Understanding the Role of Group I Paks in Thyroid Cancer DISSERTATION Presented in Partial Fulfillment of the Requirements for T

Understanding the Role of Group I PAKs in Thyroid Cancer

DISSERTATION

Presented in Partial Fulfillment of the Requirements for the Degree Doctor of Philosophy in the Graduate School of The Ohio State University

Christina Michelle Knippler, BA

Biomedical Sciences Graduate Program

The Ohio State University

2019

Dissertation Committee:

Matthew Ringel, MD, Advisor

David Carbone, MD, PhD

Joanna Groden, PhD

Michael Ostrowski, PhD

Mark Parthun, PhD

Copyrighted by

Christina Michelle Knippler

2019

Abstract

Thyroid cancer incidence has been increasing over the last several decades. Most thyroid cancers are curable, however, aggressive tumors do not respond to standard therapy and have only limited responses to recently-approved targeted therapies, especially when metastatic cancer is present. Many “driver” mutations of thyroid cancer have been determined, but it is becoming increasingly clear that these oncogenic pathways do not act as singular entities to “drive” malignancy. It is, therefore, pivotal to understand the complexity of thyroid cancer signaling in order to determine the best therapeutic approach. The MAPK pathway is overactivated in the majority of thyroid cancers, with the most common mutation causing the valine to glutamic acid point mutation at amino acid 600 in the kinase BRAF (i.e. BRAFV600E). Members of the MAPK pathway, and in particular BRAFV600E, are therapeutic targets of high interest, with ongoing clinical trials using pathway inhibitors. Preliminarily, these drugs result in only partial responses and resistance. Research is ongoing to determine the mechanisms of innate and acquired resistance to these drugs.

Our prior research has identified a family of serine/threonine kinases, the group I p21-activated kinases (PAKs), as a novel indicator of aggressive characteristics in thyroid cancer. PAK1 protein levels and activation are highly expressed in the invasive edges of thyroid cancers, compared to the center of the tumor or normal thyroid tissue. BRAF and

PAK1 signaling are interconnected, both necessary for thyroid cancer migration, with ii

PAK activation in vitro and in vivo regulated by BRAF expression. Further, BRAF and

PAK1 physically interact. Therefore, the connection between BRAF and PAK1 are significant for thyroid cancer progression. The objective of the current work was to characterize the BRAF-PAK1 interaction in order to develop novel therapeutic approaches.

This research determined the ability of PAK1 to be pharmacologically inhibited and tested the hypothesis that group I PAK inhibitors would reduce thyroid cancer growth and motility – two important hallmarks of cancer progression. Biochemical approaches tested how PAK1 physically interacts with BRAF, and its constitutively active mutated form, BRAFV600E. We hypothesized that PAK1 and BRAF physically interact as a complex with other mediating proteins and that these proteins provide additional targets for better therapeutic efficacy. This work utilized both in vitro and in vivo models to provide a comprehensive understanding of the BRAF and PAK1 complex.

In vitro studies using several different thyroid cancer cell lines determined that group I

PAKs could be pharmacologically inhibited and this reduced thyroid cancer cell growth, induced cell cycle arrest, and reduced cell invasion. Further, PAK inhibition could be combined with BRAFV600E and AKT inhibition for further synergistic reductions on thyroid cancer cell growth, useful for future clinical application. Importantly, group I

PAK inhibition was effective in vivo in a mouse model of thyroid cancer where

BRAFV600E was overexpressed in the thyroid. PAK inhibition restricted thyroid size and also carcinoma formation.

iii

In vitro biochemical approaches determined that both wild-type BRAF and

BRAFV600E can physically interact with PAK1 in precise cell contexts, most often during mitosis. The interaction is unlikely to be direct, but may be facilitated through part of the

PAK1 regulatory domain and does not require PAK1 kinase function. Proteomic evidence suggests that this interaction is mediated by chaperone proteins, allowing BRAF and PAK1 to interact transiently.

These studies expand our understanding of the signaling crosstalk between BRAF and PAK1 and identify the group I PAKs as therapeutically targetable kinases in thyroid cancer. This provides further evidence to continue efforts in drug development targeting the group I PAKs, as well as potential combination strategies with BRAF, AKT, and possibly HSP90 inhibitors, for patients with aggressive thyroid cancers.

Dedication

To my parents, sister, and fiancé for their constant love, support, and encouragement.

To Grammy for fostering my love of science.

Acknowledgments

I am truly appreciative of everyone who has helped and supported me along this journey. My mentor, Dr. Matthew Ringel, has been a cornerstone in my development as a young scientist. He taught me how to think critically and creatively about my science. I am thankful for our long meetings discussing results and developing new hypotheses, and his sincere support in my personal growth. His persistent encouragement, no matter the result of an experiment, helped motivate me to push through the challenges of research and never lose sight of the overall goal and the discoveries we make along the way. I am very thankful for all I have learned from Dr. Ringel and will use my experiences from him to mentor young scientists in the future.

I would also like to especially thank Dr. Motoyasu Saji from the Ringel Lab for his selfless guidance and expertise each day in the lab. My knowledge and skill of experimental techniques is largely due to his patience and passion for teaching. His constant care for the lab, organization, and management kept this lab running on all cylinders so that I could pursue my research goals.

The Ringel Lab would not function smoothly without the endless help and work from our administrator, Nanci Edgington. Often behind the scenes, Nanci was the glue that kept this lab together, and I am immensely grateful for her time and effort in helping me with ordering, setting up meetings, troubleshooting problems, and much more.

Additionally, I am thankful for opportunities to get feedback and inspiration from other members of the Ringel Lab, past and present, including Dr. Anisley Valenciaga, Dr.

Chaojie Wang, Dr. Steven Justiniano, Dr. Anisha Hammer, Dr. Adlina Mohd Yusof, Dr.

Kara Rossfeld, Neel Rajan, Ceimoani Bumrah, and Luis Bautista.

Thank you to my committee members, Drs. David Carbone, Joanna Groden,

Michael Ostrowski, and Mark Parthun, for their guidance and advice throughout the my graduate studies to help me advance my research and to critically analyze my work.

This research would not be possible without the expertise from our collaborators.

I’d like to thank Kyle Porter from the Department of Biomedical Informatics for his data analyses. Dr. Krista La Perle from the Department of Veterinary Biosciences was instrumental in analyzing the mouse pathology samples. I am grateful for Dr. Michael

Freitas (OSU) and Dr. Salim Merali (Temple, Philadelphia, PA) for their proteomics expertise and to Dr. Mark Parthun for helping me with the chromatography experiments.

Additionally, I am appreciative of the expertise of many of the OSU Shared Resources.

I am very appreciative of the Pelotonia Fellowship Program for not only funding my research, but also giving me the opportunities to share my passion and research with the community.

Finally, I am grateful for the support and mentorship from the Biomedical

Sciences Graduate Program, in particular the past and current program directors, Drs.

Joanna Groden, Jeffrey Parvin, and Michael Freitas.

vii

Vita

2009...... Chantilly High School, VA

2013...... B.A. Biology, B.A. Music, University of

Virginia, VA

2013-Present ...... Biomedical Sciences Graduate Program,

The Ohio State University, OH

2013-2014...... University Fellowship, The Ohio State

University, OH

2014-2016...... T32 Systems in Integrated Biology

Fellowship, The Ohio State University, OH

2016-2018...... Pelotonia Graduate Fellowship, The Ohio

State University, OH

2018-Present ...... Graduate Research Associate, The Ohio

State University, OH

Publications

Knippler CM, Saji M, Rajan N, Porter K, La Perle KMD, Ringel MD. MAPK- and

AKT-activated thyroid cancers are sensitive to group I PAK inhibition. Endocr Relat

Cancer. 2019 May 1; [Epub ahead of print]

viii

Valenciaga A, Saji M, Yu L, Zhang X, Bumrah C, Yilmaz AS, Knippler CM, Miles W,

Giordano TJ, Cote GJ, Ringel MD. Transcriptional targeting of oncogene addiction in medullary thyroid cancer. JCI Insight. 2018 Aug 23; 3(16)

Stellfox ME, Nardi IK, Knippler CM, Foltz DR. Differential binding partners of the

Mis18α/β YIPPEE domains regulate Mis18 complex recruitment to centromeres. Cell

Rep. 2016 Jun 7; 15(10): 2127-2135

Nardi IK, Zasadzińska E, Stellfox ME, Knippler CM, Foltz DR. Licensing of centromeric chromatin assembly through the Mis18α-Mis18β heterotetramer. Mol. Cell.

2016 Mar 3; 61(5): 774-787

McCarty SK, Saji M, Zhang X, Knippler CM, Kirschner LS, Fernandez S, Ringel MD.

BRAF activates and physically interacts with PAK to regulate cell motility. Endocr Relat

Cancer. 2014 Dec; 21 (6): 865-77

Fields of Study

Major Field: Biomedical Sciences Graduate Program

Emphasis: Cancer Biology

Table of Contents

Abstract ...... ii Dedication ...... v Acknowledgments...... vi Vita ...... viii List of Tables ...... xiii List of Figures ...... xiv Chapter 1: Introduction ...... 1 1.1 Overview of the Thyroid ...... 1 1.2 Thyroid Cancer ...... 2 1.2.1 Histological Classifications and Genetic Characteristics ...... 3 1.2.2 Treatment Options ...... 5 1.3 MAPK Pathway...... 7 1.3.1 Canonical Signaling and Alterations in Cancers ...... 7 1.3.2 Therapeutic BRAF Inhibitors ...... 10 1.4 Molecular Characterization of PTC Invasive Fronts...... 11 1.5 p21-activated Kinases (PAKs) ...... 12 1.5.1 Canonical Signaling and Functions ...... 12 1.5.2 Alterations in Cancers ...... 17 1.5.3 Molecular and Pharmacologic Inhibitors ...... 18 1.6 Non-canonical BRAF-PAK1 Crosstalk ...... 21 1.6.1 Signaling and Cellular Functions ...... 21 1.6.2 Physical Interaction ...... 23 1.7 Conclusions and Dissertation Overview ...... 23 Chapter 2: Group I PAK Inhibition in Thyroid Cancer ...... 25 2.1 Abstract ...... 25 2.2 Introduction ...... 26 2.3 Materials and Methods ...... 29

2.3.1 Cell lines and cell culture ...... 29 2.3.2 In vitro drug reagents ...... 30 2.3.3 Cell viability assays ...... 30 2.3.4 Protein extraction and western blots ...... 31 2.3.5 Cell cycle analysis ...... 33 2.3.6 Migration and invasion assays ...... 33 2.3.7 Thyroid-specific inducible BRAFV600E mouse model ...... 35 2.3.8 G-5555 treatment in vivo ...... 36 2.3.9 Immunohistochemistry (IHC)...... 36 2.3.10 Protein extraction from mouse thyroids ...... 37 2.3.11 Statistical analysis...... 38 2.4 Results ...... 40 2.4.1 Group I PAK inhibition reduces thyroid cancer cell viability ...... 40 2.4.2 G-5555 in combination with BRAFV600E or AKT inhibitors synergistically reduces K1 and SW1736 cancer cell viability ...... 44 2.4.3 Reduction in cell viability by group I PAK inhibition corresponds with cell cycle arrest in G0/G1...... 48 2.4.4 G-5555 reduces thyroid cancer cell migration and invasion ...... 49 2.4.5 BRAFV600E-increased thyroid size and cancer development in vivo is inhibited by G-5555 ...... 52 2.5 Discussion ...... 55 Chapter 3: Characterization of the BRAF-PAK1 Complex ...... 60 3.1 Abstract ...... 60 3.2 Introduction ...... 61 3.3 Materials and Methods ...... 64 3.3.1 Cell lines and cell culture ...... 64 3.3.2 Plasmid constructs ...... 65 3.3.3 In situ Proximity Ligation Assay (PLA) and immunofluorescence (IF) ...... 67 3.3.4 Immunoprecipitation (IP) and western blots ...... 69 3.3.5 Cell synchronization ...... 71 3.3.6 Mass spectrometry and proteomics analysis ...... 71 3.3.7 17-AAG treatment ...... 73 3.3.8 TEV Tandem co-IP ...... 74

3.3.9 Ion exchange and gel filtration/size exclusion chromatography ...... 75 3.3.10 Proximity-dependent biotin identification (BioID2) ...... 76 3.4 Results ...... 77 3.4.1 BRAFV600E and PAK1 physically interact...... 77 3.4.2 PAK1 binding to BRAF is independent of PAK1 kinase activity and the critical region may occur between PAK1 amino acids 121-140...... 81 3.4.3 Identifying BRAF and PAK1 binding partners: CCT Complex ...... 84 3.4.4 Identifying BRAF and PAK1 binding partners: HSP90 ...... 87 3.4.5 Tandem immunoprecipitation to define the BRAF-PAK1 complex ...... 90 3.4.6 The BRAF-PAK1 complex may be localized to microtubules in mitosis ...... 92 3.4.7 The BRAF-PAK1 complex is likely transient ...... 94 3.5 Discussion ...... 100 Chapter 4: Conclusions and Future Directions ...... 106 4.1 Conclusions ...... 106 4.2 Future Directions ...... 109 Appendix A: List of Abbreviations...... 114 Appendix B: Supplementary Figures and Tables ...... 117 Appendix C: Sources of Funding ...... 129 References ...... 130

xii

List of Tables

Table 2.1. Estimated IC50 concentrations (µM) for G-5555 and FRAX1036 in human thyroid cancer cell lines………………………………………………………………..42

Table B.1. Key tumor promoting mutations in human thyroid cancer cell lines…….122

Table B.2. Estimated IC50 concentrations (µM) for PLX4032 and MK2206………. 122

Table B.3. FLAG-PAK1 interacting proteins as identified by M. Freitas (OSU) proteomics………………………………………………………………………….....125

Table B.4. Interacting proteins of both MYC-BRAF and FLAG-PAK1 as analyzed by S. Merali (Temple)……………………………………………………………………....126

xiii

List of Figures

Figure 1.1. PAK family and protein structure.……………………………………….…..13 Figure 1.2. MAPK and PAK signaling…………………………………………………..22 Figure 2.1. Effects of group I PAK inhibition on cell viability………………………….41 Figure 2.2. PAK signaling with group I PAK inhibition………………………………...43 Figure 2.3. Combination treatment with BRAFV600E and PAK inhibitors in BRAFV600E cell lines………………………………………………………………………………….45 Figure 2.4. Combination treatment with AKT and PAK inhibitors in BRAFV600E cell lines………………………………………………………………………………………47 Figure 2.5. Cell cycle effects with group I PAK inhibition……………………………...49 Figure 2.6. Effects of G-5555 on cell migration and invasion………………………...…51 Figure 2.7. In vivo effects of G-5555 in a BRAFV600E-inducible thyroid cancer mouse model……………………………………………………………………………....……..54 Figure 3.1. PAK1 interactions with WT BRAF and BRAFV600E……………………...…79 Figure 3.2. PAK1 domains for BRAF interaction………………………...... 83 Figure 3.3. PAK1 and BRAF interactions with the CCT complex...... 87 Figure 3.4. PAK1 and BRAF interacts with HSP90…...... 89 Figure 3.5. Tandem-IP with TEV cleavage…...... 92 Figure 3.6. PAK1 and BRAF interacts with tubulin…...... 94 Figure 3.7. Ion exchange and gel filtration of TPC1 protein demonstrates the transient nature of the BRAF-PAK1 complex …...... 96 Figure 3.8. BioID2-PAK1 proximal interactions…...... 99 Figure 4.1. FLAG-tag knock-in cell lines…...... 113

Figure B.1. Cell viability effects of BRAFV600E and AKT inhibitor monotherapy…...... 117 Figure B.2. Combination treatment with BRAFV600E and PAK inhibitors in WT BRAF cell lines…...... 118 Figure B.3. Combination treatment with AKT and PAK inhibitors in WT BRAF cell lines…...... 119 Figure B.4. Ki67 and cleaved caspase-3 in BRAFV600E-induced thyroids…...... 120 Figure B.5. Relationship of pERK levels and thyroid size in BRAFV600E-induced mice…...... 121 Figure B.6. Variable expression of FLAG-PAK1 truncations…...... 123 Figure B.7. CCT expression in thyroid cancer cell lines…...... 123 Figure B.8. CCT-BRAF interactions…...... 124 Figure B.9. Cross-reactivity of PLA probes…...... 124

xiv

Chapter 1: Introduction

1.1 Overview of the Thyroid

The thyroid is an endocrine gland located at the base of the neck that produces thyroid hormone in order to regulate homeostasis and metabolic balance. The majority of the gland is composed of follicular cells which form round follicles surrounding colloid and express thyroid-selective proteins, such as thyroglobulin (Tg), thyroid-stimulating hormone receptor (TSH-R), and the sodium iodide symporter (NIS) that together function to secrete thyroid hormone in a tightly controlled manner (1). A smaller proportion of the thyroid consists of medullary C cells that express the hormone calcitonin that is critical for skeletal development. The follicular lumen contains several proteins, a large proportion of which is Tg, a large protein that is iodinated after dietary iodide is processed to iodine by the follicular cells, serving as the “backbone” for cleavage products that are secreted as thyroid hormones. The thyroid follicular cells are controlled by a feedback loop with the hypothalamus and the pituitary gland, in which the hypothalamus secretes thyrotropin-releasing hormone (TRH) to stimulate the pituitary gland to release thyroid-stimulating hormone (TSH). TSH binds to its receptor on the follicular cells. Upon TSH binding, iodinated Tg is reabsorbed into the follicular cells and is proteolytically cleaved to produce the thyroid hormones triiodothyronine (T3) and thyroxine (T4). T3 and T4 are then secreted into the bloodstream to regulate the body’s homeostasis and act as negative regulators of TRH and TSH. 1

1.2 Thyroid Cancer

Thyroid cancer incidence has been rising the last several decades and is the most common endocrine malignancy (2-6). Detection methods are more sensitive and screening has increased; therefore the increase in incidence is in part due to detection of smaller tumors (i.e. “overdiagnosis”). However, it is also notable that the frequency of larger tumors is also increasing. Importantly, the number of patients who die from thyroid cancer is increasing annually, even in the past few years in which the rate of diagnosis has stabilized or even decreased due to adherence to newer guidelines that recommend monitoring smaller high-risk nodules. In 2018, thyroid cancer was the 12th most common cancer in the U.S. The National Cancer Institute now estimates there will be 52,070 new cases in 2019 and 2,170 deaths in the United States (4). Thyroid cancer is three times more common in women than in men, for reasons currently not well defined. For a cancer more often seen in adults, thyroid cancer is generally detected at a younger age than most other solid tumors, with a median age of diagnosis around 50 years old. Diagnosis of thyroid cancer is usually determined by ultrasound of the neck and pathological analysis from a fine needle aspiration (FNA) sample (7). Gene-expression arrays or next- generation sequencing can be additionally performed to assess common genetic alterations in thyroid cancers that might dictate course of treatment. Risk factors include exposure to ionizing radiation, which accounts for notable increases in incidence observed after the Chernobyl accident in Ukraine in 1986 and possibly the Fukushima disaster in Japan in 2011 (8-10). Interestingly, thyroid cancer incidence was predominantly observed in children after these events, particularly in areas with iodine

2 deficiency, which influences NIS expression and iodine uptake (1). More common exposures to radiation that increase PTC risk when exposure occurs in childhood include medical external radiation in the head and neck area. Other risk factors include particular genetic variants and other cancer-related genetic syndromes, which will be described, in part, in the next section. Although the overall 5-year survival for thyroid cancer in general is over 98%, due in part to the high frequency of small thyroid cancer diagnosis, survival rates vary dramatically depending on the type of thyroid cancer and the stage of disease at which it is diagnosed.

1.2.1 Histological Classifications and Genetic Characteristics

There are five main histological types of thyroid cancer: papillary, follicular, poorly differentiated, anaplastic, and medullary. Papillary (PTC), follicular (FTC), poorly differentiated (PDTC), and anaplastic (ATC) thyroid cancers derive from the thyroid follicular cells and account for the majority of thyroid cancer cases. PTC is the most common type, constituting about 80% of thyroid cancers. Within PTC, there are several subtypes, classified by histology and genetic drivers. Most PTCs have activating mutations in the mitogen-activated protein kinase (MAPK) pathway (11, 12). The most common mutation results in the BRAFV600E mutated protein (60% of PTCs). RAS mutations occur in about 15% of PTCs, with NRAS more common than HRAS or KRAS.

Receptor tyrosine kinase (RTK) genes can be rearranged to create fusion proteins, such as

RET-PTC1 fusions, and activate MAPK signaling, in part through RAS, occurring in approximately 12% of PTCs.

FTCs are less common and occur in only 2-5% of cases. FTCs typically are driven by either the phosphatidylionositol 3-kinase (PI3K) or MAPK pathways, with activating mutations in PIK3CA, loss of function mutations in the negative regulator

PTEN, activating mutations in RAS, or PAX8-PPARγ fusions. PAX8 is a transcription factor that regulates many thyroid-specific genes (13) and PPARγ is also a transcription factor, but regulates adipocyte- and macrophage-related genes (14, 15).

In contrast to the PTC and FTC, which are classified as differentiated thyroid cancers, poorly differentiated thyroid cancers (PDTCs), accounting for 6% of thyroid cancers, are often very aggressive and lack many of the genetic characteristics of differentiated thyroid cells. They usually have mutations in BRAF, RAS, TERT, and

TP53, and are therefore thought to be derived from well-differentiated PTCs or FTCs

(16). Patients typically have a mean survival of only 3.2 years and, therefore, require more aggressive treatment regimens.

Accounting for only 1% of thyroid cancers, ATC is the most rare, yet most aggressive type of thyroid cancer. Patients with ATC have the worst prognosis, with a mean survival of merely 6 months. It is commonly believed that ATCs derive from PTCs or PDTCs that have undergone further genetic alterations to become more aggressive.

Therefore, along with BRAF and RAS mutations typically detected in PTCs, ATCs also accumulate mutations in TP53, the TERT promoter, PI3K pathway, SWI/SNF complex, histone methyltransferases, and mismatch repair genes (11, 16).

Medullary thyroid cancer (MTC) differs from the aforementioned types in that it is derived from the C cells of the thyroid, rather than the follicular cells. MTC is

4 comprises 3-5% of thyroid cancers and is associated with the multiple endocrine neoplasia (MEN) type 2 syndromes and mutations in the RET (rearranged during transfection) receptor tyrosine kinase (11). The work presented in this dissertation focuses on non-medullary thyroid cancers and, thus, will not go into detail in the etiology or treatment of MTC.

1.2.2 Treatment Options

Most thyroid cancers are first treated by surgery – either with a lobectomy, in which only one lobe of the thyroid is removed, or by total thyroidectomy. In some cases, typically with larger tumors, central neck dissection and/or removal of lymph nodes may be recommended. Radioiodine (RAI) therapy, iodine-131, is the primary initial systemic therapy administered to patients, particularly those at higher risk of recurrence, post- operatively to eliminate any remaining functioning thyroid tissue. This tactic works relatively well against differentiated tumors which have retained expression of the sodium-iodine transporter and other related genes, and have not metastasized to distant sites (ex: lungs, bone, and brain); however, many aggressive tumors, especially those with BRAFV600E, lose expression of these thyroid-specific genes and, therefore, have poor iodine uptake and are refractory to radioiodine therapy. These tumors are also more likely to metastasize beyond the neck. Additionally, thyroid-stimulating hormone (TSH) suppression with levothyroxine is recommended to reduce the risk of recurrence by reducing the activity of any remaining thyroid tissue and is necessary to replace the thyroid hormone production that was eliminated by surgery (7).

Aggressive thyroid cancers that are refractory to RAI therapy and/ or have metastasized from the primary site are treated with external beam radiation therapy, chemotherapy, and/ or targeted therapies to increase progression-free survival. Most patients with these tumors only have a 10% overall survival (7, 17). Strategies to increase

RAI uptake have only modest success; thus, alternative treatments have been studied.

Sorafenib and Lenvatinib are the two Food and Drug Administration (FDA)-approved multikinase inhibitors for RAI-refractory DTC. Sorafenib targets mainly vascular endothelial growth factor receptor 1 (VEGFR-1), VEGFR-2, VEGFR-3, RET, platelet- derived growth factor receptor (PDGFR) β, and BRAF (18). A phase 3 trial determined significant prolongation of progression-free survival in patients with RAI-refractory local or metastatic disease from 5.8 months with the placebo to 10.8 months with sorafenib)

(18). Lenvatinib also targets many receptor tyrosine kinases, including VEGFRs 1-3, fibroblast growth factor receptors (FGFRs) 1-4, PDGFRα, RET, and KIT, and was FDA- approved after a phase 3 clinical trial was reported in 2015 (19). Lenvatinib extended progression-free survival of these patients from 3.6 months with the placebo to 18.3 months and is, therefore, generally viewed as the first-line for treatment. Due to their inhibitory effects on VEGFR, it is thought that Sorafenib and Lenvatinib are disrupting tumor vasculature and angiogenesis, as one mode of action for improved patient survival.

As these inhibitors target many kinases, however, it is difficult to be certain of the exact mechanism by which these drugs are acting.

Several dozen single-agent therapies and combinations have been investigated in clinical trials as insights into the oncogenic mechanisms of thyroid cancer emerge

(cancer.gov, NCI, accessed 4/9/19), (20, 21). These include targeted therapies to oncogenic kinases (ex: the MEK inhibitor, Trametinib; RET inhibitors, LOXO-292 and

BLU-667), immune therapies (ex: Pembrolizumab, Nivolumab, and Ipilimumab), and chemotherapies (ex: docetaxel and cyclophosphamide), among others. Targeting

BRAFV600E is appealing due to its prevalence in PTCs and ATCs. This is of particular interest for the work presented here and will be discussed in further detail in the next section.

In summary, although most thyroid cancers are cured or controlled with surgery,

RAI therapy, and TSH suppression, patients with refractory cancers to these standards of care must seek alternative treatment options. Additionally, therapies tested in clinical trials have limited effectiveness and durability, and are associated with major toxicities that impact quality of life. Thus, more effective treatment therapies are needed.

1.3 MAPK Pathway

1.3.1 Canonical Signaling and Alterations in Cancers

The mitogen-activated protein kinase (MAPK) pathway is a key signal transduction pathway regulating many cell functions, including growth and proliferation.

Canonically, the pathway is activated when soluble growth factors or cytokines bind and activate receptor tyrosine kinases (RTKs) embedded in the cell membrane. These RTKs activate the small G protein, RAS, which then recruits, phosphorylates, and activates rapidly accelerated fibrosarcoma (RAF) proteins, such as BRAF, at the membrane. The

MAPK cascade continues as BRAF phosphorylates MAPK/ ERK kinase (MEK) 1 or

MEK2, which then phosphorylates extracellular signal regulated protein kinase (ERK) 1

7 or ERK2, leading to the transcription of many proliferation and cell survival genes (22-

25). Extensive studies in the past 40 years have shed light on the mechanisms and many players involved in signal transduction through the MAPK pathway. There are several isoforms of RAS, RAF, MEK, and ERK, as well as various other upstream and downstream pathways involved in the regulation of MAPK-mediated cell proliferation and inhibition of apoptosis, including the p21-activated kinases (PAKs), c-Jun amino- terminal kinase/ stress-activated protein kinase (JNK/SAPK), and transforming growth factor-β (TGF-β)-activated kinase (TAK) (24, 26). Although often depicted as a straightforward linear pathway, there is increasing evidence to support the complexity of the MAPK pathway, as many nodes can be altered in cancers to drive constitutive activation and uncontrolled cell growth.

As described previously, hyperactivation of the MAPK pathway in thyroid cancer can occur through a number of alterations, including RTK rearrangements, RAS protein mutations, and most commonly the constitutively activating BRAFV600E. Alterations within this pathway occur in many different types of cancer. For example, mutations in the RAS genes are found in approximately 27% of all cancers, most often in pancreatic ductal carcinoma, colon cancers, multiple myeloma, lung cancers, and melanomas (27).

Overall, KRAS mutations (85%) are more common than NRAS (12%) and HRAS (3%)

(28); however, in thyroid cancer, NRAS is observed more often than HRAS or KRAS, and are predominantly in FTC or follicular variant PTC (29, 30). Activated RAS can signal through both the MAPK and PI3K pathways, making it a particularly important, yet difficult, target for therapeutic approaches (27).

The most common mutation in thyroid cancers produces BRAFV600E. The rapidly accelerated fibrosarcoma (RAF) family of oncogenes was first discovered in 1983 (31) and is comprised of three family members, ARAF, BRAF, and CRAF (also referred to as

RAF-1) (32-34). All three RAFs have similar sequence and structure, and contain three conserved regions (CR1, CR2, and CR3) (34, 35). CR1 and CR2 are in the N-terminus regulatory domain and bind RAS and inhibitory 14-3-3 scaffold proteins, respectively.

CR3 is located in the C-terminus kinase domain and can bind activating 14-3-3 proteins.

CRAF, ARAF, and BRAF are regulated slightly differently due to key phosphorylation sites, and BRAF has the highest basal activity of the RAFs due to a constitutively phosphorylated S445 (35, 36). BRAF and CRAF form hetero- and homodimers to facilitate MEK activation, with heterodimers having increased kinase activity than either homodimer (37). It is important to note that this occurs in the wild-type proteins; however, BRAFV600E mutant proteins act constitutively as a monomer (38).

Additionally, BRAF is more commonly mutated in cancers than CRAF or ARAF and are found in many different cancer types, with highest incidences in melanoma (27-

70%), PTC (60%), and colon cancer (5-22%) (11, 22, 39, 40). Most of these mutations occur in the kinase domain and approximately 80% is result in BRAFV600E (39).

BRAFV600E and RAS mutations are mutually exclusive, supporting the concept that there is a balance of activated ERK levels to allow the cell to grow continuously without entering oncogene-induced senescence. Like most BRAF alterations, the valine to glutamic acid mutation at amino acid 600 (BRAFV600E) occurs in the activation segment.

This destabilizes the internal hydrophobic interactions, leading to a continuously open

9 and active conformation that is independent of RAS activation, and allows BRAFV600E to be active as a monomer (41-43). This results in about 4.6-fold increase in ERK activation than in wild-type (WT) BRAF and when the kinase domain is isolated alone, it is approximately 500-fold more active than the WT BRAF kinase domain (42). This increase in ERK activation increases transcription of ERK-regulated transcription factors and feedback effectors; however, cells with BRAFV600E are insensitive to the negative feedback regulation and continue to activate MEK (44). Taken together, BRAFV600E is a potent oncogene in many cancers and continues to be a focus for targeted therapy.

1.3.2 Therapeutic BRAF Inhibitors

Vemurafenib and Dabrafenib are two FDA-approved inhibitors of BRAFV600E

(selective over WT BRAF) that potently inhibit MEK and ERK activation by BRAFV600E

(45). Clinical trials showed remarkable short-term efficacy in BRAF-mutated melanomas and led to the approvals of both compounds by the FDA for patients with this cancer (46,

47). Similar results have been reported for BRAFV600E-mutated papillary thyroid cancer

(PTC) and the combination of Dabrafenib and Trametinib (MEK inhibitor) was recently

FDA-approved for anaplastic thyroid cancers with mutant BRAF (48, 49). Investigators tested the efficacy of Vemurafenib in BRAF-mutated thyroid cancers in a phase I clinical trial (50), leading to the implementation of a prospective phase II trial (51). However, for both melanoma and PTC, while the initial responses are often dramatic, acquired resistance is nearly universal. In melanoma, this resistance occurs through a variety of mechanisms (52). These include reactivation of the MAPK pathway via alternative splicing of the BRAFV600E gene to enhance BRAF dimerization (41), enhanced EGFR

10 and PDGFR-β expression by activation of TGF-β (53), activating NRAS mutations (54,

55), and increased eIF4F eukaryotic translation initiation complex formation (56). In

PTC, mechanisms of resistance include increased expression and activation of HER3

(57), upregulation of cMET (58-60), and the development of activating mutations in RAS genes (61). Additionally, BRAFV600E inhibition initially decreases activated ERK; however, this also decreases MAPK negative feedback, which in turn activates RAS- dependent CRAF-BRAF dimerization (62, 63). Targeting the MAPK pathway, in particular BRAFV600E, is a promising therapeutic approach. However, effects are limited and incomplete, requiring further translational studies on combination and alternative therapies for patients with these aggressive tumors.

1.4 Molecular Characterization of PTC Invasive Fronts

Although thyroid cancer is generally very curable, prognosis dramatically worsens for patients with distant metastasis. Progressive distant metastasis and local gross invasion of the primary tumor predict disease-related mortality (64). Our group compared human patient mRNA expression patterns using oligonucleotide microarrays among normal thyroid tissue, intratumoral invasive tissue, and tumor tissue from the central region in order to understand the molecular mechanisms of aggressive and invasive thyroid cancers

(65). There were a number of biological processes and molecular functions upregulated in the invasive fronts, including genes related to transcription and G protein regulation.

The overexpression of many genes and activation of pathways suggest a role in the process of epithelial-to-mesenchymal transition (EMT) (66), via the integrin, TGF-β,

Notch, NFκB, and PI3K pathways. Vimentin is also a key player in EMT, as it is an

11 intermediate filament with high expression in mesenchymal cells. Higher vimentin IHC staining was associated with invasion, nodal metastasis, and multifocality in the validation cohort of PTC samples. Because TGF-β, Notch, NFκB, and the PI3K pathways were well-characterized EMT pathways (66), the group focused on the integrin pathway, small G proteins, and vimentin phosphorylation in thyroid cancer EMT. Ingenuity pathway analysis implicated the p21-activated kinases (PAKs) as central coordinating signaling molecules for this pathway, identifying a novel kinase family involved in papillary thyroid cancer invasion.

1.5 p21-activated Kinases (PAKs)

1.5.1 Canonical Signaling and Functions

p21-activated kinases (PAKs) are serine/ threonine kinases related to the STE20 family of yeast proteins. They are highly expressed in brain tissue and bound by activated

(or GTP-bound) small G proteins, CDC42 and RAC1, members of the RAS-related GTP- binding proteins (i.e. p21 family) (67). The PAK family is divided into two groups, group

I (PAKs 1-3) and group II (PAKs 4-6), based on sequence, structure, and activation differences (Fig. 1.1) (68, 69). All PAKs have a N-terminal regulatory domain with a p21-binding domain (PBD), sometimes called a CDC42/ RAC1 interactive binding

(CRIB) domain, and a C-terminal kinase domain. The sequences and structure are more conserved within groups rather than between groups I and II, implying that regulation of the groups differ. Indeed, although PAK5 can bind CDC42 and RAC1, this does not activate PAK5; instead, PAK5 has constitutive kinase activity (70). This dissertation

12 focuses on group I PAKs, which will be rationalized in Section 1.6. Therefore, the remaining introduction will not include details regarding group II regulation and function.

In their inactive state, group I PAKs homodimerize asymmetrically, where the autoinhibitory domain (AID) in the N-terminus of one monomer overlaps the kinase domain of the other monomer (71, 72). However, recent solution-phase structures of full- length PAK1 indicate that inactive PAK1 proteins may actually be monomers that fold to auto-inhibit themselves in cis, rather than in trans (73). PAKs are typically activated by the binding of CDC42 or RAC1, which induces a conformational change to disrupt the dimer and allow autophosphorylation on several threonine and serine residues (71, 72,

74, 75). These sites, such as threonine 423 and serine 144 on PAK1, are used as markers of PAK activation in the field and in our studies, particularly in Chapter 2. Conversely,

PAKs can be inactivated by protein phosphatases, POPX1, POPX2, and PP2A (76, 77) and a number of inactivating protein-protein interactions (reviewed in (68)) to regulate proper PAK signaling.

Figure 1.1 PAK family and protein structure. Schematics of prominent domains and features of the group I and II PAKs, indicating similarities within groups. P= proline-rich region, CRIB= CDC42/RAC interactive 13 binding domain, AID= autoinhibitory domain, PIX= β-PIX binding motif, kinase= kinase domain, orange circle= phosphorylation. Numbers indicate amino acid positions. Adapted from Kumar et al (78). CDC42 and RAC1 can be activated by a number of mechanisms, most notably by integrin-mediated signaling (79) and RAS-activated RAC-selective guanine exchange factors (GEFs) (80). Activated PAK, primarily by CDC42 or RAC1, can then signal downstream to a host of substrates involved in many cellular functions and different cellular compartments – over 30 PAK substrates have been characterized thus far (68, 81,

82). Particularly, PAKs were studied extensively for their involvement in cytoskeletal rearrangements, filopodia formation, membrane ruffling, and organization of focal adhesions (83, 84). In this manner, PAKs can phosphorylate and regulate actin-, microtubule-, and intermediate filament-related proteins, such as myosin light chain kinase (MLCK) (85), LIM-kinase (86), tubulin cofactor B (TCoB) (87), Op18/ stathmin

(88), vimentin (89), among many others. It is also found in complexes with NCK (adapter protein), βPIX (guanine nucleotide exchange factor), and GIT1 (G-protein-coupled receptor kinase –interacting target 1) (90, 91).

Group I PAKs also play a role in MAPK signaling and the roles they play are becoming increasingly complex. PAK1 directly phosphorylates MEK1 at serine 298, which may be stimulated by adhesion and integrin activation (92, 93); however, the role of this phosphorylation is controversial. Early evidence reported that the phosphorylation at serine 298 “primes” MEK1 to be phosphorylated and fully activated by CRAF and, therefore, phosphorylate ERK (93, 94). Others argue that MEK phosphorylation by

PAK1 at S298 stimulates autophosphorylation of MEK at S217 and S221, independent of

CRAF, suggesting an alternative activation of the MAPK pathway via PAK1 (95).

Additionally, independent of its kinase function, PAK1 can bind to MEK1/2 at the cell membrane and facilitate CRAF activation of MEK1/2 at S217/221, thus acting as a scaffold for active MEK1/2 to phosphorylate ERK1/2 (96). While the exact mechanisms of PAK activation of MEK may be cell-type dependent or dependent on other cellular contexts or interacting proteins, each underscores the importance of PAK’s involvement in propelling MAPK signaling. The PAK-MAPK crosstalk is even further characterized by PAK1 and ERK2 direct interaction via PAK1 N-terminal regulatory domain (97).

Stimulation by adhesion signaling and platelet-derived growth factor (PDGF)-BB results in PAK1 and ERK2 forming a complex. ERK2 subsequently phosphorylates PAK1 at threonine 212; however, it is suggested that this phosphorylation negatively regulates

ERK-mediated transcription, acting as a negative feedback loop. PAKs can also bind and phosphorylate CRAF on serine 338 through interaction with the PAK C-terminus. This is regulated through PI3K activation (98-100). Additionally, PAKs promote the PI3K pathway as a scaffold protein. PAK1 can interact as a complex with PDK1 and AKT to mediate AKT phosphorylation (101). This interaction also occurred with the PAK kinase domain; however, kinase function was unnecessary for AKT activation. Thus, PAK functionally interacts with multiple components of the MAPK and PI3K pathways.

Another important role for PAKs is their involvement in cell cycle regulation.

PAKs dynamically translocate to different regions of the cell and activate different targets based on the phase of mitosis. In connection with its role in cytoskeletal rearrangements, phosphorylated PAK1 at threonine 212 (likely by CDK1/ cyclin B1 complexes in this

15 context) localizes to the microtubule organizing centers and centrosomes during mitosis and regulates microtubule dynamics throughout the different stages (102). During prophase, PAK1 localizes to the chromosomes, potentially interacting with the kinetochore or centromere components, and phosphorylates histone H3 at serine 10

(103). Tiam1, a guanine-nucleotide exchange factor for RAC1, localizes to the centrosome, where its phosphorylation is required for PAK activation at the centrosome to antagonize centrosome separation, independent of Aurora-A (104). Active PAK also forms a complex at the centrosomes with βPIX and GIT1, phosphorylating GIT1 (105).

This phosphorylation event disrupts the complex, allowing open and activated PAK to now bind and phosphorylate the centrosome-associated kinase, Aurora-A. During metaphase, a portion of PAK1 localizes to the metaphase plate, while a population is active at the centrosome. PAK colocalizes and phosphorylates polo-like kinase 1 (PLK1) at the spindles poles, central spindle, and midbody, key regulatory steps in maintaining proper spindle tension and bipolar spindle formation (106). Finally, as the cell pinches off to form two new cells in cytokinesis, PAK1 translocates to the contraction ring (103).

Taken together, PAKs play important roles in maintaining proper progression through the cell cycle.

PAKs are involved in several other signaling pathways and interactions not discussed in detail here, including roles in inhibiting apoptosis and angiogenesis, as well as nuclear-specific functions (82, 107). The diverse and expansive contributions of PAKs underscore its importance in maintaining proper cellular functions, and therefore,

16 alterations to PAK expression and regulation lead to many physiological problems, in particular, cancer.

1.5.2 Alterations in Cancers

Overactivation of group I PAKs, in particular PAK1, are observed in a number of cancers (108). Point mutations in the actual PAK genes are rare; however, increased PAK activity is often due to gene amplification (ex: 11q13 containing PAK1), protein overexpression, or mutations in positive or negative regulators of PAK, such as RAC1

(109), neurofibromatosis (NF) 1 (110), or NF2 (111, 112). Activated group I PAKs are upregulated in malignant peripheral nerve sheath tumors (MPNSTs), often with inactivation of NF1, and displayed further activation in metastatic MPNST tissue (113).

PAK1 expression is increased in meningiomas that lack NF2 and increase meningioma cell proliferation (114). Copy number gain and high expression of PAK1 have been reported in wild-type BRAF melanomas (115), and are often observed in breast cancer

(116-120) and ovarian cancer (121-123). PAK1 overexpression in these cancers is associated with poor prognosis and resistance to treatments. Other cancers with PAK amplification and/or overexpression are colon (124), pancreatic (125), and glioblastoma

(126). Based on the mRNA array finding that integrin signaling was upregulated in the invasive edges of aggressive thyroid cancers (65), our group tested an additional PTC cohort and found via immunohistochemistry that PAK1 protein and phosphorylated PAK

(pPAK) were increased in the invasive fronts compared to both the center of the tumor and normal thyroid tissue (127). It is not surprising that given the diverse roles of PAK in cellular functions and the prevalence across many tumor types, PAKs are involved in

17 several of the Hallmarks of Cancer, including evading apoptosis, self-sufficiency in growth signals, tissue invasion and metastasis, and sustained angiogenesis (128).

Therefore, there is heightened interested in targeting PAKs for therapeutic benefit.

1.5.3 Molecular and Pharmacologic Inhibitors

Since PAKs are overexpressed and/or overactivated in a number of cancers, several inhibitors have been developed to help aid in the basic understanding of PAK functions in cancer cells, as well as for potential translational therapeutic options. The isolated PAK inhibitory Domain (PID) is a small peptide covering amino acids 83-149 of

PAK1, which encompasses the autoinhibitory domain (129). The PID is currently used only for laboratory studies and inhibits the kinase activity of group I PAKs when exogenously expressed in vitro and, recently, can be induced in vivo (130).

Pharmacologic inhibitors of PAK are generally either ATP-competitive, where they bind to the active site in the kinase domain and block ATP from binding, or allosteric, where the inhibitors bind in a region outside the active site and induce structural changes to render inactivity. Allosteric inhibitors are often more selective than

ATP-competitive inhibitors, but are usually limited in their potency. One of the most widely used allosteric PAK1 inhibitors is inhibitor p21-activated kinase-3 (IPA-3) (131).

IPA-3 is selective for PAK1, but it is not well tolerated in vivo and is limited even in cell culture assays. Recently, a newer series of allosteric inhibitors, based on 1,4- naphthohydroquinone, that block the interaction between CDC42 and PAK1 have been developed and exhibit high selectively for PAKs 1 and 3, but not 2 (132). These

18 compounds are useful in cell culture and show inhibition of downstream PAK phosphorylation.

The kinase domain is highly conserved for group I PAKs and shares significant homology with the group II PAKs. Therefore, many of the ATP-competitive PAK inhibitors potently inhibit all group I PAKs and to some extent the group II PAKs. Many of the early PAK inhibitors were “broad-spectrum” and inhibited many kinases, with particular affinity for the STE20 family of kinases (133-135). This lack of specificity, has inhibited its use for laboratory studies and likely would cause toxicities as a clinical therapeutic.

Pfizer developed a series of ATP-competitive PAK inhibitors based on the pyrrolopyrazole compounds (136). One in particular, PF-3758309, is a pan-PAK inhibitor that has been utilized for in vitro and in vivo PAK inhibition in many pre-clinical studies to show reduced tumor growth across many tumor types (115, 137-141). PF-3758309 was also the only PAK inhibitor, to date, to be tested in a Phase I clinical trial for advanced solid tumors (Clinicaltrials.gov, NCT00932126); however, the study was terminated due to pharmacokinetic problems and adverse effects without positive tumor response. Many investigators still use PF-3758309 as a PAK inhibitor, even with significant off-target effects, and the contribution of PAK to the biological outcomes needs to be carefully tested by other confirmatory methods.

Several pharmaceutical companies subsequently developed and reported activity of more specific PAK1 or group I PAK inhibitors in an effort to be more selective against group II PAKs. AstraZeneca developed a series of ATP-competitive group I PAK

19 inhibitors (142, 143). AZ13705339 is very potent again PAK1 in vitro with high selectivity for PAKs 1 and 2, but lacks the capabilities for in vivo application. Its analog,

AZ13711265, had been tested in vivo for pharmacokinetic testing and was tolerated, although further testing on dosing and target validation in vivo are needed.

Afraxis/Genetech also developed a series of ATP-competitive group I PAK inhibitors, initially designed to treat certain neurological disorders associated with PAK activation.

FRAX597 is particularly potent against group I, but not group II PAKs in this screen. It did show some inhibitory activity against RET and other kinases, as well (144).

FRAX1036 had fewer off-target effects and maintained a high potency against PAK1

(114) and has been effective in in vitro and in vivo inhibition of group I PAKs (116, 145,

146). Despite the improved kinase selectivity, the FRAX compounds have off-target inhibition of hERG potassium channels, necessary for regulating electrical activity in the heart. Therefore, therapeutic application for these compounds in humans is unlikely.

Subsequently, Genentech developed a derivative of FRAX1036, G-5555, with minimal hERG inhibition and better permeability (147). G-5555 is highly selective of group I

PAKs over group II PAKs, although there are some off-target effects on a few other serine/ threonine kinases, notably SIK2, MST3, and MST4. G-5555 remains one of the most selective group I PAK inhibitors, as it is tolerated in vivo and reduces xenograft tumor growth in non-small cell lung cancer, breast cancer, and colorectal cancer mouse models (130, 147, 148). Importantly, safe in vivo doses of G-5555 must be within a

“narrow therapeutic window”, as cardiac toxicities can occur at higher doses, potentially due to on-target PAK1 and PAK2 effects in maintaining normal cardiac function (148).

Studies in Chapter 2 utilized both FRAX1036 and G-5555 to test pharmacologic group I

PAK inhibition in thyroid cancer.

1.6 Non-canonical BRAF-PAK1 Crosstalk

As explained in Section 1.4, our lab previously discovered PAK upregulation and activation in the invasive edges of thyroid cancers (65, 127). To understand the functional role of PAK in thyroid cancer, we used the PID peptide to inhibit all group I PAKs and found that thyroid cancer cell migration is decreased upon loss of PAK activity (127).

Further studies focused on the relationship between thyroid cancer and group I PAKs, rather than group II. Additionally, siRNA knock-down of individual group I PAK members demonstrated that PAK1 had the greatest effect on thyroid cancer cell migration. Thus, it was apparent that group I PAKs, specifically PAK1, play a role in thyroid cancer and further mechanistic studies were warranted. Since aberrant activation of the MAPK pathway often occurs in thyroid cancers and BRAFV600E is the most common protein mutation, our group investigated the unknown relationship between

PAK1 and BRAF in thyroid cancer.

1.6.1 Signaling and Cellular Functions

PAK has known interactions with the MAPK pathway via CRAF, MEK, and

ERK, yet interactions between PAK and BRAF were unknown. siRNA to BRAF and inhibition of PAK by PID were tested for crosstalk signaling effects (149). Interestingly,

BRAF knockdown decreases PAK activity; however, with PAK inhibition by PID, BRAF function remained intact. Both PAK and MEK can rescue migration defects after reduction of BRAF, but PAK and MEK seem to act via independent pathways, since

21 inhibition of the two together do not synergistically reduce thyroid cancer migration and loss of MEK has no effect on PAK activity. Further, in a mouse model of thyroid cancer, induction of BRAFV600E in the mouse thyroids resulted in the development of PTC-like cancer and increased group I PAK expression and activation. Therefore, in support of the observation that PAK is overexpressed in the invasive fronts of human thyroid cancer

(127), these additional findings suggest a novel “non-canonical” crosstalk between the

MAPK and PAK pathways (Fig. 1.2), in which BRAF regulates PAK function and

BRAF, PAK, and MEK coordinately regulate thyroid cancer cell motility.

Figure 1.2. MAPK and PAK signaling. The left depicts the “canonical” or traditional view of MAPK signaling and the right depicts the p21-activated kinase (PAK) signaling pathway, both involved in promoting 22 cell growth, division, and motility. Our group determined a previously unrecognized crosstalk between BRAF and PAK1, termed “non-canonical”, in which BRAF regulates PAK1 signaling.

1.6.2 Physical Interaction

Since BRAF and PAK signaling are connected, we reasoned that they might also interact physically. Co-immunoprecipitation and immunofluorescence assays determined that, indeed, wild-type BRAF and PAK1 co-localize and interact in both an exogenous overexpression setting and in an endogenous thyroid cancer cell setting (149).

Interestingly, this interaction is enhanced in mitotic thyroid cancer cells. However, the

BRAF-PAK1 interaction is likely transient, as is it difficult to consistently detect this interaction endogenously. Signaling through this complex is likely a regulated process.

1.7 Conclusions and Dissertation Overview

Although the majority of patients with thyroid cancer have an excellent prognosis, limited treatment options exist for those patients with aggressive tumors incurable by standard of care. Advancements in understanding the genetic and molecular components of thyroid cancer have led to clinical trials and approvals of targeted therapies. However, complete responses remain elusive. Recent discoveries from our lab and others have identified both MAPK and PAK pathways as key nodes in cancer development and progression. These pathways have been extensively studied and characterized, yet, the link between BRAF and PAK has only recently been discovered by our group. The details of this interaction require further investigation.

This dissertation investigates the roles BRAF and PAK coordinately play in thyroid cancer. Chapter 2 details the potential of PAK to be pharmacologically inhibited in 23 thyroid cancer. It also proposes a novel concept that PAK activity is necessary for

BRAFV600E/ ERK- mediated thyroid cancer. Chapter 3 further characterizes the BRAF-

PAK1 protein complex by identifying other proteins potentially involved in this interaction. It also highlights the transient nature of this complex. Finally, Chapter 4 summarizes these findings and discusses future directions to continue this work.

Chapter 2: Group I PAK Inhibition in Thyroid Cancer

This chapter describes the work published in Endocrine-Related Cancer:

Knippler CM, Saji M, Rajan N, Porter K, La Perle KMD, Ringel MD. MAPK- and AKT- activated thyroid cancers are sensitive to group I PAK inhibition. Endocr Relat Cancer.

2019 May 1; [Epub ahead of print]

2.1 Abstract

The number of individuals who succumb to thyroid cancer has been increasing and those who are refractory to standard care have limited effective therapeutic options, highlighting the importance of developing new treatments for patients with aggressive forms of the disease. Mutational activation of MAPK signaling, through BRAF and RAS mutations and/or gene rearrangements, and activation of PI3K signaling, through mutational activation of PIK3CA or loss of PTEN, are well described in aggressive thyroid cancer. We previously reported overactivation and overexpression of p21- activated kinases (PAKs) in aggressive human thyroid cancer invasive fronts, and determined that PAK1 functionally regulated thyroid cancer cell migration. We reported mechanistic crosstalk between the MAPK and PAK pathways that are BRAF-dependent but MEK-independent, suggesting that PAK and MEK inhibition might be synergistic. In the present study, we tested this hypothesis. Pharmacologic inhibition of group I PAKs using two PAK kinase inhibitors, G-5555 or FRAX1036, reduced thyroid cancer cell viability, cell cycle progression, and migration and invasion, with greater potency for G- 25

5555. Combination of G-5555 with Vemurafenib was synergistic in BRAFV600E-mutated thyroid cancer cell lines. Finally, G-5555 restrained thyroid size of BRAFV600E-driven murine papillary thyroid cancer by >50% (p<0.0001) and reduced carcinoma formation

(p=0.0167), despite maintenance of MAPK activity. Taken together, these findings suggest both that group I PAKs may be a new therapeutic target for thyroid cancer and that PAK activation is functionally important for BRAFV600E-mediated thyroid cancer development.

2.2 Introduction

The National Cancer Institute estimated that there will be about 52,070 new cases of thyroid cancer in 2019 and 2,170 patients were expected to die from the disease (4).

Although most thyroid cancers are curable, patients with more aggressive tumors rarely respond to standard therapy that typically includes thyroidectomy, radioactive iodine therapy, and thyroid-stimulating hormone (TSH) suppression therapy. Thus, distinct from other patients with thyroid cancer, individuals with these tumors have a poor prognosis and, until recently, no further treatments were FDA-approved. Over the past decade, there have been remarkable advances in defining the molecular underpinnings of aggressive thyroid cancer and this work has resulted in clinical trials and approvals of compounds targeting either the tumor microenvironment (e.g. VEGFR-targeting with Lenvatinib or

Sorafenib) or “genetic drivers,” such as BRAFV600E and MEK, in patients with anaplastic thyroid cancer (20). Despite these advances that result in partial remissions, none result in complete remission or cure, and progression, while delayed, is inevitable. These data

26 demonstrate that critical knowledge gaps remain to improve survival rates for patients with aggressive thyroid cancer.

Genetic alterations that result in enhanced activation of the MAPK and AKT pathways are commonly seen in thyroid cancers, and are often enriched in populations with aggressive forms of the disease. The most common form of thyroid cancer, well- differentiated papillary thyroid cancer (PTC), is driven primarily by mutually exclusive mutations that activate MAPK signaling; ~60% have activating BRAFV600E, 15% have

RAS mutations, and 12% have gene rearrangements (such as RET/ PTC) (11). By contrast, follicular thyroid cancers (FTCs) are largely driven by activation of PI3K signaling and some MAPK activation, with a high frequency of activated RAS (NRAS) mutations, PTEN loss, PIK3CA mutations, or PAX8- PPARγ fusions. Anaplastic thyroid cancers (ATCs) have the worst prognosis, with a mean survival of only 6 months, and often have mutational activation of both PI3K and MAPK pathways along with mutations in TP53, TERT promoter, and/or epigenetic genes, suggesting they derive from more well-differentiated forms of the disease (11). Of these oncogenes, there has been particular interest in targeting BRAFV600E because of its high frequency in thyroid cancer overall, its enrichment in clinical trial populations, and the availability of highly specific inhibitors with clinical activity, such as Vemurafenib and Dabrafenib. Although initial responses to these drugs are promising, resistance is inevitable (57-61).

Group I PAKs are a family of kinases that are activated by the small Rho

GTPases, RAC1 and CDC42, and signal downstream to regulate cell motility and invasion pathways as well as oncogenic growth-activating pathways, including MAPK

27 and AKT (67, 96, 101, 107). We recently identified a novel role for group I PAKs, specifically PAK1, in thyroid cancer. Initial experiments using an unbiased mRNA microarray and subsequent validation with immunohistochemistry identified activation of three pathways in the invasive fronts of aggressive human PTCs: AKT and TGFβ that had previously been reported, and PAK, a previously unstudied pathway in PTC (65,

127). Subsequent studies demonstrated that upon induction of BRAFV600E in mouse thyroids, PTCs develop and group I PAK expression and activity increased (149). We further determined that BRAF protein expression is necessary for PAK activity and that they co-localize. Unexpectedly, BRAF-mediated PAK activity was independent of MEK kinase activity or MEK expression, despite known crosstalk between PAK and

MEK/ERK. Finally, BRAF, PAK1, and MEK coordinately regulate thyroid cancer cell motility, suggesting complex but not entirely overlapping signaling pathways (149).

These data were subsequently supported in melanoma and colon cancer cells, suggesting a generalized signaling interaction (150).

PAKs are upregulated in a number of cancers and therefore, have been a target of recent interest. Over the past several years, several PAK inhibitors have been developed, but have varied efficacy (151). To date, two of the most selective ATP-competitive group

I PAK inhibitors are FRAX1036 (114, 116) and its derivative, G-5555 (147, 148). G-

5555 had shown promising results in preclinical models of breast and lung cancers, but its effectiveness in thyroid cancer or in the context of BRAFV600E-driven tumorigenesis was unknown. Therefore, we hypothesized that pharmacologic group I PAK inhibition,

28 alone or in combination with BRAFV600E or, potentially, AKT inhibitors, are effective treatment strategies, depending on the driver mutations.

In this study, we demonstrated that group I PAK activity in thyroid cancer can be inhibited by both G-5555 and FRAX1036, and that G-5555 was more potent. PAK inhibition reduced thyroid cancer cell viability, cell cycle progression, migration, and invasion in vitro. Further, combination of G-5555 and a BRAFV600E inhibitor was synergistic on thyroid cancer cell viability in BRAF-mutated thyroid cell lines and with an AKT inhibitor in a PIK3CA-mutated thyroid cancer cell line. Finally, oral treatment with G-5555 restrained BRAFV600E-driven thyroid size and reduced tumor formation in a robust mouse model of inducible thyroid-specific BRAFV600E overexpression. These data provide further pharmacologic evidence that group I PAKs are important for thyroid cancer growth and motility, that may be exploited for therapeutic intent, and that PAK likely plays an important role in BRAFV600E-mediated thyroid tumorigenesis.

2.3 Materials and Methods

2.3.1 Cell lines and cell culture

Human thyroid cancer cell lines with varying driver mutations (Table B.1, (152)) were used for in vitro studies. TPC1, FTC133, BCPAP, SW1736, and 8505C were generous gifts of Dr. Rebecca Schweppe (University of Colorado, CO) with permission from the original researchers who established the cell lines: TPC1 – H. Sato, Kanazawa

University, Japan (153), FTC133 – P. Goretzki, University of Leipzig, Germany (154),

BCPAP- N. Fabien, University of Medecine Lyon-Sud, France (155), SW1736 – N-E.

Heldin, University Hospital, Uppsala, Sweden (156), 8505C- T. Ito, Radiation Effects

Research Foundation, Japan (157). hTh74 was a generous gift from Dr. Heldin (158). K1 was purchased from Millipore Sigma (159). THJ-16T was a generous gift from J.

Copland, Mayo Clinic, Florida (160). TPC1 and FTC133 were cultured in DMEM medium with 10% fetal bovine serum (FBS), 1% glutamine, and 1% nonessential amino

o acids at 37 C and 5% CO2. BCPAP, hTh74, SW1736, 8505C, K1 and THJ-16T were cultured in RPMI medium with 10% FBS, 1% glutamine, and 1% nonessential amino

o acids at 37 C and 5% CO2. All cell lines were independently validated for identity by

DNA fingerprinting after receipt of the cells and periodically over the course of the experiments.

2.3.2 In vitro drug reagents

The drugs used in this study were the following: FRAX1036, group I PAK inhibitor (generous gift from Dr. Jonathan Chernoff, Fox Chase Cancer Center, PA); G-

5555, group I PAK inhibitor (MedChemExpress, HY-19635); PLX4032, BRAFV600E inhibitor (Selleckchem, S1267), MK2206, AKT inhibitor (MedChemExpress, HY-

10358). For in vitro assays, drugs were reconstituted in DMSO at a starting concentration of 10 mM and diluted in culture medium as needed for individual experiments.

2.3.3 Cell viability assays

Cell lines were optimized for seeding density in 96 well plates. Cells were seeded in triplicate at the following densities (cells/ well): TPC1: 1x103, FTC133: 1.25x103,

BCPAP: 5x103, HTh74: 1.25x103, SW1736: 2.5x103, 8505C: 2.5x103, K1: 5x103, and

THJ-16T: 1x103. 24 hours after seeding, cells were washed with PBS and incubated with medium containing 1% FBS for 24 hours. Cells were treated with 100 µL of the indicated

30 drug diluted in growth medium with 1% FBS for 72 hours. After 72 hours, 10 µL of

WST-8 (Cell Counting Kit-8, Dojindo Molecular Technologies, Inc.) was added per well

o and the plates were incubated for 1-3 hours at 37 C and 5% CO2. Optical density (OD) was read at 450 nm and the percent viability was calculated by comparing the OD of treated cells versus control (untreated) cells. At least three biological replicates in triplicate were performed for each cell line.

2.3.4 Protein extraction and western blots

Cells were seeded on 10 cm plates for 24 hours, washed with PBS, and incubated with medium containing 1% FBS for 24 hours. The indicated compounds were diluted in medium with 1% FBS and added to the cells for 24 hrs (G-5555, FRAX1036, and combination of G-5555+ MK2206) or 3 hrs (combination of G-5555 + PLX4032). At the end of incubation, the medium was aspirated and the cells were washed with ice-cold

PBS. Cells were scraped in 0.5 mL cold PBS, transferred to 1.5 mL tubes, and centrifuged at 500 g for 10 min. After a second wash with 0.5 mL cold PBS, cells were centrifuged at 500 g for 10 min and the supernatant was aspirated. Cell pellets were resuspended and lysed in mammalian protein extraction reagent (MPER, Thermo Fisher

Scientific) for 5 minutes on ice. After lysis, the lysates were centrifuged at 16,000 g for

10 minutes and the supernatants were collected. Protein concentration was estimated using bovine serum albumin as a standard and the Pierce BCA Protein Assay Kit

(Thermo Fisher Scientific).

For western blotting, 10-30 µg of protein for each sample was combined with 1x

NuPAGE LDS sample buffer (Thermo Fisher Scientific) and 50 mM dithiothreitol (DTT)

31 and boiled for 5 minutes. Proteins were separated on NuPAGE 4-12% Bis-Tris gels

(Thermo Fisher Scientific) and transferred to nitrocellulose membranes (Bio-Rad).

Membranes were blocked in 5% bovine serum albumin (BSA) or 5% non-fat milk (for pMEK S298 and thyroglobulin antibodies only) in 1x tris-buffered saline with 0.05%

Tween 20 (TBST) for one hour. Primary antibodies were diluted in BSA or milk (pMEK

S298, thyroglobulin) and incubated overnight at 4oC. Membranes were washed with

TBST three times for 5 minutes each and incubated at room temperature for an hour with near-infrared fluorescent secondary antibodies (1:7500 or 1:15,000 in BSA; LI-COR).

After incubation, membranes were washed three times for 5 minutes each and the bands were detected using the Odyssey CLx Imaging System (LI-COR). Image analysis was conducted using the Image Studio software (LI-COR). Membranes were blotted with

GAPDH as a loading control.

Primary antibodies used to detect the indicated proteins are as follows: pPAK1

S144/ pPAK2 S141 (1:500, 2606, Cell Signaling), PAK1 (1:1000, 2602, Cell Signaling), pVimentin S55 (1:500, D076-3, MBL), Vimentin (1:4000, V6630, Millipore Sigma), pMEK S298 (1:500, 44460G, Thermo Fisher), MEK1/2 (1:1000, 9122, Cell Signaling), pAKT1/2/3 S473 (1:1000, sc-7985-R, Santa Cruz Biotechnology), AKT1/2/3 (1:500,

9272, Cell Signaling), GAPDH (1:15,000, 2118, Cell Signaling), pERK1/2 T202/ Y204

(1:500, 9101, Cell Signaling), ERK1/2 (1:1000, 4695, Cell Signaling), pGSK-3-α/β S21/9

(1:1000, 9331, Cell Signaling), GSK-3-α/β (1:1000, 5676, Cell Signaling).

2.3.5 Cell cycle analysis

TPC1, SW1736, and FTC133 cells were seeded on 10 cm plates for 24 hours. The plates were washed with PBS and incubated with medium containing 1% FBS for 24 hours and then treated with 0.5 µM G-5555, 5 µM FRAX1036, or DMSO-only control for 24 and 48 hours. After each incubation, cells were washed with PBS, trypsinized, and counted using the Countess Automated Cell Counter (Invitrogen). Approximately 0.5-

2x106 cells were collected and resuspended in 0.5 mL PBS. 4.5 mL ice-cold 70% ethanol was added dropwise over 30 seconds to 1 minute. Cells were incubated overnight at 4oC then transferred and stored at -20oC until staining and flow cytometry analysis were performed. At that time, cells were centrifuged at 4oC and 300 g for 5 minutes. The supernatant was removed and the cells were washed twice with PBS. After the last wash and centrifugation, the supernatant was removed and the cells were resuspended in 0.5 mL propidium iodide (PI) solution (40 µg/ mL PI, 100 µg/ mL RNAse A in PBS). They were then incubated at 37oC for 30 minutes. Cells were filtered and analyzed for relative amounts of DNA using the LSR II flow cytometer (BD Biosciences) by the Analytical

Cytometry Core. Three biological replicates were performed for each cell line.

2.3.6 Migration and invasion assays

For migration assays, 1x105 TPC1 or FTC133 cells were seeded on the top of a

Boyden chamber insert (8 µm pores, 12 mm diameter, Millipore) in a 24-well dish, in

DMEM with 1% FBS in both the top and bottom well. Wells with no cells but with medium were prepared as background controls. After 3 hours at 37oC, the medium was changed and 0, 0.25 µM, or 1 µM G-5555 in DMEM with 1% FBS was added to the top

33 of the wells with the cells and DMEM with 10% FBS and no G-5555 was added to the bottom. Cells were incubated at 37oC for 16 hours. Equal numbers of cells were also seeded and treated as described above in a 48-well plate without a Boyden chamber

(similar diameter as a Boyden chamber), to be used for a WST cell viability assay to confirm that cells remained viable in the experimental conditions.

After 16 hours, medium was aspirated and the chambers were washed with distilled water. The top and bottom of the Boyden chamber were stained in crystal violet solution (0.05% crystal violet and 10% formalin) for 20 minutes at room temperature.

The chambers were washed with distilled water and dried. Cells were swabbed off the top and bottom of the chambers using a moist Q-tip and the tip was placed into a 1.5 mL tube. A final concentration of 80% methanol was used to extract the color. Aliquots of each sample were placed in a 96-well dish and the absorbance was detected on a spectrophotometer at 570 nm. Percent migration was determined by calculating the ratio of the OD of the bottom of the chamber to the OD of the total of top and bottom. At the time of staining, WST solution was added to the 48-well plate and incubated at 37oC for one hour. The absorbance was read at 450 nm and percent viability was calculated in comparison to the DMSO control. Cells were seeded in duplicate and at least three biological replicates were performed.

For invasion assays, Matrigel-coated invasion chambers were warmed and rehydrated per the manufacturer’s protocol (Corning). TPC1 or FTC133 cells were seeded and treated with G-5555 as above for the migration assays. After 16 hours, cells were washed and stained with crystal violet solution as previously described. For

FTC133 cells, staining time was extended to one hour due to faint staining from shorter incubations. The top layer of the invasion chamber with the Matrigel was swabbed off and pictures of the bottom layer were taken with a Zeiss AxioCam MRc5 camera from a

Zeiss Axiovert 40 CFL microscope at 10X magnification. Images were acquired using the AxioVision SE64 software. At least 5 randomly selected pictures were taken per sample. The images were imported into Photoshop (CS5.1, Adobe) and the number of pixels of purple color (i.e. crystal violet-stained cells) per image was calculated using the

Magic Wand tool with a tolerance of 32 and cache level of 1. Percent coverage was calculated by dividing the average number of colored pixels by the total number of pixels in the image. The average percent covered from the duplicate wells was compared to the

DMSO control to calculate the fraction of control that invaded. The cells were seeded in duplicate and there were at least three biological replicates performed.

2.3.7 Thyroid-specific inducible BRAFV600E mouse model

We utilized a thyroid-specific inducible BRAFV600E mouse model previously described by Chakravarty D et al (161) and generously gifted to us by Dr. James Fagin

(Memorial Sloan-Kettering Cancer Center, NY). Double transgenic mice on the FVB/N background were generated by Chakravarty D et al (161) to have reverse tetracycline transactivator cloned downstream of the bovine thyroglobulin promoter (Tg-rtTA) and

MYC-tagged BRAFT1799A cloned downstream of a tetO promoter (tetO-BRAFV600E). In the presence of doxycycline (dox), mice develop thyroid tumors in one week and overexpression of BRAFV600E was confirmed by pERK IHC.

2.3.8 G-5555 treatment in vivo

For in vivo G-5555 treatment, a dosing strategy was followed in accordance with

Ndubaku C et al (147). G-5555 hydrochloride (MedChemExpress, HY-19635A) was dissolved in MCT vehicle (0.5% (w/v) methylcellulose, 0.2% (w/v) Tween 80 in sterile water) at 2.5 mg/mL and sonicated for 30 minutes at 4oC. At the time mice were given doxycycline in the chow, they were administered either 25 mg/kg G-5555 or MCT vehicle with a dose volume of 10 mL/kg, via oral gavage, twice daily. After one week, mouse thyroids were harvested. One lobe was formalin-fixed for subsequent immunohistochemistry and one lobe was frozen. The thyroid volume of the frozen lobe was measured ex vivo. Volume (mm3) was calculated using the following: V = L x W x W x 0.52 (V= volume, L= diameter of long axis, W= diameter of short axis). All animal protocols, care, and studies were approved by The Ohio State University Institutional

Animal Care and Use Committee.

2.3.9 Immunohistochemistry (IHC)

Conditions were optimized for each target and antibody by the Comparative

Pathology & Mouse Phenotyping Core. Briefly, formalin-fixed, paraffin-embedded thyroid tissues were deparaffinized in xylene and rehydrated. Target retrieval was conducted at pH6, heated, and then cooled to room temperature. 3% hydrogen peroxide in methanol was used to block endogenous peroxidase. Tissues were blocked with serum- free protein and primary antibodies in Dako antibody diluent with reducing agents were added for 30 minutes. Biotinylated goat anti-rabbit secondary antibody was added for 30 minutes at 1:200 for pERK and Ki67 and 1:500 for cleaved caspase-3. The Vector RTU

ABC Elite complex was added for 30 minutes and DAB was added for 5 minutes. Slides were counterstained with hematoxylin, rinsed, dehydrated with ethanol, cleared in xylene, and mounted with coverslips.

Slides were imaged at 20x and/or 40x using the Caliper Life Sciences Vectra

2.0.8 quantitative pathology imaging system (PerkinElmer). 20x images were analyzed using the inForm 2.1.1 software. For Ki67 analysis, images were segmented by tissue

(thyrocyte vs. non-thyrocyte), as well as cell compartment (nuclei vs. cytoplasm) and percent positive thyrocyte DAB nuclear staining was calculated. For cleaved caspase-3, only cell compartment was segmented and percent positive DAB in the cytoplasm was calculated. One sample of a doxycycline and G-5555 treated mouse contained no thyroid on the slides for Ki67 and cleaved caspase-3, therefore n=16 for these analyses.

Primary antibodies used for IHC are the following: pERK1/2 T202/ Y204 (1:900,

9101, Cell Signaling), Ki67 (1:100, Thermo Fisher, RM-9106), and cleaved caspase-3

(1:180, 9661, Cell Signaling)

2.3.10 Protein extraction from mouse thyroids

Thyroid lobes from individual mice were mechanically homogenized on ice in

MPER buffer (Thermo Fisher Scientific). Lysates were centrifuged at 16,000 g for 15 minutes and the supernatants were collected. Samples were concentrated using Amicon

Ultra Centrifugal Filter devices with a 10 kDa molecular weight limit (MilliporeSigma).

Protein concentration was estimated using the Pierce BCA Protein Assay Kit (Thermo

Fisher Scientific). 10-20 µg of protein were separated on western blots and analyzed, as

37 detailed above. Quantitation of pPAK, pERK, and GAPDH bands on the western blot images was conducted using the ImageJ software (Version 1.51, NIH).

2.3.11 Statistical analysis

IC50 values were estimated by four-parameter logistic regression models. For each drug and cell line combination the weighted mean IC50 was then calculated from the individual experimental results using the inverse variance as the weight. Comparisons between G-5555 and FRAX1036 within a cell line were made by inverse variance weighted t-tests using the IC50 values from each experiment. In cell lines for which IC50 was not estimable for one or both drugs, linear dose-response slopes were compared between drugs using linear mixed models. Doses below 0.5 µM were excluded from the linear mixed models due to lack of effect and to better fit the linear form of the models.

The Holm’s procedure was applied to each family of comparisons to control the familywise type I error rate at α = 0.05.

For the effects of G-5555 in vivo, the primary endpoint was development of thyroid cancer at day 7 compared between the G-5555 treated group and the vehicle control group, with treatment starting on day 1 and simultaneous BRAFV600E induction.

The G-5555 treated group was expected to have fewer mice with carcinoma development than the control group. We expected we needed n=15 mice per group to detect a 50% decrease in carcinoma development (99% to 49%) with over 80% power at one-sided

α=0.025. Additional controls using 10 mice without BRAFV600E will control for G-5555 effects on normal thyroid (5 of each sex to assess for sex as a biological variable).

The following statistical analyses were conducted via GraphPad Prism 7. For cell cycle analysis, one-way ANOVA tests followed by Holm’s procedure were conducted for each time point separately and separately within each phase in the cell cycle (i.e. G0/G1,

S, G2/M) to compare the control, G-5555, and FRAX treatments. For migration, invasion, and in vivo thyroid size, comparisons among different treatments were analyzed using a one-way ANOVA test followed by Holm’s procedure. Analysis of carcinoma incidence was performed with a Fisher’s exact test. Difference in Ki67 IHC staining was estimated with a Mann-Whitney test. Comparisons of in vivo protein expression and thyroid volume were fit with a linear regression line. For all tests, P value ≤ 0.05 was accepted as significant. ****P ≤ 0.0001, ***P ≤ 0.001, **P ≤ 0.01, *P ≤ 0.05.

Synergy statistics

Combination drug experiments with G-5555 and PLX4032 or MK2206 were analyzed using the two-stage response surface models proposed by Zhao et al (162) building on the work of Harbon et al (163). This modeling approach is based on the

Loewe additivity model (164). The Zhao response surface models estimate a drug interaction index, τ (tau), for each non-zero dose combination. The interpretation of τ is as follows:

< 1 synergy, τ {= 1 additivity, > 1 antagonism

Confidence intervals for the estimated τ indices were constructed using non- parametric bootstrapping techniques. Model parameterizations other than the Zhao model were considered: uniform, linear dependency, separate indices without the response

39 surface relationship. AIC and BIC model fit criteria were compared between models and the Zhao model consistently had the best fit. Analyses were performed using the R package “drugCombo” (165).

2.4 Results

2.4.1 Group I PAK inhibition reduces thyroid cancer cell viability

The effects of group I PAK inhibition on thyroid cancer cell viability were determined in cell culture models. A panel of validated human thyroid cancer cell lines with activated MAPK and/or AKT pathways (detailed in methods section and Table B.1) were treated with increasing doses of G-5555 or FRAX1036 and cell viability was estimated after 72 hours using a water-soluble tetrazolium (WST) reagent (Fig. 2.1). IC50 doses were calculated (Table 2.1). In all cell lines tested, group I PAK inhibition reduced cell viability and, even in the relatively resistant FTC133 cell line, G-5555 was more potent than FRAX1036. The most sensitive cell line was TPC1, with an IC50 for G-5555 of 0.50 µM and for FRAX1036 of 3.94 µM. Most cell lines tested, with the exception of

TPC1, hTh74, and BCPAP, did not reach a calculable IC50 for FRAX1036, but had IC50 values below 7 µM for G-5555. FTC133 has an inactivating mutation in and loss of

PTEN (152, 154); to determine whether PI3K activation corresponded with PAK inhibitor resistance, we also treated the THJ-16T cell line that has an activating PIK3CA mutation (160). These cells were sensitive to G-5555 with an IC50 of 2.57 µM. These data suggest that the resistance is likely cell line-specific rather than due to activated AKT. In all cell lines tested, we confirmed that G-5555 and FRAX1036 reduced PAK signaling in a dose dependent manner by western blot (Fig. 2.2). PAK1 and PAK2

40 autophosphorylation at serine 144 and serine 141, respectively, showed a dose-dependent reduction, and downstream specific phosphorylation sites of PAK, vimentin at serine 55 and MEK1 at serine 298, showed a parallel decrease. There was no inhibition of phosphorylated AKT, consistent with the known kinase-independent role of PAK as a scaffold in AKT signaling (101). Overall, these data indicate that G-5555 and FRAX1036 inhibit PAK activity, that G-5555 is more potent than FRAX1036, and that in all but one tested thyroid cancer cell lines, the compounds are active in vitro.

Figure 2.1. Effects of group I PAK inhibition on cell viability. A panel of eight thyroid cancer cell lines was treated with increasing doses of G-5555 or FRAX1036 for 72 hours. Cell viability was determined by colorimetric WST-8 assay. The optical density of each treatment was normalized to the 0 µM control. Assays were conducted in triplicate with at least three biological replicates. Data are represented as means ± SD.

Table 2.1. Estimated IC50 concentrations (µM) for G-5555 and FRAX1036 in human thyroid cancer cell lines. The IC50 for each cell line and each drug were estimated using the viability data represented in Fig. 2.1. Significant differences between G- 5555 and FRAX1036 were calculated by inverse variance weighted t-tests when the IC50 was calculable and by linear mixed models of dose-response slopes if the IC50 was >10 µM. ***P ≤ 0.001, **P ≤ 0.01.

Cell Line G-5555 (µM) Std Err 95% CI FRAX1036 (µM) Std Err 95% CI TPC1 0.50*** 0.04 (0.41, 0.59) 3.94 0.36 (3.08, 4.80) BCPAP 2.46*** 0.49 (1.31, 3.60) 7.73 0.31 (6.96, 8.49) HTh74 3.36*** 0.25 (2.72, 4.01) 6.14 0.35 (5.17, 7.10) SW1736 1.37*** 0.10 (1.12, 1.62) >10 8505C 6.64*** 0.54 (5.40, 7.88) >10 K1 6.98*** 0.65 (5.32, 8.65) >10 FTC133 >10** >10 THJ-16T 2.60*** 0.40 (1.83, 3.38) >10

Figure 2.2. PAK signaling with group I PAK inhibition. TPC1, SW1736, K1, and FTC133 cells were treated with increasing doses of G-5555 or FRAX1036 for 24 hours to avoid significant toxicity. Western blots of downstream PAK targets were conducted to confirm drug activity. GAPDH was used as a loading control and is shown below blots on the same membrane. For SW1736 G-5555, pAKT S473 was blotted on the same membrane as the first GAPDH and is grouped accordingly. At least two biological replicates were performed per cell line and results were similar.

2.4.2 G-5555 in combination with BRAFV600E or AKT inhibitors synergistically reduces K1 and SW1736 cancer cell viability

We next tested the effects of combining G-5555 with an FDA-approved

BRAFV600E inhibitor, Vemurafenib (PLX4032), or an allosteric AKT inhibitor, MK2206

(166, 167). Six cell lines were tested for sensitivity to individual treatment with PLX4032

(Fig. B.1A, Table B.2) or MK2206 (Fig. B.1B, Table B.2) and showed varied responses in viability. In general, the cell lines with BRAFV600E mutation tended to be more sensitive to PLX4032 treatment, as previously reported (168), while in our conditions, genotype seemed to have less of a predictive effect for MK2206 sensitivity.

K1, SW1736, TPC1, and FTC133 cells were chosen for subsequent combination treatment, as they are representative of the most common genotypic alterations in thyroid cancer. In two cell lines with the BRAFV600E mutation, K1 and SW1736, combination of

PLX4032 and G-5555 synergistically decreased cell viability (Fig. 2.3A). In K1 cells, synergy was detected between 1 and 5 µM G-5555 with 0.5 – 5 µM PLX4032, indicated by the shaded combinations, with the darker shading being more synergistic (Fig 2.3B).

In SW1736 cells, synergy was detected mainly between 0.25 – 1 µM G-5555 with 0.1 – 5

µM PLX4032. Confirmation of decrease in PAK and MAPK activity was analyzed by western blot (Fig 2.3C). Cells without BRAFV600E mutation had mainly antagonistic effects with the PLX4032 and G-5555 combination, both in cell viability and signaling by western blot (Fig. B.2), as one might predict by the known increase in MAPK signaling that occurs with treatment of wild-type BRAF cells with PLX4032 (57, 63).

Figure 2.3. Combination treatment with BRAFV600E and PAK inhibitors in BRAFV600E cell lines. (A) WST-8 cell viability assay of K1 and SW1736 after 72 hours of combined BRAFV600E (PLX4032) and PAK (G-5555) inhibition at the specified combination of doses. All dose combinations were normalized to the 0 µM G-5555 + 0 µM PLX4032 control. Assays were conducted in triplicate with at least three biological replicates. Data are represented as means ± SD. (B) Synergy tables representing each PLX4032/G-5555 combination using the model from Zhao et al (162). The bolded top numbers are the drug interaction indices, τ, and the bottom numbers in parentheses are the confidence intervals. Synergy is defined as τ < 1 and upper confidence limit < 1. The darker shading indicates stronger synergy. (C) Western blot of PAK signaling (pPAK S144/S141) and BRAF signaling (pERK T202/Y204) at doses and timing selected for synergy and cell viability. 45

GAPDH is used as the loading control for the membrane above it. At least two biological replicates were performed per cell line and results were similar.

K1 cells also had a synergistic decrease in cell viability with combination of G-

5555 and the AKT inhibitor, MK2206, as one would predict since this cell line also has a

PIK3CA activating mutation (Fig. 2.4A, B). In SW1736 cells, there were very mild synergistic effects mostly at higher doses of MK2206, consistent with the absence of activation of PI3K signaling in this cell line (Fig. 2.4A, B). In both cell lines, decrease of pPAK S144/141 with addition of G-5555 and pAKT S473 and pGSK-3α/β S21/9 with

MK2206 were identified by western blot, confirming activity of the compounds (Fig.

2.4C). pPAK S144/141 also moderately decreased with MK2206 alone, but the exact mechanism remains to be elucidated further. In TPC1 and FTC133, combination of G-

5555 and MK2206 had minimal or no synergy, with some mild antagonistic combinations (Fig B.3). Taken together, these results indicate combination of G-5555 with a BRAFV600E or AKT inhibitor can have synergistic effects on decreasing cell viability in a subset of thyroid cancer cell types that appear to be predicted, in part, by driver oncogenes.

Figure 2.4. Combination treatment with AKT and PAK inhibitors in BRAFV600E cell lines. (A) WST-8 cell viability assay of K1 and SW1736 after 72 hours of combined AKT (MK2206) and PAK (G-5555) inhibition. All dose combinations were normalized to the 0 µM G-5555 + 0 µM MK2206 control. Experiments were conducted in triplicate with ≥ three biological replicates. Data are represented as means ± SD. (B) Synergy tables representing MK2206/G-5555 combinations as per Zhao et al (162) and as described in Fig. 2.3. (C) pPAK S144/S141, pAKT S473, and pGSK-3αβ S21/9 are shown at doses and timing selected for synergy and cell viability. GAPDH is loading control for each membrane above it. At least two biological replicates were performed per cell line with similar results. 47

2.4.3 Reduction in cell viability by group I PAK inhibition corresponds with cell cycle arrest in G0/G1.

PAKs activate key regulators of cell cycle progression, including histone H3

(103), GIT1 (105), Aurora-A (105), and PLK1 (106). Thus, we analyzed the effects of group I PAK inhibition on cell cycle using propidium iodide and flow cytometry. Based on initial time course and dose response experiments confirming inhibition of PAK activity and cell response on western blot and WST assays, TPC1, SW1736, and FTC133 were treated with 0.5 μM G-5555 or 5 μM FRAX1036 for 24 (Fig. 2.5A) or 48 hours

(Fig. 2.5B). In TPC1, the cell line that was most sensitive to growth inhibition by G-5555 and FRAX1036 (Fig. 2.1), both compounds induced arrest in G0/G1 phase at 24 hours, with concordant reductions of cells in S and G2/M (Fig. 2.5A). These effects remained after 48 hours, although with more variability (Fig. 2.5B). Similarly and consistent with cell viability effects (Fig. 2.1), the accumulation of SW1736 in G0/G1 after 24 and 48 hours was greater for G-5555 than FRAX1036 (Fig. 2.5). Conversely, in FTC133, the cell line that was mostly resistant to growth inhibition by both G-5555 and FRAX1036

(Fig. 2.1), neither compound altered cell cycle progression at either 24 or 48 hours (Fig.

2.5). Therefore, these data suggest that for cells whose viability is reduced with group I

PAK inhibition, there is a concomitant G0/G1 arrest.

Figure 2.5. Cell cycle effects with group I PAK inhibition. TPC1, SW1736, and FTC133 were treated with DMSO, 0.5 µM G-5555, or 5 µM FRAX1036 for (A) 24 or (B) 48 hours. Cells were stained with propidium iodide and analyzed by flow cytometry for relative DNA amount. Statistical comparisons were made within each phase of the cell cycle by one-way ANOVA followed by Holm’s procedure. Three biological replicates for each cell line and time point were performed and data are represented as means + SD. ****P ≤ 0.0001, ***P ≤ 0.001, **P ≤ 0.01, *P ≤ 0.05.

2.4.4 G-5555 reduces thyroid cancer cell migration and invasion

In addition to cell proliferation, group I PAKs are important for the rearrangement of the cytoskeleton and are key regulators of cell migration and invasion (107, 169-171).

We previously reported increased PAK activity at the invasive fronts of aggressive PTCs and that molecular inhibition of PAK1 reduced thyroid cancer migration (127). In addition, PAK inhibition by G-5555 reduced phosphorylated vimentin (Fig. 2.2), suggesting that PAK regulation of intermediate filaments might be reduced by the

49 compound. Therefore, we tested the ability of G-5555 to reduce TPC1 and FTC133 migration and invasion in transwell assays without (migration) or with (invasion)

Matrigel coating. Cells were seeded on top of a Boyden chamber and treated with G-5555 for 16 hours. The chambers were stained with crystal violet and the extracted color from the stained cells was compared between the top (non-migrated) and bottom (migrated) layers, and images of the invading cells were quantified. In both TPC1 and FTC133, G-

5555 reduced migration (Fig. 2.6A). In TPC1, 0.25 µM significantly reduced migration to ~50% and 1 µM G-5555 had even further reduction to ~24% migration. An ~47% decrease in migration for FTC133 occurred with 1 µM G-5555. Additionally, G-5555 was effective in reducing both TPC1 and FTC133 invasion (Fig. 2.6B, C), with similar trends in sensitivity as in the migration assays. Cell viability was tested in parallel via

WST assay to confirm no difference in viability in the conditions of the migration and invasion experiments (Fig. 2.6D). Together, these data indicate that G-5555 reduces thyroid cancer cell motility and invasiveness at sub-lethal doses, even in cells that are resistant to its cytotoxic effects at higher doses.

Figure 2.6. Effects of G-5555 on cell migration and invasion. TPC1 and FTC133 were seeded on top of Boyden chamber inserts without (A) or with (B) Matrigel coating. Cells were treated for 16 hours with 0, 0.25, or 1 µM G-5555. The top and bottom of the inserts were stained with crystal violet. For migration (A), percent migrated was calculated by comparing the OD of extracted crystal violet from the top and the bottom of the insert. For effects on invasion (B), the fraction of the control (i.e. 0 µM G-5555) was calculated after counting the number of purple pixels from images of the bottom layer. (C) Representative images of invading cells are shown. (D) WST-8 viability assays were conducted in parallel and indicate no cytotoxic effects in these conditions. Cells were seeded in duplicate and at least three biological replicates for each cell line were performed. Data are represented as means + SD. ****P ≤ 0.0001, ***P ≤ 0.001, **P ≤ 0.01, *P ≤ 0.05.

2.4.5 BRAFV600E-increased thyroid size and cancer development in vivo is inhibited by G-5555

The in vivo effects of G-5555 have been reported only in xenograft models of non-small cell lung cancer, breast cancer, and colon cancer (130, 147, 148). In these studies, G-5555 was effective at reducing xenograft tumor growth. To our knowledge, it has not been studied in a transgenic, immune-competent cancer model. In addition, the impact of PAK signaling on BRAFV600E effects has not been studied directly in vivo in transgenic model systems. Therefore, we determined the effects of G-5555 in an immune- competent thyroid cancer model in vivo. For these experiments, we utilized the robust and highly reproducible thyroid-specific inducible BRAFV600E overexpression mouse model of thyroid cancer (161), in which acute induction of BRAFV600E in the mouse thyroid leads to human-like PTC in one week in nearly all mice. We previously reported concurrent PAK upregulation and activation using this model (149). G-5555 or vehicle control was given orally twice a day for a week beginning on the same day as BRAFV600E induction. Remarkably, in mice harboring BRAFV600E tumors, treatment with G-5555 resulted in smaller thyroids by 52.1% compared to vehicle controls (mean volume 1.902 vs. 3.97 mm3, p<0.0001) (Fig. 2.7A). G-5555 did not have an effect on the size of non- induced thyroids compared to vehicle controls (0.591 vs. 0.429 mm3, p=0.8362).

Histopathological analysis confirmed that mice without BRAFV600E induction (no doxycycline), exhibited normal thyroid histology (Fig. 2.7B, C). In mice with BRAFV600E induction, 15 mice treated with vehicle developed thyroid carcinomas and only one mouse had thyroid hyperplasia. Surprisingly, for those mice with BRAFV600E and treated

52 with G-5555, six had hyperplasia, two formed follicular adenoma, and only nine developed carcinoma. When comparing the incidence of benign pathology and carcinoma in the mice with BRAFV600E, G-5555 significantly reduced carcinoma formation

(p=0.0167). Ki67 analysis in these mice indicated a mild trend towards a decrease in thyroid cell proliferation with G-5555 treatment (p=0.1078), consistent with the higher incidence of benign thyroid glands (Fig. B.4A). All thyroids, regardless of treatment, had less than 1% of cells with positive cytosolic staining for cleaved caspase-3 (Fig. B.4B).

The difference in thyroid size in mice with BRAFV600E and G-5555 treatment is consistent with the in vitro observation that G-5555 inhibits cell cycle progression (Fig.

2.5) and likely not due to an increase in apoptosis.

Induction of BRAFV600E in the thyrocytes resulted in activation of ERK1/2, localized nearly exclusively to the thyrocytes (Fig. 2.7D), consistent with prior observation using these mice (161). The available pPAK antibodies were unreliable for immunohistochemical analysis. Protein from individual mouse thyroids were isolated and analyzed by western blot and compared to the thyroid volume. pERK levels did not correspond with thyroid size (Fig. 2.7E, Fig. B.5); however, pPAK144/141 levels positively associated with thyroid size (Fig. 2.7E, F). Taken together, these data show that G-5555 restrains BRAFV600E-driven thyroid size and carcinoma formation, even with continued activation of MAPK signaling, and suggest that active PAK contributes to tumorigenesis in this mouse model of thyroid cancer.

Figure 2.7. In vivo effects of G-5555 in a BRAFV600E-inducible thyroid cancer mouse model. BRAFV600E was induced in mouse thyroids by doxycycline-enriched chow for one week at the same time as G-5555 or vehicle oral gavage. Thyroids were harvested and volume (A) of one lobe was measured ex vivo. Comparisons among treatments were made by one-way ANOVA followed by Holm’s procedure. Data are represented as individual plots with means ± SD. ****P ≤ 0.0001 (B) Representative H&E images of the thyroid pathologies observed. Scale bar = 100 µm. (C) Number of mice in each treatment group with the observed pathologies. Carcinoma incidence between Dox/Vehicle and Dox/G- 5555 were compared using a Fisher’s exact test. *P ≤ 0.05. (D) Representative pERK1/2 T202/Y204 IHC images showing induction of BRAFV600E in the thyrocytes with Dox. Scale bar = 100 µm. (E) Western blot of Cohort 1 mouse thyroid lysates. Each lane is lysate from an individual mouse treated with Dox and Vehicle or G-5555, as indicated. GAPDH is used as the loading control for the membrane above it. (F) Comparison of thyroid volume and quantification of the pPAK1/2 S144/141 western blot bands normalized to GAPDH for each western blot membrane. Each cohort of mice was treated 54 at different times and the lysates from mice in Cohort 2 were divided on two gels (a and b), as titled. Comparisons were fit with a linear regression and the r2 values are shown.

2.5 Discussion

Alternative treatment strategies for aggressive thyroid cancers are needed to improve clinical outcomes. Our previous studies suggested an important functional role for activated PAK1 in aggressive PTC and showed that PAK activation is functionally regulated by BRAFV600E in a MEK-independent manner using siRNA and molecular inhibitor approaches (65, 127, 149). In the present study, we extended on our previous findings and identified that pharmacologic inhibition of group I PAKs may be useful as both a monotherapy (Fig. 2.1) and in combination with BRAFV600E (Fig. 2.3) and AKT

(Fig. 2.4) inhibition at concentrations likely within the G-5555 therapeutic window in vivo [Fig. 2.7, (147)]. IC50 concentrations and the concentrations required to inhibit PAK were lower for G-5555 than FRAX1036 (Table 2.1, Fig. 2.2), consistent with the known improved relative target selectivity and potency of this compound compared to

FRAX1036 (147).

As might be predicted by the known pathways of PAK activation, cell genotype for MAPK and PI3K activation did not entirely predict the degree of sensitivity to G-

5555 or FRAX1036 monotherapy. Indeed, there was variability in sensitivity of the wild- type BRAF cells lines, the BRAFV600E mutated cell lines, and the AKT activated cell lines

(Table 2.1). In all cell lines tested, however, both G-5555 and FRAX1036 reduced PAK signaling (Fig. 2.2). Thus, these data suggest intrinsic resistance to PAK inhibitors for

55 viability of some cell lines is not due to lack of PAK inhibition by these compounds and requires further mechanistic studies.

Our data show that BRAFV600E also regulates PAK, that PAK is activated by several known mechanisms of BRAF inhibitor resistance [e.g. RAC1 mutations (172)], and that AKT and PAK functionally interact. Therefore, we reasoned that PAK inhibition might enhance the efficacy of compounds specifically targeting those pathways. Indeed, two BRAFV600E-mutated cell lines, K1 and SW1736, exhibited synergistic decreases in cell viability in response to G-5555 and PLX4032 combination therapy (Fig. 2.3), while only K1, which also harbors a PIK3CA mutation, demonstrated synergy with MK2206.

These findings suggest that PAK inhibitors may be a potential combinatorial strategy for

BRAF-mutated but treatment resistant PTC (57-60). In addition to BRAF-mediated PAK activity (149), it has also been demonstrated that PAK can be activated downstream of receptor tyrosine kinases, including HER3 or c-MET, that are involved in BRAFV600E resistance. For example, group I PAKs are an important node in the regulation of c-MET signaling by inactivating the negative regulator, merlin, via phosphorylation of merlin at

S518 (173, 174). Shrestha et al. describe that in breast cancer, PAK1 is amplified and further upregulates MAPK and MET signaling (175). In vitro and in a murine model of

BRAFV600E-induced PTC, resistance to BRAFV600E inhibition was reported to occur by c-

MET upregulation (58-60). Thus, it is possible that combination therapy with a group I

PAK inhibitor, such as G-5555, could block PAK phosphorylation of merlin, and subsequently allow merlin to negatively regulate c-MET. Further studies to determine the

56 effects of G-5555 and PLX4032 combination in vivo are needed, as this could be a clinically important approach.

Cells that were sensitive to FRAX1036 or G-5555 monotherapy for cell viability demonstrated G0/G1 arrest via flow cytometry (Fig. 2.5). These results are consistent with observed cell cycle changes with group I PAK inhibition in meningioma cells (114).

However, even in cells that survived PAK inhibition (FTC133, Fig. 2.1), G-5555 still reduced cell migration and invasion (Fig. 2.6), suggesting multiple roles for PAK in thyroid cancer cells. Consistent with this observation are the reductions in phosphorylated vimentin (Fig. 2.2) and our previous work that showed PAK-dependent cell migration using molecular inhibitors (127). Therefore, treatment with G-5555 could be a viable strategy to block the aggressive local invasion of larger thyroid cancers (127).

Importantly, we tested the effect of PAK inhibition in a transgenic model of

BRAFV600E-mediated PTC. This model was chosen due to its robust MAPK signaling effects, the highly predictable development of PTC, and because we previously reported increased PAK expression and activity in PTCs that develop in this model (149). In addition, we opted for a short-term model to avoid potential cardiac toxicity that might occur with long-term PAK inhibition in the mice (176-178). It is striking that even with such strong thyrocyte-specific activation of MAPK, inhibition of PAK signaling significantly restrained thyroid size and also reduced thyroid tumorigenesis. Further, levels of pPAK positively corresponded with thyroid size, while levels of pERK did not

(Fig. 2.7F, Fig. B.5). Taken together, these data support the prior observations from our group and others (150) of a MEK-independent role of BRAF-mediated PAK activation,

57 suggesting that both PAK and ERK activation are required for maximal BRAFV600E- effects in the thyroid. The latter hypothesis requires further studies using thyrocyte- specific approaches for confirmation.

There are several caveats to our in vivo approach. First, we recognize that further testing using a PTC model with physiological expression of BRAFV600E is optimal.

Second, G-5555 is a systemic therapy; therefore, this model does not differentiate between thyrocyte and non-thyrocyte involvement of PAK. Although G-5555 is a specific and potent PAK inhibitor and we demonstrated on-target effects, off-target effects can still occur. Third, we have not yet studied a treatment model in vivo, as we treated mice with G-5555 and doxycycline simultaneously. However, studies with delayed treatment are challenging as the inducible BRAFV600E allele in this model is

“turned-off” after a few weeks due to thyroid dedifferentiation. Despite these important caveats, we believe the striking effects of the PAK inhibitor on a potent model of

BRAFV600E-induced tumor formation, in the context of continued activation of MAPK signaling, are novel and consistent with our prior non-pharmacological data. Together, these data strongly implicate a critical role for PAK in BRAFV600E-induced thyroid tumorigenesis.

In summary, we have shown that group I PAKs are critical signaling molecules in thyroid growth and tumorigenesis and may be therapeutically targetable in several genetic contexts. PAKs are involved in several of the “Hallmarks of Cancer” (128, 179), are activated in aggressive fronts of thyroid cancer (127), and are functionally important for thyroid cancer growth and invasiveness in vitro and in vivo. Further investigation into

58 the safety and dosing strategy of G-5555 alone and in combination with BRAFV600E or

AKT inhibition is needed using additional mouse models of thyroid cancer, as this may lead to alternative treatment options for thyroid cancer patients who have cancers that are resistant to current therapies, and perhaps extend to patients with other BRAFV600E- mediated tumor types.

Chapter 3: Characterization of the BRAF-PAK1 Complex

3.1 Abstract

The MAPK pathway is frequently activated in thyroid cancers, most often from constitutively active BRAFV600E, but also by receptor tyrosine kinase fusions, such as

RET/PTC1, and RAS mutations. We recently discovered increased PAK1 expression and activity in the invasive edges of aggressive thyroid cancers and determined that BRAF regulated PAK1 activity (127, 149). Wild-type (WT) BRAF could immunoprecipitate with PAK1, but it was unknown whether or not BRAFV600E also interacts and what other proteins might be present in the BRAF-PAK1 complex. Therefore, we used biochemical approaches to test the hypotheses that BRAFV600E physically interacts with PAK1, and that BRAF and PAK1 interact as a complex with other proteins to guide BRAF-PAK1 function. We report that BRAFV600E physically interacts with PAK1 and that this interaction is enhanced in thyroid cancer cells during mitosis. We further characterized the WT BRAF-PAK1 interaction and determined that PAK1 kinase activity is dispensable for its interaction with BRAF and the interaction likely is dependent on

PAK1 amino acids 121-140. BRAF and PAK1 likely interact as a larger protein complex, potentially mediated by the CCT complex or HSP90, and may be involved in cell cycle regulation by interacting with chromosome-related microtubules. The complex is likely transient and in low abundance in thyroid cancer cells, characteristic of a regulated process involving signal transduction. Taken together, these data suggest that both WT 60

BRAF and BRAFV600E interact dynamically with PAK1 in thyroid cancer as part of a larger protein complex and that defining this complex may be important for identifying novel therapeutic strategies.

3.2 Introduction

The diversity of individual protein functions is rooted in the ability of those proteins to interact with many different proteins. Identifying protein interactomes aids in understanding its roles in various signaling pathways and cellular functions. This is important not only to understand cellular complexities, but also to identify potential novel targets for therapeutic advances. Many common methods for identifying protein-protein interactions (PPIs) have been developed over the years, such as co-immunoprecipitation

(co-IP), Förster resonance energy transfer (FRET), and chromatography methods ((180), reviewed in (181, 182). Each has varying degrees of sensitivity and limitations for use in vitro or in living cells. Two alternative techniques, utilized in this study, are based on the identification of protein interactions that occur within proximity to the protein of interest: in situ proximity ligation assay (PLA) and proximity-dependent biotin identification

(BioID). PLA was originally designed to detect proteins in solution using a DNA aptamer to amplify the signal (183). Further modifications adapted the technique to in situ analysis, allowing the visual identification of individual proteins or PPIs by fluorescence- labeled oligonucleotides (184). Primary antibodies (of different species of origin) specific for two proteins of interest are identified by secondary antibodies/probes with connector nucleotides that have free 5’ or 3’ ends. These probes are complementary and hybridize if they are within 40 nm. This DNA sequence is amplified by rolling circle amplification

61 and the addition of fluorescent oligonucleotides allow for the visualization of the interaction. In situ PLA can be combined with traditional immunofluorescence assays to identify the localization of the interaction (185). The PLA technique is attractive for identifying low-abundance protein interactions, as the signal is amplified before detection. However, this is limited by the availability of specific and usable primary antibodies.

The proximity-dependent biotin identification (BioID) method also identifies the proximal interactome without the dependence on antibody:protein interactions or maintained PPI. BioID utilizes a bacterial biotin ligase (BirA or BioID and the improved

BioID2), specifically mutated to promiscuously biotinylate proteins within 10 nm of itself

(186, 187). Biotin added to the media of cells expressing fusion proteins of BioID provides material for BioID to biotinylate proteins interacting and in proximity to the fusion protein of interest. Even if interactions are transient or lysis conditions are harsh, the biotin label will remain intact on interacting proteins. This allows the identification of the biotinylated proteins after capture using streptavidin-conjugated beads and subsequent mass spectrometry and/or western blot. The BioID approach is appealing for its application to transient or weak PPIs; however, there are concerns with determining true interactions and proper negative controls, which will be discussed further in Section

3.5.

Diedrich et al recently described the BRAF interactome in WT and BRAFV600E conditions, primarily in colon cancer cells, and also validated in BrafV600E knock-in

MEFs, using SILAC labeling (stable isotope labeling with amino acids in cell culture)

62 followed by size exclusion chromatography and mass spectrometry (188). Complex size and protein preference differed depending on the Braf mutational status. BrafV600E tended to interact more with HSP90/CDC37 and in complexes of larger molecular weights, while WT Braf interacted stronger with some of the 14-3-3 proteins and in smaller complexes. Additionally, treatment with pharmacologic inhibitors of HSP90, BRAFV600E, and MEK altered the interaction profile of these target proteins. While this technique provided a broad, unbiased approach to identifying protein complexes, limitations include detecting low abundance and transient interactions.

Mayhew et al studied PAK1-associated complexes by transfecting HEK293 cells with FLAG-tagged PAK1, followed by immunoprecipitation (IP) and mass spectrometry, with a focus on common interacting partners with βPIX and GIT1 (189). Particular

PAK1-interacting proteins detected were the 14-3-3 proteins, tubulins, and actin, among others. BRAF or other known PAK1 interactors in the MAPK pathway, such as CRAF and MEK1, were not detected. Although this study identifies many proteins likely involved in the PAK1-βPIX-GIT1 complex and highlights proteins independently interacting with PAK1, βPIX, and GIT1, gaps remain in the breadth of the PAK1 interactome, particularly in relation to MAPK signaling. Nonetheless, such studies highlight the value in identifying PPI networks in different cancer cell types and emphasize the need to perform multiple PPI detection methods in order to obtain high- confidence results.

We recently identified by co-immunoprecipitation (co-IP) and immunofluorescence (IF) that wild-type (WT) BRAF and PAK1 interact in thyroid

63 cancer cells, most apparently in mitosis (149). However, other proteins involved in this interaction, as well as the structural requirements (i.e. BRAF mutation to BRAFV600E and domains of PAK1), remained undefined. This work utilized several PPI techniques to test the hypothesis that both WT BRAF and BRAFV600E can interact with PAK1 and that this interaction is mediated by other proteins in the complex to facilitate BRAF-PAK1 function. Co-IP and PLA studies determined that PAK1 can interact with BRAF, regardless of its mutational status, and that this interaction does not require PAK1 kinase function. Proteomic analysis and subsequent confirmatory tests suggest that the BRAF-

PAK1 complex may contain chaperone proteins, such as the CCT complex or HSP90.

The interaction may also be localized to microtubules, particularly during mitosis, suggesting the BRAF-PAK1 complex aids in regulating cell cycle. Finally, we acknowledge the transient nature of this complex, as it is difficult to consistently capture in endogenous thyroid cancer cells. Overall, we have further characterized an important protein complex, providing insight to its regulation and dynamic function in thyroid cancer cells.

3.3 Materials and Methods

3.3.1 Cell lines and cell culture

Human embryonic kidney (HEK293) cells were purchased from ATCC

(Manassas, VA, USA). The human thyroid cancer cells, TPC1 (153) and 8505C (157), were generous gifts of Dr. Rebecca Schweppe (University of Colorado, CO) with permission from the original researchers who established the cell lines. WT and PAK1-/-

MEFs were generous gifts of Dr. Jonathan Chernoff (Fox Chase Cancer Center, PA).

HEK293, TPC1, and the MEFs were cultured in DMEM medium with 10% fetal bovine

o serum (FBS), 1% glutamine, and 1% nonessential amino acids at 37 C and 5% CO2. The media for the MEFs also contained penicillin and streptomycin. 8505C were cultured in

RPMI medium with 10% fetal bovine serum (FBS), 1% glutamine, and 1% nonessential

o amino acids at 37 C and 5% CO2. Cell line identity was validated by DNA fingerprinting after receipt of the cells and periodically over the course of the experiments.

3.3.2 Plasmid constructs

The MYC-tagged wild-type BRAF and MYC-tagged BRAFV600E plasmids were gifts from Dr. James Fagin (Memorial Sloan-Kettering Cancer Center, NY). The FLAG- tagged full length PAK1 was generated as described previously in our lab using PCR and the restriction enzymes BamHI and HindII to insert the PAK1 sequence into a pCMV-

Tag2B vector (149). The FLAG-tagged kinase-dead PAK1K299R plasmid was a gift of Dr.

Jonathan Chernoff (Fox Chase Cancer Center, PA). The PAK1 truncation mutants were subcloned from the pCMV-Tag2B-PAK1 using PCR with AccuPrime SuperMixII

(ThermoFisher Scientific) and restriction digests. PAK1-N (amino acids 1-206) was generated by PCR using the following primers:

CK001:

5’-G GTG GCG GCC GCC ACC ATG GAT TAC AAG GAT GAC GAC GAT AAG AGC CCG GGC GGA TCC ATG TCA AAT AAC GGC CTA-3’

CK002:

5’-ATA AAG CTT TTA AAT CAC AGA CCG TGT GTA TAC -3

PAK1-C (amino acids 207-545) was generated by PCR using the following primers:

CK003:

5’- G GTG GCG GCC GCC ACC ATG GAT TAC AAG GAT GAC GAC GAT AAG AGC CCG GGC GGA TCC GAA CCA CTT CCT GTC ACT CCA-3’

CK004:

5’-GAT AAG CTT GAT ATC GAA TTC CTC GAG GCC

The PAK1-N fragment, PAK1-C fragment, and pCMV-Tag2B vector were digested with NotI and HindIII and subsequently ligated.

PAK1 amino acids 1-140 was generated by PCR using the following primers:

CK001:

5’-G GTG GCG GCC GCC ACC ATG GAT TAC AAG GAT GAC GAC GAT AAG AGC CCG GGC GGA TCC ATG TCA AAT AAC GGC CTA-3’

CK010:

5’ - ATA AAG CTT TTA CTG GCT GTT GGA TGT CTT CTT -3’

PAK1 amino acids 121-270 was generated by PCR using the following primers:

CK011:

5’- GGC GGA TCC CCG CAG GCT GTT CTG G – 3’

CK012:

5’- ATA AAG CTT TTA ATA TTT CTT CTT AGG ATC GCC -3’

The 1-140, 121-270, and pCMV-Tag2B vector were digested with BamHI and

HindIII and subsequently ligated.

The FLAG-TEV-PAK1 construct was made by annealing the following two oligos, which contain both the FLAG sequence and TEV recognition sequence and have overhangs matching NotI and BanHI digestion products:

CK005:

5’ G GCC GCC ACC ATG GAT TAC AAG GAT GAC GAC GAT AAG GAG AAC CTC TAC TTC CAA TCG GGA GGA GGA GGA GGA GGA-3’

CK006:

5’ GA TCC TCC TCC TCC TCC TCC TCC CGA TTG GAA GTA GAG GTT CTC CTT ATC GTC GTC ATC CTT GTA ATC CAT GGT GGC-3’

The oligos were diluted in annealing buffer (0.25 M NaCl, 10 mM Tris-HCl pH7.5, 1 mM EDTA), boiled for 5 minutes, and then cooled to room temperature. The pCMV-Tag2B-PAK1 was digested with NotI and BamHI, purified to remove the original

FLAG sequence, and ligated to the FLAG-TEV oligo.

The MYC-BioID2-PAK1 construct was made using the MYC-BioID2-MCS construct purchased from Addgene (#74223). The following primers were used to add a base pair before the PAK1 cDNA sequence to make it in-frame after digestion with

BamHI and HindIII, and ligation into the BioID2 vector:

CK015:

5’-TCA G GAT CCT TCA AAT AAC GGC CTA GA -3’

CK016:

5’- GCG AAG CTT TTA GTG ATT GTT CTT TGT TGC -3’

3.3.3 In situ Proximity Ligation Assay (PLA) and immunofluorescence (IF)

For both PLA and IF, cells were seeded on coverslips for 24 hours in normal growth media. They were washed twice in PBS and fixed with 4% paraformaldehyde for

10 minutes at room temperature. Following two washes in PBS, they were then semi- permeabilized with 0.2% Triton-X in PBS for 15 minutes at room temperature. The coverslips were washed twice and incubated in blocking serum (Vectastain Universal

Quick Kit) for 30 minutes at room temperature. Primary antibodies were diluted in PBS and added to the coverslips for overnight incubation in a humidified chamber at 4oC. For

IF-only assays, the coverslips were washed and incubated with secondary antibodies conjugated with Alexa Fluor 488 or 594 (Life Technologies/ Thermo Fisher Scientific) diluted in blocking serum for 2 hours at room temperature. Coverslips were counterstained with Duolink In Situ Mounting Medium with DAPI from the Duolink

PLA kit (Sigma) and mounted on slides.

For PLA, after incubation with the primary antibodies, the assay was followed per the manufacturer’s protocol (Duolink, Sigma). Briefly, PLA anti-rabbit PLUS and anti- mouse MINUS probes were added for 1 hour at 37oC in a humidified chamber.

Coverslips were washed and the ligation mixture was added for 30 min at 37oC in a humidified chamber. After brief washes, the amplification mixture with the polymerase and fluorescent probes were added. If IF was being done with the PLA, at this step, the IF secondary antibody was also added. Coverslips were incubated in the dark for 100 minutes at 37oC in a humidified chamber. They were then washed and let dried before being mounted on coverslips with the Duolink In Situ Mounting Medium with DAPI.

Slides were imaged with a Zeiss AxioCam MRc5 camera from a Zeiss Axiovert 40 CFL microscope at 40X magnification. Images were acquired using the AxioVision SE64 software. At least 5 randomly selected pictures were taken per sample.

Primary antibodies used for PLA or IF are as follows: MYC (1:100, Cell

Signaling, 2272), FLAG (1:100, Sigma, F3165), BRAF (1:50, Santa Cruz, sc-5284),

BRAF (1:400, Santa Cruz, sc-9002), PAK1 (1:50 [1:200 for α-tubulin PLA], Cell

Signaling, 2602), CCT1 (1:50, abcam, ab109126), CCT3 (1:50, Bethyl Laboratories,

A303-459A-M), CCT5 (1:50, abcam, ab129016), α-tubulin (PLA, 1:50, Santa Cruz, sc-

5286), α-tubulin (IF, 1:100, Santa Cruz, sc-53029), pH3 S10 (1:500, Millipore Sigma,

MABE939)

3.3.4 Immunoprecipitation (IP) and western blots

For exogenous immunoprecipitation (IP) in HEK293 cells, HEK293 cells were co-transfected in 10cm plates with 3 µg each of MYC-BRAFWT or MYC-BRAFV600E and

FLAG-PAK1 using Optifect (Thermo Fisher Scientific) in OptiMEM (Thermo Fisher

Scientific) for 4 hours. The cells were then incubated in normal growth media for 24 hours. Prior to harvesting, cells were crosslinked with 1% formalin added directly to the media at room temperature for 13 minutes. The reaction was quenched with direct addition of cold 1.25M glycine diluted in PBS and incubated on ice for 5 minutes. The solution was aspirated and the cells washed with cold PBS. Cells were scraped and collected in 1 mL PBS, then an additional 1 mL PBS was added to the plate and scraped a second time to remove any cells that may be loosely crosslinked to the plate. Cells were then pelleted and lysed in Pierce IP Lysis Buffer (Thermo Fisher Scientific). Protein concentration was estimated using bovine serum albumin as a standard and the Pierce

BCA Protein Assay Kit (Thermo Fisher Scientific). 370 µg of protein at a 0.5 µg/µL concentration were used for each IP. EZview Red anti-c-MYC (Millipore Sigma, E6654), anti-FLAG (Millipore Sigma, F2426), or anti-HA (Millipore Sigma, E6779) affinity gel/beads was added individually to the protein at a 1:10 dilution and the samples rotated overnight at 4o. HA beads are used as the negative control for the MYC and FLAG pre- conjugated beads, as there is no protein expressed with the HA tag. After the incubation, the samples were centrifuged and the supernatant was collected as the “unbound”

69 fraction. The beads were washed three times with Pierce IP Lysis Buffer (Thermo Fisher

Scientific) and pelleted. The pelleted beads were resuspended at 1/4 the incubation volume in 1x NuPAGE LDS sample buffer (Thermo Fisher Scientific), vortexed, and boiled for 5 minutes. The samples were then cooled on ice and centrifuged at 16,000 g for

5 minutes to elute the protein. Soluble protein was collected and separated by western blot as detailed in Section 2.3.4.

For endogenous IP in TPC1, cells were synchronized in mitosis as detailed in

Section 3.3.5. Following release, cells were washed and scraped in cold PBS. Cells were centrifuged at 800 g for 10 min. Cell pellets were resuspended and lysed in a modified

Pierce IP Lysis Buffer (Thermo Fisher Scientific), 25 mM Tris pH 7.4, 150 mM NaCl, 1 mM EDTA pH 8, 0.1% NP-40/IGEPAL, 5% glycerol and incubated on ice for 5 minutes.

The lysates were then centrifuged at 16,000 g for 10 minutes and the supernatants were collected. Protein concentration was estimated as described above. Lysates for IP were pre-cleared to remove non-specific binding for 1 hour at 4o using washed Protein G agarose beads in a 50% slurry (Millipore Sigma, 16-266). An aliquot was saved as the input. 1 mg of pre-cleared protein at 1 mg/mL concentration (diluted in Pierce IP Lysis

Buffer, Thermo Fisher Scientific) was used for each IP. BRAF (Santa Cruz

Biotechnology, sc-5284) and PAK1 (Cell Signaling, 2602) antibodies were added to the 1 mg of protein at a 1:100 dilution. IgG controls from mouse (Santa Cruz, sc-2025) and rabbit (Cell Signaling, 2729) were used as negative controls for BRAF and PAK1, respectively. The proteins were rotated and incubated with the antibodies overnight at

4oC. Protein G agarose beads were added to the protein: antibody solution at a 1:10

70 volume ratio and rotated for 2 hours at 4oC. The beads were washed three times in Pierce

IP Lysis Buffer (Thermo Fisher Scientific). Following the final wash and centrifugation, pelleted beads were resuspended at 1/5 the incubation volume in 1x NuPAGE LDS sample buffer (Thermo Fisher Scientific) and protein was eluted as described above.

Western blot analysis was performed as detailed in Section 2.3.4.

Primary antibodies used for the western blots are as follows: MYC (1:1000, Cell

Signaling, 2276), FLAG (1:1000, Cell Signaling, 2368), CCT1 (1:2000, abcam, ab109126), CCT3 rab (1:1000, Bethyl Laboratories, A303-459A-M), CCT5 (1:10,000, abcam, ab129016), HSP90 (1:1000, Stressgen Biotechnology, SPA-830), BRAF (1:1000,

Santa Cruz Biotechnology, sc-5284), PAK1 (1:1000, Cell Signaling, 2602), AKT1/2/3

(1:500, Cell Signaling, 9272), CRAF (1:1000, Cell Signaling, 9422), GAPDH (1:15,000,

Cell Signaling, 2118)

3.3.5 Cell synchronization

TPC1 and 8505C cells synchronized as we previously described (149) with the following modifications: Cells were first treated with 2 mM thymidine in normal growth media for 12 hours then released for 3 hours in media without drugs. Cells were then treated with 100 ng/mL nocodazole in growth media for 12 hours and released in normal media for 1 hour before being harvested.

3.3.6 Mass spectrometry and proteomics analysis

HEK293 transfection, cell lysate preparation, and FLAG, MYC, and HA IP were conducted as described in Section 3.3.4. Three biological replicates for each IP and for each mass spectrometry analysis were conducted. For samples sent for mass spectrometry

71 analysis at OSU, in collaboration with Dr. Michael Freitas, interacting proteins were eluted in 1X sample buffer as described in Section 3.3.4. Each IP was separately analyzed by LC–MS/MS, as routinely conducted in the Freitas Lab. The eluted proteins were briefly run on SDS-PAGE gels, stained with Coomassie, and the bands were excised and cleaned. The IP products were individually digested with trypsin, followed by proteomic analysis using Orbitrap Fusion or Q-Exactive mass spectrometry. Peptide separation was performed using the EASY-Spray PepMap C18 column (3 μm, 100 Å, 0.75 × 150 mm) operated at 275°C and a spray voltage of 1.7 kV. Peptides were eluted along a linear gradient (5–28% buffer B) at a flow rate of 300 nl/min over 250 min followed by a column wash/equilibration step. Raw data were analyzed using the MultiSpec algorithm developed in the Freitas lab (190)). Log fold changes (expressed as fold change of normalized spectral counts comparing HA to MYC or FLAG normalized spectral counts) and respective q-values were used to distinguish significantly enriched proteins. Enriched proteins with negative fold changes (i.e. higher in FLAG or MYC IPs compared to HA) and q-values < 0.05 were identified as possible interactors for later validation.

An independent mass spectrometry and proteomics analysis was conducted in collaboration with Dr. Salim Merali (Temple University, Philadelphia, PA), as few co- immunoprecipitated proteins were detected with MYC-BRAF using the Freitas Lab approach. Three biological replicates of FLAG, MYC, and HA IPs were conducted as described in Section 3.3.4. with the exception that the proteins were not eluted from the beads by 1X sample buffer. Washed beads with the captured proteins were sent directly for preparation for mass spectrometry. The label-free proteomics analysis using modified

72 in-stage tip (iST) method was performed using the nanoelectrospray ionization (ESI) tandem MS with a LTQ Orbitrap Elite mass spectrometer (Thermo Fisher Scientific) as previously described (191). Specifically, proteins were digested with trypsin according to the Merali lab’s published protocols (191-197). The peptides were acidified and loaded onto an activated in-house-made cation stage tip. The peptides were purified and eluted into six fractions using elution buffers as previously described (191). The purified peptides were analyzed by mass spectrometer with complete system controlled by

Xcalibur software (Version 3.0.63). Mass spectra processing was carried out using

Mascot Distiller and Sequest HD with Proteome Discoverer 2.0. The generated de- isotoped peak list was submitted to an in-house Mascot server 2.2.07 for searching against the Swiss-Prot database. Mascot search parameters were set as follows: species, homo sapiens; enzyme, trypsin with maximal 2 missed cleavage; fixed modification, cysteine carboxymethylation; variable modification, phosphorylation; 20 ppm mass tolerance for precursor peptide ions; 0.2 Da tolerance for MS/MS fragment ions. All peptide matches were filtered using an ion score cutoff of 20. Quantified proteins were selected and clustered by biological functions.

Similar analyses were conducted using abundance ratios instead of spectral counts. Data are consistent from the Freitas Lab and the Merali Lab.

3.3.7 17-AAG treatment

TPC1 and 8505C cells were seeded on 10 well plates. 17-AAG was diluted at the indicated concentrations in growth media with 10% FBS and added to the cells for 24 hours. DMSO in growth media was used as the control. The cells were then harvested

73 and the protein was isolated and analyzed as described in Section 2.3.4 and using the antibodies listed in Section 3.3.4.

3.3.8 TEV Tandem co-IP

HEK293 cells were co-transfected as described in Section 3.3.4 with FLAG-TEV-

PAK1 and MYC-BRAFWT. Cells were washed in PBS, scraped in PBS, and the pelleted cells lysed in Pierce IP Lysis Buffer (Thermo Fisher Scientific). Protein concentration was estimated using the Pierce BCA Protein Assay Kit (Thermo Fisher Scientific). 4 mg of protein at 1 µg/µL was pre-cleared with Protein G beads for one hour. The pre-cleared protein was incubated with EZview FLAG beads (Millipore Sigma, F2426) at 1:10 dilution for 24 hours at 4oC. The protein-bead solution was centrifuged at 1200 g for 5 minutes and the supernatant was removed. The TEV cleavage protocol is based on the protocol by Bailey et al (198). The FLAG beads were washed in IP Lysis Buffer three times and twice in TEV wash buffer (10 mM Tris-HCl pH 7.5, 100 mM NaCl, 0.2%

IGEPAL). TEV cleavage mix (1x TEV Buffer (Thermo Fisher Scientific, 12575-015), 1 mM DTT, 40 U AcTEV protease) was added to the beads at twice the volume of the beads and rotated for 24 hours at 4oC. The cleavage reaction was centrifuged at 1200 g for 5 minutes and the supernatant (containing cleaved protein) was collected. The beads were washed with Pierce IP Lysis Buffer, centrifuged, and the supernatant was combined with the first supernatant. The cleaved proteins in the supernatant were divided for IP with the BRAF antibody (Santa Cruz Biotechnology, sc-5284) or IgG (Santa Cruz, sc-

2025) and rotated overnight at 4oC. Protein G beads were added at 1:10 dilution and incubated for 2 hours. The unbound proteins were collected after centrifugation. The

74 beads were then washed with Pierce IP Lysis Buffer three times and the proteins were eluted by resuspending the beads at 1/4 the incubation volume in 1x NuPAGE LDS sample buffer (Thermo Fisher Scientific), boiling for 5 minutes, and centrifuging at

16,000 g for 5 minutes. Eluted proteins were analyzed via western blot.

3.3.9 Ion exchange and gel filtration/size exclusion chromatography

Approximately 12 mg of lysate from TPC1 cells was extracted in IP Lysis Buffer

(Thermo Fisher Scientific) and dialyzed into a buffer containing 25 mM Tris pH 7.5, 50 mM NaCl, 1 mM EDTA, 10% glycerol with 1 µg/mL leupeptin, 1 µg/mL pepstatin, 500

µM PMSF, and 1 µg/mL aprotinin. Approximately 6 mg dialyzed protein was used for ion exchange. Ion exchange and gel filtration assays were done in collaboration with and with the expertise of Dr. Mark Parthun (OSU). Anion exchange of the TPC1 lysate was conducted over a MonoQ 5/50 GL column using the dialysis buffer as described and a high salt buffer (25 mM Tris, 1 M NaCl, 1mM EDTA, 10% glycerol) to create the increasing salt concentration gradient. The linear gradient was conducted over 10 column volumes and 0.5 mL fractions were collected and immediately stored on ice. Flow rate was 0.5 mL per minute and flow pressure was maintained below 4 mPa.

Gel filtration of anion exchange Fraction 17 was performed using a Superose 6

10/300 GL column in a buffer containing 25 mM Tris pH 7.5, 150 mM NaCl, 1 mM

EDTA, 10% glycerol. 0.5 mL fractions were collected at a flow rate of 0.5 mL/min and immediately stored on ice. Molecular weight standards (β-amylase, albumin serum, carbonic anhydrase, alcohol dehydrogenase, thyroglobulin, and apoferritin) were run prior to running the sample. Blue dextran elution defined the void fractions.

Fractions were analyzed by western blot as described in Sections 2.3.4 and 3.3.4.

Equal volume of the fractions was loaded in the NuPAGE 4-12% Bis-Tris gels (Thermo

Fisher Scientific), since total protein concentration varied based on elution profiles.

3.3.10 Proximity-dependent biotin identification (BioID2)

The following BioID2 protocol was performed based on a protocol generously shared by M. Parthun (OSU). HEK293 cells were seeded in 10cm plates and transfected with 1 µg MYC-BioID vector or MYC-BioID-PAK1, or mock transfected with no plasmid, using Optifect (Thermo Fisher Scientific) in OptiMEM (Thermo Fisher

Scientific) for 4 hours. The cells were then incubated in normal growth media for 24 hours. Media with 0, 0.5, 5, or 50 µM biotin (depending on the experiment) was added for an additional 24 hours. Cells were then washed with PBS and harvested by scraping.

Cell pellets were lysed in RIPA buffer (50 mM Tris-HCl pH 7.5, 150 mM NaCl, 1% NP-

40, 1 mM EDTA, 1 mM EGTA, 0.1% SDS, 0.5% sodium deoxycholate) with 1mM

PMSF, 1mM DTT, and Roche protease inhibitor cocktail tablets, then sonicated on ice

(3x 10s on, 2s off; 1x 6s on at 35% amplitude). Lysates were treated with 100 U benzonase for one hour, rotating at 4oC. After benzonase incubation, lysates were centrifuged at 16,000 g for 20 minutes, the supernatant was collected, and the protein quantified. Protein concentration was adjusted to the lowest concentration among the different samples in a volume of 1.5 mL. To capture the biotinylated proteins, pre-washed streptavidin-agarose bead slurry (EMD Millipore) was added for 3 hours, rotating at 4oC.

Beads were pelleted by centrifugation at 1,800 g for 3 minutes and washed once in RIPA buffer, once in 1 M NaCl, and twice again in RIPA buffer, rotating the beads for 5

76 minutes during each wash and subsequent pelleting by centrifugation. The protein was eluted in 1x NuPAGE LDS sample buffer (Thermo Fisher Scientific), vortexed, and boiled for 5 minutes. Protein was analyzed by western blot, as described Section 2.3.4.

3.4 Results

3.4.1 BRAFV600E and PAK1 interact.

Our previous work determined that wild-type (WT) BRAF and PAK1 physically interact and that BRAFV600E induction increases PAK1 expression and activity (149); however, it was unknown whether BRAFV600E also physically binds to PAK1. HEK293 cells were co-transfected with MYC-tagged BRAFV600E (MYC-BRAFV600E) or MYC-

BRAFWT and FLAG-tagged PAK1 (FLAG-PAK1). Using the proximity ligation assay

(PLA), in which proteins that are within 40 nm of each other can be identified in situ by red punctate signals, both BRAFV600E and BRAFWT interacted with PAK1 in cells that were successfully co-transfected (Fig. 3.1A). Additionally, lysate from HEK293 cells transfected with MYC-BRAFV600E and FLAG-PAK1 were subjected to immunoprecipitation (IP) with anti-MYC and anti-FLAG antibodies. Antibodies to the

HA-tag were used as negative controls. Proteins bound to these antibodies were subsequently analyzed via western blot to identify interacting proteins. In both cases, western blots indicated co-IP of BRAFV600E and PAK1: the MYC-BRAFV600E IP contained large amounts of BRAFV600E, as well as a portion of FLAG-PAK1, and the

FLAG-PAK1 IP contained not only bound PAK1, but also bound MYC-BRAFV600E (Fig.

3.1B).

The experiments in the HEK293 cells indicated that BRAFV600E and PAK1 could physically interact, either directly or indirectly, in an exogenous, overexpression system.

To determine if this interaction occurs endogenously, PLA with BRAF and PAK1 antibodies was conducted in 8505C thyroid cancer cells, which harbor only BRAFV600E and in TPC1 cells that express only BRAFWT protein. In both cases, BRAFV600E and

BRAFWT interacted with PAK1, confirming that the proteins co-localize endogenously

(Fig. 3.1C). Our previous work suggested that the BRAFWT-PAK1 interaction was enhanced during mitosis (149). To determine if this also occurs with BRAFV600E, we synchronized 8505C cells in mitosis using thymidine and nocodazole. Indeed, PLA signals for BRAF and PAK1 were strong in synchronized cells and absent in controls without primary antibodies (Fig. 3.1D). Section 3.4.5 will discuss the challenges of capturing the BRAF-PAK1 complex via endogenous IP, due to the transient nature of this complex. Therefore, we were unable to detect endogenous BRAFV600E-PAK1 interactions consistently in 8505C via IP. Nonetheless, co-IP in the exogenous HEK293 system and

PLA in both HEK293 and 8505C thyroid cancer cells indicates that BRAFV600E can interact with PAK1, as can WT BRAF.

Figure 3.1. PAK1 interactions with WT BRAF and BRAFV600E. (A) HEK293 cells were transfected with either MYC-WT or BRAFV600E and FLAG- PAK1. Red PLA signals indicated interactions between the two proteins. Nuclei were counterstained with DAPI. (B) IP with protein from HEK293 cells transfected with MYC-BRAFV600E and FLAG-PAK1. HA IP is used as a negative control. Western blot for MYC and FLAG indicated co-IP. Input is from lysate used for the IP. (C) Endogenous BRAF-PAK1 PLA in unsynchronized TPC1 and 8505C thyroid cancer cells. (D) Endogenous PLA in 8505C synchronized with thymidine and nocodazole. Bottom panel is a control with no primary antibodies. At least two biological replicates were performed with similar results.

3.4.2 PAK1 binding to BRAF is independent of PAK1 kinase activity and occurs via a critical region within PAK1 amino acids 121-140.

We previously discovered via co-IP and immunofluorescence studies that PAK1 and WT BRAF physically interact (149) and confirmed this interaction by PLA in TPC1 cells (Fig. 3.1C). As PAK1 influences cell signaling in both kinase-dependent and independent manners, we then asked whether PAK1 kinase function was required for its interaction with BRAF. HEK293 cells were transfected with MYC-BRAFWT and FLAG-

PAK1K299R, a kinase-dead mutant of PAK1, or FLAG-PAK1WT. Interestingly, both the

PAK1K299R and PAK1WT interacted with BRAF, indicating that PAK1 does not require kinase activation to interact with BRAF (Fig. 3.2A). Additionally, as prior assays used full-length PAK1 proteins, we determined the minimal region of PAK1 necessary for its interaction with BRAF. PAK1 truncation mutants were created following PAK1 structure

(i.e. N-terminal helices and beta strands, Uniprot ID: Q13153) and based on Higuchi et al’s work identifying PAK1 domains necessary for AKT and PDK1 interactions (101)

(Fig. 3.2B). Two broad truncations dividing the N and C terminal domains were first tested: N-terminal domain (PAK1 amino acids (AA) 1-206, PAK1-N) and C-terminal domain (PAK1 AA 207-545, PAK1-C). PAK1-N retained the ability to bind to BRAF, while PAK1-C had fewer PLA interactions, suggesting that the interaction domain involved the N-terminal region (Fig. 3.2C, top panels). To further narrow the interacting region, two smaller regions of PAK1 were created: PAK1 AA 1-140 (PAK11-140), which included the CRIB and autoinhibitory (AID) domains, and PAK1 AA 121-270 (PAK1121-

270), which included part of the AID and spans the region identified to interact with

CRIPAK (cysteine-rich inhibitor of PAK1), (199) (Fig. 3.2B). PLA signals indicated co- localization between BRAF and both PAK11-140and PAK1121-270 truncation mutants. Thus, when combined with the N and C-terminal data, the critical region for PAK1 co- localization with BRAF appears to be via amino acids 121-140 (Fig. 3.2C, bottom panels). This region forms a helical structure and contains the following amino acids:

PQAVLDVLEF YNSKKTSNSQ. While other regions may regulate this interaction, AA

121-140 appears to be the minimal limiting region.

One caveat is the difficulty in accurately testing for equal expression of the truncation mutants. The smaller truncations, PAK11-140 and PAK1121-270, were not always detectable via western blot without prior IP, due to either antibody epitope and/or protein stability (Fig. B.6A). Also, cloning PAK1 variants introduces stochastic point mutations for reasons still unknown, but has been confirmed by other investigators in the field

(verbal communication, J. Chernoff). The initial PAK1-C had several point mutations, which were corrected with further cloning to produce the constructs presented here; however, for reasons unknown, the expression of the “wild-type” C-terminal fragment was lower than the original mutated C fragment and the full-length PAK1 (Fig. B.6B).

Therefore, we limited our conclusions to the presence or absence of the PLA signals relative to the appropriate intra-experimental controls for our experiments.

Figure 3.2. PAK1 domains for BRAF interaction. MYC-FLAG PLA of HEK293 cells transfected with MYC-BRAFWT and (A) kinase negative FLAG-PAK1K299R or various FLAG-PAK1 truncation mutants (B, C). Numbers indicate amino acid locations. 2-3 biological replicates were performed and results were similar, with some variability with (C), as described in the text.

3.4.3 Identifying BRAF and PAK1 binding partners: CCT Complex

BRAF and PAK1 physically interact; however, other proteins involved in that complex were unknown and whether or not they interact directly was not certain. To determine other interacting members of the protein complex, we performed proteomics screens using mass spectrometry of BRAF and PAK1 interacting proteins collaboratively with the laboratories of Dr. M. Freitas (OSU) and Dr. S. Merali (Temple University,

Philadelphia, PA). For all experiments, HEK293 cells were transfected with MYC-BRAF and FLAG-PAK1. The lysate was divided for IP with MYC, FLAG, and HA antibodies, the captured proteins were analyzed by mass spectrometry, and identified proteins were confirmed by western blot. Proteomic analysis was performed by two independent methods by the two institutions using IP protein from three biological replicates for each institution (i.e. six total replicates). The results were similar between the two methods and provided confidence in the identified interactors. The list of interacting proteins for

BRAF and PAK1 were compared and the mutual proteins were used for further analysis as potential members of the BRAF-PAK1 complex (Tables B.3, B.4). As affirmation of the IPs and mass spectrometry analysis, several 14-3-3 proteins were identified to bind both BRAF and PAK1 (188, 189), as well as βPIX (also known as Rho guanine nucleotide exchange factor 7), a known RAC1 guanine exchange factor (GEF) and PAK1 interactor (189). Further analysis identified several members of the chaperonin containing

TCP-1 (CCT) complex and heat shock protein 90 (HSP90) (Tables B.3, B.4).

The CCT complex is a large chaperonin complex containing eight subunits

(CCT1-8), known for its role in helping to properly fold client proteins (estimated around

10% of cytosolic proteins), and in particular actin and tubulin (reviewed in (200, 201).

Upregulation of the CCT subunits and its interactions with oncogenic proteins have also been associated with certain cancers (200, 202). PAK4 has been shown to interact with members of the CCT complex (203); however, interactions with PAK1 or BRAF have not been reported, to our knowledge. Western blotting confirmed the co-IP of CCT3 in a portion of the same HEK293 lysate used for mass spectrometry analysis (Fig. 3.3A).

Higher levels of CCT were detected in the FLAG-PAK1 IP than in the MYC-BRAF IP, which is in agreement with the level of significance of the interaction determined by mass spectrometry. A panel of thyroid cancer cells were tested for CCT3 and CCT5 expression by western blot (Fig. B.7). All cell lines expressed CCT3 at relatively similar levels.

CCT5 was also expressed in all cell lines; however, TPC1 and HTh7 cells appeared to have the highest expression. To determine if members of the CCT complex interact with

BRAF and PAK1 endogenously in thyroid cancer cells, BRAF and PAK1 IP of lysate from TPC1 cells synchronized in mitosis were analyzed by western blot for interaction with CCT1 (Fig, 3.3B). PAK1 pulled down both BRAF and a small amount of CCT1.

BRAF did not seem to pull down a detectable amount of PAK1 and only a very small amount of CCT1 co-precipitated. Of note, this experiment was performed only once and requires replication with fresh protein. To this end, we anticipate a stronger signal of interaction between the three proteins in the endogenous setting. Additionally, due to overlapping size of all members of the CCT complex, we could not probe for more than one CCT member at a time; so therefore, other members may interact stronger than

CCT1 in TPC1 cells. IP using the CCT antibodies requires further optimization.

Additionally, in unsynchronized TPC1 cells, PLA was conducted using BRAF antibodies and antibodies to CCT1, CCT3, or CCT5 (Fig. B.8). Signals were very faint for all three conditions, and therefore, it was difficult to conclude localization of BRAF-CCT interactions. To determine if PAK1 was necessary for BRAF and CCT3 interaction, Pak1 null mouse embryonic fibroblasts (Pak1-/- MEFs) were probed for Braf and Cct3 and analyzed by immunofluorescence (Fig. 3.3C). In both wild-type and Pak1-/- MEFs, Braf and Cct3 appeared to co-localize, mainly in the cytoplasm. Therefore, BRAF and CCT3 could act as independent complexes from PAK1 and/or PAK1 is not required for BRAF to interact with CCT3. Other group I PAKs (i.e. PAK2 or PAK3) could also compensate for the PAK1 loss and facilitate the interaction between BRAF and CCT3. Taken together, BRAF and PAK1 may interact with several members of the CCT complex and the three units may bind as a complex to facilitate BRAF-PAK1 signaling.

Figure 3.3. PAK1 and BRAF interactions with the CCT complex. (A) Western blot of same IP protein analyzed by mass spectrometry (OSU) from HEK293 transfected with MYC-BRAFWT and FLAG-PAK1. Three biological replicates showed similar results. (B) Western blot of BRAF and PAK1 IP from TPC1 thyroid cells synchronized with thymidine and nocodazole. mIgG is the control for the BRAF IP and rIgG is the control for the PAK1 IP. Input for (A) and (B) are lysates used for the IP. “high” indicated a higher exposure for the western blot signal. (C) BRAF and CCT3 immunofluorescence of WT and PAK1-/- MEFs.

3.4.4 Identifying BRAF and PAK1 binding partners: HSP90

The proteomics analysis revealed HSP90 as a key interactor with both BRAF and

PAK1 (Tables B.3, B.4). STRING analysis from the OSU proteomics in collaboration with Dr. M. Freitas predicts HSP90 association with BRAF and with PAK1 (Fig. 3.4A).

Thus, HSP90 is a strong candidate to be part of the BRAF-PAK1 complex as a whole.

HSP90 has been identified as a chaperone for many proteins, including BRAF, 87 particularly BRAFV600E (188, 204, 205); however, its interaction with PAK1 has not been previously described. Western blot of MYC-BRAF and FLAG-PAK1 immunoprecipitated proteins (“bound”) confirmed HSP90 interaction with PAK1 and

BRAF (Fig. 3.4B). To test if PAK1 and BRAF are HSP90 client proteins in the thyroid cancer context, TPC1 (which is a WT BRAF cell line) and 8505C (which is a BRAFV600E cell line) were exposed to increasing concentrations of the HSP90 inhibitor, 17-AAG

(Fig. 3.4C, D). In TPC1, WT BRAF and PAK1 levels modestly decreased with HSP90 inhibition, while AKT and CRAF, which are well-known HSP90 client proteins, more dramatically decreased with 500 nM 17-AAG (Fig. 3.4C). Interestingly, HSP90 levels increased with 17-AAG, for reasons requiring further study. 8505C cells were relatively more resistant to 17-AAG; AKT and CRAF levels only decreased with 1000-5000 nM

17-AAG (Fig. 3.4D). BRAFV600E decreased with 2500 nM 17-AAG and modest decreases in PAK1 were also observed. Therefore, WT BRAF, BRAFV600E, and PAK1 may be partially dependent on HSP90 for stability in thyroid cancer cells. These results suggest that HSP90 is involved in the BRAF-PAK1 complex as a scaffold rather than a true protein chaperone required for BRAF and PAK1 stability. This interaction may be transient or depend on cell context.

Figure 3.4. PAK1 and BRAF interacts with HSP90. (A) STRING network analysis of proteins significantly interacting with FLAG-PAK1 as identified by OSU proteomics. Bold names are proteins/protein families further investigated in this study. BRAF and HSP90 were significantly interacting in the MYC- BRAF IP, as well. (B) Western blot of HEK293 IP samples sent for proteomic analysis (OSU). U= unbound to the MYC or FLAG beads. Bound= protein bound to the beads. “High” indicates a higher exposure for the western blot signal. Western blotting of 89

HSP90 in the FLAG IP was performed with two biological replicates; one for the MYC IP. Treatment with an HSP90 inhibitor, 17-AAG, for 24 hours in (C) TPC1 and (D) 8505C cells. GAPDH is the loading control for the blots above it. Two biological replicates were performed for each cell line.

3.4.5 Tandem immunoprecipitation to define the BRAF-PAK1 complex

A tandem IP system was developed to determine if the interacting proteins identified by the proteomic analysis of the MYC-BRAFWT and FLAG-PAK1 IPs are involved in the BRAF-PAK1 complex or as independent complexes with BRAF or

PAK1. A FLAG-PAK1 construct was made in which a tobacco etch virus (TEV) cleavage site was inserted between the FLAG sequence and start of the PAK1 sequence

(FLAG-TEV-PAK1) (Fig. 3.5A). This allowed for an effective removal system of the

FLAG tag from the PAK1 protein with addition of the TEV protease. HEK293 cells were transfected with FLAG-TEV-PAK1 and the lysate was immunoprecipitated with anti-

FLAG antibodies/beads. The beads with the bound proteins were incubated with TEV protease to remove the FLAG tag. The supernatant from this reaction was analyzed via western to confirm cleavage. Indeed, the input and remaining FLAG-TEV-PAK1 that did not bind to the FLAG beads (FLAG unbound) retained intact FLAG-TEV-PAK1 fusions proteins, as identified by co-localization of the FLAG and PAK1 signals (Fig. 3.5B).

Conversely, the cleaved product did not contain any FLAG signal, as the tag remained bound to the beads, although the soluble PAK1 protein without FLAG remained.

Therefore, the TEV cleavage system is effective in removing the FLAG tag and allowing further processing of un-tagged PAK1 protein complexes.

Confirmation of the effectiveness of the TEV system would permit the tandem IP of PAK1 and BRAF. HEK293 cells were co-transfected with FLAG-TEV-PAK1 and

MYC-BRAFWT. Lysates were immunoprecipitated with FLAG antibody/beads and TEV protease cleaved the tag as previously described. The lysates without the FLAG-tag were then immunoprecipitated with the BRAF antibody to identify proteins specifically in the

BRAF-PAK1 complex. IP with IgG was used as the negative control. Analysis by western blot confirmed the BRAF-PAK1 interaction in the “bound” BRAF fraction (Fig.

3.5C). This system is very specific in detecting only BRAF-PAK1 complex members, as the proteins that remain must be bound to FLAG-PAK1 and BRAF to be detected.

However, it is not yet very sensitive. The proportion of BRAF bound to PAK1 is likely very small and is not always readily detected by this system. In addition, BRAF and

PAK1 proteins are often detected in the IgG control IP (Fig. 3.5C and data not shown), perhaps due to the buffer changes required for each IP and the TEV cleavage reactions.

Therefore, this tandem-IP approach must be further optimized to limit the non-specific interactions and to enable detection of low abundance interacting partners. Nonetheless, the tandem-IP provides a useful tool to identify specific protein complexes amidst a pool of many interacting partners.

Figure 3.5. Tandem-IP with TEV cleavage. (A) Schematic of the FLAG-TEV-PAK1 construct. Numbers indicate amino acid locations. (B) HEK293 cells were transfected with FLAG-TEV-PAK1 and the lysate was used for FLAG IP. Input is lysate used for the IP. Unbound is protein not bound to the FLAG beads. Cleaved is the protein product after TEV cleavage of the FLAG tag. (C) HEK293 cells transfected with both FLAG-TEV-PAK1 and MYC-BRAF. Lysates underwent FLAG IP, then cleavage of the FLAG tag, then IP with BRAF or IgG control. Unbound and bound fractions refer to the BRAF or IgG IP. Input is lysate used for the initial IP. Two biological replicates of the tandem-IP were performed with similar results.

3.4.6 The BRAF-PAK1 complex may be localized to microtubules in mitosis

We previously observed that BRAF and PAK1 interactions were enriched in mitotic cells; BRAF and PAK1 immunofluorescent (IF) signals co-localized more often in mitotic cells and the interaction was more consistently detected by co-IP in thyroid cancer cells that were synchronized in mitosis with thymidine and nocodazole (149). In support of these observations, in situ PLA signals between BRAF and PAK1 in unsynchronized TPC1 cells were also often enriched in mitotic cells (Fig. 3.6A).

Additionally, proteomic analysis of the MYC-BRAF and FLAG-PAK1 IPs from

HEK293 cells indicated the tubulins as interacting proteins (Tables B.3, B.4, Fig. 3.4A).

Therefore, we hypothesized that the BRAF-PAK1 complex may be localized to microtubules during cell division. In TPC1 cells, both BRAF-α-tubulin and PAK1-α-

92 tubulin PLA signals were enhanced in mitotic cells, identified by IF counterstaining with phosphorylated histone 3 at serine 10 (pH3 S10), a marker of cells in mitosis (Fig. 3.6B).

The signals tend to be bright and concentrated around the chromosomes. Initial PLA studies with BRAF and PAK1 interactions and α-tubulin IF seemed to confirm that

BRAF and PAK1 interact in areas with concentrated microtubules (Fig. B.9A); however, it became apparent that the rat primary antibody for the α-tubulin IF was crossreacting with the mouse secondary probes for the PLA, and therefore, the red PLA signals were not only from BRAF-PAK1 interactions, but also PAK1-α-tubulin interactions (Fig. B.9, upper panels). Additionally, at this point in the study, the PLA signals for BRAF and

PAK1 alone were very minimal, likely from primary antibody batch differences from the initial phases of the PLA studies (Fig. B.9, lower panels). Therefore, although adjustments were made using chicken anti-α-tubulin primary antibodies for IF and various PAK1 antibodies were tested for better use in IF, BRAF-PAK1 signals, while present, were relatively weak. Thus, further studies on the localization of the BRAF-

PAK1 complex on microtubules require additional optimization for confirmation and are ongoing, as described in Chapter 4. Despite the difficulties with recent PLA studies, based on initial complementary proteomic and PLA data, BRAF and PAK1-associated complexes are likely to be localized on and interact with microtubules during mitosis.

Figure 3.6. PAK1 and BRAF interacts with tubulin. (A) BRAF-PAK1 PLA signals in unsynchronized TPC1 cells. Arrow points to a mitotic cell. (B) PLA of BRAF or PAK1 with α-tubulin in TPC1 cells. Phosphorylated histone 3 at serine 10 (pH3 S10) IF indicates mitotic cells. At least two biological replicates were formed and results were similar.

3.4.7 The BRAF-PAK1 complex is likely transient

Although we showed that BRAF and PAK1 physically interact (149), we also recognized that this interaction is not constitutive and is likely transient and dependent on cell context. Indeed, further attempts to consistently capture the endogenous BRAF-

PAK1 complex in thyroid cancer cells has been challenging. While identified in some experiments, even in cells that were synchronized in mitosis, we were unable to detect in a consistent manner that BRAF-PAK1 co-IP. Various lysis buffers were tested, as well as methods of lysis, the addition of phosphatase inhibitors in the cell media, stimulation of 94 signaling with hepatocyte growth factor (HGF), and different IP elution methods (i.e. glycine); however, the BRAF-PAK1 co-IP proved difficult to capture consistently. To this end, we pursued other methods of detecting protein-protein interactions. Protein complexes from cell lysate can be separated based on the charge and size/shape of the complex via ion exchange and gel filtration/size exclusion chromatography, respectively.

Therefore, we collaborated with Dr. M. Parthun (OSU) to pursue chromatography methods. Lysate from unsynchronized TPC1 cells were first separated by an anion exchange column, where increasing amounts of salt cause proteins to elute through the column at different rates depending on their charge. Elution fractions were collected and run on western blot to detect fractions with both BRAF and PAK1, as well as CCT5, since proteomics predicted the CCT complex as potential mediators of the BRAF-PAK1 complex. Fraction 17 contained the largest amount of all three proteins (Fig. 3.7A) and was subsequently used for further separation by gel filtration and analysis by western blot. Gel filtration fractions containing BRAF, PAK1, and CCT5 overlapped most prominently in Fraction 29 (Fig. 3.7B); however, the peaks of their elutions did not overlap and seemed to be consistent with the apparent molecular weight of each protein as a monomer. The BRAF peak occurred in Fraction 27, the PAK1 peak occurred in

Fraction 29, and the CCT5 peak occurred in Fraction 30, in agreement with their apparent molecular weights: BRAF (90 kDa), PAK1 (65 kDa), CCT5 (60 kDa). BRAF IP of

Fraction 29 was unable to detect co-IP of PAK1 or CCT5 (Fig. 3.7C). In attempt to avoid sensitivity and column resolution problems, we pooled the anion exchange Fractions 15-

20, those with the highest amount of BRAF, PAK1, and CCT5 (Fig. 3.7A), and

95 performed PAK1 IP. Western blot analysis was still unable to detect BRAF and CCT5 co-IP with PAK1 (Fig. 3.7D). Therefore, although BRAF, PAK1, and the CCT complex members were detected to interact in transfected HEK293 following crosslinking with formalin, the interaction is more transient and less abundant in TPC1 thyroid cancer cells.

Figure 3.7. Ion exchange and gel filtration of TPC1 protein demonstrates the transient nature of the BRAF-PAK1 complex. (A) Western blot of fractions from ion exchange chromatography of unsynchronized TPC1 lysate. Increased fraction number indicates increased NaCl concentration. Two separate western blot gels were run and indicated by the gap between Fraction 17 and Fraction 18. A biological replicate of the ion exchange was performed with approximately half the amount of protein and was used as a proof of concept for the replicate shown. (B) Western blot of gel filtration fractions of Fraction 17 from ion exchange in (A). Numbers represent elution fraction numbers. (C) Western blot of BRAF

IP of gel filtration Fraction 29 (D) Western blot of PAK1 IP of pooled anion exchange Fractions 15-20 from (A).

Co-IP studies require the protein-protein interaction to remain intact throughout the course of the experiment with many factors (ex: buffer composition and timing of lysis) greatly influencing the presence or absence of the interaction. Since the BRAF-

PAK1 complex may occur transiently and in low abundance, the BioID technique is a promising approach to identifying other members of this complex via proximity biotinylation. We cloned a MYC-BioID2-PAK1 fusion construct, transfected HEK293 cells with MYC-BioID2-PAK1 or the MYC-BioID2 only, and first assessed expression by immunofluorescence (IF). Both the BioID alone and the BioID-PAK1 fusion were diffusely expressed, mainly in the cytoplasm (Fig. 3.8A). To identify interacting PAK1 protein partners, HEK293 cells were transfected with the BioID2 or BioID2-PAK1 constructs and 50 µM biotin was added to the media (or a media-only control) for 24 hours. Cells were lysed and the biotinylated proteins were captured on streptavidin- conjugated agarose beads (“Strep”) and assed by western blot. Mock transfected cells, with no BioID2, and treated with biotin did not pull-down BRAF, as expected (Fig.

3.8B). Cells that were transfected, but not incubated with biotin, did not contain any detectable biotinylated proteins after streptavidin pull-down, except for the BioID2-

PAK1 itself, which readily utilizes the small amounts of biotin in the normal cell media to biotinylate itself (Fig. 3.8C, lanes 2 and 6). Once biotin was added to the media of transfected cells, streptavidin pull-down captured biotinylated BRAF protein; however, this was detectable to similar levels in cells with the BioID2 alone as in cells with the

BioID2-PAK1 (Fig. 3.8C, lanes 4 and 8), and therefore cannot be identified with certainty as a PAK1-specific interacting protein. Since the BioID2 has improved ligase efficacy from the original BioID, less biotin is needed to efficiently label proximal proteins (186). Therefore, we titrated the amount of biotin in the media to determine if the biotin concentration improved selectivity of BioID2-PAK1 compared to BioID2 alone. Transfected cells were treated with 0, 0.5, 5, or 50 µM biotin for 24 hours before harvesting and streptavidin pull-down. Lower amounts of biotin did not reduce the ability of BioID2 alone to biotinylate BRAF, and BioID2-PAK1had equal, if not lower, amounts of BRAF pull-down compared to BioID2 alone (Fig. 3.8D). Therefore, although the

BioID approach has promising potential to identify transient PAK1 interacting proteins, further optimization and insight into the proper negative control are needed to assure confidence in the selectivity of identified interacting proteins.

Figure 3.8. BioID2-PAK1 proximal interactions. (A) HEK293 cells transfected with MYC-BioID2 or MYC-BioID2-PAK1 fusion. Red indicates IF with anti-MYC antibodies. (B) Mock transfected HEK293 with no BioID2 constructs. Biotin was added and cell lysates underwent pull-down with streptavidin- conjugated beads (“strep”). Input is from lysate before streptavidin pull-down. HEK293 cells transfected with MYC-BioID2 or MYC-BioID2-PAK1 and incubated with 50 µM biotin (C) or increasing amounts of biotin (D) before streptavidin pull-down. Green bands indicate western blot signals for PAK1. Red bands indicate western blot signals for BRAF. Numbers beneath the blot in (C) indicate lane numbers. BioID2 assays as performed in (C) was done in at least two biological replicates and results were similar.

Taken together, although BRAF-PAK1 interactions can be detected both endogenously in thyroid cancer cells and in an over-expression system in HEK293 cells,

99 identification of the complex in endogenous conditions is not always consistent. The interaction may be transient and form only in certain cellular contexts, for example, during specific stages of mitosis.

3.5 Discussion

Identifying the requirements for protein-protein interactions and the members of protein complexes sheds light on the mechanisms underlying phenotypic observations and signaling connections. It is clear that BRAF and PAK1 functionally interact, from our previous work (149) and the work described in Chapter 2; however, the nature of their interaction and the other proteins involved in this interaction remained unknown.

Determining the details of protein complexes in cancer also aids in the understanding of how to better target proteins involved in complexes and potential methods of resistance to therapies. Additionally, mutated proteins may or may not interact with the same proteins or in the same manner as their wild-type counterpart and, thus, need to be independently studied. In the current study, we identified that not only wild-type (WT) BRAF, but also oncogenic BRAFV600E can interact with PAK1 (Fig. 3.1). This supports the evidence that group I PAK inhibition, as described in Chapter 2, may be useful in thyroid cancers that have either WT BRAF or BRAFV600E. While the IP experiments suggest BRAF and

PAK1 physically interact and the PLA experiments suggest this interaction is within 40 nm, they do not indicate whether this interaction is direct between BRAF and PAK1 or if there are other proteins mediating the interaction. To test direct binding of either WT

BRAF or BRAFV600E to PAK1, we expressed the individual proteins in vitro in a coupled in vitro transcription-translation assay and subsequent IP and western. Although the

100 proteins could be expressed, we were unable to detect direct binding (data not shown); however, positive controls that were expected to directly bind to each other also proved difficult to detect co-IP. Therefore, although it is unlikely that BRAF and PAK1 directly bind, this approach requires further optimization and, perhaps, other mediating proteins, such as the CCT complex or HSP90 to be co-expressed in order for BRAF to be complexed with PAK1.

Interestingly, PAK1 kinase activity was dispensable for its interaction with BRAF

(Fig. 3.2A). PAK1 has been shown to act as a scaffold in the AKT pathway (101) and to facilitate MEK activity (96), so it is feasible that PAK1 is also acting as a scaffold for

MAPK activity via BRAF. This is consistent with a lack of BRAF signaling differences with the addition of PID, which inhibits PAK kinase activity, but the full length PAK1 protein is still expressed (149). Additionally, we previously determined that BRAF expression regulates PAK signaling, since PAK activity decreases with BRAF siRNA

(149). This regulation likely occurs via BRAF (or another adapter protein) interacting with PAK1 in a regulatory region outside of K299 and is not dependent on PAK phosphorylation status, since interaction between BRAF and PAK1K299R remained intact

(Fig. 3.2A). Further, this is supported by our PAK1 domain mapping studies, in which it is likely that PAK1 amino acids 121-140 is a critical region of PAK1 for its interaction with BRAF, whether direct or indirect (Fig. 3.2C). It is possible that an adapter protein, such as HSP90 or the CCT complex, interacts with PAK1 and has allosteric effects that modulate PAK1 conformation to facilitate the BRAF-PAK1 interaction and, thus, PAK1 activity. Interestingly, CCT3 colocalized with BRAF in Pak1-/- MEFs (Fig. 3.3C). It is

101 possible that CCT3 could be acting as a facilitator for the BRAF-PAK1 interaction and that BRAF binds to the CCT complex independently or before it interacts with PAK1.

Therefore, even without PAK1 present, BRAF can still bind to CCT3. Further studies using the PAK1 truncation mutants and immunoprecipitation or PLA will aid in identifying these key adapter proteins and their role in mediating the BRAF-PAK1 interaction.

Additionally, BRAF and PAK1 have known roles in the progression of mitosis, including maintaining proper spindles and centrosomes through regulation of microtubules and activating key cell cycle-related proteins (87, 102, 103, 105, 106, 206-

210). However, their role in regulating mitotic progression as a complex has not yet been defined. Since the BRAF-PAK1 complex is enhanced in mitotic thyroid cancer cells and can interact with microtubules around the dividing chromosomes (Figs. 3.1 and 3.6), the interactions BRAF and PAK1 make with the microtubules may be important for proper chromosome separation. Further, the CCT complex also has known roles in the progression of mitosis, most notably as a chaperonin for tubulins, but also by regulating a number of cell cycle-dependent proteins (200, 211-214). The CCT complex is also emerging as a key interacting player in oncogenesis (201, 202). Therefore, it is possible that BRAF and PAK1 are acting as a complex with the CCT proteins, and this complex may be localized to microtubules during mitosis. Further studies with improved PLA with BRAF, PAK1, and the CCT members with microtubule IF will aid in clarifying the localization of this complex. Initial studies to disrupt proper microtubule formation with nocodazole indicated the BRAF-PAK1 complex remained intact (data not shown);

102 however, these were performed with the cross-reacting tubulin antibodies and need to be confirmed.

Another potential link in the BRAF-PAK1 complex may be HSP90. Proteomic analysis suggested HSP90 bound to both BRAF and PAK1 (Fig. 3.4). BRAF has been previously identified as a binding partner of HSP90, particularly when mutated to

BRAFV600E (188, 204, 215); however, PAK1 was previously determined not to bind

HSP90 (215). This could be due to differences in methods, as we co-expressed FLAG-

PAK1 and MYC-BRAF in HEK293 cells, crosslinked cells with formalin prior to lysis, and therefore, may be able to capture more transient or weak interactions. It is also possible that PAK1, and potentially BRAF, do not require HSP90 for stability, but rather as a mediator of their interaction. Indeed, western blot analysis after 17-AAG treatment resulted in only a modest decrease in BRAF and PAK1 proteins levels, in comparison to known HSP90 client proteins, AKT and CRAF (Fig. 3.4C). This is supported by the large-scale analysis of HSP90 interactions that determined non-client proteins were generally more stable than client proteins (215), thus the latter hypothesis may be more correct. BRAF-PAK1 PLA signals were very weak in experiments with both control and

17-AAG treatments in TPC1 cells (data not shown), and therefore, the effect of HSP90 inhibition on BRAF-PAK1 complex formation could not be determined with confidence.

Future studies using the PAK1 truncation mutants and probing for co-IP of BRAF and

HSP90 will aid in determining if PAK1 influences BRAF-HSP90 binding, and therefore, supporting the hypothesis that the three are required for stable complex formation.

Optimization of the TEV tandem-IP assay, for example with IP of HSP90 instead of

103

BRAF IP after the FLAG IP and probing for all three proteins on western blot, will further clarify if BRAF and PAK1 are binding to HSP90 as a complex together or as separate complexes.

As BRAF and PAK1 are both kinases involved in major signaling pathways, one would not expect them to be constitutively interacting, but rather, interacting with the proper stimuli. As mentioned above, this could be related to cell cycle. Therefore, it is not surprising that we could not detect the complex by anion exchange and gel filtration in unsynchronized TPC1 cells (Fig. 3.7), as the proportion of BRAF-bound PAK1 is likely very small. Therefore, it may be necessary to synchronize the cells in mitosis and crosslink the proteins before separating the complexes by the various columns in order to detect the complex. Alternatively, the protein complexes could be separated by sucrose or glycerol gradients, or run on a non-denaturing polyacrylamide gel electrophoresis.

The BioID approach seemed very promising to detect the transient BRAF-PAK1 interaction without the need of synchronizing the cells or relying on the complex to stay physically intact throughout the course of the experiment. However, the difficulty lies in determining the proper negative control. In our studies, both BioID2 alone and BioID2-

PAK1 readily came within proximity to BRAF (Fig. 3.8). This is not surprising given the diffuse over-expression of both BioID2 and BioID2-PAK1 in the cytoplasm (Fig. 3.8A), giving BioID2 alone chances to be in proximity of many cytosolic proteins. Mock-treated cells did not contain detectable amounts of biotinylated BRAF (Fig. 3.8B), so the enrichment of BRAF in the BioID2-transfected cells is not due to endogenous biotinylated BRAF, but is a product of the BioID2 reaction. Due to the catalytic nature of

104 the BioID2, it is not possible to determine if BRAF is interacting with the BioID2 protein itself or is simply within 10 nm and being biotinylated. We attempted to improve the enrichment of PAK1-interacting proteins by lowering the amount of biotin (Fig. 3.8D); however, BRAF remained biotinylated by the BioID2 itself. It is possible that shortening the incubation time with biotin may improve substrate selectivity. Alternatively, a

CRISPR-based approach to knock-in BioID2 before the PAK1 locus to attain physiologic levels of the fusion protein may also improve substrate selectivity; however, the question of the proper negative control for this type of system still remains difficult, as it might be best to express BioID2 alone under the PAK1 promoter, without affecting PAK1 expression. Taken together, the BioID system provides a useful tool for detecting protein- protein interactions if there is enrichment of interacting proteins with the fusion protein over the BioID alone; however, subsequent validation studies are required by other techniques due to limitations of controlling for false positives.

Despite the important limitations of detecting protein-protein interactions, we have further characterized the BRAF-PAK1 interaction. Both wild-type and oncogenic

BRAFV600E can interact with PAK1, and this interaction in enriched in mitotic cells. The interaction may be mediated by CCT or HSP90 proteins, likely via amino acids 121-140 in PAK1. We have identified novel interacting proteins of BRAF and PAK1, furthering our knowledge of how BRAF and PAK1 may regulate many signaling pathways within the cell. Overall, the interaction between these two oncogenic proteins is important in understanding the signaling crosstalk between BRAF and PAK1, and additional sensitive assays are needed to parse the precise biochemistry.

105

Chapter 4: Conclusions and Future Directions

4.1 Conclusions

Aggressive forms of thyroid cancer remain resistant to current therapy and require a better understanding of the complexities of these cancers to aid in the development of better therapeutic approaches. This work encompasses a broad translational study utilizing in vitro and in vivo methods to identify group I PAKs as important nodes in aggressive behavior in thyroid cancer. PAK1 can physically interact with both WT BRAF and BRAFV600E, supporting our pharmacologic studies showing that G-5555 is effective at reducing cell viability in a number of thyroid cancer cell lines with varying genetic drivers. These BRAF-PAK1 interactions are enhanced in mitotic cells and localized to microtubules. Treatment with either of the group I PAK inhibitors, G-5555 or

FRAX1036, arrests TPC1 and SW1736 cells in G0/G1, suggesting that the BRAF-PAK1 complex is involved in regulating cancer cell progression through G1, and that this may be due to altered microtubule dynamics.

Only one among the eight thyroid cancer cell lines tested did not have significant reduction in cell viability with G-5555, despite a decrease in PAK signaling. This cell line, FTC133, still displayed reduced migration and invasion, indicating that inhibition of group I PAKs can negatively affect thyroid cancer cells in multiple ways. This supports the idea that drug screens using viability as the only marker of efficacy may be missing evidence of functional effectiveness in other aspects of cell behavior. Phosphorylated 106 vimentin also decreased with G-5555, indicating that PAK inhibition may be effective at treating cases similar to those studied by our group previously, in which increased vimentin and EMT were found at the edges of invasive thyroid cancers.

We identified a combination therapy with G-5555 and the BRAFV600E inhibitor,

PLX4032, with synergistic decreases in thyroid cancer cell viability in cells harboring

BRAFV600E. These observations support the hypothesis that the BRAF-PAK interaction is likely independent of MEK. Agreeably, in vivo, we observed that treatment with G-5555 did not affect pERK levels and that pPAK, but not pERK, levels corresponded with thyroid size in mice with BRAFV600E expression.

In the thyroid-specific inducible BRAFV600E mouse model of thyroid cancer, one week treatment with G-5555 at the time of BRAFV600E induction restrained thyroid size by over 50% compared to vehicle-treated mice. This agrees with a trend toward decreased Ki67 in thyroid cells of mice treated with G-5555 and the inhibition of cell cycle progression in SW1736 cells, which harbor BRAFV600E. This also suggests the role of the BRAF-PAK1 complex localized to mitotic cells observed from the PLA studies.

Importantly, one week of G-5555 treatment in vivo reduced carcinoma formation in this mouse model in which BRAFV600E is highly overexpressed. None of the mice with

BRAFV600E induction exhibited normal thyroid pathology, as even those that did not develop carcinoma displayed thyroid hyperplasia or developed follicular adenomas. This suggests that although G-5555 reduces carcinoma formation, it does not completely inhibit tumor initiation. Therefore, PAKs may have a larger role in tumor progression rather than initiation in this model. This hypothesis is supported by our human data,

107 where PAKs were upregulated at the invasive edges rather than the centers of aggressive thyroid tumors. Together, these results demonstrate that pharmacologic group I PAK inhibition by G-5555 is a promising therapeutic approach for thyroid cancers and that further efforts to test safety and dosing strategy are needed.

Domain mapping of PAK1 identified the amino acids between 121 and 140 as one region to maintain PAK1 interaction with BRAF. Additionally, PAK1 kinase activity was not required for the interaction with BRAF. Together with our previous data indicating that BRAF regulates PAK activity, these findings suggest that BRAF may interact with PAK1 prior to phosphorylation and activation by complex formation to activate PAK1. This complex may also incorporate the CCT chaperonin complex or

HSP90 chaperone to promote the BRAF-PAK1 complex in specific cellular contexts, such as mitosis. It is likely that BRAF and PAK1 bind these chaperones, particularly the

CCT complex, prior to interacting with each other, as BRAF and CCT3 could maintain interaction in Pak1-/- MEFs. It is also possible that BRAF and PAK1 interact with the

CCT complex or HSP90 independently of each other. The tandem-IP and column separation approaches were designed to clarify this. However, further optimization is needed. Additionally, we have shown that the BRAF-PAK1 complex is likely very transient and is difficult to capture consistently, consistent with the very specific role that

BRAF and PAK1 coordinately play in the cell.

The work presented here has expanded our knowledge of the role group I PAKs play in thyroid cancer. Group I PAKs can be pharmacologically inhibited. Biochemical studies suggest further avenues of therapeutic approaches, potentially through the use of

108 combination treatment with HSP90 inhibitors or microtubule-disrupting agents, such as taxanes. BRAFV600E and AKT inhibitors also lead to synergistic decreases in thyroid cancer cell growth when combined with PAK inhibition. Therefore, our studies have identified several alternative strategies for treating aggressive thyroid cancers, changing the course of therapeutic advancements for these and potentially other cancers with

BRAF or PAK1 activation.

4.2 Future Directions

There are several avenues for continuation of this work. Our in vivo model of group I PAK inhibition with G-5555 tested the involvement of PAKs in thyroid cancer initiation; however, G-5555 was given simultaneously with the dox to induce BRAFV600E.

Therefore, future models in which G-5555 is administered following tumor establishment will provide a more accurate treatment model for testing the ability of G-5555 to reduce or maintain thyroid size, a more clinically applicable approach. Thyroid size could be monitored by ultrasound and comparisons of growth rates and final thyroid volumes can be made. Additionally, this transgenic BRAFV600E mouse model, under the Tg promoter, so robustly overexpresses BRAFV600E, and thus has very high activation of ERK, that after approximately two weeks, BRAFV600E expression is abrogated. This is likely due to dedifferentiation and loss of Tg promoter activity. Thus, this model is useful to study primary tumor development, but it is unable to study longer term effects on metastasis.

Since our data indicates PAK activity is involved with metastatic characteristics, it will be important to study the effects of PAK inhibition in vivo on metastatic growth,

109 potentially using a knock-in BRAFV600E mouse model (216), which expresses more physiological levels of BRAFV600E and is useful for longer term thyroid cancer progression studies. These mice are currently breeding in our facility.

Since G-5555 was administered by oral gavage and disseminated throughout the mouse, it is unclear whether the restriction in thyroid size and carcinoma formation are due to thyrocyte-specific inhibition of group I PAKs or if inhibition in other cell types are also contributing. To this end, we can test thyrocyte-specific PAK effects by utilizing a mouse model with thyroid-specific, inducible expression of the PID peptide, which molecularly inhibits group I PAK activity. This model is currently being crossed with the thyroid-specific, inducible BRAFV600E mice from this study. Studies with the PID mouse in a colon cancer model support our observations that reduced PAK activity also reduces carcinoma formation (130). Therefore, group I PAK inhibition may be a useful therapeutic approach across several types of cancers.

Our results suggest that PAKs are important for thyroid cancer progression rather than initiation and that both ERK and PAK activation are necessary for thyroid cancer growth. But, further experiments are required to test this hypothesis. In vivo experiments using the treatment model of G-5555 after tumor establishment or models with delayed induction of the PID peptide in thyrocytes with BRAFV600E may support the role of PAKs in tumor progression. Transformation assays using NIH/3T3 cells transfected with

BRAFV600E and/or the PID peptide may determine if group I PAKs have a role in cancer cell transformation, and thus initiation.

110

Additionally, our in vitro data suggest that combining G-5555 with the

BRAFV600E inhibitor, PLX4032, has synergistic effects on cell growth. Therefore, it is important to study this combination in vivo, as it could be therapeutically advantageous.

Combining the two drugs and using lower doses of both could reduce unwanted off-target effects, and limit the potential adverse effects on cardiac function with PAK inhibition.

Further, if PAK activation is a mechanism of resistance to BRAFV600E inhibition (141), this combination may eliminate the resistant cell populations before they can expand.

Different dosing strategies, such as administering PLX4032 before G-5555 versus giving the two drugs together at the same time, should also be tested for this reason.

It will also be interesting to test the effects of G-5555 and PLX4032, and the combination, on BRAF-PAK1 complex formation. Our results with the kinase-dead

PAK1K299R suggest that PAK1 does not need its kinase function to interact with BRAF.

However, inhibition with a pharmacologic inhibitor of PAK1 may disrupt the ability of

PAK1 to interact with other mediating proteins, such at the CCT complex or HSP90.

Therefore, co-IP, proteomic, and PLA studies after thyroid cancer cell treatment with G-

5555 will shed light on further requirements for the BRAF-PAK1 interaction.

We had difficulties in consistently detecting the BRAF-PAK1 complex, particularly in thyroid cancer cells with endogenous levels of the proteins. It is possible that stimulation by adhesion is necessary for the BRAF-PAK complex to form, as cytoskeletal rearrangements and microtubules are important for the BRAF-PAK1 complex. PAK activation or stimulation by adhesion results in a complex with ERK2

(97). This may also be true for its interaction with BRAF. This hypothesis is also

111 supported by the observation that the BRAF-PAK1 complex occurs in mitotic cells, as these cells are undergoing cytoskeletal rearrangements and loosely detaching from the plates to divide. Cells can be seeded on fibronectin- or collagen-coated plates/ slides and co-IP or PLA assays can be done to probe for BRAF-PAK1 interactions. Alternatively, time course experiments can be done after detaching the cells from regular, non-coated plates, allowing them to settle, and then probing for the BRAF-PAK1 interaction.

Our PLA studies had difficulty detecting BRAF-PAK1 PLA signals, in part due to

PAK1 antibody complications, that have occurred for many in the field based on personal communications. To improve the detection of endogenous PAK1 protein by antibodies, we worked with Biocytogen Co., Ltd. to generate TPC1 and 8505C cell lines with FLAG tags knocked-in to the endogenous PAK1 locus to generate N-terminal FLAG-PAK1 fusion proteins. These would be expressed under the regulation of the endogenous PAK1 promoter, providing the ability to use anti-FLAG antibodies to detect endogenous levels of PAK1 in thyroid cells. We have been validating these cell lines (Fig. 4.1) for future co-IP and PLA studies to confirm our current proteomics results in thyroid cancer cells.

These cells will also be useful in comparing the differences in BRAF-PAK1 complex members between WT and BRAFV600E cells (TPC1 cells only express WT BRAF and the

8505C cells only express BRAFV600E).

112

Figure 4.1. FLAG-tag knock-in cell lines. TPC1 and 8505C cells created by Biocytogen to contain FLAG-tagged PAK1 in the endogenous PAK1 locus. “WT” are the original cell lines with no FLAG insertion; “FLAG” are the FLAG-PAK1 cell lines. Cell lysates from each cell line were used for FLAG IP and western blot for FLAG confirms the creation of the FLAG-PAK1 fusion. The first lane contains the molecular weight marker, indicating 70 kDa. This assay was run by Luis Bautista, a graduate student in the Ringel Lab, with guidance from Christina Knippler.

Finally, although both BRAF and PAK1 may interact with HSP90 and the CCT complex, current detection methods cannot eliminate the possibility that BRAF and

PAK1 interact with these proteins in separate complexes. The tandem-IP approach, in which IP of PAK1 is followed by IP of BRAF, will ideally purify only the BRAF-PAK1 complex and its associated members. Although initial studies yielded low sensitivity and high background, further improvements on this method, perhaps by crosslinking the proteins prior to isolation, will greatly impact the confidence in identifying interacting partners of the BRAF-PAK1 complex. Additional methods may also include further optimization of the ion exchange/ gel filtration studies or separating thyroid cell lysate by native PAGE, instead of by denaturing conditions, to probe for co-migration of BRAF,

PAK1, HSP90, and/or the CCT complex.

Many exciting opportunities remain to understand the roles group I PAKs play in thyroid cancer.

113

Appendix A: List of Abbreviations

AID autoinhibitory domain AKT v-akt murine thymoma viral oncogene homolog 1 ANOVA analysis of variance ARAF v-raf murine sarcoma 3611 viral oncogene homolog 1 ATC anaplastic thyroid cancer ATP adenosine triphosphate BioID proximity-dependent biotin identification BRAF v-raf murine sarcoma viral oncogene homolog B1 BSA bovine serum albumin CCT chaperonin containing TCP-1 CDC42 cell division cycle 42 CDK1 cyclin-dependent kinase 1 cMET hepatocyte growth factor receptor Co-IP co-immunoprecipitation CR conserved region CRAF v-raf-1 murine leukemia viral oncogene homolog 1 CRIB CDC42/RAC1 interactive binding CRIPAK cysteine-rich inhibitor of PAK1 Dox doxycycline DTT dithiothreitol EGFR epidermal growth factor receptor EMT epithelial-to-mesechymal transition ERK extracellular signal regulated protein kinase FBS fetal bovine serum FDA Food and Drug Administration FGFR fibroblast growth factor receptor FNA fine needle aspiration FRET Förster resonance energy transfer FTC follicular thyroid cancer GEF guanine exchange factor GIT1 G-protein-coupled receptor kinase-interacting target 1 GSKα/β glycogen synthase kinase-3 alpha/beta GTP guanosine triphosphate HER3 erb-b2 receptor tyrosine kinase 3 hERG human ether-à-go-go-related gene HGF hepatocyte growth factor 114

HSP90 heat shock protein 90 IC50 half maximal inhibitory concnetration IF immunofluorescence IHC immunohistochemistry IP immunoprecipitation IPA-3 inhibitor p21-activated kinase-3 JNK/SAPK c-Jun amino-terminal kinase/ stress-activated protein kinase KIT cellular homolog of the feline sarcoma viral oncogene v-kit MAPK mitogen-activated protein kinase MEF mouse embryonic fibroblast MEK MAPK/ERK kinase MEN multiple endocrine neoplasia MLCK myosin light chain kinase MPER mammalian protein extraction reagent MPNST malignant peripheral nerve sheath tumors MTC medullary thyroid cancer NCI National Cancer Institute NCK NCK adapter protein NF neurofibromatosis NIH National Institutes of Health NIS sodium iodide symporter OD optical density PAGE polyacrylamide gel electrophoresis PAK p21-activated kinase PAX8 paired box 8 PBS phosphate buffered saline PDGF platelet-derived growth factor PDGFR platelet-derived growth factor receptor PDTC poorly differentiated thyroid cancer PI propidium iodide PI3K phosphatidylionositol 3-kinase PID PAK inhibitory domain PIK3CA phosphatidylinositol-4,5-bisphosphate 3-kinase catalytic subunit PLA proximity ligation assay PLK1 polo-like kinase 1 pPAK phosphorylated PAK PPARγ peroxisome proliferator-activated receptor gamma PPI protein-protein interaction PTC papillary thyroid cancer PTEN phosphatase and tensin homolog RAC1 RAS-related C3 botulinum toxin substrate 1 RAF rapidly accelerated fibrosarcoma RAI radioiodine RAS rat sarcoma viral oncogene homolog 115

RET rearranged during transfection RTK receptor tyrosine kinase SD standard deviation SIK2 salt inducible kinase 2 SILAC stable isotope labeling with amino acids in cell culture Strep streptavidin-conjugated agarose beads Sync synchronized T3 triiodothyronine T4 thyroxine TAK transforming growth factor-β (TGF-β)-activated kinase TBST tris-buffered saline with Tween 20 TCoB tubulin cofactor B TERT telomerase reverse transcriptase TEV tobacco etch virus Tg thyroglobulin TRH thyrotropin-releasing hormone TSH thyroid-stimulating hormone TSH-R thyroid-stimulating hormone receptor VEGFR vascular endothelial growth factor receptor WB western blot WST water-soluble tetrazolium WT wild-type

116

Appendix B: Supplementary Figures and Tables

Figure B.1. Cell viability effects of BRAFV600E and AKT inhibitor monotherapy. Six thyroid cancer cell lines were treated with increasing doses of PLX4032 (A) or MK2206 (B) for 72 hours. Cell viability was determined by colorimetric WST-8 assay. The optical density of each treatment was normalized to the 0 µM control. Assays were conducted in triplicate with at least three biological replicates. Data are represented as means ± SD.

117

Figure B.2. Combination treatment with BRAFV600E and PAK inhibitors in WT BRAF cell lines. (A) WST-8 cell viability assay of TPC1 and FTC133 after 72 hours of combined BRAFV600E (PLX4032) and PAK (G-5555) were performed. All dose combinations were normalized to the 0 µM G-5555 + 0 µM PLX4032 control. Assays were performed in triplicate with at least three biological replicates. Data are represented as means ± SD. (B) Synergy tables representing each PLX4032/G-5555 combination using the model proposed by Zhao et al (Zhao et al. 2012). The bolded top numbers are the drug interaction indices, τ, and the bottom numbers in parentheses are the confidence intervals. Synergy is defined as τ < 1 and upper confidence limit < 1. Antagonism is defined as τ > 1 and lower confidence limit > 1. The darker red indicates stronger antagonism. (C) Western blot of PAK signaling (pPAK S144/S141) and BRAF signaling (pERK T202/Y204) at doses and timing selected for antagonism and cell viability. 118

GAPDH is used as the loading control for the membranes above it. At least two biological replicates were performed per cell line and results were similar.

Figure B.3. Combination treatment with AKT and PAK inhibitors in WT BRAF cell lines. (A) WST-8 cell viability assay of TPC1 and FTC133 after 72 hours of combined AKT (MK2206) and PAK (G-5555) inhibition at the specified combination of doses. All dose combinations were normalized to the 0 µM G-5555 + 0 µM MK2206 control. Assays were conducted in triplicate with at least three biological replicates. Data are represented as means ± SD. (B) Synergy tables representing each MK2206/G-5555 119 combination using the model proposed by Zhao et al (Zhao et al. 2012) and as described in Supplementary Fig. 2. The darker red indicates stronger antagonism; the darker blue indicates stronger synergy. (C) Western blot of PAK signaling (pPAK S144/S141) and AKT signaling (pAKT S473 and pGSK-3αβ S21/9) at doses and timing selected for potential synergy and cell viability. GAPDH is used as the loading control for the membrane above it. At least two biological replicates were performed per cell line and results were similar.

Figure B.4. Ki67 and cleaved caspase-3 in BRAFV600E-induced thyroids. (A) Quantitation of Ki67-positive thyrocytes by IHC in mice with BRAFV600E induction and treated as noted. Orange dots indicate benign pathologies. Arrows in images point to examples of Ki67-positive thyrocytes. Comparisons between groups were analyzed by Mann-Whitney test. (B) Quantitation of cleaved caspase-3-positive cells by IHC in mice with BRAFV600E induction and treated as noted. Representative images are shown. Scale bar indicates 50 µm. Data are represented as individual plots with means ± SD.

120

Figure B.5. Relationship of pERK levels and thyroid size in BRAFV600E-induced mice. Comparison of thyroid volume and quantification of the pERK1/2 T202/Y204 western blot bands normalized to GAPDH for each western blot membrane. Each cohort of mice was treated at different times and the lysates from mice in Cohort 2 were divided on two gels (a and b), as titled. Comparisons were fit with a linear regression and the r2 values are shown.

121

Table B.1. Key tumor promoting mutations in human thyroid cancer cell lines.

V600E Cell Line Histology BRAF RET/PTC PI3K/ PTEN Other TPC1 PTC WT Fusion BCPAP PTC V600E/ WT None p53 (D259Y, K286E) HTh74 ATC WT None NF1 (L732fs), p53 (K286E) SW1736 ATC V600E/ WT None 8505C ATC V600E/- None p53 (R248G, R273C)

K1 PTC V600E/ WT None PI3K (E542K) p53 (R213R FTC133 FTC WT PTEN (R130*) p53 (R273H) THJ-16T ATC WT None PI3K (E545K) MKRN1/BRAF fusion

Table B.2. Estimated IC50 concentrations (µM) for PLX4032 and MK2206.

Cell Line PLX4032 Std Err 95% CI MK2206 Std Err 95% CI TPC1 2.55 0.09 (2.36, 2.74) 6.24 0.23 (5.78, 6.70) BCPAP 0.38 0.09 (0.20, 0.56) 2.66 0.10 (2.46, 2.87) SW1736 0.53 0.28 (<0.01, 1.09) >10 K1 8.44 1.04 (6.40, 10.48) >10 FTC133 5.50 0.19 (5.14, 5.87) 4.71 0.10 (4.52, 4.90) THJ-16T 3.41 0.19 (3.05, 3.78) 3.22 0.45 (2.34, 4.10) 122

Figure B.6. Variable expression of FLAG-PAK1 truncations. HEK293 cells were transfected with the indicated FLAG-PAK1 constructs. (A) FLAG IP of PAK1 truncations and HA as the negative control. Note that different amounts of input for the western were used: 20 µg for FL, 10 µg for PAK11-140, and 20 µg for PAK1121-270. However, same amount of protein was used for each IP assay and western. First arrow indicates FL PAK1, second arrow indicates the apparent size of PAK1121-270, third arrow indicates the apparent size of PAK11-140. (B) Western blot of different plasmids of the FLAG-PAK1-C. “C” indicates the initial construct with mutations. “WT” indicates constructs without mutations, and those used in the study. “New” indicates a new clone of the WT C plasmid. GAPDH is used as a loading control. First arrow indicates FL PAK1, second arrow indicates the apparent size of PAK1-C.

Figure B.7. CCT expression in thyroid cancer cell lines. Western blot of CCT3 and CCT5 expression in a panel of thyroid cancer cell lines. Those cells with BRAFV600E are indicated. GAPDH is used as a loading control for the panel above it.

123

Figure B.8. CCT-BRAF interactions. Unsynchronized TPC1 cells were used for PLA between BRAF and CCT1, CCT3, or CCT5. Faint red signals indicate potential interactions.

Figure B.9. Cross-reactivity of PLA probes. TPC1 cells were used for PLA and IF. (A) PLA signals with BRAF-PAK1 antibodies (mouse and rabbit, respectively) and IF with a rat primary antibody to α-tubulin. (B) Upper panels are PLA with the rat primary antibody to α-tubulin and the rabbit PAK1 antibody, indicating cross-reactivity with rat primary antibodies and mouse secondary PLA probes, interfering with signals in (A). Lower panels in (B) indicate the loss of actual BRAF-PAK1 PLA signals without counterstaining with α-tubulin IF, likely from PAK1 antibody batch differences.

124

Table B.3. FLAG-PAK1 interacting proteins as identified by M. Freitas (OSU) proteomics. Bolded proteins were also significantly interacting with MYC-BRAF. HA/FLAG log FC is the fold change of HA (negative control) interacting proteins compared to FLAG proteins. Note: TCPD= CCT4.

UniprotAC Protein Name HA/FLAG logFC q value Q13153 PAK1 -6.501998899 1.43E-15 P21333 FLNA -7.245760051 6.51E-07 P68366 TBA4A -2.591915055 7.70E-06 P08238 HS90B -2.173899524 3.21E-05 P04350 TBB4A -1.815524824 0.000488104 P68371 TBB4B -1.704362213 0.000953602 Q13885 TBB2A -1.717719809 0.001228375 Q9BVA1 TBB2B -1.702971405 0.001339431 P07437 TBB5 -1.616837762 0.001406338 P07900 HS90A -1.817577911 0.001521286 P23526 SAHH -3.957773225 0.001886561 Q9NY65 TBA8 -1.951079862 0.002006542 P50991 TCPD -3.928563666 0.002019208 P60842 IF4A1 -5.541639602 0.00205335 Q58FF7 H90B3 -2.062523877 0.00205335 Q13748 TBA3C -1.520191146 0.002980228 P13639 EF2 -2.400062124 0.003135063 P31948 STIP1 -5.35159834 0.004277103 Q9BUF5 TBB6 -1.787066177 0.004311199 A6NNZ2 TBB8L -1.92763749 0.004333451 Q13509 TBB3 -1.591491246 0.004333451 Q12931 TRAP1 -2.188941105 0.005231837 Q3ZCM7 TBB8 -1.613462853 0.005231837 P14625 ENPL -2.320943117 0.005526569 P11021 GRP78 -2.284460939 0.006547173 Q14240 IF4A2 -5.056721799 0.008625497 P15056 BRAF -5.125876877 0.008785109 Q13162 PRDX4 -2.977637318 0.009390521 Q58FG1 HS904 -2.687306641 0.010015381 P11142 HSP7C -1.604035363 0.013466103 Q6PEY2 TBA3E -1.389339418 0.013466103 Q58FF8 H90B2 -1.732329656 0.013466103 P63261 ACTG -1.187275577 0.017781839 A6NHL2 TBAL3 -4.885906131 0.021627392 P60709 ACTB -1.15113141 0.021627392 Q9H4B7 TBB1 -1.767168957 0.022534546 Q14568 HS902 -1.77246996 0.023195794 A6NMY6 AXA2L -2.315682955 0.02516338 Q58FF6 H90B4 -2.423856184 0.028430324 P08107 HSP71 -0.927203089 0.029358865 P07355 ANXA2 -1.534715632 0.033158925 P68363 TBA1B -1.089562275 0.033158925 Q71U36 TBA1A -1.050412415 0.042829955

125

Table B.4. Interacting proteins of both MYC-BRAF and FLAG-PAK1 as analyzed by S. Merali (Temple).

Abundance Ratio: q-value: (FLAG+MYC) / UniprotAC Description Combined LogFC HA Q562R1 Beta-actin-like protein 2 0 6.6438562 100 P49902 Cytosolic purine 5'-nucleotidase 0 6.6438562 100 Q9Y2H1 serine/threonine-protein kinase 38-like 0 6.6438562 100 Q9BQA1 Methylosome protein 50 0 6.6438562 100 Q9Y6Y0 influenza virus NS1A-binding protein 0 6.6438562 100 Q04917 14-3-3 protein eta 0 6.6438562 100 Q16769 glutaminyl-peptide cyclotransferase 0 6.6438562 100 P50213-1 Isocitrate dehydrogenase [NAD] subunit alpha, mitochondrial 0 6.6438562 100 Q15392 Delta(24)-sterol reductase 0 6.6438562 100 O60884 DnaJ homolog subfamily A member 2 0 6.6438562 100 P98175-2 Isoform 2 of RNA-binding protein 10 0 6.6438562 100 Q99460 26S proteasome non-ATPase regulatory subunit 1 0 6.6438562 100 Q13190 Syntaxin-5 0 6.6438562 100 Q9NYL9 tropomodulin-3 0 6.6438562 100 P23921 Ribonucleoside-diphosphate reductase large subunit 0 6.6438562 100 P12110 Collagen alpha-2(VI) chain 0 6.6438562 100 O96008 Mitochondrial import receptor subunit TOM40 homolog 0 6.6438562 100 P23258 tubulin gamma-1 chain 0 6.6438562 100 O15212 Prefoldin subunit 6 0 6.6438562 100 Q15363 Transmembrane emp24 domain-containing protein 2 0 6.6438562 100 126

Q14318-2 Isoform 2 of Peptidyl-prolyl cis-trans isomerase FKBP8 0 6.6438562 100 Q96DV4 39S ribosomal protein L38, mitochondrial 0 6.6438562 100 Q15181 Inorganic pyrophosphatase 0 6.6438562 100 Q15208 serine/threonine-protein kinase 38 0 3.1070182 8.616 O60313-2 Isoform 2 of Dynamin-like 120 kDa protein, mitochondrial 0 3.0292762 8.164 Q9UBS4 DnaJ homolog subfamily B member 11 0 1.9286546 3.807 Q13153 Serine/threonine-protein kinase PAK 1 0 1.9187678 3.781 P47756-1 F-actin-capping protein subunit beta 0 1.8698714 3.655 P31946 14-3-3 protein beta/alpha 0 1.7714631 3.414 P15056 serine/threonine-protein kinase B-raf 0 1.7337882 3.326 Q86TI2-2 Isoform 2 of Dipeptidyl peptidase 9 0 1.7202785 3.295 O75688-2 Isoform Beta-2 of Protein phosphatase 1B 0 1.6364506 3.109 Q71U36 tubulin alpha-1A chain 0 1.6257381 3.086 Q14257-2 Isoform 2 of Reticulocalbin-2 0 1.6144743 3.062 P36776 Lon protease homolog, mitochondrial 0 1.5420104 2.912 P52907 F-actin-capping protein subunit alpha-1 0 1.4992719 2.827 P40925-3 Isoform 3 of Malate dehydrogenase, cytoplasmic 0 1.4833644 2.796 Q14155 Rho guanine nucleotide exchange factor 7 0 1.4531224 2.738 Q13162 Peroxiredoxin-4 0 1.4206171 2.677 P05387 60S acidic ribosomal protein P2 0 1.3201958 2.497 Q9BR76 Coronin-1B 0 1.3062622 2.473 P46821 microtubule-associated protein 1B 0 1.2803625 2.429 P07900-2 Isoform 2 of Heat shock protein HSP 90-alpha 0 1.2636354 2.401 Q6UB35-1 Monofunctional C1-tetrahydrofolate synthase, mitochondrial 0 1.2454957 2.371 127

O00232-1 26s proteasome non-atpase regulatory subunit 12 0 1.2252749 2.338 O95163 Elongator complex protein 1 0 1.2159892 2.323 P06753-2 Isoform 2 of Tropomyosin alpha-3 chain 0 1.2141248 2.32 P40227-1 T-complex protein 1 subunit zeta 0 1.1769610 2.261 P62195-1 26S protease regulatory subunit 8 0 1.1615653 2.237 P09493-10 Isoform 10 of Tropomyosin alpha-1 chain 0 1.1609202 2.236 P08238 Heat shock protein HSP 90-beta 0 1.1583370 2.232 P07437 tubulin beta chain 0 1.1368476 2.199 P60709 Actin, cytoplasmic 1 0 1.1342209 2.195 P49368-1 T-complex protein 1 subunit gamma 0 1.1256511 2.182 P49327 Fatty acid synthase 0 1.1070182 2.154 P63104-1 14-3-3 protein zeta/delta 0 1.0989585 2.142 P68133 Actin, alpha skeletal muscle 0 1.0820213 2.117 P30041 Peroxiredoxin-6 0 1.0628125 2.089 Q99832 T-complex protein 1 subunit eta 0 1.0545012 2.077 P67936 Tropomyosin alpha-4 chain 0 1.0531113 2.075 P54105 Methylosome subunit pICln 0 1.0306892 2.043 P31948-2 Isoform 2 of Stress-induced-phosphoprotein 1 0 1.0143553 2.02

128

Appendix C: Sources of Funding

This work was supported by an NCI grant (1R01CA227847-01) to Dr. Matthew

Ringel, NIH cancer center grant (P30CA016058) to The Ohio State University, National

Institute of General Medical Sciences of the NIH Training Grant (T32GM068412) to The

Ohio State University College of Medicine and Christina Knippler, and a Pelotonia

Fellowship Program award to Christina Knippler.

129

References

1. J. L. Jameson, S. J. Mandel, A. P. Weetman, in Harrison's Principles of Internal Medicine, 20e, J. L. Jameson, A. S. Fauci, D. L. Kasper, S. L. Hauser, D. L. Longo, J. Loscalzo, Eds. (McGraw-Hill Education, New York, NY, 2018). 2. L. Enewold, K. Zhu, E. Ron, A. J. Marrogi, A. Stojadinovic, G. E. Peoples, S. S. Devesa, Rising thyroid cancer incidence in the United States by demographic and tumor characteristics, 1980-2005. Cancer Epidemiol Biomarkers Prev 18, 784- 791 (2009). 3. C. K. Jung, M. P. Little, J. H. Lubin, A. V. Brenner, S. A. Wells, Jr., A. J. Sigurdson, Y. E. Nikiforov, The increase in thyroid cancer incidence during the last four decades is accompanied by a high frequency of BRAF mutations and a sharp increase in RAS mutations. J Clin Endocrinol Metab 99, E276-285 (2014). 4. N. Howlader, A. M. Noone, M. Krapcho, D. Miller, A. Brest, M. Yu, J. Ruhl, Z. Tatalovich, A. Mariotto, D. R. Lewis, H. S. Chen, E. J. Feuer, K. A. Cronin, SEER Cancer Statistics Review, 1975-2016. based on November 2018 SEER data submission, posted to SEER website, April 2019, (2019). 5. L. G. Morris, R. M. Tuttle, L. Davies, Changing Trends in the Incidence of Thyroid Cancer in the United States. JAMA Otolaryngol Head Neck Surg 142, 709-711 (2016). 6. H. Lim, S. S. Devesa, J. A. Sosa, D. Check, C. M. Kitahara, Trends in Thyroid Cancer Incidence and Mortality in the United States, 1974-2013. JAMA 317, 1338-1348 (2017). 7. B. R. Haugen, E. K. Alexander, K. C. Bible, G. M. Doherty, S. J. Mandel, Y. E. Nikiforov, F. Pacini, G. W. Randolph, A. M. Sawka, M. Schlumberger, K. G. Schuff, S. I. Sherman, J. A. Sosa, D. L. Steward, R. M. Tuttle, L. Wartofsky, 2015 American Thyroid Association Management Guidelines for Adult Patients with Thyroid Nodules and Differentiated Thyroid Cancer: The American Thyroid Association Guidelines Task Force on Thyroid Nodules and Differentiated Thyroid Cancer. Thyroid 26, 1-133 (2016). 8. H. M. Rabes, E. P. Demidchik, J. D. Sidorow, E. Lengfelder, C. Beimfohr, D. Hoelzel, S. Klugbauer, Pattern of radiation-induced RET and NTRK1 rearrangements in 191 post-chernobyl papillary thyroid carcinomas: biological, phenotypic, and clinical implications. Clin Cancer Res 6, 1093-1103 (2000). 9. J. C. Ricarte-Filho, S. Li, M. E. Garcia-Rendueles, C. Montero-Conde, F. Voza, J. A. Knauf, A. Heguy, A. Viale, T. Bogdanova, G. A. Thomas, C. E. Mason, J. A. Fagin, Identification of kinase fusion oncogenes in post-Chernobyl radiation- induced thyroid cancers. J Clin Invest 123, 4935-4944 (2013). 10. S. Yamashita, S. Suzuki, S. Suzuki, H. Shimura, V. Saenko, Lessons from Fukushima: Latest Findings of Thyroid Cancer After the Fukushima Nuclear Power Plant Accident. Thyroid 28, 11-22 (2018). 11. J. A. Fagin, S. A. Wells, Jr., Biologic and Clinical Perspectives on Thyroid Cancer. N Engl J Med 375, 1054-1067 (2016). 130

12. H. Chen, R. Luthra, M. J. Routbort, K. P. Patel, M. E. Cabanillas, R. R. Broaddus, M. D. Williams, Molecular Profile of Advanced Thyroid Carcinomas by Next- Generation Sequencing: Characterizing Tumors Beyond Diagnosis for Targeted Therapy. Mol Cancer Ther 17, 1575-1584 (2018). 13. M. Pasca di Magliano, R. Di Lauro, M. Zannini, Pax8 has a key role in thyroid cell differentiation. Proc Natl Acad Sci U S A 97, 13144-13149 (2000). 14. C. Corzo, P. R. Griffin, Targeting the Peroxisome Proliferator-Activated Receptor-gamma to Counter the Inflammatory Milieu in Obesity. Diabetes Metab J 37, 395-403 (2013). 15. E. D. Rosen, P. Sarraf, A. E. Troy, G. Bradwin, K. Moore, D. S. Milstone, B. M. Spiegelman, R. M. Mortensen, PPAR gamma is required for the differentiation of adipose tissue in vivo and in vitro. Mol Cell 4, 611-617 (1999). 16. I. Landa, T. Ibrahimpasic, L. Boucai, R. Sinha, J. A. Knauf, R. H. Shah, S. Dogan, J. C. Ricarte-Filho, G. P. Krishnamoorthy, B. Xu, N. Schultz, M. F. Berger, C. Sander, B. S. Taylor, R. Ghossein, I. Ganly, J. A. Fagin, Genomic and transcriptomic hallmarks of poorly differentiated and anaplastic thyroid cancers. J Clin Invest 126, 1052-1066 (2016). 17. C. Durante, N. Haddy, E. Baudin, S. Leboulleux, D. Hartl, J. P. Travagli, B. Caillou, M. Ricard, J. D. Lumbroso, F. De Vathaire, M. Schlumberger, Long-term outcome of 444 patients with distant metastases from papillary and follicular thyroid carcinoma: benefits and limits of radioiodine therapy. J Clin Endocrinol Metab 91, 2892-2899 (2006). 18. M. S. Brose, C. M. Nutting, B. Jarzab, R. Elisei, S. Siena, L. Bastholt, C. de la Fouchardiere, F. Pacini, R. Paschke, Y. K. Shong, S. I. Sherman, J. W. Smit, J. Chung, C. Kappeler, C. Pena, I. Molnar, M. J. Schlumberger, D. investigators, Sorafenib in radioactive iodine-refractory, locally advanced or metastatic differentiated thyroid cancer: a randomised, double-blind, phase 3 trial. Lancet 384, 319-328 (2014). 19. M. Schlumberger, M. Tahara, L. J. Wirth, B. Robinson, M. S. Brose, R. Elisei, M. A. Habra, K. Newbold, M. H. Shah, A. O. Hoff, A. G. Gianoukakis, N. Kiyota, M. H. Taylor, S. B. Kim, M. K. Krzyzanowska, C. E. Dutcus, B. de las Heras, J. Zhu, S. I. Sherman, Lenvatinib versus placebo in radioiodine-refractory thyroid cancer. N Engl J Med 372, 621-630 (2015). 20. S. Haraldsdottir, M. H. Shah, An update on clinical trials of targeted therapies in thyroid cancer. Curr Opin Oncol 26, 36-44 (2014). 21. R. Laursen, M. Wehland, S. Kopp, J. Pietsch, M. Infanger, J. Grosse, D. Grimm, Effects and Role of Multikinase Inhibitors in Thyroid Cancer. Curr Pharm Des 22, 5915-5926 (2016). 22. M. J. Garnett, R. Marais, Guilty as charged: B-RAF is a human oncogene. Cancer Cell 6, 313-319 (2004). 23. M. A. Rahman, A. Salajegheh, R. A. Smith, A. K. Lam, B-Raf mutation: a key player in molecular biology of cancer. Exp Mol Pathol 95, 336-342 (2013). 24. M. J. Robinson, M. H. Cobb, Mitogen-activated protein kinase pathways. Curr Opin Cell Biol 9, 180-186 (1997). 131

25. B. H. Zhang, K. L. Guan, Activation of B-Raf kinase requires phosphorylation of the conserved residues Thr598 and Ser601. EMBO J 19, 5429-5439 (2000). 26. L. Goitre, E. Trapani, L. Trabalzini, S. F. Retta, The Ras superfamily of small GTPases: the unlocked secrets. Methods Mol Biol 1120, 1-18 (2014). 27. G. A. Hobbs, C. J. Der, K. L. Rossman, RAS isoforms and mutations in cancer at a glance. J Cell Sci 129, 1287-1292 (2016). 28. D. K. Simanshu, D. V. Nissley, F. McCormick, RAS Proteins and Their Regulators in Human Disease. Cell 170, 17-33 (2017). 29. R. T. Liu, C. Y. Hou, H. L. You, C. C. Huang, L. Hock, F. F. Chou, P. W. Wang, J. T. Cheng, Selective occurrence of ras mutations in benign and malignant thyroid follicular neoplasms in Taiwan. Thyroid 14, 616-621 (2004). 30. V. Vasko, M. Ferrand, J. Di Cristofaro, P. Carayon, J. F. Henry, C. de Micco, Specific pattern of RAS oncogene mutations in follicular thyroid tumors. J Clin Endocrinol Metab 88, 2745-2752 (2003). 31. U. R. Rapp, M. D. Goldsborough, G. E. Mark, T. I. Bonner, J. Groffen, F. H. Reynolds, Jr., J. R. Stephenson, Structure and biological activity of v-raf, a unique oncogene transduced by a retrovirus. Proc Natl Acad Sci U S A 80, 4218-4222 (1983). 32. T. I. Bonner, S. B. Kerby, P. Sutrave, M. A. Gunnell, G. Mark, U. R. Rapp, Structure and biological activity of human homologs of the raf/mil oncogene. Mol Cell Biol 5, 1400-1407 (1985). 33. K. Huebner, A. ar-Rushdi, C. A. Griffin, M. Isobe, C. Kozak, B. S. Emanuel, L. Nagarajan, J. L. Cleveland, T. I. Bonner, M. D. Goldsborough, et al., Actively transcribed genes in the raf oncogene group, located on the X chromosome in mouse and human. Proc Natl Acad Sci U S A 83, 3934-3938 (1986). 34. S. Ikawa, M. Fukui, Y. Ueyama, N. Tamaoki, T. Yamamoto, K. Toyoshima, B- raf, a new member of the raf family, is activated by DNA rearrangement. Mol Cell Biol 8, 2651-2654 (1988). 35. R. Roskoski, Jr., RAF protein-serine/threonine kinases: structure and regulation. Biochem Biophys Res Commun 399, 313-317 (2010). 36. C. S. Mason, C. J. Springer, R. G. Cooper, G. Superti-Furga, C. J. Marshall, R. Marais, Serine and tyrosine phosphorylations cooperate in Raf-1, but not B-Raf activation. EMBO J 18, 2137-2148 (1999). 37. L. K. Rushworth, A. D. Hindley, E. O'Neill, W. Kolch, Regulation and role of Raf-1/B-Raf heterodimerization. Mol Cell Biol 26, 2262-2272 (2006). 38. J. R. Haling, J. Sudhamsu, I. Yen, S. Sideris, W. Sandoval, W. Phung, B. J. Bravo, A. M. Giannetti, A. Peck, A. Masselot, T. Morales, D. Smith, B. J. Brandhuber, S. G. Hymowitz, S. Malek, Structure of the BRAF-MEK complex reveals a kinase activity independent role for BRAF in MAPK signaling. Cancer Cell 26, 402-413 (2014). 39. H. Davies, G. R. Bignell, C. Cox, P. Stephens, S. Edkins, S. Clegg, J. Teague, H. Woffendin, M. J. Garnett, W. Bottomley, N. Davis, E. Dicks, R. Ewing, Y. Floyd, K. Gray, S. Hall, R. Hawes, J. Hughes, V. Kosmidou, A. Menzies, C. Mould, A. Parker, C. Stevens, S. Watt, S. Hooper, R. Wilson, H. Jayatilake, B. A. Gusterson, 132

C. Cooper, J. Shipley, D. Hargrave, K. Pritchard-Jones, N. Maitland, G. Chenevix-Trench, G. J. Riggins, D. D. Bigner, G. Palmieri, A. Cossu, A. Flanagan, A. Nicholson, J. W. Ho, S. Y. Leung, S. T. Yuen, B. L. Weber, H. F. Seigler, T. L. Darrow, H. Paterson, R. Marais, C. J. Marshall, R. Wooster, M. R. Stratton, P. A. Futreal, Mutations of the BRAF gene in human cancer. Nature 417, 949-954 (2002). 40. N. Cancer Genome Atlas Research, Integrated genomic characterization of papillary thyroid carcinoma. Cell 159, 676-690 (2014). 41. P. I. Poulikakos, Y. Persaud, M. Janakiraman, X. Kong, C. Ng, G. Moriceau, H. Shi, M. Atefi, B. Titz, M. T. Gabay, M. Salton, K. B. Dahlman, M. Tadi, J. A. Wargo, K. T. Flaherty, M. C. Kelley, T. Misteli, P. B. Chapman, J. A. Sosman, T. G. Graeber, A. Ribas, R. S. Lo, N. Rosen, D. B. Solit, RAF inhibitor resistance is mediated by dimerization of aberrantly spliced BRAF(V600E). Nature 480, 387- 390 (2011). 42. P. T. Wan, M. J. Garnett, S. M. Roe, S. Lee, D. Niculescu-Duvaz, V. M. Good, C. M. Jones, C. J. Marshall, C. J. Springer, D. Barford, R. Marais, P. Cancer Genome, Mechanism of activation of the RAF-ERK signaling pathway by oncogenic mutations of B-RAF. Cell 116, 855-867 (2004). 43. Z. Yao, N. M. Torres, A. Tao, Y. Gao, L. Luo, Q. Li, E. de Stanchina, O. Abdel- Wahab, D. B. Solit, P. I. Poulikakos, N. Rosen, BRAF Mutants Evade ERK- Dependent Feedback by Different Mechanisms that Determine Their Sensitivity to Pharmacologic Inhibition. Cancer Cell 28, 370-383 (2015). 44. C. A. Pratilas, B. S. Taylor, Q. Ye, A. Viale, C. Sander, D. B. Solit, N. Rosen, (V600E)BRAF is associated with disabled feedback inhibition of RAF-MEK signaling and elevated transcriptional output of the pathway. Proc Natl Acad Sci U S A 106, 4519-4524 (2009). 45. E. W. Joseph, C. A. Pratilas, P. I. Poulikakos, M. Tadi, W. Wang, B. S. Taylor, E. Halilovic, Y. Persaud, F. Xing, A. Viale, J. Tsai, P. B. Chapman, G. Bollag, D. B. Solit, N. Rosen, The RAF inhibitor PLX4032 inhibits ERK signaling and tumor cell proliferation in a V600E BRAF-selective manner. Proc Natl Acad Sci U S A 107, 14903-14908 (2010). 46. P. B. Chapman, A. Hauschild, C. Robert, J. B. Haanen, P. Ascierto, J. Larkin, R. Dummer, C. Garbe, A. Testori, M. Maio, D. Hogg, P. Lorigan, C. Lebbe, T. Jouary, D. Schadendorf, A. Ribas, S. J. O'Day, J. A. Sosman, J. M. Kirkwood, A. M. Eggermont, B. Dreno, K. Nolop, J. Li, B. Nelson, J. Hou, R. J. Lee, K. T. Flaherty, G. A. McArthur, B.-S. Group, Improved survival with vemurafenib in melanoma with BRAF V600E mutation. N Engl J Med 364, 2507-2516 (2011). 47. A. Hauschild, J. J. Grob, L. V. Demidov, T. Jouary, R. Gutzmer, M. Millward, P. Rutkowski, C. U. Blank, W. H. Miller, Jr., E. Kaempgen, S. Martin-Algarra, B. Karaszewska, C. Mauch, V. Chiarion-Sileni, A. M. Martin, S. Swann, P. Haney, B. Mirakhur, M. E. Guckert, V. Goodman, P. B. Chapman, Dabrafenib in BRAF- mutated metastatic melanoma: a multicentre, open-label, phase 3 randomised controlled trial. Lancet 380, 358-365 (2012).

133

48. G. S. Falchook, M. Millward, D. Hong, A. Naing, S. Piha-Paul, S. G. Waguespack, M. E. Cabanillas, S. I. Sherman, B. Ma, M. Curtis, V. Goodman, R. Kurzrock, BRAF inhibitor dabrafenib in patients with metastatic BRAF-mutant thyroid cancer. Thyroid 25, 71-77 (2015). 49. V. Subbiah, R. J. Kreitman, Z. A. Wainberg, J. Y. Cho, J. H. M. Schellens, J. C. Soria, P. Y. Wen, C. Zielinski, M. E. Cabanillas, G. Urbanowitz, B. Mookerjee, D. Wang, F. Rangwala, B. Keam, Dabrafenib and Trametinib Treatment in Patients With Locally Advanced or Metastatic BRAF V600-Mutant Anaplastic Thyroid Cancer. J Clin Oncol 36, 7-13 (2018). 50. K. B. Kim, M. E. Cabanillas, A. J. Lazar, M. D. Williams, D. L. Sanders, J. L. Ilagan, K. Nolop, R. J. Lee, S. I. Sherman, Clinical responses to vemurafenib in patients with metastatic papillary thyroid cancer harboring BRAF(V600E) mutation. Thyroid 23, 1277-1283 (2013). 51. M. S. Brose, M. E. Cabanillas, E. E. Cohen, L. J. Wirth, T. Riehl, H. Yue, S. I. Sherman, E. J. Sherman, Vemurafenib in patients with BRAF(V600E)-positive metastatic or unresectable papillary thyroid cancer refractory to radioactive iodine: a non-randomised, multicentre, open-label, phase 2 trial. Lancet Oncol 17, 1272-1282 (2016). 52. P. Lito, N. Rosen, D. B. Solit, Tumor adaptation and resistance to RAF inhibitors. Nat Med 19, 1401-1409 (2013). 53. C. Sun, L. Wang, S. Huang, G. J. Heynen, A. Prahallad, C. Robert, J. Haanen, C. Blank, J. Wesseling, S. M. Willems, D. Zecchin, S. Hobor, P. K. Bajpe, C. Lieftink, C. Mateus, S. Vagner, W. Grernrum, I. Hofland, A. Schlicker, L. F. Wessels, R. L. Beijersbergen, A. Bardelli, F. Di Nicolantonio, A. M. Eggermont, R. Bernards, Reversible and adaptive resistance to BRAF(V600E) inhibition in melanoma. Nature 508, 118-122 (2014). 54. R. Nazarian, H. Shi, Q. Wang, X. Kong, R. C. Koya, H. Lee, Z. Chen, M. K. Lee, N. Attar, H. Sazegar, T. Chodon, S. F. Nelson, G. McArthur, J. A. Sosman, A. Ribas, R. S. Lo, Melanomas acquire resistance to B-RAF(V600E) inhibition by RTK or N-RAS upregulation. Nature 468, 973-977 (2010). 55. M. Lidsky, G. Antoun, P. Speicher, B. Adams, R. Turley, C. Augustine, D. Tyler, F. Ali-Osman, Mitogen-activated protein kinase (MAPK) hyperactivation and enhanced NRAS expression drive acquired vemurafenib resistance in V600E BRAF melanoma cells. J Biol Chem 289, 27714-27726 (2014). 56. L. Boussemart, H. Malka-Mahieu, I. Girault, D. Allard, O. Hemmingsson, G. Tomasic, M. Thomas, C. Basmadjian, N. Ribeiro, F. Thuaud, C. Mateus, E. Routier, N. Kamsu-Kom, S. Agoussi, A. M. Eggermont, L. Desaubry, C. Robert, S. Vagner, eIF4F is a nexus of resistance to anti-BRAF and anti-MEK cancer therapies. Nature 513, 105-109 (2014). 57. C. Montero-Conde, S. Ruiz-Llorente, J. M. Dominguez, J. A. Knauf, A. Viale, E. J. Sherman, M. Ryder, R. A. Ghossein, N. Rosen, J. A. Fagin, Relief of feedback inhibition of HER3 transcription by RAF and MEK inhibitors attenuates their antitumor effects in BRAF-mutant thyroid carcinomas. Cancer Discov 3, 520-533 (2013). 134

58. H. K. Byeon, H. J. Na, Y. J. Yang, S. Ko, S. O. Yoon, M. Ku, J. Yang, J. W. Kim, M. J. Ban, J. H. Kim, D. H. Kim, J. M. Kim, E. C. Choi, C. H. Kim, J. H. Yoon, Y. W. Koh, Acquired resistance to BRAF inhibition induces epithelial-to- mesenchymal transition in BRAF (V600E) mutant thyroid cancer by c-Met- mediated AKT activation. Oncotarget 8, 596-609 (2017). 59. H. K. Byeon, H. J. Na, Y. J. Yang, H. J. Kwon, J. W. Chang, M. J. Ban, W. S. Kim, D. Y. Shin, E. J. Lee, Y. W. Koh, J. H. Yoon, E. C. Choi, c-Met-mediated reactivation of PI3K/AKT signaling contributes to insensitivity of BRAF(V600E) mutant thyroid cancer to BRAF inhibition. Mol Carcinog 55, 1678-1687 (2016). 60. J. A. Knauf, K. A. Luckett, K. Y. Chen, F. Voza, N. D. Socci, R. Ghossein, J. A. Fagin, Hgf/Met activation mediates resistance to BRAF inhibition in murine anaplastic thyroid cancers. J Clin Invest 128, 4086-4097 (2018). 61. T. Ofir Dovrat, E. Sokol, G. Frampton, E. Shachar, S. Pelles, R. Geva, I. Wolf, Unusually long-term responses to vemurafenib in BRAF V600E mutated colon and thyroid cancers followed by the development of rare RAS activating mutations. Cancer Biology & Therapy 19, 871-874 (2018). 62. P. Lito, C. A. Pratilas, E. W. Joseph, M. Tadi, E. Halilovic, M. Zubrowski, A. Huang, W. L. Wong, M. K. Callahan, T. Merghoub, J. D. Wolchok, E. de Stanchina, S. Chandarlapaty, P. I. Poulikakos, J. A. Fagin, N. Rosen, Relief of profound feedback inhibition of mitogenic signaling by RAF inhibitors attenuates their activity in BRAFV600E melanomas. Cancer Cell 22, 668-682 (2012). 63. P. I. Poulikakos, C. Zhang, G. Bollag, K. M. Shokat, N. Rosen, RAF inhibitors transactivate RAF dimers and ERK signalling in cells with wild-type BRAF. Nature 464, 427-430 (2010). 64. S. I. Sherman, J. D. Brierley, M. Sperling, K. B. Ain, S. T. Bigos, D. S. Cooper, B. R. Haugen, M. Ho, I. Klein, P. W. Ladenson, J. Robbins, D. S. Ross, B. Specker, T. Taylor, H. R. Maxon, 3rd, Prospective multicenter study of thyroiscarcinoma treatment: initial analysis of staging and outcome. National Thyroid Cancer Treatment Cooperative Study Registry Group. Cancer 83, 1012- 1021 (1998). 65. V. Vasko, A. V. Espinosa, W. Scouten, H. He, H. Auer, S. Liyanarachchi, A. Larin, V. Savchenko, G. L. Francis, A. de la Chapelle, M. Saji, M. D. Ringel, Gene expression and functional evidence of epithelial-to-mesenchymal transition in papillary thyroid carcinoma invasion. Proc Natl Acad Sci U S A 104, 2803- 2808 (2007). 66. M. A. Huber, N. Kraut, H. Beug, Molecular requirements for epithelial- mesenchymal transition during tumor progression. Curr Opin Cell Biol 17, 548- 558 (2005). 67. E. Manser, T. Leung, H. Salihuddin, Z. S. Zhao, L. Lim, A brain serine/threonine protein kinase activated by Cdc42 and Rac1. Nature 367, 40-46 (1994). 68. L. E. Arias-Romero, J. Chernoff, A tale of two Paks. Biol Cell 100, 97-108 (2008). 69. Z. M. Jaffer, J. Chernoff, p21-activated kinases: three more join the Pak. Int J Biochem Cell Biol 34, 713-717 (2002). 135

70. S. Cotteret, Z. M. Jaffer, A. Beeser, J. Chernoff, p21-Activated kinase 5 (Pak5) localizes to mitochondria and inhibits apoptosis by phosphorylating BAD. Mol Cell Biol 23, 5526-5539 (2003). 71. M. Lei, W. Lu, W. Meng, M. C. Parrini, M. J. Eck, B. J. Mayer, S. C. Harrison, Structure of PAK1 in an autoinhibited conformation reveals a multistage activation switch. Cell 102, 387-397 (2000). 72. J. Wang, J. W. Wu, Z. X. Wang, Structural insights into the autoactivation mechanism of p21-activated protein kinase. Structure 19, 1752-1761 (2011). 73. F. J. Sorrell, L. M. Kilian, J. M. Elkins, Solution structures and biophysical analysis of full-length group A PAKs reveal they are monomeric and autoinhibited in cis. Biochem J 476, 1037-1051 (2019). 74. G. A. Martin, G. Bollag, F. McCormick, A. Abo, A novel serine kinase activated by rac1/CDC42Hs-dependent autophosphorylation is related to PAK65 and STE20. EMBO J 14, 1970-1978 (1995). 75. G. E. Benner, P. B. Dennis, R. A. Masaracchia, Activation of an S6/H4 kinase (PAK 65) from human placenta by intramolecular and intermolecular autophosphorylation. J Biol Chem 270, 21121-21128 (1995). 76. C. G. Koh, E. J. Tan, E. Manser, L. Lim, The p21-activated kinase PAK is negatively regulated by POPX1 and POPX2, a pair of serine/threonine phosphatases of the PP2C family. Curr Biol 12, 317-321 (2002). 77. R. S. Westphal, R. L. Coffee, Jr., A. Marotta, S. L. Pelech, B. E. Wadzinski, Identification of kinase-phosphatase signaling modules composed of p70 S6 kinase-protein phosphatase 2A (PP2A) and p21-activated kinase-PP2A. J Biol Chem 274, 687-692 (1999). 78. R. Kumar, R. Sanawar, X. Li, F. Li, Structure, biochemistry, and biology of PAK kinases. Gene 605, 20-31 (2017). 79. C. D. Lawson, K. Burridge, The on-off relationship of Rho and Rac during integrin-mediated adhesion and cell migration. Small GTPases 5, e27958 (2014). 80. N. M. Baker, H. Yee Chow, J. Chernoff, C. J. Der, Molecular pathways: targeting RAC-p21-activated serine-threonine kinase signaling in RAS-driven cancers. Clin Cancer Res 20, 4740-4746 (2014). 81. G. M. Bokoch, Biology of the p21-activated kinases. Annu Rev Biochem 72, 743- 781 (2003). 82. R. R. Singh, C. Song, Z. Yang, R. Kumar, Nuclear localization and chromatin targets of p21-activated kinase 1. J Biol Chem 280, 18130-18137 (2005). 83. E. Manser, H. Y. Huang, T. H. Loo, X. Q. Chen, J. M. Dong, T. Leung, L. Lim, Expression of constitutively active alpha-PAK reveals effects of the kinase on actin and focal complexes. Mol Cell Biol 17, 1129-1143 (1997). 84. M. A. Sells, U. G. Knaus, S. Bagrodia, D. M. Ambrose, G. M. Bokoch, J. Chernoff, Human p21-activated kinase (Pak1) regulates actin organization in mammalian cells. Curr Biol 7, 202-210 (1997). 85. L. C. Sanders, F. Matsumura, G. M. Bokoch, P. de Lanerolle, Inhibition of myosin light chain kinase by p21-activated kinase. Science 283, 2083-2085 (1999). 136

86. D. C. Edwards, L. C. Sanders, G. M. Bokoch, G. N. Gill, Activation of LIM- kinase by Pak1 couples Rac/Cdc42 GTPase signalling to actin cytoskeletal dynamics. Nature cell biology 1, 253-259 (1999). 87. R. K. Vadlamudi, C. J. Barnes, S. Rayala, F. Li, S. Balasenthil, S. Marcus, H. V. Goodson, A. A. Sahin, R. Kumar, p21-activated kinase 1 regulates microtubule dynamics by phosphorylating tubulin cofactor B. Mol Cell Biol 25, 3726-3736 (2005). 88. T. Wittmann, G. M. Bokoch, C. M. Waterman-Storer, Regulation of microtubule destabilizing activity of Op18/stathmin downstream of Rac1. J Biol Chem 279, 6196-6203 (2004). 89. Q. F. Li, A. M. Spinelli, R. Wang, Y. Anfinogenova, H. A. Singer, D. D. Tang, Critical role of vimentin phosphorylation at Ser-56 by p21-activated kinase in vimentin cytoskeleton signaling. J Biol Chem 281, 34716-34724 (2006). 90. A. Hammer, P. Oladimeji, L. E. De Las Casas, M. Diakonova, Phosphorylation of tyrosine 285 of PAK1 facilitates betaPIX/GIT1 binding and adhesion turnover. FASEB J 29, 943-959 (2015). 91. C. E. Turner, M. C. Brown, J. A. Perrotta, M. C. Riedy, S. N. Nikolopoulos, A. R. McDonald, S. Bagrodia, S. Thomas, P. S. Leventhal, Paxillin LD4 motif binds PAK and PIX through a novel 95-kD ankyrin repeat, ARF-GAP protein: A role in cytoskeletal remodeling. J Cell Biol 145, 851-863 (1999). 92. J. K. Slack-Davis, S. T. Eblen, M. Zecevic, S. A. Boerner, A. Tarcsafalvi, H. B. Diaz, M. S. Marshall, M. J. Weber, J. T. Parsons, A. D. Catling, PAK1 phosphorylation of MEK1 regulates fibronectin-stimulated MAPK activation. J Cell Biol 162, 281-291 (2003). 93. J. A. Frost, H. Steen, P. Shapiro, T. Lewis, N. Ahn, P. E. Shaw, M. H. Cobb, Cross-cascade activation of ERKs and ternary complex factors by Rho family proteins. EMBO J 16, 6426-6438 (1997). 94. L. C. Coles, P. E. Shaw, PAK1 primes MEK1 for phosphorylation by Raf-1 kinase during cross-cascade activation of the ERK pathway. Oncogene 21, 2236- 2244 (2002). 95. E. R. Park, S. T. Eblen, A. D. Catling, MEK1 activation by PAK: a novel mechanism. Cell Signal 19, 1488-1496 (2007). 96. Z. Wang, M. Fu, L. Wang, J. Liu, Y. Li, C. Brakebusch, Q. Mei, p21-activated kinase 1 (PAK1) can promote ERK activation in a kinase-independent manner. The Journal of biological chemistry 288, 20093-20099 (2013). 97. L. J. Sundberg-Smith, J. T. Doherty, C. P. Mack, J. M. Taylor, Adhesion stimulates direct PAK1/ERK2 association and leads to ERK-dependent PAK1 Thr212 phosphorylation. J Biol Chem 280, 2055-2064 (2005). 98. A. Chaudhary, W. G. King, M. D. Mattaliano, J. A. Frost, B. Diaz, D. K. Morrison, M. H. Cobb, M. S. Marshall, J. S. Brugge, Phosphatidylinositol 3- kinase regulates Raf1 through Pak phosphorylation of serine 338. Curr Biol 10, 551-554 (2000).

137

99. A. J. King, H. Sun, B. Diaz, D. Barnard, W. Miao, S. Bagrodia, M. S. Marshall, The protein kinase Pak3 positively regulates Raf-1 activity through phosphorylation of serine 338. Nature 396, 180-183 (1998). 100. M. Zang, C. Hayne, Z. Luo, Interaction between active Pak1 and Raf-1 is necessary for phosphorylation and activation of Raf-1. J Biol Chem 277, 4395- 4405 (2002). 101. M. Higuchi, K. Onishi, C. Kikuchi, Y. Gotoh, Scaffolding function of PAK in the PDK1-Akt pathway. Nature cell biology 10, 1356-1364 (2008). 102. M. Banerjee, D. Worth, D. M. Prowse, M. Nikolic, Pak1 phosphorylation on t212 affects microtubules in cells undergoing mitosis. Curr Biol 12, 1233-1239 (2002). 103. F. Li, L. Adam, R. K. Vadlamudi, H. Zhou, S. Sen, J. Chernoff, M. Mandal, R. Kumar, p21-activated kinase 1 interacts with and phosphorylates histone H3 in breast cancer cells. EMBO Rep 3, 767-773 (2002). 104. H. J. Whalley, A. P. Porter, Z. Diamantopoulou, G. R. White, E. Castaneda- Saucedo, A. Malliri, Cdk1 phosphorylates the Rac activator Tiam1 to activate centrosomal Pak and promote mitotic spindle formation. Nat Commun 6, 7437 (2015). 105. Z. S. Zhao, J. P. Lim, Y. W. Ng, L. Lim, E. Manser, The GIT-associated kinase PAK targets to the centrosome and regulates Aurora-A. Mol Cell 20, 237-249 (2005). 106. B. Maroto, M. B. Ye, K. von Lohneysen, A. Schnelzer, U. G. Knaus, P21- activated kinase is required for mitotic progression and regulates Plk1. Oncogene 27, 4900-4908 (2008). 107. M. Radu, G. Semenova, R. Kosoff, J. Chernoff, PAK signalling during the development and progression of cancer. Nature reviews. Cancer 14, 13-25 (2014). 108. D. Z. Ye, J. Field, PAK signaling in cancer. Cell Logist 2, 105-116 (2012). 109. D. Araiza-Olivera, Y. Feng, G. Semenova, T. Y. Prudnikova, J. Rhodes, J. Chernoff, Suppression of RAC1-driven malignant melanoma by group A PAK inhibitors. Oncogene 37, 944-952 (2018). 110. Y. Tang, S. Marwaha, J. L. Rutkowski, G. I. Tennekoon, P. C. Phillips, J. Field, A role for Pak protein kinases in Schwann cell transformation. Proc Natl Acad Sci U S A 95, 5139-5144 (1998). 111. Y. Hirokawa, A. Tikoo, J. Huynh, T. Utermark, C. O. Hanemann, M. Giovannini, G. H. Xiao, J. R. Testa, J. Wood, H. Maruta, A clue to the therapy of neurofibromatosis type 2: NF2/merlin is a PAK1 inhibitor. Cancer J 10, 20-26 (2004). 112. J. L. Kissil, E. W. Wilker, K. C. Johnson, M. S. Eckman, M. B. Yaffe, T. Jacks, Merlin, the product of the Nf2 tumor suppressor gene, is an inhibitor of the p21- activated kinase, Pak1. Mol Cell 12, 841-849 (2003). 113. G. Semenova, D. S. Stepanova, C. Dubyk, E. Handorf, S. M. Deyev, A. J. Lazar, J. Chernoff, Targeting group I p21-activated kinases to control malignant peripheral nerve sheath tumor growth and metastasis. Oncogene 36, 5421-5431 (2017).

138

114. H. Y. Chow, B. Dong, S. G. Duron, D. A. Campbell, C. C. Ong, K. P. Hoeflich, L. S. Chang, D. B. Welling, Z. J. Yang, J. Chernoff, Group I Paks as therapeutic targets in NF2-deficient meningioma. Oncotarget 6, 1981-1994 (2015). 115. C. C. Ong, A. M. Jubb, D. Jakubiak, W. Zhou, J. Rudolph, P. M. Haverty, M. Kowanetz, Y. Yan, J. Tremayne, R. Lisle, A. L. Harris, L. S. Friedman, M. Belvin, M. R. Middleton, E. M. Blackwood, H. Koeppen, K. P. Hoeflich, P21- activated kinase 1 (PAK1) as a therapeutic target in BRAF wild-type melanoma. J Natl Cancer Inst 105, 606-607 (2013). 116. C. C. Ong, S. Gierke, C. Pitt, M. Sagolla, C. K. Cheng, W. Zhou, A. M. Jubb, L. Strickland, M. Schmidt, S. G. Duron, D. A. Campbell, W. Zheng, S. Dehdashti, M. Shen, N. Yang, M. L. Behnke, W. Huang, J. C. McKew, J. Chernoff, W. F. Forrest, P. M. Haverty, S. F. Chin, E. A. Rakha, A. R. Green, I. O. Ellis, C. Caldas, T. O'Brien, L. S. Friedman, H. Koeppen, J. Rudolph, K. P. Hoeflich, Small molecule inhibition of group I p21-activated kinases in breast cancer induces apoptosis and potentiates the activity of microtubule stabilizing agents. Breast Cancer Res 17, 59 (2015). 117. S. Bekri, J. Adelaide, S. Merscher, J. Grosgeorge, F. Caroli-Bosc, D. Perucca- Lostanlen, P. M. Kelley, M. J. Pebusque, C. Theillet, D. Birnbaum, P. Gaudray, Detailed map of a region commonly amplified at 11q13-->q14 in human breast carcinoma. Cytogenet Cell Genet 79, 125-131 (1997). 118. J. Bostner, M. Ahnstrom Waltersson, T. Fornander, L. Skoog, B. Nordenskjold, O. Stal, Amplification of CCND1 and PAK1 as predictors of recurrence and tamoxifen resistance in postmenopausal breast cancer. Oncogene 26, 6997-7005 (2007). 119. S. Balasenthil, A. A. Sahin, C. J. Barnes, R. A. Wang, R. G. Pestell, R. K. Vadlamudi, R. Kumar, p21-activated kinase-1 signaling mediates cyclin D1 expression in mammary epithelial and cancer cells. J Biol Chem 279, 1422-1428 (2004). 120. C. Holm, S. Rayala, K. Jirstrom, O. Stal, R. Kumar, G. Landberg, Association between Pak1 expression and subcellular localization and tamoxifen resistance in breast cancer patients. J Natl Cancer Inst 98, 671-680 (2006). 121. L. A. Brown, S. E. Kalloger, M. A. Miller, M. Shih Ie, S. E. McKinney, J. L. Santos, K. Swenerton, P. T. Spellman, J. Gray, C. B. Gilks, D. G. Huntsman, Amplification of 11q13 in ovarian carcinoma. Genes Chromosomes Cancer 47, 481-489 (2008). 122. P. Schraml, G. Schwerdtfeger, F. Burkhalter, A. Raggi, D. Schmidt, T. Ruffalo, W. King, K. Wilber, M. J. Mihatsch, H. Moch, Combined array comparative genomic hybridization and tissue microarray analysis suggest PAK1 at 11q13.5- q14 as a critical oncogene target in ovarian carcinoma. Am J Pathol 163, 985-992 (2003). 123. B. Davidson, M. Shih Ie, T. L. Wang, Different clinical roles for p21-activated kinase-1 in primary and recurrent ovarian carcinoma. Hum Pathol 39, 1630-1636 (2008).

139

124. J. H. Carter, L. E. Douglass, J. A. Deddens, B. M. Colligan, T. R. Bhatt, J. O. Pemberton, S. Konicek, J. Hom, M. Marshall, J. R. Graff, Pak-1 expression increases with progression of colorectal carcinomas to metastasis. Clin Cancer Res 10, 3448-3456 (2004). 125. W. Zhou, A. M. Jubb, K. Lyle, Q. Xiao, C. C. Ong, R. Desai, L. Fu, F. Gnad, Q. Song, P. M. Haverty, D. Aust, R. Grutzmann, M. Romero, K. Totpal, R. M. Neve, Y. Yan, W. F. Forrest, Y. Wang, R. Raja, C. Pilarsky, A. de Jesus-Acosta, M. Belvin, L. S. Friedman, M. Merchant, E. M. Jaffee, L. Zheng, H. Koeppen, K. P. Hoeflich, PAK1 mediates pancreatic cancer cell migration and resistance to MET inhibition. J Pathol 234, 502-513 (2014). 126. H. Aoki, T. Yokoyama, K. Fujiwara, A. M. Tari, R. Sawaya, D. Suki, K. R. Hess, K. D. Aldape, S. Kondo, R. Kumar, Y. Kondo, Phosphorylated Pak1 level in the cytoplasm correlates with shorter survival time in patients with glioblastoma. Clin Cancer Res 13, 6603-6609 (2007). 127. S. K. McCarty, M. Saji, X. Zhang, D. Jarjoura, A. Fusco, V. V. Vasko, M. D. Ringel, Group I p21-activated kinases regulate thyroid cancer cell migration and are overexpressed and activated in thyroid cancer invasion. Endocr Relat Cancer 17, 989-999 (2010). 128. D. Hanahan, R. A. Weinberg, The hallmarks of cancer. Cell 100, 57-70 (2000). 129. Z. S. Zhao, E. Manser, X. Q. Chen, C. Chong, T. Leung, L. Lim, A conserved negative regulatory region in alphaPAK: inhibition of PAK kinases reveals their morphological roles downstream of Cdc42 and Rac1. Mol Cell Biol 18, 2153- 2163 (1998). 130. H. Y. Chow, B. Dong, C. A. Valencia, C. T. Zeng, J. N. Koch, T. Y. Prudnikova, J. Chernoff, Group I Paks are essential for epithelial- mesenchymal transition in an Apc-driven model of colorectal cancer. Nat Commun 9, 3473 (2018). 131. S. W. Deacon, A. Beeser, J. A. Fukui, U. E. Rennefahrt, C. Myers, J. Chernoff, J. R. Peterson, An isoform-selective, small-molecule inhibitor targets the autoregulatory mechanism of p21-activated kinase. Chem Biol 15, 322-331 (2008). 132. D. J. Kim, C. K. Choi, C. S. Lee, M. H. Park, X. Tian, N. D. Kim, K. I. Lee, J. K. Choi, J. H. Ahn, E. Y. Shin, I. Shin, E. G. Kim, Small molecules that allosterically inhibit p21-activated kinase activity by binding to the regulatory p21-binding domain. Exp Mol Med 48, e229 (2016). 133. U. T. Ruegg, G. M. Burgess, Staurosporine, K-252 and UCN-01: potent but nonspecific inhibitors of protein kinases. Trends Pharmacol Sci 10, 218-220 (1989). 134. M. W. Karaman, S. Herrgard, D. K. Treiber, P. Gallant, C. E. Atteridge, B. T. Campbell, K. W. Chan, P. Ciceri, M. I. Davis, P. T. Edeen, R. Faraoni, M. Floyd, J. P. Hunt, D. J. Lockhart, Z. V. Milanov, M. J. Morrison, G. Pallares, H. K. Patel, S. Pritchard, L. M. Wodicka, P. P. Zarrinkar, A quantitative analysis of kinase inhibitor selectivity. Nat Biotechnol 26, 127-132 (2008). 135. T. V. Nheu, H. He, Y. Hirokawa, K. Tamaki, L. Florin, M. L. Schmitz, I. Suzuki- Takahashi, R. N. Jorissen, A. W. Burgess, S. Nishimura, J. Wood, H. Maruta, The 140

K252a derivatives, inhibitors for the PAK/MLK kinase family selectively block the growth of RAS transformants. Cancer J 8, 328-336 (2002). 136. C. Guo, I. McAlpine, J. Zhang, D. D. Knighton, S. Kephart, M. C. Johnson, H. Li, D. Bouzida, A. Yang, L. Dong, J. Marakovits, J. Tikhe, P. Richardson, L. C. Guo, R. Kania, M. P. Edwards, E. Kraynov, J. Christensen, J. Piraino, J. Lee, E. Dagostino, C. Del-Carmen, Y. L. Deng, T. Smeal, B. W. Murray, Discovery of pyrroloaminopyrazoles as novel PAK inhibitors. J Med Chem 55, 4728-4739 (2012). 137. B. W. Murray, C. Guo, J. Piraino, J. K. Westwick, C. Zhang, J. Lamerdin, E. Dagostino, D. Knighton, C. M. Loi, M. Zager, E. Kraynov, I. Popoff, J. G. Christensen, R. Martinez, S. E. Kephart, J. Marakovits, S. Karlicek, S. Bergqvist, T. Smeal, Small-molecule p21-activated kinase inhibitor PF-3758309 is a potent inhibitor of oncogenic signaling and tumor growth. Proc Natl Acad Sci U S A 107, 9446-9451 (2010). 138. T. M. Pitts, G. N. Kulikowski, A. C. Tan, B. W. Murray, J. J. Arcaroli, J. J. Tentler, A. Spreafico, H. M. Selby, M. I. Kachaeva, K. L. McPhillips, B. C. Britt, E. L. Bradshaw-Pierce, W. A. Messersmith, M. Varella-Garcia, S. G. Eckhardt, Association of the epithelial-to-mesenchymal transition phenotype with responsiveness to the p21-activated kinase inhibitor, PF-3758309, in colon cancer models. Front Pharmacol 4, 35 (2013). 139. L. E. Arias-Romero, O. Villamar-Cruz, M. Huang, K. P. Hoeflich, J. Chernoff, Pak1 kinase links ErbB2 to beta-catenin in transformation of breast epithelial cells. Cancer Res 73, 3671-3682 (2013). 140. E. L. Bradshaw-Pierce, T. M. Pitts, A. C. Tan, K. McPhillips, M. West, D. L. Gustafson, C. Halsey, L. Nguyen, N. V. Lee, J. L. Kan, B. W. Murray, S. G. Eckhardt, Tumor P-Glycoprotein Correlates with Efficacy of PF-3758309 in in vitro and in vivo Models of Colorectal Cancer. Front Pharmacol 4, 22 (2013). 141. H. Lu, S. Liu, G. Zhang, W. Bin, Y. Zhu, D. T. Frederick, Y. Hu, W. Zhong, S. Randell, N. Sadek, W. Zhang, G. Chen, C. Cheng, J. Zeng, L. W. Wu, J. Zhang, X. Liu, W. Xu, C. Krepler, K. Sproesser, M. Xiao, B. Miao, J. Liu, C. D. Song, J. Y. Liu, G. C. Karakousis, L. M. Schuchter, Y. Lu, G. Mills, Y. Cong, J. Chernoff, J. Guo, G. M. Boland, R. J. Sullivan, Z. Wei, J. Field, R. K. Amaravadi, K. T. Flaherty, M. Herlyn, X. Xu, W. Guo, PAK signalling drives acquired drug resistance to MAPK inhibitors in BRAF-mutant melanomas. Nature 550, 133-136 (2017). 142. W. McCoull, E. J. Hennessy, K. Blades, M. R. Box, C. Chuaqui, J. E. Dowling, C. D. Davies, A. D. Ferguson, F. W. Goldberg, N. J. Howe, P. D. Kemmitt, G. M. Lamont, K. Madden, C. McWhirter, J. G. Varnes, R. A. Ward, J. D. Williams, B. Yang, Identification and optimisation of 7-azaindole PAK1 inhibitors with improved potency and kinase selectivity. MedChemComm 5, 1533-1539 (2014). 143. W. McCoull, E. J. Hennessy, K. Blades, C. Chuaqui, J. E. Dowling, A. D. Ferguson, F. W. Goldberg, N. Howe, C. R. Jones, P. D. Kemmitt, G. Lamont, J. G. Varnes, R. A. Ward, B. Yang, Optimization of Highly Kinase Selective Bis- anilino Pyrimidine PAK1 Inhibitors. ACS Med Chem Lett 7, 1118-1123 (2016). 141

144. S. Licciulli, J. Maksimoska, C. Zhou, S. Troutman, S. Kota, Q. Liu, S. Duron, D. Campbell, J. Chernoff, J. Field, R. Marmorstein, J. L. Kissil, FRAX597, a small molecule inhibitor of the p21-activated kinases, inhibits tumorigenesis of neurofibromatosis type 2 (NF2)-associated Schwannomas. J Biol Chem 288, 29105-29114 (2013). 145. T. Y. Prudnikova, J. Chernoff, The Group I Pak inhibitor Frax-1036 sensitizes 11q13-amplified ovarian cancer cells to the cytotoxic effects of Rottlerin. Small GTPases 8, 193-198 (2017). 146. T. Y. Prudnikova, O. Villamar-Cruz, S. J. Rawat, K. Q. Cai, J. Chernoff, Effects of p21-activated kinase 1 inhibition on 11q13-amplified ovarian cancer cells. Oncogene 35, 2178-2185 (2016). 147. C. O. Ndubaku, J. J. Crawford, J. Drobnick, I. Aliagas, D. Campbell, P. Dong, L. M. Dornan, S. Duron, J. Epler, L. Gazzard, C. E. Heise, K. P. Hoeflich, D. Jakubiak, H. La, W. Lee, B. Lin, J. P. Lyssikatos, J. Maksimoska, R. Marmorstein, L. J. Murray, T. O'Brien, A. Oh, S. Ramaswamy, W. Wang, X. Zhao, Y. Zhong, E. Blackwood, J. Rudolph, Design of Selective PAK1 Inhibitor G-5555: Improving Properties by Employing an Unorthodox Low-pK a Polar Moiety. ACS Med Chem Lett 6, 1241-1246 (2015). 148. J. Rudolph, L. J. Murray, C. O. Ndubaku, T. O'Brien, E. Blackwood, W. Wang, I. Aliagas, L. Gazzard, J. J. Crawford, J. Drobnick, W. Lee, X. Zhao, K. P. Hoeflich, D. A. Favor, P. Dong, H. Zhang, C. E. Heise, A. Oh, C. C. Ong, H. La, P. Chakravarty, C. Chan, D. Jakubiak, J. Epler, S. Ramaswamy, R. Vega, G. Cain, D. Diaz, Y. Zhong, Chemically Diverse Group I p21-Activated Kinase (PAK) Inhibitors Impart Acute Cardiovascular Toxicity with a Narrow Therapeutic Window. J Med Chem 59, 5520-5541 (2016). 149. S. K. McCarty, M. Saji, X. Zhang, C. M. Knippler, L. S. Kirschner, S. Fernandez, M. D. Ringel, BRAF activates and physically interacts with PAK to regulate cell motility. Endocr Relat Cancer 21, 865-877 (2014). 150. M. Babagana, S. Johnson, H. Slabodkin, W. Bshara, C. Morrison, E. S. Kandel, P21-activated kinase 1 regulates resistance to BRAF inhibition in human cancer cells. Mol Carcinog 56, 1515-1525 (2017). 151. G. Semenova, J. Chernoff, Targeting PAK1. Biochem Soc Trans 45, 79-88 (2017). 152. I. Landa, N. Pozdeyev, C. Korch, L. A. Marlow, R. C. Smallridge, J. A. Copland, Y. C. Henderson, S. Y. Lai, G. L. Clayman, N. Onoda, A. C. Tan, M. E. R. Garcia-Rendueles, J. A. Knauf, B. R. Haugen, J. A. Fagin, R. E. Schweppe, Comprehensive genetic characterization of human thyroid cancer cell lines: a validated panel for preclinical studies. Clin Cancer Res, (2019). 153. J. Tanaka, T. Ogura, H. Sato, M. Hatano, Establishment and biological characterization of an in vitro human cytomegalovirus latency model. Virology 161, 62-72 (1987). 154. P. E. Goretzki, A. Frilling, D. Simon, H. D. Roeher, Growth regulation of normal thyroids and thyroid tumors in man. Recent Results Cancer Res 118, 48-63 (1990). 142

155. N. Fabien, A. Fusco, M. Santoro, Y. Barbier, P. M. Dubois, C. Paulin, Description of a human papillary thyroid carcinoma cell line. Morphologic study and expression of tumoral markers. Cancer 73, 2206-2212 (1994). 156. X. Xu, R. M. Quiros, P. Gattuso, K. B. Ain, R. A. Prinz, High prevalence of BRAF gene mutation in papillary thyroid carcinomas and thyroid tumor cell lines. Cancer Res 63, 4561-4567 (2003). 157. T. Ito, T. Seyama, Y. Hayashi, T. Hayashi, K. Dohi, T. Mizuno, K. Iwamoto, N. Tsuyama, N. Nakamura, M. Akiyama, Establishment of 2 human thyroid- carcinoma cell-lines (8305c, 8505c) bearing p53 gene-mutations. Int J Oncol 4, 583-586 (1994). 158. N. E. Heldin, D. Cvejic, S. Smeds, B. Westermark, Coexpression of functionally active receptors for thyrotropin and platelet-derived growth factor in human thyroid carcinoma cells. Endocrinology 129, 2187-2193 (1991). 159. C. Challeton, F. Branea, M. Schlumberger, N. Gaillard, F. de Vathaire, C. Badie, P. Antonini, C. Parmentier, Characterization and radiosensitivity at high or low dose rate of four cell lines derived from human thyroid tumors. Int J Radiat Oncol Biol Phys 37, 163-169 (1997). 160. L. A. Marlow, J. D'Innocenzi, Y. Zhang, S. D. Rohl, S. J. Cooper, T. Sebo, C. Grant, B. McIver, J. L. Kasperbauer, J. T. Wadsworth, J. D. Casler, P. W. Kennedy, W. E. Highsmith, O. Clark, D. Milosevic, B. Netzel, K. Cradic, S. Arora, C. Beaudry, S. K. Grebe, M. L. Silverberg, D. O. Azorsa, R. C. Smallridge, J. A. Copland, Detailed molecular fingerprinting of four new anaplastic thyroid carcinoma cell lines and their use for verification of RhoB as a molecular therapeutic target. J Clin Endocrinol Metab 95, 5338-5347 (2010). 161. D. Chakravarty, E. Santos, M. Ryder, J. A. Knauf, X. H. Liao, B. L. West, G. Bollag, R. Kolesnick, T. H. Thin, N. Rosen, P. Zanzonico, S. M. Larson, S. Refetoff, R. Ghossein, J. A. Fagin, Small-molecule MAPK inhibitors restore radioiodine incorporation in mouse thyroid cancers with conditional BRAF activation. J Clin Invest 121, 4700-4711 (2011). 162. W. Zhao, L. Zhang, L. Zeng, H. Yang, A Two-Stage Response Surface Approach to Modeling Drug Interaction. Statistics in Biopharmaceutical Research 4, 375- 383 (2012). 163. C. Harbron, A flexible unified approach to the analysis of pre-clinical combination studies. Stat Med 29, 1746-1756 (2010). 164. S. Loewe, The problem of synergism and antagonism of combined drugs. Arzneimittelforschung 3, 285-290 (1953). 165. M. Nazarov, N. Goeyvaerts, C. Harbron. (R package, https://CRAN.R- project.org/package=drugCombo, 2018). 166. R. Liu, D. Liu, E. Trink, E. Bojdani, G. Ning, M. Xing, The Akt-specific inhibitor MK2206 selectively inhibits thyroid cancer cells harboring mutations that can activate the PI3K/Akt pathway. J Clin Endocrinol Metab 96, E577-585 (2011). 167. T. A. Yap, L. Yan, A. Patnaik, I. Fearen, D. Olmos, K. Papadopoulos, R. D. Baird, L. Delgado, A. Taylor, L. Lupinacci, R. Riisnaes, L. L. Pope, S. P. Heaton, G. Thomas, M. D. Garrett, D. M. Sullivan, J. S. de Bono, A. W. Tolcher, First-in- 143

man clinical trial of the oral pan-AKT inhibitor MK-2206 in patients with advanced solid tumors. J Clin Oncol 29, 4688-4695 (2011). 168. J. Xing, R. Liu, M. Xing, B. Trink, The BRAFT1799A mutation confers sensitivity of thyroid cancer cells to the BRAFV600E inhibitor PLX4032 (RG7204). Biochem Biophys Res Commun 404, 958-962 (2011). 169. H. Goto, K. Tanabe, E. Manser, L. Lim, Y. Yasui, M. Inagaki, Phosphorylation and reorganization of vimentin by p21-activated kinase (PAK). Genes Cells 7, 91- 97 (2002). 170. D. D. Tang, Y. Bai, S. J. Gunst, Silencing of p21-activated kinase attenuates vimentin phosphorylation on Ser-56 and reorientation of the vimentin network during stimulation of smooth muscle cells by 5-hydroxytryptamine. Biochem J 388, 773-783 (2005). 171. A. Nayal, D. J. Webb, C. M. Brown, E. M. Schaefer, M. Vicente-Manzanares, A. R. Horwitz, Paxillin phosphorylation at Ser273 localizes a GIT1-PIX-PAK complex and regulates adhesion and protrusion dynamics. J Cell Biol 173, 587- 589 (2006). 172. I. R. Watson, L. Li, P. K. Cabeceiras, M. Mahdavi, T. Gutschner, G. Genovese, G. Wang, Z. Fang, J. M. Tepper, K. Stemke-Hale, K. Y. Tsai, M. A. Davies, G. B. Mills, L. Chin, The RAC1 P29S hotspot mutation in melanoma confers resistance to pharmacological inhibition of RAF. Cancer Res 74, 4845-4852 (2014). 173. J. L. Kissil, K. C. Johnson, M. S. Eckman, T. Jacks, Merlin phosphorylation by p21-activated kinase 2 and effects of phosphorylation on merlin localization. J Biol Chem 277, 10394-10399 (2002). 174. G. H. Xiao, A. Beeser, J. Chernoff, J. R. Testa, p21-activated kinase links Rac/Cdc42 signaling to merlin. J Biol Chem 277, 883-886 (2002). 175. Y. Shrestha, E. J. Schafer, J. S. Boehm, S. R. Thomas, F. He, J. Du, S. Wang, J. Barretina, B. A. Weir, J. J. Zhao, K. Polyak, T. R. Golub, R. Beroukhim, W. C. Hahn, PAK1 is a breast cancer oncogene that coordinately activates MAPK and MET signaling. Oncogene 31, 3397-3408 (2012). 176. Y. Wang, H. Tsui, Y. Ke, Y. Shi, Y. Li, L. Davies, E. J. Cartwright, L. Venetucci, H. Zhang, D. A. Terrar, C. L. Huang, R. J. Solaro, X. Wang, M. Lei, Pak1 is required to maintain ventricular Ca(2)(+) homeostasis and electrophysiological stability through SERCA2a regulation in mice. Circ Arrhythm Electrophysiol 7, 938-948 (2014). 177. W. Liu, M. Zi, R. Naumann, S. Ulm, J. Jin, D. M. Taglieri, S. Prehar, J. Gui, H. Tsui, R. P. Xiao, L. Neyses, R. J. Solaro, Y. Ke, E. J. Cartwright, M. Lei, X. Wang, Pak1 as a novel therapeutic target for antihypertrophic treatment in the heart. Circulation 124, 2702-2715 (2011). 178. Y. Wang, S. Wang, M. Lei, M. Boyett, H. Tsui, W. Liu, X. Wang, The p21- activated kinase 1 (Pak1) signalling pathway in cardiac disease: from mechanistic study to therapeutic exploration. Br J Pharmacol 175, 1362-1374 (2018). 179. D. Hanahan, R. A. Weinberg, Hallmarks of cancer: the next generation. Cell 144, 646-674 (2011).

144

180. K. J. Kirkwood, Y. Ahmad, M. Larance, A. I. Lamond, Characterization of native protein complexes and protein isoform variation using size-fractionation-based quantitative proteomics. Mol Cell Proteomics 12, 3851-3873 (2013). 181. V. S. Rao, K. Srinivas, G. N. Sujini, G. N. Kumar, Protein-protein interaction detection: methods and analysis. Int J Proteomics 2014, 147648 (2014). 182. I. Weibrecht, K. J. Leuchowius, C. M. Clausson, T. Conze, M. Jarvius, W. M. Howell, M. Kamali-Moghaddam, O. Soderberg, Proximity ligation assays: a recent addition to the proteomics toolbox. Expert Rev Proteomics 7, 401-409 (2010). 183. S. Fredriksson, M. Gullberg, J. Jarvius, C. Olsson, K. Pietras, S. M. Gustafsdottir, A. Ostman, U. Landegren, Protein detection using proximity-dependent DNA ligation assays. Nat Biotechnol 20, 473-477 (2002). 184. O. Soderberg, M. Gullberg, M. Jarvius, K. Ridderstrale, K. J. Leuchowius, J. Jarvius, K. Wester, P. Hydbring, F. Bahram, L. G. Larsson, U. Landegren, Direct observation of individual endogenous protein complexes in situ by proximity ligation. Nat Methods 3, 995-1000 (2006). 185. R. Pacchiana, M. Abbate, U. Armato, I. Dal Pra, A. Chiarini, Combining immunofluorescence with in situ proximity ligation assay: a novel imaging approach to monitor protein-protein interactions in relation to subcellular localization. Histochem Cell Biol 142, 593-600 (2014). 186. D. I. Kim, S. C. Jensen, K. A. Noble, B. Kc, K. H. Roux, K. Motamedchaboki, K. J. Roux, An improved smaller biotin ligase for BioID proximity labeling. Mol Biol Cell 27, 1188-1196 (2016). 187. K. J. Roux, D. I. Kim, M. Raida, B. Burke, A promiscuous biotin ligase fusion protein identifies proximal and interacting proteins in mammalian cells. J Cell Biol 196, 801-810 (2012). 188. B. Diedrich, K. T. Rigbolt, M. Roring, R. Herr, S. Kaeser-Pebernard, C. Gretzmeier, R. F. Murphy, T. Brummer, J. Dengjel, Discrete cytosolic macromolecular BRAF complexes exhibit distinct activities and composition. EMBO J 36, 646-663 (2017). 189. M. W. Mayhew, D. J. Webb, M. Kovalenko, L. Whitmore, J. W. Fox, A. F. Horwitz, Identification of protein networks associated with the PAK1-betaPIX- GIT1-paxillin signaling complex by mass spectrometry. J Proteome Res 5, 2417- 2423 (2006). 190. O. E. Branson, M. A. Freitas, A multi-model statistical approach for proteomic spectral count quantitation. J Proteomics 144, 23-32 (2016). 191. H. Zhang, S. Pandey, M. Travers, H. Sun, G. Morton, J. Madzo, W. Chung, J. Khowsathit, O. Perez-Leal, C. A. Barrero, C. Merali, Y. Okamoto, T. Sato, J. Pan, J. Garriga, N. V. Bhanu, J. Simithy, B. Patel, J. Huang, N. J. Raynal, B. A. Garcia, M. A. Jacobson, C. Kadoch, S. Merali, Y. Zhang, W. Childers, M. Abou- Gharbia, J. Karanicolas, S. B. Baylin, C. A. Zahnow, J. Jelinek, X. Grana, J. J. Issa, Targeting CDK9 Reactivates Epigenetically Silenced Genes in Cancer. Cell 175, 1244-1258 e1226 (2018).

145

192. C. A. Barrero, O. Perez-Leal, M. Aksoy, C. Moncada, R. Ji, Y. Lopez, K. Mallilankaraman, M. Madesh, G. J. Criner, S. G. Kelsen, S. Merali, Histone 3.3 participates in a self-sustaining cascade of apoptosis that contributes to the progression of chronic obstructive pulmonary disease. Am J Respir Crit Care Med 188, 673-683 (2013). 193. G. Boden, P. Cheung, K. Kresge, C. Homko, B. Powers, L. Ferrer, Insulin resistance is associated with diminished endoplasmic reticulum stress responses in adipose tissue of healthy and diabetic subjects. Diabetes 63, 2977-2983 (2014). 194. G. Boden, X. Duan, C. Homko, E. J. Molina, W. Song, O. Perez, P. Cheung, S. Merali, Increase in endoplasmic reticulum stress-related proteins and genes in adipose tissue of obese, insulin-resistant individuals. Diabetes 57, 2438-2444 (2008). 195. G. Boden, C. Homko, C. A. Barrero, T. P. Stein, X. Chen, P. Cheung, C. Fecchio, S. Koller, S. Merali, Excessive caloric intake acutely causes oxidative stress, GLUT4 carbonylation, and insulin resistance in healthy men. Sci Transl Med 7, 304re307 (2015). 196. S. Shanmughapriya, D. Tomar, Z. Dong, K. J. Slovik, N. Nemani, K. Natarajaseenivasan, E. Carvalho, C. Lu, K. Corrigan, V. N. S. Garikipati, J. Ibetti, S. Rajan, C. Barrero, K. Chuprun, R. Kishore, S. Merali, Y. Tian, W. Yang, M. Madesh, FOXD1-dependent MICU1 expression regulates mitochondrial activity and cell differentiation. Nat Commun 9, 3449 (2018). 197. D. Tomar, Z. Dong, S. Shanmughapriya, D. A. Koch, T. Thomas, N. E. Hoffman, S. A. Timbalia, S. J. Goldman, S. L. Breves, D. P. Corbally, N. Nemani, J. P. Fairweather, A. R. Cutri, X. Zhang, J. Song, F. Jana, J. Huang, C. Barrero, J. E. Rabinowitz, T. S. Luongo, S. M. Schumacher, M. E. Rockman, A. Dietrich, S. Merali, J. Caplan, P. Stathopulos, R. S. Ahima, J. Y. Cheung, S. R. Houser, W. J. Koch, V. Patel, V. M. Gohil, J. W. Elrod, S. Rajan, M. Madesh, MCUR1 Is a Scaffold Factor for the MCU Complex Function and Promotes Mitochondrial Bioenergetics. Cell Rep 15, 1673-1685 (2016). 198. D. Bailey, L. Urena, L. Thorne, I. Goodfellow, Identification of protein interacting partners using tandem affinity purification. J Vis Exp, (2012). 199. A. H. Talukder, Q. Meng, R. Kumar, CRIPak, a novel endogenous Pak1 inhibitor. Oncogene 25, 1311-1319 (2006). 200. K. I. Brackley, J. Grantham, Activities of the chaperonin containing TCP-1 (CCT): implications for cell cycle progression and cytoskeletal organisation. Cell Stress Chaperones 14, 23-31 (2009). 201. S. H. Roh, M. Kasembeli, D. Bakthavatsalam, W. Chiu, D. J. Tweardy, Contribution of the Type II Chaperonin, TRiC/CCT, to Oncogenesis. Int J Mol Sci 16, 26706-26720 (2015). 202. Y. Zhang, Y. Wang, Y. Wei, J. Wu, P. Zhang, S. Shen, H. Saiyin, R. Wumaier, X. Yang, C. Wang, L. Yu, Molecular chaperone CCT3 supports proper mitotic progression and cell proliferation in hepatocellular carcinoma cells. Cancer Lett 372, 101-109 (2016).

146

203. M. Zhao, M. Spiess, H. J. Johansson, H. Olofsson, J. Hu, J. Lehtio, S. Stromblad, Identification of the PAK4 interactome reveals PAK4 phosphorylation of N- WASP and promotion of Arp2/3-dependent actin polymerization. Oncotarget 8, 77061-77074 (2017). 204. O. M. Grbovic, A. D. Basso, A. Sawai, Q. Ye, P. Friedlander, D. Solit, N. Rosen, V600E B-Raf requires the Hsp90 chaperone for stability and is degraded in response to Hsp90 inhibitors. Proc Natl Acad Sci U S A 103, 57-62 (2006). 205. R. K. Jaiswal, E. Weissinger, W. Kolch, G. E. Landreth, Nerve growth factor- mediated activation of the mitogen-activated protein (MAP) kinase cascade involves a signaling complex containing B-Raf and HSP90. J Biol Chem 271, 23626-23629 (1996). 206. S. I. Borysov, A. W. Cheng, T. M. Guadagno, B-Raf is critical for MAPK activation during mitosis and is regulated in an M phase-dependent manner in Xenopus egg extracts. J Biol Chem 281, 22586-22596 (2006). 207. S. I. Borysov, T. M. Guadagno, A novel role for Cdk1/cyclin B in regulating B- raf activation at mitosis. Mol Biol Cell 19, 2907-2915 (2008). 208. M. K. Borysova, Y. Cui, M. Snyder, T. M. Guadagno, Knockdown of B-Raf impairs spindle formation and the mitotic checkpoint in human somatic cells. Cell Cycle 7, 2894-2901 (2008). 209. Y. Cui, M. K. Borysova, J. O. Johnson, T. M. Guadagno, Oncogenic B- Raf(V600E) induces spindle abnormalities, supernumerary centrosomes, and aneuploidy in human melanocytic cells. Cancer Res 70, 675-684 (2010). 210. Y. Cui, T. M. Guadagno, B-Raf(V600E) signaling deregulates the mitotic spindle checkpoint through stabilizing Mps1 levels in melanoma cells. Oncogene 27, 3122-3133 (2008). 211. A. Camasses, A. Bogdanova, A. Shevchenko, W. Zachariae, The CCT chaperonin promotes activation of the anaphase-promoting complex through the generation of functional Cdc20. Mol Cell 12, 87-100 (2003). 212. M. B. Yaffe, G. W. Farr, D. Miklos, A. L. Horwich, M. L. Sternlicht, H. Sternlicht, TCP1 complex is a molecular chaperone in tubulin biogenesis. Nature 358, 245-248 (1992). 213. S. Kaisari, D. Sitry-Shevah, S. Miniowitz-Shemtov, A. Teichner, A. Hershko, Role of CCT chaperonin in the disassembly of mitotic checkpoint complexes. Proc Natl Acad Sci U S A 114, 956-961 (2017). 214. X. Liu, C. Y. Lin, M. Lei, S. Yan, T. Zhou, R. L. Erikson, CCT chaperonin complex is required for the biogenesis of functional Plk1. Mol Cell Biol 25, 4993- 5010 (2005). 215. M. Taipale, I. Krykbaeva, M. Koeva, C. Kayatekin, K. D. Westover, G. I. Karras, S. Lindquist, Quantitative analysis of HSP90-client interactions reveals principles of substrate recognition. Cell 150, 987-1001 (2012). 216. R. P. Charles, G. Iezza, E. Amendola, D. Dankort, M. McMahon, Mutationally activated BRAF(V600E) elicits papillary thyroid cancer in the adult mouse. Cancer Res 71, 3863-3871 (2011).

147