FRS2 Is an Oncogene in High Grade Ovarian Cancer

The Harvard community has made this article openly available. Please share how this access benefits you. Your story matters

Citation Luo, Leo Y. 2015. FRS2 Is an Oncogene in High Grade Ovarian Cancer. Doctoral dissertation, Harvard Medical School.

Citable link http://nrs.harvard.edu/urn-3:HUL.InstRepos:15821602

Terms of Use This article was downloaded from Harvard University’s DASH repository, and is made available under the terms and conditions applicable to Other Posted Material, as set forth at http:// nrs.harvard.edu/urn-3:HUL.InstRepos:dash.current.terms-of- use#LAA

Abstract

Ovarian cancer is the most common cause of gynecologic cancer death in the United

States. Despite aggressive surgical cytoreduction and chemotherapy, ovarian cancer remains one of the most lethal cancer types due to advanced stages at diagnosis and lack of effective systemic therapy.

High-grade serous ovarian cancers (HGSOC) are characterized by widespread recurrent regions of copy number gain and loss. Here we interrogated 50 genes that are recurrently amplified in HGSOC and essential for cancer proliferation and survival in ovarian cancer cell lines. FRS2 is one of the 50 genes located on chromosomal region 12q15 that is focally amplified in 12.5% of HGSOC.

We found that FRS2 amplified cancer cell lines are dependent on FRS2 expression and that FRS2 overexpression in immortalized human cell lines conferred the ability to grow in an anchorage independent manner and as tumors in immunodeficient mice. FRS2, an adaptor protein in the FGFR pathway, induces downstream activation of Ras-MAPK pathway. These observations identify FRS2 as an oncogene in a subset of HGSOC that harbor FRS2 amplifications.

This study underlines the power of complementary genomic approaches that use efficient high-throughput methods to assess functional consequences of genomic alterations and develop therapeutic targets toward clinical translation. The discovery of FRS2 as an oncogene also highlights adaptor proteins as a new class of oncogenes that can become the next generation of therapeutic targets.

2

Acknowledgments

This thesis could not have happened without the help of the following people. First of all,

I am deeply grateful to my advisor, Professor William Hahn for his mentorship and unfailing support. He leads by example and sets a high bar for future physician-scientists. He challenges me to approach science with rigor and fearlessness, taught me the importance of precision in communicating scientific results and ideas, and showed me the rapidly changing landscape of cancer research and its transformative power in the clinic. I want to thank my collaborator, friend,

Hiu Wing (Tony) Cheung, for instilling excitement into this project and being a great mentor.

Although we only worked side-by-side for two months, I have learned from Tony through many discussions over the experimental techniques and research directions. This project would not have been completed without my colleague Eejung Kim, who carried the project over its final stretch toward publication. I thank the other co-authors of the manuscript, Barbara Weir, Gavin

Dunn, and Rhine Shen, and fellow members of the Hahn lab, Susan Moody, Diane Shao,

Xiaoxing Wang, for providing valuable guidance on experiments and scientific directions.

I have benefited greatly from discussions on science, medicine, and life with my classmates, faculty members, and staffs of the Health Sciences and Technology program at

Harvard Medical School, especially my advisors David Ting and Rick Mitchell. This project was generously supported by a research fellowship from the Howard Hughes Medical Institute. I thank Melanie Daub and William Galey at HHMI for their roles in making the fellowship year a fantastic experience. Lastly I am forever indebted to my parents, who offer unconditional support and taught me the value of scholarship.

3 Table of Contents

Page No.

Abstract 1

Acknowledgement 3

Table of Contents 4

Glossary of Abbreviations 5

Introduction 6

Materials and Methods 11

Results 17

Discussion 22

Conclusion 29

Summary 30

Figures and Legends 32

References 45

4 Glossary

HEK293 Human embryonic kidney 293 cells

AKT AK-strain thymoma gene, also known as protein kinase B

CCLE Cancer Cell Line Encyclopedia

CRKL Crk-like protein

ERK1/2 Extracellular signal-regulated kinases 1 and 2

FGFR Fibroblast growth factor receptor

FRS2 Fibroblast growth factor receptor Substrate 2

GAB2 GRB2-associated-binding protein 2

GRB2 Growth factor receptor-bound protein 2

HA1E Human embryonic kidney cells immortalized with SV40TAg and

hTERT

HA1E-A Human embryonic kidney cells immortalized with SV40TAg and

hTERT, transformed with myristolated AKT

HA1E-M Human embryonic kidney cells immortalized with SV40TAg and

hTERT, transformed with MEK-DD (constitutively active MEK)

IOSE Immortalized ovarian surface epithelium

PARP-1 Poly-ADP-ribose polymerase 1

SH2 Src Homology 2 domain

SH3 Src Homology 3 domain shRNA short hairpin RNA

TCGA The Cancer Genome Atlas project

5 Introduction

Ovarian cancer is the second most common gynecologic malignancy and the most common cause of gynecologic cancer death in the United States (1). In 2014, ovarian cancer accounts for 21,980 estimated new cases and 14,270 estimated deaths (2). Histologically, ovarian epithelial carcinomas can be divided into high-grade serous, low-grade serous, endometrioid, mucinous, and clear cell subtypes. Clinically, high-grade serous ovarian cancer (HGSOC) accounts for 70-80% of all ovarian carcinomas and is characterized by its de novo invasive nature. The origin of high-grade serous ovarian cancer (HGSOC) has thought to be from ovarian epithelial cells, however, more recent studies have proposed the origin to be in fallopian tube fimbria (3). Due to similar clinical behavior, epithelial carcinoma, fallopian tubal, and peritoneal carcinomas are considered a single clinical entity for treatment. The first-line chemotherapy treatment for advanced epithelial ovarian cancer is a platinum plus taxane combination.

HGSOCs are initially sensitive to platinum-taxane treatment but develop resistance in approximately 25% of patients within six months (4). To this date, ovarian cancer remains one of the most lethal cancer types due to the advanced stage at diagnosis and lack of effective systemic therapy.

It is known that ovarian cancer is characterized by a combination of germline and somatic mutations. Germline mutations in BRCA1/2 account for approximately 13% of HGSOCs, while somatic mutations are largely dominated by TP53 mutations (5, 6). A smaller proportion of

HGSOCs are attributable to germline mutations in mismatch repair genes such as MSH2, MSH6,

MLH2, and PMS2 (Lynch syndrome) (5). Several common chromosomal region copy number variations have also been observed in ovarian cancers. However, most of the somatic aberrations have not been well characterized.

6

Comprehensive characterization and analysis of the human cancer genome

Due to a significant decrease in the cost of high-throughput sequencing technologies, it has become possible to comprehensively characterize and analyze the cancer genome. The

Cancer Genome Atlas (TCGA) project, sponsored by the National Cancer Institute and the

National Human Genome Research Institute, began a series of large-scale sequencing studies to catalogue all genetic alterations responsible for cancer. The pilot studies focused on the characterization of two tumor types, glioblastoma multiforme and ovarian carcinoma. Over 500 primary tumor samples from each tumor type underwent whole-exome sequencing, copy number variation analysis, gene expression profiling, and DNA methylation analysis (7, 8). All data from the publication were made available to the public through the TCGA Data Portal. Since the pilot phase, TCGA project has expanded the characterization effort and published the data on other major cancer types, including colon and rectal cancer, breast cancer, squamous cell lung cancer, and lung adenocarcinoma (9-12).

The information generated by sequencing thousands of tumor samples presents an opportunity as well as a challenge. It has pushed for development of newer computational analytic tools to decode the cancer genome, including newer methods to detect mutations and copy number variations. The wealth of data points and robust analytic methods provide a unique opportunity to investigate the biology and pathogenesis of cancer.

Genomic landscape of ovarian carcinoma

To catalog all the molecular aberrations present in HGSOC, The TCGA network performed a large-scale, multiplatform genomic profiling study (13). A total of 489 clinically-

7 annotated, stage II-IV, HGSOC tumors were analyzed for copy number variation, mRNA expression, microRNA expression, and DNA promoter methylation. Whole-exome sequencing was performed on 316 of these tumor samples. The results and correlated clinical outcomes were made accessible to the public.

The mutation analysis from sequenced HGSOCs has revealed a predominance of TP53 mutations in 96% of the sequenced tumors, similar to what has been described in previous literature. Approximately 9% of BRCA1 and 8% of BRCA2 contained germline mutations.

Another six genes were found to be significantly mutated, although to a much lesser extent: RB1

(2%), NF1 (4%), FAT3 (6%), CSMD3 (6%), GABRA6 (2%), and CKD12 (3%). Further analysis has shown that these mutations, in addition to mRNA expression aberrations, represent alterations in key cancer-associated signaling pathways, such as RB signaling, RAS/PI3K signaling, and NOTCH signaling pathways.

In contrast to a surprisingly simple spectrum of mutations, the copy number analysis of

HGSOC identified a large number of recurrent somatic copy number alterations including eight recurrent chromosomal gains and 22 recurrent losses. The GISTIC (Genomic Identification of

Significant Targets In Cancer) (see Methods and Materials section) analysis revealed 31 focal amplifications that contained eight or fewer genes and 50 focal deletions (Fig. 1). These amplified regions encode 1825 genes including highly-amplified known oncogenes such as

CCNE1 and MYC, each amplified in more than 20% of tumors. However, the driver genes in the majority of the recurrently amplified regions remain unidentified.

The mutation spectrum and copy number variation distinguish HGSOC from other histologic subtypes of ovarian cancer. Clear-cell ovarian cancer and endometrioid ovarian cancer tumors contain fewer TP53 mutations but more recurrent ARID1A and PIK3CA mutations (14-

8 16). Mucinous ovarian cancer tumors have higher frequency of KRAS mutations (17). These differences mark HGSOC as a distinct subtype of ovarian cancer that should be treated with different therapeutic approaches.

Identifying therapeutic targets through integrated genomics approach

The Cancer Genome Atlas project has provided a comprehensive survey of genetic and epigenetic alterations that can play a causal role in ovarian cancer. However, only a subset of the large number of these alterations contributes to the cancer phenotype, mixed with other benign alterations as results of the genomic instability in cancer cells. In order to decipher which molecular events are important in driving cancer initiation and/or progression, we need functional studies to provide biological context and clues to the mechanical basis for oncogenesis.

In parallel to the genome characterization efforts, the Hahn laboratory has initiated

Project Achilles, a systematic effort to identify cancer dependencies (18, 19). We performed a genome-scale, pooled short hairpin RNA (shRNA) screen in 102 cell lines that include 25 ovarian cancer cell lines. Each cell line was lentivirally infected with a pool of shRNAs consisted of 54,020 shRNAs targeting 11,194 genes. The cell line was propagated for at least 16 doublings.

The abundance of shRNA sequences at the endpoint was measured by microarray hybridization and compared to the initial reference pool.

This complementary approach to genome sequencing effort has identified ovarian cancer lineage-specific dependencies such as PAX8 (20). PAX8 is focally amplified in 16% of HGSOCs and expressed at a higher level in ovarian cancer cell lines than non-ovarian cell lines. Through

Project Achilles, PAX8 has found to be an essential gene for ovarian cancer cell survival by three shRNA scoring methods (see Methods and Materials section). Suppression of PAX8 with individual shRNAs targeting PAX8 induced apoptosis selectively in ovarian cancer cells. The

9 evidence strongly suggests PAX8 as a lineage-specific oncogene and a potential therapeutic target in HGSOC.

RNA interference represents one of the many approaches to functionally annotate the cancer genome (21). Gain-of-function approaches with overexpression systems can also be used to explore gene functions in cancer. An increasingly complete collection of human ORFs (open reading frames) has become available (22-24). The most recent collection includes 16,100 fully sequenced human ORFs, representing over 13,500 human genes (23). It enables genome-scale high-throughput screen in mammalian cells. The application of the human ORF expression library has contributed to a better understanding of many aspects of cancer biology such as oncogenic RAS signaling and resistance to MAPK pathway inhibition (25, 26).

Oncogene validation through models of transformation

In addition to genetic alterations in primary tumors and essentiality to cancer cell survival, oncogenes should demonstrate ability to drive malignant transformation in a cell or animal model. Previous work has shown that a limited set of genetic changes is sufficient to transform human cells (27). Specifically, co-expression of telomerase catalytic subunit (hTERT), the SV40 large T (LT) and small t antigens (ST), and an activated H-RAS (H-RASV12) can induce transformation in human HEK cells and ovarian epithelial cells (28). HRAS can be replaced by a combination of its downstream signaling components, MEK and AKT1. The model has facilitated the identification of oncogenes that emerged from genomic alteration characterization effort. An example of integrating the approach is the discovery of IKBKE as a novel breast cancer oncogene (29).

The oncogenic potential of IKBKE is discovered through an overexpression screen with a library of 354 human kinases and kinase-related ORFs. The screen searched for genes that could

10 transform HEK cells with a constitutively active MEK (MEK-DD) (HA1E-M cells) to replace activated H-RAS. IKBKE and AKT1 independently induced colony formation in soft agar growth assays and mouse xenograft assays with HA1E-M cells. Since AKT1 is known to cooperate with

MEK in transformation, IKBKE is able to replace AKT1 via activation of the NF--κB pathway.

Furthermore, IKBKE is amplified and overexpressed in primary breast tumors and suppression of

IKKε (IKBKE) expression induced cell death in IKBKE-amplified cell lines.

Here with a similar approach, by combining the output of ovarian cancer genome analysis with Project Achilles, we systematically interrogated 1825 recurrently amplified genes in ovarian cancer to identify genes that are essential in ovarian cancer cell lines that harbor such amplifications and identified FRS2 as an amplified and essential gene in HGSOC. FRS2 expression leads to transformation in immortalized HEK cells and ovarian epithelial cells.

11 Materials and methods

Analysis of shRNA screening data was performed by Barbara Weir and Hiu Wing Cheung. Flow cytometry and tumorigenicity studies were done by Eejung Kim. All other experiments described in the methods section below were done by me.

Analysis of TCGA primary tumor data

Regions of copy number amplification identified by Genomic Identification of Significant

Targets in Cancer (GISTIC) analyses were used from the TCGA study on high-grade serous ovarian cancer (13). All RefSeq genes within these regions of amplification (n = 1825) were identified and cross-referenced with genes interrogated in the Achilles screening library (n =

582). All primary HGSOC data were downloaded from the TCGA portal (http://tcga- data.nci.nih.gov/tcga). Genomic characterization data were visualized using the Integrative

Genome Browser (http://www.broadinstitute.org/igv). Mutual exclusivity analysis was performed using the cBio Portal for Cancer Genomics (30, 31), which uses different thresholds for scoring regions of copy number alteration.

Analysis of shRNA screening data

Data from genome-scale loss of function screening was processed as described (18).

Briefly, 54,000 shRNAs were lentivirally delivered to 102 cancer cell lines and the degree of representation of each shRNAs in the final cell population was measured by custom Affymetrix array. Normalization, variance stabilization and expression score calculation were conducted as specified in modified dCHIP method (19). Scores were median-adjusted per cell lines. Ovarian- specific gene dependencies were determined with three complementary methods: (i) 150 best

12 single shRNA or (ii) 300 second best shRNA or (iii) composite of all shRNAs for the gene using

KS statistics. 582 genes (5.2%) were selected from the union of three methods above.

To identify genes that were both amplified in ovarian tumors and essential in amplified cancer cell lines, each gene identified as amplified in primary ovarian tumors (1,825 genes) was tested across the entire panel of 102 cell lines screened. Only genes with more than 5 amplified cell lines were included in the study. Amplified genes that had mapped shRNAs with a P < 0.05 were identified as candidate genes.

Cell culture and generation of stable cell lines

All human cancer cell lines were cultured in previously described media supplemented with 10% fetal bovine serum (FBS, Sigma) (18). Immortalized human ovarian surface epithelial cells

(IOSE) (32) were maintained in 1:1 Medium 199: DMEM supplemented with 10% FBS.

CAL120, COV644, COV362, and CAOV3 cells were cultured in Dulbecco’s modification of

Eagle’s medium (Invitrogen) with 10% FBS. HCC1143, EFO21 cells were cultured in RPMI-

1640 medium (Invitrogen) with 10% FBS. NIH/3T3 cells were cultured in DMEM with 10% bovine calf serum. Lentiviruses were produced by transfection of 293T packaging cells with a three-plasmid system. To generate stable cell lines, cells were seeded into 6-well dishes for 24 h before infection with 0.3 ml of lentiviruses for 12 h in the presence of 8 μg/ml polybrene. After the incubation, medium was replaced with fresh medium for another 24 h before selection in media containing 2 μg/ml of puromycin or 10 μg/ml of blasticidin until the control cells were no longer viable.

13 Plasmids

Human FRS2 from the CCSB human ORFeome collection (22) was cloned into pLenti6.3-blast

(BamHI and BsrGI sites). The pLX304-LacZ was used as a control vector. The human

MEKD218, D222 (or MEKDD) fragment was removed from pBabe-puro-MEKDD plasmid (29) with BamHI and SalI and inserted into pLX304-Blasticidin. Lentiviral pLKO.1-puro-shRNA constructs were obtained from The RNAi Consortium or designed by custom oligo synthesis

(IDT). The shRNA constructs used are as follows: control shRNA targeting LacZ

(TRCN0000231710), FRS2-specific shRNAs (shFRS2#1: TRCN0000370440, shFRS2#2: 5’-

CTCTAAATGGCTACCATAATA-3’)

Cell proliferation assay

CAL120, COV644, HCC1143, EFO21, CAOV3, and COV362 cells (3 X 103) were seeded into each well of 96-well plates 24 h prior to infection. Six replicate infections were performed for control shRNAs and each gene-specific shRNA in the presence of 8 μg/ml polybrene for 24 h followed by selection with 2 μg/ml of puromycin. The ATP content was measured at 6 days post-infection by using CellTiter-Glo luminescent cell viability assay (Promega).

Anchorage-independent growth assay

Growth in soft agar was determined by plating 5×104 cells in triplicate in 4 ml of medium containing 0.35% Noble agar (BD Biosciences), which was placed on top of 4 ml of solidified

0.6% agar. Unstained colonies greater than 100 μm in diameter were counted 4 weeks after plating using Cell Profiler software (33).

14

Immunoblotting

Cell lysates were prepared by scraping cells in lysis buffer (50 mM Tris HCl (pH 8), 150 mM

NaCl, 1% Nonidet P40, 0.5% sodium deoxycholate, and 0.1% SDS) containing complete protease inhibitors (Roche) and phosphatase inhibitors (10 mM Sodium Floride and 5 mM

Sodium Orthovanadate). Protein concentration was measured by using BCA Protein Assay kit

(Pierce). An equal amount of protein (20 μg) was separated by NuPAGE Novex Bis-Tris 4-12% gradient gels (Invitrogen) and then transferred onto a polyvinylidene difluoride membrane

(Amersham). Antibodies against FRS2 (sc-8318) were purchased from Santa Cruz

Biotechnology. Antibodies for PARP (#9532), phospho-ERK1/2 (#9101), ERK1/2(#9102) were purchased from Cell Signaling Technology and antibody specific for β-actin was obtained from

Santa Cruz (sc-8432-HRP).

After incubation with the appropriate HRP-linked secondary antibodies (Bio-Rad), signals were visualized by enhanced chemiluminescence plus Western blotting detection reagents (Amersham). Alternatively, membrane was incubated with IRDye fluorescent secondary antibodies (LI-COR) and visualized by Odyssey quantitative fluorescence imaging system (LI-COR).

Real-time quantitative RT-PCR

Total RNA was extracted with RNeasy mini kit (Qiagen). Reverse transcription was performed using SuperScript III First-Strand Synthesis System (Invitrogen). Quantitative RT-PCR reactions were performed using SYBR green PCR Master Mix (Applied Biosystems). The primer

15 sequences used were obtained from MGH PrimerBank: FRS2 (forward: 5’‐

CTGTCCAGATAAAGACACTGTCC‐3’, reverse: 5’‐ CACGTTTGCGGGTGTATAAAATC

‐3’); GAPDH (forward: 5’-CCTGTTCGACAGTCAGCCG-3’, reverse: 5’-

CGACCAAATCCGTTGACTCC-3’). Triplicate reactions for the gene of interest and the endogenous control (GAPDH) were performed separately on the same cDNA samples by using the ABI 7900HT real-time PCR instrument (Applied Biosystems). The mean cycle threshold (Ct) was used for the ddCt analysis method.

Flow cytometry

Cells were collected, washed, and fixed with 70% ethanol at -20C for 4 hours. Fixed cells were washed, re-hydrated and re-suspended in propidium iodide staining solution (25ug/ml propidium iodide, Sigma P4862, 50ug/ml RNase A, Invitrogen 12091-021, in PBS) at room temperature for

30 minutes. Flow cytometry was done on BD LSR II flow cytometer (BD Biosciences). Debris and aggregates were gated out and the sub-G1 population was analyzed using FlowJo software.

Tumorigenicity assay

Female NCR/nude mice (Charles River Laboratories) were obtained at 6 weeks of age. All animal experiments were approved by the Dana-Farber Institutional Animal Care and Use

Committee. Tumor xenograft experiments were performed as described (29). NIH/3T3 cells expressing indicated constructs were trypsinized and collected in fresh media. Cells were washed and re-suspended in PBS at 106 cells per 100 ul. Cells were injected subcutaneously on left and right flanks, and upper back. Two mice were used for each experimental condition. 2 X 106 cells

16 were injected per site, three sites per mouse. Tumor injection sites were monitored for 3 months for tumor formation. Mice were euthanized when the largest tumor on mouse reached 2 cm in largest dimension. We attempted tumor experiments with HA1E-A cells but encountered a high background in control cells.

Statistical analysis

Unless otherwise indicated, one-way ANOVA was used (GraphPad). P < 0.05 was considered statistically significant. Fisher’s exact test was used for tumor formation assays and mutual exclusivity analysis. Two-tailed Student’s t test was used for pairwise comparisons. A log-rank test was performed for animal survival studies.

17 Results

Identification of FRS2 as an amplified and essential gene in ovarian cancer

High-grade serous ovarian cancers are characterized by high frequency, recurrent regions of copy number gain and loss. Recent genome-scale effort to characterize structural alterations in

HGSOC has identified 31 recurrently amplified chromosomal regions containing total of 1,825 genes (13). To systematically study previously unknown lineage-specific dependencies, we initiated a genome-scale effort (Project Achilles) to identify genes essential for proliferation/survival of a large number of well characterized cancer cell lines using loss-of- function genetics with short hairpin RNAs (shRNA) (19). Although recent studies suggest that established ovarian cancer cell lines do not fully recapitulate the genetic alterations found in high grade ovarian cancers (34, 35), here we have focused on those alterations found by the TCGA in human cancers and shared by these ovarian cancer cell lines. Using data from 102 cell lines of which 25 were from the ovarian lineage, we identified 582 ovarian-lineage specific gene dependencies (18). By looking at the intersection of genes involved in regions of recurrent copy number and essential in ovarian cancer cell lines, we identified 50 genes (Fig. 1, Supplementary

Table 1). Two of the 50 genes were previously identified as ovarian specific oncogenes (PAX8,

CCNE1) with similar analytical approach (18, 36).

Among the remaining genes, we focused on fibroblast growth receptor substrate 2 (FRS2) because FRS2 is (i) adaptor protein in the Fibroblast Growth Factor Receptor (FGFR) pathway,

(ii) is located on chromosomal region 12q15, which is focally amplified in 12.5% of 559 primary high-grade serous ovarian cancers characterized by TCGA (Fig. 2), and (iii) was among the top

100 genes that scored by our analysis of Project Achilles and copy number data in HGSOC. We also found a structurally similar chromosomal region amplification in other cancer types such as

18 breast invasive carcinoma, lung adenocarcinoma, lung squamous cell carcinoma, head and neck squamous cell carcinoma, gastric adenocarcinoma, and bladder urothelial carcinoma. We used

Genomic Identification of Significant Targets in Cancer Version 2.0 (GISTIC 2.0) algorithm to identify the peak of amplification, which corresponds to the highest level of copy number gain.

In ovarian cancer samples, we observed the overlap between the peak of amplification and the location of FRS2 gene. Furthermore, the focal amplification of 12q15 region in HGSOC is correlated with increased mRNA expression of FRS2, suggesting the functional relevance of the copy number gain (Fig. 3A and 3B). In addition, we also observed frequent amplification of

FGFR family of tyrosine kinase receptor genes in HGSOC. Strikingly, using the cBio portal, we found that HGSOC samples that harbor 12q15 amplifications were often mutually exclusive with

HGSOC that harbor FGFR1, FGFR2, FGFR3, and FGFR4 amplifications (Fisher’s exact test

P=0.028) (Fig. 4). This pattern of mutations is observed in commonly mutated genes in the same pathway, such as KRAS and EGFR mutations or TP53 and MDM2 mutations. These observations implicate FGF signaling through amplifications of FGFRs and FRS2 as a common event in

HGSOCs.

FRS2 is essential in cancer cell lines that harbor 12q15 amplification

To confirm that FRS2 was essential in FRS2 amplified cancer cell lines, we used two independent shRNAs to suppress FRS2 expression in three cell lines with 12q15 amplification

(CAL120_BREAST, COV644_OVARY, HCC1143_BREAST) and three cancer cell lines that contain normal copies of 12q15 (CAOV3_OVARY, EFO21_OVARY, COV362_OVARY). We used both breast and ovarian cancer cell lines since we found focal amplification of 12q15 in a large subset of the primary breast cancers (Fig. 2). Copy number data for these cell lines were obtained from the Broad Institute/Novartis Cancer Cell Line Encyclopedia (CCLE) (37) (Fig.

19 5A). We found that FRS2 suppression by two independent shRNAs significantly decreased the proliferation of cancer cell lines that harbor the 12q15 amplification, when compared to cells that exhibit diploid copy number at 12q15 or cells infected with control shRNA (Fig. 5B). The degree of FRS2 suppression in 12q15 amplified cell lines was validated by quantitative real-time PCR

(Supplemental Fig. 1). To demonstrate that FRS2 suppression induced apoptotic cell death in

12q15 amplified cell lines, we interrogated poly ADP-ribose polymerase (PARP) cleavage after suppression of FRS2. We found increased level of cleaved PARP in 12q15 amplified cell lines compared to cell lines without 12q15 amplification (Fig. 6A). Analysis of apoptosis by staining cells with propidium iodide confirmed higher level of apoptosis in 12q15 amplified cell lines

(Fig. 6B). Together, these findings demonstrate that cancer cells that harbor 12q15 amplification require FRS2 expression for proliferation and survival.

FRS2 induces oncogenic transformation

To determine if FRS2 can contribute to tumorigenesis by inducing transformation, we performed anchorage-independent growth assays and tumor xenograft experiments. In our prior studies, we have shown that human kidney epithelial cells are immortalized by co-expression of the human catalytic subunit of telomerase (hTERT) and the SV40 Early Region (HA1E cell) and the expression of oncogenic alleles of RAS confers the ability to grow in anchorage-independent manner (27). We had previously demonstrated the RAS oncogene can be replaced by combination of downstream effectors of RAS signaling pathway, such as constitutively-activated

MEK1 (MEK-DD) and AKT1 (myristoylated AKT) (29). In addition, we used the same genetic elements to immortalize human ovarian epithelial cells (IOSE) and fallopian tube epithelial cells and used this cell line to identify ovarian cancer oncogenes such as ID4 (38). We note that recent reports suggest that both fallopian tube and ovarian surface epithelial cells can serve as the origin

20 for HGSOC and have not noted differences in the transformation potential of cells from either lineage (38-42). As previous studies have shown that FRS2 preferentially activates MAPK pathway, we overexpressed FRS2 in HA1E cell lines expressing constitutively active myristoylated AKT (HA1E-A) to determine whether FRS2-mediated MAPK pathway activation complemented AKT pathway activation to induce transformation. We measured anchorage independent growth with FRS2 overexpression and found that FRS2 overexpression was sufficient to induce anchorage independent colony formation of HA1E-A cells compared to cells expressing the control LacZ (Fig. 7A). The number of colonies formed with FRS2 overexpression is significantly higher (P<0.001) compared to constitutively activated MEK, suggesting possible activation of additional pathways that contribute to the transformation process. We also conducted the same experiment in IOSE cells to show that FRS2 also induced transformation in ovarian epithelial cells (Fig. 7B).

Next, we determined whether expression of FRS2 also induced tumor formation in vivo by expressing FRS2 in NIH3T3 mouse fibroblast cells and implanting these cells subcutaneously in immunodeficient mice. At 11 weeks, we observed that tumors formed in 33% (2 out of 6) of the injection sites harboring cells expressing FRS2 but failed to observe any tumors in sites harboring control cells (Fig. 8). We note that since we implanted tumors in several sites in each mouse, and we terminated the experiment prior to observing tumor growth in all sites, these experiments may underestimate the tumorigenicity of these cells. These observations confirm that FRS2 overexpression can induce oncogenic transformation in human kidney fibroblasts or mouse fibroblasts by promoting anchorage-independent growth or in vivo xenograft tumor formation.

21 FRS2 amplification activates the MAPK pathway

Previous studies have shown that FRS2 is a critical mediator of FGFR signaling and plays an important role in activating MAPK and PI3K pathways (Fig. 9)(43, 44). We have confirmed FRS2 overexpression induces activation of MAPK pathway in 293 HEK cells and

IOSE cells by assessing phospho-Thr202/Tyr204 ERK1/2 levels (Fig. 10A). Conversely, suppression of FRS2 in FRS2-amplified cancer cell line resulted in a decrease in phospho-ERK levels (Fig. 10B). In contrast, we failed to observe a change in phospho-AKT when we overexpressed FRS2 (Fig. 10A). These observations suggest that FRS2 overexpression preferentially activates MAPK pathway in this context. This finding corroborates anchorage- independent growth assays where we observed that FRS2 was able to induce increased colony growth when expressed with Myr-AKT as compared to co-expression with MEK-DD in HEK cells.

22 Discussion

Here we identified FRS2 as one of the 50 genes that are recurrently amplified in high-grade serous ovarian cancers (HGSOC) and essential to survival in ovarian cancer cell lines. FRS2 belongs to the 12q15 genomic region that is focally amplified in 12.5% of HGSOC. Using independent shRNAs targeting against FRS2, we showed the expression of FRS2 was essential for survival in cancer cells with 12q15 amplification. We also discovered that overexpression of

FRS2 in immortalized kidney fibroblast or ovarian epithelial cells promoted anchorage independent growth and tumorigenesis in mice. Together these observations nominate FRS2 as an amplified oncogene in a subset of high-grade serous ovarian cancers.

FRS2 and FGFR signaling in cancer

The discovery of FRS2 as an amplified oncogene adds to the family of FGFR signaling components that are critical to tumorigenesis. There is ample evidence for aberrant FGF signaling in the pathogenesis of many cancer types. It is known that mutations or amplifications of FGFRs are frequent and inhibitor-sensitive in bladder cancer (45), gastric cancer (46), endometrial cancer (47), and non-small cell lung cancers (48, 49). Large-scale genome-wide association studies have also linked breast cancer risk loci to FGFR2 (50).

FGF receptor family comprises of 4 highly conserved transmembrane tyrosine kinase receptors. A fifth receptor, FGFR5, does not have tyrosine kinase domain but may play a role in negative signaling regulation (51). FGF receptors function by dimerization similar to epidermal growth factor receptors. Ligand-dependent dimerization induces a shift in receptor protein structure that activates the intracellular kinase domain, which results in trans-phosphorylation of the intracellular tyrosine residues. Phosphorylated tyrosine residues on the intracellular tail recruit adaptor proteins, leading to propagation of signals to multiple pathways, including the

23 MAPK pathway (Fig. 9). One of the key adaptor proteins in the FGFR family is fibroblast growth factor receptor substrate 2 (FRS2).

FRS2 functions as adaptor protein largely specific for FGFRs, although it can interact with other tyrosine kinase receptors such as RET, anaplastic lymphoma kinase (ALK), and neurotrophic tyrosine kinase receptor type 1 (NTRK1) (52). Upon FGFR activation, FRS2 attaches to the juxtamembrane region of FGFRs through its phosphotyrosine-binding (PTB) domain, and functions close to the intracellular cell surface. Several tyrosine residues on FRS2 are phosphorylated upon binding to activated FGFRs, serve as recruitment sites for adaptor proteins sons of sevenless (SOS) and growth factor receptor-bound 2 (GRB2) to activate

RAS/RAF/MAPK pathway. A separate complex involving GRB2-associated binding protein 1

(GAB1) activates PI3K/AKT pathway. These signaling pathways lead to expression of target genes that control critical cellular functions such as proliferation, survival, migration, and differentiation.

In addition to FGFR alterations, more recent studies revealed the importance of FGF ligands, such as FGF19 amplification in liver cancer (53) and the therapeutic effect of neutralizing anti-FGF antibodies (54). The mammalian FGF family consists of 18 ligands that interact with 4 FGFRs (FGFR1-4). FGFs are glycoproteins that are secreted from the cell to the extracellular matrix. Cell surface heparin sulphate proteoglycans (HPSGs) sequester FGFs and stabilize the FGF-FGFR interaction (55, 56). The specificity of FGF-FGFR interaction depends on the ligand-binding capacity of FGFR, alternative splicing of FGFR, as well as tissue-specific expression of ligands and receptors (57, 58).

Deregulation of FGF signaling in cancer occurs mostly in a ligand-independent manner, through activating mutation, translocation, or amplifications in FGF receptors. A second

24 oncogenic mechanism of FGF signaling is aberrant ligand-dependent signaling. There is strong evidence for autocrine FGF signaling loop in a subset of melanoma tumors, where high levels of both FGFR1 and FGF2 are expressed (59). Another example of autocrine loop is the frequent amplification of FGF1 reported in ovarian cancer with associated poor survival (60).

Adaptor protein as a novel class of oncogenes

Here we show a new functional class of adaptor proteins as driver oncogene in ovarian cancer. The adaptor proteins lack intrinsic enzymatic activities but mediate protein-protein interactions that drive protein complex formation. Classic examples of adaptor proteins include

GRB2 in receptor tyrosine kinase signaling (61) and MYD88 in NF-κB signaling (62). FRS2 was originally discovered as a docking site for coordinated assembly of a multi-protein complex that include GRB2, GAB1, and SOS1, and serves a critical role in the FGFR signaling pathway

(Fig. 4A)(43, 63). Unlike the signaling-amplifying activity of kinases, adaptor proteins are bottlenecks of the signaling pathway due to their stoichiometric relationship with interacting partners. Therefore, amplification or overexpression of the adaptor proteins can significantly alter the flux of the signal, thus carry important therapeutic implications such as mediating resistance to targeted therapy against receptor tyrosine kinase or conferring de novo sensitivity to signaling pathway inhibitors. Our laboratory has previously identified CRKL, an adaptor protein involved in RAS and RAP signaling, as an amplified oncogene in NSCLC (64). It was demonstrated that CRKL overexpression can mediate resistance to EGFR inhibitor in EGFR- mutant lung cancer cells and its amplification has been observed in gefitinib-resistant lung tumors. More recently, through a multiplexed in vivo transformation screen, we found another adaptor protein, GAB2, as an amplified ovarian cancer oncogene that activates PI3K signaling

(65). Ovarian cancer cells with GAB2 alteration are sensitive to PI3K-pathway inhibition. An

25 independent analysis of TCGA datasets across 16 cancer types has generated 75 amplified genes with druggable properties, including FRS2 and EGFR family adaptors GRB2 and GRB7 (66).

These findings suggest a new class of targetable oncogenes that are sensitive to exhibiting RTK signaling pathway inhibitors and present a new therapeutic opportunity to those patients with such genetic alterations.

FRS2 is amplified in multiple cancer types

In addition to HGSOC, 12q15 amplification containing FRS2 is found in other cancer types. 12q15 amplicon containing FRS2 is focally amplified in 9.2% of breast invasive carcinomas. Indeed, we found that breast cancer cell lines that harbor 12q15 amplification are also sensitive to suppression of FRS2 (Fig. 5). Furthermore, new evidence has suggested the oncogenic role of FRS2 and 12q15 amplification in high-grade liposarcomas (67, 68). Amplified region containing FRS2 gene was found via high-resolution SNP/copy number variation microarray of 47 well-differentiated and dedifferentiated liposarcomas (68). A separate study demonstrated sensitivity of FRS2-amplified high-grade liposarcoma cell lines to FRS2 suppression through shRNAs (69). In 40 glioma tumor samples, FRS2 is among 10 genes that are recurrently amplified and overexpressed (70). These studies support FRS2 as a bona fide oncogene in a variety of cancers and a potential therapeutic target for a subset of cancers that harbor such amplification.

12q15 amplified region contains multiple genes

The 12q15 genomic region contains 15 genes with FRS2 residing at the copy number peak of the amplicon (Supplementary Table 2). Prior work in high-grade liposarcoma, which has a broader region of amplification (12q13-12q15) than HGSOC, has suggested in addition to

FRS2, other genes such as CDK4 and MDM2 may be driving events (67). Although neither

26 CDK4 nor MDM2 is located within the 12q15 amplified region in HGSOC, we do not preclude the possibility that other genes in the genomic region may cooperate to drive various stages of tumorigenesis. Indeed, we recently demonstrated that multiple genes resident in a recurrently amplified region (3q26) contribute to cell transformation by inducing different cancer associated phenotypes, suggesting that further studies involving other assays will be necessary to investigate the function of these other genes (71).

Therapeutic approaches to targeting FRS2 and FGFR pathway

Several small molecular tyrosine kinase inhibitors against FGFRs exist in the preclinical or early clinical trial phase. Majority of the inhibitors are ATP-competitive VEGFR2 inhibitors due to structural similarity between VEGFR and FGFR kinase domains (72). Dual inhibition of

FGFR and VEGFR targets both tumor angiogenesis and cell proliferation. Pazopanib (Votrient,

GlaxoSmithKline) was approved by Food and Drug Administration in 2009 for renal cell carcinoma and soft tissue sarcoma. However, further clinical trial to extend therapy to advanced ovarian cancer did not suggest any benefit with overall survival or progression-free survival (73).

Dovitinib (TKI258), an inhibitor of FGFR1, FGFR2, FGFR3, has presented antitumor activity in

FGFR-amplified breast cancers in preclinical and early clinical data (74). Dovitinib monotherapy inhibits proliferation in FGFR1 and FGFR2 amplified breast cancer cell lines and tumor growth in FGFR1-amplified breast cancer xenografts. Dovitinib also showed activity in a subset of 81 patients with metastatic breast cancer. 5 patients (25%) with FGFR1-amplified tumor had unconfirmed response or stable disease for more than 6 months, compared to 1 patient (3%) with

FGFR1-non-amplified tumor (74).

Therapeutic antibodies can be specific to a particular FGFR and minimize the side effects of pan-FGFR inhibition. Single-chain Fv antibody (scFv) fragments that target FGFR3 has

27 shown anti-proliferative effect in bladder cancer cell lines (75). A separate study examined the effect of anti-FGFR3 monoclonal antibody (R3Mab) in FGFR3-mutated bladder carcinoma and t(4;14)-positive multiple myeloma (76). R3Mab blocks both FGF ligand binding and FGFR3 dimerization, resulting in inhibition of both mutated FGFR3 in bladder tumors and wild-type, translocated FGFR3 in multiple myeloma.

Adaptor proteins, such as FRS2, lack tyrosine or serine/threonine kinase domains and present a unique challenge to rationalized drug design. A new approach to cancer therapy has been the development of small molecules targeting protein-protein interactions (77). Targeting protein-protiein interface has been difficult in the past due to several reasons. The interaction surface is often smooth and void of any clefts and pockets for drug binding. The surface often comprises non-contiguous amino acid residues in the polymer chain. Peptides derived from short contiguous chain are poor chemical starting sequences. Furthermore, protein-protein interaction surfaces are relatively large (1500-3000 square Angstrom) compared to protein-small molecule interactions (300-1000 square Angstrom) (78).

Despite the challenges, several studies have discovered small subsets of contact surface amino acid residues, named “hotspots”, are critical for the protein-protein interaction (79-82).

ABT-737 is an example of newly emerged protein-protein interaction inhibitors that targets members of the B-cell lymphoma 2 (BCL-2) family. ABT-737 and its derivative ABT-263 bind to the hydrophobic helical domains of BCL-XL, BCL-2 and BCL-W, which are important regulators of apoptotic cell death (83). ABT-263 (Navitoclax, Abbott Laboratories) is currently undergoing Phase I/II trial for multiple lymphoid and solid malignancies.

Targeting FRS2 with protein-protein interaction inhibitor adds another layer of specificity toward targeted therapy. Conformations of the tyrosine kinase domain and serine/threonine

28 kinase domain are highly conserved across species, especially in the substrate-binding region.

Therefore kinase inhibitors are often non-specific and present with wide range of toxicities.

Since FRS2 demonstrates specificity toward tyrosine kinase receptors and adaptor proteins, inhibition of FRS2 should have limited toxicity and increased efficacy for patients compared to kinase inhibition.

29 Conclusion

Here by combining the output of ovarian cancer genome analysis with Project Achilles, we systematically interrogated 1825 recurrently amplified genes in ovarian cancer to identify genes that are essential in ovarian cancer cell lines that harbor such amplifications and identified

FRS2 as an amplified and essential gene in HGSOC. FRS2 expression leads to transformation in immortalized HEK cells and ovarian epithelial cells. Together, these observations strongly support the conclusion that FRS2 amplifications play an important role in a subset of ovarian tumors and identify FRS2 as a potential therapeutic target.

This study is an example and underlines the strengths of complementary genomic approaches that use efficient high-throughput methods to assess functional consequences of genomic alterations and develop therapeutic targets toward clinical translation. The discovery of

FRS2 as an oncogene also highlights adaptor proteins as a new class of oncogenes that were previously less studied in the cancer biology context. Lastly, the recent advancement in disrupting protein-protein interaction creates an exciting future that may become the therapeutic strategy toward this new class of oncogenes.

30 Summary

Ovarian cancer is the most common cause of gynecologic cancer death in the United

States. Despite aggressive surgical cytoreduction and chemotherapy, ovarian cancer remains one of the most lethal cancer types due to advanced stages at diagnosis and lack of effective systemic therapy. High-grade serous ovarian cancers (HGSOC) contribute 70-80% of all ovarian cancers and are characterized by widespread recurrent regions of copy number gain and loss.

In this study, I interrogated 50 genes that are recurrently amplified in HGSOC and essential for cancer proliferation and survival in ovarian cancer cell lines. FRS2 is one of the 50 genes located on chromosomal region 12q15 that is focally amplified in 12.5% of HGSOC. The focal amplification of 12q15 region in HGSOC is correlated with increased mRNA expression of

FRS2, suggesting the functional relevance of the copy number gain. We also found a structurally similar chromosomal region amplification in other cancer types such as breast invasive carcinoma, lung adenocarcinoma, lung squamous cell carcinoma.

FRS2 amplified cancer cell lines are dependent on FRS2 expression and FRS2 suppression in 12q15 amplified cell lines induced apoptotic cell death. FRS2 overexpression in immortalized human cell lines conferred the ability to grow in an anchorage independent manner and as tumors in immunodeficient mice. FRS2, an adaptor protein in the FGFR pathway, induces downstream activation of Ras-MAPK pathway. These observations identify FRS2 as an oncogene in a subset of HGSOC that harbor FRS2 amplifications.

Together, these observations strongly support the conclusion that FRS2 amplifications play an important role in a subset of ovarian tumors and identify FRS2 as a potential therapeutic target. This study underlines the power of complementary genomic approaches that use efficient high-throughput methods to assess functional consequences of genomic alterations and develop

31 therapeutic targets toward clinical translation. The discovery of FRS2 as an oncogene also highlights adaptor proteins as a new class of oncogenes and potential next generation of therapeutic targets.

32 Figures and Legends

Figures 6, 8, and Supplemental Figure 1 were generated by my colleague Eejung Kim and used here with her permission. I generated all other figures.

Amplified genes in Ovarian-specific ovarian cancer 182550 582 essential genes (TCGA) (Project Achilles)

POOLED shRNA PLASMID LIBRARY PACKAGED IN VIRUS

Infect cells

REFERENCE TEST CONDITIONS CONDITIONS

Harvest genomic DNA

Amplify hairpin regions by PCR

Cut with Xho I

Hybridize to microarray

Identify hairpins

NO CHANGE ENRICHED DEPLETED

Identify genes

Figure 1. Schematic diagram of oncogene discovery through integrated cancer genomics.

Bottom left: copy number profiles of ovarian carcinoma representing 31 focally amplified and 22 focally deleted regions (TCGA network, 2011) (7). Within the amplified regions, 1825 amplified genes have been identified. Bottom right: flow chart showing a pooled RNAi screening strategy

33 to identify cancer vulnerabilities (Luo B et al., 2008) (84). 582 genes are essential to ovarian cancer cell lines from a genome-wide shRNA screen in 102 cancer cell lines. FRS2 is one of the

50 genes that are recurrently amplified in primary ovarian tumors and essential for ovarian cancer cell proliferation and survival.

q21.31

q21.2

q21.1

q15

q14.2 Chromosome 12

q13.3

q13.13

Ovarian Breast Lung Lung Squamous AC H&N Deletion Neutral Amplification Bladder Stomach

Figure 2. Copy number profile along chromosome 12q of human tumor samples. FRS2 was amplified in multiple cancer types including ovarian, breast, lung squamous, lung adenocarcinoma, stomach, head and neck (H&N), and bladder. Each vertical line represents one tumor sample. Red represents copy number gain, Blue represents copy number loss (Luo et al.,

2014) (85).

34

A B

15 3.5

3

10 2.5

2

1.5 5 1 0.5 R = 0.33598 0 0 FRS2 mRNA Z-score mRNA FRS2 -0.5 FRS2 mRNA Z-score mRNA FRS2

-1 5 -1.5 Loss Normal Gain Amp -1.5 -0.5 0.5 1.5 2.5 3.5 4.5 FRS2 copy number FRS2 copy number

Figure 3. Level of FRS2 mRNA expression in primary tumors correlates with the copy number.

A. Copy number is divided into 4 categories based on log2 of copy numbers. “Amplification” is defined as Log2(Copy number) more than 1; “Gain” is between 0.2 and 1; “Normal” is between -

0.2 and 0.2; “Loss” is less than -0.2.

B. Positive linear correlation between FRS2 expression and copy number in primary high-grade ovarian serous adenocarcinoma.

35 Ovarian Serous Cystadenocarcinoma

FRS2

FGFR1

FGFR2

FGFR3

FGFR4

Figure 4. FRS2 amplification and FGFR1, FGFR2, FGFR3, and FGFR4 amplifications are mutually exclusive in high-grade serous ovarian cancers. Each red column represents a primary tumor with FRS2 amplification. Each grey column with green square represents a mutation within FRS2 gene locus. Data was analyzed using the cBio portal rather than GISTIC algorithm.

36 A B ** 0.8 CAL120

Genomic Position on Chromosome 12 HCC1143 FRS2 COV644 66 Mb 68 Mb70 Mb 72 Mb 74 Mb 0.6 CAL120 CAOV3 COV644 HCC1143 COV362 EFO21 0.4 COV362 EFO21 CAOV3

Relative Proliferation Relative 0.2 ShFRS2 #1 Deletion Neutral Amplification ShFRS2 #2

0.0

Amplified Non-amplified

Figure 5. Suppression of FRS2 decreases the proliferation of ovarian and breast cancer cells harboring 12q15 amplification.

A. SNP array colorgram showing genomic amplification of chromosome 12q15 in ovarian and breast cancer cell lines. Red represents copy number amplification, blue represents copy number deletion.

B. Consequences of FRS2 suppression on the proliferation of cancer cell line that either harbor

12q15 amplification (CAL120, HCC1143, COV644) or normal copy number of 12q15 (CAOV3,

COV362, EFO21) normalized to cells treated with shLacZ. Red: cell lines treated shFRS2 #1.

Black: cell lines treated with shFRS2#2. **P < 0.01 compared to control shLacZ, Student’s t test was used.

37 A CAL120 COV644 HCC1143 COV362 EFO21 CAOV3 shLacZ #2 shFRS2 shLacZ shFRS2 #1 shFRS2 #2 shLacZ shFRS2 #1 shFRS2 #2 shLacZ #1 shFRS2 #2 shFRS2 shLacZ #2 shFRS2 shLacZ #2 shFRS2 shFRS2 #1 shFRS2 #1 shFRS2 shFRS2 #1 shFRS2

PARP Cleaved PARP

FRS2

Actin

B 25

20

15

10

5 Sub-G1 fraction (%)

0 shLacZ shLacZ shLacZ shLacZ shLacZ shLacZ shFRS2#1 shFRS2#2 shFRS2#1 shFRS2#2 shFRS2#1 shFRS2#2 shFRS2#1 shFRS2#2 shFRS2#1 shFRS2#2 shFRS2#1 shFRS2#2 CAL120 COV644 HCC1143 COV362 EFO21 CAOV3

Figure 6. Suppression of FRS2 induces cancer cell death through apoptosis.

A. Increased apoptosis in FRS2 amplified cell lines (red) upon FRS2 suppression, shown by increased PARP cleavage (Eejung Kim).

B. FRS2 suppression induces apoptosis in FRS2 amplified cell lines, shown by propidium iodide staining and flow cytometry (Eejung Kim).

38

300 *** A *** 250

200 LacZ

150

Colony Number 100 FRS2 50

0 LacZ FRS2 MEK-DD HA1E + AKT

60 B 50 GAB2 40 ** 30 20 FRS2

Colony Number 10 0 LacZ FRS2 GAB2 KRAS LacZ IOSE

IOSE

Figure 7. FRS2 overexpression potentiates tumorigenicity.

A. FRS2 promotes anchorage-independent growth in HA1E-A cells compared to LacZ control.

MEK-DD, a constitutively active MEK, is positive control. Right, images of soft agar colonies formed by HA1E-A with either FRS2 or control vector overexpression.

B. FRS2 promotes anchorage independent growth of IOSE (immortalized human ovarian epithelial) cells. GAB2 is a similar adaptor protein known to transform ovarian epithelial cells.

39 **P < 0.01, ***P < 0.001 compared to respective control vectors, Student’s t test.

) 3 Tumor volume (mm 10 005000 3000 01000

Control FRS2 KRAS G13D

Figure 8. FRS2 overexpression promotes tumorigenicity in vivo. 3T3 cells with FRS2 overexpression were able to form tumor in mouse xenograft models compared to LacZ control.

Constitutively active KRAS G13D was used as positive control. Within each condition, tumors from the same mouse were annotated with same color (Eejung Kim).

40 FGFR1 FGFR2 FGFR3 FGFR4

FGF FGF

FRS2 GRB2 Sos Ras Raf

P P P P MEK P P MAPK P P Pathway

MAPK

Figure 9. FRS2 promotes tumorigenesis via activation of MAPK pathway. FRS2 functions as an adaptor protein in the fibroblast growth factor receptor signaling pathway (Luo et al., 2014, adapted from Turner and Grose) (85).

41 A B shLacZ shFRS2 #1 shFRS2 #2 FRS2 LacZ LacZ FRS2 FRS2 FRS2 FRS2

pERK1/2 pERK1/2 pERK1/2

ERK1/2 ERK1/2 ERK1/2 pAKT pAKT Actin AKT AKT 293T CAL120 Actin

IOSE

Figure 10. FRS2 overexpression and suppression affects phosphor-ERK levels.

A. Effect of FRS2 overexpression on phosphorylation of ERK in 293T cells and ovarian epithelial cells.

B. Effect of FRS2 suppression on phosphorylation of ERK in cancer cell line with 12q15 amplification.

42 Supplemental Table 1.

1 PAX8 26 MYST3 2 APOBEC2 27 OR2T1 3 CDK2 28 PABPC1 4 CXCL1 29 PABPC4 5 HRASLS 30 PEX13 6 NPC1 31 PNPT1 7 OR14A16 32 PRB3 8 PRKCQ 33 PRKCE 9 ABHD8 34 PTPRB 10 ADAMTSL4 35 RANBP9 11 AFP 36 RHAG 12 AGPS 37 RPTOR 13 AKR1CL1 38 SENP5 14 ALG8 39 SLC38A2 15 APOD 40 SLC45A2 16 C18orf45 41 SLC4A4 17 CCNE1 42 TACC3 18 FRS2 43 TBC1D16 19 GUCA1A 44 TFEB 20 HNRNPA3 45 TTR 21 HNRNPU 46 TXNDC5 22 LASS2 47 UQCR10 23 MRPL12 48 ZFP36L2 24 MTA3 49 ZNF398 25 MYLIP 50 ZNF623

Supplemental Table 1. List of 50 genes that are specifically essential and amplified in ovarian cancer. The list is generated by overlapping two lists of genes, one containing 1825 recurrently amplified genes in 559 high-grade serous ovarian cancers, another containing 582 genes essential to 25 ovarian cell lines. The order of the 50 genes does not denote a functional ranking.

43 Supplemental Table 2. Ovarian chr12:69692322-71120515 #bin name chrom strand txStart txEnd cdsStart cdsEnd exonCount exonStarts exonEnds name2 1117ENST00000261267.2 chr12 + 69742120 69748014 69742188 69746999 469742120,69769742324,697LYZ 1117ENST00000549690.1 chr12 + 69742163 69747275 69742188 69747045 369742163,69769742324,697LYZ 1117ENST00000548839.1 chr12 + 69742165 69744286 69742188 69744066 269742165,69769742324,697LYZ 1117ENST00000548900.1 chr12 - 69747272 69748005 69748005 69748005 269747272,69769747388,697RP11-1143G9.4 1117ENST00000247843.2 chr12 + 69753482 69784576 69753752 69784096 769753482,69769753803,697YEATS4 1117ENST00000548020.1 chr12 + 69753489 69784411 69753752 69784096 569753489,69769753803,697YEATS4 1117ENST00000549261.1 chr12 + 69843249 69854504 69854504 69854504 369843249,69869843468,698RP11-956E11.1 1118ENST00000299293.2 chr12 + 69864128 69973559 69962810 69968735 1069864128,69969864310,699FRS2 1118ENST00000549921.1 chr12 + 69864154 69968744 69962810 69968735 969864154,69969864310,699FRS2 1118ENST00000550389.1 chr12 + 69864185 69973562 69962810 69968735 769864185,69969864310,699FRS2 1118ENST00000397997.2 chr12 + 69924602 69973559 69962810 69968735 869924602,69969924740,699FRS2 139 ENST00000299300.6 chr12 + 69979113 69995350 69979301 69995105 16 69979113,699 69979304,699CCT2 139 ENST00000544368.2 chr12 + 69979239 69995305 69979301 69995232 15 69979239,69969979304,699CCT2 139 ENST00000543146.2 chr12 + 69979445 69995345 69980595 69995105 16 69979445,69969979789,699CCT2 1118ENST00000550871.1 chr12 + 69982764 69985840 69985840 69985840 369982764,69969982851,699CCT2 1119ENST00000361484.3 chr12 - 70002350 70004942 70003784 70004618 170002350, 70004942, LRRC10 1119ENST00000331471.4 chr12 - 70037333 70093141 70037470 70091578 1070037333,70070037567,700BEST3 1119ENST00000488961.1 chr12 - 70047388 70083056 70048686 70072668 670047388,70070049593,700BEST3 1119ENST00000330891.5 chr12 - 70047388 70093196 70048686 70091578 1070047388,70070049593,700BEST3 1119ENST00000553096.1 chr12 - 70047598 70093065 70048686 70087616 870047598,70070049593,700BEST3 1119ENST00000476098.1 chr12 - 70063458 70093131 70064225 70072668 770063458,70070064358,700BEST3 1119ENST00000266661.4 chr12 - 70077018 70093256 70078332 70087616 370077018,70070078388,700BEST3 1119ENST00000551160.1 chr12 - 70078187 70093175 70078332 70087616 470078187,70070078388,700BEST3 1119ENST00000393365.1 chr12 - 70078332 70093141 70078332 70087616 570078332,70070078388,700BEST3 1119ENST00000533674.1 chr12 - 70081038 70093124 70093124 70093124 470081038,70070081239,700BEST3 139 ENST00000501387.1 chr12 - 70107412 70132348 70132348 70132348 4 70107412,70170109367,701RP11-588G21.2 139 ENST00000501300.1 chr12 - 70116101 70132342 70132342 70132342 3 70116101,70170116351,701RP11-588G21.2 1120ENST00000247833.7 chr12 + 70132460 70211157 70149188 70209226 1170132460,70170132811,701RAB3IP 1120ENST00000378815.6 chr12 + 70132641 70190427 70149188 70189184 670132641,70170132811,701RAB3IP 1120ENST00000483530.2 chr12 + 70132641 70209503 70149188 70206801 1070132641,70170132811,701RAB3IP 1120ENST00000325555.9 chr12 + 70132641 70209503 70188228 70209226 1170132641,70170132811,701RAB3IP 1120ENST00000550536.1 chr12 + 70133169 70216984 70133626 70209226 1170133169,70170133649,701RAB3IP 1120ENST00000362025.5 chr12 + 70133179 70209503 70133626 70206801 1070133179,70170133649,701RAB3IP 1120ENST00000551641.1 chr12 + 70172729 70210961 70188228 70209226 970172729,70170172961,701RAB3IP 1120ENST00000553099.1 chr12 + 70172746 70210897 70188228 70209226 970172746,70170172876,701RAB3IP 1120ENST00000550847.1 chr12 + 70190352 70209330 70190414 70209226 670190352,70170190423,701RAB3IP 1120ENST00000550437.1 chr12 + 70195448 70249143 70195448 70219110 570195448,70270195501,702AC025263.3 17 ENST00000552032.1 chr12 + 70219083 70352503 70284895 70352311 25 70219083,70270219343,702C12orf28 1121ENST00000299350.4 chr12 + 70320436 70352503 70326367 70352311 1270320436,70370320514,703C12orf28 1121ENST00000535034.1 chr12 + 70326315 70352387 70326367 70352311 970326315,70370326378,703C12orf28 1121ENST00000547547.1 chr12 - 70340322 70340861 70340861 70340861 270340322,70370340575,703RP11-611E13.3 1123ENST00000552324.1 chr12 + 70574117 70595784 70595784 70595784 370574117,705 70574318,705RP11-320P7.2 1123ENST00000552998.1 chr12 - 70612911 70615642 70615642 70615642 270612911,70670613377,706RP11-320P7.1 1123ENST00000549651.1 chr12 - 70636085 70637140 70637140 70637140 270636085,70670636673,706RP11-611E13.2 140 ENST00000229195.3 chr12 + 70636776 70748773 70672006 70747695 16 70636776,70670637260,706CNOT2 140 ENST00000418359.3 chr12 + 70636809 70748773 70672006 70747695 17 70636809,70670637017,706CNOT2 1124ENST00000548230.1 chr12 + 70721286 70729246 70729246 70729246 570721286,70770721495,707CNOT2 1124ENST00000551483.1 chr12 + 70728214 70747717 70732471 70747695 770728214,70770732343,707CNOT2 140 ENST00000258111.4 chr12 + 70760055 70828072 70760514 70824433 3 70760055,70770760850,707KCNMB4 1125ENST00000410473.1 chr12 - 70837563 70837703 70837703 70837703 170837563, 70837703, U4 140 ENST00000549460.1 chr12 + 70861859 70931840 70931840 70931840 6 70861859,70870862107,708RP11-588H23.3 140 ENST00000548687.1 chr12 + 70861864 70932859 70932859 70932859 9 70861864,70970862107,709RP11-588H23.3 1125ENST00000548924.1 chr12 + 70861889 70905002 70905002 70905002 570861889,70970862107,709RP11-588H23.3 140 ENST00000549616.1 chr12 + 70861902 70914619 70914619 70914619 6 70861902,70970862107,709RP11-588H23.3 140 ENST00000549359.1 chr12 + 70861902 70921529 70921529 70921529 6 70861902,70970862107,709RP11-588H23.3 140 ENST00000551438.1 chr12 + 70861974 70931985 70931985 70931985 6 70861974,70970862107,709RP11-588H23.3 1126ENST00000451516.2 chr12 - 70910629 71003594 70915268 71003594 3170910629,70970915291,709PTPRB 1126ENST00000334414.6 chr12 - 70910629 71031220 70915268 71031175 3470910629,70970915291,709PTPRB 1126ENST00000547656.1 chr12 + 70913970 70932279 70932279 70932279 270913970,70970914047,709RP11-588H23.3 1126ENST00000546836.1 chr12 + 70913985 70932443 70932443 70932443 370913985,70970914047,709RP11-588H23.3 1126ENST00000550358.1 chr12 - 70915095 71031201 70915268 71031175 3370915095,70970915291,709PTPRB 1126ENST00000544694.1 chr12 - 70915096 71031201 70965684 71031175 3470915096,70970915291,709PTPRB 1126ENST00000538708.1 chr12 - 70915182 71003623 70915268 71003594 3170915182,70970915291,709PTPRB 1126ENST00000550857.1 chr12 - 70915182 71003623 70915268 71003594 3170915182,70970915291,709PTPRB 1126ENST00000261266.5 chr12 - 70915182 71003624 70915268 71003594 3270915182,70970915291,709PTPRB 1126ENST00000551525.1 chr12 - 70952567 71031200 70953081 71031175 1870952567,70970953404,709PTPRB 1126ENST00000538174.2 chr12 - 70978744 71031194 71031194 71031194 1070978744,70970979075,709PTPRB 140 ENST00000440835.2 chr12 - 71031852 71148441 71032963 71147973 10 71031852,71071033057,710PTPRR 140 ENST00000537619.2 chr12 - 71031858 71058457 71058457 71058457 5 71031858,71071033057,710PTPRR 140 ENST00000378778.1 chr12 - 71031861 71148373 71032963 71148373 11 71031861,71071033057,710PTPRR 17 ENST00000283228.2 chr12 - 71031861 71314623 71032963 71314170 14 71031861,71071033057,710PTPRR 140 ENST00000342084.4 chr12 - 71032710 71182762 71032963 71182616 13 71032710,71071033057,710PTPRR 140 ENST00000549308.1 chr12 - 71032839 71148496 71032963 71147973 11 71032839,71071033057,710PTPRR

Supplemental Table 2. List of genes contained in genomic region 12q15 (chr12: 69692322-

71120515) that are recurrently amplified in high-grade serous ovarian cancers.

44 CAL120 COV644 HCC1143 COV362 EFO21 CAOV3 1

Relative FRS2 mRNA level Relative FRS2 mRNA 0 shLacZ shLacZ shLacZ shLacZ shLacZ shLacZ shFRS2#2 shFRS2#1 shFRS2#2 shFRS2#1 shFRS2#2 shFRS2#1 shFRS2#2 shFRS2#1 shFRS2#2 shFRS2#1 shFRS2#2 shFRS2#1

Supplemental Figure 1. Quantitative RT-PCR of FRS2 expression in FRS2 amplified (red) and non-amplified (black) cell lines (Eejung Kim).

HA1E-M HA1E-A LacZ FRS2 Myr-AKT LacZ FRS2 MEK-DD

FRS2

pAKT (S473)

AKT

MEK

Actin

Supplemental Figure 2. Immunoblot confirms the overexpression of FRS2, Myr-AKT,

MEK-DD in HA1E-M and HA1E-A cells.

45 References

1. Bowtell DD. The genesis and evolution of high-grade serous ovarian cancer. Nat Rev Cancer. 2010;10(11):803-8. PubMed ID: 20944665 2. Siegel R, Ma J, Zou Z, Jemal A. Cancer statistics, 2014. CA Cancer J Clin. 2014;64(1):9- 29. PubMed ID: 24399786 3. Karst AM, Drapkin R. The new face of ovarian cancer modeling: better prospects for detection and treatment. F1000 Med Rep. 2011;3:22. PubMed ID: 22076125 4. Miller DS, Blessing JA, Krasner CN, Mannel RS, Hanjani P, Pearl ML, et al. Phase II evaluation of pemetrexed in the treatment of recurrent or persistent platinum-resistant ovarian or primary peritoneal carcinoma: a study of the Gynecologic Oncology Group. J Clin Oncol. 2009;27(16):2686-91. PubMed ID: 19332726 5. Pal T, Permuth-Wey J, Betts JA, Krischer JP, Fiorica J, Arango H, et al. BRCA1 and BRCA2 mutations account for a large proportion of ovarian carcinoma cases. Cancer. 2005;104(12):2807-16. PubMed ID: 16284991 6. Ahmed AA, Etemadmoghadam D, Temple J, Lynch AG, Riad M, Sharma R, et al. Driver mutations in TP53 are ubiquitous in high grade serous carcinoma of the ovary. J Pathol. 2010;221(1):49-56. PubMed ID: 20229506 7. Cancer Genome Atlas Research N. Integrated genomic analyses of ovarian carcinoma. Nature. 2011;474(7353):609-15. PubMed ID: 21720365 8. Cancer Genome Atlas Research N. Comprehensive genomic characterization defines human glioblastoma genes and core pathways. Nature. 2008;455(7216):1061-8. PubMed ID: 18772890 9. Cancer Genome Atlas N. Comprehensive molecular characterization of human colon and rectal cancer. Nature. 2012;487(7407):330-7. PubMed ID: 22810696 10. Cancer Genome Atlas N. Comprehensive molecular portraits of human breast tumours. Nature. 2012;490(7418):61-70. PubMed ID: 23000897 11. Cancer Genome Atlas Research N. Comprehensive genomic characterization of squamous cell lung cancers. Nature. 2012;489(7417):519-25. PubMed ID: 22960745 12. Cancer Genome Atlas Research N. Comprehensive molecular profiling of lung adenocarcinoma. Nature. 2014;511(7511):543-50. PubMed ID: 25079552 13. Bell D, Berchuck A, Birrer M, Chien J, Cramer DW, Dao F, et al. Integrated genomic analyses of ovarian carcinoma. Nature. 2011;474(7353):609-15. PubMed ID: WOS:000292204300032 14. Ho ES, Lai CR, Hsieh YT, Chen JT, Lin AJ, Hung MH, et al. p53 mutation is infrequent in clear cell carcinoma of the ovary. Gynecol Oncol. 2001;80(2):189-93. PubMed ID: 11161858 15. Wiegand KC, Shah SP, Al-Agha OM, Zhao Y, Tse K, Zeng T, et al. ARID1A mutations in endometriosis-associated ovarian carcinomas. N Engl J Med. 2010;363(16):1532-43. PubMed ID: 20942669 16. Kuo KT, Mao TL, Jones S, Veras E, Ayhan A, Wang TL, et al. Frequent activating mutations of PIK3CA in ovarian clear cell carcinoma. Am J Pathol. 2009;174(5):1597-601. PubMed ID: 19349352 17. Cuatrecasas M, Villanueva A, Matias-Guiu X, Prat J. K-ras mutations in mucinous ovarian tumors: a clinicopathologic and molecular study of 95 cases. Cancer. 1997;79(8):1581-6. PubMed ID: 9118042

46 18. Cheung HW, Cowley GS, Weir BA, Boehm JS, Rusin S, Scott JA, et al. Systematic investigation of genetic vulnerabilities across cancer cell lines reveals lineage-specific dependencies in ovarian cancer. Proceedings of the National Academy of Sciences of the United States of America. 2011;108(30):12372-7. PubMed ID: WOS:000293129900039 19. Luo B, Cheung HW, Subramanian A, Sharifnia T, Okamoto M, Yang XP, et al. Highly parallel identification of essential genes in cancer cells. Proceedings of the National Academy of Sciences of the United States of America. 2008;105(51):20380-5. PubMed ID: WOS:000261995600066 20. Cheung HW, Cowley GS, Weir BA, Boehm JS, Rusin S, Scott JA, et al. Systematic investigation of genetic vulnerabilities across cancer cell lines reveals lineage-specific dependencies in ovarian cancer. Proc Natl Acad Sci U S A. 2011;108(30):12372-7. PubMed ID: 21746896 21. Boehm JS, Hahn WC. Towards systematic functional characterization of cancer genomes. Nat Rev Genet. 2011;12(7):487-98. PubMed ID: 21681210 22. Lamesch P, Li N, Milstein S, Fan C, Hao T, Szabo G, et al. hORFeome v3.1: a resource of human open reading frames representing over 10,000 human genes. Genomics. 2007;89(3):307-15. PubMed ID: 17207965 23. Yang X, Boehm JS, Yang X, Salehi-Ashtiani K, Hao T, Shen Y, et al. A public genome- scale lentiviral expression library of human ORFs. Nat Methods. 2011;8(8):659-61. PubMed ID: 21706014 24. Park J, Hu Y, Murthy TV, Vannberg F, Shen B, Rolfs A, et al. Building a human kinase gene repository: bioinformatics, molecular cloning, and functional validation. Proc Natl Acad Sci U S A. 2005;102(23):8114-9. PubMed ID: 15928075 25. Shao DD, Xue W, Krall EB, Bhutkar A, Piccioni F, Wang X, et al. KRAS and YAP1 converge to regulate EMT and tumor survival. Cell. 2014;158(1):171-84. PubMed ID: 24954536 26. Johannessen CM, Johnson LA, Piccioni F, Townes A, Frederick DT, Donahue MK, et al. A melanocyte lineage program confers resistance to MAP kinase pathway inhibition. Nature. 2013;504(7478):138-42. PubMed ID: 24185007 27. Hahn WC, Counter CM, Lundberg AS, Beijersbergen RL, Brooks MW, Weinberg RA. Creation of human tumour cells with defined genetic elements. Nature. 1999;400(6743):464-8. PubMed ID: WOS:000081715000054 28. Liu J, Yang G, Thompson-Lanza JA, Glassman A, Hayes K, Patterson A, et al. A genetically defined model for human ovarian cancer. Cancer Res. 2004;64(5):1655-63. PubMed ID: 14996724 29. Boehm JS, Zhao JJ, Yao J, Kim SY, Firestein R, Dunn IF, et al. Integrative genomic approaches identify IKBKE as a breast cancer oncogene. Cell. 2007;129(6):1065-79. PubMed ID: WOS:000247390400012 30. Gao J, Aksoy BA, Dogrusoz U, Dresdner G, Gross B, Sumer SO, et al. Integrative analysis of complex cancer genomics and clinical profiles using the cBioPortal. Sci Signal. 2013;6(269):pl1. PubMed ID: 23550210 31. Cerami E, Gao J, Dogrusoz U, Gross BE, Sumer SO, Aksoy BA, et al. The cBio cancer genomics portal: an open platform for exploring multidimensional cancer genomics data. Cancer Discov. 2012;2(5):401-4. PubMed ID: 22588877 32. Liu JS, Yang G, Thompson-Lanza JA, Glassman A, Hayes K, Patterson A, et al. A genetically defined model for human ovarian cancer. Cancer Research. 2004;64(5):1655-63. PubMed ID: WOS:000189357800016

47 33. Carpenter AE, Jones TR, Lamprecht MR, Clarke C, Kang IH, Friman O, et al. CellProfiler: image analysis software for identifying and quantifying cell phenotypes. Genome Biol. 2006;7(10):R100. PubMed ID: 17076895 34. Domcke S, Sinha R, Levine DA, Sander C, Schultz N. Evaluating cell lines as tumour models by comparison of genomic profiles. Nat Commun. 2013;4:2126. PubMed ID: 23839242 35. Anglesio MS, Wiegand KC, Melnyk N, Chow C, Salamanca C, Prentice LM, et al. Type- specific cell line models for type-specific ovarian cancer research. PLoS One. 2013;8(9):e72162. PubMed ID: 24023729 36. Etemadmoghadam D, Weir BA, Au-Yeung G, Alsop K, Mitchell G, George J, et al. Synthetic lethality between CCNE1 amplification and loss of BRCA1. Proc Natl Acad Sci U S A. 2013;110(48):19489-94. PubMed ID: 24218601 37. Barretina J, Caponigro G, Stransky N, Venkatesan K, Margolin AA, Kim S, et al. The Cancer Cell Line Encyclopedia enables predictive modelling of anticancer drug sensitivity. Nature. 2012;483(7391):603-7. PubMed ID: 22460905 38. Ren Y, Cheung HW, von Maltzhan G, Agrawal A, Cowley GS, Weir BA, et al. Targeted tumor-penetrating siRNA nanocomplexes for credentialing the ovarian cancer oncogene ID4. Sci Transl Med. 2012;4(147):147ra12. Epub 2012/08/17. PubMed ID: 22896676 39. Karst AM, Levanon K, Drapkin R. Modeling high-grade serous ovarian carcinogenesis from the fallopian tube. Proc Natl Acad Sci U S A. 2011;108(18):7547-52. PubMed ID: 21502498 40. Levanon K, Ng V, Piao HY, Zhang Y, Chang MC, Roh MH, et al. Primary ex vivo cultures of human fallopian tube epithelium as a model for serous ovarian carcinogenesis. Oncogene. 2010;29(8):1103-13. PubMed ID: 19935705 41. Flesken-Nikitin A, Hwang CI, Cheng CY, Michurina TV, Enikolopov G, Nikitin AY. Ovarian surface epithelium at the junction area contains a cancer-prone stem cell niche. Nature. 2013;495(7440):241-5. PubMed ID: 23467088 42. Dunn GP, Cheung HW, Agarwalla PK, Thomas S, Zektser Y, Karst AM, et al. In vivo multiplexed interrogation of amplified genes identifies GAB2 as an ovarian cancer oncogene. Proc Natl Acad Sci U S A. 2014. Epub 2014/01/05. PubMed ID: 24385586 43. Hadari YR, Gotoh N, Kouhara H, Lax I, Schlessinger J. Critical role for the docking- protein FRS2 alpha in FGF receptor-mediated signal transduction pathways. Proceedings of the National Academy of Sciences of the United States of America. 2001;98(15):8578-83. PubMed ID: WOS:000169967000068 44. Ong SH, Hadari YR, Gotoh N, Guy GR, Schlessinger J, Lax I. Stimulation of phosphatidylinositol 3-kinase by fibroblast growth factor receptors is mediated by coordinated recruitment of multiple docking proteins. Proceedings of the National Academy of Sciences of the United States of America. 2001;98(11):6074-9. PubMed ID: WOS:000168883700027 45. Cappellen D, De Oliveira C, Ricol D, de Medina SGD, Bourdin J, Sastre-Garau X, et al. Frequent activating mutations of FGFR3 in human bladder and cervix carcinomas. Nature Genetics. 1999;23(1):18-20. PubMed ID: WOS:000082337300009 46. Kunii K, Davis L, Gorenstein J, Hatch H, Yashiro M, Di Bacco A, et al. FGFR2- amplified gastric cancer cell lines require FGFR2 and Erbb3 signaling for growth and survival. Cancer Research. 2008;68(7):2340-8. PubMed ID: WOS:000254738500036 47. Dutt A, Salvesen HB, Chent TH, Ramos AH, Onofrio RC, Hatton C, et al. Drug-sensitive FGFR2 mutations in endometrial carcinoma. Proceedings of the National Academy of Sciences of the United States of America. 2008;105(25):8713-7. PubMed ID: WOS:000257185700042

48 48. Weiss J, Sos L, Seidel D, Peifer M, Zander T, Heuckmann JM, et al. Frequent and Focal FGFR1 Amplification Associates with Therapeutically Tractable FGFR1 Dependency in Squamous Cell Lung Cancer (vol 3, 66er5, 2011). Science Translational Medicine. 2010;2(62). PubMed ID: WOS:000288444300005 49. Dutt A, Ramos AH, Hammerman PS, Mermel C, Cho J, Sharifnia T, et al. Inhibitor- Sensitive FGFR1 Amplification in Human Non-Small Cell Lung Cancer. Plos One. 2011;6(6). PubMed ID: WOS:000291607000005 50. Easton DF, Pooley KA, Dunning AM, Pharoah PDP, Thompson D, Ballinger DG, et al. Genome-wide association study identifies novel breast cancer susceptibility loci. Nature. 2007;447(7148):1087-U7. PubMed ID: WOS:000247564600032 51. Wiedemann M, Trueb B. Characterization of a novel protein (FGFRL1) from human cartilage related to FGF receptors. Genomics. 2000;69(2):275-9. PubMed ID: 11031111 52. Gotoh N. Regulation of growth factor signaling by FRS2 family docking/scaffold adaptor proteins. Cancer Sci. 2008;99(7):1319-25. PubMed ID: 18452557 53. Sawey ET, Chanrion M, Cai CL, Wu GM, Zhang JP, Zender L, et al. Identification of a Therapeutic Strategy Targeting Amplified FGF19 in Liver Cancer by Oncogenomic Screening. Cancer Cell. 2011;19(3):347-58. PubMed ID: WOS:000288642500008 54. Harding TC, Long L, Palencia S, Zhang HB, Sadra A, Hestir K, et al. Blockade of Nonhormonal Fibroblast Growth Factors by FP-1039 Inhibits Growth of Multiple Types of Cancer. Science Translational Medicine. 2013;5(178). PubMed ID: WOS:000316795300003 55. Harmer NJ, Ilag LL, Mulloy B, Pellegrini L, Robinson CV, Blundell TL. Towards a resolution of the stoichiometry of the fibroblast growth factor (FGF)-FGF receptor-heparin complex. J Mol Biol. 2004;339(4):821-34. PubMed ID: 15165853 56. Mohammadi M, Olsen SK, Ibrahimi OA. Structural basis for fibroblast growth factor receptor activation. Cytokine Growth Factor Rev. 2005;16(2):107-37. PubMed ID: 15863029 57. Ornitz DM, Xu J, Colvin JS, McEwen DG, MacArthur CA, Coulier F, et al. Receptor specificity of the fibroblast growth factor family. J Biol Chem. 1996;271(25):15292-7. PubMed ID: 8663044 58. Zhang X, Ibrahimi OA, Olsen SK, Umemori H, Mohammadi M, Ornitz DM. Receptor specificity of the fibroblast growth factor family. The complete mammalian FGF family. J Biol Chem. 2006;281(23):15694-700. PubMed ID: 16597617 59. Wang Y, Becker D. Antisense targeting of basic fibroblast growth factor and fibroblast growth factor receptor-1 in human melanomas blocks intratumoral angiogenesis and tumor growth. Nat Med. 1997;3(8):887-93. PubMed ID: 9256280 60. Birrer MJ, Johnson ME, Hao K, Wong KK, Park DC, Bell A, et al. Whole genome oligonucleotide-based array comparative genomic hybridization analysis identified fibroblast growth factor 1 as a prognostic marker for advanced-stage serous ovarian adenocarcinomas. J Clin Oncol. 2007;25(16):2281-7. PubMed ID: 17538174 61. Lowenstein EJ, Daly RJ, Batzer AG, Li W, Margolis B, Lammers R, et al. The Sh2 and Sh3 Domain Containing Protein Grb2 Links Receptor Tyrosine Kinases to Ras Signaling. Cell. 1992;70(3):431-42. PubMed ID: WOS:A1992JH12400008 62. Warner N, Nunez G. MyD88: A Critical Adaptor Protein in Innate Immunity Signal Transduction. Journal of Immunology. 2013;190(1):3-4. PubMed ID: WOS:000312832500001 63. Kouhara H, Hadari YR, SpivakKroizman T, Schilling J, BarSagi D, Lax I, et al. A lipid- anchored Grb2-binding protein that links FGF-receptor activation to the Ras/MAPK signaling pathway. Cell. 1997;89(5):693-702. PubMed ID: WOS:A1997XB92500006

49 64. Cheung HW, Du JY, Boehm JS, He F, Weir BA, Wang XX, et al. Amplification of CRKL Induces Transformation and Epidermal Growth Factor Receptor Inhibitor Resistance in Human Non-Small Cell Lung Cancers. Cancer Discovery. 2011;1(7):608-25. PubMed ID: WOS:000299412300032 65. Dunn GP, Cheung HW, Agarwalla PK, Thomas S, Zektser Y, Karst AM, et al. In vivo multiplexed interrogation of amplified genes identifies GAB2 as an ovarian cancer oncogene. Proc Natl Acad Sci U S A. 2014;111(3):1102-7. PubMed ID: 24385586 66. Chen Y, McGee J, Chen X, Doman TN, Gong X, Zhang Y, et al. Identification of Druggable Cancer Driver Genes Amplified across TCGA Datasets. PLoS One. 2014;9(5):e98293. PubMed ID: 24874471 67. Zhang KQ, Chu KV, Wu XW, Gao HL, Wang JH, Yuan YC, et al. Amplification of FRS2 and Activation of FGFR/FRS2 Signaling Pathway in High-Grade Liposarcoma. Cancer Research. 2013;74(4):1298-307. PubMed ID: WOS:000315029800005 68. Wang XK, Asmann YW, Erickson-Johnson MR, Oliveira JL, Zhang HY, Moura RD, et al. High-Resolution Genomic Mapping Reveals Consistent Amplification of the Fibroblast Growth Factor Receptor Substrate 2 Gene in Well-Differentiated and Dedifferentiated Liposarcoma. Genes Chromosomes & Cancer. 2011;50(11):849-58. PubMed ID: WOS:000295208800001 69. Zhang K, Chu K, Wu X, Gao H, Wang J, Yuan YC, et al. Amplification of FRS2 and activation of FGFR/FRS2 signaling pathway in high-grade liposarcoma. Cancer Res. 2013;73(4):1298-307. PubMed ID: 23393200 70. Fischer U, Keller A, Leidinger P, Deutscher S, Heisel S, Urbschat S, et al. A different view on DNA amplifications indicates frequent, highly complex, and stable amplicons on 12q13- 21 in glioma. Mol Cancer Res. 2008;6(4):576-84. PubMed ID: 18403636 71. Hagerstrand D, Tong A, Schumacher SE, Ilic N, Shen RR, Cheung HW, et al. Systematic interrogation of 3q26 identifies TLOC1 and SKIL as cancer drivers. Cancer Discov. 2013;3(9):1044-57. PubMed ID: 23764425 72. Brooks AN, Kilgour E, Smith PD. Molecular pathways: fibroblast growth factor signaling: a new therapeutic opportunity in cancer. Clin Cancer Res. 2012;18(7):1855-62. PubMed ID: 22388515 73. du Bois A, Floquet A, Kim JW, Rau J, del Campo JM, Friedlander M, et al. Incorporation of pazopanib in maintenance therapy of ovarian cancer. J Clin Oncol. 2014;32(30):3374-82. PubMed ID: 25225436 74. Andre F, Bachelot T, Campone M, Dalenc F, Perez-Garcia JM, Hurvitz SA, et al. Targeting FGFR with dovitinib (TKI258): preclinical and clinical data in breast cancer. Clin Cancer Res. 2013;19(13):3693-702. PubMed ID: 23658459 75. Martinez-Torrecuadrada J, Cifuentes G, Lopez-Serra P, Saenz P, Martinez A, Casal JI. Targeting the extracellular domain of fibroblast growth factor receptor 3 with human single- chain Fv antibodies inhibits bladder carcinoma cell line proliferation. Clin Cancer Res. 2005;11(17):6280-90. PubMed ID: 16144932 76. Qing J, Du X, Chen Y, Chan P, Li H, Wu P, et al. Antibody-based targeting of FGFR3 in bladder carcinoma and t(4;14)-positive multiple myeloma in mice. J Clin Invest. 2009;119(5):1216-29. PubMed ID: 19381019 77. Wells JA, McClendon CL. Reaching for high-hanging fruit in drug discovery at protein- protein interfaces. Nature. 2007;450(7172):1001-9. PubMed ID: 18075579

50 78. Sato T, Gotoh N. The FRS2 family of docking/scaffolding adaptor proteins as therapeutic targets of cancer treatment. Expert Opin Ther Targets. 2009;13(6):689-700. PubMed ID: 19456272 79. Clackson T, Wells JA. A hot spot of binding energy in a hormone-receptor interface. Science. 1995;267(5196):383-6. PubMed ID: 7529940 80. Muller YA, Li B, Christinger HW, Wells JA, Cunningham BC, de Vos AM. Vascular endothelial growth factor: crystal structure and functional mapping of the kinase domain receptor binding site. Proc Natl Acad Sci U S A. 1997;94(14):7192-7. PubMed ID: 9207067 81. Moreira IS, Fernandes PA, Ramos MJ. Hot spots--a review of the protein-protein interface determinant amino-acid residues. Proteins. 2007;68(4):803-12. PubMed ID: 17546660 82. Thanos CD, DeLano WL, Wells JA. Hot-spot mimicry of a cytokine receptor by a small molecule. Proc Natl Acad Sci U S A. 2006;103(42):15422-7. PubMed ID: 17032757 83. Oltersdorf T, Elmore SW, Shoemaker AR, Armstrong RC, Augeri DJ, Belli BA, et al. An inhibitor of Bcl-2 family proteins induces regression of solid tumours. Nature. 2005;435(7042):677-81. PubMed ID: 15902208 84. Luo B, Cheung HW, Subramanian A, Sharifnia T, Okamoto M, Yang X, et al. Highly parallel identification of essential genes in cancer cells. Proc Natl Acad Sci U S A. 2008;105(51):20380-5. PubMed ID: 19091943 85. Luo LY, Kim E, Cheung HW, Weir BA, Dunn GP, Shen RR, et al. The tyrosine kinase adaptor protein FRS2 is oncogenic and amplified in high-grade serous ovarian cancer. Mol Cancer Res. 2014. PubMed ID: 25368431

51