanchez-Vega
    142. C Sander
    143. N Schultz
    144. R Sheridan
    145. SO Sumer
    146. Y Sun
    147. BS Taylor
    148. J Wang
    149. H Zhang
    150. P Anur
    151. M Peto
    152. P Spellman
    153. C Benz
    154. JM Stuart
    155. CK Wong
    156. C Yau
    157. DN Hayes
    158. JS Parker
    159. MD Wilkerson
    160. A Ally
    161. M Balasundaram
    162. R Bowlby
    163. D Brooks
    164. R Carlsen
    165. E Chuah
    166. N Dhalla
    167. R Holt
    168. SJM Jones
    169. K Kasaian
    170. D Lee
    171. Y Ma
    172. MA Marra
    173. M Mayo
    174. RA Moore
    175. AJ Mungall
    176. K Mungall
    177. AG Robertson
    178. S Sadeghi
    179. JE Schein
    180. P Sipahimalani
    181. A Tam
    182. N Thiessen
    183. K Tse
    184. T Wong
    185. AC Berger
    186. R Beroukhim
    187. AD Cherniack
    188. C Cibulskis
    189. SB Gabriel
    190. GF Gao
    191. G Ha
    192. M Meyerson
    193. SE Schumacher
    194. J Shih
    195. MH Kucherlapati
    196. RS Kucherlapati
    197. S Baylin
    198. L Cope
    199. L Danilova
    200. MS Bootwalla
    201. PH Lai
    202. DT Maglinte
    203. DJ Van Den Berg
    204. DJ Weisenberger
    205. JT Auman
    206. S Balu
    207. T Bodenheimer
    208. C Fan
    209. KA Hoadley
    210. AP Hoyle
    211. SR Jefferys
    212. CD Jones
    213. S Meng
    214. PA Mieczkowski
    215. LE Mose
    216. AH Perou
    217. CM Perou
    218. J Roach
    219. Y Shi
    220. JV Simons
    221. T Skelly
    222. MG Soloway
    223. D Tan
    224. U Veluvolu
    225. H Fan
    226. T Hinoue
    227. PW Laird
    228. H Shen
    229. W Zhou
    230. M Bellair
    231. K Chang
    232. K Covington
    233. CJ Creighton
    234. H Dinh
    235. H Doddapaneni
    236. LA Donehower
    237. J Drummond
    238. RA Gibbs
    239. R Glenn
    240. W Hale
    241. Y Han
    242. J Hu
    243. V Korchina
    244. S Lee
    245. L Lewis
    246. W Li
    247. X Liu
    248. M Morgan
    249. D Morton
    250. D Muzny
    251. J Santibanez
    252. M Sheth
    253. E Shinbrot
    254. L Wang
    255. M Wang
    256. DA Wheeler
    257. L Xi
    258. F Zhao
    259. J Hess
    260. EL Appelbaum
    261. M Bailey
    262. MG Cordes
    263. L Ding
    264. CC Fronick
    265. LA Fulton
    266. RS Fulton
    267. C Kandoth
    268. ER Mardis
    269. MD McLellan
    270. CA Miller
    271. HK Schmidt
    272. RK Wilson
    273. D Crain
    274. E Curley
    275. J Gardner
    276. K Lau
    277. D Mallery
    278. S Morris
    279. J Paulauskis
    280. R Penny
    281. C Shelton
    282. T Shelton
    283. M Sherman
    284. E Thompson
    285. P Yena
    286. J Bowen
    287. JM Gastier-Foster
    288. M Gerken
    289. KM Leraas
    290. TM Lichtenberg
    291. NC Ramirez
    292. L Wise
    293. E Zmuda
    294. N Corcoran
    295. T Costello
    296. C Hovens
    297. AL Carvalho
    298. AC de Carvalho
    299. JH Fregnani
    300. A Longatto-Filho
    301. RM Reis
    302. C Scapulatempo-Neto
    303. HCS Silveira
    304. DO Vidal
    305. A Burnette
    306. J Eschbacher
    307. B Hermes
    308. A Noss
    309. R Singh
    310. ML Anderson
    311. PD Castro
    312. M Ittmann
    313. D Huntsman
    314. B Kohl
    315. X Le
    316. R Thorp
    317. C Andry
    318. ER Duffy
    319. V Lyadov
    320. O Paklina
    321. G Setdikova
    322. A Shabunin
    323. M Tavobilov
    324. C McPherson
    325. R Warnick
    326. R Berkowitz
    327. D Cramer
    328. C Feltmate
    329. N Horowitz
    330. A Kibel
    331. M Muto
    332. CP Raut
    333. A Malykh
    334. JS Barnholtz-Sloan
    335. W Barrett
    336. K Devine
    337. J Fulop
    338. QT Ostrom
    339. K Shimmel
    340. Y Wolinsky
    341. AE Sloan
    342. A De Rose
    343. F Giuliante
    344. M Goodman
    345. BY Karlan
    346. CH Hagedorn
    347. J Eckman
    348. J Harr
    349. J Myers
    350. K Tucker
    351. LA Zach
    352. B Deyarmin
    353. H Hu
    354. L Kvecher
    355. C Larson
    356. RJ Mural
    357. S Somiari
    358. A Vicha
    359. T Zelinka
    360. J Bennett
    361. M Iacocca
    362. B Rabeno
    363. P Swanson
    364. M Latour
    365. L Lacombe
    366. B Têtu
    367. A Bergeron
    368. M McGraw
    369. SM Staugaitis
    370. J Chabot
    371. H Hibshoosh
    372. A Sepulveda
    373. T Su
    374. T Wang
    375. O Potapova
    376. O Voronina
    377. L Desjardins
    378. O Mariani
    379. S Roman-Roman
    380. X Sastre
    381. M-H Stern
    382. F Cheng
    383. S Signoretti
    384. A Berchuck
    385. D Bigner
    386. E Lipp
    387. J Marks
    388. S McCall
    389. R McLendon
    390. A Secord
    391. A Sharp
    392. M Behera
    393. DJ Brat
    394. A Chen
    395. K Delman
    396. S Force
    397. F Khuri
    398. K Magliocca
    399. S Maithel
    400. JJ Olson
    401. T Owonikoko
    402. A Pickens
    403. S Ramalingam
    404. DM Shin
    405. G Sica
    406. EG Van Meir
    407. H Zhang
    408. W Eijckenboom
    409. A Gillis
    410. E Korpershoek
    411. L Looijenga
    412. W Oosterhuis
    413. H Stoop
    414. KE van Kessel
    415. EC Zwarthoff
    416. C Calatozzolo
    417. L Cuppini
    418. S Cuzzubbo
    419. F DiMeco
    420. G Finocchiaro
    421. L Mattei
    422. A Perin
    423. B Pollo
    424. C Chen
    425. J Houck
    426. P Lohavanichbutr
    427. A Hartmann
    428. C Stoehr
    429. R Stoehr
    430. H Taubert
    431. S Wach
    432. B Wullich
    433. W Kycler
    434. D Murawa
    435. M Wiznerowicz
    436. K Chung
    437. WJ Edenfield
    438. J Martin
    439. E Baudin
    440. G Bubley
    441. R Bueno
    442. A De Rienzo
    443. WG Richards
    444. S Kalkanis
    445. T Mikkelsen
    446. H Noushmehr
    447. L Scarpace
    448. N Girard
    449. M Aymerich
    450. E Campo
    451. E Giné
    452. AL Guillermo
    453. N Van Bang
    454. PT Hanh
    455. BD Phu
    456. Y Tang
    457. H Colman
    458. K Evason
    459. PR Dottino
    460. JA Martignetti
    461. H Gabra
    462. H Juhl
    463. T Akeredolu
    464. S Stepa
    465. D Hoon
    466. K Ahn
    467. KJ Kang
    468. F Beuschlein
    469. A Breggia
    470. M Birrer
    471. D Bell
    472. M Borad
    473. AH Bryce
    474. E Castle
    475. V Chandan
    476. J Cheville
    477. JA Copland
    478. M Farnell
    479. T Flotte
    480. N Giama
    481. T Ho
    482. M Kendrick
    483. J-P Kocher
    484. K Kopp
    485. C Moser
    486. D Nagorney
    487. D O’Brien
    488. BP O’Neill
    489. T Patel
    490. G Petersen
    491. F Que
    492. M Rivera
    493. L Roberts
    494. R Smallridge
    495. T Smyrk
    496. M Stanton
    497. RH Thompson
    498. M Torbenson
    499. JD Yang
    500. L Zhang
    501. F Brimo
    502. JA Ajani
    503. AMA Gonzalez
    504. C Behrens
    505. J Bondaruk
    506. R Broaddus
    507. B Czerniak
    508. B Esmaeli
    509. J Fujimoto
    510. J Gershenwald
    511. C Guo
    512. AJ Lazar
    513. C Logothetis
    514. F Meric-Bernstam
    515. C Moran
    516. L Ramondetta
    517. D Rice
    518. A Sood
    519. P Tamboli
    520. T Thompson
    521. P Troncoso
    522. A Tsao
    523. I Wistuba
    524. C Carter
    525. L Haydu
    526. P Hersey
    527. V Jakrot
    528. H Kakavand
    529. R Kefford
    530. K Lee
    531. G Long
    532. G Mann
    533. M Quinn
    534. R Saw
    535. R Scolyer
    536. K Shannon
    537. A Spillane
    538. O Stretch
    539. M Synott
    540. J Thompson
    541. J Wilmott
    542. H Al-Ahmadie
    543. TA Chan
    544. R Ghossein
    545. A Gopalan
    546. DA Levine
    547. V Reuter
    548. S Singer
    549. B Singh
    550. NV Tien
    551. T Broudy
    552. C Mirsaidi
    553. P Nair
    554. P Drwiega
    555. J Miller
    556. J Smith
    557. H Zaren
    558. J-W Park
    559. NP Hung
    560. E Kebebew
    561. WM Linehan
    562. AR Metwalli
    563. K Pacak
    564. PA Pinto
    565. M Schiffman
    566. LS Schmidt
    567. CD Vocke
    568. N Wentzensen
    569. R Worrell
    570. H Yang
    571. M Moncrieff
    572. C Goparaju
    573. J Melamed
    574. H Pass
    575. N Botnariuc
    576. I Caraman
    577. M Cernat
    578. I Chemencedji
    579. A Clipca
    580. S Doruc
    581. G Gorincioi
    582. S Mura
    583. M Pirtac
    584. I Stancul
    585. D Tcaciuc
    586. M Albert
    587. I Alexopoulou
    588. A Arnaout
    589. J Bartlett
    590. J Engel
    591. S Gilbert
    592. J Parfitt
    593. H Sekhon
    594. G Thomas
    595. DM Rassl
    596. RC Rintoul
    597. C Bifulco
    598. R Tamakawa
    599. W Urba
    600. N Hayward
    601. H Timmers
    602. A Antenucci
    603. F Facciolo
    604. G Grazi
    605. M Marino
    606. R Merola
    607. R de Krijger
    608. A-P Gimenez-Roqueplo
    609. A Piché
    610. S Chevalier
    611. G McKercher
    612. K Birsoy
    613. G Barnett
    614. C Brewer
    615. C Farver
    616. T Naska
    617. NA Pennell
    618. D Raymond
    619. C Schilero
    620. K Smolenski
    621. F Williams
    622. C Morrison
    623. JA Borgia
    624. MJ Liptay
    625. M Pool
    626. CW Seder
    627. K Junker
    628. L Omberg
    629. M Dinkin
    630. G Manikhas
    631. D Alvaro
    632. MC Bragazzi
    633. V Cardinale
    634. G Carpino
    635. E Gaudio
    636. D Chesla
    637. S Cottingham
    638. M Dubina
    639. F Moiseenko
    640. R Dhanasekaran
    641. K-F Becker
    642. K-P Janssen
    643. J Slotta-Huspenina
    644. MH Abdel-Rahman
    645. D Aziz
    646. S Bell
    647. CM Cebulla
    648. A Davis
    649. R Duell
    650. JB Elder
    651. J Hilty
    652. B Kumar
    653. J Lang
    654. NL Lehman
    655. R Mandt
    656. P Nguyen
    657. R Pilarski
    658. K Rai
    659. L Schoenfield
    660. K Senecal
    661. P Wakely
    662. P Hansen
    663. R Lechan
    664. J Powers
    665. A Tischler
    666. WE Grizzle
    667. KC Sexton
    668. A Kastl
    669. J Henderson
    670. S Porten
    671. J Waldmann
    672. M Fassnacht
    673. SL Asa
    674. D Schadendorf
    675. M Couce
    676. M Graefen
    677. H Huland
    678. G Sauter
    679. T Schlomm
    680. R Simon
    681. P Tennstedt
    682. O Olabode
    683. M Nelson
    684. O Bathe
    685. PR Carroll
    686. JM Chan
    687. P Disaia
    688. P Glenn
    689. RK Kelley
    690. CN Landen
    691. J Phillips
    692. M Prados
    693. J Simko
    694. K Smith-McCune
    695. S VandenBerg
    696. K Roggin
    697. A Fehrenbach
    698. A Kendler
    699. S Sifri
    700. R Steele
    701. A Jimeno
    702. F Carey
    703. I Forgie
    704. M Mannelli
    705. M Carney
    706. B Hernandez
    707. B Campos
    708. C Herold-Mende
    709. C Jungk
    710. A Unterberg
    711. A von Deimling
    712. A Bossler
    713. J Galbraith
    714. L Jacobus
    715. M Knudson
    716. T Knutson
    717. D Ma
    718. M Milhem
    719. R Sigmund
    720. AK Godwin
    721. R Madan
    722. HG Rosenthal
    723. C Adebamowo
    724. SN Adebamowo
    725. A Boussioutas
    726. D Beer
    727. T Giordano
    728. A-M Mes-Masson
    729. F Saad
    730. T Bocklage
    731. L Landrum
    732. R Mannel
    733. K Moore
    734. K Moxley
    735. R Postier
    736. J Walker
    737. R Zuna
    738. M Feldman
    739. F Valdivieso
    740. R Dhir
    741. J Luketich
    742. EMM Pinero
    743. M Quintero-Aguilo
    744. CG Carlotti
    745. JS Dos Santos
    746. R Kemp
    747. A Sankarankuty
    748. D Tirapelli
    749. J Catto
    750. K Agnew
    751. E Swisher
    752. J Creaney
    753. B Robinson
    754. CS Shelley
    755. EM Godwin
    756. S Kendall
    757. C Shipman
    758. C Bradford
    759. T Carey
    760. A Haddad
    761. J Moyer
    762. L Peterson
    763. M Prince
    764. L Rozek
    765. G Wolf
    766. R Bowman
    767. KM Fong
    768. I Yang
    769. R Korst
    770. WK Rathmell
    771. JL Fantacone-Campbell
    772. JA Hooke
    773. AJ Kovatich
    774. CD Shriver
    775. J DiPersio
    776. B Drake
    777. R Govindan
    778. S Heath
    779. T Ley
    780. B Van Tine
    781. P Westervelt
    782. MA Rubin
    783. JI Lee
    784. ND Aredes
    785. A Mariamidze
    (2018)
    Cell Reports 23:194–212.
    https://doi.org/10.1016/j.celrep.2018.03.063
  13. 13
  14. 14
  15. 15
  16. 16
  17. 17
  18. 18
  19. 19
  20. 20
  21. 21
  22. 22
  23. 23
  24. 24
  25. 25
  26. 26
  27. 27
  28. 28
  29. 29
  30. 30
  31. 31
  32. 32
  33. 33
  34. 34
  35. 35
  36. 36
  37. 37
  38. 38
  39. 39
  40. 40
  41. 41
  42. 42
  43. 43
  44. 44
  45. 45
  46. 46
  47. 47
  48. 48
  49. 49
  50. 50
  51. 51
  52. 52
  53. 53
  54. 54
  55. 55
  56. 56
  57. 57
  58. 58
  59. 59
  60. 60
  61. 61
  62. 62
  63. 63
  64. 64
  65. 65
  66. 66
  67. 67
  68. 68
  69. 69
  70. 70
  71. 71
    Tumorigenic keratinocyte lines requiring anchorage and fibroblast support cultured from human squamous cell carcinomas
    1. JG Rheinwald
    2. MA Beckett
    (1981)
    Cancer Research 41:1657–1663.
  72. 72
  73. 73
  74. 74
  75. 75
  76. 76
  77. 77
  78. 78
  79. 79
  80. 80
  81. 81
  82. 82
    Establishment and characterization of a cell line (SAS) from poorly differentiated human squamous cell carcinoma of the tongue
    1. K Takahashi
    2. H Kanazawa
    3. Y Akiyama
    4. S Tazaki
    5. M Takahara
    6. T Muto
    7. H Tanzawa
    8. K Sato
    (1989)
    Journal of the Japanese Stomatological Society 38:20–28.
  83. 83
  84. 84
  85. 85
  86. 86
  87. 87
  88. 88
  89. 89
    Comprehensive molecular characterization of the hippo signaling pathway in Cancer
    1. Y Wang
    2. X Xu
    3. D Maglic
    4. MT Dill
    5. K Mojumdar
    6. PK-S Ng
    7. KJ Jeong
    8. YH Tsang
    9. D Moreno
    10. VH Bhavana
    11. X Peng
    12. Z Ge
    13. H Chen
    14. J Li
    15. Z Chen
    16. H Zhang
    17. L Han
    18. D Du
    19. CJ Creighton
    20. GB Mills
    21. F Camargo
    22. H Liang
    23. SJ Caesar-Johnson
    24. JA Demchok
    25. I Felau
    26. M Kasapi
    27. ML Ferguson
    28. CM Hutter
    29. HJ Sofia
    30. R Tarnuzzer
    31. Z Wang
    32. L Yang
    33. JC Zenklusen
    34. JJ Zhang
    35. S Chudamani
    36. J Liu
    37. L Lolla
    38. R Naresh
    39. T Pihl
    40. Q Sun
    41. Y Wan
    42. Y Wu
    43. J Cho
    44. T DeFreitas
    45. S Frazer
    46. N Gehlenborg
    47. G Getz
    48. DI Heiman
    49. J Kim
    50. MS Lawrence
    51. P Lin
    52. S Meier
    53. MS Noble
    54. G Saksena
    55. D Voet
    56. H Zhang
    57. B Bernard
    58. N Chambwe
    59. V Dhankani
    60. T Knijnenburg
    61. R Kramer
    62. K Leinonen
    63. Y Liu
    64. M Miller
    65. S Reynolds
    66. I Shmulevich
    67. V Thorsson
    68. W Zhang
    69. R Akbani
    70. BM Broom
    71. AM Hegde
    72. Z Ju
    73. RS Kanchi
    74. A Korkut
    75. J Li
    76. H Liang
    77. S Ling
    78. W Liu
    79. Y Lu
    80. GB Mills
    81. K-S Ng
    82. A Rao
    83. M Ryan
    84. J Wang
    85. JN Weinstein
    86. J Zhang
    87. A Abeshouse
    88. J Armenia
    89. D Chakravarty
    90. WK Chatila
    91. I de Bruijn
    92. J Gao
    93. BE Gross
    94. ZJ Heins
    95. R Kundra
    96. K La
    97. M Ladanyi
    98. A Luna
    99. MG Nissan
    100. A Ochoa
    101. SM Phillips
    102. E Reznik
    103. F Sanchez-Vega
    104. C Sander
    105. N Schultz
    106. R Sheridan
    107. SO Sumer
    108. Y Sun
    109. BS Taylor
    110. J Wang
    111. H Zhang
    112. P Anur
    113. M Peto
    114. P Spellman
    115. C Benz
    116. JM Stuart
    117. CK Wong
    118. C Yau
    119. DN Hayes
    120. JS Parker
    121. MD Wilkerson
    122. A Ally
    123. M Balasundaram
    124. R Bowlby
    125. D Brooks
    126. R Carlsen
    127. E Chuah
    128. N Dhalla
    129. R Holt
    130. SJM Jones
    131. K Kasaian
    132. D Lee
    133. Y Ma
    134. MA Marra
    135. M Mayo
    136. RA Moore
    137. AJ Mungall
    138. K Mungall
    139. AG Robertson
    140. S Sadeghi
    141. JE Schein
    142. P Sipahimalani
    143. A Tam
    144. N Thiessen
    145. K Tse
    146. T Wong
    147. AC Berger
    148. R Beroukhim
    149. AD Cherniack
    150. C Cibulskis
    151. SB Gabriel
    152. GF Gao
    153. G Ha
    154. M Meyerson
    155. SE Schumacher
    156. J Shih
    157. MH Kucherlapati
    158. RS Kucherlapati
    159. S Baylin
    160. L Cope
    161. L Danilova
    162. MS Bootwalla
    163. PH Lai
    164. DT Maglinte
    165. DJ Van Den Berg
    166. DJ Weisenberger
    167. JT Auman
    168. S Balu
    169. T Bodenheimer
    170. C Fan
    171. KA Hoadley
    172. AP Hoyle
    173. SR Jefferys
    174. CD Jones
    175. S Meng
    176. PA Mieczkowski
    177. LE Mose
    178. AH Perou
    179. CM Perou
    180. J Roach
    181. Y Shi
    182. JV Simons
    183. T Skelly
    184. MG Soloway
    185. D Tan
    186. U Veluvolu
    187. H Fan
    188. T Hinoue
    189. PW Laird
    190. H Shen
    191. W Zhou
    192. M Bellair
    193. K Chang
    194. K Covington
    195. CJ Creighton
    196. H Dinh
    197. H Doddapaneni
    198. LA Donehower
    199. J Drummond
    200. RA Gibbs
    201. R Glenn
    202. W Hale
    203. Y Han
    204. J Hu
    205. V Korchina
    206. S Lee
    207. L Lewis
    208. W Li
    209. X Liu
    210. M Morgan
    211. D Morton
    212. D Muzny
    213. J Santibanez
    214. M Sheth
    215. E Shinbrot
    216. L Wang
    217. M Wang
    218. DA Wheeler
    219. L Xi
    220. F Zhao
    221. J Hess
    222. EL Appelbaum
    223. M Bailey
    224. MG Cordes
    225. L Ding
    226. CC Fronick
    227. LA Fulton
    228. RS Fulton
    229. C Kandoth
    230. ER Mardis
    231. MD McLellan
    232. CA Miller
    233. HK Schmidt
    234. RK Wilson
    235. D Crain
    236. E Curley
    237. J Gardner
    238. K Lau
    239. D Mallery
    240. S Morris
    241. J Paulauskis
    242. R Penny
    243. C Shelton
    244. T Shelton
    245. M Sherman
    246. E Thompson
    247. P Yena
    248. J Bowen
    249. JM Gastier-Foster
    250. M Gerken
    251. KM Leraas
    252. TM Lichtenberg
    253. NC Ramirez
    254. L Wise
    255. E Zmuda
    256. N Corcoran
    257. T Costello
    258. C Hovens
    259. AL Carvalho
    260. AC de Carvalho
    261. JH Fregnani
    262. A Longatto-Filho
    263. RM Reis
    264. C Scapulatempo-Neto
    265. HCS Silveira
    266. DO Vidal
    267. A Burnette
    268. J Eschbacher
    269. B Hermes
    270. A Noss
    271. R Singh
    272. ML Anderson
    273. PD Castro
    274. M Ittmann
    275. D Huntsman
    276. B Kohl
    277. X Le
    278. R Thorp
    279. C Andry
    280. ER Duffy
    281. V Lyadov
    282. O Paklina
    283. G Setdikova
    284. A Shabunin
    285. M Tavobilov
    286. C McPherson
    287. R Warnick
    288. R Berkowitz
    289. D Cramer
    290. C Feltmate
    291. N Horowitz
    292. A Kibel
    293. M Muto
    294. CP Raut
    295. A Malykh
    296. JS Barnholtz-Sloan
    297. W Barrett
    298. K Devine
    299. J Fulop
    300. QT Ostrom
    301. K Shimmel
    302. Y Wolinsky
    303. AE Sloan
    304. A De Rose
    305. F Giuliante
    306. M Goodman
    307. BY Karlan
    308. CH Hagedorn
    309. J Eckman
    310. J Harr
    311. J Myers
    312. K Tucker
    313. LA Zach
    314. B Deyarmin
    315. H Hu
    316. L Kvecher
    317. C Larson
    318. RJ Mural
    319. S Somiari
    320. A Vicha
    321. T Zelinka
    322. J Bennett
    323. M Iacocca
    324. B Rabeno
    325. P Swanson
    326. M Latour
    327. L Lacombe
    328. B Têtu
    329. A Bergeron
    330. M McGraw
    331. SM Staugaitis
    332. J Chabot
    333. H Hibshoosh
    334. A Sepulveda
    335. T Su
    336. T Wang
    337. O Potapova
    338. O Voronina
    339. L Desjardins
    340. O Mariani
    341. S Roman-Roman
    342. X Sastre
    343. M-H Stern
    344. F Cheng
    345. S Signoretti
    346. A Berchuck
    347. D Bigner
    348. E Lipp
    349. J Marks
    350. S McCall
    351. R McLendon
    352. A Secord
    353. A Sharp
    354. M Behera
    355. DJ Brat
    356. A Chen
    357. K Delman
    358. S Force
    359. F Khuri
    360. K Magliocca
    361. S Maithel
    362. JJ Olson
    363. T Owonikoko
    364. A Pickens
    365. S Ramalingam
    366. DM Shin
    367. G Sica
    368. EG Van Meir
    369. H Zhang
    370. W Eijckenboom
    371. A Gillis
    372. E Korpershoek
    373. L Looijenga
    374. W Oosterhuis
    375. H Stoop
    376. KE van Kessel
    377. EC Zwarthoff
    378. C Calatozzolo
    379. L Cuppini
    380. S Cuzzubbo
    381. F DiMeco
    382. G Finocchiaro
    383. L Mattei
    384. A Perin
    385. B Pollo
    386. C Chen
    387. J Houck
    388. P Lohavanichbutr
    389. A Hartmann
    390. C Stoehr
    391. R Stoehr
    392. H Taubert
    393. S Wach
    394. B Wullich
    395. W Kycler
    396. D Murawa
    397. M Wiznerowicz
    398. K Chung
    399. WJ Edenfield
    400. J Martin
    401. E Baudin
    402. G Bubley
    403. R Bueno
    404. A De Rienzo
    405. WG Richards
    406. S Kalkanis
    407. T Mikkelsen
    408. H Noushmehr
    409. L Scarpace
    410. N Girard
    411. M Aymerich
    412. E Campo
    413. E Giné
    414. AL Guillermo
    415. N Van Bang
    416. PT Hanh
    417. BD Phu
    418. Y Tang
    419. H Colman
    420. K Evason
    421. PR Dottino
    422. JA Martignetti
    423. H Gabra
    424. H Juhl
    425. T Akeredolu
    426. S Stepa
    427. D Hoon
    428. K Ahn
    429. KJ Kang
    430. F Beuschlein
    431. A Breggia
    432. M Birrer
    433. D Bell
    434. M Borad
    435. AH Bryce
    436. E Castle
    437. V Chandan
    438. J Cheville
    439. JA Copland
    440. M Farnell
    441. T Flotte
    442. N Giama
    443. T Ho
    444. M Kendrick
    445. J-P Kocher
    446. K Kopp
    447. C Moser
    448. D Nagorney
    449. D O’Brien
    450. BP O’Neill
    451. T Patel
    452. G Petersen
    453. F Que
    454. M Rivera
    455. L Roberts
    456. R Smallridge
    457. T Smyrk
    458. M Stanton
    459. RH Thompson
    460. M Torbenson
    461. JD Yang
    462. L Zhang
    463. F Brimo
    464. JA Ajani
    465. AMA Gonzalez
    466. C Behrens
    467. J Bondaruk
    468. R Broaddus
    469. B Czerniak
    470. B Esmaeli
    471. J Fujimoto
    472. J Gershenwald
    473. C Guo
    474. AJ Lazar
    475. C Logothetis
    476. F Meric-Bernstam
    477. C Moran
    478. L Ramondetta
    479. D Rice
    480. A Sood
    481. P Tamboli
    482. T Thompson
    483. P Troncoso
    484. A Tsao
    485. I Wistuba
    486. C Carter
    487. L Haydu
    488. P Hersey
    489. V Jakrot
    490. H Kakavand
    491. R Kefford
    492. K Lee
    493. G Long
    494. G Mann
    495. M Quinn
    496. R Saw
    497. R Scolyer
    498. K Shannon
    499. A Spillane
    500. J Stretch
    501. M Synott
    502. J Thompson
    503. J Wilmott
    504. H Al-Ahmadie
    505. TA Chan
    506. R Ghossein
    507. A Gopalan
    508. DA Levine
    509. V Reuter
    510. S Singer
    511. B Singh
    512. NV Tien
    513. T Broudy
    514. C Mirsaidi
    515. P Nair
    516. P Drwiega
    517. J Miller
    518. J Smith
    519. H Zaren
    520. J-W Park
    521. NP Hung
    522. E Kebebew
    523. WM Linehan
    524. AR Metwalli
    525. K Pacak
    526. PA Pinto
    527. M Schiffman
    528. LS Schmidt
    529. CD Vocke
    530. N Wentzensen
    531. R Worrell
    532. H Yang
    533. M Moncrieff
    534. C Goparaju
    535. J Melamed
    536. H Pass
    537. N Botnariuc
    538. I Caraman
    539. M Cernat
    540. I Chemencedji
    541. A Clipca
    542. S Doruc
    543. G Gorincioi
    544. S Mura
    545. M Pirtac
    546. I Stancul
    547. D Tcaciuc
    548. M Albert
    549. I Alexopoulou
    550. A Arnaout
    551. J Bartlett
    552. J Engel
    553. S Gilbert
    554. J Parfitt
    555. H Sekhon
    556. G Thomas
    557. DM Rassl
    558. RC Rintoul
    559. C Bifulco
    560. R Tamakawa
    561. W Urba
    562. N Hayward
    563. H Timmers
    564. A Antenucci
    565. F Facciolo
    566. G Grazi
    567. M Marino
    568. R Merola
    569. R de Krijger
    570. A-P Gimenez-Roqueplo
    571. A Piché
    572. S Chevalier
    573. G McKercher
    574. K Birsoy
    575. G Barnett
    576. C Brewer
    577. C Farver
    578. T Naska
    579. NA Pennell
    580. D Raymond
    581. C Schilero
    582. K Smolenski
    583. F Williams
    584. C Morrison
    585. JA Borgia
    586. MJ Liptay
    587. M Pool
    588. CW Seder
    589. K Junker
    590. L Omberg
    591. M Dinkin
    592. G Manikhas
    593. D Alvaro
    594. MC Bragazzi
    595. V Cardinale
    596. G Carpino
    597. E Gaudio
    598. D Chesla
    599. S Cottingham
    600. M Dubina
    601. F Moiseenko
    602. R Dhanasekaran
    603. K-F Becker
    604. K-P Janssen
    605. J Slotta-Huspenina
    606. MH Abdel-Rahman
    607. D Aziz
    608. S Bell
    609. CM Cebulla
    610. A Davis
    611. R Duell
    612. JB Elder
    613. J Hilty
    614. B Kumar
    615. J Lang
    616. NL Lehman
    617. R Mandt
    618. P Nguyen
    619. R Pilarski
    620. K Rai
    621. L Schoenfield
    622. K Senecal
    623. P Wakely
    624. P Hansen
    625. R Lechan
    626. J Powers
    627. A Tischler
    628. WE Grizzle
    629. KC Sexton
    630. A Kastl
    631. J Henderson
    632. S Porten
    633. J Waldmann
    634. M Fassnacht
    635. SL Asa
    636. D Schadendorf
    637. M Couce
    638. M Graefen
    639. H Huland
    640. G Sauter
    641. T Schlomm
    642. R Simon
    643. P Tennstedt
    644. O Olabode
    645. M Nelson
    646. O Bathe
    647. PR Carroll
    648. JM Chan
    649. P Disaia
    650. P Glenn
    651. RK Kelley
    652. CN Landen
    653. J Phillips
    654. M Prados
    655. J Simko
    656. K Smith-McCune
    657. S VandenBerg
    658. K Roggin
    659. A Fehrenbach
    660. A Kendler
    661. S Sifri
    662. R Steele
    663. A Jimeno
    664. F Carey
    665. I Forgie
    666. M Mannelli
    667. M Carney
    668. B Hernandez
    669. B Campos
    670. C Herold-Mende
    671. C Jungk
    672. A Unterberg
    673. A von Deimling
    674. A Bossler
    675. J Galbraith
    676. L Jacobus
    677. M Knudson
    678. T Knutson
    679. D Ma
    680. M Milhem
    681. R Sigmund
    682. AK Godwin
    683. R Madan
    684. HG Rosenthal
    685. C Adebamowo
    686. SN Adebamowo
    687. A Boussioutas
    688. D Beer
    689. T Giordano
    690. A-M Mes-Masson
    691. F Saad
    692. T Bocklage
    693. L Landrum
    694. R Mannel
    695. K Moore
    696. K Moxley
    697. R Postier
    698. J Walker
    699. R Zuna
    700. M Feldman
    701. F Valdivieso
    702. R Dhir
    703. J Luketich
    704. EMM Pinero
    705. M Quintero-Aguilo
    706. CG Carlotti
    707. JS Dos Santos
    708. R Kemp
    709. A Sankarankuty
    710. D Tirapelli
    711. J Catto
    712. K Agnew
    713. E Swisher
    714. J Creaney
    715. B Robinson
    716. CS Shelley
    717. EM Godwin
    718. S Kendall
    719. C Shipman
    720. C Bradford
    721. T Carey
    722. A Haddad
    723. J Moyer
    724. L Peterson
    725. M Prince
    726. L Rozek
    727. G Wolf
    728. R Bowman
    729. KM Fong
    730. I Yang
    731. R Korst
    732. WK Rathmell
    733. JL Fantacone-Campbell
    734. JA Hooke
    735. AJ Kovatich
    736. CD Shriver
    737. J DiPersio
    738. B Drake
    739. R Govindan
    740. S Heath
    741. T Ley
    742. B Van Tine
    743. P Westervelt
    744. MA Rubin
    745. JI Lee
    746. ND Aredes
    747. A Mariamidze
    (2018)
    Cell Reports 25:1304–1317.
    https://doi.org/10.1016/j.celrep.2018.10.001
  90. 90
  91. 91
  92. 92
  93. 93
  94. 94
  95. 95
  96. 96

Decision letter

  1. Erica A Golemis
    Reviewing Editor; Fox Chase Cancer Center, United States
  2. Päivi M Ojala
    Senior Editor; University of Helsinki, Finland
  3. Marius Sudol
    Reviewer

In the interests of transparency, eLife publishes the most substantive revision requests and the accompanying author responses.

Acceptance summary:

This work defines genes required for survival and robust growth by oral squamous cell carcinoma (OSCC) cells in Western and Asian populations. Following a genome-wide CRISPR screen, the authors identified components of the Hippo/Yap pathway as important for growth of multiple OSCC cell lines. In dissecting the pathway, the authors determined that some OSCC cell lines depend more on YAP1, some more on WWTR1/TAZ, while some depend on both, suggesting more heterogeneity in response than was previously appreciated. Importantly, the work also suggests some specific features of transcriptional pathways, including activation of the NF-κB pathway, in OSCC derived from Asians with a habit of betel quid chewing. These different dependencies may have clinical consequences, as the data suggests that OSCC cells specifically dependent on WWTR1 are associated with a distinct transcriptome and immune signature associated with response to immunotherapy.

Decision letter after peer review:

Thank you for submitting your article "Genome-wide CRISPR screens of oral squamous cell carcinoma reveal fitness genes in the Hippo pathway" for consideration by eLife. Your article has been reviewed by two peer reviewers, and the evaluation has been overseen by a Reviewing Editor and Päivi Ojala as the Senior Editor. The following individual involved in review of your submission has agreed to reveal their identity: Marius Sudol (Reviewer #2).

The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.

As the editors have judged that your manuscript is of interest, but as described below that additional experiments are required before it is published, we would like to draw your attention to changes in our revision policy that we have made in response to COVID-19 (https://elifesciences.org/articles/57162). First, because many researchers have temporarily lost access to the labs, we will give authors as much time as they need to submit revised manuscripts. We are also offering, if you choose, to post the manuscript to bioRxiv (if it is not already there) along with this decision letter and a formal designation that the manuscript is "in revision at eLife". Please let us know if you would like to pursue this option. (If your work is more suitable for medRxiv, you will need to post the preprint yourself, as the mechanisms for us to do so are still in development.)

Summary:

This manuscript by Annie Chai and colleagues from the laboratory of Sok Ching Cheong reports in the CRISPR/Cas9-based screen of cells derived from oral squamous cell carcinoma (OSCC), which were predominantly of Asian, Malaysian origin. The screen was aimed at the identification of "fitness genes". Several of these genes, some known and some new, were given non-essential status and a new set of essential genes was identified as worthwhile targets of a potential therapy. The authors identified components of the Hippo/Yap pathway as important for growth of multiple OSCC cell lines. For more than a decade, the two main pro-oncogene effectors of the Hippo pathway, namely YAP and TAZ, have been considered acting in unison in cancers. The current report indicates that YAP and TAZ may play specific roles in cancer. In dissecting the pathway, the authors determined that some OSCC cell lines depend more on YAP1, some more on WWTR1/TAZ, and in some either protein can complement the function of the other, suggesting more heterogeneity in response than was previously appreciated. The copy number amplification of WWTR1/TAZ was shown as likely to constitute a functional oncogenic role of this gene and its protein in OSCC. However, non-genomic mechanisms (with some exceptions) could likely activate YAP. These different dependencies may have clinical consequences, as the data suggests that OSCC cells specifically dependent on WWTR1 appear to be associated with a distinct transcriptome and immune signature associated with response to immunotherapy. Given the considerable interest in therapeutically targeting immune pathways, these findings could provide a valuable response biomarker, as indicating that two Hippo pathway endpoints that are often grouped together have quite distinct roles. The results also could guide the use of therapies under development for targeting YAP1 or WWTR1/TAZ. For example, another of the important conclusions of this study is that TAZ-dependent cell lines may represent fiduciaries of their original tumors, which could be primary candidates for checkpoint inhibitors treatment via modern immunotherapy.

There are some points that should be addressed by the authors to solidify the conclusions and improve the study.

Essential revisions:

(1) The authors emphasize that a rationale for their study is the use of cell materials developed from Asian populations with different risk factors, such as use of betel. They hypothesize that these features of the OSCC cell set may result in different dependency profiles than those already published for OSCC from the DepMap project (Behan et al. 2019, Meyers et al., 2017). However, they do not explicitly compare results with those from the DepMap project. This analysis should be added, to determine if the hypothesized difference exists.

(2) A lot of the analysis pathway appears to be derivative of approaches from Behan et al., 2019. This includes data exclusion for core fitness genes, assignment of tractability groups, and other points. The authors need to make clear the novelty of the current work.

(3) From data shown in Figure 1C, most of the cell lines have a limited number of non-core fitness genes (40 or less), whereas 4 lines have a very large number of dependencies. What genomic features characterize those with large numbers of dependencies? If these are excluded, how does this affect the statistical analysis present in Figure 2? Similarly, 10 of the 21 cell lines appeared to have <10 core dependencies. Do these models specifically have unique genomic features, and YAP1 or WWTR1 dependency?

(4) In Figure 3—figure supplement 3, the authors note correlation between WWTR1 gene essentiality and expression across 273 cell models. What are the cancer types that are most dependent on WWTR1 (the point being, how specific is the observation of dependency to OSCC, versus being a general feature of squamous cancers)?

(5) In Figure 6, the authors analyze the YAP1 and WWTR1 gene signatures across a large group of OSCCs in the TCGA. Earlier in the study, the authors note that WWTR1 is often amplified as part of an amplicon including PIK3CA, TP63, and SOX2. To what extent is the transcriptional signature associated with amplification of these other genes, and dependent on their transcriptional activity, rather than that of WWTR1? Using their cell line models, they should complement TCGA analysis by using shRNA or drug inhibition to deplete WWTR1 versus other key genes in the amplicon, to determine which causes a specific loss of the immune-associated transcriptional signature. This is a really important data point for interpretation of the results.

(6) Besides papers cited by the authors, work relevant to the topic of this study includes Omori et al., 2020, for YAP1 as a driver, Saloura et al., 2019, showing YAP1 mutation is associated with a low CD8+ T cell inflamed phenotype, Martin et al., 2018, dissecting the Hippo pathway and demonstrating the targetability of YAP1 in head and neck cancer in a detailed Nat Comm paper, and an extensive review of the pathway at the end of 2019 by Santos-de-Frutos and colleagues. This work should be cited and discussed by the authors. It is important that the Discussion addresses more about how the generated data relate to previously published reports. For example, the fact that YAP, TAZ, PI3K, TP63, and SOX2 are amplified in OSCC is curious as all these genes are part of the extended signaling network of the Hippo pathway. Please see the examples of relevant references and consider a short paragraph discussing this point. [Huang et al., 2017; Fan , Kim and Gumbiner, 2013. And Li et al., 2019] Also when the role of TA in eliciting the immune response is mentioned, it would be important to reference one of the first publications (as far as the reviewer is aware of) that indicated the role of TAZ (and YAP, perhaps) via LATS1 KO in mice. The Hippo Pathway Kinases LATS1/2 Suppress Cancer Immunity. Moroishi T, (many authors) Guan KL. Cell. 2016 Dec 1;167(6):1525-1539.

(7) The authors should specifically state how many of their cell lines were highly dependent on WWTR1, how many on YAP1, and how many on both, out of the group of 21.

(8) The cell lines used in the screen are critical for the study. Therefore, it is suggested to provide in the Materials and methods and/or in the Results more information about how these lines were derived. It would help the readers a lot without referring to a previous publication (Fadlullah et al., 2016) and various commercial data for cell repositories. Please address if these cell lines were derived spontaneously (ala HeLa cells) or were engineered by transfections of immortalizing genes (ala HEK293 cell line, for example). If the latter was the case, please comment if the process of immortalization did not affect the actual transcriptional profile of the lines.

(9) As stated above, one of the critical conclusions this study is that TAZ-dependent OSCC lines may represent fiduciary (direct references) of their original tumors, and therefore patients with tumors driven by TAZ could be primary candidates for checkpoint inhibitors treatment via immunotherapy. Please address directly in the Discussion the fact that most of the tumors are quite heterogeneous in terms of oncogenic drivers and either the tumor biopsy or the cell lines derived from the tumor might not represent the "main" oncogenic drivers of the entire tumor.

https://doi.org/10.7554/eLife.57761.sa1

Author response

Essential revisions:

(1) The authors emphasize that a rationale for their study is the use of cell materials developed from Asian populations with different risk factors, such as use of betel. They hypothesize that these features of the OSCC cell set may result in different dependency profiles than those already published for OSCC from the DepMap project (Behan et al., 2019, Meyers et al., 2017). However, they do not explicitly compare results with those from the DepMap project. This analysis should be added, to determine if the hypothesized difference exists.

We thank the reviewers for the suggestion. We have added the analysis of enriched dependencies seen among betel quid-associated OSCC in the revised manuscript and discussed the results accordingly. In particular, by comparing the gene dependencies identified from betel quid-associated OSCC lines with those non-betel quid associated OSCCs, we found that the NF-κB pathway is significantly enriched in the betel-quid associated lines. When cross checking the fitness genes in this pathway with those from the DepMap project, where models not known to be associated with betel quid were used, all except one were not identified as fitness genes (CSNK2A1 being marginally depleted in one of the cell lines included in DepMap; SCC-4). Notably, the literature that reported on betel quid as a risk factor in OSCC have demonstrated that aberrant NF-κB pathway activation is common in these cancers. Furthermore, extracts from betel quid have been shown to directly activate NF-κB signaling in OSCC as reported in previous studies. By contrast, this is rarely reported among the OSCC studies from the Western countries, where betel quid is uncommon.

We have added the results of this analysis to Figure 2 (Figure 2D), and explained this in the Results section (subsection “Identification of unique dependencies in betel-quid associated OSCC”). We also discussed our findings in the Discussion section (third paragraph). The description of the analysis has been added to the Materials and methods section of the revised manuscript (subsection “Differential fitness genes analysis for betel quid associated OSCCs”). Figure legends and figure source data have also been included accordingly.

2) A lot of the analysis pathway appears to be derivative of approaches from Behan et al., 2019. This includes data exclusion for core fitness genes, assignment of tractability groups, and other points. The authors need to make clear the novelty of the current work.

The publication from Behan et al. is one of the most comprehensive essential screens that was conducted across many types of cancers (pan-cancer) and in their project, they have developed comprehensive and robust methods in filtering and analyzing data from such screens particularly in identifying tractable genes for further therapeutic development. As one of our aims in conducting the essential screens for OSCC was to identify gene targets that could be translated to clinical benefit through the development or identification of new therapies, we applied similar filtering criteria to our data including removing the core-fitness genes (as described by Hart et al. Cell. 2015 Dec 3;163(6):1515-26. doi: 10.1016/j.cell.2015.11.015), and assigning tractability groups as described by Behan et al. Whilst we may have adopted a similar approach, our work is novel in that (i) our study focuses on OSCC and it would be the first to describe non-core fitness genes for OSCC and it is also the largest dataset of dependencies in OSCC models; (ii) we included models derived from Asians (particularly those with betel-quid chewing habit) where the disease is most prevalent and describe possible distinctiveness in the non-core fitness genes in models that are representative of OSCC in Asians; (iii) we looked specifically at known oncogenes and recurrently amplified regions to interrogate the function of OSCC cancer genes, and (iv) we report distinct dependencies on YAP1 and WWTR1 among the different subsets of OSCC where these dependencies could have implications in cancer therapies including immunotherapy.

3) From data shown in Figure 1C, most of the cell lines have a limited number of non-core fitness genes (40 or less), whereas 4 lines have a very large number of dependencies. What genomic features characterize those with large numbers of dependencies? If these are excluded, how does this affect the statistical analysis present in Figure 2? Similarly, 10 of the 21 cell lines appeared to have <10 core dependencies. Do these models specifically have unique genomic features, and YAP1 or WWTR1 dependency?

We would like to clarify that the y-axis on Figure 1C refers to the number of dependent cell lines; while x-axis refers to the number of non-core fitness genes. For example, there are 366 genes that are essential in one out of 21 cell line (bottom most bar) – indicating fitness genes unique to a single cell line. In other words, we found 366 unique genes that is essential in only one cell line. On the other hand, only one gene (NELFCD) is a fitness gene for all the 21 cell lines (top most bar). This demonstrates the heterogeneity of the essentiality across the cell lines. To make this clearer, we have changed the y-axis label to “Number of dependent cell lines” and added more description in the figure legend.

4) In Figure 3—figure supplement 3, the authors note correlation between WWTR1 gene essentiality and expression across 273 cell models. What are the cancer types that are most dependent on WWTR1 (the point being, how specific is the observation of dependency to OSCC, versus being a general feature of squamous cancers)?

We observed the correlation of WWTR1 gene dependency with its gene expression in several cancer types including squamous cell lung carcinoma, breast carcinoma, glioblastoma, ovarian carcinoma and low-grade glioma in addition to OSCC. Therefore, this observation is not entirely unique to OSCC or squamous cell carcinoma. We have added this information on other cancers in Figure 3—figure supplement 3D and included the results in the fourth paragraph of the subsection “Differential dependency pattern on YAP1 and WWTR1”.

5) In Figure 6, the authors analyze the YAP1 and WWTR1 gene signatures across a large group of OSCCs in the TCGA. Earlier in the study, the authors note that WWTR1 is often amplified as part of an amplicon including PIK3CA, TP63, and SOX2. To what extent is the transcriptional signature associated with amplification of these other genes, and dependent on their transcriptional activity, rather than that of WWTR1? Using their cell line models, they should complement TCGA analysis by using shRNA or drug inhibition to deplete WWTR1 versus other key genes in the amplicon, to determine which causes a specific loss of the immune-associated transcriptional signature. This is a really important data point for interpretation of the results.

We thank the reviewer for the suggestion. We agree that it would be important to show that the correlation between WWTR1-dependency gene signatures with the immune signatures are indeed specific to WWTR1, and not due to other co-amplified genes such as PIK3CA, TP63 and SOX2. We did not previously include this analysis as using cell line model, most of the immune-associated transcriptional signature might not be directly measurable. Nevertheless, we have now taken two approaches to address this question as best as we could. First, we used a WWTR1-dependent model, ORL-214 to knockout WWTR1, PIK3CA, TP63 and SOX2 using two individual sgRNAs for each gene, and examined the down-regulation effect on PD-L1 gene expression. As we anticipated, only WWTR1-knock-out (KO) cells showed significant down-regulation of PD-L1 expression, but not in PIK3CA-, TP63-, nor SOX2-KO cells. These results are now reported in the subsection “ OSCC with WWTR1 dependency signature and immune biomarkers” and Figure 6—figure supplement 3A.

In addition, we also examined the correlation between the expression of PIK3CA, TP63 and SOX2 with PD-L1 expression, using an OSCC microarray dataset from Hiemer et al., 2015. Consistently, only WWTR1 gene expression showed significant correlation with PD-L1 gene expression. This correlation was not observed between PIK3CA, TP63, or SOX2 with PD-L1. These results are now reported in the aforementioned subsection and Figure 6—figure supplement 3B.

6) Besides papers cited by the authors, work relevant to the topic of this study includes Omori et al., 2020, for YAP1 as a driver, Saloura et al., 2019, showing YAP1 mutation is associated with a low CD8+ T cell inflamed phenotype, Martin et al., 2018, dissecting the Hippo pathway and demonstrating the targetability of YAP1 in head and neck cancer in a detailed Nat Comm paper, and an extensive review of the pathway at the end of 2019 by Santos-de-Frutos and colleagues. This work should be cited and discussed by the authors. It is important that the Discussion addresses more about how the generated data relate to previously published reports. For example, the fact that YAP, TAZ, PI3K, TP63, and SOX2 are amplified in OSCC is curious as all these genes are part of the extended signaling network of the Hippo pathway. Please see the examples of relevant references and consider a short paragraph discussing this p