AN ANALYSIS OF P21-ACTIVATED -5 (PAK5) FUNCTION

BAHAREH TABANIFAR

NATIONAL UNIVERSITY OF SINGAPORE

2016

AN ANALYSIS OF P21-ACTIVATED KINASE-5 (PAK5) FUNCTION

BAHAREH TABANIFAR

A THESIS SUBMITTED

FOR THE DEGREE OF DOCTOR OF PHILOSOPHY

DEPARTMENT OF PHARMACOLOGY

NATIONAL UNIVERSITY OF SINGAPORE

2016

Declaration

I hereby declare that the thesis is my original work and it has been written by me in its entirety. I have duly acknowledged all the sources of information which have been used in the thesis.

This thesis has also not been submitted for any degree in any university previously.

______Bahareh Tabanifar 2016

I

Acknowledgments

Foremost, I would like to express my sincere gratitude to my supervisor Professor Edward Manser for the continuous support of my Ph.D study and research, for his patience, motivation, enthusiasm, and immense knowledge. His guidance and support helped me in all the time of research and writing of this thesis.

My deepest gratitude goes to my family: My beloved husband who has been there for me every step of the way… My parents for their unconditional support throughout my life and my studies.

My special thanks to my mentors:

Dr.Yohendran Baskaran and Dr. Ling Goh, for always being patient to answer my questions, for all of their support and help.

I would like to thank the rest of my thesis committee: Professor. Shazib Pervaiz and Professor Giulia Rancati, for their encouragements and insightful comments.

My sincere thanks also goes to the fellows in sGSK.IMCB lab and IMB: Dr. Jing Ming Dong, Dr.Zhoushen Zhao, Dr. Praju Vikas Anekal, Dr. Pao Chun Lin, Dr Ivan Khang Nyee TAN for their invaluable knowledge and generous guidance.

Also many thanks to my friends in the lab: Ms.Elsa, Ms. Felicia and Ms.Christy for being supportive, kind and fun!

I have to thank Agency for Science, Technology And Research (A*STAR) for granting me Singapore International Graduate Award (SINGA).

II

Table of Contents

Chapter 1: Introduction ...... 1

1.1 Overview of Rho GTPases ...... 1

1.2 Overview of P21 activated (PAKs) ...... 2

1.2.1 Domain structure of group I and II PAKs ...... 3

1.2.2 Regulation of group I and II PAKs ...... 4

1.2.3 Expression pattern and subcellular localization of PAK5...... 7

1.2.4 Phosphorylation targets of PAK5 ...... 9

1.2.5 The role of PAK5 in organization ...... 10

1.2.6 Group II PAKs and adherens junctions ...... 11

1.2.6.1 Background on adherens junctions ...... 11

1.2.6.2 Localization of group II PAKs at cell-cell junctions ...... 13

1.2.7 PAK5 and the regulation of apoptosis ...... 16

1.2.7.1 Background ...... 16

1.2.7.2 PAK5 plays a role in cell survival ...... 19

1.2.8 The role of PAK5 on MAPK pathway ...... 22

1.2.8.1 Overview of Raf-1 kinase ...... 22

1.2.8.2 PAK5 role on Raf-1 activation...... 23

1.2.9 The role of PAK5 on cancer progression ...... 26

1.2.10 PAK5 role in neuronal vesicle trafficking ...... 28

1.3 Overview of the LZTS family ...... 28

1.3.1 The role of LZTS2 on tumour suppression ...... 29

1.3.2 LZTS2 and regulation of the cytoskeleton ...... 33

1.3.3 LZT2 and cell migration ...... 34

1.3.4 The role of LZTS2 in development ...... 34

1.4 Aims of the study ...... 35

III

Chapter 2: Materials and methods ...... 37

2.1 Materials ...... 37

2.1.1 Plasmids and siRNAs ...... 37

2.1.2 Antibodies and reagents ...... 38

2.2 DNA manipulation methods ...... 38

2.2.1 Plasmid DNA purification ...... 38

2.2.2 Polymerase Chain Reaction (PCR) and gel electrophoresis ...... 39

2.2.3 Preparation of competent cells for transformation via heat shock method ...... 40

2.2.4 ...... 40

2.3 Protein analysis methods ...... 42

2.3.1 SDS polyacrylamide gel electrophoresis (SDS-PAGE) ...... 42

2.3.2 Western blotting ...... 42

2.3.3 Expression and purification of bacterial expressed recombinant proteins ...... 42

2.3.4 Antibody production ...... 43

2.3.4.1 Coupling to Cyanogen Bromide-Activated agarose for Antibody purification ...... 43

2.3.4.2 Affinity purification of antibody ...... 44

2.4 Cell culture techniques and manipulations ...... 44

2.4.1 Cell culture ...... 44

2.4.2 Transient transfection ...... 45

2.4.3 siRNA transfection...... 45

2.4.4 Establishing stable lines ...... 45

2.4.5 Immuno-precipitation ...... 47

2.4.6 In vitro kinase assay ...... 47

2.4.7 Immunostaining ...... 47

2.5 Methods for detections of protein-protein interactions ...... 48

IV

2.5.1 Establishing stable lines for BioID and GFP pull down analysis ...... 48

2.5.2 Adapting cells in SILAC media ...... 48

2.5.3 Purification of biotin-labelled proteins from stable GFP-Bir*A cells ...... 49

2.5.4 GFP Purification of proteins from stable GFP-Bir*A cells ...... 50

Chapter 3: Regulation of PAK5 activity ...... 52

3.1 Introduction ...... 52

3.2 PAK5 contains a functional auto inhibitory domain ...... 54

3.3 PAK5 is likely a Cdc42-specific effector ...... 54

3.4 PAK5 auto-regulatory domain can inhibit PAK4 ...... 59

3.5 Different cellular distributions of PAK4 and PAK5 ...... 61

3.6 The non-catlytic domain of PAK5 is sufficient for puncta formation ...... 62

3.7 Establishing the regions of PAK5 involved in self-association ...... 67

3.8 PAK5 self-association contributes to its high basal activity ...... 73

3.9 Discussion ...... 78

3.9.1 PAK5 activity is partially regulated by the presence of a functional AID at N-terminal ...... 78

3.9.2 The punctate distribution of PAK5 reflects kinase oligomers ...... 80

3.9.3 The central region of PAK5 plays role in elevated activity of the protein ...... 81

Chapter 4: Analysis of potential PAK5 substrates BAD and Raf-1 ...... 84

4.1 Introduction ...... 84

4.2 Subcellular distribution of PAK5 ...... 86

4.3 Analysis of BAD as a PAK5 target ...... 92

4.3.1 PAK5 does not co-localize with BAD in the cells ...... 92

4.3.2 Camptothecin treatment does not affect PAK5 localization ...... 95

4.3.3 PAK5 promotes BAD Ser112 phosphorylation ...... 95

4.4 Analysis of PAK5 interaction and localization with Raf-1 ...... 98

V

4.4.1 PAK5 can co-localize with Raf-1 and phosphorylate it at Ser338 ...... 98

4.4.2 Investigating the interaction site of PAK5 with Raf-1 ...... 104

4.5 Summary ...... 106

4.6 Characterization of endogenous PAK5 ...... 106

4.6.1 GFP-PAK5 localizes in junctions of U2OS cells ...... 111

4.7 Discussion ...... 115

4.7.1 PAK5 phosphorylate BAD and Raf-1 in vivo ...... 115

4.7.2 Localization of PAK5 in mammalian cell lines ...... 117

Chapter 5. Screening and characterization of proteins co-localizing with PAK5 in U2OS cells ...... 120

5.1 Introduction ...... 120

5.2. Classical methods for identification of protein-protein interactions ...... 124

5.3 Proximal protein labelling by BirA* biotinylation in mammalian cells ...... 125

5.4 Establishing GFP-BirA*-PAK5 line for BioID ...... 127

5.5 Identification and analysis of PAK5 proximal proteins ...... 132

5.6 Identification and analysis of PAK5 interacting partners using GFP-pull down ...... 138

5.7 Analysis of the results of BioID and GFP-PD ...... 145

5.8 LZTS2 is a direct partner of PAK5 at cell-cell junctions...... 147

5.8.1 Identification of the region of PAK5 involved in binding LZTS2 ...... 151

5.8.2 Identification and localization of a new form of LZTS2 ...... 156

5.8.3 LZTS2 and PAK5 co-localize to cell-cell junctions ...... 161

5.8.4 Junctional localization of PAK5 is independent of LZTS2 ...... 163

5.8.5 Afadin is required for junctional localization of LZTS2 and PAK5 ...... 163

5.8.6 Binding to LZTS2 does not regulate PAK5 activity ...... 164

5.9 Analyzing the potential interaction between LZTS2 and Raf-1 ...... 169

5.10 LZTS2 mediates PAK5 and Raf-1 co-localization at mitochondria ...... 171

VI

5.11 LZTS2 does not affect Raf-1 Ser338 phosphorylation...... 173

5.11 Discussion ...... 175

5.11.1 Identification of PAK5 interacting partners in U2OS cells ...... 175

5.11.2 Is LZTS2 a key partner for PAK5? ...... 181

Chapter 6. General Conclusions ...... 186

6.1 The group II PAK5 oligomerization underlies its high basal activity ...... 186

6.2 Identification of PAK5 as a junctional kinase ...... 189

VII

Summary

Rho (Ras homology) small GTPases (or p21s) play important roles in a wide variety of cellular process most importantly the organization of the actin cytoskeleton and establishment of cell polarity, which allows coordinated movement of cells and tissues. The PAKs (p21 activated kinases) are an important class of effector for Cdc42 and Rac-like Rho proteins, with PAK1-3 constituting the conventional or 'group I' kinases and PAK4-6 the group II kinases. Both groups are represented in all metazoans and are characterized by a C-terminal catalytic domain and an N-terminal p21 binding domain (PBD) and auto-inhibitory domain (AID). The 85 kDa human

PAK5 does not appear to be auto-inhibited since it is active when purified from mammalian cells, even without Cdc42 bound. I revealed that PAK5 has the ability to undergo oligomerization through PAK5-unique residues 120-425 which are responsible for the elevated basal activity of the kinase relative to the monomeric 65 kDa PAK4. Analysis of PAK5 localization in various mammalian cell lines indicated that regardless of 'tag', the protein equilibrates between the and characteristic PAK5 puncta which are driven by its oligomerization. This does not appear to result from non-specific aggregation. Although some previous studies found that these puncta co-localized with mitochondria, I could not reproduce this result.

Instead, I show convincingly in multiple stable cell lines expressing GFP-PAK5 that the kinase is junctional. This unexpected observation was followed up by an analysis of PAK5 proximal proteins using a technique termed BioID that utilizes a mutant version of the E.coli protein BirA fused to PAK5 to modify nearby cellular proteins with biotin. To this end, I generated osteosarcoma U2OS stable cell lines expressing

GFP-BirA*-PAK5 or control GFP-BirA*. By using stable isotopic labelling of cells

(SILAC), it was possible to greatly increase the specificity of protein identification

VIII using LC-MS/MS mass spectrometry. The majority of proteins identified with BioID in the GFP-BirA*-PAK5 assay were indeed associated with cell-cell junctions, including Afadin and Nectin-2. In addition, it was interesting to note that PAK5 can be seen in the centrosomal region where it may associate with CEP85. The leucine zipper tumor suppressor protein LZTS2 was identified as proximal to PAK5, which was of interest given the proposed role of PAK5 in cancer progression. Biochemical experiments confirmed the direct interaction between PAK5 and LZTS2. Interestingly

PAK5 self-association promoted this interaction with LZTS2, which binds much more weakly to PAK4. Based on previous work from our laboratory and new experiments using an LZTS2 specific antibody, I showed that the junctional form of LZTS2 is derived from an alternate translational start site that generates a shorter membrane bound myristoylated isoform (LZTS2b). Knockdown of LZTS2 by siRNA indicated that PAK5 was not dependent on this adaptor for its junctional localization.

Conversely, LZTS2 did not affect PAK5 activity in vitro. Interestingly a fraction of

PAK5 was translocated to the mitochondria when was co-expressed with the myristoylated LZTS2b. This is because the myristolated signal of LZTS2b can target the protein both to the plasma membrane and mitochondria. Further I suggest an interplay between Raf-1 and LZTS2 may allow specific function of PAK5 and Raf-1 kinases on mitochondria, consistent with PAK5 playing a role to inactivate BAD.

IX

List of Figures

Figure 1.1 Schematic diagram of Rho-GTPases activation cycle ...... 2

Figure 1.2 Schematic diagrams of PAK1 and PAK4 domains ...... 4

Figure 1.3 Schematic of group I PAKs activation ...... 6

Figure 1.4 A model for regulation of group II PAKs activity through binding to Cdc42...... 7

Figure 1.5 Schematic diagram of the known targets of PAK5 ...... 10

Figure 1.6 Schematic diagram of key components of adherens junctions and tight junctions. (Adapted from the review by Kooistra, et al. 2007 [63]) ...... 13

Figure 1.7 Schematic overview of -catenin functional domain ...... 15

Figure 1.8 Schematic overview of two main apoptotic pathways ...... 18

Figure 1.9 The Bcl-2 family of proteins ...... 19

Figure 1.10 schematic diagram of apoptosis regulation through the regulation of BAD phosphorylation...... 21

Figure 1.11 schematic of the classic MAPK kinase pathway ...... 24

Figure 1.12 Schematic overview of Raf-1 domains (A) and mechanism of regulation. (B)...... 25

Figure 1.13 ClustalW sequence alignment of human LZTS1 and LZTS2 ...... 31

Figure 1.14 Schematic view of LZTS2 mediated regulation of β-catenin/Tcf pathway ...... 32

Figure 3.1. Comparison of N-terminal and C-terminal sequences of human group II PAKs...... 53

Figure 3.2 N-terminal deletion induces PAK5 activity ...... 56

Figure 3.2 N-terminal deletion induces PAK5 activity ...... 57

Figure 3.3 Co-transfection of Cdc42 (G12V) increases the in vitro activity of PAK5 ...... 57

Figure 3.3 Co-transfection of Cdc42 (G12V) increases the in vitro activity of PAK5 ...... 58

Figure 3.4 The PAK5 auto-inhibitory domain is functionally active towards PAK4 catalytic domain ...... 60

Figure 3.5 Subcellular distribution of PAK4 and PAK5 in COS-7 cells ...... 63

X

Figure 3.6 ClustalW sequence alignment of the central region of PAK4 and PAK5 ...... 64

Figure 3.7 The central region of PAK5 is involved in puncta formation ...... 65

Figure 3.8 The PAK5-specific central region is primarily involved in puncta formation ... 66

Figure 3.9. Biochemical experiments show PAK5 has the ability of self-interaction ...... 69

Figure 3.9. Biochemical experiments show PAK5 has the ability of self-interaction ...... 70

Figure 3.10 The central PAK5-specific region is required for self-interaction ...... 70

Figure 3.10 The central PAK5-specific region is required for self-interaction ...... 71

Figure 3.11. Gel-filtration chromatography of Flag-tagged PAK1, PAK4, PAK5 proteins 72

Figure 3.12. The central PAK5-specific region is responsible for its elevated activity ...... 75

Figure 3.13. The PAK5 specific central region may interfere with the AID function ...... 76

Figure 3.13. The PAK5 specific central region may interfere with the AID function ...... 77

Figure 4.1 The effect of relevant kinase inhibitors on BAD and Raf-1 phosphorylation .... 87

Figure 4.2 Subcellular localization of PAK5 ...... 88

Figure 4.3 PAK5 contains a cryptic mitochondrial localization sequence ...... 89

Figure 4.4 The effect of on putative nuclear export sequences of PAK4 and PAK5 ...... 91

Figure 4.5 Analysis of co-localization of transiently expressed PAK5 and BAD ...... 93

Figure 4.6 Mitochondrial localization of mutated BAD does not induce PAK5 translocation to the mitochondria ...... 94

Figure 4.7 Camptothecin treatment does not induce PAK5 re-localization to the mitochondria ...... 96

Figure 4.8 PAK5 phosphorylates BAD on Ser-112 in vivo ...... 97

Figure 4.9 PAK5 phosphorylates Raf-1 on Ser338 in vivo ...... 100

Figure 4.10 Analysis of PAK5 and Raf-1 co-localization in COS-7 cells ...... 101

Figure 4.10 Analysis of PAK5 and Raf-1 co-localization in COS-7 cells ...... 102

Figure 4.10 Analysis of PAK5 and Raf-1 co-localization in COS-7 cells ...... 103

Figure 4.11 Raf-1 can bind to the N-terminal of PAK5 ...... 105

XI

Figure 4.12 Comparison of sensitivity of PAK5 antibodies ...... 109

Figure 4.13 Analysis of endogenous PAK5 expression in different types of lines ...... 110

Figure 4.14 PAK5 expression in PAK5 knockdown stable line NCI-H446 ...... 110

Figure 4.15 PAK5 localization in GFP-PAK5 U2OS stable line ...... 112

Figure 4.15 PAK5 localization in GFP-PAK5 U2OS stable line ...... 113

Figure 4.16 Partial co-localization of GFP PAK5 with p120 catenin ...... 114

Figure 5.1 Analysis of U2OS stable lines expressing GFP-BirA*-PAK5 (FL) or GFP- BirA* controls ...... 129

Figure 5.2 Schematic of Bio-ID work-flow ...... 130

Figure 5.3 Labelling proteins using SILAC for quantitative mass spectrometry...... 131

Figure 5.4 The scatter plot analysis of the SILAC ratio versus coverage derived from BioID analysis ...... 136

Figure 5.5 The identification of PAK5 proximal proteins by SILAC BioID ...... 137

Figure 5.6 Schematic of the SILAC GFP pull down experimental work-flow ...... 139

Figure 5.7 The scatter plot of the complete set of proteins identified in the GFP-PAK5 pull down...... 140

Figure 5.8 Potential PAK5 binding proteins assessed by GFP-PAK5 pull-down with SILAC normalization ...... 144

Figure 5.9 Weak overlap between PAK5-asscoaietd proteins extracted from databases and those identified by BioID and GFP pull-down...... 147

Figure 5.10 Sequence alignment of human LZTS family isoforms. A...... 150

Figure 5.11 PAK5 interacts directly with LZTS2 ...... 152

Figure 5.12 PAK5 oligomerization promotes interaction with LZTS2 ...... 153

Figure 5.13 Investigating the basis for an interaction between PAK5 and LZTS2 ...... 154

Figure 5.13 Investigating the basis for an interaction between PAK5 and LZTS2 ...... 155

Figure 5.14 Subcellular localization of the two isoforms of LZTS2 in the cell ...... 158

Figure 5.15 Analysis of endogenous LZTS2 expression and localization ...... 159

XII

Figure 5.16 Subcellular localization of the non-tagged constructs of LZTS2a and LZTS2b in U2OS and COS-7 cells ...... 160

Figure 5.17 Analysis of PAK5 and LZTS2 co-localization in the cell ...... 162

Figure 5.18 Junctional localization of PAK5 is independent of LZTS2 expression ...... 165

Figure 5.19 Afadin-dependent localization of PAK5 and LZTS2 at cell-cell junction ..... 166

Figure 5.19 Afadin-dependent localization of PAK5 and LZTS2 at cell-cell junction ..... 167

Figure 5.20 PAK5 activity is not regulated by binding to LZTS2 ...... 168

Figure 5.21 Investigating the interaction site of LZTS2 with Raf-1 ...... 170

Figure 5.22 Analysis of colocalization of expressed PAK5, LZTS2b and Raf-1 in COS-7 cells ...... 172

Figure 5.23 The effect of LZTS2 suppression on Raf-1 and BAD phosphorylation ...... 174

Figure 6.1. A model for PAK5 oligomerization and subsequent activation ...... 188

Figure 6.2 A proposed model for PAK5 localization in the cell and its association with neighboring proteins ...... 191

XIII

Nomenclature

2-ME β-mercaptoethanol APS Ammonium persulfate ATP Adenosine 5'-triphosphate BSA Bovine serum albumin CA Constitutively active cDNA Complementary deoxyribonucleic acid Ci Curie CRIB Cdc42/Rac-interactive binding CT Carboxy-terminus DMEN Dulbecco’s minimal essential medium DMSO Dimethyl sulfoxide DNA Deoxyribonucleic acid dNTP Deoxyribonucleotide triphosphate DTT Dithiothreitol E. coli Escherichia coli ECL Enhanced chemiluminescence EDTA Ethylenediaminetetraacetic acid FBS Fetal bovine serum Fig Figure GAP GTPase-activating protein GDP Guanosine 5'-diphosphate GST Glutathione S-transferase GTP Guanosine 5'-triphosphate HA Hemagglutinin HCl Hydrochloric acid HEPES N-[2-Hydroxyethyl]piperazine-N’-[2-ethanesulfonic acid] IgG Immunoglobulin G IP Immunoprecipitation

XIV

IPTG Isopropylthio-b-D-galactoside Kb Kilo base pairs KCl Potassium chloride KD Kinase dead kD Kilo Dalton KO Knock out LB Luria broth

Na2PO4 Sodium dihydrogen orthophosphate M Molar MAPK Mitogen-activated protein kinase MEM Minimal essential media

MgCl2 Magnesium chloride μM Micromolar mM Milimolar NaCl Sodium chloride NES Nuclear export sequence NLS Nuclear localization signal NT Amino-terminus OD Optical density PAGE Polyacrylamide gel electrophoresis PBD p21-binding domain PBS Phosphate buffered saline PCR Polymerase chain reaction pfu Plaque forming unit PVDF Polyvinylidene difluride RT Room temperature SDS Sodium dodecyl sulfate SDS-PAGE Sodium dodecyl sulphate-polyacrylamide gel electrophoresis TE Tris-EDTA TEMED Tetramethylethylethane-1-2-diamine

XV

Tris Tris[hydroxymethyl]aminomethane Tris-HCl Tris-hydroxymethyl aminomethane hydrochloride UV Ultraviolet WB Western blot w/v Weight per volume wt Wild type

XVI

Chapter 1: Introduction

1.1 Overview of Rho GTPases

Rho (Ras homology) small GTPases are small (21-25kD) proteins which belong to the Ras superfamily of GTPases. They are evolutionarily conserved proteins that are expressed in all eukaryotes. The 20 members of this family are sub divided to

8 groups based on their , structure, and biological functions; including Rho proteins (RhoA, RhoB and RhoC), Rac proteins (Rac1,

Rac2, Rac3 and RhoG), Cdc42 proteins (Cdc42, TC10 and TCL), Rnd proteins

(Rnd1, Rnd2, and Rnd3), RhoH, RhoUV (RhoU and RhoV), RhoBTB proteins

(RhoBTB1, RhoBTB2) and the RhoF (RhoD and RhoF) [1]. Like other GTP-ases, most of the Rho proteins family members switch between active GTP bound and inactive GDP bound forms that are mainly regulated by three groups of regulatory factors (Figure 1.1) [2]:1. Guanine Nucleotide exchange factors (GEFs) that catalyse exchange of GDP to GTP, 2. GTPase activating proteins (GAPs) that enhance activity of the GTPases and converse GTP to GDP and 3. Guanine nucleotide dissociation inhibitors (GDIs) that bind to the inactive cytosolic GTPases and prevent their membrane association. Some of the members of this family are known to be constitutively activated by permanent binding to GTP. In these cases, alternative mechanisms of regulation such as modulating expression, post translational modification like phosphorylation and protein degradation are involved in regulation of these proteins [1, 2].

Activated Rho GTPases interact with many downstream effectors and regulate a wide variety of cellular functions such as reorganization of the actin cytoskeleton and cell morphology, adhesion and migration, cytokinesis, cell proliferation and

1 survival [3-5]. PAKs (p21 activated kinases) are one of the well-known effectors of

Rho GTpases [6].

Figure 1.1 Schematic diagram of Rho-GTPases activation cycle (Adapted from the review by Huveneers and Danen. 2009 [7]). Inactive Rho-GDP is targeted to the membrane when dissociated from Rho-GDP dissociation inhibitor (Rho-GDI). In the membrane, guanine Nucleotide exchange factor (GEF) induces activation of Rho-GDP by replacing GDP to GTP. In GTP-bound form, the Rho protein is active and interacts with the downstream effectors. GTPase activating protein (GAP) catalyses GTPase activity of the Rho-GTP leads to hydrolysis of GTP to GDP and inactivation of the Rho protein.

1.2 Overview of P21 activated kinases (PAKs)

PAKS are a family of kinases that were initially discovered in

1994 via interaction with GTP bound forms of Rac1 and Cdc42 [8]. So far, 6 mammalian PAKs have been identified that are classified into two groups based on their sequence homology and structure. The conventional group I PAKs contains

PAK1-3 and non-conventional group II PAKs includes PAK4-6 [9]. PAK1 and PAK4 are the best characterized members of the PAK family. PAKs play an important role in a variety of cellular functions including cytoskeleton organization, cell proliferation

2 and survival. Overexpression of PAKs is contributed to some diseases such as cancer and inflammatory diseases [10, 11]. Group I and II PAKs differ in their binding partners and mechanisms of regulation.

1.2.1 Domain structure of group I and II PAKs

Basically PAKs family domains structure can be divided to three main domains: 1. The Regulatory domain at the N-terminal containing PBD (p21 binding domain) and AID (Autoinhibitory domain); 2. The central region, and 3. Serine threonine kinase domain at C-terminal (Figure 1.2).

PBD of Group I PAKs binds to Cdc42 and Rac1 whereas group II preferentially binds Cdc42 [12]. The catalytic domain contains an activation loop and phosphorylation on this residue is essential for full activity of the protein. The AID interacts with the catalytic domain in the inactive state of the kinase. Upon PBD interaction with Rho GTPases, the AID dissociates from the kinase domain which leads to activation of the protein.

Group I PAKs contains three proline rich conserved sequences at the N- terminal and Central region that interact with the SH3 domain of NCK (non-catalytic region of tyrosine kinase adaptor protein 1), Grb2 (growth-factor-receptor-bound protein 2) and PIX (PAK interacting exchange factor) [6, 13-15].

3

PIX

NCK Grb2

250 Kinase domain 545 PAK1 pT423

AID PBD 300 591 PAK4

pS474

Figure 1.2 Schematic diagrams of PAK1 and PAK4 domains (Adapted from the review by Zhao, et.al. 2012 [6], with some modifications) All isoforms of PAKs have an N-terminal regulatory domain; contains the P21 binding domain (PBD) and an auto inhibitory domain (AID), a central region and the C- terminal catalytic domain. The N-terminal and central region have several proline rich domains that are marked in yellow. In group I PAKs the proline rich domains are SH3 binding sites which interact with the indicated proteins.

1.2.2 Regulation of group I and II PAKs

Regulatory mechanism of group I PAKs activity was uncovered through identification of its AID domain, and the crystal structure of the auto-inhibited PAK1 suggested a dimeric conformation in which the AID sequence itself dimerizes [16]: later this was found to be a crystallization artifact. Group I PAKs are likely all regulated via a trans auto-inhibited mechanism in which the AID of one PAK molecule binds and inhibits the kinase domain of the other (Figure 1.3) [17].

Conformational changes in the AID, upon binding to active RhoGTPases, drive its dissociation from the kinase domain and allows autophosphorylation at multiple sites including the activation loop (T423) that is essential for full kinase activity [18-20].

PAK1 is a cytoplasmic kinase and its translocation to the cell membrane regulates its activity [21]. Binding to PIX proteins recruits PAK1 to the focal adhesions in a close association with Cdc42 and Rac1 [22]. Binding of SH3

4 containing adaptor proteins Grb2 and Nck may also recruits PAK1 in response to trans-membrane receptors such as EGFR [13, 14, 23]. PAK1 activity is negatively regulated by binding to and de-phosphorylation by protein phosphatases such as protein phosphatase type 2A (PP2A) and partner of PIX 1 and 2 (POPX1 and POPX2)

[24]. In addition, direct binding of hPIP (human PAK1-interacting protein 1),

CRIPAK (Cys-rich inhibitor of PAK1) or Nischarin has been suggested to prevent

PAK1 activation [25-27].

The group II PAKs do not have a homologous AID resembling PAK1, and their mechanism of regulation was unrecognized for more than 10 years after their discovery. In 2012, our group and others identified an AID in group II PAKs that binds in a pseudo-substrate-like manner to PAK4 catalytic domain [28, 29]. In addition, the realization that PAK4 is constitutively phosphorylated on its activation loop Ser-474, showed that the kinase domain of group II PAKs is always in an active state. The mechanism by which the PAK4 AID is regulated is not currently understood since in vitro Cdc42 binding to the flanking CRIB sequence does not change the binding of the AID to the kinase (Figure 1.4) [28]. One possibility is that the proline containing sequence (R49PKPLV) interacts with the SH3 domains [29] to relieve the auto-inhibited state of PAK4.

Group II PAKs preferentially bind to Cdc42: some studies have shown that

Cdc42 binding regulates both PAK4 and PAK5 activity [28, 30], but as mentioned above this interaction does not significantly affect PAK4 activity in vitro [29].

Interaction with Cdc42 might serve primarily to affect kinase localization [31, 32].

For example, it has been reported that co-expression of PAK5 with active

Cdc42(G12V) targets PAK5 to the filopodia, or that co-expression of RhoD leads to

PAK5 translocation to mitochondria [31].

5

Among group II PAKs, it is generally agreed that PAK5 and PAK6 are basally more active than PAK4 [32-34]. In the case of PAK6 binding to androgen receptor may increase its activity [34, 35]. The p38 MAPK (mitogen-activated protein kinase)

MKK6 stimulates PAK6 activity [36] by phosphorylating Ser165 and Tyr 566 of

PAK6. Co-expression of activated MKK6 has been shown to stimulate PAK4 and

PAK5 activity by phosphorylating Tyr 480 and 608 respectively [36] although it is not clear this represent a physiological modification.

Figure 1.3 Schematic of group I PAKs activation (Adapted from the review by Zhao and Manser. 2012 [6]). PAK1 forms dimer in inactive conformation. AID of one molecule binds to the kinase domain of another molecule and prevents auto-phosphorylation of the kinase. Binding to PIX targets PAK1 for activation. Rac1.GTP binding to the PBD domain induces some conformational changes that allow auto-phosphorylation and activation of the protein. Auto-phosphorylated PAK1 has reduced affinity for PIX.

6

Figure 1.4 A model for regulation of group II PAKs activity through binding to Cdc42. (Baskaran, et.al. 2012 [28], used with permission). PAK4 is constitutively phosphorylated on Ser 474 (activation loop). In inactive conformation, the AID binds to the kinase domain and prevents access to substrate. Binding to the Cdc42.GTP induces kinase activation by dissociation of the AID from the kinase domain which allows to substrate binding.

1.2.3 Expression pattern and subcellular localization of PAK5

PAK4 is the first member of group II PAKs to be identified [32] and is the most investigated member of this group. PAK4 is a ubiquitously expressed protein

[37] and has been implicated in many signalling pathways including reorganization of the actin cytoskeleton, cell migration and cell survival [12]. It drives the oncogenic transformation of cells and therefore involves in tumorigenesis [38-42]. PAK6 was identified as an androgen receptor (AR) interacting partner in yeast two hybrid assay

[34]. Later, it was revealed that PAK6 can interact with both AR and estrogen receptor α (ER α) and inhibits their transcriptional activity independent of binding to

RhoGTPases [43, 44]. Unlike PAK4, PAK6 has restricted expression pattern and is

7 mostly expressed in the , placenta, testis and prostate based on mRNA analysis data [9].

PAK5, also sometimes named PAK7, was isolated and characterized by

Pandey et al at 2002 [45]. The human PAK5 gene is located at 20p12 and is encoded by 12 . The protein is expressed only in vertebrates and unlike the ubiquitous

PAK4 [32], its expression is limited to neuronal tissues however; northern blot analysis has been detected its mRNA expression in several tissues such as brain, eye, muscles, pancreas, adrenal gland, prostate, ovary and testis [33, 37, 46].

PAK4 is highly expressed in the early stages of development and the expression is reduced in the adult tissues, therefore PAK4 is suggested to plays a key role in early development [47]. Loss of PAK4 is embryonic lethal due to defects in heart, however other abnormalities for example in the nervous system are reported as well [48]. In contrast, PAK5 and PAK6 are expressed in the brain later in development and they might be required for neuronal system maintenance. PAK5-null mice develop normally as has been reported for PAK6 knockout mice [46, 49].

Because both PAK5 and PAK6 are highly expressed in the brain, a functional redundancy is suggested between these two kinases. Although Pak5/Pak6 double knockout mice are also viable and fertile they have deficits in locomotive activity, as well as impaired and functions [49]. Primary cortical neurons from the double knockout mice have impaired neurite outgrowth. Thus, the decreased motor function in the Pak5/Pak6 KO mice likely results from decreased synaptic function among motor neurons.

The subcellular localization of PAK5 is ambiguous. Similar to other members of group II PAKs, PAK5 has nuclear localization sequences (NLS) at N-terminal

(amino acids 5-10) but the protein is exported from the nucleus to the cytoplasm by

8 the function of nuclear export signal (NES) which is mapped in residues 400-411 [32,

50]. Mitochondrial localization sequences have been mapped within the PAK5 N- terminal and C-terminal and the protein is reported to localize on mitochondria by a number of studies [31, 50-52]. However, other studies have shown a dotted appearance of transiently expressed PAK5 in CHO cells and NIH-3T3 cells which is not associated with mitochondrial signal [53, 54].

1.2.4 Phosphorylation targets of PAK5

Group II PAKs kinases including PAK4-6, differ in structure and function from group I PAKs. Although the PBD and kinase domains are very well conserved in PAK4, 5 and 6, these domains are more diverged between the two groups [37]; therefore it is not unexpected that group I and II show different substrate specificity although there are some overlaps as well. A few specific targets of PAK5 have been identified (Figure 1.5), however PAK5 likely shares many substrates in common with

PAK4 and PAK6, as discussed in the next sections.

9

Cdc42 MKK6

?

PAK5

Raf-1 E47

GATA1 BAD

MARK2 P120- Pacsin-1 catenin Synaptoja Epithelial- Apoptosis/ nin-1 cell survival mesenchymal transition (EMT) Cytoskeleton Neurotransmission organization

Figure 1.5 Schematic diagram of the known targets of PAK5 A few numbers of PAK5 interacting partners (or substrates) have been identifies that implicated the protein in some signalling pathways. Cdc42.GTP binds to PAK5 and may regulate PAK5 activity. MKK6 has also been reported to act upstream of PAK5 and [36], however the in vivo effect is not clear [12].

1.2.5 The role of PAK5 in cytoskeleton organization

Rho small GTPases play a key role in the organization of the actin

cytoskeleton [1] and PAKs , as a well-known effectors of these proteins, are

implicated in this phenomenon. In contrast with PAK4, little is known about the

PAK5 interplay with the actin cytoskeleton [12]. While PAK4 inhibits neurite

outgrowth by activating LIMK which in turn deactivates Cofilin [55], over-expression

of PAK5 in N1E-115 cells promotes filopodia formation and neurite outgrowth;

10 however the underlying mechanism is not understood. One suggestion is that PAK5 acts through the inhibition of Rho-A [33] which, in turn, stimulates neuritis production in N1E115 cells [56, 57]; but PAK5 has not been shown to interact with any of the Rho-A GEFs. Instead, another study has shown that GEFT which is an activator of Rho, Rac and Cdc42 and is highly expressed in the brain, induces neurite outgrowth through the Rac-PAK5 pathway [58].

Furthermore, PAK5 has been implicated in stabilization via negative regulation of MARK-2 in non-neural cells [53]. PAK5 and MARK2 interact via their kinase domain and this interaction is independent of kinase activity of the two proteins.

1.2.6 Group II PAKs and adherens junctions

1.2.6.1 Background on adherens junctions

The adherens junction is a cell-cell adhesion in which the neighbouring plasma membranes are bridged by transmembrane cell adhesion molecules. Adherens junctions localise to the basal side of tight junctions in mammalian cells and play an important role in regulating tissue formation and morphogenesis during development as well as maintaining the physical association between cells in adult organism [59,

60]. Adherens junctions can involve typical cadherins (such as E-cadherin) or nectin which are different trans-membrane proteins. They mediate adhesive contacts through the extracellular region while their intracellular region interacts with the cytosolic proteins that connect them to the actin cytoskeleton (Figure 1.6).

All 20 members of mammalian classical cadherins share a common domain structure: the extracellular region composed of five repeats called the EC

(extracellular cadherin) domain that binds calcium. The adhesive function of cadherin

11 is dependent to its interaction with the calcium ions since this interaction is needed for conformational stability of the EC. Classical cadherins connect neighbouring cells through homophilic interactions of extracellular domains [61]. The cytoplasmic domain of classical cadherins interacts with two important catenins. Both beta-catenin and p120 catenin can bind simultaneously to cadherin 'tails'. The p120 catenin interacts with the juxta membrane domain of cadherin while β-catenin is associated with the carboxy terminal of the cytoplasmic domain [59, 60].

Nectins are a family of four proteins with extracellular immunoglobulin related repeats. Nectins do not bind calcium and their interaction with the nectins of neighbouring cells can occur in either homophilic or heterophilic manner, though heterophilic interaction leads to stronger cell-cell adhesions [59]. The cytoplasmic domain of nectins bind an important adaptor called afadin, via its PDZ domain.

Afadin also interacts with the small G-protein Rap1, and can interact with actin filaments [60]. Nectin and afadin complex may initiate adherens junction formation by recruiting cadherin-catenin complex [62].

12

ZO-1 Occludin ZO-2 TJ Claudin ZO-3 JAMs AF6

Nectin

AF6 AJ

p120 (V)E-cadherin β α

Figure 1.6 Schematic diagram of key components of adherens junctions and tight junctions. (Adapted from the review by Kooistra, et al. 2007 [63])

1.2.6.2 Localization of group II PAKs at cell-cell junctions

The association of group II PAKs with adherens junction has been shown. For example PAK4 is localized at cell-cell junctions of U2OS and MDCK cells in a

Cdc42 dependent manner, where it plays role in cell polarity [64]. PAK6 forms a complex with IQGAP and E-cadherin at cell-cell contacts of DU145 cells, leads to disassembly of cell-cell junctions [65]. In addition, both PAK4 and PAK6 phosphorylate Ser675 of β-catenin at cell-cell boundaries [62, 66]. P120-catenin has been reported as a binding partner and substrate for group II PAKs, although PAK5 binding seems to be most efficient. Ectopic expression of PAK5 and p120 catenin in

NIH-3T3 cells, showed their partial colocalization on cytoplasmic puncta. [54]. P120-

13 catenin is a well-established component of adherence junction and regulates several processes such as cell adhesion, reorganization of the cytoskeleton and cell movement

[67, 68]. Nucleocytoplasmic trafficking of p120-catenin plays role in

[69].

p120-catenin has four isoforms resulting from alternate translational start sites or post translational modifications [70]. The long isoform of p120-catenin is composed of four distinct functional regions, including a coiled-coil domain, a regulatory or phosphorylation domain, an armadillo domain containing ten armadillo repeats and a short C-terminal tail (Figure 1.7). The short isoform contains only the armadillo repeats and c-terminal tail. p120-catenin is ubiquitously expressed and despite the other members of the armadillo repeat proteins, it lacks the PDZ domain at

C-terminal. P120-catenin interacts with cadherin and increases cadherin stabilization in cell surface by regulating endocytosis and degradation of this protein, thus increases cell adhesiveness. Furthermore, p120 regulates Rho family small GTPases signalling through different mechanisms: it binds to and suppresses RhoA activity; although different isoforms of p120 vary in this ability. The p120-RhoA interaction is further stabilized by p120 phosphorylation on Tyr217 and 228, or destabilized by its phosphorylation at Tyr 112. The other mechanism is through its association with Rho

GAPs and Rho GEFs which leads to indirect regulation of downstream Rho GTP- ases. For example p120 binding to Vav-2 activates Rac-1 and Cdc42. P120 and Rho

GTPases signalling are involved in regulation of cell motility, however the function of p120 on regulation of Rho GTPases seems to be tissue dependent [70].

Besides its physiological function, p120 plays role in cancer progression and downregulation or mislocalization of p120 is shown in most of human cancers [70,

14

71]. For example translocation of p120-catenin from junctions to the cytoplasm or nucleus is observed in majority of breast cancers [72].

Coiled-coil ******

Armadillo Repeats

Figure 1.7 Schematic overview of p120-catenin functional domain (Adapted from the review by Daniel. 2007 [69]). Armadillo domain is composed of ten 42 amino acids armadillo repeats which mediate interaction with E-cadherin. The N- terminal coiled domain is involved in protein-protein interactions. Asterisks represent phosphorylation domain involved in protein regulation.

PAK5 has been shown to bind and phosphorylate p120 catenin mainly on

Ser288; however the biological function of the PAK5-p120 complex has not been assigned. As mentioned above, p120 is complicated in cytoskeleton reorganization through inhibition of Rho-A [73-75] and overexpressed p120 induces dendrite-like processes in NIH-3T3 cells through the Rho-A inhibition [76]. On the other hand,

PAK5 is also shown to inhibit Rho-A [33], although the underlying mechanism has not been completely understood. One suggestion is that p120 acts downstream of

PAK5 to drive Rho-A inhibition.

15

1.2.7 PAK5 and the regulation of apoptosis

1.2.7.1 Background

Apoptosis or programmed cell death is a physiological event which is applied for disposal of unwanted cells. It occurs as a normal process during development and aging and as a defence mechanism in a diseases condition [77]. Two major mechanisms of apoptosis have been identified including extrinsic or death receptor pathway and intrinsic or mitochondrial pathway [78] (Figure 1.8). The apoptosis mediated by the extrinsic pathway requires the binding of death receptor to the respective ligand on the cell surface. The death receptors are the members of the tumour necrosis factor (TNF) receptors family such as TNFR-1 and FasR that bind to their ligands TNF-α and FasL respectively. Ligand binding to its death receptor leads to recruitment of cytoplasmic adaptor proteins that are associated with the pro- caspases such as pro-caspase 8 and 10, causing to formation of death-inducing signalling complex (DISC) that finally results to activation of caspase 3 [79]. The intrinsic pathway is activated by a variety of receptor-independent stimuli such as radiation or serum withdrawal that produce some intracellular signals. These signals mainly target the inner membrane of the mitochondria, leading to loss of the function of mitochondrial permeability transition (MPT) pore as well as mitochondrial trans- membrane. This follows by release of pro-apoptotic proteins such as cytochrome c to the cytosol and subsequent formation of the 'apoptosome' that activates caspases such as caspase 9 and finally leads to the activation of caspase 3 [77, 79, 80]. Intrinsic and extrinsic pathways are linked and are concurrent in the same termination pathway which initiated with the cleavage of caspase 3. Caspase 3 cleavage subsequently leads to DNA fragmentation and cytoskeleton degradation.

16

The B-cell lymphoma-2 proteins (Bcl-2) family are a group of proteins involved in regulation of the intrinsic pathway by governing the permeabilization of mitochondrial outer membrane [81, 82]. They constitute BH (Bcl-2 homology) domains blocks and share homology in at least one of the four BH domains (Figure

1.9). Bcl-2 family are basically divided to three groups based on the composition of homology domain. 1. The anti-apoptotic proteins such as Bcl-2 and Bcl-XL that possess four BH domains. 2. Pro-apoptotic members contain BH domain 1-3 including Bax and Bak3. BH3 only domain proteins such as BAD, BID and BIM.

Pro-apoptotic and anti-apoptotic members can form heterodimers and this interaction influences their function [83]. For example, heterodimerization of BH3-only protein

BAD with Bcl-2, neutralizes anti-apoptotic function of Bcl-2. Mutation of BH1, BH2 and BH3 affect homo and hetero dimerization.

Most members of Bcl-2 family contain a trans-membrane domain that seems to be responsible for membrane localization of these proteins [84].

17

Figure 1.8 Schematic overview of two main apoptotic pathways (Taylor, et.al. 2008 [80]. Used with permission). Extrinsic pathway is activated by engagement of the death receptor to its ligand. This promotes recruitment of the Fas- associated death domain protein (FADD) which in turn recruits and activates caspase 8. Activated caspase 8 then activates caspase 3 and caspase 7 via proteolytic activity which finally leads to cell death. The extrinsic pathway can crosstalk with the intrinsic pathway through the proteolysis of BH3 only protein BID by caspase 8. Truncated BID (tBID) promotes release of mitochondrial cytochrome c. The intrinsic pathway is initiated by diverse receptor independent death signals such as serum withdrawal, which activates one or more BH3 only proteins. Activated BH3 only proteins can prevent anti-apoptotic activity of BCL-2 family members such as BCL-2. This promotes oligomerization of BAX and BAK proteins and formation of pores within the mitochondrial outer-membrane and releasing cytochrome c to the cytosol which promotes assembly of proteasome containing caspase 9. Caspase 9 can provoke activation of the other caspases through the proteolytic activity.

18

Figure 1.9 The Bcl-2 family of proteins (Danial NN. 2008 [85]. used with permission). Bcl-2 family members are divided to three groups based on the composition of the BH domains as indicated. They contain between 1-4 BH domains. Most of the members also contain trans-membrane domains (TM) and are therefore usually associated with the membrane. BH domains have crucial role for homo or hetero dimerization of the BCL-2 proteins.

1.2.7.2 PAK5 plays a role in cell survival

PAK family members are involved in apoptosis regulation through several pathways. There is evidence that PAK1, PAK2, PAK4, PAK5 and PAK6 promote cell survival [40, 52, 86, 87]. On the other hand, PAK2 is cleaved and activated by the caspase 3 activity during apoptosis and promotes certain morphological changes characteristic of apoptotic cells [88], therefore PAK2 is suggested to play antagonistic role in apoptosis [89]. One common pathway applied by PAKs to exert their role in cell survival is phosphorylation and inactivation of pro-apoptotic member of the BCL-

2 family, BAD. BAD triggers apoptosis by binding and neutralizing anti-apoptotic

19 partner BCL-2 which in turn promotes BAX-BAK oligomerization and formation of pore within mitochondrial outer membrane and release of cytochrome c [80]. The pro- apoptotic function of BAD is latent until its activation. Phosphorylation by survival kinases inactivates BAD by promoting BAD interaction with 14-3-3 protein and thus its maintenance in the cytoplasm (Figure 1.10). Cytoplasmic BAD cannot bind and neutralize mitochondrial Bcl-2. BAD contains three major evolutionarily conserved serine residues S112, S136 and S155 (mouse BAD enumeration) and their phosphorylation regulates the anti-apoptotic activity of BAD [85]. To be inactivated,

BAD requires phosphorylation in at least two of these residues.

A sequential phosphorylation model has suggested in which Bad Ser136 is the apical phosphorylation site, leading to phosphorylation of Ser112 and Ser136 [90].

Phosphorylated Ser112 and Ser136 allow Bad to bind various isoforms of 14-3-3 [91].

A nearby site, Ser155, is located in the BH3 domain which interacts with the hydrophobic groove of anti-apoptotic partners like BCl-2. Ser155 phosphorylation can prevents this interaction, but may first require phosphorylation of Ser112 and 136

[90]. This suggests that 14-3-3 binding to BAD induces conformational changes that expose Ser155 for phosphorylation. Binding to 14-3-3 also maintains BAD in the cytoplasm, so BAD phosphorylation not only influences its association with the mitochondrial partners, but also regulates its subcellular localization.

Various kinases have been reported to phosphorylate BAD. AKT and p70S6 target the Ser 136 residue [92, 93]. Ser 112 can be phosphorylated by different kinases such as protein kinase A (PKA), p90 ribosomal S6 kinase (p90RSK), Raf-1, PIM kinase and PAKs [86, 90, 94-97]. PKA and p90RSK are able to phosphorylate Ser155 residue as well [98]. Unlike the survival effect of phosphorylated residues 112, 136 and 155, phosphorylation of Ser128 abrogates 14-3-3 binding by interfering with

20

Ser136 phosphorylation. Cdc2 has been proposed to phosphorylate Ser128 of BAD and induces apoptosis in primary neurons [99].

Group II PAKs directly phosphorylate BAD on Ser112, whereas this is mediates through activation of Raf-1 in group I PAKs (PAK1) [40, 52, 100]. PAK5 has been reported to phosphorylate BAD on Ser136 as well; however this might be an indirect effect through Raf-1 [50].

14-3-3 p p p p p p Ser112 Ser136 Ser155 Ser112 Ser136 Ser155

BAD BAD

Cytosol

Survival Bcl-Xl BAD BAD

Mitochondria

Deathl

Figure 1.10 schematic diagram of apoptosis regulation through the regulation of BAD phosphorylation. (Adapted from the review by Liu and Lin. 2005 [101] with some modifications). BAD contains three major conserved residues S112, S136 and S155 whose phosphorylation regulate pro-apoptotic effect of BAD. Phosphorylated BAD is bound to phospho-site adaptor protein 14-3-3 and retained in the cytoplasm. The de- phosphorylation of BAD allows it to translocate to the mitochondria where it can form a complex with BCL-2 and Bcl-XL and finally leads to release of cytochrome c, thus promoting apoptosis.

21

1.2.8 The role of PAK5 on MAPK pathway

1.2.8.1 Overview of Raf-1 kinase

The Raf family kinases act as an effectors of Ras to mediate the cellular effect of growth factor signals to their downstream effectors MEK and ERK to regulate some cellular processes such as cell proliferation, differentiation, apoptosis and migration (Figure 1.11). The Raf family embraces three isoforms: A-Raf, B-Raf and

C-Raf (Raf-1). Raf-1 is the most investigated isoform of this family, however despite the extensive studies; Raf-1 regulation seems a complex process that is not completely understood. Raf-1 contains three conserved regions (CR); CR1 comprises of Ras binding domain (RBS) and cysteine-rich domain (CRD), CR2 contains several

Ser/Thr residues and CR3 that is the C-terminal kinase domain (Figure 1.12A). The

Raf-1 C-terminal contains residues including Ser 491 and Thr 494 in activation loop and Ser338 and Tyr 341 near the kinase domain whose phosphorylations mostly stimulate Raf-1 activation [102, 103]. In contrast, most of the phosphorylated residues required for negative regulation of Raf-1 are located at the N-terminal. These residues comprising Ser43, 233 and 259 are phosphorylated by PKA or PKB to create additional sites for 14-3-3 binding [104, 105], as well as Ser 29, 43, 289, 296 and 301 whose phosphorylation by ERK negatively regulate Raf-1 kinase activity [106]. The other ERK phosphorylation site, Ser 624, located at C-terminal. Negative regulation of Raf-1 by ERK phosphorylation suggests the existence of a negative regulatory feedback loop that is actually limiting the ERK activation in the cell.

As illustrated in figure 1.12B, in the inactive conformation, Raf-1 is auto- inhibited by binding of the N-terminal to the catalytic domain. 14-3-3 binding to the phosphorylated residues of the N-terminal and the C-terminal stabilizes this inactive conformation. The kinase activation of Raf-1 by growth factors requires

22 dephosphorylation of Ser259 by phosphatase PP2A and subsequent release of 14-3-3 protein from the N-terminal which recruits Raf-1 to the membrane where it binds to active Ras and becomes phosphorylated on at least four residues including Ser338,

Tyr341, Ser491 and Thr494 [107-110]. Phosphorylation of these residues stabilizes the Raf-1 active conformation. [111]. Some recent papers have suggested Raf-1 dimerization as a key part of activation. The mechanism is not clear but it is suggested that the active Ras induces Raf-1 dimerization through conformational changes. Raf-1 dimerization then triggers trans-autophosphorylation of Ser338 that stabilizes Raf-1 dimerization and its binding to the Ras.GTP [112, 113].

A negative feedback loop terminates Raf-1 activation via phosphorylation of inhibitory residues by ERK and PKB (Figure 1.12B).

1.2.8.2 PAK5 role on Raf-1 activation

Serine residue 338 is one of the important regulatory residues as its phosphorylation has been proposed to be necessary for Raf-1 activity [114]. In addition to auto-phosphorylation of Ser338, other kinases have been shown to target this site. All the PAK isoforms have been reported to phosphorylate Raf-1 on Ser338; moreover, PAK1 and PAK5 have been shown to interact with and activate Raf-1 [51,

115, 116]. The role of PAK4 in activating Raf-1 is not consistent; while one study has shown that PAK4 significantly activate Raf-1[116] another one has demonstrated that

PAK5 is the only member of group II PAKs which is able to activate Raf-1 [51].

PAK1 and PAK5 are reported to translocate Raf-1 to the mitochondria. PAK1 kinase activity is required for interaction and translocation of Raf-1 to the mitochondria.

Conversely, PAK5 interact with Raf-1 and translocate it to the mitochondria

23 independent of kinase activity [51, 117]. Mitochondrial Raf-1 is believed to protect cell from apoptosis either in a MEK dependent or MEK independent manner [117].

Besides PAKs, calcium/calmodulin-dependent kinase II (CaMKII) has been reported to induce Raf-1 phosphorylation on Ser338 and subsequent activation in response to Ras activation [118].

R Cell membrane T Ras-GDP Ras-GTP K P Raf SOS

P P MEK 1/2

P Erk 1/2 P

Nuclear substrates Cytosolic substrates

Figure 1.11 schematic of the classic MAPK kinase pathway (Adapted from the review by Matallanas, et al. 2011 [109] with some modifications). Activated receptor tyrosine kinase (RTK) recruits GEF SOS to activate Ras protein. Ras-GTP binds to and phosphorylate and activate Raf. Activated Raf phosphorylates and activates MEK which in turn phosphorylates and activates the downstream effector ERK. ERK has more than 150 substrates in the cytosol and nucleus; indicating the ERK involvement in variety of signalling pathways.

24

A S289 S29 S233 S296 S43 S259 S301 S642

CR1 CR2 CR3 RBD CRD S338 T491 S621 B Y341 S494

Figure 1.12 Schematic overview of Raf-1 domains (A) and mechanism of regulation. (B). A. (Adapted from a review from Baccarini. 2005 [111]). Raf-1 contains three conserved regions (CR); CR1, CR2 and CR3. CR1 comprises of Ras binding domain (RBS) and cysteine-rich domain (CRD). CR2 contains several Ser/Thr residues that their phosphorylations negatively regulate Raf-1 activity (Red residues). CR3 is the C-terminal kinase domain and contains residues that their phosphorylation induces Raf-1 activity (Green residues). B. (Matallanas, et.al. 2011 [109]. used with permission) Raf-1 phosphorylation on S259 and S621 stabilizes the inactive conformation of Raf-1 via binding to 14-3-3 protein in the cytosol of quiescent cells. Activated Ras recruits the inactive Raf-1 to the membrane where it is dephosphorylated on S259. This leads to the release of intermolecular bridge of 14- 3-3 protein. Subsequent phosphorylation on activation loop and C-terminal residues promotes Raf-1 homo-dimerization or hetero-dimerization (with B-Raf), leading to the full activation of Raf-1. Deactivation is initiated by de-phosphorylation of S338 as well as ERK-mediated negative feedback phosphorylation, which in turn leads to re- phosphorylation of S259 and subsequent intra-molecular binding of 14-3-3 protein.

25

1.2.9 The role of PAK5 on cancer progression

Enhanced cell proliferation, cell survival and motility are the important hallmarks of cancers. Since PAKs are involved in a number of oncogenic signalling pathways, they are widely considered as promoters of cancer progression. Among the group II PAKs, the importance of PAK4 in cancer development is well established

[119-121]. PAK4 has the potential of cell transformation and its over-expression promotes anchorage independent of [122]. Moreover, PAK4 mRNA has been up-regulated in many types of cancers including breast, ovarian, colon, lung and skin. PAK4 plays role in migration and invasion of several types of cancer cells such as prostate and pancreatic cancer cells [47, 89, 123]. Up-regulation of PAK6 protein has been demonstrated in some prostate and cell lines as well as the clinical samples from the patients with [124, 125]. Down regulation of PAK6 in prostate cancer cell lines decreases cell proliferation and enhances sensitivity to the Docetaxel [126]. In lung cancer, high expression of PAK6 is correlated with the poor survival of the patients [124].

In recent years, PAK5 has also been implicated in progression of several types of cancers. One study indicates that among all the kinases, knockdown of PAK5 has the most effect on apoptosis induction of pancreatic cancer cell lines MiaPaCa-2 and another study by the same group shows that the combined inhibition of both PAK5 and MAP3K7 slows down pancreatic adenocarcinoma progression in a mouse xenograft model [127, 128]. A gain of function mutation (GOF) in Tyr 538 (T538N) of PAK5 has been reported in lung cancer cell lines H2087 and H157. Overexpression of this mutated form of PAK5 increases Raf-1 and ERK phosphorylation versus the wild type protein. Consistent with this result, PAK5 suppression reduced cell proliferation and induced apoptosis in H2087 cells line [129].

26

High expression of PAK5 is contributed to development of , breast cancer, , osteosarcoma, and glioma [130-134] as well as resistance to Camptothecin and Paclitaxel treatment of colorectal and ovarian cancers respectively [132, 135].

The mechanism underlying PAK5 role in cancer progression has not been investigated in details. Abolished expression of PAK5 in gastric cancer cell lines

SGC-7901 and MGC-803 induced cell cycle arrest in G0/G1 phase which was in concordance with the down regulation of cell cycle’s regulator CDK2, CDC25A and cyclin D1; however the underlying molecular mechanism is yet to be addressed [136].

Similar to gastric cancer, PAK5 suppression in glioma cell lines U87 and U251 leads to G0/G1 cell cycle arrest as well as down regulation of cyclin D1 and increased expression of p21 [134].

PAK5 up-regulation has been shown to increase cell migration through the

PAK5-Egr1-MMP2 pathway in glioma and breast cancer cells [131, 134]. MMp2 is a matrix metalloproteinase that is involved in breakdown of extracellular matrix. MMp2 activity is inhibited by binding of the transcription factor Egr1 to its promoter [137].

Egr1 expression is increased in glioma and breast cancer cells upon PAK5 suppression.

Two recent studies indicate the PAK5 role in epithelial-mesenchymal transition (EMT) through phosphorylation of cadherin transcriptional repressors E47 and GATA1 in colon cancer and breast cancer respectively. EMT is involved in cancer metastasis which is characterized by the reduced expression of components of cell-cell adhesions such as cadherin.

Under HGF (hepatocytes growth factor) stimulation, PAK5 phosphorylates

E47 on Ser39. This phosphorylation increased cell migration and invasion in vitro and

27 promoted metastasis in colon cancer xenograft mice. Ser39 phosphorylation of E47 is shown to induce nuclear accumulation of E47 and its subsequent effect on cadherin suppression [138].

Another study showed that GATA1 acts as a suppressor of E-cadherin. PAK5 mediated phosphorylation of GATA1 decreased cadherin expression in breast cancer cells and induced cell migration in vivo [139].

1.2.10 PAK5 role in neuronal vesicle trafficking

PAK5 has been implicated in synaptic vesicle trafficking by binding to and phosphorylating the proteins synaptojanin-1 and pacsin-1 [140]. These two proteins were identified as PAK5 interacting partners in the murine brain extract using yeast two hybrid assay and indeed, they are the specific substrates for group II PAKs.

Synaptojanin has phosphatase activity and dephosphorylates phosphatidylinositol 4,5- bisphosphate (PIP2) that is required for un-coating and recycling of the synaptic vesicles [141]. Pacsin-1 is the brain specific isoform of pacsin family which plays role in endocytosis and vesicle trafficking. Pacsin-1 binds to proline rich domain of synaptojanin via its C-terminal SH3 domain and this interaction regulates vesicle dynamics at the synapse. Pacsin-1 and Synaptojanin phosphorylation by PAK5 promotes their interaction in vitro and in vivo. This role of PAK5 is contributed to the

PAK5/PAK6 knockout mice phenotype which shows cognitive deficits.

1.3 Overview of the LZTS family

The leucine zipper tumor suppressor (LZTS) proteins family is a group of related proteins termed LZTS1 (FEZ1), LZTS2 (LAPSER1) and LZTS3

(ProSAPiP1). Mapped on human chromosomal region 8p22, LZTS1 was the first

28 member of this group to be identified that is ubiquitously expressed in normal tissues

[142]. LZTS1 plays an important role in tumor suppression and its deletion significantly increases the rate of spontaneous or carcinogen induced tumors [143].

LZTS1 chromosomal region is frequently deleted and protein expression is reduced in human tumors [144].

LZTS2, named also LAPSER1, was identified based on homology with

LZTS1 gene [145]. It is located on human 10q24.3 near the PTEN gene and its deletion has been reported in various cancers [145]. LZTS2 contains three leucine zipper motifs at the C-terminal responsible for interaction with other proteins.

The protein is ubiquitously expressed and high expression has been detected in prostate and testis. Northern blot analysis demonstrated the expression of multiple isoforms of LZTS2 that are differentially expressed in different tissues and cell lines

[145]. LZTS2 expression is suggested to be regulated by NF-κB. Indeed, NF-κB directly binds to the LZTS2 promoter and regulates its transcription but this regulation differs among different cell lines. This might be the results of NF-κB interaction with different transcriptional co-regulators in different cell lines [146].

LZTS3 is the less investigated member of LZTS family which is predominantly expressed in the brain and kidney [147].

1.3.1 The role of LZTS2 on tumour suppression

The sequences alignment of LZTS1 and LZTS2 shows significant conserved sequences in these proteins suggesting that they are functionally similar (Figure 1.13).

Since LZTS1 is known as a tumour suppressor gene, several studies investigated the importance of LZTS2 as a tumour suppressor. The results of one study showed that overexpression of LZTS2 in several cell lines including human and Rat cancer cell lines slows down cell growth and its down regulation increases cell proliferation

29

[145]. Another study revealed that LZTS2 negatively regulates Wnt signalling pathway through modulating β-catenin localization [148]. They demonstrated that endogenous LZTS2 directly binds to β-catenin in colorectal cancer cell line SW480 and prevents nuclear accumulation of β-catenin, thereby repressing β-catenin mediated transcription. In neurons, LZTS2 targets to the post synaptic compartment via its interaction with the PDZ domain of ProSAP-Shank proteins. In this compartment LZTS2 interacts directly with β-catenin and regulates nuclear translocation and export of β-catenin in response to N-methyl-D-aspartate (NMDA)

[149]. However, the interaction of LZTS2 and β-catenin was not observed in lung cancer cell lines H1299 and A549, though overexpression of LZTS2 reduced the rate of cell proliferation in these cell lines. [150]. LZTS2 was shown to negatively regulate

AKT phosphorylation in these cells which could enhance the ability of Gsk3β to phosphorylate and promote β-catenin degradation (Figure 1.14).

In addition, LZTS2 mediates the crosstalk of β-catenin/Tcf and NF-κB pathways, however the effect differs in various cell lines [151]. LZTS2 inhibition consistently promotes β-catenin/Tcf pathway in various cancer cells but its suppression inhibits NF-κB activation in breast, colon and prostate cancers whereas increases it in glioma. This suggests that LZTS2 expression modulates tumorigenicity differently in different cell lines. However, it has been shown that LZTS2 repression increases tumour volume regardless of its effect on NF-κB activity in nude mice subcutaneously injected with colon cancer, prostate cancer or glioma cells [146].

Therefore, the function of LZTS2 on NF-κB pathway is ambiguous.

LZTS2 deficient mice are susceptible to spontaneous and carcinogen induced tumour development [152]. However, compared to LZTS1, LZTS2 knockout induces

30 lower rate of tumorigenicity. This suggests that additional genetic or epi-genetics changes are required to drive LZTS2 mediated tumorigenesis.

hLZTS1 ------MGSVSSLISGHSFHSKHCRASQYKL 25 hLZTS2 MAIVQTLPVPLEPAPEAATAPQAPVMGSVSSLISGRPCPGGPAPPRHHGP 50 **********:. . . . :: hLZTS1 RKSSHLKKLNRYSDGLLRFGFSQDSGHGKSSSKMGKSEDFFYIKVSQKAR 75 hLZTS2 PGPTFFRQQ----DGLLRGGYEAQEPLCPAVPPRKAVPVTSFTYINEDFR 96 .:.::: ***** *:. :. : . : :.:. * hLZTS1 GSHHPDYTALSSGDLGGQAGVDFDPSTPPKLMPFSNQLEMGSEKGAVRPT 125 hLZTS2 TESPPSPSSDVEDAREQRAHNAHLRGPPPKLIPVSGKLEKNMEKILIRPT 146 . *. :: .. :* . ..****:*.*.:** . ** :*** hLZTS1 AFKPVLPRSG------AILHSSPESASHQLHP 151 hLZTS2 AFKPVLPKPRGAPSLPSFMGPRATGLSGSQGSLTQLFGGPASSSSSSSSS 196 *******:. :: ....*:* . . hLZTS1 APPDKPKEQE-LKPGLCSGALSDSGRNSMSSLPTHSTSSSYQLDPLVTPV 200 hLZTS2 SAADKPLAFSGWASGCPSGTLSDSGRNSLSSLPTYSTGGAEPTTSSPGGH 246 :..*** . .* **:********:*****:**..: . hLZTS1 GPTSRFG-----GSAHNITQGIVLQDSNMMSLKALSFSDGGSKLG----- 240 hLZTS2 LPSHGSGRGALPGPARGVPTGPSHSDSGRSSSSKSTGSLGGRVAGGLLGS 296 *: * *.*:.:. * .**. * . : * ** * hLZTS1 ------HSNKADKGPSCVRSPISTDECSIQELEQKLLEREGALQKLQRS 283 hLZTS2 GTRASPDSSSCGERSPPPPPPPPSDEALLHCVLEGKLRDREAELQQLRDS 346 *. .::.*. .* * : ** ** :**. **:*: * hLZTS1 FEEKELASSLAYEERPRRCRDELEGPEPKGGNKLKQASQKSQRAQQVLHL 333 hLZTS2 LDENEATMCQAYEERQRHWQREREALR----EDCAAQAQRAQRAQQLLQL 392 ::*:* : . ***** *: : * *. . :. :*::*****:*:* hLZTS1 QVLQLQQEKRQLRQELESLMKEQDLLETKLRSYEREKTSFGPALEETQWE 383 hLZTS2 QVFQLQQEKRQLQDDFAQLLQEREQLERRCATLEREQRELGPRLEETKWE 442 **:*********:::: .*::*:: ** : : ***: .:** ****:** hLZTS1 VCQKSGEISLLKQQLKESQTEVNAKASEILGLKAQLKDTRGKLEGLELRT 433 hLZTS2 VCQKSGEISLLKQQLKESQAELVQKGSELVALRVALREARATLRVSEGRA 492 *******************:*: *.**::.*:. *:::*..*. * *: hLZTS1 QDLEGALRTKGLELEVCENELQRKKNEAELLREKVNLLEQELQELRAQAA 483 hLZTS2 RGLQEAARARELELEACSQELQRHRQEAEQLREKAGQLDAEAAGLREPPV 542 :.*: * *:: ****.*.:****:::*** ****.. *: * ** .. hLZTS1 LARDMGPPTFPEDVPALQR------ELERLRAELREERQGHDQM 521 hLZTS2 PPATADPFLLAESDEAKVQRAAAGVGGSLRAQVERLRVELQRERRRGEEQ 592 . .* :.*. * : ::****.**:.**: :: hLZTS1 SSGFQHERLVWKEEKEKVIQYQKQLQQSYVAMYQRNQRLEKALQQLARGD 571 hLZTS2 RDSFEGERLAWQAEKEQVIRYQKQLQHNYIQMYRRNRQLEQELQQLSLEL 642 ..*: ***.*: ***:**:******:.*: **:**::**: ****: hLZTS1 SAGEPLEVDLEGAD--IPYEDIIATEI 596 hLZTS2 EARELADLGLAEQAPCICLEEITATEI 669 .* * ::.* * *:* ****

Figure 1.13 ClustalW sequence alignment of human LZTS1 and LZTS2 ClustalW alignment depicts significant conserved sequences in LZTS1 and LZTS2. Asterisks show the identical sequences and dots indicate the conservative amino acids substitution.

31

Wnt

FRAT Dvl

LZTS2 AKT

A Axin Axin P GSK3 C FRAT Dvl Β-Catenin A GSK3 p P C p Β-Catenin Β-Catenin

LZTS2 Proteasome degradation Β-Catenin Nucleus TCF/LEF Gene transcription

Resting cells Stimulated cells

Figure 1.14 Schematic view of LZTS2 mediated regulation of β-catenin/Tcf pathway (Adapted from van wauwe and Haefner. 2003 [153]. with some modifications). In resting cells, β-catenin is targeted for ubiquitination and degradation through formation of the destruction complex. Destruction complex including APC/Axin/GSK3β binds to β-catenin and leads to phosphorylation of β-catenin by GSK3β and subsequent degradation. In stimulated cells, Wnt binding to its receptor disrupts degradation complex and prevents β-catenin degradation. Thus, β-catenin accumulates in the nucleus and drives its transcriptional activity by binding to transcription factor TCF/LEF. LZTS2 interferes with β-catenin function via two reported mechanisms: 1. LZTS2 binds to β-catenin and exports it to the cytoplasm through CRM1/exportin-regulated NES sequences located at the C terminus of the LZTS2. Thus, LZTS2 prevents β-catenin accumulation in the nucleus. 2. LZTS2 prevents AKT mediated accumulation of β-catenin in the nucleus. Activated AKT leads to phosphorylation and inactivation of GSK3β and subsequent accumulation of β-catenin in the nucleus. LZTS2 has been suggested to negatively regulate AKT activity via an unknown mechanism and prevents β-catenin mediated transcriptional activity of TCF/LEF.

32

1.3.2 LZTS2 and regulation of the cytoskeleton

A previous study has shown that LZTS2 localizes in either centrosome or midbody of osteosarcoma cell line Saos-2 depending to the cell cycle phase [154].

During interphase, metaphase and anaphase, LZTS2 co-localizes with γ-tubulin in the centrosome and then translocates to the centre of the mid-body during cytokinesis.

Ectopic expression of LZTS2 abrogates central spindle formation during cytokinesis and induces bi-nucleation in Saos-2 cells [154]. This effect is the consequence of

LZTS2 interaction with the microtubule severing protein katanin.

Katanin is composed of two subunits: 1. Catalytic P60 subunit (P60 katanin) and regulatory P80 subunit (P80 katanin). P80 katnin enhances the P60 severing activity [155, 156]. P80 Katanin localizes in the centrosome in prometa/metaphase and in the mid body during cytokinesis. LZTS2 has been shown to interact with the

P80 katanin via its C-terminal and inhibits severing activity of the katanin, therefore

LZTS2 disrupts central spindle formation which leads to bi-nucleation [157]. This is contributed to the tumour suppression activity of LZTS2, because bi-nucleated cells have the reduced ability of proliferation.

An shRNA screen suggested that LZTS2 knockdown decreases cell survival and sensitizes A549 cell line to paclitaxel treatment [158]. In agreement with this result, another study has shown that stable expression of LZTS2 in Rat-1 cells decreases sensitivity to the paclitaxel. This effect is contributed to the LZTS2 induced-cell arrest in prophase which prevents cells from entering into the metaphase thus inhibiting paclitaxel induced apoptosis [157].

33

1.3.3 LZT2 and cell migration

LZTS2 suppression increases cell migration in Rat-1 and SaOs-2 cells [157].

LZTS2 is suggested to negatively regulate cell migration through inhibition of katanin mediated microtubule severing [157] because inhibition of microtubule dynamics attenuates cell migration [159]. These results are controversial because wound healing assay in breast cancer, colon cancer and glioma cell lines showed that LZTS2 expression has no effect on cell migration [146],

1.3.4 The role of LZTS2 in development

LZTS2 and LZTS1 are expressed in mice embryo at stage E9.5 and E11.5 respectively [160]. LZTS2 knockout mice show sever defects in kidney and urinary tract development such as renal/ureteral duplication, hydroureter and hydronephrosis

[160]. Such a defects are very similar to the abnormalities detected in β-catenin loss of functions or gain of functions [161]. LZTS2 is expressed in ureteric bud and developing nephron structure which is similar to the expression pattern of β- catenin. This suggests that LZTS2 acts as a β-catenin regulator in nephrogenesis. In addition, β-catenin subcellular localization is affected in LZTS2 null mice. In wild type kidney, β-catenin is expressed in the cell junctions of the collecting duct epithelium as well as basolateral side of the renal tubule epithelium, while β-catenin is mainly cytoplasmic in LZTS2 null mice [160].

LZTS2 has also been implicated in PSD (post synaptic density) through interaction with the ProSAP/Shank family [162]. One study demonstrated the spatio- temporal expression of ProSAP/Shank family during xenopus laveis development and highlighted the similar expression pattern of LZTS2. They showed the co-localization of LZTS2 with the all three members of proSAP/Shank family when they were

34 overexpressed in COS-7 cells. Co-expression of LZTS2 with ProSAP/Shank family in forebrain, isthmus, nerves, Hypochord, Somites, pronephros and proctodeum suggests that LZTS2 plays role in PSD formation as well as embryonic development in xenopus laveis. [161]. Nevertheless, the function of the interplay of LZTS2 with

ProSAP/Shank family members has not been investigated.

LZTS2 is required for the specification of organ precursors of heart, pancreas and liver in zebrafish and negatively regulates the midline convergence. LZTS2 exerts this function through direct interaction with β-catenin and regulation of several marker that are implicated in regulation of convergence and extension movements in zebrafish [163].

1.4 Aims of the study

The identification of a conserved auto-inhibitory domain (AID) in group II PAKs

(isoforms 4-6) uncovered the mechanism of regulation of this group by the polarity protein Cdc42 [28, 29]. However, an outstanding question regarding PAK5 (and

PAK6) remained - why is PAK5 then constitutively active? Since its discovery in

2002, it was clear that PAK5 has much higher activity relative to PAK4 when isolated from mammalian cells whether or not it was bound to Cdc42 [33]. One object of my study was to identify the differences between the regulation of PAK4 and PAK5, and the mechanisms that are responsible for human PAK5 being basally active.

PAK5 is predominantly expressed in the brain: knockout studies in mice suggested it acts in a redundant pathway with PAK6 in brain development [49]. The kinase has received more attention because PAK5 is linked to cancer progression in a variety of situations [130-132, 135], however the basis for such activities are unclear.

One reason is that a few targets of PAK5 have been discovered. These targets are

35 likely to overlap with PAK4 given the high degree of conservation in the kinase domain [37]. In order to find valid kinase target for PAK5, I have sought to identify proteins which are closely associated with PAK5 in cancer cells. By identifying such a network it should now be possible to establish the protein network acting downstream of Cdc42 in this context.

36

Chapter 2: Materials and methods

2.1 Materials

2.1.1 Plasmids and siRNAs

The pXJ-40 mammalian expression vector [164] was applied as the base vector for generation of most of the constructs. The vector contains a CMV promoter and AmpR gene with N-terminal Flag, HA, GST and GFP tag. Furthermore, pXJ-GFP-

BirA* containing the mutated version of biotin ligase gene (R118G) was utilized to generate pXJ-GFP-BirA* -PAK5 that used for the BioID assay and GFP-pull down analysis. The pBABE.puro and pSUPER.neo contains puromycin and neomycin resistance genes respectively and were co-transfected with the target gen for establishing stable lines. The pGEX4T1 (GE Healthcare) was used for bacterial expression.

Smart-pool human PAK5 siRNA was purchased from Dharmacon. In other cases, the siRNAs (Table 2.1) were designed by siRNA design tool from Dharmacon

(http://dharmacon.gelifesciences.com/design-center). The oligos were synthesized by

Sigma.

Table 2.1 List of siRNAs

siRNA Sequences (5'-3'), Sense

siLZTS2-ORF CAACCUGUGGUCCAUAAACAAUU siLZTS2-3' UTR-1 CAAAGGAAGAACYCAGAAAUU siLZTS2-3' UTR-2 CCACUUGUCUCCUAGGAUCUU siAfadin GAAGAAAGAAGAUUGGAUAUU

37

2.1.2 Antibodies and reagents

Mouse monoclonal and rabbit polyclonal antibodies against Flag and HA were from Sigma. Mouse anti GFP obtained from Santa Cruz and rabbit GFP was from

Invitrogen. Mouse and rabbit anti GST obtained from Santa Cruz. Mouse monoclonal anti-Raf-1 and anti-p120 antibodies were from Upstate and Santa Cruz respectively.

Rabbit poly-clonal antibodies against phosphoS338 Raf-1, p44/42 ERK and BAD were purchased from Cell Signaling. The rabbit polyclonal antibodies against phosphoS112-BAD and phosphoS474, phosphoS602, phosphoS560 of PAK4/5/6 were from Epitomics. The anti-LZTS2 and anti-afadin rabbit polyclonal antibodies were from Protein Tech Group and Sigma Respectively. Secondary horse radish peroxidase conjugated antibodies were from Dako.

Restriction and DpnI were obtained from NEB. PFU was from Promega and T4 ligase from Thermo Scientific.

The PKA inhibitor (H89), pan Raf inhibitor (TAK-632) and Akt inhibitor

(MK-2206) purchased from “Selleck Chemicals”. PI3 kinase inhibitor (LY 294002) obtained from “Calbiochem”.

Puromycin and neomycin were obtained from Sigma.

2.2 DNA manipulation methods

2.2.1 Plasmid DNA purification

Plasmid DNA was extracted from transformed bacterial pellet using the

QIAprep (QIAGEN) columns according to the manufacture’s protocol. Briefly, the bacterial cells were lysed under alkaline condition and the released DNA was adsorbed on a silica matrix of the provided column under high salt conditions. DNA

38 washed with ethanol and then recovered from the column by elution in low salt buffer

TE.

2.2.2 Polymerase Chain Reaction (PCR) and gel electrophoresis

The standard PCR reaction mixture was provided as following:

10X PFU buffer+MgCl2 5μl (1X) dNTP Mix 200μM Foreword and reverse primers 0.5μM Template DNA 0.1μg PFU 1U Distilled water Up to 50μl

The standard PCR performed in these conditions: Step Temperature Time (min) Number of cycles (°C) Initial denaturation 95 2 1 Denaturation 95 0.5-1 Annealing Tm-5 0.5-1 25-30 Extension 72 2min/kb Final extension 72 10 1

Quick-change (QIAGEN) PCR performed in these conditions: Step Temperature Time (min) Number of cycles (°C) Initial denaturation 95 5 1 Denaturation 95 1 Annealing 55 1 20 Extension 68 1.5min/kb Final extension 68 10 1

After PCR, the samples were subjected to agarose gel electrophoresis. DNA was separated on an agarose gels (1-1.5% w/v) containing ethidium bromide

(0.5μg/ml) or gel green (1/10000v/v) by electrophoresis in TAE buffer and the separated bands were visualized by UV illumination. Electrophoresis was carried out at a constant voltage of 90-110 V for 1-1/5 h.

39

Quick-change PCR with 50 µl reactions was followed by DpnI digestion. 10 units of DpnI enzyme was added to 25μl of PCR sample and incubated in 37°C for

1h. 10μl of sample was used to transform 50μl of competent cells.

2.2.3 Preparation of competent cells for transformation via heat shock method

E.coli strains XL-1 Blue and BL-21 were used for DNA amplification and protein production respectively. A single fresh colony of E.coli was inoculated in 10 ml of LB agar and incubated in 37°C for overnight. The following day, the starter culture was diluted in 400 ml of LB media and incubated in shaker incubator at 37°C for around 2h till the OD600 reached to 0.4-0.6. The cells were collected with centrifugation at 2500 rpm for 15 min at 4°C and re-suspended in 60 ml of ice-cold

CCMB and incubated on ice for 20 min. The cells were recovered by centrifugation and re-suspended in 16 ml of CCMB buffer and aliquoted into 1.5ml Eppendorf tubes and stored at -80°C.

For DNA transformation, the frozen competent cells were thawed on ice and then mixed with the either plasmid DNA or ligated DNA reaction and incubated on ice for 10 minutes. The cells were heat shocked by incubating at 42 °C for 90s followed by chilling in 4 °C for 5 minutes. Transformed cells then were grown in LB media for 1h at 37°C and then plated on LB agar supplemented with the appropriate antibiotics.

2.2.4 Cloning

To clone a gene of interest in a plasmid with particular restriction sites, the gene was amplified using primers containing the sequence of suitable restriction enzymes. This allows adding the desired restriction sites to flank the insert. The PCR products as well as recipient plasmid were digested with appropriate restriction

40 enzymes for 2-4h at 37 °C. Digested samples were run in an agarose gel (0.5 -1.0%) electrophoresis and the band of the correct size was excised and the DNA was recovered using QIAquick gel extraction kit according to the manufacture’s protocol.

The concentration of purified plasmid and insert were calculated by NanoDrop 1000 and a molar ratio of 1:3 or 1:5 of vector to insert was used in a 10μl of ligation mixture containing T4 DNA ligase buffer and T4 ligase (New England BioLabs) and incubated overnight at 16 °C for ligation.

The following day, 5-7μl of ligation mixture was transformed to the E.coli

XL-1Blue via heat shock method and then plated in LB agar containing Ampicillin

(100μg/ml). The plate was then incubated in 37 °C for overnight and the next day 10-

15 colonies were selected and inoculated in LB media containing 100μg/ml of ampicillin. After overnight growing of bacteria in shaker incubator at 37 °C, the plasmids were purified after recovering the cells with centrifugation. Purified plasmids were subjected to digestion with the same restriction enzyme and the products were analyzed with gel electrophoresis to observe whether gene of interest ligated to the recipient plasmid.

To confirm the success of ligation and assess the accuracy of manipulated gene of interest, DNA sequencing was performed using a primer flanking the 5’ or 3’ end of the insert in a mixture containing BigDye (composed of DNA polymerase and tagged nucleotides) and the purified plasmid. The reaction was performed as following:

step Temperature (°C) Time (second) cycle 1 96 30 1 2 96 10

3 50 5 30 4 60 240 (4 mins) 5 4

41

2.3 Protein analysis methods

2.3.1 SDS polyacrylamide gel electrophoresis (SDS-PAGE)

8-12% acrylamide gel were prepared in mini gel cast and used for protein electrophoresis in BioRad mini-protean system. Protein samples mixed with the SDS sample buffer before loading. Electrophoresis was performed in SDS Running buffer at a constant voltage of 100-120 V for 1-2h.

2.3.2 Western blotting

SDS-PAGE separated proteins were transferred to PVDF membrane using tank blotting or BioRad semi-dry trans-blotting system. In most cases the membrane was stained in 0.05% Coomassie Blue before pre-blocked with 5% milk for 30 minutes. The membrane then was incubated with diluted primary antibody (1/100-

1/5000 dilution from 1mg/ml stock solution) in 3% milk or 1% BSA/PBS(or TBS) for

1-2h at room temperature or overnight at 4 °C and then washed 3 times in 0.1%

Tween 20/PBS (or TBS) for 30 minutes. The membrane was then incubated with diluted horse radish peroxidase (HRP) conjugated secondary antibody for 30-45 minutes. After washing the membrane for 20-30 minutes, signals were visualized with enhanced chemiluminescence reagents.

2.3.3 Expression and purification of bacterial expressed recombinant proteins

Bacterial expression vector pGEX4T1 containing the gene of interest was transfected into E.coli BL-21. A fresh single colony of recombinant bacteria was inoculated in 10ml of LB media containing 100μg/ml ampicillin and incubated overnight at 37 °C. The following day, the starter culture was diluted in 200ml of LB containing ampicillin and incubated in 37 °C till the OD 600 was 0.6. The culture was then induced with 0.5mM IPTG for 5h at room temperature. Cells were pelleted by

42 centrifugation at 4000 rpm for 20 minutes and re-suspended in ice-chilled bacterial lysis buffer contains lysozyme and allowed to equilibrate for 10 minutes on ice. The samples were then sonicated at 4°C in a triple cycle of 20 seconds pulse and 10 seconds interval and centrifuged at 35000rpm for 30 minutes at 4 °C. Clarified lysate was passed through the glutathione Sepharose beads column to purify GST tagged protein. The protein was then eluted using reduced glutathione containing buffer.

2.3.4 Antibody production

Two polypeptides were chosen from N-terminal (amino acids 7-15) and central region (amino acids 142-159) of PAK5. The synthesized (Genemed) used as a source of antigen for production of polyclonal rabbit antibodies (Genemed).

The crude sera were obtained and affinity purification was performed to gain the purified antibody against PAK5.

2.3.4.1 Coupling peptide to Cyanogen Bromide-Activated agarose for Antibody purification

Cyanogen bromide-activated resin was used for coupling to peptide according to the manufacture’s protocol. Briefly, 2 mg of peptide was dissolved in coupling buffer. Cyanogen bromide-activated resin was swelled and washed in cold 1mM HCl for 30 minutes, followed by washing with distilled water and then equilibrating with coupling buffer. The resin was mixed with dissolved peptide immediately and incubated at 4 °C overnight. The next day, un-reacted ligand was washed away using coupling buffer before blocking the resin with 0.2M glycine for 2 hours. The resin then washed with high and low pH buffer several times to remove the blocking buffer.

43

2.3.4.2 Affinity purification of antibody

5 ml of serum was centrifuged for 5 minutes at 13000 rpm, the supernatant then diluted with 5 ml of PBS and pre-cleared by rolling with Sepharose 4B beads for

1h at 4ºC. The serum passed through the column to remove the beads and flow through was collected. 500μl of peptide coupled resin was added to the serum and rolled at 4°C overnight. The suspension was passed through the column and washed with 5ml of cold PBS. Four tubes of 15 ml conical tubes were prepared by adding 30

μl of 1M Tris Ph 8.5 to the first tube and 50 μl to tubes 2-4. 500 μl of elution buffer added to column and flow through was collected in tube 1. This step was repeated for the tubes 2-4 very quickly. Then the pH of the collected fractions was checked by litmus paper. The ideal pH should be around 7-8. The specificity and sensitivity of the purified fractions of antibody were analyzed using Western blotting and antibodies with good immune-reactivity were mixed with 30-40% glycerol and kept in -20°C.

2.4 Cell culture techniques and manipulations

2.4.1 Cell culture

COS-7, HEK-293, MDCK and U2OS cells were grown in Dulbecco’s modified eagle’s medium (DMEM) supplemented with 10% fetal bovine serum (FBS) and penicillin-streptomycin (100I.U/ml and 100μg/ml respectively) (Pen Strep). HeLa cells were grown in DMEM supplemented with L-glutamine (2mM), sodium pyruvate

(1mM), sodium bicarbonate (1.5g/liter), non-essential amino acids (1X from 100X stock), FBS and Pen Strep. NCI-H446 cells were grown in RPMI 1640 medium supplemented with D- (4.5g/liter), L-glutamine (2mM), Sodium pyruvate

(1mM), sodium bicarbonate (1.5g/liter), FBS (10%) and Pen Strp. The other

44 mammalian cancer derived cell lines were grown in RPMI 1640 medium supplemented with 10% FBS. All cell lines were grown in 37°C and 5% CO2

2.4.2 Transient transfection

COS-7, HEK 293 and HeLa cells were transfected with GeneJuice (Novagen) reagent and U2OS, MDCK and NCI-H446 cells were transfected with Lipofectamine

2000 (Invitrogen) according to the manufacture’s protocol. 2-5μg of DNA was used to transfect cells with 70-90% confluency in 6cm plates. Cells were incubated for 16-

20 hours before harvesting.

2.4.3 siRNA transfection

Typically reverse transfections were used to transfect cells with 40-60nM of siRNAs using Lipofectamine 2000. Briefly siRNA and Lipofectamine 2000 diluted in serum free DMEM media separately and after 5minutes incubation the content of two tubes were combined and incubated for additional 20 minutes. Meanwhile the cells were trypsinized and the appropriate volume of cell solution in complete growing media was added to 35mm plates. The siRNA mixture then was added dropwise to the cells and cells were incubated for 48h at 37°C.

2.4.4 Establishing stable lines

Two types of stable lines were established: 1. U2OS cells that stably expressing GFP PAK5 and GFP-BirA*-PAK5 as well as corresponding controls 2.

NCI-H446 with stable knockdown of endogenous PAK5.

U2OS cells were grown in 35mm plate to around 70% confluency and then co- transfected with 2μg of target plasmid ( pXJ-GFP-PAK5 or pXJ-GFP-BirA*-PAK5) or control plasmids (pXJ-GFP or pXJ-GFP-BirA*) and 0.04μg of pBABE-puro using

45

Lipofectamine 2000 and were incubated for 48 hours. The expression of GFP PAK5 and GFP vector were analyzed using fluorescence microscope. Cells then were trypsinized and split in four 10cm plates with low density. After overnight incubation, the media were changed with the complete media containing 0.8μg/ml puromycin and cells were incubated for two weeks in the presence of puromycin. During this period, culture media were changed with a fresh media containing puromycin every 4-5 days to remove the dead cells. Cells which have integrated the resistance gene forms clones after 2-3 weeks which were examined for GFP expression with fluorescence microscope. Colonies with different intensity of GFP expression were selected and transferred using sterile whatman filter papers separately to media containing puromycin in 6cm plates. Confluent cultures then were used for further analysis using immunofluorescence staining to analysis protein localization and Western blotting to assess the protein expression.

To stably knockdown PAK5 in NCI-H446 cells, I used the pSUPER RNAi system. A siRNA sequence which specifically targets PAK5 and siScramble sequence were obtained from previously published data [132]. The oligos were purchased from sigma and cloned to the pSUPER.neo vector according to pSuper RNAi System manual (OligoEngine) to generate shRNAs. NCI-H446 cells were plated in 35mm plate and transfected with sh-PAK5 or sh-scramble using Lipofectamin 2000. Cells then were trypsinized in 48h and split with low confluency in several 10cm plates.

The following day, the media were changed with a fresh media containing 300μg/ml

G418. The remaining steps were identical to establishing U2OS stable line.

46

2.4.5 Immuno-precipitation

Cells were harvested and lysed in lysis buffer (Appendix) and then centrifuged in 13000 rpm for 15 minutes at 4°C. The appropriate volume of clarified supernatant were incubated with M2-Flag beads (Sigma), glutathione beads or streptavidin conjugated beads and rolled for 2h at 4°C. The beads then washed three times with lysis buffer and then either used for in vitro kinase assay or mixed with SDS sample buffer to use directly for Western blotting.

2.4.6 In vitro kinase assay

The immunoprecipitated kinase was subjected for in vitro kinase assay. 30μl of kinase mixture containing kinase buffer (1X), substrate (5-10μg/reaction), cold

ATP (500μM) and distilled water was added to the beads and incubated in 30°C for

20 minutes. For radioactive in vitro kinase assay, I used 10μg/reaction substrate,

40μM cold ATP and 2µCi 32P-γATP in 1X kinase buffer. The reaction was terminated by addition of hot SDS sample buffer (2X). In most of the cases, cold ATP was used and the samples were subjected to SDS-PAGE and Western blotting to detect phosphorylated substrate using specific antibody. When 32P-γATP was used for the kinase assay, the intensity of radiation detected by exposure to photographic film.

2.4.7 Immunostaining

Cells were seeded on cover slip and transfected with 0.5μg of desired gene for overnight. Cells were then fixed with either paraformaldehyde (PFA) 4% for 20 minutes at room temperature or cold methanol for 5 minutes at -20°C followed by washing with PBS for 10 minutes. Cells were permeabilized using PBS containing

0.2% Triton X-100 for 10 minutes and then blocked with 5% BSA in 0.1% Triton X-

100/PBS for 10 minutes and then incubated with relevant primary antibody diluted in

47

0.1% Triton X-100/PBS for 1-2 hours. The cells then were washed 3 times with 0.1%

Triton X-100/PBS for 30 minutes before incubating with the appropriate Alexa fluor conjugated secondary antibody (Alexa 488, 546 and 635) and DAPI (1.5μg/ml) for an hour. Cells were finally washed 3 times and coverslips were dried and mounted on glass slides using mounting solution.

For visualization of mitochondria, transfected cells were incubated with

Mitotracker CMXRose (life technologies) (150nM final concentration) for 20-30 minutes before fixation.

2.5 Methods for detections of protein-protein interactions

2.5.1 Establishing stable lines for BioID and GFP pull down analysis

PAK5 was cloned in pXJ-GFP-BirA* vector and this vector along with empty vector were used to generate U2OS stable lines under puromycin selection as described in section 2.4.3. Several clones of each line were analyzed using immunofluorescence staining and Western blotting for proper localization and expression of protein. Finally, two clones of each line with low expression of proteins were selected for BioID and GFP pull down (GFP-PD) analysis.

2.5.2 Adapting cells in SILAC media

To increase reliability of the results and reduce the rate of false positive data in mass spectrometry, SILAC strategy was adapted to isotopically label PAK5 lines. To this end, GFP-BirA*PAK5 lines were grown in SILAC DMEM media (deficient in both L-lysine and L-arginine, Thermo Scientific) supplemented with heavy isotopes

(13C,15N) of arginine and lysine (Cambridge Isotope Laboratories) and 10% dialyzed serum (Thermo Scientific) for two weeks. The control lines were adapted in the

48 identical media that was supplemented with the normal isoforms of arginine and lysine for a few days. The cells then were used for immuno-precipitation with

Neutravidin beads (Thermo Scientific) or GFP-Trap beads (ChromoTek).

2.5.3 Purification of biotin-labelled proteins from stable GFP-Bir*A cells

Two different clones of U2OS stable cells expressing GFP-Bir*A-PAK5 along with two clones of GFP-Bir*A stable cells were adapted in SILAC media for two weeks. The confluent culture of cells in 10cm plates were then grown in SILAC media containing 100μM of Biotin (Invitrogen) for 5 hours before they were used for immune-precipitation. Cells were rinsed with PBS and harvested in 500 µl of lysis buffer containing 100mM KCl, 25mM Tris pH 8, 0.5% Triton X-100, 0.5% sodium deoxycholate (DOC) and protease cocktail containing EDTA(Roche). Cells were spun for 20 seconds at 13000 rpm to pellet nuclei. The supernatants were collected and protein concentrations assessed using Bradford assay (40µl of the lysates were kept for further analysis). The supernatants were then mixed with SDS (final concentration of 0.2%) and sonicated briefly (14% power for 20 Sec) before centrifugation at 13000 rpm for 10 minutes at 4°C. The cleared lysates of the target (Heavy) and control

(Light) were pooled in one tube. Neutravidin agarose beads were pretreated with lysis buffer+0.2% SDS and then incubated with the clarified lysate at 4°C for overnight by rolling. The following day, the `flow through` was removed (50µ of `flow through` was kept for further analysis) and beads were washed in buffer comprising 25mM

Tris pH 8, 1% SDS and proteases cocktail for 5 minutes at room temperature and then in buffer containing 25mM Tris pH 7.4, 0.5% deoxycholate, 0.5% triton X-100,

100mM KCl and protease cocktail for additional 10 minutes at 4°C. 80μl of elution buffer (1X LDS sample buffer from novex) was added to the sample and incubated for 5 minutes at 100°C. Supernatant was collected and this step was repeated with

49

40μl of dH2O. The supernatants then were pooled in one tube (120μl total) and the tube was vacuumed to reduce the volume to 50μl. 40μl of the sample was run on 1mm

10% SDS-PAGE for 30 minutes till the dye front moved half way down the gel and remaining 10μl was kept for further analysis. The gel was submitted for mass spectrometry analysis.

2.5.4 GFP Purification of proteins from stable GFP-Bir*A cells

Two different clones of U2OS GFP-Bir*A-PAK5 along with two different clones of GFP-Bir*A vector lines (the identical clones that were used for BioID) were adapted in heavy media and light media respectively as previously described in section 2.5.1. The cells then harvested in 500µl of lysis buffer containing 100mM KCl

, 25mM Tris pH 8, 0.5% Triton X-100, 5mM MgCl2, 5mM DTT and protease cocktail (Roche) and then were cleared by centrifuge at 13200 rpm for 15min at 4°C.

The two Heavy lysates and the two light lysates were combined in the separate microtubes and protein concentration was assessed using Bradford assay (OD595 should be ~ 0.3) (50µl of the heavy lysate and 50µl of the light lysate were kept and maked up with 50µl 2x SDS buffer for further analysis). GFP-Trap beads were washed with 1ml of 0.1% Triton X-100 in PBS and incubated with the cleared lysate of the target (H) sample or control sample (L) in 4 °C for 3 hours by rolling. The

'flow-through' lysates were removed (50µl was kept for further analysis) and the beads were washed with 1 ml of lysis buffer, 3 times (5 minutes each). The proteins were eluted by addition of 40μl elution buffer (1X LDS sample buffer from novex) and incubation for 5 minutes at 100°C. This step was repeated with 20µl of H2O and the supernatants then were combined (60µl) and 10µl of each of heavy and light samples were kept for further analysis. The remaining 50µl of H sample and 50 µl of

L sample (total = 100µl) were combined and concentrated to 50µl in a Speed

50

Vacuum. 40µl of the elute proteins were run in 1mm 10% SDS-PAGE 30 minutes at

100V.

51

Chapter 3: Regulation of PAK5 activity

3.1 Introduction

Previously a PAK4 auto-inhibitory domain (AID) was identified through biochemical and structural analysis [28, 29]. This AID sequence which is located C- terminal to the Cdc42 binding domain is conserved across group II PAKs (Figure 3.1).

These sequences can adopt a pseudo-substrate like conformation in which a conserved proline (P52) occupies the position that the substrate serine would bind. The AID binding to the substrate binding site of the catalytic domain prevents access of substrate(s) and although the activation loop Ser474 is constitutively phosphorylated, the kinase is thus inactive. When expressed in mammalian cells, full length PAK4 has low activity compared to that of the catalytic domain [40] and this situation seems to prevail when the kinase is purified from E.coli [28]. In contrast to PAK4, PAK5 has much higher activity when isolated from mammalian cells [33]. Given that the PAK5

AID is 90% identical to that of PAK4 (Figure 3.1), it seems puzzling as to why the kinase is apparently constitutively active. In this chapter, I show that the PAK5 AID is indeed functional with respect to its ability to inhibit the kinase domain of PAK4, and show how sequences outside this region are responsible for the pronounced difference in basal activity of PAK5 when compared with PAK4.

52

PBD AID

hPAK4 –MFGKRKKRVEISAPSNFEHRVHTGFDQHEQKFTGLPRQWQSLIEESARR 49 hPAK5 –MFGKKKKKIEISGPSNFEHRVHTGFDPQEQKFTGLPQQWHSLLADTANR 49 hPAK6 MFRKKKKKRPEISAPQNFQHRVHTSFDPKEGKFVGLPPQWQNILDTLR-R 49 : *:**: ***.*.**:*****.** :* **.*** **:.:: *

hPAK4 PKPLVDPACITSIQPGAPKTIVRGSKGAKDGALTLLLDEFENMSVTRSNS 99 hPAK5 PKPMVDPSCITPIQLAPMKTIVRGNKPCKETSINGLLEDFDNISVTRSNS 99 hPAK6 PKPVVDPSRITRVQLQPMKTVVRGSAMPVDGYISGLLNDIQKLSVISSNT 99 Regulatory domain ***:***: ** :* . **:***. : :. **::::::** **:

Kinase domain hPAK4 -----ACTPAAPAVPGPPGPRSPQREPQRVSHEQFRAALQLVVDPGDPRS 314 hPAK5 ASYLSSLSLSSSTYPPPSWGSSSDQQPSRVSHEQFRAALQLVVSPGDPRE 442 hPAK6 -----TSNLYLPQDPTVAKGALAGEDTGVVTHEQFKAALRMVVDQGDPRL 400 : . . * . . .:. *:****:***::**. ****

hPAK4 YLDNFIKIGEGSTGIVCIATVRSSGKLVAVKKMDLRKQQRRELLFNEVVI 365 hPAK5 YLANFIKIGEGSTGIVCIATEKHTGKQVAVKKMDLRKQQRRELLFNEVVI 493 hPAK6 LLDSYVKIGEGSTGIVCLAREKHSGRQVAVKMMDLRKQQRRELLFNEVVI 451 * .::***********:* : :*: **** ******************

hPAK4 MRDYQHENVVEMYNSYLVGDELWVVMEFLEGGALTDIVTHTRMNEEQIAA 416 hPAK5 MRDYHHDNVVDMYSSYLVGDELWVVMEFLEGGALTDIVTHTRMNEEQIAT 544 hPAK6 MRDYQHFNVVEMYKSYLVGEELWVLMEFLQGGALTDIVSQVRLNEEQIAT 502

****:* ***:**.*****:****:****:********::.*:******:

hPAK4 VCLAVLQALSVLHAQGVIHRDIKSDSILLTHDGRVKLSDFGFCAQVSKEV 467

domain hPAK5 VCLSVLRALSYLHNQGVIHRDIKSDSILLTSDGRIKLSDFGFCAQVSKEV 595 hPAK6 VCEAVLQALAYLHAQGVIHRDIKSDSILLTLDGRVKLSDFGFCAQISKDV 553 ** :**:**: ** **************** ***:**********:**:*

Catalytic hPAK4 PRRKSLVGTPYWMAPELISRLPYGPEVDIWSLGIMVIEMVDGEPPYFNEP 518 hPAK5 PKRKSLVGTPYWMAPEVISRLPYGTEVDIWSLGIMVIEMIDGEPPYFNEP 646 hPAK6 PKRKSLVGTPYWMAPEVISRSLYATEVDIWSLGIMVIEMVDGEPPYFSDS 603 *:**************:*** *..**************:*******.:.

hPAK4 PLKAMKMIRDNLPPRLKNLHKVSPSLKGFLDRLLVRDPAQRATAAELLKH 569 hPAK5 PLQAMRRIRDSLPPRVKDLHKVSSVLRGFLDLMLVREPSQRATAQELLGH 696 hPAK6 PVQAMKRLRDSPPPKLKNSHKVSPVLRDFLERMLVRDPQERATAQELLDH 654 *::**: :**. **::*: ****. *:.**: :***:* :**** *** *

hPAK4 PFLAKAGPPASIVPLMRQNRTR---- 591 hPAK5 PFLKLAGPPSCIVPLMRQYRHH---- 719 hPAK6 PFLLQTGLPECLVPLIQLYRKQTSTC 681 *** :* * .:***:: * :

Figure 3.1. Comparison of N-terminal and C-terminal sequences of human group II PAKs. An alignment of human PAK4, PAK5 and PAK6 sequences using ClustalW. This highlights the N-terminal regulatory domain that contains a p21(Cdc42) binding domain (PBD) and auto inhibitory domain (AID). Identical residues are marked (*) while conservative residues are indicated as (:) or (.). The pseudo-substrate like sequence which binds the PAK4 substrate binding pocket and forms the major AID contact with the catalytic domain is shaded green.

53 3.2 PAK5 contains a functional auto inhibitory domain

Both PAK4 and PAK5 which belong to the group II (non-conventional) PAKs are highly related at their N-terminal regulatory domain and C-terminal kinase domain. As illustrated in Figure 3.1, the p21 binding domain (PBD), which binds

Cdc42, and the auto inhibitory domain (AID) lie within the first 60 residues. The critical contact of the AID with the kinase domain involves residues 49-52 (Figure

3.1, green mark). Constructs of PAK4 lacking residues 1-50 or mutated in the AID show >10 fold increase in activity relative to the wild type kinase [28].

To test if the PAK5 N-terminal similarly regulates kinase activity, I generated a series of deletion constructs (removing 20, 30, 40, 50 residues from the N-terminus).

These proteins were expressed in mammalian cells and immuno-precipitated to test their activity (Figure 3.2A). In vitro kinase assays were carried out in the presence of

ATP (0.5mM) and recombinant GST-BAD, which has been shown previously as a substrate for PAK5 [52]. As can be seen in Figure 3.2B, the PAK5 lacking residues 1-

50 (PAK5 Δ50) was several-fold more active than the full-length kinase. Comparing the kinase activity of PAK5 with PAK4 clearly indicates that PAK5 is indeed constitutively active (Figure 3.2D) and as previously reported [33]. Thus, although

PAK4 and PAK5 contain conserved regulatory and catalytic domains, the behaviour of PAK5 is quite different.

3.3 PAK5 is likely a Cdc42-specific effector

In the group I PAKs, Cdc42 binding to the PBD is thought to induce a large change in the conformation of the AID; thus displacing it from the kinase domain [8,

19]. Previously it has been shown that PAK5 binds preferentially to Cdc42 rather than

Rac1 [45], unlike PAK1 [8]. The higher activity of PAK5 relative to PAK4 might be explained by a different sensitivity to the different Cdc42 related GTPases. Therefore,

54

I tested the ability of PAK5 to interact with Cdc42 related GTPases (mutationally activated) when co-expressed in COS-7 cells. Surprisingly, although TCL (RhoJ) and

TC10 (RhoQ) are highly related to Cdc42, they could not efficiently bind PAK5

(Figure 3.3A). Next, using an in vitro kinase assay I examined whether binding to

Cdc42 can regulate PAK5 activity. Indeed, Cdc42 could bind to and enhances PAK5 activity ~2.5 times; confirming that in-vivo PAK5 is likely to be regulated by Cdc42

(Figure 3.3B,C), as was reported for PAK4 [28]. It should be noted that this kinase regulation cannot be demonstrated in vitro [28]

55

A Isoform Kinase 108 specific 420 domain 719 PAK5-FL PAK5∆20 PAK5∆30 PAK5∆40 PAK5∆50

B

Flag constructs

Vector

PAK5 ∆ 30 ∆ PAK5 PAK5 ∆ 20 ∆ PAK5 40 ∆ PAK5 50 ∆ PAK5 PAK5 (FL) PAK5 + + + + + + GST BAD (FL) kDa Anti-p-S112 BAD 50

< GST BAD

Flag IP: anti-Flag (PAK5) 75 in vitro kinase assay C

3.1

activity kinase 1.2 0.9

1.0 0.7

Relative Relative

50 20 30 40

FL

Δ - Δ Δ Δ

PAK5 PAK5 PAK5 PAK5 PAK5

Figure 3.2 N-terminal deletion induces PAK5 activity. A. Schematic of PAK5 constructs used for expression in COS-7 cells B. The various Flag- PAK5 constructs were purified and tested in vitro for their kinase activity toward BAD Ser-112. Kinase activity was assayed in vitro using full-length recombinant GST-BAD (5μg) and 0.5mM ATP in 30ul kinase buffer. Phosphorylated BAD was detected using anti-pS112 BAD antibody that detects the major site of PAK5 phosphorylation [52]. C. The relative in vitro kinase activity of purified PAK5 constructs averaged in three independent experiments. The data were normalized relative to full-length PAK5 (FL = 1).

56

D 50 Δ

Flag constructs

+ PAK4 PAK4 + PAK5 + kDa PAK4 + GST BAD

50 anti-pS112 BAD

< GST BAD

75 Flag IP: IB: anti-Flag

in vitro kinase assay

Figure 3.2 N-terminal deletion induces PAK5 activity. D. COS-7 cells were transfected with the indicated constructs and immunoprecipitated constructs were subjected to in vitro kinase assay ( in Sepharose.)

A

1(G12 V) 1(G12 Flag constructs

-

+ + (G18V) TC10 + TCL(G30V) + Rac kDa (G12V) + Cdc42 HA PAK5

HA-PAK5 (co-IP) 75 Flag IP

25 IB: anti-Flag

Input: anti-HA

Figure 3.3 Co-transfection of Cdc42 (G12V) increases the in vitro activity of PAK5. A. HA-PAK5 was co-expressed with the mutationally activated form of GTPases most closely related to Cdc42. Rac1 used as a negative control. The interaction was examined by anti-Flag pull-down.

57 significant The PAK Sepharose assessed PAK Flag activity Figure 5 5 relative beads : Flag Cdc42.G12VFlag C B Cdc was Flag IP 3 of . 3 42 BAD using . PAK indirectly Co and C Flag activity GST GST complex . HA - kinase transfection 5 The an . Flag PAK5 PAK5 PAK5 PAK5 kDa 25 B 75 BAD 50 . in recovered is - relative Co PAK (data vitro were Relative kinase activity shown - immunoprecipitation 5 + + - - not kinase in vitro in vitro was of tested 1.0 in . using shown) Cdc Cdc PAK5 vitro + + + directly - kinase kinase assay assay in 42 42 58 Flag 2.6 kinase . V

three PAK5+ + + (G - - 12 Cdc42(GV12) - immune 12 of Cdc by V) the independent was + - - - activity 42 itself increases immobilized - V performed precipitated anti PAK5 IB: IB: IB: anti Flag < 12 GST did . anti - of p The - S112 BAD - not BAD PAK p experiments the - activity Ser602 using pull kinase 5 and in or - down vitro anti was HA the in - - . 3.4 PAK5 auto-regulatory domain can inhibit PAK4

Full-length PAK5 cannot be expressed in E.coli (data not shown) and the mammalian expressed kinase is clearly very active (Figure 3.2D). Since PAK4 and

PAK5 have highly conserved catalytic and regulatory domains, why is PAK5 already active when expressed in mammalian cells? One study has shown that mutationally activated MKK6 (EE) phosphorylates PAK5 on Tyr 608 and up-regulates the basal activity of PAK5 as shown for PAK4 and PAK6 as well [36]; however, I found no change on activity of PAK4 and PAK5, when co-expressed with constitutive active

MKK6(EE) (data not shown). Another untested possibility is that the PAK5 auto- inhibitory domain has become ineffective because of the differences in amino acids residues (Figure 3.1). The best way to test this seemed to exchange the regulatory domain of the two kinases, since the region on the catalytic domain to which the AID is bound [29] is essentially identical. I therefore created two chimaeras in which the regulatory domains (residues 1-72) were exchanged (Figure 3.4A). The results shown in figure 3.4B demonstrate that the ‘PAK4(72)-PAK5’ chimaera was equally active to wild type PAK5 and conversely ‘PAK5(72)-PAK4’ was fully inhibited relative to wild type PAK4. Thus PAK5 AID is fully functional and the determinant responsible for the high activity of PAK5 must reside outside the auto-inhibitory domain.

59

isoform Catalytic CRIB/AID A specific 419 domain PAK5 719 291 PAK4 591

switch AID - chimaera

72 419 PAK4(72)-PAK5 719

72 291

PAK5(72)-PAK4 591

PAK4 PAK5

- - B

Flag constructs

PAK4 PAK5(72) PAK5 PAK4(72) kDa 50 anti-p-S112 BAD

< GST BAD

Flag IP: 75 IB: anti Flag in vitro kinase assay

Figure 3.4 The PAK5 auto-inhibitory domain is functionally active towards PAK4 catalytic domain. A. Schematic diagram of two PAK4 and PAK5 chimaeras. The first 72 amino acids of PAK4 and PAK5 contains PBD and AID were exchanged to make a chimaera of PAK5 contains PAK4 regulatory domain Flag-PAK4(72)-PAK5 or Flag- PAK5(72)-PAK4 B. The indicated Flag tagged constructs were expressed in COS-7 cells and the fusion proteins recovered by immunoprecipitation. in vitro kinase assay was performed using GST- BAD as a substrate in the presence of 0.5mM ATP. The GST-Bad (10μg / lane) was then probed with anti-pSer112.

60 3.5 Different cellular distributions of PAK4 and PAK5

Could the differences in activities of PAK4 and PAK5 relate to their intra- cellular location? For example there are reports that PAK4 and PAK5 can be found in the nucleus [50, 165]. Epitope tagged versions of PAK4 localize to the cytoplasm, at cell-cell junctions and in the nucleus [26, 66]; the situation with PAK5 is less clear

[50, 52, 53]. Some studies have suggested that the kinase is found in clusters that mainly co-localizes with mitochondria [31, 52], while this punctate distribution of

PAK5 in other circumstances does not appear to be mitochondrial [53, 54]. In my hands, the PAK4 and PAK5 clearly have very different basal level of activity, and I was interested to investigate whether this might relate to their compartmentalization.

In COS-7 cells, Flag-PAK4 showed diffused cytoplasmic distribution whereas Flag-

PAK5 demonstrated a punctate distribution even when cells were transfected with

0.2μg of Flag-PAK5 plasmid for 8 hours (Figure 3.5). These puncta are regular in size and do not resemble aggregates that typically form by unfolded proteins [53, 166].

Therefore, I hypothesized that PAK5 contains sequences that promote self-assembly.

Comparison of the primary sequence of PAK4 and PAK5 indicates the presence of a central region in PAK5 which is evolutionary conserved and is not found in PAK4

(Figure 3.6).

To test which PAK5-specific region is involved in targeting the protein to puncta, I generated a series of N-terminal deletion PAK5 constructs which were expressed at low level in COS-7 cells (Figure 3.7A). The removal of the N-terminal

109 residues clearly has two effects; firstly I noticed that in contrast to PAK5∆50 construct, the protein was not completely excluded from the nucleus, suggesting that residues 50-109 contain a nuclear export signal. Secondly, the puncta formed by

Flag-PAK5∆109 were more diffuse (Figure 3.7B). This suggested that this N-terminal

61 region, which is well conserved in PAK4, participates in puncta formation although it is unlikely to be responsible for this phenomenon. Only by removing residues 1-236 from PAK5, I was able to completely alter the distribution of PAK5 such that it remained largely cytoplasmic, with some accumulation at peripheral membrane ruffles

(which is typical for this cell type). By contrast PAK5∆178 construct was relatively punctate. This data suggested the importance of region 178-236 in puncta formation.

To test this, I generated flag-PAK5 lacking the region 178-268. The protein

PAK5∆178-268 showed more diffuse cytoplasmic distribution with a few puncta but was not distributed as evenly as PAK4 (Figure 3.7B).

3.6 The non-catlytic domain of PAK5 is sufficient for puncta formation

An inactive version of PAK5 (PAK5 S602M/T606M) was found to localize similar to the active kinase (data not shown). Removal of the catalytic domain

(PAK5(1-418) construct) did not disrupt the ability for puncta formation of PAK5 but resulted in ir-regularly sized punctate (Figure 3.8B) suggesting the participation of the kinase domain in formation and stabilization of the normal puncta. I therefore, generated smaller N-terminal constructs to establish the limits required for this process given that PAK5 residues 109-425 either are not found or not conserved in

PAK4 (Figure 3.6). This data suggested that the whole central region of PAK5 is required for puncta formation (Figure 3.8B).

62

A Anti-Flag Anti-Flag

PAK4 PAK5

Anti-Flag B

PAK5

Figure 3.5 Subcellular distribution of PAK4 and PAK5 in COS-7 cells. A. COS-7 cells were transiently transfected with 0.5μg of either Flag-PAK4 or Flag-PAK5 constructs in 35mm culture dishes containing glass coverslips; At 16h the cells were fixed in 4% paraformaldehyde (20 min, RT) and immunostained with mouse anti-Flag antibody. B. COS-7 cells were transfected with 0.2μg of Flag-PAK5 and after 8h expression the cells were fixed and processed for staining with anti-Flag antibody. (Scale bar = 10 μM)

63 hPAK5 RKESPPTPDQGASSHGPGHAEENGFITFSQYSSESDTTADYTTEKYREKS 150 cPAK5 RKESPPTPDQGASSHGQGHREENGFITFSQYSSESDTTADYTTEKYREKS 150 gPAK5 RKESPPAHHQGNANHVPRHQEENGYITYSQYSSESDTTTDYVMDKYRDKT 150 hPAK4 RRDSPPPP------ARARQENGMP------EEPATTARGGPGKAGS-- 134 *::***. :*** .*. **: * .

hPAK5 LYGDDLDPYYRGSHAAKQNGHVMKMKHGEAYYSEVKPLKSDFARFSADYH 200 cPAK5 LYGDDLDPFYKGSHVAKQNGHVMKTKHGEAHYPEMKPWKSDLARFPLDYH 200 gPAK5 LYGEELDRYYKGGYAAKQNGHMMKVTSRDIYYSEMTPLKSDLSRFLPDYH 200 hPAK4 ------RGRFAG------HSEAGGGSGDRRRAGP--- 156 :* ... :.* ..* *

hPAK5 SHLDSLSKPSEYSDLKWEYQRASSSSPLDY--SFQFTPSRTAGTSGCSKE 248 cPAK5 THLDSLSKPSEYSDLKWEYQRASSSSPLDY--PFQFTPSRTAGTSGCSKE 248 gPAK5 AQVEAKPKPLEYGGLKLEYQRIPSGSSLDYRDPFPYALSRASVQSECPKE 250 hPAK4 ------

hPAK5 SLAYSESEWG--PSLDDYDRRPKSSYLNQTSPQPTMR-QRSRSGSGLQEP 295 cPAK5 SLVFSESEWG--PSLDDYDRRPKSSYLNQTSPQPTMR-QRSRSGSGLQEP 295 gPAK5 RLEYSDGDWGHSPAKDDYDKRPKSSYVDPASPQPTMR-QRSRSGSGLQEP 299 hPAK4 ------EKRPKSSREGSGGPQESSRDKRPLSGPDVGTP 188 ::***** . .** : * :*. **..: *

hPAK5 MMPFGASAFKTHPQGHSYNSYTYPRLSEPTMCIPKVDYDRAQMVLSPPLS 345 cPAK5 MMPFGASAFKTHPQGHSYNSYTYPRLSEPTMCIPKVDYDRAQMVLSPPLS 345 gPAK5 AMPYGASAFKAHQQGHSYSSYTYPRLSETAAGISKVDYDRAQLVVSPPLS 349 hPAK4 -QPAGLASGAKLAAGRPFNTYPRADTDHPSRGAQGEPHDVAPNGPS---- 233 * * :: *:.:.:*. . ...: :* * *

hPAK5 GSDTYPRGPAKLPQSQSKSGYSSSSHQYPSGYHKATLYHHPSLQSSSQYI 395 cPAK5 GSDTYPRGPAKLPQSQSKAGYSSSSHQYPSGYHKATLYHHPSLQTSSQYI 395 gPAK5 GSDTYPRGPVKLPQSQSKVSYSSSSYQYPLVYHKASHYHQPSLQPSSPYI 399 hPAK4 ------AGGLAIPQS------SSSSSRPPTRARGAPSPGVLGPHASEPQL 271 * :*** **** : * : *. . :.*. :

hPAK5 STASYLSSLSLSSSTYPPPSWGSS 419 cPAK5 STASYLSSLSISSSTYPPPSWGSS 419 gPAK5 STASYPSSPSITSSAYPPPSWGSS 423 hPAK4 APPACTP----AAPAVPGPPGPRS 291 :..: . ::.: * *. *

Figure 3.6 ClustalW sequence alignment of the central region of PAK4 and PAK5. The central region of PAK5 is moderately conserved across vertebrates, however this sequence is not found in human PAK4. Identical sequences are marked (*) and conserved amino acids are shown with (.) or (:). In PAK5 species, dark grey and light grey highlights the identical amino-acids or conserved amino-acids respectively. h, Homo sapiens; c, Canis lupus familiaris; g: Gallus gallus.

64 A isoform Catalytic puncta 108 specific 419 domain 719 PAK5-FL +++ PAK5 (51-719) +++ PAK5 (109-719) ++ PAK5 (179-719) ++ PAK5 (237-719) - PAK5∆178-268 ++

B Anti Flag Anti

PAK5 (full length) PAK5 (51-719) PAK5 (109-719) Anti Flag Anti

PAK5 (179-719) PAK5 (237-719) PAK5Δ178-268

Figure 3.7 The central region of PAK5 is involved in puncta formation. A. The diagram indicates the various constructs of PAK5 were used to test cellular localization by immunostaining. B, C. COS-7 cells were transfected with 0.5μg of the indicated constructs for overnight then fixed in 4% PFA and processed for indirect immunostaining using mouse anti-Flag antibody. Removal of residues 1-236 significantly changes the kinase distribution

65 A isoform Catalytic 108 specific 419 domain PAK5-FL 719

PAK5 (1-268) 268 PAK5 (1-310) 310 PAK5 (1-418) 418

B

PAK5 (1-268) PAK5 (1-310)

PAK5 (1-418) PAK5–FL (1-719)

Figure 3.8 The PAK5-specific central region is primarily involved in puncta formation. A. A schematic of C-terminal truncated constructs of PAK5 used to test cellular localization. B. Plasmids encoding tagged versions of PAK5 (1-268), PAK5(1-310) and PAK5(1-418) were generated and expressed in COS-7 cells. Indirect immuno-fluorescent staining was performed.

66 3.7 Establishing the regions of PAK5 involved in self-association

Puncta distribution of a protein is contributed to its oligomerization in some reports. For instance, overexpressed dishevelled (dvl), one of the component of the

Wnt signalling pathway, form puncta that represents its oligomerization [167]. Given that PAK5 contains residues not found in PAK4 and are conserved across evolution

(Figure 3.6), it seemed possible that these play an important role in self-association.

To examine this hypothesis and map the possible self-interaction site of PAK5, co- transfection experiments were carried out using HA-PAK5 and testing if this protein could be co-immunoprecipitated with Flag N-terminal truncated constructs of PAK5.

These experiments suggested the PAK5 kinase domain was not involved in PAK5 self-association (Figure 3.9B) and that residues 268-420 in this context seem critical for PAK5 self-association. To determine whether the residues 1-108 of PAK5, which are largely conserved in PAK4, play a key role in self-association of PAK5, the similar experiment was performed using HA-PAK5∆109. The results demonstrated that HA-PAK5∆108 co-immunoprecipitated with various truncated PAK5 constructs with roughly the same efficiency (Figure 3.9C). Further experiments were then carried out to further assess region required for PAK5 self-interaction. N-terminal truncated constructs of PAK5 were tested for their ability to interact with HA-PAK5(1-268).

This contains approximately 50% of the PAK5-specific central region and yet retained the ability to self-associate albeit not as efficiently as the full length protein (Figure

3.10A). The results showed that although HA-PAK5(1-268) interacts with Flag-

PAK5∆108, it could not be co-immunoprecipitated with Flag-PAK5∆236 implicating the importance of 109-236 region in self-interaction (Figure 3.10A).

Taken together the immuno-precipitation analysis suggested two critical regions in PAK5 self-association: residues 109-236 and residues 268-420. To test

67 more directly the importance of central region of PAK5 in self-association, I generated PAK5 lacking residues 268-420 (Figure 3.10B). As can be seen in figure

3.10C, Flag-PAK5(∆268-420) failed to co-immunoprecipitate HA-PAK5(∆268-420).

Thus the biochemical data support the model that residues in the PAK5-specific region play key role to promote PAK5 self-association. In agreement with these results, gel filtration analysis in our lab has suggested that despite monomeric PAK4,

PAK5 is dimer or oligomer (Figure 3.11)

68

A

isoform Catalytic 108 specific 419 domain 719 PAK5-FL PAK5 (41-719) PAK5 (109-719) PAK5 (179-719) PAK5 (237-719) PAK5 (269-719) PAK5 (420-719)

B Flag constructs

+ PAK5 ∆108 PAK5 + ∆178 PAK5 + ∆236 PAK5 + ∆268 PAK5 + ∆419 PAK5 + Lysate + + PAK5 ∆40 PAK5 + HA PAK5 kD 75

50 IB: anti Flag Flag IP

37

IB: anti HA (co-IP) 75

75 Input: anti HA

Figure 3.9. Biochemical experiments show PAK5 has the ability of self-interaction. A. Schematic of N-terminal deletion constructs of human PAK5 that used for co-immunoprecipitation. B. HA tagged PAK5 were co-expressed with the indicated Flag tagged N-terminal deletion constructs of PAK5. Anti-Flag immunoprecipitation was performed after cell lysis and clarification. The bound proteins were eluted from sepharose by adding SDS sample buffer. These were subjected to SDS-PAGE and transferred to PDVF membranes for Western blot analysis.

69

268 360 418

C 108

FL

Δ Δ Δ

Δ Flag constructs -

HA-PAK5 Δ108

+ PAK5 PAK5 + + PAK5 PAK5 + PAK5 + PAK5 + + kD PAK5 + 75 IB: anti-HA

75 Flag IP 50 IB: anti-Flag 37

75 Input: anti-HA

Figure 3.9. Biochemical experiments show PAK5 has the ability of self-interaction. C. HA-PAK5 ∆108 was co-transfected with Flag tagged N-terminal truncated constructs of PAK5. Cell lysate were harvested and Co-immunoprecipitation was performed using anti-Flag Sepharose. Indirect binding of HA-PAK5∆108 was tested.

A

Flag constructs

+ PAK5 ∆N236 PAK5 + ∆N268 PAK5 + kDa ∆N108 PAK5 + HA PAK5 (1-268) 37 anti-HA (co-IP)

75 Flag IP IB: anti-Flag

Input: anti-HA

Figure 3.10 The central PAK5-specific region is required for self- interaction. A. The Flag tagged N-terminal truncated construct of PAK5 were co-transfected with HA-PAK5(1-268). Cell lysates were harvested and immuno-precipitation was performed using anti-Flag Sepharose. The samples then subjected to SDS-PAGE to detect co-immunoprecipitation of HA-PAK5 1-268.

70 immunoprecipitation constructs in interaction Figure self C B - assembly 3 . 10 . of PAK5∆268 PAK5 B The . PAK A Flag IP was - PAK central FL 5 analysis (∆ generated 5 - 108 (∆ 420 268 PAK ) 75 75 75 or as - 420 5 . indicated - PAK specific C ) + - . construct PAK5Δ108 COS 71 5 (∆ 268 - + - by PAK5∆268-420 7 region - cells Western lacking 420 < Anti Anti Anti HAPAK5∆108 HAPAK5∆268 NS Flag constructs ) were is - - - and HA(co HA(Input) Flag the required blotting transfected region tested - IP) . - 420 for involved for self with co - - Void 440kD 150kD 66kD Mr Markers

kDa 3 4 5 6 7 8 9 10 11 12 13 14 15 16 Fractions 97 Flag-Pak1 < PAK1 (61 kDa) 66 97 Flag-Pak4 < PAK4 (64 kDa) 66 97 < PAK5 (79 kDa) Flag-Pak5 66 Superose 6 fraction: IB: anti-Flag

Figure 3.11. Gel-filtration chromatography of Flag-tagged PAK1, PAK4, PAK5 proteins. PAK1, PAK4 and PAK5 were expressed in HEK293 cells by transient transfection from pXJ-Flag vectors. Transfected cells were lysed with lysis buffer containing 25mM HEPES pH=7.3, 100mM KCl, 5mM DTT, 10% glycerol +protease inhibitor cocktail (Roche). Cell lysates was clarified at 60,000rpm (30min, 4oC) using a TLA-100 rotor in a Beckman TL-100 ultracentrifuge (~100,000g). Fractions of 50ul were injected in a Superdex™ 200 column on a SMART system (Pharmacia) pre-equilibrated with running buffer (25mM HEPES pH=7.3, 100mM KCl). A total of 30 fractions (50ul) were collected for each sample. Fractions collected were assessed by anti- Flag western blotting. Molecular weight standards as shown were run under identical conditions and the peak fraction assessed by absorption at 280 nm.

72 3.8 PAK5 self-association contributes to its high basal activity

Based on the localization of various truncated versions of PAK5 (Figures 3.7,

3.8) and the results of co-immunoprecipitation experiments (Figures 3.9, 3.10), I suggested that a key role of the PAK5-specific domain is self-interaction. In a number of other contexts, dimerization or oligomerization of protein kinases is associated with their (auto) activation [168-171]. The construct PAK5(∆268-420), with ~50% of the

PAK5-specific sequences deletion, showed a fully dispersed protein in the cytoplasm

(Figure 3.12B). The construct has lost the ability of self-interaction (Figure 3.10C) and can be considered essentially monomeric (ie. like PAK4). I therefore tested the in vitro kinase activity of this truncated kinase versus the full-length PAK5 (Figure

3.12C). This smaller protein had much lower activity showing a correlation between self-association and a high basal activity of PAK5 constructs (Figure 3.12C, lane 4 versus lane 2). Other mechanisms could be responsible however; for example the deleted region might contain sequences required for interaction with cellular activators

(protein or otherwise), or this region could promote binding of substrates, and thereby affect the efficiency of the catalytic domain to target substrates as has been reported for other protein kinases [172-174]. To test these possibilities, I also generated a

PAK5(∆116-260) construct (Figure 3.12A) that lacks the remaining ‘PAK5-specific’ sequences (ie not present in PAK4). Immuno-fluorescence staining showed diffuse pattern of cytoplasmic localization of this protein in COS-7 cells (Figure 3.12B). In vitro kinase assay indicated this protein exhibited a reduced activity versus the wild type protein (Figure 3.12C, lane 3 versus lane 2). These results further support the notion that PAK5 oligomerization can be abolished by removing all or part of the

PAK5-specific region which in turn affects basal activity.

73

How might PAK5 self-association lead to elevated kinase activity? The first issue was whether all these constructs were constitutively phosphorylated on activation loop Ser602 as was suggested previously for group II PAKs [28].

Therefore, I tested the phosphorylation status of this residue for PAK5(∆116-260) and

PAK5(∆268-420) compared to wild type PAK5, using a phospho-specific antibody that recognizes the sequence around the unique phosphorylated serine in activation loop of group II PAKs. The result showed the activation loop is phosphorylated with the same extent in all PAK5 constructs (figure 3.13A). A second possibility is that

PAK5 self-interaction interferes (in trans) with the AID binding to the catalytic domain. This hypothesis would be consistent with reduced activity of ‘monomeric’

PAK5(∆268-420). To test this, I investigated the effect of mutating the AID at equivalent residues to those in PAK4 that cause activation [28] (Figure 3.13B). This

PAK5(∆268-420)-mAID protein exhibited elevated activity relative to the non- mutated version, indicating that this shorter PAK5(∆268-420) is indeed auto-inhibited by the functional AID (Figure 3.12C). This data also corroborates the original domain swapping experiment that indicated the PAK5 AID is indeed functional (Figure 3.4).

74

in deletion Fig precipitation staining deletion elevated Figure B C A PAK

3 anti-Flag . 2 5 ) 3 constructs constructs in . of PAK5 PAK5 PAK5 12 in activity Flag 50 . PAK and the The % 75 50 - Δ Δ - PAK5 FL L2 L1 of 5 subsequent lysate . central . of the deletion A Δ . PAK in vitro in vitro 116

central + PAK4 Schematic of 108 115 PAK 5 - 260 . COS + PAK5 isoform constructs in kinase kinase assay specific B 5 267 region . vitro - - specific 261 Diagram 7 + PAK5Δ116-260 diagram cells kinase 419 75 . + PAK5ΔL268-420 B 420 Flag . in expressed domain Catalytic region Indirect illustrates + assay - COS PAK5 of Flag GST GST BAD < IB: IB: anti IB Flag IP: PAK - GST is 7 : : anti Immuno was - Δ constructs 268 responsible the cells 5 Flag 719 - - - BAD pS112 BAD performed Flag - constructs regions 420 . - - PAK fluorescence C . 5 Immuno (FL) deleted for (as with its or in - western plasmid AID Figure function 3 blotting . constructs 13 A . The . using A . PAK 75 50 50 75 kDa COS for anti 5 16 specific - 7 - h Flag PAK4 . cells Cells PAK5 and central were

were PAK5ΔL1 anti 76

- PAK5 ΔL2 pS transfected then region 602 PAK5- ΔL2-mAID - < lysed PAK anti (Ser602) anti PAK5 Flag may - - 5 Flag p with and antibodies constructs - interfere FL PAK5 processed the indicated with . the for B A E AID mutation

hPAK4 PRQWQSLIEESARRPKPLVDPACIT 60

hPAK5 PQQWHSLLADTANRPKPMVDPSCIT 60

mAID

-

420

420 -

C -

268

268

mAID

Δ

- Δ

Flag constructs

+ PAK4 + PAK5 + PAK5 + PAK5 + + + PAK4 + GST BAD kDa

50 anti-pS112 BAD

< GST BAD

75 Flag IP: IB: anti-Flag

in vitro kinase assay D 13

11 10.5

activity kinase 2.5

1

Relative Relative

L2

Δ

-

mAID

PAK4 PAK5

-

mAID

-

L2

PAK5

Δ

-

PAK4 PAK5 Figure 3.13. The PAK5 specific central region may interfere with the AID function B. The position of the point mutations generated in the PAK5 AID sequence are indicated and as described previously [28]. C. COS-7 cells were transfected with the indicated constructs of PAK5 and PAK4. The in vitro kinase assay was performed directly on the anti-Flag Sepharose beads with recombinant GST-BAD. The relative in vitro kinase activity of the purified PAK5 constructs tested in three independent experiments. The data were normalized and Flag-PAK5(FL) value set to 1.

77 3.9 Discussion

Identification of a conserved AID domain at N-terminal of group II PAKs [28,

29], has suggested that the activity of group II PAKs are regulated in a similar manner to group I PAKs. In its auto-inhibited conformation, a key region of N-terminal AID

(amino acids 49-52), which acts as a pseudo-substrate bound to catalytic domain, prevents kinase activity. Binding to Cdc42 induces conformational changes results in relief from auto-inhibition. Although this mechanism applies to PAK4, Cdc42 itself cannot interfere with pseudo-substrate binding and the mechanism by which PAK4 is activated remains to be addressed. It has been suggested that the proline-rich sequence in PAK4 could interact with SH3 domains [29] which would allow substrate access.

Another complexity is that although the regulatory domain of group II PAKs

(including AID) is conserved (Figure 3.1), PAK5 is clearly much more active than

PAK4 when expressed and purified from mammalian cells (Figure 3.2D) [45]. Ching et al, suggested that an alternate AID exists in PAK5 (amino acids 60-180) which is not conserved among other members of group II PAKs [30], however this observation does not explain the enhanced activity of PAK5 relative to PAK4. One of the questions I wished to address in my studies was, if PAK5 contains an active and functional AID, why is it much more active?

3.9.1 PAK5 activity is partially regulated by the presence of a functional AID at

N-terminal

The presence of a in group II that serves as an AID in

PAK4 has been noted [29]. Removal of 50 amino acids of PAK4 N-terminal, which contains the AID, leads to ~10 fold increased activity [28]. Since PAK5 is much more

78 active than PAK4, when purified from mammalian cells (Figure 3.2D), the outstanding question is whether the PAK5 indeed contains a functional AID? I have been able to show here that human PAK5 does contain an AID which has strong similarities with that found in PAK4. These issues are discussed further in the next section:

1. Removal of 50 amino acids of PAK5 N-terminal increased the basal activity of the protein ~3 times (Figure 3.2B, C) which is significant though less dramatic than that seen with activity of PAK4 [28].

2. Binding to activated Cdc42(G12V) in vivo can increase kinase activity > 2.5 times (Figure 3.3B, C). Because the group II PAKs activity is not regulated in-vitro by binding to Cdc42(G12V), literatures regarding PAK5 regulation by Rho-GTPases is confusing [31, 33]. Certainly the ability of Cdc42 to regulate PAK5 localization

(for example at filopodia) might provide the right cues to drive kinase activation - say by lipids. The suggestion that RhoD is a partner for PAK5 at mitochondria [31] is not borne out in my studies (Table 5.2).

3. Swapping the first 72 amino acids of PAK4 and PAK5 was informative (Figure

3.4). Firstly, a chimaera of PAK4 replacing with the PAK5 N1-72 was clearly fully inhibited, indicating that the conserved PAK5 AID is likely to bind tightly to the

PAK4 catalytic domain. Secondly, in the context of the full-length PAK5 protein,

PAK4 AID was not effective in suppressing catalytic activity to the level seen with full-length PAK4. This suggests these kinases may have diverse mechanisms of regulation. I conclude that human PAK5 contains a sequence that can act as a functional auto-inhibitory domain (against PAK4 catalytic activity).

79

3.9.2 The punctate distribution of PAK5 reflects kinase oligomers

In multiple cell types, GFP-PAK5 or Flag-PAK5 show unusual puncta distribution in the cytoplasm, but not PAK4 which is diffuse (Figure 3.5) or presents in cell-cell junctions [175]. Other studies have also demonstrated such a distribution for PAK5 using various tags in mammalian cell lines such as CHO, HeLa, HEK293 and NIH-3T3 cells [31, 51, 53, 54]. One hypothesis I developed was that puncta observed with PAK5 were connected to its elevated activity. A trivial explanation for puncta formation is aggregation as an unfolded protein [166]. Matenia et al [53] indicated an absence of the Vimentin cage-like structure typical under these conditions [53]. Thus, such aggresome formation is unlikely in the case of PAK5. I was able to show that puncta formation requires the presence of a specific sequences

(residues 109-420) mapped in the central region of PAK5 which are largely conserved in PAK5 in vertebrates but not present in PAK4 (Figures 3.6). My results suggest that most of the central region is required for puncta formation as PAK5(1-310) is unable to form the puncta (Figure 3.8). In addition, immuno-fluorescent staining of the truncated constructs of PAK5 suggests that neither the conserved N-terminal regulatory nor catalytic region is critical for puncta formation (Figures 3.7 and 3.8) suggesting they may play role to stabilize puncta.

Most interestingly, my studies show that PAK5 has the ability to self-associate

(Figures 3.9 and 3.10). Co-immunoprecipitation experiments demonstrated that the full length PAK5 can bind to N-terminal truncated constructs of PAK5 but not the catalytic domain (Figure 3.9B). The removal of residues 1-109 did not affect PAK5 self-interaction (Figure 3.9C), whereas removing the central region disrupted PAK5 dimerization (Figure 3.10C). This region has no homology to other proteins in the database, and does not resemble coiled-coil sequences which are often used for self-

80 association [176]. Thus, the regions of PAK5 that are conserved in PAK4 are clearly not critical for PAK5 dimerization, although we would expect the AID at the N- terminus to bind the catalytic domain (in cis or trans).

Here, PAK5 has clearly been shown to undergo self-association and/or oligomerization, as noted for a number of other proteins [177, 178]. For example

Dishevelled (Dvl) protein, a critical adaptor for Wnt signaling pathway, shows punctate distribution in endogenous or ectopic expression which is contributed to its oligomerization and subsequent induction of Wnt pathway [179]. The result of gel filtration analysis suggests PAK5 exhibits both dimeric and oligomer states after extraction without detergent (Figure 3.11). However, I do not exclude this possibility that a portion of PAK5 might be presented in monomeric state.

3.9.3 The central region of PAK5 plays role in elevated activity of the protein

Protein dimerization or oligomerization could have physiological outcomes: it places proteins in proximity to each other, perhaps to allow regulation of kinase activity or enhancing the specific protein-protein interaction site [180, 181]. Based on the experiments conducted in section 3.7, it appears PAK5 self-association is the key event that regulates the observed increased kinase activity. For example PAK5(∆268-

420) which lacks self-interaction (Figure 3.10C) has much lower in vitro kinase activity than the full-length protein (Figure 3.12C). One possible mechanism is that these residues increasing binding with substrates as reported for the other proteins

[14, 182]. However this model is complicated by the fact that PAK5(∆116-260), has also lower kinase activity compared to the full length protein. (Figure 3.12C). The phosphorylation status of Ser602, likely the only residue modified in the activation loop, is not influenced by deletion of central region (Figure 3.13a). Given that this

81 residue is constitutively phosphorylated [28], changes in activity are unlikely to be regulated through A-loop phosphorylation.

Interestingly PAK5(∆268-420) activity can be increased through an established mutation in the AID that prevents its binding to the catalytic domain (Figure

3.13.B,C). The significant lower activity of PAK5(∆268-420) coupled to its loss of oligomerization shows a good correlation between these two properties. Examples of similar phenomena include the Raf-1 homo- or hetero dimerization that is required for

Raf-1 signaling in several cellular contexts [183]. Similarly most receptor tyrosine kinases such as epidermal growth factor receptor (EGFR) undergo oligomerization that leads to enhanced activity of the kinase [169, 184].

Interestingly, PAK5 activity is increased a few-fold by preventing the AID function through different strategies (Figures 3.2, 3.3). For example, removal of the first 50 amino acids or binding to Cdc42 increases PAK5 activity ~ 3-fold. There are two hypotheses that could explain this observation: 1. The PAK5 is in equilibrium between monomeric and oligomeric kinase, and therefore the mutation selectively activates the monomeric pool. 2. Oligomerization could partially interfere with the function of the AID; therefore removal of the AID would give higher activity. The second hypothesis is supported by the observation that both PAK5Δ50 (Figure 3. 2B) and PAK5 bound to Cdc42(G12V) (Figure 3.3C) showed a mobility shift on SDS-

PAGE indicating additional phosphorylation (at as yet unidentified sites) after full loss of the AID function.

Overall, the elevated activity of PAK5 clearly correlates with its oligomerization. In this proposal a previous study found that PAK5 activity could be suppressed by a fragment from the central region of PAK5 [30]. Here, GST-PAK5 was incubated with

82 the N-terminal fragment (N180) and there was reduced auto-phosphorylation and activity towards MBP however, the N60 fragment had no effect. The interpretation suggests that residues 60-180 (not conserved in other members of group II PAKs) are inhibitory. This suggestion can also be explained if the N180 fragments can interfere with PAK5 oligomerization as I have shown in a model presented in figure 6.1B.

83

Chapter 4: Analysis of potential PAK5 substrates BAD and Raf-1

4.1 Introduction

In general protein kinases are activated at specific cell locations, for example

PKA acts at specific sites in the cell because it binds to AKAPs (A kinase anchor proteins) [185, 186]. It is not clear if PAK5 is targeted through intrinsic signals or requires specific adaptor proteins for its subcellular localization. In common with

PAK4, it has been reported that PAK5 shuttles between the cytoplasm and nucleus due to the presence of nuclear localization sequences (NLS) and nuclear export sequence (NES) [50]. The NES seems to be dominant over NLS and the protein is observed primarily in the cytoplasm. Previously it had been suggested that

'mitochondrial localization sequences' are present at the N-terminal region (residues

30-83 and 400-411) [50] and within the kinase domain of PAK5 [31] .The underlying mechanism to establish mitochondrial targeting has not been explored in detail. Nonetheless, given the low level of PAK5 expression in cell lines

(http://www-test.ebi.ac.uk/) and the poor quality of commercial anti-PAK5 antibodies (my observations) it is unclear where endogenous PAK5 is localized.

When tagged PAK5 was expressed stably in CHO cells or transiently in neuronal cell line HMN1, the protein was found to be mitochondrial [52]. In agreement with this report another study showed mitochondrial localization of overexpressed PAK5 in

HeLa cells [31]. However, a number of studies have failed to detect mitochondrial

PAK5 [53, 54].

Mitochondrial localization of PAK5 could be important for its observed anti- apoptotic activity through phosphorylation of BAD [50] and indeed PAK5 was suggested to translocate Raf-1 to the mitochondria [51], where it is known to

84 promote pro-survival pathway. BAD and Raf-1 are common substrates of group I and II PAKs. While the both groups of PAKs phosphorylate Raf-1 on a regulatory site Ser338 [51, 187], the group I kinases may have different preference site on BAD and actually PAK1 preferred the adjacent Ser111 rather than Ser112[100]. To test if group II PAKs are important kinases for BAD and Raf-1 in U2OS cells, a panel of kinase inhibitors were used; H89 (PKA and ROCK inhibitor) [188, 189], MK-2206

(Akt inhibitor) [190], TAK-632 (pan Raf inhibitor) [191] and PF3758309 (PAK inhibitor) [192]. First, U2OS cells were treated with different concentrations of Akt and Raf inhibitors to assess the effective concentration. As I have shown in Figure

4.1, treatment of U2OS cells (lacking PAK5, Figure 4.13) with PAK inhibitor

PF3759309 strongly affected BAD Ser112 phosphorylation. It was notable that Raf-1

Ser338 phosphorylation and the level of p-ERK1/2 were significantly reduced by treatment with PF3758309. The effect of the PAK inhibitor on BADSer112 phosphorylation was much more pronounced than inhibiting PKA or Raf-1. This result needs to be validated with other PAK inhibitors such as FRAX120[193].

It has long been known that phosphorylation of Ser112 and Ser136 of BAD allows its interaction with 14-3-3 protein and retention in the cytoplasm [91, 194]. It has been established that many apoptotic signals lead to BAD de-phosphorylation, and its subsequent translocation to the mitochondria where it binds to BCl-2 and prevents the anti-apoptotic properties of this protein by disrupting its binding to the

BAX protein. BAX-BAK oligomerization forms some pores within mitochondrial outer membrane that leads to release of cytochrome C and subsequent apoptosis [77,

80] (Figures 1.8 and 1.10). PAK5 has been shown to directly phosphorylate BAD

Ser112 and indirectly increase the level of phospho-Ser136 [52] which is the gatekeeper site for 14-3-3 binding [91]. Given that PAK5 clearly localizes in puncta

85 in multiple mammalian cell types (Figure 4.2), I was interested to know if this represents a subset of mitochondria (these never resemble typical elongated and interconnected mitochondria [195]).

4.2 Subcellular distribution of PAK5

Indirect immuno-staining was performed on COS-7, HeLa and HEK-293 cells expressing full-length Flag-PAK5 or GFP-PAK5. Mitochondrial localization of full-length PAK5 in HeLa and HEK-293 cells has been reported [31, 51]. Figure 4.2 shows the typical distribution of PAK5 in distinct puncta throughout the cytoplasm.

In cells co-stained with Mitotracker there was little co-localization between the two markers (Figure 4.2). Interestingly truncated constructs such as PAK5(1-236) did show some mitochondrial co-localization suggesting that in certain contexts PAK5 might be able to associate with mitochondria [31, 50] (Figure 4.3A).

86

for then of ( ( were serum to concentrations phosphorylation Figure (DMSO independent were MK 5 10 C B A Serum Serum μM) 10 assess μM) - SDS 2206 % lysed serum or carried starved 4 FCS or treatment stimulation . - PAK 1 PAGE the AKT in The Anti - - starved (where experiments RIPA DMSO 25 50 75 effective inhibitor for of out inhibitor - effect + and pS473 . AKT 4 =

1μM 1h buffer hours 1 using A + + indicated) western ) - for . . of inhibitor 5μM - PF concentration Akt + + MK DMSO 4 relevant U 3758309 before and

2 10μM hours, - - image OS 2206 blot were + + 50

(MK H89 for TAK addition µg - cells analyses ( kinase additional then 2 - ( J 2206 . 5 + + normalized of . 5 - 87 μM) 632 + + + + + + of μM), MK-2206 The total treated - the were ) of for inhibitors or or C + + the protein inhibitors

pan TAK-632 10 averaged . 15

pan 1μM Quantitation - inhibitors 1h minutes treated with minutes relative Raf 5μM Raf + + PF-3758309 Anti loaded on inhibitor PKA . 10μM P - P inhibitor P U p - . - values - BAD 2 - S112 . S338 ERK1/2 before ERK1/2 The to with OS B 1μM inhibitor of in . 2h the U (TAK and - cells cells - the each 5μM 2 BAD Raf TAK OS different addition from control bands

- 10μM Raf - 1 were were cells 632 - lane H 632 89 - - DMSO-1h 2 1 )

- DMSO-2h

Anti-Flag

7

- COS

PAK5 WT Mitotracker

Merge

Anti-Flag

293

-

HeLa HEK

PAK5 WT GFP PAK5 WT

Figure 4.2 Subcellular localization of PAK5. Flag or GFP-PAK5 constructs were transfected in COS-7, HEK-293 and HeLa cells as indicated. The following day, cells were fixed in 4% PFA, permeabilized and processed for immunostaining with mouse anti-Flag antibody as indicated. GFP-PAK5 transfected cells were analyzed after fixation without further processing. Mitochondrial staining of live COS-7 cells were performed using Red Mitotracker (CMXRose) for 20 min, the cells were then washed in PBS and fixed in 4% PFA. Puncta distribution of PAK5 did not overlap the typical elongated mitochondrial network. (Scale bar: 10µm)

88 Anti-Flag Mitotracker Merge

PAK5 1-236

PAK5 1-268

Figure 4.3 PAK5 contains a cryptic mitochondrial localization sequence. A. COS-7 cells were transfected with truncated constructs of Flag-PAK5. the following day, cells were live stained with the red Mitotracker (CMX Ros, 150nM), then fixed and immunostained with mouse anti-Flag antibody. PAK5 1-236 is largely localized to the mitochondria. (Scale bar:10µm)

89 In previous chapter I demonstrated that PAK5 has an ability to oligomerize, which promotes kinase activation. To investigate whether PAK5 oligomerization can affect protein compartmentalization, I tested the effect of oligomerization on the nuclear localization of PAK5. PAK5 contain a basic nuclear localization sequence

(NLS) in residues 5-10 as well as a nuclear export sequence (NES) [50]. The reported NES is mapped in residues 400-411 of PAK5 which is not conserved in

PAK4 and PAK6. Several truncated constructs of PAK5 in my study suggested that

PAK5 should contain another dominant NES. For example PAK5 ∆268-420 (which lacks the reported NES at residue 400-411) is still cytoplasmic (Figure 3.12B), though its molecular weight is less than the allowable size of 90-110kD for diffusion through the nuclear pore [196]. Work in our laboratory has suggested that group II PAKs contain a conserved functional NES at residues 84-94, and selective alanine mutations in this motif interfere with the NES function (Yohendran

Baskaran, unpublished data). A similar set of mutations in PAK5 and PAK5∆268-

420 were tested (Figure 4.4A). The indirect immunofluorescent staining demonstrated that PAK4 with a mutated NES accumulated in the nucleus. However, although the PAK5 NES mutant was not nuclear enriched, the monomeric

PAK5∆268-420 (with NES mutant) behaved like the PAK4 mutant (Figure 4.4B).

This result suggests PAK5 self-association (oligomerization) largely prevents the nuclear import of the kinase.

90

A hPAK4 VRGSKGAKDGALTLLLDEFENMSVTRSNSL 100

hPAK5 VRGNKPCKETSINGLLEDFDNISVTRSNSL 100

B Anti-Flag Anti-Flag

PAK4 WT PAK4 mNES

PAK5 WT PAK5 mNES

PAK5∆268-420 PAK5∆268-420 mNES

Figure 4.4 The effect of mutation on putative nuclear export sequences of PAK4 and PAK5. A. Schematic showing the alignment of PAK4 and PAK5 putative NES sequences. Alanine mutations of PAK4 at the sites shown leads to nuclear accumulation of PAK4 (Yohendran Baskaran, unpublished data). B. The plasmids encoding wild type or NES mutated Flag-PAK4, Flag-PAK5 and Flag PAK5∆268-420 were transfected in COS-7 cells. Indirect immunofluorescence analysis was performed 16h after transfection. While PAK4mNES accumulates in the nucleus, PAK5mNES retains its cytoplasmic localization. However, PAK5∆268-420-mNES which is not able to form puncta is partially localized to the nucleus. 91 4.3 Analysis of BAD as a PAK5 target

4.3.1 PAK5 does not co-localize with BAD in the cells

Mitochondrial localization of PAK5 has been suggested as one mechanism to allow access to its substrate BAD [50]. I observed under normal culture conditions

(ie in 10% serum) PAK5 shows no enrichment on the mitochondrial compartment.

Since over-expressed BAD tends to accumulate at mitochondria [197], I was interested to see if PAK5 would also translocate under this condition. Both COS-7 and HEK-293 cells were transfected with low levels of HA-PAK5 and Flag-BAD for overnight, fixed and analyzed by indirect immune-fluorescent. As can be seen in Fig

4.5 some Flag-BAD accumulated at mitochondria; however there was no obvious change in PAK5 distribution. In normal cells, BAD is phosphorylated and sequestered by 14-3-3 protein and therefore the protein is predominantly found in the cytosol [85]. I hypothesized that this binding to 14-3-3 protein could prevent PAK5 binding to the BAD which would result in only a transient interaction (before its phosphorylation). Since phosphorylated Ser136 residue is critical for 14-3-3 binding

[91], I generated a construct encoding BAD(S136A) and then examined its distribution with respect to PAK5 (Fig 4.6A,B). The mutated BAD was much more efficiently localized to mitochondria but again PAK5 remained in puncta at which did not co-localize with Mitotracker.

92

COS-7 HEK-293

HA

- Anti

PAK5 PAK5

Flag

- Anti

BAD BAD

Merge Merge

Figure 4.5 Analysis of co-localization of transiently expressed PAK5 and BAD. 0.5μg of Flag-BAD and HA-PAK5 were co-expressed in COS-7 or HEK 293 cells for overnight. Cells were then fixed in 4% PFA and co-immunostained with anti-Flag and anti-HA antibodies to analyse co-localization of PAK5 and BAD. (Scale bar:10µm)

93 A mBAD MGTPKQPSLAPAHALGLRKSDPGIRSLGSDAGGRRWRPAAQSMFQIPEFE 50 hBAD ------MFQIPEFE ********

mBAD PSEQEDASATDRGLGPSLTEDQPGP-----YLAPGLLGSNIHQQGRAATN 95 hBAD PSEQEDSSSAERGLGPSPAGDGPSGSGKHHRQAPGLLWDASHQQEQPTSS 58 ******:*:::****** : * *. ***** . *** :.::. Serine 112 Serine 136 mBAD SHHGGAGAMETRSRHSSYPAGTEEDEGMEEELSPFRGRSRSAPPNLWAAQ 145 hBAD SHHGGAGAVEIRSRHSSYPAGTEDDEGMGEEPSPFRGRSRSAPPNLWAAQ 108 ********:* ************:**** ** ****************** Serine 155 mBAD RYGRELRRMSDEFEGSFKG-LPRPKSAGTATQMRQSAGWTRIIQSWWDRN 194 hBAD RYGRELRRMSDEFVDSFKKGLPRPKSAGTATQMRQSSSWTRVFQSWWDRN 158 ************* .*** ****************:.***::*******

mBAD LGKGGSTPSQ 204 hBAD LGRGSSAPSQ 168 **:*.*:*** B Anti-HA Anti-Flag

PAK5 BAD

Mitotracker Merge

Figure 4.6 Mitochondrial localization of mutated BAD does not induce PAK5 translocation to the mitochondria. A. Sequence alignment of BAD protein in human and mouse. Phosphorylation of the conserved serine 112 and 136 residues (shaded yellow) promotes 14-3- 3 binding which prevents mitochondrial targeting of Bad. The Ser155 lies within the BH3 domain of BAD and its phosphorylation negatively regulates binding to the mitochondrial BCL-2. BH3 domain marked gray. B. COS-7 were co- transfected with 0.5μg of HA-PAK5 and Flag-BAD- S136A constructs. After overnight expression, mitochondrial staining using Mitotracker (150nM) was performed. Cells were then fixed and immuno-stained with the indicated antibodies. (Scale bar: 10µm) 94 4.3.2 Camptothecin treatment does not affect PAK5 localization

A previous report suggested that PAK5 expression can promote

Camptothecin resistance in colorectal cancer cells [135]. I was therefore interested to see if drug treatment (which induces apoptosis) affects the localization of PAK5 under this stressed condition. COS-7 cells were treated with Camptothecin and Flag-

PAK5 distribution assessed by immunofluorescent staining. Although Camptothecin clearly induced apoptosis in COS-7 cells as suggested [198], such a signal (after

DNA damage) does not influence PAK5 distribution (Figure 4.7 A, B).

4.3.3 PAK5 promotes BAD Ser112 phosphorylation

Since in my hands there is little PAK5 associated with mitochondria, I was interested to see whether in fact PAK5 promotes BAD phosphorylation nonetheless.

First, I tried to find an agent that could induce apoptosis coupled to BAD de- phosphorylation. For example Camptothecin treatment induced BAD dephosphorylation in CHO cells [52]. Camptothecin is a topoisomerase I inhibitor and induces DNA damage in the cells (during S-phase) [199]. Treatment of HEK293 cells with 1μM of Camptothecin for 24h induces apoptosis in 60% of cells [200].

Using this system, I co-transfected HEK-293 cells with HA-BAD and either Flag-

PAK5 or empty vector for 16h then treated them with 10μM, 25μM and 50μM

Camptothecin for a further 24h. As can be seen in figure 4.8, this treatment failed to reduce BAD pS112 signal (compare lane1 with lanes 2,3,4 or lane 5 with lanes 6,7,8) under these conditions. Most importantly, it is clear that ectopic expression of PAK5 increases the basal level of BAD phosphorylation on Ser112 in vivo.

95

DMSO 10 μM Camptothecin A

Anti-Flag Mitotracker Merge+DNA

B DMSO

PAK5 WT

Camptothecin

M M μ

10 10 PAK5 WT

Camptothecin M M

μ PAK5 WT 25

Figure 4.7 Camptothecin treatment does not induce PAK5 re-localization to the mitochondria. COS-7 cells transiently expressing Flag PAK5 were treated with either DMSO or indicated concentration of Camptothecin for 24h. Next, cells were stained with mitotracker and then fixed and stained with anti- flag antibody to examine PAK5 localization. A. Increased nuclear fragmentation is consistent with apoptosis in COS-7 cells treated with 10μM of Camptothecin for 24h. B. Micrographs indicate Flag-PAK5 distribution is not significantly altered after treatment with Camptothecin. (Scale bar:10µm)

96 pXJ-Flag vector Flag PAK5

Camptothecin (µM) - 10 25 50 - 10 25 50 HA BAD + + + + + + + + 75 IB: anti- Flag

25 IB: anti- p-S112 BAD

25 IB: anti-BAD

1 2 3 4 5 6 7 8

Figure 4.8 PAK5 phosphorylates BAD on Ser-112 in vivo. HEK293 cells were co-transfected with HA-BAD and either empty vector or Flag-PAK5 overnight. Then cells were treated with either DMSO or various concentration of Camptothecine as indicated. After 24h, cells were harvested and lysates separated on SDS-PAGE to examine the pSer112 status of BAD. Ser-112 phosphorylation in the cells treated with DMSO was compared with camptothecin treated cells (lane 1 compared to lanes 2,3 and 4 or lane 5 compared to lane 6,7 and 8). Camptothecin treatment did not induce BAD de-phosphorylation on Ser112. PAK5 transfected cells showed an increased phosphorylation of Ser112 compared to the control.

97 4.4 Analysis of PAK5 interaction and localization with Raf-1

4.4.1 PAK5 can co-localize with Raf-1 and phosphorylate it at Ser338

As I described in the main introduction, Raf-1 activation is a complex mechanism involving a number of upstream kinases, small G-proteins and adaptors

(Figure 1.12). Ras.GTP promotes Raf-1 (auto)phosphorylation on several residues that are essential for full Raf-1 activity towards MEK1 [109]. Ser338 residue phosphorylation has been suggested to be critical for Raf-1 activation although the equivalent residue in B-Raf is constitutively phosphorylated [114, 201, 202]. PAKs have been reported as one of the potential Raf-1 ser338 kinase [51, 115, 116, 203-

205]. As can be seen in figure 4.9, I confirmed that Raf-1 is phosphorylated by over- expressed PAK5 in cells. Although PAK4 was also reported to phosphorylate Ser338

[51, 116], it has been suggested that PAK5 is much more effective in this regard because it can directly bind Raf-1[205]; this observation has not been confirmed by other laboratories. Co-expression of Raf-1 with PAK5 has been suggested to translocate Raf-1 to the mitochondria independent of PAK5 activity [205].

I have shown that the punctate distribution of PAK5 in HEK-293, HeLa and

COS-7 cells is not significantly co-localized with Mitotracker (Figure 4.3). One possibility, given that Raf-1 action at mitochondria [206], is that PAK5 requires interaction with Raf-1 in order to form a productive complex at the mitochondrial surface. Since the level of Raf-1, A-Raf and B-Raf vary in different cells; it might explain the difference in PAK5 localization. COS-7 and HeLa cells were transfected with either exogenous HA-Raf-1 or co-transfected with HA-Raf-1 and Flag-PAK5.

The cells were serum-starved for 16h in order to synchronize them in G1 and to reduce basal phosphorylation of Raf-1. Immuno-staining analysis revealed diffuse

98 expression of Raf-1 in the cytoplasm with some concentration in the perinuclear region that partially colocalized with mitochondria (Figure 4.10A,B). When Raf-1 co-expressed with PAK5, in some cells the two proteins were poorly co-localized

(Figure 4.10C,D), however in some other cells the both proteins were present in puncta concentrated in perinuclear region (Figure 4.10E). At higher resolution it was apparent that there was not convincing co-staining of the two on mitochondria

(Figure 4.10E,F). These results were similar when I used HeLa cells (data not shown). In summary my results suggest that the two proteins can interact in vivo but their interaction may not be stable.

99

Flag PAK5 - + + HA Raf-1 + + +

IB: anti-p-S338 Raf-1 75

IB: anti-Raf-1

Figure 4.9 PAK5 phosphorylates Raf-1 on Ser338 in vivo. COS-7 cells were co-transfected with HA-Raf-1 and either empty vector or 0.4μg and 0.8μg. of Flag-PAK5 expression plasmids. The following day, cells were serum starved for another 16h and then harvested and lysed in RIPA buffer. SDS-PAGE and western blotting was performed to analyze Raf-1 phosphorylation with the antibody specific to phosphorylated Ser 338.

100 A Anti-HA

Raf-1 Mitotracker Merge

B 120 Mitotracker 100 Raf-1 80

60

Intensity 40

20

0 1 5 9 13 17 21 25 29 33 37 41 45 49 Line length

Figure 4.10 Analysis of PAK5 and Raf-1 co-localization in COS-7 cells. COS-7 cells were transfected with either HA-Raf-1 or Flag-PAK5 in combination with HA-Raf-1. After overnight expression, cells were serum starved for 16h and then stained with Mitotracker (150nM) before fixation (4% PFA). Immuno-staining was carried out with mouse anti HA and rabbit anti Flag antibodies. A. Representative image of immunofluorescence staining of COS-7 cells transfected with HA tagged Raf-1. Scale bar:10µm. The boxed were magnified (3 folds) to better assess the mitochondrial localization of Raf-1 B. Line scan analysis of Raf-1 and Mitotracker colocalization was performed along the indicated white arrow.

101 C

Raf-1 Mitotracker PAK5

D

60 PAK5 50 Mitotrcaker Raf-1 Merge 40

30

20

10

0 1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33

Figure 4.10 Analysis of PAK5 and Raf-1 co-localization in COS-7 cells. C. Representative image of COS-7 cells transiently co-expressing HA-Raf-1 and Flag-PAK5 with no colocalization between the two proteins. The boxed area was enlarged (3 folds) in the lower panel. D. Line scan analysis of signals from three channels was performed using ImageJ along the white arrow.

102 E

Raf-1 PAK5 Mitotracker

F

140

120 Mitotracker 100 Merge Raf-1 80 PAK5 60

40

20

0 1 4 7 10 13 16 19 22 25 28 31

Figure 4.10 Analysis of PAK5 and Raf-1 co-localization in COS-7 cells. E. Representative image of COS-7 cells co-expressing PAK5 and Raf-1 with some of the signals in puncta that is not associated with the mitochondria. scale bar: 10µm F. colocalization of the signals from three channels was quantified using line scan analysis.

103 4.4.2 Investigating the interaction site of PAK5 with Raf-1

Since Raf-1 was reported to selectively binds PAK5 versus PAK4 [51], I was interested to map the interaction site(s) of Raf-1 with PAK5. Using a Flag tagged N- terminal deletion construct of PAK5 generated previously, I tested Raf-1 interaction by standard co-immunoprecipitation. Full length Flag-PAK5 is able to pull down

Raf-1 while Flag-PAK5∆268 has largely lost this ability (Figure 4.11A). Because removing 268 amino acids from PAK5 disrupts its oligomerization (Figure 3.7) there is a possibility that Raf-1 binding requires an oligomeric form of PAK5 or that the first 268 amino acids of PAK5 includes the Raf-1 interaction site. To evaluate these models, I switched to using glutathione-S-transferase (GST) fusion constructs of PAK5 since GST itself is prominently a dimer [207]. The results (Figure 4.11B) suggest that the residues 1-268 of PAK5 are required for binding to Raf-1 and PAK5 oligomerization may not be important for this interaction.

104

FL - Flag constructs

A

PAK5 PAK5∆268 - HA Raf-1 + + + IB: anti-HA 75 (co-IP)

Flag IP 75 IB: Anti Flag

75 Input: anti-HA

B 268 -

GST constructs

Vector

PAK5 1 PAK5 ∆360 PAK5 HA Raf-1 + + + 75 IB: anti-HA

75

GST pull down

50 IB: anti GST

37

75 Input: anti-HA

Figure 4.11 Raf-1 can bind to the N-terminal of PAK5. A. COS-7 cells were co-transfected with HA-Raf-1 and either indicated Flag tagged constructs of PAK5 or empty vector. The lysates were used for co- immunoprecipitation analysis using anti-Flag Sepharose (see methods) and then subjected to western blotting. HA-Raf-1 was detected using anti-HA coupled to HRP B. COS-7 cells were co-transfected with HA- Raf-1 and either mutated GST-PAK5 constructs or pXJ-GST vector. The GST fusion proteins were recovered using glutathione Sepharose and HA-Raf-1 binding was detected after Western blotting.

105 4.5 Summary

I have demonstrated that ectopic expressed Flag-PAK5 or GFP-PAK5 shows an unusual punctate distribution in COS-7, HeLa and HEK293 cells (Figure 4.2).

This is unlike protein aggregates that commonly seen when mis-folded proteins are expressed in mammalian cells. In no case, I was able to localize the full-length kinase to mitochondria using Mitotracker. The Flag-PAK5 retains a punctate distribution after co-expression with BAD (Figure 4.5 and 4.6) or the initiation of apoptosis by Camptothecin (Figure 4.7). I could however confirm that PAK5 contains cryptic mitochondrial localization signals (Figure 4.3) that might allow the protein to associate with mitochondria under specific conditions.

PAK5 is able to drive in vivo BAD phosphorylation at Ser112 (Figure 4.8) but it is not co-localized with BAD (when over-expressed) on mitochondria (Figures

4.5 and 4.6). This is not untypical since many kinase substrates interact with low affinity with their kinases. By contrast, I found that PAK5 could interact with Raf-1 which requires residues 1-268 and this leads to enhanced Raf-1 phosphorylation at

Ser338 (Figures 4.9 and 4.11). Co-immunostaining results suggest that PAK5 and

Raf-1 are poorly associated in COS-7 and HeLa cells (Figure 4.10).

4.6 Characterization of endogenous PAK5

My previous results suggested that even when expressed at low level (0.2 ug/ml (see Fig 3.5) PAK5 oligomerized and that this process is responsible for the elevated level of PAK5 activity when compared to the fully auto-inhibited monomeric PAK4. These PAK5 puncta however were not specifically associated with any larger-scale structure in the cell (mitochondria, actin stress fibers, endosomes, or ). Therefore, I was interested to compare where

106 endogenous PAK5 localizes in cultured cells and the nature of its oligomerization.

Previous work has noted that PAK5 can be associated with a number of compartment but most of these have been based on ectopic expression of tagged version of the kinase [31, 50, 52, 53].

In order to find a cell line with high levels of PAK5, the gene expression data set was probed (http://www-test.ebi.ac.uk/). Based on a data derived from 950 samples of 317 different cancer cell lines [208], I found that PAK5 mRNA is higher in NCI-H446 and SW962. The small lung cancer cell line NCI-H446 exhibited the highest PAK5 mRNA level expression; therefore this cell line was included in the analysis. A commercial PAK5 antibody (Abcam 37753) was used to detect PAK5 expression in NCI-H446 (ATCC) but as can be seen in figure 4.12A and B, the antibody was not sensitive enough to detect endogenous level of PAK5. Conversely, an antibody that detects the activation loop of PAK4, PAK5 and PAK6 (pS474, pS602 and pS560) was extremely sensitive compared to specific PAK5 antibody

(Figure 4.12A). Because group II PAKs are constitutively phosphorylated on activation loop [28], we can assume that the antibody can detect the relative level of these proteins. We also raised a specific antibody against PAK5 (residues 142-159) which was extensively characterized (Figure 4.12C), however it was not sensitive enough to detect endogenous level of PAK5 in NCI-H446 cell line (data not shown).

Therefore, the antibody against the activation loop of group II PAKs were used to compare PAK5 expression in NCI-H446 cell line and a few other cancer derived lines that some of them were reported to express PAK5 [3, 130, 131, 209]. As can be seen in figure 4.13A, a band corresponding to the expected size of phospho-PAK5 was only detected in NCI-H446 cells. Unfortunately an unrelated protein band of ~75 kDa protein is found in most cells. I also tested PAK5 expression in a number of cell

107 lines using another commercial PAK5 antibody (ab110069). This antibody detected a non-specific band with a close size to PAK5 in most of the cell lines (Fig 4.13B) but not endogenous PAK5. As I have shown, there is no sensitive (and specific) polyclonal or monoclonal commercial anti-PAK5 antibody; it is therefore difficult to compare results from different labs. The anti-activation loop antibody indicated that

PAK5 expression is several times lower than PAK4 expression in NCI-H446 cells.

This cell line was found to be poorly attached on plastic and glass surfaces and was difficult to dissociate during trypsinization; indicating the cell-cell junctions were selectively stable. These cells are not easily transfected using standard liposome preparations and therefore not suitable for siRNA or transient transfection analysis.

In order to analyze the effect of PAK5 expression in this line, I therefore used shRNA to derive NCI-H446 clonal lines in which the PAK5 level was reduced (Fig

4.14). Different clones of PAK5 stable knockdown line, exhibited no homogeneity in my experiments (such as migration assay and soft agar assay), therefore the results of these experiments are not included in this thesis.

108

A

H446 H446

- -

Flag PAK4 Flag PAK5 Flag NCI PAK4 Flag PAK5 Flag NCI PAK4 Flag PAK5 Flag

75kD 75kD Non-specific bands Anti-p-PAK4/5/6 Commercial Anti-Flag

(1/1000 v/v) anti-PAK5(2/1000 v/v) (1/2000v/v)

7

-

H446

-

COS Flag PAK5 Flag PAK6 Flag NCI B PAK4 Flag

Commercial 75kD anti-PAK5(1/100)

C GFP-PAK5 - + - + - + - + - + - + GFP-Vector + - + - + - + - + - + -

100kD

F1 F2 F3 F4

PAK4/5/6

-

p

-

purified serum purified

-

Anti Un Figure 4.12 Comparison of sensitivity of PAK5 antibodies. A. B Lysate of COS-7 cells expressing low levels of Flag-PAK4 or Flag-PAK5 and the lysate of NCI-H446 cells with endogenous expression of PAK4 and PAK5 were used to compare the sensitivity and specificity of two commercial anti-PAK5 antibodies using Western blot. Anti-phospho-PAK4(Ser474)/ PAK5(Ser602)/ PAK6(Ser560) (Cell signaling) or anti-PAK5 (Abcam,37753) antibodies were used at 1/1000 and 2/1000 v/v respectively. The anti-p-PAK4/5/6 antibody clearly detects Flag tagged PAK4 and PAK5 and endogenous PAK4 (65 kDa) and PAK5 (~85kD), although it picks up an unrelated band at 70 kDa. The anti-PAK5 (Abcam) detects overexpressed PAK5 with much less sensitivity versus anti-p- PAK4/5/6. It cannot detect endogenous PAK5 even in higher concentration (1/100 v/v) (B). C. Polyclonal antibody raised against the residues 142-159 of PAK5 was affinity purified in 4 fractions (see material and methods). Lysate of U2OS cells stably expressing GFP-PAK5 were used to test the sensitivity of the purified antibody versus the commercial anti-p-PAK4/5/6 antibody. 109

*

MB231 * MB231

7

116 * 116

-

7 * 7

-

7

H446 -

-

- 145

- 29 * 29

A -

-

Cos NCI U2OS HeLa Huh HCT HT MCF MDA DU LNCaP

kD p-PAK5

75 Anti-p-PAK4/5/6 p-PAK4

37 Anti-β actin

231*

-

*

MB

7

7

116 * 116

-

7* -

-

H446 H187 -

-

- - 29 * 29

B -

NCI NCI HCT HT MCF MDA HeLa Huh COS

< PAK5 75 < non-specific band

anti-PAK5 (ab110069)

Figure 4.13 Analysis of endogenous PAK5 expression in different types of cancer cell lines. A,B. Lung cancer (NCI-H446,NCI-H187), Osteosarcoma (U2OS), Cervical cancer (HeLa), Hepato cellular carcinoma (Huh-7), Colorectal cancer (HCT-116, HT-29), Breast cancer (MCF-7, MDA-MB231) and Prostate cancer (DU-145, LNCap) cell lines were used to analyze endogenous PAK5 expression. Cells were lysed with RIPPA buffer and 50μg of proteins were loaded into SDS-PAGE. The antibody that detects phosphorylated activation loop of PAK4, PAK5 and Pak6 and a commercial PAK5 antibody (ab110069) were used for Western blotting. β actin expression was used as a loading control. The asterisks show the cell lines that previously reported for endogenous PAK5 expression.

NCI-H446-Sh-PAK5- NCI-H446-Sh-Scramble- clones clones

kD p-Ser602 PAK5 75 75 p-Ser474 PAK4 Anti p-PAK4/ 5/ 6

Figure 4.14 PAK5 expression in PAK5 knockdown stable line NCI- H446. NCI-H446 clones with stable expression of sh-PAK5 (see methods) or sh-control were established under G418 selection (300μg/ml). Sh-PAK5 suppressed PAK5 protein levels but not that of PAK4 as assessed by antibody directed at the identical p-Ser474/pS-602 sequence.

110 4.6.1 GFP-PAK5 localizes in junctions of U2OS cells

Because of the lake of specific and sensitive PAK5 antibody, I was not able to analyze endogenous PAK5 localization; therefore I established a number of U2OS cell lines which express GFP-PAK5 under Puromycin selection. The PAK5 expression level could again be compared with endogenous level of PAK4 by western blotting (Figure 4.15A). Based on these results, both examined clones seem to express PAK5 3 fold of endogenous PAK4.

When cells were fixed and imaged, it was obvious that the protein was found in puncta and also distributed on cell junctions (Figure 4.15B). To find out whether

PAK5 might move to mitochondria in the absence of cell-cell junctions, the cells were grown in low density. But I was not able to co-localize PAK5 significantly with

Mitotracker (Figure 4.15C,D).

To investigate the junctional localization of PAK5 in a well characterized epithelial cell line, I transfected the kinase in MDCK cells which exhibit well organized and stratified junctions [210, 211]. Transient expression of GFP-PAK5 indicated junctional localization of the protein (Figure 4.16A). p120 catenin has been suggested as a binding partner and substrate for PAK5 and a constitutively activated

GFP PAK5 was found to phosphorylate p120 at Ser288 [54]. As can be seen in

Figures 4.16B and C, the GFP-PAK5 is partially co-localized with p120 in the junctions of MDCK and U2OS cells, indicating that the kinase might be involved with several junctional compartments. The cells tended to exhibit a single or doublet spot often above the nucleus (arrows). Further experiments showed colocalization of these spots with γ-tubulin, suggesting another new localization site for PAK5 in the centrosome (Fig 4.16D).

111

A B fixation Typical antibody PAK control A Figure . The 5 was 4 lines with distribution lysate . U2OS GFP (clone 4 GFP 15 directed 4 compared PAK were % - PAK5 (clone 14) from PFA 100 100 75 5 at subjected localization . of the two (Scale to GFP identical

GFP GFP-vector-1 endogenous - bar ) PAK to GFP-vector-4 - PAK : in western 10 5 GFP-PAK5-9 GFP p µm) 5 - 112 in S

474 GFP-PKA5-22 stable - GFP U GFP PAK PAK 2 blot OS IB: IB: anti IB: anti IB: anti /pS - - 5 . 4 Endogenous PAK4 GFP PAK5 (clone 22) PAK5 (clone 9) cell The 602 U stable expression 2 - - - - OS PAK5 β P GFP lines level sequence - - actin /5/6 stable lines and of using GFP after line . two B - . . C GFP-PAK5-Clone 14 GFP-Vector-Clone 1

Mitotracker Mitotracker

Merge Merge

D

GFP-PAK5 80 150 GFP-vector Mitotracker Mitotracker 60 100 40 50 20 0 0 1 4 7 10131619222528 1 4 7 10131619222528

Figure 4.15 PAK5 localization in GFP-PAK5 U2OS stable line. C. Cells were seeded at low density and treated with 150nM of Mitotracker for 20 min before fixation in 4% PFA. Immuno-staining was performed with rabbit anti-GFP antibody. Scale bar: 10µm.The boxed area was magnified (3 folds) to better assess the distribution of mitochondria versus GFP-PAK5. D. colocalization of GFP-PAK5 and Mitotrcaker was quantitatively analysed by line scan using ImageJ.

113 A

Anti-GFP Anti-p120 MDCK

GFP-PAK5 P120-catenin Merge

300 B GFP-PAK5 200 Endo-p120

100 Intensity 0 1 6 11 16 21 26 31 36 Distance (Pix) C

Anti-GFP Anti-p120

PAK5 (Clone 9) (Clone PAK5 -

GFP GFP PAK5 P120 catenin Merge

D

PAK5(Clone 22) PAK5(Clone -

GFP PAK5 γ-tubulin Merge UGFP

Figure 4.16 Partial co-localization of GFP PAK5 with p120 catenin. A. MDCK cells with transient expression of GFP-PAK5 were fixed in PFA for 20 minutes followed by methanol for 2 minutes and then co-immunostained with rabbit anti-GFP and mouse anti-p120 catenin antibodies (B). Line scan analysis of PAK5 and p120 along the indicated arrow to assess GFP-PAK5 and p120 catenin colocalization.C,D. U2OS cells stably expressing GFP PAK5 were fixed and co-immunostaining was performed using rabbit anti-GFP and either mouse anti-p120 catenin (C) or mouse anti-γtubulin antibodies (D). (Scale bar: 10µm) 114 4.7 Discussion

4.7.1 PAK5 phosphorylate BAD and Raf-1 in vivo

The results in this chapter showed a punctate distribution of ectopically expressed Flag- and GFP-PAK5 in multiple cell types (COS-7, HEK-293 and HeLa).

These puncta do not resemble the typical mitochondrial staining pattern nor do these overlap with Mitotracker (Figure 4.2). Such puncta PAK5 have been noted previously but was suggested to colocalize with mitochondria [52]; although other failed to reproduce this result in the same cell line [53]. Interestingly PAK5 (1-236) but not full-length PAK5 clearly localizes to mitochondria (Figure 4.3A) which is consistent with a reported mitochondrial localization sequence within residues 30-83 [50].

Established mutation that leads to inactivation of the NES in PAK4, did not lead to nuclear localization of PAK5 (Figure 4.4B), although PAK4 harboring the same mutations is nuclear (Fig 4.4B). One reason could be that PAK5 puncta greatly exceed 39nm that does not allow a protein to freely pass through the nuclear pore complex [212]. Based on these observations, I proposed that PAK5 self-association

(oligomerization) may prevent nuclear import of the protein.

In no condition tested in this study, such as serum starvation or Camptothecin induced apoptosis (Figure 4.7), full length PAK5 showed mitochondrial localization.

For example, Akt is usually cytoplasmic, however under cell stress the protein can accumulate on mitochondria where it can promote cell survival. My results also rule out the possibility of PAK5 recruitment to mitochondria via interaction with BAD.

PAK5 distribution was not co-localized with wild type BAD or predominantly mitochondrial BAD(S136A) (Figures 4.5 and 4.6). Nonetheless, PAK5 can clearly phosphorylate BAD Ser112 in vivo, as shown with overexpressed BAD (Fig 4.8).

115

Indeed BAD phosphorylation may be a common feature of both group I and II PAKs.

As it is demonstrated in figure 4.1, U2OS cells (lacking PAK5) treatment with

PF3758309, leads to dramatic reduction in BAD Ser112 phosphorylation; thus PAKs are key BAD kinases in this cell line. Work in our lab has established that PF3758309 inhibits both group I and II PAKs in vivo (Elsa Ng, unpublished). At 5μM concentration, which was used in this experiment, PF3758309 partially inhibits group

I PAKs and fully inhibits group II PAKs suggesting that the observed effect is mainly related to group II PAKs. Further, Ser111 appears to be the preferred site for PAK1 phosphorylation rather than Ser112 [100]. The former does not generate a 14-3-3 binding site, thus antagonizing phosphorylation of Ser112. By contrast the PKA and

ROCK inhibitor (H89) or the pan Raf inhibitor (TAK-632) has little effect; although both are reported as BAD Ser112 kinases [206, 213]. One reason could be the kinase selectivity of PF3758309 that has been previously reported [121]. Additional PAK inhibitors such as FRAX120 should be tested to validate this result. Unfortunately these compounds are not currently available from commercial sources.

Raf-1 has been established by a number of studies as an important kinase for

BAD phosphorylation and its anti-apoptotic effect [100, 206]. I have shown that

PAK5 can bind to Raf-1 through the N-terminal of PAK5 (Figure 4.11) and that this event leads to Raf-1 Ser338 phosphorylation (Figure 4.9). Immunofluorescence staining showed PAK5 expression can lead to Raf-1 recruitment to the puncta (Figure

4.10E) however; this is not always apparent (Figure 4.10C). It is possible that another protein partner stabilizes this interaction in the cell (my results in the next chapter, suggest that LZTS2 could play this role (Figure 5.22). The ability of PAK5 to phosphorylate Raf-1 on Ser338, has been noted for other PAK kinases [115, 187,

205], but if PAK5 binds to Raf-1 it may be more effective in this regard. PAK5 can

116 activate Raf-1 independent of growth factors such as EGF [51]. Raf-1 Ser338 is an important site of kinase regulation [114]. Although PAKs have been long known to phosphorylate Raf-1 on Ser338, this role is not widely accepted [109, 113]. Indeed

PAKs may not be necessarily activated by the stimuli that phosphorylate Ser338 [214] and may not have the ability to bind Raf-1 at the same region of the plasma membrane, where Raf-1 interacts with Ras [214]. Since Ser338 is reported as an auto- phosphorylated site upon dimerization of Raf-1 [113], the role of PAKs is questionable. Based on published data, PF3758309 does not show activity to Raf-1

[121], which makes it difficult to explain the effects of PF3758309 on Raf-1 phosphorylation and activation in U2OS cells (Figure 4.1). The effect of the PAK inhibitor on ERK phosphorylation was comparable to effect of pan-RAF inhibitor

TAK-632 (Figure 4.1 B, lanes 7 and 9). Although Sarofenib (Raf inhibitor) has significantly reduced phosphorylation of Ser338 in HEK-293 cells [113], TAK-632 treatment did not significantly affect Ser338 phosphorylation indicating a mechanism in this case, not involving auto-phosphorylation. Taken together, these results suggest that PAK4 could be an important Raf-1 kinase in U2OS cells; although PF3758309 has some inhibitory effect on CaMKII kinase [121]. CaMKII kinase is reported to act downstream of Ras (via calcium signaling) to mediate the Raf-1 phosphorylation in response to several stimuli [118] and the use of a specific CaMKII inhibitor would be needed to assess this in U2OS cells.

4.7.2 Localization of PAK5 in mammalian cell lines

PAK5 largely localizes in junctions in U2OS cells in which the kinase is stably expressed as a GFP protein (Figure 4.15, B). This localization has not been described previously. I have also confirmed the junctional localization of GFP-PAK5 by transiently expressing the kinase in commonly used epithelial MDCK cells (Figure

117

4.16A). GFP-PAK5 staining was similar to, but may not identical, with p120 catenin localization in these cells, as well as in the U2OS cells (Figure 4.16 A, C). The punctate distribution of PAK5 appeared to accumulate in a perinuclear region, however since this region of the cell contains a large volume of the cytoplasm in adherent cells, one must be cautious with the interpretation. Some protein concentration was seen in the centrosomal region which was colocalized with the centrosomal marker γ-tubulin (Figure 4.16D). The cell-cell junctional localization of

PAK5 is consistent with the presence of the Drosophila PAK4 orthologue Mbt [215] in junctions of photoreceptor cells. PAKs have been suggested to regulate a number of junctional components [26, 65, 216]. For example PAK4 localizes in cell-cell junctions of nascent lung epithelial cell 16HBE and promotes maturation of these junctions in calcium switch experiments [26]. PAK4 can phosphorylate β-catenin at

Ser675 [66] which is suggested to promote dissociation of β-catenin from E-cadherin and subsequent junctional disassembly [65]. A recent report found GFP-PAK6 concentrated in cell-cell junctions of DU145 prostate cancer and HT29 colorectal cancer cells [65]. Here, PAK6 was required for HGF-mediated dissociation of junctions by forming a complex with E-cadherin and IQGAP1 (IQ motif containing

GTPase activating protein). Moreover, they again suggested β-catenin as one junctional substrate for PAK6 that promotes junctional disassembly.

Partial co-localization of PAK5 with p120 catenin in MDCK and U2OS cells

(Figure 4.16A and C), might indicate p120 catenin phosphorylation by PAK5 [54].

The role of p120-catenin is to stabilize conventional cadherins at the adherent junctions [217]. p120-catenin binding regulates the level of E-cadherin by preventing its internalization and degradation [70, 217]. The increased level of p120 has been associated with increased cell migration, particularly of cancer cells [218, 219].

118

Likewise in neurons, p120 catenin is not only associated with N-cadherin cell-cell adhesion, but also plays distinct roles in synaptogenesis [220]. Since PAK5 is mainly expressed in the brain, PAK5 may influence p120's role in synaptic formation, which is consistent with the severe nervous system phenotype of PAK5/6 null mice [49].

In conclusion, I have shown that both BAD and Raf-1 are PAK5 substrates in vivo.

My results suggest that mitochondria is not the main resident compartment for PAK5,

Most importantly, PAK5 accumulates in cell-cell junctions and centrosomal region though it remains to be seen which regions of the kinase are important for the compartmentalization.

119

Chapter 5. Screening and characterization of proteins co-localizing with PAK5 in U2OS cells

5.1 Introduction

In the previous chapter, I described the junctional localization of PAK5, which has not been previously reported although the co-localization of PAK5 with cytoplasmic phosphorylated p120 catenin was shown in NIH-3T3 cells [54]. Previous attempts to identify partners of PAK5 in the brain used a biochemical purification approach to identify Pacsin1 and Synaptojanin as substrates for PAK5 [140]. Both

Pacsin1 and Synaptojanin regulate the endocytosis and recycling of synaptic vesicles.

Synaptojanin plays role in un-coating and recycling of synaptic vesicles via its phosphatase activity. Pacsin1 is a brain isoform of Pacsin family of proteins containing F-BAR and SH3 domain and can drive membrane deformation via its F-

BAR domain [221]. Pacsin1 binds to proline rich domain of Synaptojanin via its SH3 domain and thus the two form a complex in vivo [222]. PAK5 phosphorylates both

Pacsin1 and Synaptojanin: it is suggested that PAK5 phosphorylates Pacsin1 and releases its auto-inhibited conformation which promotes its interaction with

Synaptojanin.

Several public databases of compiled protein-protein interactions exist, and from this, the partners for PAK5 have been summarized (Table 5.1). In this chapter, I describe a new method that allows local modification of PAK5 neighbors in suitably generated cell lines. This method could successfully identify proximal proteins of

PAK5 primarily located at cell-cell junctions, and that one of these proteins named

LZTS2 is shown to be a direct interacting protein with PAK5.

120

Table 5.1 The list of PAK5 interacting partners extracted from data bases (1.BioGRID 2.IntAct-EBI 3.Mentha 4.SPIKE).

Data No Gene Name Protein name Type of experiment Role Ref. base

AP-2 complex 1 AP2S1 Affinity Capture-MS HIT 1, 2, 3 [223] subunit sigma

armadillo repeat- 2 ARMC7 Two-Hybrid HIT 1 [224] containing protein 7

Beta-arrestin-2 3 ARRB2 Affinity Capture-MS Bait 1, 2, 3 [225]

ATP synthase mitochondrial F1 4 ATPAF2 Two-hybrid HIT 1 [224] complex assembly factor 2

caspase recruitment 5 CARD9 domain-containing Two hybrid HIT 1 [224] protein 9

thyroid transcription 6 CCDC59 factor 1-associated Affinity Capture-MS HIT 1,2,3 [223] protein 26

cyclin G1 7 CCNG1 Two hybrid Bait 1 [224]

Affinity Capture- [33] Bait/ 8 CDC42 cell division cycle 42 western/ 1, 2, 3 [45] HIT Two- hybrid [226]

protein cereblon 9 CRBN Two-hybrid Bait 1 [224]

121

Table 5.1

Data No Gene Name Protein name Type of experiment Role Ref. base

LEM domain 10 EMD containing 5 Two-hybrid Bait 1 [224]

homolog of yeast 11 EXOSC1 exosomal core Two-hybrid Bait 1 [224] protein CSL4

family with 12 FAM212A sequence similarity Two-hybrid HIT 1 [224] 212, member A

four and a half LIM 1, 2, 13 FHL3 Two-hybrid Bait domains protein 3 3, 4 [227]

GRIP and coiled- coil domain- 14 GCC1 Two-hybrid HIT 1 [224] containing protein 1

HIG1 domain 2, 3, 15 HIGD1A family member 1A, Affinity Capture-MS HIT 4 [223] mitochondrial

leukocyte receptor 16 LENG1 cluster (LRC) Two-hybrid HIT 1 [224] member 1

leucine zipper, 1, 2, 17 LZTS2 putative tumor Two-hybrid HIT 3, 4 [227] suppressor 2

P21 activated kinase-4 18 PAK4 Affinity Capture-MS Bait 1 [228]

122

Table 5.1

Data No Gene Name Protein name Type of experiment Role Ref. base

PDZ and LIM 1, 2, 19 PDLIM7 Two-hybrid Bait domain 7 (enigma) 3, 4 [227]

substr 20 PKM Kinase assay 2, 3 Pyruvate kinase ate [229] PKM

Proteasome 1, 2, 21 PSMF1 inhibitor PI31 Two-hybrid Bait 3 [230] subunit

Ras-related C3 Affinity Capture- 1, 3, 22 RAC1 botulinum toxin HIT western 4 [33] substrate 1

replication protein 1, 2, 23 RPA3 Affinity Capture-MS HIT A 14 kDa subunit 3, 4 [223]

Protein S100-A7 1, 2, 24 S100A7 Affinity Capture-MS HIT 3, 4 [223]

Protein S100-A9 1, 2, 25 S100A9 Affinity Capture-MS HIT 3, 4 [223]

vinexin beta (SH3- 26 SORBS3 containing adaptor Two-hybrid Bait 1 [224] molecule-1)

gamma tubulin ring 1, 2, 27 TUBGCP4 complex protein Two-hybrid Bait 3, 4 [227]

ubiquitously- expressed, 28 UXT Two-hybrid Bait 1 [224] prefoldin-like chaperone

123

5.2. Classical methods for identification of protein-protein interactions

Several methods have been developed to identify protein-protein interactions

(PPIs). Yeast two hybrid assay (Y2H), affinity complex purification and chemical crosslinking method are traditional approaches for investigation of PPIs [231]. Y2H method was developed by Fields and Songs in 1989 [232]. To perform the classical

Y2H, a bait cDNA is fused in frame to the binding domain (BD) of a transcription factor and a library of prey cDNAs, which are fused to an activating domain (AD), are tested against this construct [233]. The interaction between bait and prey proteins, leads to connection of AD and BD of transcription factor and subsequent transcription of a reporter gene which could be detected. The advantage of this method (over the affinity purification) is that a weak interactions can be amplified by the reporter gene

[234]. However the technique suffers from a high rate of false positives [235]. The classical Y2H method cannot be used for membrane-bound proteins or for those requiring modification that occurs in mammalian cells but not yeast cells [233].

Variants of this method can overcome some of these limitations, for example split- ubiquitin membrane yeast two-hybrid system (MYTH) allows analysis of integral membrane proteins [236].

The affinity based methods rely on isolation of the protein of interest (bait) with a range of binding proteins (preys), via different methods such as a specific antibody. The use of epitope tag sequences fused to the target protein allows a generic system to immobilize the 'bait' [237]. However, the tag may interfere with protein localization or activation, and over-expression can lead to the identification of non- physiological targets. Affinity purification allows one to identify multi-protein complexes (that cannot be detected by Y2H), or probe interactions that require post translational modifications such as phosphorylation: for example 14-3-3 protein

124 family requires targets to be phosphorylated [238]. The disadvantage of methods that require capture and washing is the bias towards strong interactions, while transient interactions may not be detected. The binding affinity of two proteins in the cellular environment may not survive in the diluted buffer used to assay the protein-protein interaction [234]. Multiple forms of affinity purification method have been developed to improve the efficiency of this method [237], for example Tandem affinity purification (TAP) was developed to increase the specificity of this method [239].

TAP method relies on two step affinity purification which reduces non-specific binding but one disadvantage of this method is that transient interactions are most likely lost during the two step purifications [234].

Chemical cross-linking can be used to stabilize protein-protein interactions; by adding two or more reactive groups [240], the cross-linker can react randomly with amine or sulfhydryl groups. In combination with affinity tagging, one can preserve weak in vivo interactions [231]. This method can detect transient interactions and provide information of the site of protein-protein interaction [241] however, random cross-linking of proteins and the increased complexity of resultant mass spectra are problematic [237].

5.3 Proximal protein labelling by BirA* biotinylation in mammalian cells

In this study, a new affinity capturing based method termed BioID [242], has been applied to allow proximity labelling of a target protein of choice. In this protocol, one needs to generate a fusion protein that includes the E.coli BirA*

(R118G) biotin ligase. This modified version of BirA* contains a mutation that make it more promiscuous [243] compared to the wild type enzyme that only modifies a single target (in E.coli). When biotin levels are increased, the BirA* in mammalian

125 cells generates reactive bioAMP in the location of the fusion protein which can modify lysine residues on interacting and proximal proteins; thus providing information about the average direct and indirect interacting proteins. The unique attribute of this method is that biotinylation of local proteins take place in the physiological cellular context, and the biotinylation of target proteins remain intact during cell lysis. The ability of the biotin-streptavidin interact to survive SDS treatment [244], allows most contaminant proteins to be removed. The sensitivity of the BioID assay has not yet been fully assessed [237]. For most proteins in which

GFP fusions have been tested, the addition of BirA (30 kDa) should not affect localization.

In order to increase the selectivity of the method, we used SILAC (stable isotope labelling with amino acids in cell culture) strategy which allows for protein quantification by mass spectrometry. This greatly reduces the rate of false positive and increases the confidence of positive data [245, 246]. In this method, stable isotope labelled amino acids (13C, 2H, 15N) are introduced to growth media (heavy media), which are taken up by the cells to synthesis new proteins. When incorporated in the peptides, these amino acids could provide the mass difference compared to non- labelled proteins. Using this strategy, proteins can be purified from cells which are labelled (H) and then combined with 'non-specific' samples with normal isotopic (L) composition (typically control). A fractionation protocol is then performed after lysates are combined, in order to avoid bias, and single mass spectrometry used to quantify the SILAC peptide pairs which can be expressed as a ratio of H/L.

SILAC have been used for a variety of affinity applications [247, 248]. It can be used as adjunct to protein-protein interactions assays or for the analysis of proteins binding to specific DNA or RNA sequences (used as 'bait' [249-251]). It is also useful

126 for identification of interacting proteins of small molecules, such as inhibitors, for drug discovery [251, 252].

For example, SILAC can be used to monitor quantitative differences or changes in protein compositions or interactions under two or three different conditions [253]. One of the most important applications of SILAC is for discovery of diseases biomarkers. For example the secretome of pancreatic cancer cells has been compared with those non-neoplastic pancreatic ductal cells [254]. In addition, SILAC has been used to detect significant changes in the nuclear proteome during apoptosis

[255]. SILAC is particularly powerful as a means of studying intracellular signaling pathways, as assessed by changes in post translation modifications of proteins such as phosphorylation [256]. For example the change in phosphorylation upon treatment with EGF can provide a global view of the EGFR pathway [257]. Specific protein modification by palmitoylation [258] or ubiquitination [259] can also be improved by using SILAC. Labelling of 13C2H3-methionine has been used to assess methylation sites on protein and DNA [260]. This strategy has been used for study of histone methylation that is critical in the process of gene regulation[260].

5.4 Establishing GFP-BirA*-PAK5 line for BioID

I established Puromycin-resistant U2OS cell line expressing either GFP-BirA*-

PAK5 or the parental GFP-BirA*vector (gift of Dr. Dong Jing Ming) and chose two clones of each line based on appropriate low expression level of the fusion protein

(Figure 5.1). In order to carry out high sensitivity mass spectrometry (MS) of the

PAK5 proximal proteins, I adopted the strategy of stable isotope labelling of cells - more widely termed SILAC [245]. The GFP-BirA*-PAK5 lines were isotopically labelled by passaging them for 2 weeks in 'heavy' media containing C13/N15 (H)

127 labelled lysine and arginine, while the control lines were adapted to the same identical media with 'light' (L) lysine and arginine (see schematic Figure 5.2). Lys is an essential and Arg a semi-essential amino acids for mammalian cells. The application of isotope analog of essential amino acids in SILAC ensuring that cells only incorporate the isotope analog of these amino acids. Labelled Arg and Lys are preferred amino acids applied for SILAC because most of time trypsin is used as proteolytic enzyme for mass spectrometry. Trypsin specifically cleaves proteins on C- terminal side of Arg or Lys residues. Following trypsin digestion of protein each peptide typically contains either Lys or Arg at its C-terminus thus ensuring that the

'heavy' peptides can be paired with their 'light' partner as illustrated in Figure 5.3. The presence of heavy Lysine (13C6H1415N2O2) or Arginine (13C6H1415N4O2) results to

8 or 10 Dalton increasing in the mass of heavy peptides compared to light peptides which could be distinguished by mass spectrometry. For those proteins that are not enriched after streptavidin Sepharose pull down (Figure 5.3), the SILAC ratio for corresponding peptides will have a ratio of H/L=1. Higher ratios show an enrichment indicating these proteins are 'PAK5- proximal'. After protein fragmentation by in-gel trypsin treatment of the acrylamide gel slices, the recovered peptide mixture is separated by HPLC and peptides are introduced in-line to the mass spectrometer. The charge ratio (m/z) of peptides and fragmentation pattern (MS/MS) are used to determine a probability of a peptide sequence from the relevant database [261].

Multiple peptides assigned to the same protein have H/L ratios which indicate the origin of the same sequence.

128

R118G A GFP-BirA* GFP BirA * R118G PAK5 GFP-BirA*-PAK5 (FL) GFP BirA *

GFP-BirA*vector GFP-BirA*vector 17

B 16 C

2 6 -

- Clone (16) Clone (17)

- - line

GFPGFP-BirA*vector-BirA*PAK5line -17-6

line line -

- GFP-BirA*

- -

constructs

PAK5 PAK5 vector vector kD

150

IB: anti-GFP

50

GFP-BirA*PAK5 GFP-BirA*PAK5 (clone 2) (clone 6)

Figure 5.1 Analysis of U2OS stable lines expressing GFP-BirA*- PAK5 (FL) or GFP-BirA* controls. A. PAK5(FL) were fused to the BioID vector (kindly provided by Dr. Dong Jing Ming) to generate the GFP- BirA*-PAK5 construct. These were used to generate stable lines in U2OS cells under Puromycin selection (0.8 ug/ml). B,C. The two clones of GFP-BirA*-PAK5 and two control GFP-BirA*-lines (with roughly similar expression between PAK5 and control lines) were analyzed by western blot (B) and the correct localization of the proteins assessed by immunofluorescence (C).

129 Weak GFP-BirA* PAK5

GFP-BirA*

GFP-BirA*PAK5 GFP-BirA* Biotin Biotin Adding 1x LDS sample buffer Arg/Lys 13C,15N-media(H) Arg/Lys 12C,14N-media(L) to the beads and heating GFP-BirA*PAK5 Running the samples Extraction with in 1mm 10% SDS-PAGE GFP-BirA*vector 0.5% DOC 0.5% Tx100 Spin out nuclei

Spin down the beads Washing with Trypsin digestion buffers 1 and 2

Mixing H and L supernatants+0.2%SDS Sonication and spin Mass spectrometry

Lysate+Neutravidin H Sepharose Rotate overnight L H

L Intensity

m/z Streptavidin sepharose Non-Specific protein Specific protein Transient interacting protein Biotin

Figure 5.2 Schematic of Bio-ID work-flow. GFP-BirA* PAK5 stable lines are grown and adapted in heavy media (H) containing heavy isotopes of Arg and Lys (13C,15N). Control lines are grown in the light media (L) containing the light (12C,14N) isotopes of Arg and Lys. After two weeks adaption of the U2OS stable lines (brown color for heavy isotope, and blue color for light isotope) cells then are incubated with 100μM biotin (5h) to allow BirA* activity in vivo. GFP-BirA*-PAK5 and GFP-BirA* are assumed to label proximal proteins (indicated with red spots). The cells are lysed in 0.5% deoxycholate and 0.5% Triton x-100 and the nuclei removed by centrifugation at low speed. Supernatants of PAK5 and control lines are collected and made to 0.2% SDS to dissociate complexes and then mixed. The mixed lysate are then incubated with Neutravidin beads (Thermo Scientific) for overnight to capture the biotinylated proteins. The beads then are washed in 1% SDS to completely remove contaminants and 1x LDS sample buffer (Novex) added at 95°C to release proteins. These are run in 1mm 10% SDS-PAGE and the gel stained with colloidal CCB (Novax). The region >28 kDa is processed into spots for standard trypsin digestion and peptide extraction. The on-line LC-MS/MS analysis is set to find (H) and (L) pairs of peptides whose ratio was determined from the relative intensity value. 130 Arg: 13C6H1415N4O2 Arg: 12C6H1414N4O2 Lys: 13C6H1415N2O2 Lys: 12C6H1414N2O2

Heavy media Light media

Heavy proteins Light proteins

XXXXXKXXXXXKXXKXXXXRXX… XXXXXKXXKXXXXRXX…

Trypsin digestion

XXXXXK XXXXXK Heavy peptides XXXXXK Light peptides XXK XXK XXXXR XXXXR XX XX

L H

+8-10 D Intensity

m/z

Figure 5.3 Labelling proteins using SILAC for quantitative mass spectrometry. Cells growing in heavy media containing Lys (13C, 15N) and Arg (13C, 15N) results to incorporation of heavy amino acids in all proteins. Protein digestion with trypsin yields peptide fragments which usually contain C-terminal of Lys or Arg (except mis-cut peptides or the C- terminal fragment of the protein). The (H) and (L) peptides migrate identically in the LC fractionation and are processed together during the MS/MS analysis. The graph shows an example of a non-specific peptide.

131 5.5 Identification and analysis of PAK5 proximal proteins

In total, 390 proteins were identified by MS/MS in the Bio-ID. Figure 5.4 displays the scatter plot of log2 value of SILAC ratio versus percentage of sequence coverage for these proteins. The H/L ratios of data were normalized by median and proteins with SILAC ratio less than 2 and one peptide only removed. The resultant set of proteins is given in Table S1 of the supplementary section.

Proteins with SILAC ratio ≥3 and the peptide number ≥4 are listed in table 5.2 as high-confidence 'hits'. Figure 5.5 shows a diagram of classification of these proteins. The BioID method identified many junctional proteins as PAK5 proximal partners consistent with the junctional localization of PAK5 in U2OS stable lines

(Figure 4.15B). Of these, p120 catenin and Cdc42 are previously known as PAK5 interacting partners, providing us with some confidence with the analysis. In this

SILAC dataset, LZTS2 is suggested to be proximal to PAK5 since it has one of the highest H/L ratios, and comparable to nectin-2 and afadin (Table 5.2). There is also extensive overlap with data generated using a PAK4 cell line (Felicia Tay, personal communication). LZTS2 is a member of FEZ family of proteins that are poorly characterized: one study has suggested a role for LZTS2 in cytokinesis by binding to the microtubule severing protein katanin [154]. Here, I find that LZTS2 is predominantly at cell-cell junctions, as will be discussed with more details later

(section 5.8). The centrosomal protein CEP85 is a functionally uncharacterized protein that is part of this protein complex known as the microtubule organizing

Centre (MTOC). Interestingly, the presence of PAK4-unique peptides in my datasets suggests PAK5 and PAK4 localization overlap. Afadin and the trans-membrane protein nectin-2 are part of several complexes found next to adherens junctions as reviewed [262, 263]. Nectin is a trans-membrane protein whose C-terminus recruits

132 afadin via its PDZ domain [264]. Afadin can bind to F-actin but is also associated with other components of the cadherin-catenin complex through interaction with ponsin and vinculin proteins [262]. It is suggested that nectin-afadin complex initiates formation of adherens junctions by recruiting several proteins including cadherins to this compartment [264, 265]. The presence of direct or indirect afadin interacting partners such as p120 catenin [266], ZO-1 [267] and ERBB2IP (Erbin) [268] with somewhat lower SILAC ratios in the PAK5 BioID data strongly indicates the kinase is present in these junctions. The transmembrane protein Neph1 (KIRREL-1) is a member of the immuno-globulin superfami1y and is also known to interact with ZO-

1. Neph-1 knockout mice dye in 3-8 weeks of age because of a severe nephrotic syndrome [269], suggesting defects in the development and function of the glomerular filtration barrier. The C-terminal cytoplasmic domain of Neph-1 interacts with the

PDZ domain of the tight junction protein ZO-1 in a specialized epithelial cell-cell contact called the “slit diaphragm” [270]. Interestingly, Neph1 is extensively expressed in the central nervous system and with Neph2 plays important roles in synaptogenesis [271], again a very specialized cell-cell contact.

It was particularly interesting to find Kif7 as a unique motor protein associated with PAK5. Like most kinesins, it binds to microtubules via an N-terminal globular motor domain [272], and this isoform is implicated in cilia formation and thereby affects Hh (Hedgehog) signaling. Kif7 has been found to act as both positive and negative regulator of Hh signaling through its effect on the Gli transcription factor

[273]. Kif7 knockdown in retinal epithelial cells also indicates that it is required for

Golgi structure and centrosomal integrity [274]. The Golgi apparatus is usually positioned by the MTOC (microtubule-organizing centre) [275]. In summary, the list of proteins which were identified by BioID is consistent with a role of PAK5 at

133 specialized cell-cell junctions and perhaps at the centrosome. I did not find evidence for PAK5 being associated with proteins involved with the acto-myosin system, not proteins such as formins which are important for F-actin nucleation [276].

Table 5.2 High-confidence proteins identified by BioID as PAK5 proximal proteins in stable cell lines expressing BirA* PAK5

Seq. Ratio Pep. No Gene Name Protein name MW (kD) Cov. H/L No [%]

1 PAK5 PAK5 81 137.2 35 57

2 LZTS2 LZTS2 73 29.7 11 23

Centrosomal protein of 85 3 CEP85 86 19.4 9 17 kDa

4 PVRL2 Nectin-2 51 16.6 6 15

5 PAK4 PAK4 64 14.1 5 8

6 KIF7 Kinesin-like protein KIF7 151 13.8 6 5

7 MLLT4 Afadin 208 13.1 52 38

8 Tubuline alpha 3C chain 50 11.4 12 42 TUBA3C

9 FAM83H Protein FAM83H 127 11.3 13 19

10 MAGED1 NRAGE 92 10.9 7 11

Trinucleotide repeat 11 TNRC6B 194 9.6 11 7 containing gene 6B protein

12 PHLDB2 LL5-beta protein 142 9.2 5 6

13 TBKBP1 TBK1 binding protein 68 8.8 4 7

14 KIRREL Neph 1 85 7.1 4 8

Protein transport protein 15 SEC16A 252 6.5 8 4 Sec16A

16 DLG5 Disks large homolog 5 214 6.3 37 21

17 SIPA1L1 SPAR 200 6.3 7 4

18 CTNND1 P120 catenin 108 5.9 11 15

19 SLC25A13 Citrin 74 5.1 4 6

134

Table 5.2

Ratio Pep. Seq. No Gene Name Protein name MW (kD) H/L No [%]

20 RPS15A 40s ribosomal protein S15a 15 5.1 4 31

21 TP53BP2 ASPP2 126 4.1 4 3

22 ERBB2IP Erbin 159 4.1 25 23

23 STAU2 Staufen protein homolog 2 63 3.8 6 11

24 CAD CAD protein 243 3.3 14 8

25 EIF4ENIF1 eIF4E 108 3.3 4 5

ATP-dependent RNA 26 DDX3X 73 3.2 20 40 helicase DDX3X

Tight junctional protein 27 TJP1 197 3.2 37 25 ZO-1

28 CDC42 Cdc42 21 2.0 6 34

135

80

70

60

50

40

30

20

Sequence coverage (%) coverage Sequence 10

0 -8 -7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 8 Log2 of ratio GFP-BirA*PAK5/GFP-BirA* control

Figure 5.4 The scatter plot analysis of the SILAC ratio versus coverage derived from BioID analysis. The Log2 value of the SILAC ratio is here plotted against sequence coverage (%) for each protein identified in the Bio-ID dataset (393 proteins). A SILAC ratio >3 was used as a cutoff for significance in which 4 peptide were required to provide confidence in the median SILAC H/L value (as indicated with the black lines). The yellow filled spots show some of the important proximal proteins to PAK5: 1. LZTS2, 2. CEP85, 3.Nectin-2, 4.PAK4, 5.KIF7, 6.Afadin, 7.p120

136 Microtubule associated proteins

LL5β TUBA3C CEP85 LZTS2 KIF7 Junctional proteins

Afadin p120 catenin Nectin-2 RNA associated proteins CDC42 Erbin FAM83H TNRC6B DDX3X PAK4 ZO-1 Neph 1 DLG5 PAK5 RPS15A EIF4ENIF1 ASPP2 LZTS2 STAU2 SPAR

TBKBP1 MAGED1 Citrin ASPP2 CAD SEC16A

Others Mitochondrial proteins

Figure 5.5 The identification of PAK5 proximal proteins by SILAC BioID. The mass spectrometry data that were obtained from SILAC BioID experiment using U2OS cells expressing GFP-BirA*PAK5 were analysed and proteins with ≥3 H/L SILAC ratio and ≥4 peptides were selected. These proteins were classified manually based on their localization. Physical interactions between proteins are shown as annotated by STRING.

137 5.6 Identification and analysis of PAK5 interacting partners using GFP-pull down

In order to compare sensitivity and efficiency of BioID with traditional tag affinity pull down, the GFP-pull down (GFP-PD) were carried out. Because the cell lines used for BioID expressed GFP (GFP-BirA*-PAK5 or GFP-BirA*) I was able to use the same lines to perform standard GFP pull-down, though using recombinant

GFP-Trap beads which avoid IgG contamination [277]. The work flow of GFP pull down has been illustrated in figure 5.6. After adapting GFP-BirA*-PAK5 cells in heavy (H) media and the control lines in light (L) media, the lysates from GFP-BirA*-

PAK5 or GFP-BirA*control lines were separately incubated with GFP-Trap

Sepharose beads. The proteins were released by heating in SDS and then mixed for

SDS-PAGE, in-gel digestion and subsequent MS/MS analysis. The scatter plot of sequence coverage versus the log2 value of the SILAC ratio for the whole dataset of

GFP-PD is illustrated in figure 5.7. All those proteins assigned by 2 or more, high confidence peptides and SILAC ratio more than 2 are shown in table S2 of

Supplementary data. As for the BioID data, a working list of high confidence proteins were selected based on SILAC ratio (ratio H/L≥3) and peptide number (containing ≥4 informative peptides) (Table 5.3). Figure 5.8 illustrates the known interactions of these proteins based on the STRING 9.1 database [278].

138

GFP-BirA* PAK5 Weak GFP-BirA* PAK5

GFP-BirA*PAK5 Extraction with Collecting the Arg/Lys 13C,15N-media(H) Supernatant 0.5% Tx100 Mixing +Anti-GFP-Trap Beads and Washing Spin at 13000rpm, Eluting in 1x LDS H and L 15 mins beads Rolling 3h sample buffer GFP-BirA*

Arg/Lys 12C,14N-media(L) Concentrating the elutes in speed vacuum GFP-BirA* Running the samples in 1mm 10% SDS-PAGE GFP-BirA*

H GFP-BirA*PAK5 L GFP-BirA*vector

Trypsin digestion Intensity m/z Mass spectrometry

GFP-Trap beads Specific protein Non-Specific protein Transient interacting protein

Figure 5.6 Schematic of the SILAC GFP pull down experimental work-flow. U2OS cells stably expressing GFP-BirA* PAK5 or GFP-BirA* were cultivated in heavy (H) and light (L) medium respectively (see methods). After adaption, the cells were harvested in a lysis buffer containing 0.5% triton X-100. The solubilized proteins were clarified (Eppendorf centrifuge at 13000 rpm, 15min). The supernatant fractions of the H and L were separately added to GFP-Trap beads (recombinant anti-GFP) and incubated for 3h at 4°C. The Sepharose beads were collected and washed with the same lysis buffer (3 x 10 min) and bound proteins eluted in 1X LDS sample buffer at 95°C. The eluates were mixed and concentrated. The samples were subjected to 15% SDS-PAGE and processed for tryptic digest and mass spectrometry.

139 90 80 70 60 50 40 30 20 10 Sequence Coverage (%) Coverage Sequence 0 -7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 Log2 of Ratio GFP-PAK5/GFP-Control

Figure 5.7 The scatter plot of the complete set of proteins identified in the GFP-PAK5 pull down. The Log2 SILAC H/L ratio was plotted against sequence coverage (%) for each of the 377 proteins identified by SILAC GFP pull down. A 3 SILAC ratio cut off and 4 peptide number cut off were used as indicated with the red lines. These proteins are shown in table 5.3.

140 Table 5.3 Potential PAK5 interacting partners identified by GFP-PAK5 pull down

Seq. Ratio Pep. No Gene Name Protein name MW (kD) Cov. H/L No [%]

1 PAK5 PAK5 81 66.3 37 57

Prolyl 4-hydroxylase subunit 2 P4HA2 61 30.2 16 39 alpha-2

Prolyl 4-hydroxylase subunit 3 P4HA1 61 15.6 16 38 alpha-1

BAG family molecular 4 BAG2 24 11.8 12 60 chaperone regulator 2

5 CAD CAD protein 243 11.7 68 41

6 SEC13 Protein SEC13 homolog 41 11.6 5 17

Protein transport protein 7 SEC16A 252 11.3 22 12 Sec16A

8 P4HB Protein disulfide-isomerase 57 11.2 18 38

9 FAM83H Protein FAM83H 127 10.3 19 23

10 CDC42 Cdc42 21 10.2 4 23

ATPase family AAA 11 ATAD3A domain-containing protein 66 9.2 7 12 3A

Apoptosis-inducing factor 1, 12 AIFM1 67 9.1 10 19 mitochondrial

KN motif and ankyrin repeat 13 KANK2 92 8.5 4 5 domain-containing protein 2

T-complex protein 1 subunit 14 TCP1 60 7.5 21 44 alpha

Low-density lipoprotein 15 LRP1 receptor-related protein 1 505 7.5 16 4 85 kDa subunit

T-complex protein 1 subunit 16 CCT7 59 7.5 10 23 eta

T-complex protein 1 subunit 17 CCT3 61 7.3 16 32 gamma

141

Table 5.3

Seq. Cov. MW (kD) Ratio Pep. No Gene Name Protein name [%] H/L No

T-complex protein 1 subunit 18 CCT4 58 7.2 17 40 delta

T-complex protein 1 subunit 19 CCT5 60 7.0 16 39 epsilon

20 PKP2 Plakophilin-2 97 7.0 6 8

21 CCT6A T-complex protein 1 subunit 58 7.0 12 30 zeta

22 CCT2 T-complex protein 1 subunit 57 6.9 11 29 beta

23 YWHAH 14-3-3 protein eta 28 6.7 6 26

24 YWHAG 14-3-3 protein gamma 28 6.0 10 37

25 CCT8 T-complex protein 1 subunit 60 6.0 19 39 theta

26 UBR5 E3 ubiquitin-protein ligase 309 5.3 13 7 UBR5

27 PGAM5 Serine/threonine-protein 32 5.3 12 40 phosphatase PGAM5, mitochondrial

28 POLR2B DNA-directed RNA 134 5.0 5 4 polymerase II subunit RPB2

29 EMD Emerin 29 5.0 10 46

30 YWHAE 14-3-3 protein epsilon 29 4.9 12 48

31 FLNC Filamin-C 291 4.9 53 27

32 MAGED1 NRAGE 92 4.7 5 7

33 LEPRE1 Prolyl 3-hydroxylase 1 91 4.6 5 8

142

Table 5.3

Seq. Cov. No Gene Name Protein name MW (kD) Ratio Pep. H/L [%] No

34 USP9X; Probable ubiquitin carboxyl- 292 4.6 8 4 terminal hydrolase FAF-X USP9Y

35 LOXL4 Lysyl oxidase homolog 4 84 4.4 5 8

36 RUVBL2 RuvB-like 2 51 4.3 5 12

37 HSPA8 Heat shock cognate 71 kDa 71 4.2 34 60 protein

37 FLNB Filamin-B 282 3.6 5 2

38 PLOD1 Procollagen-lysine,2- 88 3.5 5 8 oxoglutarate 5-dioxygenase 1

39 CHCHD3 Coiled-coil-helix-coiled-coil- 27 3.5 5 24 helix domain-containing protein 3, mitochondrial

40 HSPA1A Heat shock 70 kDa protein 70 3.4 21 37 1A/1B

41 IMMT Mitochondrial inner 84 3.4 11 17 membrane protein

42 MYO1C Unconventional myosin-I c 120 3.2 12 14

43 PHB2 Prohibitin-2 33 3.2 9 35

44 HSPA6 Heat shock 70 kDa protein 71 3.1 11 17 6

45 YWHAB 14-3-3 protein beta/alpha 28 3.1 9 46

46 CYC1 Cytochrome c1, heme 35 3.1 5 15 protein, mitochondrial

47 HNRNPM Heterogeneous nuclear 78 3.0 8 15 ribonucleoprotein M

143

14-3-3 proteins family Chaperons

Figure 5.8 Potential PAK5 binding proteins assessed by GFP-PAK5 pull-down with SILAC normalization. The STRING annotation diagram shows the various known interactions between all the proteins isolated using GFP-Trap beads, and account for background contamination by selecting only those proteins with ratio H/L ≥3 and peptide number ≥4. The cluster of protein chaperones (CTT family) are prominent. PAK5 is annotated as PAK7, and is only known to associate with Cdc42 in this set.

144 5.7 Analysis of the results of BioID and GFP-PD

In contrast to the BioID sub-list, the GFP-PD data did not yield junctional proteins associated with PAK5; instead we observed many chaperone proteins and scaffolding proteins of the 14-3-3 family. Typically, the abundant 14-3-3 proteins bind phosphorylated Ser and Thr residues based on conserved flanking sequence. The interaction with group II PAKs has been previously described [279]. Figure 5.9 shows a comparison between PAK5 'interactomes' extracted from public databases with the high-confidence hits that I found using either BioID or GFP pull-down. The single overlapping protein among the three data sets is Cdc42, whose binding site in group II

PAKs is well characterized [8, 28]. The overlap between BioID list and those proteins supposedly interacting with PAK5 (GFP-PD list) is limited. Further experiments are needed to test which of these proteins might bind PAK5 directly.

FAM83H is among the five proteins which were identified through these two different methods, BioID and GFP-PD. The FAM83 (family with sequence similarity

83) is a well conserved set of proteins with FAM83H implicated genetically in amelogenesis [280] through a mechanism that is likely to involve cargo exocytosis

[281]. One microarray analysis has shown over-expression of this protein in variety of cancers [282]. Other members of FAM83 family are also poorly understood but these proteins are likely involved in regulating MAPK pathway [283]. Another interesting candidate is MAGED1, which I will refer to as NRAGE (neurotrophin receptor- interacting melanoma antigen). This is a multi-functional protein originally identified as one of the many melanoma antigens [284-287]. Overexpression of NRAGE in

U2OS disrupts E-cadherin/ β-catenin complex [285] which suggests the proteins is a modulator of cell-cell junctions. NRAGE is also thought to bind p75NTR (p75 neurotrophin receptor) and promote apoptosis in sympathetic neuron precursors cells

145

[288]. How NRAGE potentially promotes lung and kidney cancers [289, 290] is unclear. The fact that NRAGE interacts with ankyrin-G, a component of cell adhesion complex, points to a junctional role [291]. One model for NRAGE function is that during epithelial-mesenchymal transition (EMT), the loss of ankyrin-G leads to

NRAGE movement to the nucleus where it binds to the oncogenic transcriptional suppressor protein TBX2 and results in repression of tumor suppressor gene p14ARF

[291]. Both PAK5 and NRAGE are highly expressed in the nervous system [33, 292] and the both proteins can regulate mitochondrial cell death [52, 293].

Finally both Sec16A and CAD proteins are potential PAK5 partners. Sec16A is involved in cargo transport from ER to Golgi by organizing the ER exit site (ERES)

[294]. Since we have no evidence for PAK5 localization at the ER, this connection is unexpected. While depletion of Sec16A in HeLa cells attenuates cell proliferation and its over-expression increases cell numbers [295], the mechanism is obscure. CAD is a multi-functional protein involved in de novo synthesis of pyrimidine nucleotides by three enzymatic activity; carbamoyl phosphate synthesis, aspartate transcarbamoylase and dihydroorotase [296]. Induction of cell proliferation, leads to translocation of a fraction of CAD into the nucleus, where its activity is regulated by MAPK, ensuring synthesis of proper amount of pyrimidine which is required in cell proliferation [297].

CAD has also been contributed with cancer progression and its expression is increased in several cancers such as breast cancer [298] and hepatocellular carcinoma

[299]. A recent study has suggested CAD as a biomarker of prostate cancer. CAD regulates transcriptional activity of androgen receptor (AR) in prostate cancer through interaction with and nuclear translocation of AR (androgen receptor) [300]. Therefore,

CAD silencing in prostate cancer leads to reduced cell proliferation.

146

PAK5 FAM83H MAGED1 Sec16A CAD

BioID GFP-PD 5 20 40

Cdc42 2 1

24 PAK4 EMD LZTS2 Database

Figure 5.9 Weak overlap between PAK5-asscoaietd proteins extracted from databases and those identified by BioID and GFP pull-down. Comparison of the complete list of PAK5 'associated' proteins extracted from four public databases (BioGRID, IntAct-EBI, Mentha, SPIKE) with the PAK5 associated proteins (Ratio H/L>3 and peptide number>4 ) identified by either BioID or GFP-PD. Although PAK5 is predominantly located at cell-cell junctions, neither the GFP-PD nor the database extracted proteins reflect this. but a well conserved coiled-coil regions and C-terminal PDZ domain binding sequence which can interact with the scaffolding protein Shank of postsynaptic density [162]. LZTS1 is the best studied member of this family and is ubiquitously expressed. Its role as a tumor suppressor is widely accepted based on the results using

LZTS1 knockout mice [301]. LZTS2 is expressed in most tissues with the higher expression in prostate and testis [145]. LZTS1 and LZTS2 share all the domain features that have been mapped (Figure 5.10). LZTS2 knockout mice are susceptible to both spontaneous and carcinogen induced tumor development [152]; however with the lower incidence compared to LZTS1 deficient mice [301]. LZTS2 is reported as a negative regulator of Wnt signaling via interaction with β-catenin [148]. The role of

LZTS2 in development could be mediated in part through its effect on Wnt signaling

[160, 163]. Indeed, LZTS2 knock-out mice show kidney abnormalities, and here alteration in β-catenin localization is observed in the renal tubule epithelium [160].

N-terminal of LZTS2 contains a region, 25 amino-acids downstream of the proposed initiation codon (ATG), almost identical to the N-terminal of LZTS1 which contains a consensus N-myristoylation sequence (Figure 5.10) [302]. Thus, LZTS2 mRNA appears to have the potential for alternate translation from two ATG sequences

(Figure 5.10) in which the second (LZTS2b) potentially generates a membrane-bound form of the protein.

148

A hLZTS1 ------MGSVSSLISGHSFHSKHCRASQY 23 hLZTS2 MAIVQTLPVPLEP--APEAATAPQAPVMGSVSSLISGRPCPGGPAPPRHH 48 hLZTS3 MAKLETLPVRADPGRDPLLAFAPRPSELGPPDPRLAMGSVGSGVAHAQEF 50 :*. .. :: ...... hLZTS1 KLRKSSHLKKLNRYSDGLLRFGFSQDSGHGKSSSKMGKSED---FFYIKV 70 hLZTS2 GPPGPTFFRQQ----DGLLRGGYEAQEPLCPAVPPRKAVPV---TSFTYI 91 hLZTS3 AMKS-VGTRTGGGGSQGSFPGPRGSGSGASRERPGRYPSEDKGLANSLYL 99 : :* : . . : hLZTS1 SQKARGSHHPDYTALSSGDLGGQAG------VDFDPSTPPKLMPFSN 111 hLZTS2 NEDFRTESPPSPSSDVEDAREQRAH------NAHLRGPPPKLIPVSG 132 hLZTS3 NGELRGSDHTDVCGNVVGSSGGSSSSGGSDKAPPQYREPSHPPKLLATSG 149 . . * . .. . . : . ****:. *. hLZTS1 QLEMGSEKGAVRPTAFKPVLPRSG------AI 137 hLZTS2 KLEKNMEKILIRPTAFKPVLPKPRGAPSLPSFMGPRATGLSGSQGSLTQL 182 hLZTS3 KLDQCSEP-LVRPSAFKPVVPKNFHS------MQN 177 :*: * :**:*****:*: hLZTS1 LHSSPESASHQLHPAPPDKPKEQE-LKPGLCSG----ALSDSGRNSMSSL 182 hLZTS2 FGGPASSSSSSSSSSAADKPLAFSGWASGCPSG----TLSDSGRNSLSSL 228 hLZTS3 LCPPQTNGTPEGRQGPGGLKGGLDKSRTMTPAGGSGSGLSDSGRNSLTSL 227 : . ..: ...... :* ********::** hLZTS1 PTHSTSSSYQLDPLVTPVGPTSRFG-----GSAHNITQGIVLQDSNMMSL 227 hLZTS2 PTYSTGGAEPTTSSPGGHLPSHGSGRGALPGPARGVPTGPSHSDSGRSSS 278 hLZTS3 PTYSSSYSQHLAPLSASTSHINRIG----TASYGSGSGGSSGGGSGYQDL 273 **:*:. : . * .. . . * .*. . hLZTS1 KALSFSDGGSKLG------HSNKADKGPSCVRSPISTDECSIQE 265 hLZTS2 SKSTGSLGGRVAGGLLGSGTRASPDSSSCGERSPPPPPPPPSDEALLHCV 328 hLZTS3 GTSDSGRASSKSGSSSSMGRPGHLGSGEGGGGGLPFAACSPPSPSALIQE 323 . .. * ...... hLZTS1 LEQKLLEREGALQKLQRSFEEKELASSLAYEERPRRCRDELEGPEPKGGN 315 hLZTS2 LEGKLRDREAELQQLRDSLDENEATMCQAYEERQRHWQREREALR----E 374 hLZTS3 LEERLWEKEQEVAALRRSLEQSEAAVAQVLEERQKAWERELAELRQGCSG 373 ** :* ::* : *: *:::.* : . . *** : . * . hLZTS1 KLKQASQKSQRAQQVLHLQVLQLQQEKRQLRQELESLMKEQDLLETKLRS 365 hLZTS2 DCAAQAQRAQRAQQLLQLQVFQLQQEKRQLQDDFAQLLQEREQLERRCAT 424 hLZTS3 KLQQVARRAQRAQQGLQLQVLRLQQDKKQLQEEAARLMRQREELEDKVAA 423 . ::::***** *:***::***:*:**::: *::::: ** : : hLZTS1 YEREKTSFGPALEETQWEVCQKSGEISLLKQQLKESQTEVNAKASEILGL 415 hLZTS2 LEREQRELGPRLEETKWEVCQKSGEISLLKQQLKESQAELVQKGSELVAL 474 hLZTS3 CQKEQADFLPRIEETKWEVCQKAGEISLLKQQLKDSQADVSQKLSEIVGL 473 ::*: .: * :***:******:***********:**::: * **::.* hLZTS1 KAQLKDTRGKLEGLELRTQDLEGALRTKGLELEVCENELQRKKNE----- 460 hLZTS2 RVALREARATLRVSEGRARGLQEAARARELELEACSQELQRHRQE----- 519 hLZTS3 RSQLREGRASLREKEEQLLSLRDSFSSKQASLELGEGELPAACLKPALTP 523 : *:: *..*. * : .*. : :: .** . ** : hLZTS1 --AELLREKVNLLEQELQELRAQAALARDMGPPTFPEDVPALQR------502 hLZTS2 --AEQLREKAGQLDAEAAGLREPPVPPATADPFLLAESDEAKVQRAAAGV 567 hLZTS3 VDPAEPQDALATCESDEAKMRRQAGVAAAASLVSVDGEAEAGGESGTR-- 571 . :: : : :* . . . . . * .

149

hLZTS1 ------ELERLRAELREERQGHDQMSSGFQHERLVWKEEKEKVIQYQKQL 546 hLZTS2 GGSLRAQVERLRVELQRERRRGEEQRDSFEGERLAWQAEKEQVIRYQKQL 617 hLZTS3 --ALRREVGRLQAELAAERRARERQGASFAEERRVWLEEKEKVIEYQKQL 619 :: **:.** **: :. .* ** .* ***:**.***** hLZTS1 QQSYVAMYQRNQRLEKALQQLARGDSAGEPLEVDLEGAD----IPYEDII 592 hLZTS2 QHNYIQMYRRNRQLEQELQQLSLELEARELADLGLAEQAPC--ICLEEIT 665 hLZTS3 QLSYVEMYQRNQQLERRLRERGAAGGASTPTPQHGEEKKAWTPSRLERIE 669 * .*: **:**::**: *:: . * * * hLZTS1 ATEI 596 hLZTS2 ATEI 669 hLZTS3 STEI 673 :***

B

hLZTS2 GCCTCTGCCACCATGGCCATTGTGCAGACTCTGCCAGTGCCACTGGAGCCTGC 53 mLZTS2 GCCTCTGCCACCATGGCCATTGTGCACACTCTGCCAGTGCCTCTGGAGCCCGC 53

hLZTS2 TCCTGAAGCTGCCACTGCCCCACAAGCTCCAGTCATGGGTAGTGTGAGCAGC 105 mLZTS2 ACGTGAGACAGCCACTGCCCCCAAAACTCCAGCAATGGGCAGCGTGAGCAGT105

Figure 5.10 Sequence alignment of human LZTS family isoforms. A. ClustalW sequence alignment of human LZTS1, LZTS2 and LZTS3. Asterisks indicate identical amino acids while colon and periods indicate the conservative amino acids substitutions. Green highlights show the initiation codon (methionine): alternate start codons are suggested for both LZTS2 and LZTS3 based on manual alignment. Isoform one, LZTS2a, is 25 amino acids longer than LZTS2b. A conserved myristoylation consensus sequence (M-G-X-X-X-(S/T)-X-X-X) is highlighted gray in LZTS1 and the shorter isoforms of LZTS2 and LZTS3. The yellow highlights show the most conserved sequences among these proteins which are predicted coiled coil helices. The PDZ binding motif is highlighted in pink. B. Alignment of parts of mRNA sequence from human (h) and mouse (m) LZTS2. The yellow shaded residues indicate start codon and the red shaded residues is representative of kozak sequence (GccRccATGG).

150

5.8.1 Identification of the region of PAK5 involved in binding LZTS2

In order to map the regions of PAK5 involved in LZTS2 binding, I carried out a series of immunoprecipitation experiments using GST-LZTS2 and the PAK5 N- terminal deletion constructs described previously (section 3.7). N-terminal deletions greater than 236 amino acids were found to abolish the interaction of LZTS2 (Figure

5.11A,B). Because I have previously shown that removing 236 amino acids from

PAK5 disrupts its oligomerization (Figure 3.7), there is a possibility that LZTS2 binding requires an oligomeric form of PAK5 or that the first 236 amino acids of

PAK5 includes the LZTS2 interaction site. To evaluate these hypothesizes; I used instead GST-tagged versions of series of PAK5 constructs in combination with Flag

LZTS2 construct to test their interaction using co-immunoprecipitation. As mentioned previously (Section 4.4.2), the GST protein under physiological conditions is completely dimeric [207]. The data in figure 5.12 showed that all the N-terminal truncated forms of GST PAK5 were able to co-precipitate with Flag-LZTS2 with similar efficiency. Thus, PAK5 self-association aids its interaction with LZTS2 and suggesting that the LZTS2 binding site(s) were more C-terminal. To evaluate these potential interaction sites, I compared Flag-LZTS2 interaction with GST-PAK5 (1-

268) and GST-PAK5 (360-719). As seen in figure 5.13A, both N-terminal and C- terminal regions of PAK5 appeared to bind with similar efficiency. To find out which part of LZTS2 plays role in PAK5 interaction, I tested co-immunoprecipitation of

PAK5 full length with a series of truncated constructs of LZTS2. The results suggested that the FEZ domain of LZTS2 is enough for interaction with PAK5

(Figure 5.13B,C).

151

A Flag vector Flag

- Flag constructs

+ PAK5 ∆159 PAK5 + + PAK5 ∆109 PAK5 + ∆178 PAK5 + ∆236 PAK5 + ∆268 PAK5 + ∆419 PAK5 + GFP + kD ∆50 PAK5 + GST LZTS2 150 IB: Anti-GST (co-IP)

Flag IP 75

IB: Anti-Flag

37

Input: Anti-GST

B isoform Catalytic LZTS2 Dimerization interaction ability 108 specific 419 domain 719 +++ +++ PAK5 (51-719) PAK5 (109-719) +++ +++ PAK5 (160-719) +++ +++ PAK5 (179-719) ++ ++ PAK5 (237-719) - - PAK5 (269-719) - - PAK5 (420-719) - -

Figure 5.11 PAK5 interacts directly with LZTS2. A. HEK-293 cells were co-transfected with a series of N-terminal truncated constructs of Flag-PAK5 and full-length GST-LZTS2 as indicated. Anti-Flag Sepharose were used to isolate the complex between Flag-PAK5 and GST-tagged proteins. The samples then were subjected to SDS-PAGE and Western blotting to detect LZTS2 co-immunoprecipitation. B. Graph shows the truncated constructs of PAK5. The results of PAK5 constructs interaction with LZTS2 and the ability of self-association of these constructs have been summarized which show a correlation between these two properties of PAK5.

152

GST constructs

PAK5 ∆360 PAK5

+ PAK5 ∆268 PAK5 + + PAK5 PAK5 ∆109 + PAK5 ∆310 + PAK5 ∆360 + - Vector + + PAK5 PAK5 FL + Flag LZTS2 kD

75

IB: Anti-GST (Co-IP)

Flag IP

100 IB: Anti-Flag

150

75 Input: Anti-GST

25

Figure 5.12 PAK5 oligomerization promotes interaction with LZTS2. Flag-LZTS2 was co-expressed with GST-PAK5 constructs as indicated in HEK-293 cells. The Flag tagged proteins were recovered using anti- Flag Sepharose followed by SDS-PAGE and Western blot to analyze GST-PAK5 constructs co-immunoprecipitation. The experiment indicates that the PAK5 C-terminal construct (PAK5∆360) can bind LZTS2 when is tagged with GST but not Flag (Figure 5.11).

153

719

-

268

268 - A -

GST constructs

PAK5 1 PAK5

- + PAK5 PAK5 360 + Vector + + PAK5 PAK5 1 + Flag LZTS2 kD

50 IgG band IB: Anti-GST Flag IP (Co-IP)

100 IB: Anti-Flag

Input: Anti-GST

Figure 5.13 Investigating the basis for an interaction between PAK5 and LZTS2. A. Flag LZTS2 co-transfected with either GST constructs of PAK5 or pXJ-GST vector in HEK-293 cells. Immunoprecipitation with anti-flag beads were performed and GST-PAK5 constructs binding were identified after Western blotting using mouse anti-GST antibody.

154 B LZTS2 (1-669) PAK5 binding CC FEZ ++++

LZTS2 ∆N (322-669) ++++ LZTS2 ∆C (1-540) ++++

LZTS2 1-CC (1-434) +

LZTS2 FEZ (435-639)

++++

CC

-

FL

FL - - GST constructs

C LZTS2

+ LZTS2 ∆N LZTS2 + ∆C LZTS2 + LZTS21 + FEZ LZTS2 + - + LZTS2 + Flag PAK5 kD IB: Anti-Flag 75

GST-pull down 75 IB: Anti-GST

50

75 Input: Anti-Flag

Figure 5.13 Investigating the basis for an interaction between PAK5 and LZTS2. FEZ domain is sufficient to bind PAK5. B. Schematic of LZTS2 constructs used to test their interaction with full length PAK5 and summary of these interactions. C. Flag-PAK5 was co-expressed with GST LZTS2 constructs in HEK-293 cells. After overnight expression the cells were lysed and the GST fusion proteins were recovered using glutathione Sepharose and co-immunoprecipitation of PAK5 was detected after Western blot. Image J were used to quantify the intensity of bands for PAK5 binding.

155 5.8.2 Identification and localization of a new form of LZTS2

A previous study suggested that LZTS2 could be found at the centrosome and midbody, depending on the stage of the cell cycle [154]. To examine subcellular localization of these alternate LZTS2 isoforms, I generated an N-terminally tagged

GST-LZTS2a construct and a C-terminal GST tagged construct corresponding to

LZTS2b. These two proteins localized quite differently as one would expect. GST-

LZTS2 was found in puncta that was mainly concentrated near the nucleus whereas the shorter N-myristoylated LZTS2b was clearly localized to mitochondria, as assessed by the Mitotracker signal (Figure 5.14). Given the lack of consensus regarding the localization of endogenous LZTS2, I examined commercial antibodies for their ability of detecting endogenous LZTS2 (given that the protein must be relatively abundant in U2OS cells). Of these antibodies, only a rabbit polyclonal antibody (Protein Tech Group) appeared to detected several bands in the correct marker range (Figure 5.15A). siRNA experiments in U2OS cells indicated that only the middle band of the triplet (arrow) shows reduced intensity (Figure 5.15B), suggesting that the other bands were related LZTS isoforms or unrelated cross- reacting antigens. LZTS2 expression was assessed in multiple cell lines (Figure

5.14A). U2OS showed relatively high expression of LZTS2.

Although the tested LZTS2 antibody did not specifically identify LZTS2, I decided to test indirect-immunostaining on the basis that siRNA treatment would lead to specific deletion of the protein in U2OS cells. I noted consistent junctional staining in this cell line (Figure 5.15C) and siRNA treatment diminished the cell junctional staining but not the peri-nuclear signal. Thus it appears LZTS2 is predominantly in cell-cell junctions of U2OS cells. To evaluate which of the two isoforms of LZTS2 (Figure

5.14) is most likely to be expressed in U2OS cells, I tested non-tagged versions of

156

LZTS2a and the shorter LZTS2b and assessed their localization by using indirect immunofluorescence staining in which endogenous LZTS2 mRNA was suppressed by using an LZTS2 siRNA that target the 3'UTR of the mRNA (siRNA-U-1: residues

2683-2701 siRNA-U-2: residues 2289-2307). The LZTS2a demonstrated cytoplasmic staining as well as punctate with a concentration in the peri-nuclear region, whereas myristoylated LZTS2b was found on mitochondria and at cell-cell junctions (Figure

5.16). Although LZTS2a isoform is 25 amino-acids longer, the presence of the N- myristoyl generates a form that migrates at a very similar position on SDS- polyacrylamide gels and therefore it is not possible to distinguish the two isoforms on

SDS-PAGE. However, junctional localization of endogenous LZTS2 in U2OS cells suggests that LZTS2b isoform is abundant in U2OS cells.

157

Anti-GST Merge+DNA

< LZTS2

LZTS2-a GST

Anti-GST Mitotracker Merge

GST LZTS2

LZTS2-b Myristoylated

Figure 5.14 Subcellular localization of the two isoforms of LZTS2 in the cell. The N-terminal GST-tagged construct of LZTS2a and C-terminal GST- fusion construct of LZTS2b were generated. COS-7 cells were transfected with 0.5µg of either of these constructs. After 16h the cells were live stained with Mitotracker (150nM) for 20min prior to fixation in 4% PFA. The cells then were subjected to indirect immunofluorescence staining with Mouse anti-GST antibody. (Scale bar: 10µm)

158 A 1

B -

1

U

-

-

MB231

7

293 116

-

7

-

7

- -

H446

-

-

145

-

29

-

-

MDA HEK MDCK NCI U2OS HeLa Huh HCT HT MCF DU LNCaP

kD Cos

U2OS U2OS+siLZTS2 siLZTS2 U2OS+ 75 kD 75

Anti-LZTS2 Anti-LZTS2

C si-Scramble si-LZTS2-U-1

LZTS2

- Anti

Figure 5.15 Analysis of endogenous LZTS2 expression and localization. A. LZTS2 expression was assessed in several cell lines: COS-7, HEK-293, Lung cancer (NCI-H446), osteosarcoma (U2OS), cervical cancer (HeLa), Hepatocellular carcinoma (Huh-7), colorectal cancer (HCT-116, HT-29), Breast cancer (MCF-7, MDA-MB231) and Prostate cancer (DU-145, LNCap) cell lines. U2OS, HEK-293 and HCT- 116 cell lines showed high expression of LZTS2 compared to other cell lines. B. Two different siRNAs that target either ORF (siLZTS2-1) or 3’UTR (siLZTS2-U-1) of LZTS2 mRNA were generated. U2OS cells were transfected with 50nM of either of these siRANAs or si-scramble. After 48h, cells were lysed using RIPA buffer. A total of 50μg proteins were loaded per lane for SDS-PAGE and Western blotting. Rabbit anti-LZTS2 (Proteintech Group) were used to detect LZTS2 expression. The LZTS2 antibody detected several bands but only the intensity of the middle band was weakened after siRNAs treatment C. U2OS cells were transfected with either si-LZTS2-1 or si-scramble for 48h. Cells were washed and fixed in methanol for 5 minutes and then subjected to indirect immunofluorescence staining using rabbit anti-LZTS2 antibody (ProteinTech Group). (Scale bar: 10µm)

159

Anti- LZTS2 Anti- LZTS2 U2OS

LZTS2-a LZTS2-b

7

- COS

LZTS2-a LZTS2-b

Figure 5.16 Subcellular localization of the non-tagged constructs of LZTS2a and LZTS2b in U2OS and COS-7 cells. U2OS and COS-7 cells were transfected with siLZTS2-2 that targets 3’ UTR of the endogenous LZTS2 mRNA. The next day, cells were transfected with 0.5μg/ml of non- tagged constructs of LZTS2a or LZTS2b for16h. Cells then were fixed in methanol and immunostained with rabbit anti-LZTS2 antibody. (Scale bar: 10µm)

160 5.8.3 LZTS2 and PAK5 co-localize to cell-cell junctions

To determine if PAK5 was co-localized with one or both potential isoforms of

LZTS2, I expressed Flag-PAK5 and either GST-LZTS2a or LZTS2b-GST constructs at low level (0.2ug/ml DNA). Under these conditions, I could observe PAK5 co- localizes with GST-LZTS2a in puncta or with LZTS2b at mitochondria (Figure 5.17

A). Given that transient protein expression can alter protein behavior, I decided instead to investigate their colocalization in U2OS cells stably expressing GFP-PAK5

(as described previously in section 4.6.1). Endogenous LZTS2 and GFP-PAK5 clearly co-localized at cell-cell junctions (Fig 5.17 B). A function for LZTS2 at cell-cell junctions is consistent with the phenotype of LZTS2 knockout mice, which exhibit defects in kidney formation [160]. One of the abnormalities noted for LZTS2 null mice was cysts formation in renal tissues. This could result from alterations in adhesive properties of ureteric epithelium cells, probably as a result of reduced expression of adherent junction components [160].

161

A Anti-Flag Anti-GST+DAPI Merge+ DAPI

PAK5 LZTS2-a Merge

Anti-Flag Anti-GST

Merge

PAK5 LZTS2-b Merge

B

Anti-GFP Anti-LZTS2

clone 9 clone

-

PAK5 -

GFP PAK5 LZTS2 Merge

Figure 5.17 Analysis of PAK5 and LZTS2 co-localization in the cell. A. COS-7 cells were co-transfected with 0.2μg/ml of Flag-PAK5 and either GST-LZTS2a or LZTS2b-GST (C-terminal tag). After 16h, cells were fixed in 4% PFA and co-stained with rabbit anti-Flag and mouse anti-GST antibodies. B. 80-90% confluent culture of U2OS GFP-PAK5 stable lines were fixed in methanol for 5 minutes and used for co-immunnostaining with anti-mouse GFP and anti-rabbit LZTS2 antibodies. (Scale bar: 10µm)

162 5.8.4 Junctional localization of PAK5 is independent of LZTS2

My previous results showed a direct interaction between PAK5 and LZTS2 and their co-localization at cell-cell junctions. Because GFP-PAK5 exhibited the both junctional and cytoplasmic puncta localization in U2OS stable lines, I asked whether junctional PAK5 requires LZTS2 interaction. I then transfected U2OS cells stably expressing GFP-PAK5 with si-LZTS2 and assessed PAK5 localization (Figure 5.18).

The results showed that PAK5 localization is not affected by LZTS2 knock down.

5.8.5 Afadin is required for junctional localization of LZTS2 and PAK5

As previously mentioned, nectin and afadin were characterized as PAK5 proximal proteins using BioID (Table 5.2). Afadin is a junctional protein that binds to the

Nectin and the complex of these two proteins along Cadherin-catenin complex form the major components of cell adhesions (Figure 1.6). I was therefore interested to analyze afadin interaction with PAK5 and check whether afadin is required for junctional recruitment of PAK5. Since afadin is a quite large protein, the expression vector of this protein is constructed in two fragments as shown in Figure 5.19A [303].

Co-immunoprecipitation experiment was carried out to test the HA-PAK5 interaction with the Flag tagged afadin constructs (Gift from Prof.Yoshimi Takai). The results showed no interaction between the two proteins (Figure 5.19B). Next, I tested the requirement of afadin for junctional localization of GFP-PAK5. U2OS stable lines expressing GFP-PAK5 were transfected with si-RNA against afadin. Figure 5.19C

(left panel) shows that si- afadin is extremely efficient to reduce the level of afadin expression in U2OS cells. Knockdown of afadin was largely diminished localization of PAK5 at cell-cell junctions but not at the centrosomal region (Figure 5.19C, right panel). These results indicate that junctional localization of PAK5 required afadin and

163 suggest that other protein(s) mediates afadin and PAK5 connection. Because LZTS2 is one of the direct interacting partners of PAK5 complex in the junctions, I tested whether PAK5 is connected to afadin via LZTS2 though the previous results suggested that LZTS2 is not required for recruitment of PAK5 to the junctions. Co- immunoprecipotation experiment suggested that LZTS2 might not be directly interacted with afadin (Figure 5.19D). Afadin si-RNA treatment of U2OS cells does not significantly affect β-catenin but was severely reduced LZTS2 at cell-cell junctions (Figure 5.19E). These results strongly suggest that PAK5 and LZTS2 are more proximal to nectin-afadin complex rather than Cadherin-catenin.

5.8.6 Binding to LZTS2 does not regulate PAK5 activity

Given that PAK5 is a constitutively active kinase (Figure 3.2), I was interested to see if interaction with LZTS2 could affect PAK5 activity. To this end, I tested the activity of PAK5 co-immunoprecipitated with LZTS2a In vitro (Figure 5.17). The result showed no difference in activity when PAK5 was recovered with or without

LZTS2.

164

Si-Scramble Si-LZTS2

)

22

-

Clone

PAK5 ( PAK5

- GFP

Figure 5.18 Junctional localization of PAK5 is independent of LZTS2 expression. U2OS cells stably expressing GFP-PAK5 were transfected with either 50nM of si-scramble. The next day, cells were seeded on coverslip and incubated for another 24h. Cells then were fixed in 4% PFA and analyzed for PAK5 localization using fluorescent microscope. (Scale bar: 10µm)

165 RA1 PR2 A RA2 PDZ PR1 PR3

PDZ-C construct

N-PDZ construct (1-1100 aa) (1016-1829)

C C

- PDZ PDZ B - Flag Afadin constructs

HA PAK5

+ N + PDZ + kD IB: anti-HA (co-IP) 75 150 IB: anti-Flag Flag IP 100

Input: anti-HA 75

C U2OS U2OS-GFP-PAK5 (Clone-9)

Scramble

- Si

Afadin

Afadin

- Si

Afadin

Figure 5.19 Afadin-dependent localization of PAK5 and LZTS2 at cell-cell junction. A. Schematic diagram of Afadin fragment constructs. Both constructs contain PDZ domain. B. COS-7 cells were co-transfected with HA-PAK5 and either Flag tagged N-PDZ or PDZ-C constructs of Afadin. Immunoprecipitation was performed with anti-Flag Sepharose and co-immunoprecipitation of PAK5 was analyzed using SDS-PAGE and Western blotting. C. U2OS cells or U2OS cells stably expressing GFP-PAK5 were transfected with either si-Afadin or si-scramble. The next day, cells were seeded on coverslip and after additional 24h, cells were fixed and analyzed. U2os cells were subjected to indirect immunostaining using Rabbit anti-Afadin antibody while U2os stable line were analyzed for PAK5 localization directly after fixation. (Scale bar: 10µm)

166

C)

-

PDZ)

-

(PDZ (N D

Flag constructs

Afadin Afadin

+ + PAK5 + + + HA-LZTS2 kD

IB: Anti-HA (co-IP) 75

Flag IP 150 IB: Anti-Flag 100

Input: Anti-HA 75 E

U2OS U2OS

Scramble

- Si

LZTS2 Beta catenin

Afadin

- Si

LZTS2 Beta catenin

Figure 5.19 Afadin-dependent localization of PAK5 and LZTS2 at cell-cell junction. D. COS-7 cells were co-transfected with HA-LZTS2 and either one of the Flag tagged constructs of afadin or Flag-PAK5. Co- immunoprecipitation experiment with anti-Flag Sepharose following SDS-PAGE and Western blotting was applied to analyze afadin and LZTS2 binding. PAK5 and LZTS2 interaction was considered as a positive control. E. U2OS cells were transfected with either si-Afadin or si-scramble as described in panel C. After cell fixation in methanol, immunostaining was performed using Rabbit anti-LZTS2 or mouse anti- beta catenin antibodies. 167 - 0.5µg 2µg GST-LZTS2 + + + Flag PAK5 kD 50 Anti p-S112 BAD

< GST BAD

75 Flag IP: Anti Flag 100 IB: Anti GST

In vitro kinase assay

Figure 5.20 PAK5 activity is not regulated by binding to LZTS2. COS-7 cells were co-transfected with Flag tagged PAK5 and either GST- LZTS2a or pXJ-GST vector. The next day, cells were lysed and co- immunoprecipitation was performed using anti-Flag Sepharose. In vitro kinase assay was applied to compare the kinase activity of PAK5 or PAK5:LZTS2 complex in the presence of 500mM ATP and GST-BAD as substrate (5μg). GST-BAD phosphorylation on Ser112 was detected after SDS-PAGE and Western blotting.

168 5.9 Analyzing the potential interaction between LZTS2 and Raf-1

The BioID and immuno-localization data indicated that LZTS2 (most likely the LZTS2b isoform), and a number of other junctional proteins can form a complex with PAK5. Co-localization of LZTS2b and PAK5 on mitochondria, indicates they might form also a complex at this site. Raf-1 is a reported interacting partner of PAK5 that can localize to mitochondria via binding to Bcl-2 [304]. I hypothesized that Raf-1 could be a component of PAK5 and LZTS2 mitochondrial complex. To test whether

Raf-1 could interact with LZTS2 and to map the possible interaction site of the two proteins, full length HA-Raf-1 was co-transfected with various truncated constructs of

LZTS2 (Figure 5.21A) and binding assessed by co-immunoprecipitation. As for

PAK5, I noted that the minimal LZTS2 construct capable of binding Raf-1 largely requires the coiled-coil FEZ (Figure 5.21B). Raf-1 interaction with other coiled-coil proteins is reported. For example Raf-1 and ROCK interaction plays role in regulating cell migration [305].

169

A LZTS2 (1-669) CC FEZ

LZTS2 ∆N (322-669)

LZTS2 ∆C (1-540)

LZTS2 1-CC (1-434)

LZTS2 FEZ (435-639) CC

B -

N N C

Δ Δ

GST constructs

+ LZTS2 + LZTS2 + FEZ LZTS2 + 1 LZTS2 + kD LZTS2 + HA Raf-1

75 IB: anti-Raf-1

100 GST-pull down 75

IB : anti GST

75 Input: anti-Raf-1

Figure 5.21 Investigating the interaction site of LZTS2 with Raf-1. A. Schematic of LZTS2 constructs were used to test their interaction with full length Raf-1. B.HEK-293 cells were co-transfected with GST tagged truncated constructs of LZTS2 and HA-Raf-1 as indicated. The next day, cells were lysed and harvested. Glutathione Sepharose were used to recover the GST-LZTS2 protein. HA-Raf-1 binding were detected after SDS-PAGE and Western blotting with mouse anti-Raf-1 antibody (Upstate).

170 5.10 LZTS2 mediates PAK5 and Raf-1 co-localization at mitochondria

It has been reported previously that full-length PAK5 and Raf-1 can co- localize on mitochondria [51]. My results did not show a consistent co-localization of the two proteins (Figure 4.10) and in low expression the two proteins were not co- localized. These results suggested that PAK5 and Raf-1 interaction may not be stable in COS-7 cells. Since I showed that LZTS2 could interact with both PAK5 and Raf-1, one explanation might be the presence (or not) of sufficient LZTS2b protein whose translation might be cell-type specific. To test whether LZTS2b expression stabilizes

PAK5 and Raf-1 interaction, COS-7 cells were transfected with GFP-PAK5 and

LZTS2b isoform in addition to low levels of HA-Raf-1. All three proteins accumulated in peri-nuclear region (Figure 5.22 A,B); however in some of these cells the mitochondrial distribution is disrupted making the localization ambiguous (data not shown). In the previous section (Figure 5.14) the localization of LZTS2b was clearly mitochondrial suggesting these proteins (PAK5, Raf-1 and LZTS2b) are associated with mitochondria. These results suggest that myristoylated isoform of

LZTS2 (LZTS2b) may stabilize PAK5 and Raf-1 interaction and mediate their co- localization at mitochondria.

171

A

GFP-PAK5 Raf-1 LZTS2-b

B

80 70 Raf-1 60 PAK5 50 LZTS2-b Merge 40 30 20 10 0 1 3 5 7 9 11 13 15 17 19 21 23

Figure 5.22 Analysis of colocalization of expressed PAK5, LZTS2b and Raf-1 in COS-7 cells. A. Representative confocal microscopic image of cells transfected with 0.3 µg of each of HA-Raf-1, GFP-PAK5 and GST tagged LZTS2b constructs. After 16h, the cells were fixed in 4% PFA and co-immunostained for Raf-1 and LZTS2b using mouse anti-HA and rabbit anti-GST antibodies respectively. B. Line scan analysis of signals from three channels was performed along the indicated white arrow. (Scale bar: 10µm)

172 5.11 LZTS2 does not affect Raf-1 Ser338 phosphorylation.

The mechanism of Raf-1 activation is complex and involves the participation of scaffolding proteins such as KSR and 14-3-3 proteins [112]. Since there seemed to be evidence of LZTS2 interaction with Raf-1 (as with PAK5), thus at cell-cell junctions LZTS2 could play role on regulation of Raf-1 activity. I then treated U2OS cells with siRNA to LZTS2 and investigated the phosphorylation status of Ser338 of

Raf-1 and its downstream kinase (ERK1/2) phosphorylation [112, 306]. I also probed the status of phosphorylation on BAD Ser112, because Raf-1 acts upstream of BAD

[307]. Lowering of LZTS2 levels did not appear to directly play an important role in

Raf-1 activation and BAD phosphorylation (Figure 5.23).

173

loaded buffer(see adding or phosphorylation Figure Si - scramble 10 in 5 % each . 23 methods) 37 75 kD FCS 25 25 37 75 75 kD The lane ( 50 . for nM) U for 2 effect OS 10 + containing - SDS for min cells

24 Serum starvation - of PAGE + - h as . were Cells LZTS 0 indicated . 1 and - + transfected % 174 2 were Western SDS 10 min Serum stimulation suppression + Si - . then . Cell Si Anti Anti A Anti Anti - Anti Loading control staining:CBB Scramble with - blotting LZTS2 serum total lysate - - - - p - LZTS2 p BAD p - - either S338 Raf - S112 BAD ERK of starved . on prepared 50 si Raf μg - LZTS - 1 - for proteins 1 using 2 20 and - 2 h ( 50 before RIPA were BAD nM ) 5.11 Discussion

5.11.1 Identification of PAK5 interacting partners in U2OS cells

Protein-protein interactions are a fundamental requirement for many cellular processes. The ability of proteins to form specific associations with neighboring proteins often underlies their function (for adaptor proteins) or allows for targeting of enzymes to the correct location. In the recent years, several studies have reported that increased PAK5 expression promotes or is associated with the cancer development

[130-132, 134, 209], but the signaling pathway(s) underlying this phenomenon are poorly understood. One reason might be that the intracellular site of PAK5 activity is not known, and only few partners of PAK5 have been uncovered (Figure 1.5). A summary of published PAK5 partners is given below.

Affinity pull-down mass spectrometry (AP-MS) demonstrate p120-catenin as a PAK5 partner in HeLa cells [54]. PAK5 was shown to phosphorylate p120 on

Ser288; though the function of this interaction is unknown. A yeast two hybrid (Y2H) screen against a brain cDNA library identified MARK-2 kinase and transcription factor E47 as interactors of PAK5 [53, 138]. PAK5 was found to interact with

MARK-2 via the kinase domain and down-regulated its activity. In vitro kinase assay suggested E47 is a substrate for PAK5 and its phosphorylation correlates with PAK5 expression and colon cancer progression in tissue samples. Synaptojanin and Pacsin-1 were found in vitro substrates for recombinant PAK5 isolated from mouse brain lysate

[140]. These two proteins act in a complex to regulate synaptic vesicle recycling and

PAK5 was suggested to promote Pacsin-1 and synaptojanin interaction by their phosphorylation. PAK5 was found to phosphorylate transcription factor GATA1 in vitro. This phosphorylation was connected to the increased epithelial-mesenchymal

175 transition in breast cancer cells [139]. Raf-1 was identified as a PAK5 interactor by co-immunoprecipitation analysis of overexpressed proteins in HEK-293 cells and in the context of brain lysate [51]. Finally on the basis of what was known about PAK4

[40], BAD was suggested as a PAK5 substrate [52].

This study is the first to implicate PAK5 in cell-cell junctions. This new localization at cell-cell junctions of osteosarcoma cell line U2OS (Figure 4.15) has been used to identify proximal 'partners' for PAK5 in this compartment. The two methods that I employed include the recently described BioID method [242] which has been used to define the ZO-1 compartment [308], and a more traditional GFP pull down (GFP-PD), but using SILAC analysis for mass spectrometry (MS) in order to remove 'background' signal. The SILAC approach allows one to distinguish non- specific light proteins [245] such as keratins and proteins that non-specifically bind

Sepharose (Referred to as the Crapome, http://www.crapome.org).

Two high confidence data sets (ratio H/L≥3 and the peptide number≥4) were prepared from the BioID and GFP-PD data (Table 5.2 and 5.3). The majority of proteins were not shared between the two methods (Figure 5.9) and since the BioID yields primarily junctional proteins (where PAK5 was enriched), we must assume that this is the most accurate list.

Among the previously reported PAK5 partners, p120 catenin was identified by BioID method (SILAC ratio= 5.9) and Cdc42 was seen in both BioID and GFP-PD data sets.

The PAK5 Bio-ID method identified a relatively small number of cell-cell junction proteins set (Figure 5.5) compared with the reported ZO-1 associated set

[308]. The fact that junctional proteins are totally absent in GFP-PAK5 PD data set indicates the pull-down method is unreliable, perhaps because the genuine 'interactors'

176 cannot be solubilized effectively (Table 5.3). The PAK5 is likely to form a protein complex in the cell membrane which may not be extracted under mild lysis buffer

(0.5 % Triton X-100) that were used for cell lysis in GFP-PD. Additional ionic detergents for example sodium deoxycholate (DOC) might help in this respect. The

BioID process does not require the protein-protein interactions to be maintained, and indeed SDS is used to ensure these are disrupted (see figure 5.2). Standard affinity purification methods need to maintain protein-protein interactions, which is why a mild lysis buffer was used. In the case of BioID, the labelling of proximal proteins occurs in the cell and covalently attaches the label. The capture of biotin with streptavidin is one of the most stable non-covalent interactions and can survive 1%

SDS [244].

Among the junctional proteins identified with BioID (Table 5.2), Nectin-2 and

Afadin show a SILAC ratio >10 suggesting they are the most proximal to PAK5 versus ZO-1 with much lower SILAC ratio (ratio H/L 3.2). In this cell type, the adhesions are not well organized consistent with immature junctions in these cancer cells that may undergo continuous remodeling. The ZO-1 compartment cannot be found as a 'tight' band typical for say MDCK cells [309]. Figure 5.5 shows that most of the junctional proteins are associated together as annotated by STRING

(http://string-db.org). Beside the well validated junctional proteins (as classified in

Figure 5.5) some other proteins in our BioID set have also been found associated with junctions such as NRAGE [285], ASPP2 [310] and LL5-β [311] suggesting these proteins could be part of a subset of proteins neighboring PAK5. Comparison of neighboring proteins of PAK5 (in my study) with those identified for ZO-1[308] showed partially overlaps: Afadin, p120-catenin, Nectin-2 and Aspp2 were identified as common neighboring proteins for PAK5 and ZO-1 in junctions. ASPP2 (TP53BP2)

177 is an interesting PAK5 proximal protein because it is reported to interact with both junctional and mitochondrial proteins [310, 312]. This protein was first identified as a tumor suppressor protein that interacts with p53 and Bcl-2 and induces p53 mediated mitochondrial death pathway [312]. ASPP2 was also shown to form a complex with

β-catenin and E-cadherin in the junctions of multiple epithelial cell lines [310].

Identification of some microtubule associated proteins in the PAK5 BioID set, such as CEP85, could indicate that the kinase associates with the centrosome. This is consistence with co-localization of PAK5 with centrosomal marker γ-tubulin (Figure

4.16D). It would be interesting to test CEP85 as LZTS2 interacting partner because

LZTS2 is thought to localize to the centrosome and bind the centrosomal protein

“katanin” [154].

Since PAK5 has previously been associated with mitochondria [50-52], any putative mitochondrial proteins are of interest. The protein Citrin is a Ca2+ binding mitochondrial carrier of Aspartate-Glutamate found within the inner membrane of mitochondria [268] and which does not contain biotin (though it might bind the mitochondrial biotinylated proteins such as PCCA (Propionyl-CoA carboxylase) and

PC (Pyruvate carboxylase)). Since PAK5 probably only accesses the mitochondrial outer membrane [52], one would not expect these PAK5 and citrin to interact directly.

The estimated radius of activity for BirA* is 20-30nm in vivo [242] while the distance between the cytoplasmic surface of outer membrane and the matrix surface of inner membrane of mitochondria is 17-20nm [313]. However, given the large number of proteins that should be detected on the mitochondrial surface (which are not), I suggest citrin is an artifact.

178

The GFP-PAK5 PD detected several mitochondrial proteins (like AIFM and

ATAD3A) which are likely to be false positives typically found with tag affinity pull- down [234]: it is easy to imagine that after cell lysis mitochondrial proteins are often released into the lysate, allowing them to interact non-specifically.

Analysis of the two data sets of SILAC-enriched proteins identified using

BioID and GFP-PD from the same cell-line shows the presence of 5 common proteins, FAM83H, MAGE-D1, Sec16A, CAD and Cdc42 (Figure5.9). These proteins could well represent those can complex with PAK5 (directly or indirectly), which could be considered as interesting candidates for further studies. Among these proteins, only Cdc42 is a known interacting partner of the group II PAKs. The GFP-

PD identified Cdc42 with relatively high SILAC ratio (=10.3) but this was not the case with the BioID analysis (SILAC ratio <2). This might be because the interaction with PAK5 masks key residues involved in Cdc42 modification and actually prevents its biotinylation. Cdc42 is a small single domain protein with multiple interactors in its active and inactive state [314, 315]. We would not expect the protein to be easily labelled by active AMP-biotin.

Of the five proteins mentioned above, NRAGE (MAGE-D1) could be an interesting candidate for further studies. Both PAK5 and NRAGE are involved in apoptosis regulation pathway [52, 288, 316]. Although PAK5 is generally considered anti-apoptotic, NRAGE is reported to be both pro- and anti-apoptotic under different conditions [293, 316, 317]. The primary sequence of NRAGE indicates that it has a cysteine-rich N-terminus that is likely palmitoylated and a hydrophobic C-terminus that could act as a membrane anchor. Both NRAGE and PAK5 expression are linked to increased tumor cell migration [130, 131, 134, 285]. NRAGE may interfere with cell adhesions by disrupting E-cadherin-β-catenin interactions [285]. How PAK5

179 promotes migration is unclear: it could affect cell-cell junctions or up-regulate the metalloproteinase MMP2 that can disrupt the extracellular matrix [131, 134].

The biotinylation of some ribosomal and RNA associated proteins is expected but should be controlled for the biotin ligase [308]. In addition, (random) clonal differences in the level of translational proteins might well be picked up by SILAC.

The protein TNRC6B exhibited a SILAC ratio >9 in the PAK5 BioID set (Table 5.2).

This protein encodes a component of RNA interference machinery which is required for the gene silencing mediated with microRNAs/small interfering [318]. One would expect other members of the complex to be identified, if they were the 'PAK5- proximal' complex.

Taken together I have seen that the bulk of PAK5 BioID derived 'partners' are junctional, suggesting that the method as applied by our lab is a sensitive and reliable technique for identification of proximal proteins. The PAK5 neighbors seen here have not been identified previously and overlap in part with those identified using a similar analysis of PAK4 (Felicia Tay, personal communication). Although in this study I have only a single biological experiment to determine PAK5 GFP-PD or BioID targets, the technique which was developed in the Manser lab has an estimated reliability of 90% based on the studies using Paxillin and Kindlin-2 (Dong Jing-Ming, personal communication). This means that on average 10% of 'hits' will not fulfill the

H/L SILAC ratio >3 in different experiments. Other non-SILAC studies have also suggested BioID is potentially more reproducible than protein-protein interaction based techniques [242, 308, 319].

180

5.11.2 Is LZTS2 a key partner for PAK5?

The result of BioID suggested LZTS2 is closely associated with PAK5. The higher SILAC ratio of LZTS2 versus to other hits (Table 5.2) suggests that this protein is the most proximal interacting partner of PAK5.

Co-immunoprecipitation analysis (Figure 5.11) demonstrated that the so called

FEZ domain of LZTS2 (which is unrelated to the protein FEZ1 [320]) interacts with both N-terminal and C-terminal regions of PAK5 (Figure5.13). PAK5 oligomerization promotes this interaction; again highlighting the importance of PAK5 self-association to its localization (Figure 3.7) and interaction with other proteins (Figure 5.12).

Binding to LZTS2 did not affect PAK5 activity as measured in vitro (Figure 5.20).

Nonetheless, co-expression of PAK5 and LZTS2 leads to an obvious mobility shift of

LZTS2 on SDS-PAGE suggesting LZTS2 phosphorylation (Figure 5.12). Further analysis suggested PAK5 sites were present throughout the protein.' Phospho Site

Plus' showed several phosphorylation sites in LZTS2. Based on optimal substrate sequence of group II PAKs [37], I suggest that Ser 99, Ser 276, Ser278 and Ser281 have the potential of phosphorylation by group II PAKs.

I have proposed that two translational isoforms of LZTS2 exist in cells based on the different localization of these isoforms (Figure 5.14). The N-terminal residues of shorter isoform contain a consensus myristoylation sequences almost identical to

N-terminal of LZTS1 (Figure 5.10), and shown to be N-myristoylated [302].

Myristoylation is an irreversible attachment of myristic acid to an N-terminal glycine

[321] after removal of the (initiation) methionine. Rare examples of modification to internal Lysines have also been reported [322]. Myristoylation strongly affect protein localization in the same way as C-terminal farnesylation [321]. The post-translational modification of Gly (at position +2) is the most common form of modification; but

181 myristoylation can be occurred in exposed internal Glycine(s). For example caspase cleavage of Bid leads to exposure of an internal myristoylation motif in C-terminal fragment (ctBid) [323]. Myristoylated ctBid translocates to the mitochondrial membrane where it promotes release of cytochrome c. In the case of LZTS2, we propose that myristoylation should be a co-translational modification with ribosomal selection of the second ATG sequence which fits the Kozak consensus [324] (Figure

5.10B). We hypothesize that the conservation of this internal N-myristoylation sequence (with LZTS1) only occurs because this ATG is functional. The two isoforms which I term LZTS2a and LZTS2b show quite different localization: LZTS2a demonstrated a punctate distribution particularly near the nucleus (Figures 5.14 and

5.16) and is consistent with a centrosomal-enriched localization of LZTS2 in osteosarcoma cell line SaOs-2 [154]. I have not tested the ability of LZTS2 to interact with the p80 subunit of katanin at this site as reported [157]. The LZTS2b isoform localizes on mitochondria and cell membrane (Figures 5.14 and 5.16) consistent with the membrane localization of LZTS1 in hippocampal neurons [302]. Although the anti-LZTS2 antibody shows significant non-specific signal (based on siRNA effects), clearly, endogenous LZTS2 is largely junctional in U2OS cells (Figure 5.15) suggesting that myristoylated isoform might be the dominant isoform of LZTS2 in this cell line. The mitochondrial localization of the protein was not seen by indirect immunofluorescent staining because of this background signal in peri-nuclear region.

Cell fractionation is probably required to reveal any mitochondrial localization of the protein.

My results are the first report of a junctional localization of LZTS2, which fits with a role in regulating β-catenin location in the kidney cells of LZTS2 null mice

[160]. My current model suggests that the LZTS2b may stabilize the β-catenin

182 complex in cell junctions (of epithelium of ductal tubes) and decrease the amount of

β-catenin in the cytoplasm. The interaction of the LZTS2b with β-catenin, should be tested in the light of these results

Interestingly ASPP2, another tumor suppressor, is also though to enhance β-catenin levels at cell-cell junctions [310]. It should be noted that we did not detect β-catenin in the PAK5 BioID in U2OS cells although β-catenin is abundant in these cells

(Tables 5.2 and S1).

Interestingly, PAK5 was shown to localize on mitochondria when co- expressed with LZTS2b in COS-7 cells where mitochondria are well resolved (Figure

5.17A). Endogenous LZTS2 was clearly co-localized with GFP-PAK5 at junctions; however, the lower level of protein at mitochondria could not be resolved (Figure

5.17B). The PAK5 BioID analysis (Table 5.2) suggests LZTS2 is a close partner for

PAK5 in cell-cell junctions. Since I did not observe a significant change of the GFP-

PAK5 signal with LZTS2 siRNA (Figure 5.18), it seems likely that PAK5 does not depend on LZTS2b for its junctional localization. Instead, I found that this localization is dependent to the presence of afadin, another ‘proximal PAK5’ protein identified using BioID. In addition, afadin is appeared to be required for junctional recruitment of LZTS2, although neither PAK5 nor LZTS2 could interact with afadin.

It would be interesting to test Nectin interaction with PAK5 and LZTS2, because except LZTS2, this protein is supposed to be the most proximal junctional protein to PAK5.

I have shown that LZTS2 can also directly interact with Raf-1 through the coiled-coil region known as the FEZ domain (Figure 5.21) and demonstrated a mitochondrial ternary complex between PAK5, Raf-1 and LZTS2 (Figure 5.22). Since

183 we did not observe Raf-1 in the PAK5 BioID analysis, future experiments to validate this potential interaction using GFP-BirA*-LZTS2 need to be carried out. If LZTS2 indeed interacts with Raf-1, a ternary complex between PAK5, Raf-1 and LZTS2 might be important for the biology of both PAK5 and LZTS2 particularly in the context of mitochondria. Co-immunoprecipitation experiment using brain lysate has suggested an interaction between PAK5 and Raf-1 [51]; but this could be indirect because PAK5 and Raf-1 do not appear to form a stable complex in COS-7 cells

(Figure 4.10C). The addition of LZTS2 to the system might well stabilize this complex as I have shown (Figure 5.23). One should note that LZTS2 levels are low in

COS-7 cells (Figure 5.15A); but the Raf-1 connection could be important in tissues in which LZTS2 level are high.

The mitochondrial localization of PAK5/Raf-1/LZTS2b in COS-7 cells (Figure

5.22) which do not normally form cell-cell junctions should be re-tested in cells with well-organized junctions. This would allow us to test the hypothesis that if LZTS2 mediates Raf-1 and PAK5 interaction in the junctions where it may play roles other than classical ERK activation, although presently Raf-1 is not known to be a junctional component. Attempts to detect endogenous Raf-1 either at junctions or on mitochondria have been unsuccessful (data not shown).

On the other hand, the jucntional protein ASPP2, which promotes oncogenic-induced senescence, is reported to act through Ras/Raf/MEK/ERK pathway [325]. It is notable that LZTS2 knockdown did not affect Raf-1 phosphorylation in U2OS cells (Figure

5.23), which might indicate that only a pool of Raf-1 is affected.

Taken together, I have confirmed the direct interaction of PAK5 and LZTS2 and mapped the regions responsible for their mutual interaction. I found that LZTS2b

184 could target mitochondria in addition to cell-cell junctions and find that LZTS2 co- localized with PAK5 in GFP-PAK5 U2OS cells. Although the LZTS2 interaction does not affect PAK5 activity, it is likely to be involved in substrate recruitment: the C- terminal region of LZTS proteins bind to PDZ domain containing proteins such as

ProSAP2/Shank3 [149, 162]. The role of LZTS2 in linking PAK5 to Raf-1 required further investigation. My current model suggests that both LZTS1 and LZTS2 binds to a variety of proteins through their coiled-coil domains and act as an adaptor for as yet un-determined pathways.

185

Chapter 6. General Conclusions

6.1 The group II PAK5 oligomerization underlies its high basal activity

Based on several lines of evidence from biochemical experiments (Figures 3.9,

3.10), indirect immunofluorescence staining (Figure 3.7) and size estimation based on gel filtration (Figure 3.11), it appears that PAK5 is an oligomeric protein and its cellular puncta distribution reflects this. I have shown that PAK5 contains a conserved central specific-region (residues 109-420) that is not found (or not conserved) in

PAK4 (Figure 3.6) and this region is critical for the protein self-association. PAK5 oligomerization is directly implicated in the known elevated levels of PAK5 activity

(in the absence of co-transfected activated Cdc42). PAK5 has long been known to be a constitutively activated protein [33] and here I have demonstrated that it contains a functional auto-inhibitory domain at an N-terminal position as for PAK4 [28, 29] which seems paradoxical. I have shown that elevated basal activity of Flag-PAK5 results from kinase oligomerization (Figure 3.12).

My current model (Figure 6.1A) suggests that PAK5 is in equilibrium between monomer, dimer and oligomeric state. Self-association antagonizes the auto-inhibited conformation of the kinase in which the N-terminal AID normally binds to the C- terminal catalytic domain [28]. This probably occurs by preventing the AID binding to the kinase domain, perhaps by conformationally blocking the interaction (Figure

6.1A). Based on this model, PAK5 is more active as an oligomer, and the loss of self- association likely drives an auto-inhibited state.

Based on careful biochemical studies, I showed that PAK5 is ~35% active relative to the full activity of PAK5 when it is bound to Cdc42 (Figure 3.3). Cdc42(G12V) expression leads to the loss of PAK5 puncta (data not shown), indicating that Cdc42

186 also blocks kinase oligomerization. The complex between Cdc42(G12V) and PAK5 is difficult to probe by gel filtration because we require detergent to solubilize the protein. However, all of this data is consistent with the distribution of Flag or GFP-

PAK5 in the cytoplasm where it forms the puncta (Figures 3.5, 3.7, 4.2) and GFP-

PAK5 behavior at junctions of U2OS stable lines where it likely binds to Cdc42 and normally do not form the puncta (Figure 4.15B).

The gene expression data base (http://www.ebi.ac.uk/) suggests that PAK5 expression is always significantly lower than PAK4. Using the phospho-specific antibody (recognizes phospho-Ser602 and phospho-Ser474 of PAK5 and PAK4 activation loop respectively), I estimate that NCI-H446 cells contain 30% PAK5 relative to PAK4 and actually this cell line expresses one of the highest levels of

PAK5 mRNA (Based on gene expression data base).

At low levels of the kinase, PAK5 is likely to be predominantly at cell-cell junctions and very little in 'active' punctate state. The ability of PAK5 to be active without

Cdc42 interaction may be one reason that its level is generally low (compared with

PAK4). Nonetheless, PAK5 has been found to be overexpressed in variety of cancers

[130, 132, 133]; but whether it is a 'driver' mutation in cancer is not resolved.

My suggested model provides an explanation for the previous proposal that a fragment of central region (residues 60-180) acts as an 'auto-inhibitory domain' [30].

Incubation of the full length PAK5 with GST-PAK5(1-180) blocked PAK5 activity in vitro. As shown in figure 6.1B, this fragment in excess is likely to compete for binding with other full length PAK5 molecules. This binding however, may not interfere with the (monomeric) auto-inhibited conformation, thus shifting the balance to an auto-inhibited state.

187

A auto-inhibited partial auto-inhibition active

monomer Dimer Oligomer

B inhibited inhibited

PAK5-FL

PAK5 (1-180)

PAK5 (1-180) CRIB

AID

Figure 6.1. A model for PAK5 oligomerization and subsequent activation. A. in monomeric state PAK5 is autoinhibited by the binding of the AID to the catalytic domain. The presence of a conserved specific central region (residues 109-420) in PAK5 promotes its oligomerization which antagonizes the AID association with the catalytic domain. B. PAK5 1-180 fragment reduces PAK5 activity in vitro.[30]. We suggest that this can occur by promoting the 'monomeric' PAK5 state (left side) or by promoting AID binding with the full-length (active) PAK5.

188 6.2 Identification of PAK5 as a junctional kinase

In contrast to previous studies [31, 52], my findings suggest that GFP-PAK5 is not primarily resident of mitochondria (Figure 4.2). Instead, PAK5 is clearly present at cell-cell junctions (Figures 4.15 and 4.16) and may be directly associated with junctional proteins such as LZTS2 in this compartment (Figure 5.17B). The BioID assay also indicated that afadin and a number of other junctional proteins including

Nectin-2, PAK4 and p120 catenin are proximal to PAK5 in U2OS cells (Table 5.2).

Because this analysis will reveal the 'average' behavior of the protein in cell junctions

(Figure 5.16), we may not be able to detect proteins proximal to a minority of the kinase (such as at mitochondria). My current model (Figure 6.2) suggests that PAK5 is associated with the myristoylated isoform of LZTS2 (and possibly LZTS1) in junctions, however its localization does not appear to depend on this interaction

(Figure 5.18). The junctional PAK5 is likely to be in vicinity of nectin-afadin because afadin knockdown had a great effect on junctional localization of PAK5 but not β- catenin (Figure 5.19). I have not seen a significant difference in the overall structure of cell-cell adhesions in the GFP-PAK5 cell lines. Since PAK5 is not directly associated with the plasma membrane, it is likely to be in equilibrium with cytoplasmic PAK5 and is also enriched at centrosome where it may associate with

CEP85. At present, it is possible that the labelling of Kif7 is the result of its accumulation at cell-cell junctions in these cells. The presence of PAK5 at mitochondria (via interaction with LZTS2b) has yet to be demonstrated. PAK5 activity at the mitochondrial surface might be expected to protect the organelle from

BAD. At mitochondria, PAK5 phosphorylates BAD on Ser112 directly. If interacts with Raf-1, PAK5 can induce BAD phosphorylation on Ser136 via Raf-1/Akt pathway. Phosphorylated BAD on Ser112/136 is inactive and will be sequestered by

189

14-3-3 protein in the cytoplasm. It should be noted however that Raf-1 has not been identified by BioID (Table 5.2).

This model (Figure 6.2) is based on the expression of GFP-BirA* PAK5 in the

U2OS cells. Because PAK5 expression is dominant in the brain, it is interesting to speculate that similar proteins associate with PAK5 in neurons: cell-cell junctions between neurons occur primarily at synapses, and both afadin and cadherins play a key role in establishing and maintaining these structures [326, 327]. A number of

PAK5 proximal proteins (listed in Table 5.2) are shown to localize at post synaptic density (PSD), such as LZTS2 [149], SPAR [149] and Erbin [328]. The phenotype of

PAK5/PAK6 double knockout mice indicates defects in neurite outgrowth resulting in deficits in memory and learning [49, 329] which could be the consequence of defects in synaptic contacts [330]. The proteins that are assumed to be proximal to PAK5 in vivo (Table 5.2), are potentially kinase targets of PAK5, and it would be an important goal to test these as PAK5 substrates in the next phase of the project.

190

junctions mitochondria association F reported mitochondria, may PAK nectin surrounded

igure Basal 5 target LZTS2b - at afadin 6 [

. 50 cell 2 and

+ several Nucleus ] by A . ? - with cell compartment [ the 52 centrosomal proposed a Cadherin ] complex . kinase junctions

neighboring α-catenin β-catenin α-catenin Here, β-catenin proteins Adherens may I of model . MTOC and show LZTS region in proteins play ? the probably Kif7 proteins - junction 2 for new . role junctions ? Nectin is In + 191 which PAK - one 42 Cdc on cell localization . centrosome - regulation 2 5 PAK of junctions, are

which Afadin

localization Afadin the 5 most is direct has sites reported of likely and I apoptosis yet suggest interacting for Tight junction Tight in AKT mitochondria to associated PAK to the be localize

S112 ZO-1 ZO-1 that identified as 5 cell p partners at previously S136 p PAK with and cell . to

PAK

. 5 Apical - cell the the On its of is 5 14 p BAD - 3 - 3 p References:

1. Murali, A. and K. Rajalingam, Small Rho GTPases in the control of cell shape and mobility. Cell Mol Life Sci, 2014. 71(9): p. 1703-21.

2. Sadok, A. and C.J. Marshall, Rho GTPases: masters of cell migration. Small GTPases, 2014. 5: p. e29710.

3. Bastea, L.I., et al., Protein kinase D-mediated phosphorylation at Ser99 regulates localization of p21-activated kinase 4. Biochem J, 2013. 455(2): p. 251-60. 4. Geiger, T., et al., Comparative proteomic analysis of eleven common cell lines reveals ubiquitous but varying expression of most proteins. Mol Cell Proteomics, 2012. 11(3): p. 25. 5. Schwanhausser, B., et al., Global quantification of mammalian gene expression control. Nature, 2011. 473(7347): p. 337-42. 6. Zhao, Z.S. and E. Manser, PAK family kinases: Physiological roles and regulation. Cell Logist, 2012. 2(2): p. 59-68. 7. Huveneers, S. and E.H. Danen, Adhesion signaling - crosstalk between integrins, Src and Rho. J Cell Sci, 2009. 122(Pt 8): p. 1059-69. 8. Manser, E., et al., A brain serine/threonine protein kinase activated by Cdc42 and Rac1. Nature, 1994. 367(6458): p. 40-6. 9. Jaffer, Z.M. and J. Chernoff, p21-activated kinases: three more join the Pak. Int J Biochem Cell Biol, 2002. 34(7): p. 713-7. 10. Dammann, K., V. Khare, and C. Gasche, Tracing PAKs from GI inflammation to cancer. Gut, 2014. 63(7): p. 1173-84. 11. Chan, P.M. and E. Manser, PAKs in human disease. Prog Mol Biol Transl Sci, 2012. 106: p. 171-87. 12. Wells, C.M. and G.E. Jones, The emerging importance of group II PAKs. Biochem J, 2010. 425(3): p. 465-73. 13. Bokoch, G.M., et al., Interaction of the Nck adapter protein with p21-activated kinase (PAK1). J Biol Chem, 1996. 271(42): p. 25746-9. 14. Puto, L.A., et al., p21-activated kinase 1 (PAK1) interacts with the Grb2 adapter protein to couple to growth factor signaling. J Biol Chem, 2003. 278(11): p. 9388-93. 15. Manser, E., et al., PAK kinases are directly coupled to the PIX family of nucleotide exchange factors. Mol Cell, 1998. 1(2): p. 183-92. 16. Lei, M., et al., Structure of PAK1 in an autoinhibited conformation reveals a multistage activation switch. Cell, 2000. 102(3): p. 387-97.

192

17. Parrini, M.C., et al., Pak1 kinase homodimers are autoinhibited in trans and dissociated upon activation by Cdc42 and Rac1. Mol Cell, 2002. 9(1): p. 73-83. 18. Yu, J.S., et al., Identification of the regulatory autophosphorylation site of autophosphorylation-dependent protein kinase (auto-kinase). Evidence that auto- kinase belongs to a member of the p21-activated kinase family. Biochem J, 1998. 334 ( Pt 1): p. 121-31. 19. Chong, C., et al., The mechanism of PAK activation. Autophosphorylation events in both regulatory and kinase domains control activity. J Biol Chem, 2001. 276(20): p. 17347-53. 20. Zhao, Z.S., et al., A conserved negative regulatory region in alphaPAK: inhibition of PAK kinases reveals their morphological roles downstream of Cdc42 and Rac1. Mol Cell Biol, 1998. 18(4): p. 2153-63. 21. Kumar, R., A.E. Gururaj, and C.J. Barnes, p21-activated kinases in cancer. Nat Rev Cancer, 2006. 6(6): p. 459-71. 22. Brown, M.C., K.A. West, and C.E. Turner, Paxillin-dependent paxillin kinase linker and p21-activated kinase localization to focal adhesions involves a multistep activation pathway. Mol Biol Cell, 2002. 13(5): p. 1550-65. 23. Galisteo, M.L., et al., The adaptor protein Nck links receptor tyrosine kinases with the serine-threonine kinase Pak1. J Biol Chem, 1996. 271(35): p. 20997-1000. 24. Koh, C.G., et al., The p21-activated kinase PAK is negatively regulated by POPX1 and POPX2, a pair of serine/threonine phosphatases of the PP2C family. Curr Biol, 2002. 12(4): p. 317-21. 25. Xia, C., et al., Regulation of the p21-activated kinase (PAK) by a human Gbeta -like WD-repeat protein, hPIP1. Proc Natl Acad Sci U S A, 2001. 98(11): p. 6174-9. 26. Wallace, S.W., et al., Cdc42 regulates apical junction formation in human bronchial epithelial cells through PAK4 and Par6B. Mol Biol Cell, 2010. 21(17): p. 2996-3006. 27. Alahari, S.K., P.J. Reddig, and R.L. Juliano, The integrin-binding protein Nischarin regulates cell migration by inhibiting PAK. Embo j, 2004. 23(14): p. 2777-88. 28. Baskaran, Y., et al., Group I and II mammalian PAKs have different modes of activation by Cdc42. EMBO Rep, 2012. 13(7): p. 653-9. 29. Ha, B.H., et al., Type II p21-activated kinases (PAKs) are regulated by an autoinhibitory pseudosubstrate. Proc Natl Acad Sci U S A, 2012. 109(40): p. 16107- 12. 30. Ching, Y.P., et al., Identification of an autoinhibitory domain of p21-activated protein kinase 5. J Biol Chem, 2003. 278(36): p. 33621-4. 31. Wu, X. and J.A. Frost, Multiple Rho proteins regulate the subcellular targeting of PAK5. Biochem Biophys Res Commun, 2006. 351(2): p. 328-35.

193

32. Abo, A., et al., PAK4, a novel effector for Cdc42Hs, is implicated in the reorganization of the actin cytoskeleton and in the formation of filopodia. Embo j, 1998. 17(22): p. 6527-40. 33. Dan, C., et al., PAK5, a new brain-specific kinase, promotes neurite outgrowth in N1E-115 cells. Mol Cell Biol, 2002. 22(2): p. 567-77. 34. Yang, F., et al., Androgen receptor specifically interacts with a novel p21-activated kinase, PAK6. J Biol Chem, 2001. 276(18): p. 15345-53. 35. Liu, X., et al., Direct interaction between AR and PAK6 in androgen-stimulated PAK6 activation. PLoS One, 2013. 8(10): p. e77367. 36. Kaur, R., et al., Activation of p21-activated kinase 6 by MAP kinase kinase 6 and p38 MAP kinase. J Biol Chem, 2005. 280(5): p. 3323-30. 37. Arias-Romero, L.E. and J. Chernoff, A tale of two Paks. Biol Cell, 2008. 100(2): p. 97-108. 38. Dan, C., et al., Cytoskeletal changes regulated by the PAK4 serine/threonine kinase are mediated by LIM kinase 1 and cofilin. J Biol Chem, 2001. 276(34): p. 32115-21. 39. Li, Z., et al., p21-activated kinase 4 phosphorylation of integrin beta5 Ser-759 and Ser-762 regulates cell migration. J Biol Chem, 2010. 285(31): p. 23699-710. 40. Gnesutta, N., J. Qu, and A. Minden, The serine/threonine kinase PAK4 prevents caspase activation and protects cells from apoptosis. J Biol Chem, 2001. 276(17): p. 14414-9. 41. Callow, M.G., et al., Requirement for PAK4 in the anchorage-independent growth of human cancer cell lines. J Biol Chem, 2002. 277(1): p. 550-8. 42. Liu, Y., et al., The pak4 protein kinase plays a key role in cell survival and tumorigenesis in athymic mice. Mol Cancer Res, 2008. 6(7): p. 1215-24. 43. Lee, S.R., et al., AR and ER interaction with a p21-activated kinase (PAK6). Mol Endocrinol, 2002. 16(1): p. 85-99. 44. Schrantz, N., et al., Mechanism of p21-activated kinase 6-mediated inhibition of androgen receptor signaling. J Biol Chem, 2004. 279(3): p. 1922-31. 45. Pandey, A., et al., Cloning and characterization of PAK5, a novel member of mammalian p21-activated kinase-II subfamily that is predominantly expressed in brain. Oncogene, 2002. 21(24): p. 3939-48. 46. Li, X. and A. Minden, Targeted disruption of the gene for the PAK5 kinase in mice. Mol Cell Biol, 2003. 23(20): p. 7134-42. 47. Minden, A., PAK4-6 in cancer and neuronal development. Cell Logist, 2012. 2(2): p. 95-104. 48. Qu, J., et al., PAK4 kinase is essential for embryonic viability and for proper neuronal development. Mol Cell Biol, 2003. 23(20): p. 7122-33.

194

49. Nekrasova, T., et al., Targeted disruption of the Pak5 and Pak6 genes in mice leads to deficits in learning and locomotion. Dev Biol, 2008. 322(1): p. 95-108. 50. Cotteret, S. and J. Chernoff, Nucleocytoplasmic shuttling of Pak5 regulates its antiapoptotic properties. Mol Cell Biol, 2006. 26(8): p. 3215-30. 51. Wu, X., et al., p21 activated kinase 5 activates Raf-1 and targets it to mitochondria. J Cell Biochem, 2008. 105(1): p. 167-75. 52. Cotteret, S., et al., p21-Activated kinase 5 (Pak5) localizes to mitochondria and inhibits apoptosis by phosphorylating BAD. Mol Cell Biol, 2003. 23(16): p. 5526-39. 53. Matenia, D., et al., PAK5 kinase is an inhibitor of MARK/Par-1, which leads to stable microtubules and dynamic actin. Mol Biol Cell, 2005. 16(9): p. 4410-22. 54. Wong, L.E., et al., p120-catenin is a binding partner and substrate for Group B Pak kinases. J Cell Biochem, 2010. 110(5): p. 1244-54. 55. Soosairajah, J., et al., Interplay between components of a novel LIM kinase-slingshot phosphatase complex regulates cofilin. Embo j, 2005. 24(3): p. 473-86. 56. Kozma, R., et al., Rho family GTPases and neuronal growth cone remodelling: relationship between increased complexity induced by Cdc42Hs, Rac1, and acetylcholine and collapse induced by RhoA and lysophosphatidic acid. Mol Cell Biol, 1997. 17(3): p. 1201-11. 57. Leeuwen, F.N., et al., The guanine nucleotide exchange factor Tiam1 affects neuronal morphology; opposing roles for the small GTPases Rac and Rho. J Cell Biol, 1997. 139(3): p. 797-807. 58. Bryan, B., et al., GEFT, a Rho family guanine nucleotide exchange factor, regulates neurite outgrowth and dendritic spine formation. J Biol Chem, 2004. 279(44): p. 45824-32. 59. Meng, W. and M. Takeichi, Adherens junction: molecular architecture and regulation. Cold Spring Harb Perspect Biol, 2009. 1(6): p. a002899. 60. Ebnet, K., Organization of multiprotein complexes at cell-cell junctions. Histochem Cell Biol, 2008. 130(1): p. 1-20. 61. Troyanovsky, S., Cadherin dimers in cell-cell adhesion. Eur J Cell Biol, 2005. 84(2- 3): p. 225-33. 62. Takai, Y., et al., Nectins and nectin-like molecules: roles in cell adhesion, migration, and polarization. Cancer Sci, 2003. 94(8): p. 655-67. 63. Kooistra, M.R., N. Dube, and J.L. Bos, Rap1: a key regulator in cell-cell junction formation. J Cell Sci, 2007. 120(Pt 1): p. 17-22. 64. Selamat, W., et al., The Cdc42 Effector Kinase PAK4 Localizes to Cell-Cell Junctions and Contributes to Establishing Cell Polarity. PLoS One, 2015. 10(6).

195

65. Fram, S., et al., A PAK6-IQGAP1 complex promotes disassembly of cell-cell adhesions. Cell Mol Life Sci, 2014. 71(14): p. 2759-73. 66. Li, Y., et al., Nucleo-cytoplasmic shuttling of PAK4 modulates beta-catenin intracellular translocation and signaling. Biochim Biophys Acta, 2012. 1823(2): p. 465-75. 67. Noren, N.K., et al., p120 catenin regulates the actin cytoskeleton via Rho family GTPases. J Cell Biol, 2000. 150(3): p. 567-80. 68. Kourtidis, A., S.P. Ngok, and P.Z. Anastasiadis, p120 catenin: an essential regulator of cadherin stability, adhesion-induced signaling, and cancer progression. Prog Mol Biol Transl Sci, 2013. 116: p. 409-32. 69. Daniel, J.M., Dancing in and out of the nucleus: p120(ctn) and the transcription factor Kaiso. Biochim Biophys Acta, 2007. 1773(1): p. 59-68. 70. Peglion, F. and S. Etienne-Manneville, p120catenin alteration in cancer and its role in tumour invasion. Philos Trans R Soc Lond B Biol Sci, 2013. 368(1629): p. 20130015. 71. Thoreson, M.A. and A.B. Reynolds, Altered expression of the catenin p120 in human cancer: implications for tumor progression. Differentiation, 2002. 70(9-10): p. 583-9. 72. Shibata, T., et al., Cytoplasmic p120ctn regulates the invasive phenotypes of E- cadherin-deficient breast cancer. Am J Pathol, 2004. 164(6): p. 2269-78. 73. Anastasiadis, P.Z. and A.B. Reynolds, Regulation of Rho GTPases by p120-catenin. Curr Opin Cell Biol, 2001. 13(5): p. 604-10. 74. Anastasiadis, P.Z., et al., Inhibition of RhoA by p120 catenin. Nat Cell Biol, 2000. 2(9): p. 637-44. 75. Anastasiadis, P.Z., p120-ctn: A nexus for contextual signaling via Rho GTPases. Biochim Biophys Acta, 2007. 1773(1): p. 34-46. 76. Reynolds, A.B., et al., The novel catenin p120cas binds classical cadherins and induces an unusual morphological phenotype in NIH3T3 fibroblasts. Exp Cell Res, 1996. 225(2): p. 328-37. 77. Portt, L., et al., Anti-apoptosis and cell survival: a review. Biochim Biophys Acta, 2011. 1813(1): p. 238-59. 78. Duprez, L., et al., Major cell death pathways at a glance. Microbes Infect, 2009. 11(13): p. 1050-62. 79. Elmore, S., Apoptosis: a review of programmed cell death. Toxicol Pathol, 2007. 35(4): p. 495-516. 80. Taylor, R.C., S.P. Cullen, and S.J. Martin, Apoptosis: controlled demolition at the cellular level. Nat Rev Mol Cell Biol, 2008. 9(3): p. 231-41.

196

81. Danial, N.N. and S.J. Korsmeyer, Cell death: critical control points. Cell, 2004. 116(2): p. 205-19. 82. Chipuk, J.E. and D.R. Green, How do BCL-2 proteins induce mitochondrial outer membrane permeabilization? Trends Cell Biol, 2008. 18(4): p. 157-64. 83. Adams, J.M. and S. Cory, The Bcl-2 protein family: arbiters of cell survival. Science, 1998. 281(5381): p. 1322-6. 84. Czabotar, P.E., et al., Control of apoptosis by the BCL-2 protein family: implications for physiology and therapy. Nat Rev Mol Cell Biol, 2014. 15(1): p. 49-63. 85. Danial, N.N., BAD: undertaker by night, candyman by day. Oncogene, 2008. 27 Suppl 1: p. S53-70. 86. Schurmann, A., et al., p21-activated kinase 1 phosphorylates the death agonist bad and protects cells from apoptosis. Mol Cell Biol, 2000. 20(2): p. 453-61. 87. Li, X., et al., Phosphorylation of caspase-7 by p21-activated protein kinase (PAK) 2 inhibits chemotherapeutic drug-induced apoptosis of breast cancer cell lines. J Biol Chem, 2011. 286(25): p. 22291-9. 88. Rudel, T. and G.M. Bokoch, Membrane and morphological changes in apoptotic cells regulated by caspase-mediated activation of PAK2. Science, 1997. 276(5318): p. 1571-4. 89. Eswaran, J., et al., UnPAKing the class differences among p21-activated kinases. Trends Biochem Sci, 2008. 33(8): p. 394-403. 90. Datta, S.R., et al., 14-3-3 proteins and survival kinases cooperate to inactivate BAD by BH3 domain phosphorylation. Mol Cell, 2000. 6(1): p. 41-51. 91. Masters, S.C., et al., 14-3-3 inhibits Bad-induced cell death through interaction with serine-136. Mol Pharmacol, 2001. 60(6): p. 1325-31. 92. Datta, S.R., et al., Akt phosphorylation of BAD couples survival signals to the cell- intrinsic death machinery. Cell, 1997. 91(2): p. 231-41. 93. Harada, H., et al., p70S6 kinase signals cell survival as well as growth, inactivating the pro-apoptotic molecule BAD. Proc Natl Acad Sci U S A, 2001. 98(17): p. 9666- 70. 94. Houslay, M.D., A RSK(y) relationship with promiscuous PKA. Sci STKE, 2006. 2006(349): p. pe32. 95. Fang, X., et al., Regulation of BAD phosphorylation at serine 112 by the Ras- mitogen-activated protein kinase pathway. Oncogene, 1999. 18(48): p. 6635-40. 96. Fox, C.J., The serine/threonine kinase Pim-2 is a transcriptionally regulated. 2003. 17(15): p. 1841-54. 97. Macdonald, A., et al., Pim kinases phosphorylate multiple sites on Bad and promote 14-3-3 binding and dissociation from Bcl-XL. BMC Cell Biol, 2006. 7: p. 1.

197

98. Tan, Y., et al., BAD Ser-155 phosphorylation regulates BAD/Bcl-XL interaction and cell survival. J Biol Chem, 2000. 275(33): p. 25865-9. 99. Konishi, Y., et al., Cdc2 phosphorylation of BAD links the cell cycle to the cell death machinery. Mol Cell, 2002. 9(5): p. 1005-16. 100. Ye, D.Z., et al., p21-Activated Kinase 1 (Pak1) Phosphorylates BAD Directly at Serine 111 In Vitro and Indirectly through Raf-1 at Serine 112. PLoS One, 2011. 6(11). 101. Liu, J. and A. Lin, Role of JNK activation in apoptosis: a double-edged sword. Cell Res, 2005. 15(1): p. 36-42. 102. Chong, H., H.G. Vikis, and K.L. Guan, Mechanisms of regulating the Raf kinase family. Cell Signal, 2003. 15(5): p. 463-9. 103. Chong, H. and K.L. Guan, Regulation of Raf through phosphorylation and N terminus-C terminus interaction. J Biol Chem, 2003. 278(38): p. 36269-76. 104. Dumaz, N. and R. Marais, Protein kinase A blocks Raf-1 activity by stimulating 14-3- 3 binding and blocking Raf-1 interaction with Ras. J Biol Chem, 2003. 278(32): p. 29819-23. 105. Hafner, S., et al., Mechanism of inhibition of Raf-1 by protein kinase A. Mol Cell Biol, 1994. 14(10): p. 6696-703. 106. Dougherty, M.K., et al., Regulation of Raf-1 by direct feedback phosphorylation. Mol Cell, 2005. 17(2): p. 215-24. 107. Terai, K. and M. Matsuda, Ras binding opens c-Raf to expose the docking site for mitogen-activated protein kinase kinase. EMBO Rep, 2005. 6(3): p. 251-5. 108. Raabe, T. and U.R. Rapp, Ras signaling: PP2A puts Ksr and Raf in the right place. Curr Biol, 2003. 13(16): p. R635-7. 109. Matallanas, D., et al., Raf family kinases: old dogs have learned new tricks. Genes Cancer, 2011. 2(3): p. 232-60. 110. Dhillon, A.S., et al., Regulation of Raf-1 activation and signalling by dephosphorylation. Embo j, 2002. 21(1-2): p. 64-71. 111. Baccarini, M., Second nature: biological functions of the Raf-1 "kinase". FEBS Lett, 2005. 579(15): p. 3271-7. 112. Baljuls, A., B.N. Kholodenko, and W. Kolch, It takes two to tango--signalling by dimeric Raf kinases. Mol Biosyst, 2013. 9(4): p. 551-8. 113. Zang, M., et al., Characterization of Ser338 phosphorylation for Raf-1 activation. J Biol Chem, 2008. 283(46): p. 31429-37. 114. Mason, C.S., et al., Serine and tyrosine phosphorylations cooperate in Raf-1, but not B-Raf activation. Embo j, 1999. 18(8): p. 2137-48.

198

115. Beeser, A., et al., Role of group A p21-activated kinases in activation of extracellular-regulated kinase by growth factors. J Biol Chem, 2005. 280(44): p. 36609-15. 116. Cammarano, M.S., et al., Pak4 induces premature senescence via a pathway requiring p16INK4/p19ARF and mitogen-activated protein kinase signaling. Mol Cell Biol, 2005. 25(21): p. 9532-42. 117. Jin, S., et al., p21-activated Kinase 1 (Pak1)-dependent phosphorylation of Raf-1 regulates its mitochondrial localization, phosphorylation of BAD, and Bcl-2 association. J Biol Chem, 2005. 280(26): p. 24698-705. 118. Salzano, M., et al., Calcium/calmodulin-dependent protein kinase II (CaMKII) phosphorylates Raf-1 at serine 338 and mediates Ras-stimulated Raf-1 activation. Cell Cycle, 2012. 11(11): p. 2100-6. 119. Kumar, R., A.E. Gururaj, and C.J. Barnes, p21-activated kinases in cancer. Nat Rev Cancer, 2006. 6(6): p. 459-471. 120. Ye, D.Z. and J. Field, PAK signaling in cancer. Cell Logist, 2012. 2(2): p. 105-116. 121. Radu, M., et al., PAK signalling during the development and progression of cancer. Nat Rev Cancer, 2014. 14(1): p. 13-25. 122. Qu, J., et al., Activated PAK4 regulates cell adhesion and anchorage-independent growth. Mol Cell Biol, 2001. 21(10): p. 3523-33. 123. Eswaran, J., M. Soundararajan, and S. Knapp, Targeting group II PAKs in cancer and metastasis. Cancer Metastasis Rev, 2009. 28(1-2): p. 209-17. 124. Chen, H., et al., Expression and prognostic significance of p21-activated kinase 6 in hepatocellular carcinoma. J Surg Res, 2014. 189(1): p. 81-8. 125. Kaur, R., et al., Increased PAK6 expression in prostate cancer and identification of PAK6 associated proteins. Prostate, 2008. 68(14): p. 1510-6. 126. Wen, X., et al., Knockdown of p21-activated kinase 6 inhibits prostate cancer growth and enhances chemosensitivity to docetaxel. Urology, 2009. 73(6): p. 1407-11. 127. Giroux, V., J. Iovanna, and J.C. Dagorn, Probing the human kinome for kinases involved in pancreatic cancer cell survival and gemcitabine resistance. Faseb j, 2006. 20(12): p. 1982-91. 128. Giroux, V., et al., Combined inhibition of PAK7, MAP3K7 and CK2alpha kinases inhibits the growth of MiaPaCa2 pancreatic cancer cell xenografts. Cancer Gene Ther, 2009. 16(9): p. 731-40. 129. Fawdar, S., et al., Targeted genetic dependency screen facilitates identification of actionable mutations in FGFR4, MAP3K9, and PAK5 in lung cancer. Proc Natl Acad Sci U S A, 2013. 110(30): p. 12426-31.

199

130. Gong, W., et al., P21-activated kinase 5 is overexpressed during colorectal cancer progression and regulates colorectal carcinoma cell adhesion and migration. Int J Cancer, 2009. 125(3): p. 548-55. 131. Wang, X.X., et al., PAK5-Egr1-MMP2 signaling controls the migration and invasion in breast cancer cell. Tumour Biol, 2013. 34(5): p. 2721-9. 132. Li, D., X. Yao, and P. Zhang, The overexpression of P21-activated kinase 5 (PAK5) promotes paclitaxel-chemoresistance of epithelial ovarian cancer. Mol Cell Biochem, 2013. 383(1-2): p. 191-9. 133. Han, K., et al., p21-activated kinase 7 is an oncogene in human osteosarcoma. Cell Biol Int, 2014. 134. Han, Z.X., et al., Downregulation of PAK5 inhibits glioma cell migration and invasion potentially through the PAK5-Egr1-MMP2 signaling pathway. Brain Tumor Pathol, 2013. 135. Wang, X., et al., p21-activated kinase 5 inhibits camptothecin-induced apoptosis in colorectal carcinoma cells. Tumour Biol, 2010. 31(6): p. 575-82. 136. Gu, J., et al., A role for p21-activated kinase 7 in the development of gastric cancer. Febs j, 2013. 280(1): p. 46-55. 137. Zcharia, E., et al., Inhibition of matrix metalloproteinase-2 by halofuginone is mediated by the Egr1 transcription factor. Anticancer Drugs, 2012. 23(10): p. 1022- 31. 138. Zhu, G., et al., PAK5-mediated E47 phosphorylation promotes epithelial- mesenchymal transition and metastasis of colon cancer. Oncogene, 2015. 139. Li, Y., et al., GATA1 induces epithelial-mesenchymal transition in breast cancer cells through PAK5 oncogenic signaling. Oncotarget, 2015. 6(6): p. 4345-56. 140. Strochlic, T.I., et al., Identification of neuronal substrates implicates Pak5 in synaptic vesicle trafficking. Proc Natl Acad Sci U S A, 2012. 109(11): p. 4116-21. 141. McPherson, P.S., et al., A presynaptic inositol-5-phosphatase. Nature, 1996. 379(6563): p. 353-7. 142. Vecchione, A., et al., Fez1/lzts1 alterations in gastric carcinoma. Clin Cancer Res, 2001. 7(6): p. 1546-52. 143. Vecchione, A., et al., Fez1/Lzts1 absence impairs Cdk1/Cdc25C interaction during mitosis and predisposes mice to cancer development. Cancer Cell, 2007. 11(3): p. 275-89. 144. Ishii, H., et al., The FEZ1 gene at chromosome 8p22 encodes a leucine-zipper protein, and its expression is altered in multiple human tumors. Proc Natl Acad Sci U S A, 1999. 96(7): p. 3928-33.

200

145. Cabeza-Arvelaiz, Y., et al., LAPSER1: a novel candidate tumor suppressor gene from 10q24.3. Oncogene, 2001. 20(46): p. 6707-17. 146. Kim, J.M., et al., Effect of the modulation of leucine zipper tumor suppressor 2 expression on proliferation of various cancer cells functions as a tumor suppressor. Mol Cell Biochem, 2011. 346(1-2): p. 125-36. 147. Teufel, A., et al., In silico characterization of LZTS3, a potential tumor suppressor. Oncol Rep, 2005. 14(2): p. 547-51. 148. Thyssen, G., et al., LZTS2 is a novel beta-catenin-interacting protein and regulates the nuclear export of beta-catenin. Mol Cell Biol, 2006. 26(23): p. 8857-67. 149. Schmeisser, M.J., et al., Synaptic cross-talk between N-methyl-D-aspartate receptors and LAPSER1-beta-catenin at excitatory synapses. J Biol Chem, 2009. 284(42): p. 29146-57. 150. Cui, Q.Z., et al., Leucine zipper tumor suppressor 2 inhibits cell proliferation and regulates Lef/Tcf-dependent transcription through Akt/GSK3beta signaling pathway in lung cancer. J Histochem Cytochem, 2013. 61(9): p. 659-70. 151. Hyun Hwa, C., et al., Crossregulation of beta-catenin/Tcf pathway by NF-kappaB is mediated by lzts2 in human adipose tissue-derived mesenchymal stem cells. Biochim Biophys Acta, 2008. 1783(3): p. 419-28. 152. Johnson, D.T., et al., Deletion of Leucine Zipper Tumor Suppressor 2 (Lzts2) Increases Susceptibility to Tumor Development. J Biol Chem, 2013. 288(6): p. 3727- 38. 153. Van Wauwe, J. and B. Haefner, Glycogen synthase kinase-3 as drug target: from wallflower to center of attention. Drug News Perspect, 2003. 16(9): p. 557-65. 154. Sudo, H. and Y. Maru, LAPSER1 is a putative cytokinetic tumor suppressor that shows the same centrosome and midbody subcellular localization pattern as p80 katanin. Faseb j, 2007. 21(9): p. 2086-100. 155. Hartman, J.J., et al., Katanin, a microtubule-severing protein, is a novel AAA ATPase that targets to the centrosome using a WD40-containing subunit. Cell, 1998. 93(2): p. 277-87. 156. McNally, K.P., O.A. Bazirgan, and F.J. McNally, Two domains of p80 katanin regulate microtubule severing and spindle pole targeting by p60 katanin. J Cell Sci, 2000. 113 ( Pt 9): p. 1623-33. 157. Sudo, H. and Y. Maru, LAPSER1/LZTS2: a pluripotent tumor suppressor linked to the inhibition of katanin-mediated microtubule severing. Hum Mol Genet, 2008. 17(16): p. 2524-40.

201

158. Ji, D., S.L. Deeds, and E.J. Weinstein, A screen of shRNAs targeting tumor suppressor genes to identify factors involved in A549 paclitaxel sensitivity. Oncol Rep, 2007. 18(6): p. 1499-505. 159. Abal, M., et al., Microtubule release from the centrosome in migrating cells. J Cell Biol, 2002. 159(5): p. 731-7. 160. Peng, Y., et al., The leucine zipper putative tumor suppressor 2 protein LZTS2 regulates kidney development. J Biol Chem, 2011. 286(46): p. 40331-42. 161. Bridgewater, D., et al., Canonical WNT/beta-catenin signaling is required for ureteric branching. Dev Biol, 2008. 317(1): p. 83-94. 162. Gessert, S., et al., The spatio-temporal expression of ProSAP/shank family members and their interaction partner LAPSER1 during Xenopus laevis development. Dev Dyn, 2011. 240(6): p. 1528-36. 163. Li, Y., et al., Lzts2 regulates embryonic cell movements and dorsoventral patterning through interaction with and export of nuclear beta-catenin in zebrafish. J Biol Chem, 2011. 286(52): p. 45116-30. 164. Manser, E., et al., Expression of constitutively active alpha-PAK reveals effects of the kinase on actin and focal complexes. Mol Cell Biol, 1997. 17(3): p. 1129-43. 165. Siu, M.K., et al., p21-activated kinase 4 regulates ovarian cancer cell proliferation, migration, and invasion and contributes to poor prognosis in patients. Proc Natl Acad Sci U S A, 2010. 107(43): p. 18622-7. 166. Gething, M.J. and J. Sambrook, Protein folding in the cell. Nature, 1992. 355(6355): p. 33-45. 167. Rothbacher, U., et al., Dishevelled phosphorylation, subcellular localization and multimerization regulate its role in early embryogenesis. Embo j, 2000. 19(5): p. 1010-22. 168. Cheng, H., et al., Regulation of c-Fes tyrosine kinase and biological activities by N- terminal coiled-coil oligomerization domains. Mol Cell Biol, 1999. 19(12): p. 8335- 43. 169. Murali, R., et al., Structural analysis of p185c-neu and epidermal growth factor receptor tyrosine kinases: oligomerization of kinase domains. Proc Natl Acad Sci U S A, 1996. 93(13): p. 6252-7. 170. Polley, S., et al., A structural basis for IkappaB kinase 2 activation via oligomerization-dependent trans auto-phosphorylation. PLoS Biol, 2013. 11(6): p. e1001581. 171. Schwarz, J.K., C.M. Lovly, and H. Piwnica-Worms, Regulation of the Chk2 protein kinase by oligomerization-mediated cis- and trans-phosphorylation. Mol Cancer Res, 2003. 1(8): p. 598-609.

202

172. Morgan, D.O., Principles of CDK regulation. Nature, 1995. 374(6518): p. 131-4. 173. Bayliss, R., et al., Structural basis of Aurora-A activation by TPX2 at the mitotic spindle. Mol Cell, 2003. 12(4): p. 851-62. 174. Mozumdar, S. and J.H. Wang, Differentiation of substrate-binding sites in pyruvate kinase by selective photoaffinity labeling. Biochem Mol Biol Int, 1997. 43(3): p. 465- 75. 175. Selamat, W., et al., The Cdc42 Effector Kinase PAK4 Localizes to Cell-Cell Junctions and Contributes to Establishing Cell Polarity. PLoS One, 2015. 10(6): p. e0129634. 176. Harbury, P.B., P.S. Kim, and T. Alber, Crystal structure of an isoleucine-zipper trimer. Nature, 1994. 371(6492): p. 80-3. 177. Faux, M.C., et al., Recruitment of adenomatous polyposis coli and [beta]-catenin to axin-puncta. Oncogene, 2008. 27(44): p. 5808-5820. 178. Schwarz-Romond, T., C. Metcalfe, and M. Bienz, Dynamic recruitment of axin by Dishevelled protein assemblies. J Cell Sci, 2007. 120(Pt 14): p. 2402-12. 179. Schwarz-Romond, T., et al., The DIX domain of Dishevelled confers Wnt signaling by dynamic polymerization. Nat Struct Mol Biol, 2007. 14(6): p. 484-92. 180. Pellicena, P. and J. Kuriyan, Protein–protein interactions in the allosteric regulation of protein kinases. Current Opinion in Structural Biology, 2006. 16(6): p. 702-709. 181. Klemm, J.D., S.L. Schreiber, and G.R. Crabtree, Dimerization as a regulatory mechanism in signal transduction. Annu Rev Immunol, 1998. 16: p. 569-92. 182. Pavletich, N.P., Mechanisms of cyclin-dependent kinase regulation: structures of Cdks, their cyclin activators, and Cip and INK4 inhibitors. J Mol Biol, 1999. 287(5): p. 821-8. 183. Freeman, A.K., D.A. Ritt, and D.K. Morrison, The importance of Raf dimerization in cell signaling. Small GTPases, 2013. 4(3): p. 180-185. 184. Needham, S.R., et al., Structure-function relationships and supramolecular organization of the EGFR (epidermal growth factor receptor) on the cell surface. Biochem Soc Trans, 2014. 42(1): p. 114-9. 185. Edwards, A.S. and J.D. Scott, A-kinase anchoring proteins: protein kinase A and beyond. Curr Opin Cell Biol, 2000. 12(2): p. 217-21. 186. Welch, E.J., B.W. Jones, and J.D. Scott, Networking with AKAPs: context-dependent regulation of anchored enzymes. Mol Interv, 2010. 10(2): p. 86-97. 187. Zang, M., et al., Microtubule integrity regulates Pak leading to Ras-independent activation of Raf-1. insights into mechanisms of Raf-1 activation. J Biol Chem, 2001. 276(27): p. 25157-65.

203

188. Geilen, C.C., et al., A selective inhibitor of cyclic AMP-dependent protein kinase, N- [2-bromocinnamyl(amino)ethyl]-5-isoquinolinesulfonamide (H-89), inhibits phosphatidylcholine biosynthesis in HeLa cells. FEBS Lett, 1992. 309(3): p. 381-4. 189. Lochner, A. and J.A. Moolman, The many faces of H89: a review. Cardiovasc Drug Rev, 2006. 24(3-4): p. 261-74. 190. Hirai, H., et al., MK-2206, an allosteric Akt inhibitor, enhances antitumor efficacy by standard chemotherapeutic agents or molecular targeted drugs in vitro and in vivo. Mol Cancer Ther, 2010. 9(7): p. 1956-67. 191. Nakamura, A., et al., Antitumor activity of the selective pan-RAF inhibitor TAK-632 in BRAF inhibitor-resistant melanoma. Cancer Res, 2013. 73(23): p. 7043-55. 192. Murray, B.W., et al., 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, 2010. 107(20): p. 9446-51. 193. Hayashi-Takagi, A., et al., PAKs inhibitors ameliorate schizophrenia-associated dendritic spine deterioration in vitro and in vivo during late adolescence. Proc Natl Acad Sci U S A, 2014. 111(17): p. 6461-6. 194. Chiang, C.W., et al., Protein phosphatase 2A activates the proapoptotic function of BAD in interleukin- 3-dependent lymphoid cells by a mechanism requiring 14-3-3 dissociation. Blood, 2001. 97(5): p. 1289-97. 195. Karbowski, M. and R.J. Youle, Dynamics of mitochondrial morphology in healthy cells and during apoptosis. Cell Death Differ, 2003. 10(8): p. 870-80. 196. Wang, R. and M.G. Brattain, The maximal size of protein to diffuse through the nuclear pore is larger than 60 kDa. FEBS Letters, 2007. 581(17): p. 3164- 3170. 197. Adachi, M. and K. Imai, The proapoptotic BH3-only protein BAD transduces cell death signals independently of its interaction with Bcl-2. Cell Death Differ, 2002. 9(11): p. 1240-7. 198. Hann, C., et al., Increased camptothecin toxicity induced in mammalian cells expressing Saccharomyces cerevisiae DNA topoisomerase I. J Biol Chem, 1998. 273(14): p. 8425-33. 199. Liu, L.F., et al., Mechanism of action of camptothecin. Ann N Y Acad Sci, 2000. 922: p. 1-10. 200. Li, H.L., et al., Phosphorylation of tau antagonizes apoptosis by stabilizing beta- catenin, a mechanism involved in Alzheimer's neurodegeneration. Proc Natl Acad Sci U S A, 2007. 104(9): p. 3591-6. 201. Xiang, X., et al., Phosphorylation of 338SSYY341 regulates specific interaction between Raf-1 and MEK1. J Biol Chem, 2002. 277(47): p. 44996-5003.

204

202. Edin, M.L. and R.L. Juliano, Raf-1 Serine 338 Phosphorylation Plays a Key Role in Adhesion-Dependent Activation of Extracellular Signal-Regulated Kinase by Epidermal Growth Factor. Mol Cell Biol, 2005. 25(11): p. 4466-75. 203. King, A.J., et al., The protein kinase Pak3 positively regulates Raf-1 activity through phosphorylation of serine 338. Nature, 1998. 396(6707): p. 180-3. 204. Sun, H., et al., Regulation of the protein kinase Raf-1 by oncogenic Ras through phosphatidylinositol 3-kinase, Cdc42/Rac and Pak. Curr Biol, 2000. 10(5): p. 281-4. 205. Chaudhary, A., et al., Phosphatidylinositol 3-kinase regulates Raf1 through Pak phosphorylation of serine 338. Curr Biol, 2000. 10(9): p. 551-4. 206. Wang, H.G. and J.C. Reed, Bc1-2, Raf-1 and mitochondrial regulation of apoptosis. Biofactors, 1998. 8(1-2): p. 13-6. 207. Maru, Y., et al., The dimerization property of glutathione S-transferase partially reactivates Bcr-Abl lacking the oligomerization domain. J Biol Chem, 1996. 271(26): p. 15353-7. 208. Krupp, M., et al., CellLineNavigator: a workbench for cancer cell line analysis. Nucleic Acids Research, 2013. 41(D1): p. D942-D948. 209. Fang, Z.P., et al., P21-activated kinase 5 plays essential roles in the proliferation and tumorigenicity of human hepatocellular carcinoma. Acta Pharmacol Sin, 2014. 35(1): p. 82-8. 210. Gaush, C.R., W.L. Hard, and T.F. Smith, Characterization of an established line of canine kidney cells (MDCK). Proc Soc Exp Biol Med, 1966. 122(3): p. 931-5. 211. Dukes, J.D., P. Whitley, and A.D. Chalmers, The MDCK variety pack: choosing the right strain. BMC Cell Biol, 2011. 12: p. 43. 212. Panté, N. and M. Kann, Nuclear Pore Complex Is Able to Transport Macromolecules with Diameters of ∼39 nm. Mol Biol Cell, 2002. 13(2): p. 425-34. 213. Harada, H., et al., Phosphorylation and inactivation of BAD by mitochondria- anchored protein kinase A. Mol Cell, 1999. 3(4): p. 413-22. 214. Chiloeches, A., C.S. Mason, and R. Marais, S338 phosphorylation of Raf-1 is independent of phosphatidylinositol 3-kinase and Pak3. Mol Cell Biol, 2001. 21(7): p. 2423-34. 215. Menzel, N., et al., The Drosophila p21-activated kinase Mbt modulates DE-cadherin- mediated cell adhesion by phosphorylation of Armadillo. Biochem J, 2008. 416(2): p. 231-41. 216. Lozano, E., et al., PAK is required for the disruption of E-cadherin adhesion by the small GTPase Rac. J Cell Sci, 2008. 121(Pt 7): p. 933-8. 217. Davis, M.A., R.C. Ireton, and A.B. Reynolds, A core function for p120-catenin in cadherin turnover. J Cell Biol, 2003. 163(3): p. 525-34.

205

218. Macpherson, I.R., et al., p120-catenin is required for the collective invasion of squamous cell carcinoma cells via a phosphorylation-independent mechanism. Oncogene, 2007. 26(36): p. 5214-28. 219. Yanagisawa, M. and P.Z. Anastasiadis, p120 catenin is essential for mesenchymal cadherin-mediated regulation of cell motility and invasiveness. J Cell Biol, 2006. 174(7): p. 1087-96. 220. Elia, L.P., et al., p120 catenin regulates dendritic spine and synapse development through Rho-family GTPases and cadherins. Neuron, 2006. 51(1): p. 43-56. 221. Wang, Q., et al., Molecular mechanism of membrane constriction and tubulation mediated by the F-BAR protein Pacsin/Syndapin. Proc Natl Acad Sci U S A, 2009. 106(31): p. 12700-5. 222. Qualmann, B., et al., Syndapin I, a synaptic dynamin-binding protein that associates with the neural Wiskott-Aldrich syndrome protein. Mol Biol Cell, 1999. 10(2): p. 501-13. 223. Ewing, R.M., et al., Large-scale mapping of human protein–protein interactions by mass spectrometry. Mol Syst Biol, 2007. 3: p. 89. 224. Rolland, T., et al., A proteome-scale map of the human interactome network. Cell, 2014. 159(5): p. 1212-26. 225. Xiao, K., et al., Functional specialization of beta-arrestin interactions revealed by proteomic analysis. Proc Natl Acad Sci U S A, 2007. 104(29): p. 12011-6. 226. Bandyopadhyay, S., et al., A human MAP kinase interactome. Nat Methods, 2010. 7(10): p. 801-5. 227. Rual, J.F., et al., Towards a proteome-scale map of the human protein-protein interaction network. Nature, 2005. 437(7062): p. 1173-8. 228. Huttlin, E.L., et al., The BioPlex Network: A Systematic Exploration of the Human Interactome. Cell, 2015. 162(2): p. 425-40. 229. Keller, K.E., et al., SAICAR induces protein kinase activity of PKM2 that is necessary for sustained proliferative signaling of cancer cells. Mol Cell, 2014. 53(5): p. 700-9. 230. Venkatesan, K., et al., An empirical framework for binary interactome mapping. Nat Methods, 2009. 6(1): p. 83-90. 231. Ngounou Wetie, A.G., et al., Investigation of stable and transient protein-protein interactions: Past, present, and future. Proteomics, 2013. 13(3-4): p. 538-57. 232. Fields, S. and O. Song, A novel genetic system to detect protein-protein interactions. Nature, 1989. 340(6230): p. 245-6. 233. Van Criekinge, W. and R. Beyaert, Yeast Two-Hybrid: State of the Art. Biol Proced Online, 1999. 2: p. 1-38.

206

234. Berggard, T., S. Linse, and P. James, Methods for the detection and analysis of protein-protein interactions. Proteomics, 2007. 7(16): p. 2833-42. 235. Deane, C.M., et al., Protein interactions: two methods for assessment of the reliability of high throughput observations. Mol Cell Proteomics, 2002. 1(5): p. 349- 56. 236. Fetchko, M. and I. Stagljar, Application of the split-ubiquitin membrane yeast two- hybrid system to investigate membrane protein interactions. Methods, 2004. 32(4): p. 349-62. 237. Miteva, Y.V., H.G. Budayeva, and I.M. Cristea, Proteomics-based methods for discovery, quantification, and validation of protein-protein interactions. Anal Chem, 2013. 85(2): p. 749-68. 238. Dubois, T., et al., Structure and sites of phosphorylation of 14-3-3 protein: role in coordinating signal transduction pathways. J Protein Chem, 1997. 16(5): p. 513-22. 239. Rigaut, G., et al., A generic protein purification method for protein complex characterization and proteome exploration. Nat Biotechnol, 1999. 17(10): p. 1030-2. 240. Tang, X. and J.E. Bruce, A new cross-linking strategy: protein interaction reporter (PIR) technology for protein-protein interaction studies. Mol Biosyst, 2010. 6(6): p. 939-47. 241. Kluger, R. and A. Alagic, Chemical cross-linking and protein-protein interactions-a review with illustrative protocols. Bioorg Chem, 2004. 32(6): p. 451-72. 242. Roux, K.J., et al., A promiscuous biotin ligase fusion protein identifies proximal and interacting proteins in mammalian cells. J Cell Biol, 2012. 196(6): p. 801-10. 243. Choi-Rhee, E., H. Schulman, and J.E. Cronan, Promiscuous protein biotinylation by Escherichia coli biotin protein ligase. Protein Sci, 2004. 13(11): p. 3043-50. 244. Bayer, E.A., S. Ehrlich-Rogozinski, and M. Wilchek, Sodium dodecyl sulfate- polyacrylamide gel electrophoretic method for assessing the quaternary state and comparative thermostability of avidin and streptavidin. Electrophoresis, 1996. 17(8): p. 1319-24. 245. Ong, S.E., et al., Stable isotope labeling by amino acids in cell culture, SILAC, as a simple and accurate approach to expression proteomics. Mol Cell Proteomics, 2002. 1(5): p. 376-86. 246. Zhu, H., et al., Amino acid residue specific stable isotope labeling for quantitative proteomics. Rapid Commun Mass Spectrom, 2002. 16(22): p. 2115-23. 247. Ong, S.E., The expanding field of SILAC. Anal Bioanal Chem, 2012. 404(4): p. 967- 76. 248. Mann, M., Functional and quantitative proteomics using SILAC. Nat Rev Mol Cell Biol, 2006. 7(12): p. 952-8.

207

249. Mittler, G., F. Butter, and M. Mann, A SILAC-based DNA protein interaction screen that identifies candidate binding proteins to functional DNA elements. Genome Res, 2009. 19(2): p. 284-93. 250. Butter, F., et al., Unbiased RNA-protein interaction screen by quantitative proteomics. Proc Natl Acad Sci U S A, 2009. 106(26): p. 10626-31. 251. Ong, S.E., et al., Identifying the proteins to which small-molecule probes and drugs bind in cells. Proc Natl Acad Sci U S A, 2009. 106(12): p. 4617-22. 252. Zhou, Q., et al., Screening for Therapeutic Targets of Vorinostat by SILAC-based Proteomic Analysis in Human Breast Cancer Cells. Proteomics, 2010. 10(5): p. 1029- 39. 253. Blagoev, B., et al., Temporal analysis of phosphotyrosine-dependent signaling networks by quantitative proteomics. Nat Biotechnol, 2004. 22(9): p. 1139-45. 254. Gronborg, M., et al., Biomarker discovery from pancreatic cancer secretome using a differential proteomic approach. Mol Cell Proteomics, 2006. 5(1): p. 157-71. 255. Hwang, S.I., et al., Systematic characterization of nuclear proteome during apoptosis: a quantitative proteomic study by differential extraction and stable isotope labeling. Mol Cell Proteomics, 2006. 5(6): p. 1131-45. 256. Amanchy, R., D.E. Kalume, and A. Pandey, Stable isotope labeling with amino acids in cell culture (SILAC) for studying dynamics of protein abundance and posttranslational modifications. Sci STKE, 2005. 2005(267): p. pl2. 257. Olsen, J.V., et al., Global, in vivo, and site-specific phosphorylation dynamics in signaling networks. Cell, 2006. 127(3): p. 635-48. 258. Martin, B.R., et al., Global profiling of dynamic protein palmitoylation. Nat Methods, 2012. 9(1): p. 84-9. 259. Wagner, S.A., et al., A proteome-wide, quantitative survey of in vivo ubiquitylation sites reveals widespread regulatory roles. Mol Cell Proteomics, 2011. 10(10): p. M111.013284. 260. Sweet, S.M., et al., Kinetics of re-establishing H3K79 methylation marks in global human chromatin. J Biol Chem, 2010. 285(43): p. 32778-86. 261. Steen, H. and M. Mann, The ABC's (and XYZ's) of peptide sequencing. Nat Rev Mol Cell Biol, 2004. 5(9): p. 699-711. 262. Mandai, K., et al., Afadin/AF-6 and canoe: roles in cell adhesion and beyond. Prog Mol Biol Transl Sci, 2013. 116: p. 433-54. 263. Ogita, H., et al., Cell adhesion molecules nectins and associating proteins: Implications for physiology and pathology. Proc Jpn Acad Ser B Phys Biol Sci, 2010. 86(6): p. 621-9.

208

264. Samanta, D. and S.C. Almo, Nectin family of cell-adhesion molecules: structural and molecular aspects of function and specificity. Cell Mol Life Sci, 2014. 265. Tachibana, K., et al., Two cell adhesion molecules, nectin and cadherin, interact through their cytoplasmic domain-associated proteins. J Cell Biol, 2000. 150(5): p. 1161-76. 266. Hoshino, T., et al., Regulation of E-cadherin endocytosis by nectin through afadin, Rap1, and p120ctn. J Biol Chem, 2005. 280(25): p. 24095-103. 267. Birukova, A.A., et al., Afadin controls p120-catenin-ZO-1 interactions leading to endothelial barrier enhancement by oxidized phospholipids. J Cell Physiol, 2012. 227(5): p. 1883-90. 268. Laura, R.P., et al., The Erbin PDZ domain binds with high affinity and specificity to the carboxyl termini of delta-catenin and ARVCF. J Biol Chem, 2002. 277(15): p. 12906-14. 269. Donoviel, D.B., et al., Proteinuria and perinatal lethality in mice lacking NEPH1, a novel protein with homology to NEPHRIN. Mol Cell Biol, 2001. 21(14): p. 4829-36. 270. Huber, T.B., et al., The carboxyl terminus of Neph family members binds to the PDZ domain protein zonula occludens-1. J Biol Chem, 2003. 278(15): p. 13417-21. 271. Gerke, P., et al., Neuronal expression and interaction with the synaptic protein CASK suggest a role for Neph1 and Neph2 in synaptogenesis. J Comp Neurol, 2006. 498(4): p. 466-75. 272. Klejnot, M. and F. Kozielski, Structural insights into human Kif7, a kinesin involved in Hedgehog signalling. Acta Crystallogr D Biol Crystallogr, 2012. 68(Pt 2): p. 154- 9. 273. Cheung, H.O., et al., The kinesin protein Kif7 is a critical regulator of Gli transcription factors in mammalian hedgehog signaling. Sci Signal, 2009. 2(76): p. ra29. 274. Dafinger, C., et al., Mutations in KIF7 link Joubert syndrome with Sonic Hedgehog signaling and microtubule dynamics. J Clin Invest, 2011. 121(7): p. 2662-7. 275. Wehland, J., et al., Role of microtubules in the distribution of the Golgi apparatus: effect of taxol and microinjected anti-alpha-tubulin antibodies. Proc Natl Acad Sci U S A, 1983. 80(14): p. 4286-90. 276. Zigmond, S.H., Formin-induced nucleation of actin filaments. Curr Opin Cell Biol, 2004. 16(1): p. 99-105. 277. Rothbauer, U., et al., A versatile nanotrap for biochemical and functional studies with fluorescent fusion proteins. Mol Cell Proteomics, 2008. 7(2): p. 282-9. 278. von Mering, C., et al., STRING: a database of predicted functional associations between proteins. Nucleic Acids Res, 2003. 31(1): p. 258-61.

209

279. Tinti, M., et al., ANIA: ANnotation and Integrated Analysis of the 14-3-3 interactome. Database (Oxford), 2014. 2014: p. bat085. 280. Kim, J.W., et al., FAM83H mutations in families with autosomal-dominant hypocalcified amelogenesis imperfecta. Am J Hum Genet, 2008. 82(2): p. 489-94. 281. Zhang, C., Y. Song, and Z. Bian, Ultrastructural analysis of the teeth affected by amelogenesis imperfecta resulting from FAM83H mutations and review of the literature. Oral Surg Oral Med Oral Pathol Oral Radiol, 2015. 119(2): p. e69-76. 282. Sasaroli, D., et al., Novel surface targets and serum biomarkers from the ovarian cancer vasculature. Cancer Biol Ther, 2011. 12(3): p. 169-80. 283. Cipriano, R., et al., Conserved oncogenic behavior of the FAM83 family regulates MAPK signaling in human cancer. Mol Cancer Res, 2014. 12(8): p. 1156-65. 284. Barker, P.A. and A. Salehi, The MAGE proteins: emerging roles in cell cycle progression, apoptosis, and neurogenetic disease. J Neurosci Res, 2002. 67(6): p. 705-12. 285. Xue, B., et al., Human NRAGE disrupts E-cadherin/beta-catenin regulated homotypic cell-cell adhesion. Biochem Biophys Res Commun, 2005. 336(1): p. 247-51. 286. Masuda, Y., et al., Dlxin-1, a novel protein that binds Dlx5 and regulates its transcriptional function. J Biol Chem, 2001. 276(7): p. 5331-8. 287. Mouri, A., et al., MAGE-D1 regulates expression of depression-like behavior through serotonin transporter ubiquitylation. J Neurosci, 2012. 32(13): p. 4562-80. 288. Salehi, A.H., et al., NRAGE, a novel MAGE protein, interacts with the p75 neurotrophin receptor and facilitates nerve growth factor-dependent apoptosis. Neuron, 2000. 27(2): p. 279-88. 289. Boer, J.M., et al., Identification and classification of differentially expressed genes in renal cell carcinoma by expression profiling on a global human 31,500-element cDNA array. Genome Res, 2001. 11(11): p. 1861-70. 290. Bhattacharjee, A., et al., Classification of human lung carcinomas by mRNA expression profiling reveals distinct adenocarcinoma subclasses. Proc Natl Acad Sci U S A, 2001. 98(24): p. 13790-5. 291. Kumar, S., et al., A pathway for the control of anoikis sensitivity by E-cadherin and epithelial-to-mesenchymal transition. Mol Cell Biol, 2011. 31(19): p. 4036-51. 292. Bertrand, M., et al., Comparative expression analysis of the MAGED genes during embryogenesis and brain development. Dev Dyn, 2004. 230(2): p. 325-34. 293. Salehi, A.H., S. Xanthoudakis, and P.A. Barker, NRAGE, a p75 neurotrophin receptor-interacting protein, induces caspase activation and cell death through a JNK-dependent mitochondrial pathway. J Biol Chem, 2002. 277(50): p. 48043-50.

210

294. Budnik, A. and D.J. Stephens, ER exit sites--localization and control of COPII vesicle formation. FEBS Lett, 2009. 583(23): p. 3796-803. 295. Tillmann, K.D., et al., Regulation of Sec16 levels and dynamics links proliferation and secretion. J Cell Sci, 2014. 296. Coleman, P.F., D.P. Suttle, and G.R. Stark, Purification from hamster cells of the multifunctional protein that initiates de novo synthesis of pyrimidine nucleotides. J Biol Chem, 1977. 252(18): p. 6379-85. 297. Kotsis, D.H., et al., Protein kinase A phosphorylation of the multifunctional protein CAD antagonizes activation by the MAP kinase cascade. Mol Cell Biochem, 2007. 301(1-2): p. 69-81. 298. Sigoillot, F.D., S.M. Sigoillot, and H.I. Guy, Breakdown of the regulatory control of pyrimidine biosynthesis in human breast cancer cells. Int J Cancer, 2004. 109(4): p. 491-8. 299. Reardon, M.A., J.E. Dixon, and G. Weber, Increased messenger RNA concentration for carbamoyl-phosphate synthase II in hepatoma 3924A. Biochem Biophys Res Commun, 1987. 147(1): p. 494-500. 300. Morin, A., et al., Identification of CAD as an androgen receptor interactant and an early marker of prostate tumor recurrence. Faseb j, 2012. 26(1): p. 460-7. 301. Baffa, R., et al., Fez1/Lzts1-deficient mice are more susceptible to N-butyl-N-(4- hydroxybutil) nitrosamine (BBN) carcinogenesis. Carcinogenesis, 2008. 29(4): p. 846-8. 302. Konno, D., et al., The postsynaptic density and dendritic raft localization of PSD- Zip70, which contains an N-myristoylation sequence and leucine-zipper motifs. J Cell Sci, 2002. 115(Pt 23): p. 4695-706. 303. Ooshio, T., et al., Involvement of the interaction of afadin with ZO-1 in the formation of tight junctions in Madin-Darby canine kidney cells. J Biol Chem, 2010. 285(7): p. 5003-12. 304. Wang, H.G., U.R. Rapp, and J.C. Reed, Bcl-2 targets the protein kinase Raf-1 to mitochondria. Cell, 1996. 87(4): p. 629-38. 305. Ehrenreiter, K., et al., Raf-1 regulates Rho signaling and cell migration. J Cell Biol, 2005. 168(6): p. 955-64. 306. Macdonald, S.G., et al., Reconstitution of the Raf-1-MEK-ERK signal transduction pathway in vitro. Mol Cell Biol, 1993. 13(11): p. 6615-20. 307. Baccarini, M., An old kinase on a new path: Raf and apoptosis. Cell Death Differ, 2002. 9(8): p. 783-5. 308. Van Itallie, C.M., et al., The N and C termini of ZO-1 are surrounded by distinct proteins and functional protein networks. J Biol Chem, 2013. 288(19): p. 13775-88.

211

309. Gonzalez-Mariscal, L., et al., Molecular characterization of the tight junction protein ZO-1 in MDCK cells. Exp Cell Res, 1999. 248(1): p. 97-109. 310. Wang, Y., et al., ASPP2 controls epithelial plasticity and inhibits metastasis through beta-catenin-dependent regulation of ZEB1. Nat Cell Biol, 2014. 16(11): p. 1092- 104. 311. Astro, V., et al., Liprin-alpha1, ERC1 and LL5 define polarized and dynamic structures that are implicated in cell migration. J Cell Sci, 2014. 127(Pt 17): p. 3862- 76. 312. Bergamaschi, D., et al., ASPP1 and ASPP2: common activators of p53 family members. Mol Cell Biol, 2004. 24(3): p. 1341-50. 313. Reichert, A.S. and W. Neupert, Contact sites between the outer and inner membrane of mitochondria-role in protein transport. Biochim Biophys Acta, 2002. 1592(1): p. 41-9. 314. Chen, C., A. Wirth, and E. Ponimaskin, Cdc42: an important regulator of neuronal morphology. Int J Biochem Cell Biol, 2012. 44(3): p. 447-51. 315. Johnson, J.L., J.W. Erickson, and R.A. Cerione, New insights into how the Rho guanine nucleotide dissociation inhibitor regulates the interaction of Cdc42 with membranes. J Biol Chem, 2009. 284(35): p. 23860-71. 316. Reddy, E.M., et al., Dlxin-1, a MAGE family protein, induces accelerated neurite outgrowth and cell survival by enhanced and early activation of MEK and Akt signalling pathways in PC12 cells. Exp Cell Res, 2010. 316(14): p. 2220-36. 317. Bertrand, M.J., et al., NRAGE, a p75NTR adaptor protein, is required for developmental apoptosis in vivo. Cell Death Differ, 2008. 15(12): p. 1921-9. 318. Tao, S., et al., A genome-wide search for loci interacting with known prostate cancer risk-associated genetic variants. Carcinogenesis, 2012. 33(3): p. 598-603. 319. Dingar, D., et al., BioID identifies novel c-MYC interacting partners in cultured cells and xenograft tumors. J Proteomics, 2014. 320. Eckler, M.J. and B. Chen, Fez family transcription factors: controlling neurogenesis and cell fate in the developing mammalian nervous system. Bioessays, 2014. 36(8): p. 788-97. 321. Martin, D.D., E. Beauchamp, and L.G. Berthiaume, Post-translational myristoylation: Fat matters in cellular life and death. Biochimie, 2011. 93(1): p. 18- 31. 322. Hedo, J.A., E. Collier, and A. Watkinson, Myristyl and palmityl acylation of the receptor. J Biol Chem, 1987. 262(3): p. 954-7. 323. Zha, J., et al., Posttranslational N-myristoylation of BID as a molecular switch for targeting mitochondria and apoptosis. Science, 2000. 290(5497): p. 1761-5.

212

324. Kozak, M., An analysis of 5'-noncoding sequences from 699 vertebrate messenger RNAs. Nucleic Acids Res, 1987. 15(20): p. 8125-48. 325. Wang, Z., et al., N terminus of ASPP2 binds to Ras and enhances Ras/Raf/MEK/ERK activation to promote oncogene-induced senescence. Proc Natl Acad Sci U S A, 2013. 110(1): p. 312-7. 326. Takai, Y. and H. Nakanishi, Nectin and afadin: novel organizers of intercellular junctions. J Cell Sci, 2003. 116(Pt 1): p. 17-27. 327. Shimono, Y., et al., Immunoglobulin superfamily receptors and adherens junctions. Subcell Biochem, 2012. 60: p. 137-70. 328. Huang, Y.Z., et al., Erbin is a protein concentrated at postsynaptic membranes that interacts with PSD-95. J Biol Chem, 2001. 276(22): p. 19318-26. 329. Furnari, M.A., et al., Functional deficits in PAK5, PAK6 and PAK5/PAK6 knockout mice. PLoS One, 2013. 8(4): p. e61321. 330. Kleim, J.A., et al., Synapse formation is associated with memory storage in the cerebellum. Proc Natl Acad Sci U S A, 2002. 99(20): p. 13228-31.

213

Appendix

List of the buffers and media

Mammalian cell lysis buffer:

50mM Tris HCL pH 7.4-8 150mM NaCl

20mM MgCl2 0.5% Triton X-100 RIPA buffer

50mM Tris HCL pH 7.4-8 150mM NaCl 1% Triton X-100 0.5% Sodium Deoxycholate (DOC) 0.1% SDS 1mM EDTA Bacterial lysis buffer

50mM Tris HCL pH 8

0.5% Triton X-100

0.5mM MgCl2

1mg/ml Lysozyme

0.5mM PMSF Add freshly

5mM DTT

Protease inhibitor

214

CCMB buffer

11.8g/L CaCl2.2H2O

4g/L Mn.Cl2.4H2O

2g/L Mg.Cl2.6H2O pH 6.4 10ml/L1M KOAc pH 7 100ml/L Glycerol 10X SDS running buffer 0.25M Tris Base 2M Glycine 1% SDS

10X transfer buffer 0.8M Glycine 0.25M Tris Base

50X TAE buffer 2M Tris Base 0.05 M EDTA, pH 7.5-8 57.1ml Glacial Acetic Acid

2X SDS sample buffer 50mM Tris HCL pH 6.8 6% SDS 100 mM DTT 0.1% bromophenol blue 50% Glycerol 2X Kinase buffer 50mM HEPES pH 7.3 100mM KCl

20mM MgCl2 0.2% Triton X-100 2mM DTT

215

Coupling buffer 0.1 M NaHCO3 0.5 M NaCl, pH 8.3–8.5

Low pH washing buffer

0.1 M acetate buffer pH 4 0.5 M NaCl TBS buffer 24g/L Tris HCL 5.6g/L Tris Base pH 6.4 88g NaCl

SILAC DMEM: (Light)

444 ml of SILAC DMEM 50 ml of Dialyzed Fetal Bovine Serum 0.5 ml of light Arginine stock (85mg/ml) 0.5 ml of light Lysine (146 mg/mL) 5 ml of Penicillin Streptomycin

SILAC DMEM: (Heavy)

444 ml of SILAC DMEM 50 ml of Dialyzed Fetal Bovine Serum 0.5 ml of heavy Arginine (85mg/ml) 0.5 ml of heavy Lysine (146 mg/mL) 5 ml of Penicillin Streptomycin

216

Table S1. List of the proteins identified by BioID as PAK5 proximal proteins in stable cell lines expressing BirA* PAK5

Gene MW. Ratio Pep. Seq. No Protein names names kD H/L No (%) Serine/threonine-protein 1 PAK5 81 137.2 35 57 kinase PAK 5 Leucine zipper putative 2 LZTS2 73 29.7 11 23 tumor suppressor 2 Centrosomal protein of 85 3 CEP85 86 19.4 9 17 kDa 4 PVRL2 Nectin 2 51 16.6 6 15 Serine/threonine-protein 5 PAK4 64 14.1 5 8 kinase PAK 4 6 KIF7 Kinesin-like protein KIF7 151 13.8 6 5 7 MLLT4 Afadin 208 13.1 52 38 Inactive ubiquitin carboxyl- 8 USP54 187 12.0 2 2 terminal hydrolase 54 9 TUBA3C Tubulin alpha-3C/D chain 50 11.4 12 42 10 FAM83H Protein FAM83H 127 11.3 13 19 Melanoma-associated 11 MAGED1 92 10.9 7 11 antigen D1 Trinucleotide repeat- 12 TNRC6A 210 10.8 3 2 containing gene 6A protein 13 PKP2 Plakophilin-2 97 9.8 2 3 Trinucleotide repeat- 14 TNRC6B 194 9.6 11 7 containing gene 6B protein Pleckstrin homology-like 15 PHLDB2 142 9.2 5 6 domain family B member 2 TANK-binding kinase 1- 16 TBKBP1 68 8.8 4 7 binding protein 1 17 KIRREL Neph 1 85 7.1 4 8 5-azacytidine-induced 18 AZI1 122 6.9 2 2 protein 1 PERQ amino acid-rich with 19 GIGYF1 GYF domain-containing 115 6.7 2 3 protein 1 Protein transport protein 20 SEC16A 252 6.5 8 4 Sec16A 21 DLG5 Disks large homolog 5 214 6.3 37 21 Signal-induced 22 SIPA1L1 proliferation-associated 1- 200 6.3 7 4 like protein 1 23 ODF2 Outer dense fiber protein 2 95 6.0 3 4 24 CTNND1 p120 Catenin 108 5.9 11 15 25 RPS11 40S ribosomal protein S11 18 5.8 3 17 Nucleoporin NUP188 26 NUP188 196 5.1 2 2 homolog

217

Gene MW. Ratio Pep. Seq. No Protein names names kD H/L No (%) Calcium-binding 27 SLC25A13 mitochondrial carrier 74 5.1 4 6 protein Aralar2 28 RPS15A 40S ribosomal protein S15a 15 5.1 4 31 Single-stranded DNA- 29 SSBP1 binding protein, 17 4.9 2 22 mitochondrial Apoptosis-stimulating of 30 TP53BP2 126 4.1 4 3 p53 protein 2 31 ERBB2IP Protein LAP2 159 4.1 25 23 Double-stranded RNA- 32 STAU2 binding protein Staufen 63 3.8 6 11 homolog 2 Eukaryotic translation 33 EIF3E 52 3.8 2 4 initiation factor 3 subunit E CCR4-NOT transcription 34 CNOT1 267 3.6 3 1 complex subunit 1 35 CAD CAD protein 243 3.3 14 8 Eukaryotic translation 36 EIF4ENIF1 initiation factor 4E 108 3.3 4 5 transporter DDX3X ATP-dependent RNA 37 73 3.2 20 40 helicase DDX3X 38 TJP1 Tight junction protein ZO-1 197 3.2 37 25 39 ERC2 ERC protein 2 111 3.2 3 4 Translocon-associated 40 SSR4 19 3.2 2 17 protein subunit delta 4F2 cell-surface antigen 41 SLC3A2 71 2.9 7 14 heavy chain Pleckstrin homology 42 PLEKHA5 domain-containing family A 147 2.9 3 3 member 5 ELKS/Rab6- 43 ERC1 interacting/CAST family 128 2.9 12 12 member 1 Centrosomal protein of 192 44 CEP192 279 2.9 3 1 kDa 45 RPL23 60S ribosomal protein L23 15 2.8 6 50 SKT; 46 Sickle tail protein homolog 214 2.8 8 5 KIAA1217 KN motif and ankyrin 47 KANK2 repeat domain-containing 92 2.8 6 9 protein 2 Pleckstrin homology 48 PLEKHA7 domain-containing family A 144 2.8 2 3 member 7

218

Gene MW. Ratio Pep. Seq. No Protein names names kD H/L No (%) PDZ and LIM domain 49 PDLIM7 50 2.8 2 5 protein 7 RanBP2-like and GRIP RGPD 50 domain-containing protein 198 2.7 26 18 1/2/3/4 1/2/3/4 51 RPL9 60S ribosomal protein L9 22 2.7 2 13 52 PPP1R13L RelA-associated inhibitor 89 2.6 3 4 Rho GTPase-activating 53 ARHGAP21 217 2.6 2 1 protein 21 NME1/2 Nucleoside diphosphate 54 33 2.6 2 14 kinase A/ B 55 RPL22 60S ribosomal protein L22 15 2.5 2 19 Unconventional myosin- 56 MYO18B 285 2.5 2 1 XVIIIb 57 TUBB3 Tubulin beta-3 chain 50 2.5 10 29 Uncharacterized protein 58 C19orf21 75 2.5 8 14 C19orf21 YTH domain family protein 59 YTHDF1/3 61 2.4 4 8 1/ 3 60 SERPINH1 Serpin H1 46 2.4 6 19 YTH domain family protein 61 YTHDF2 62 2.4 3 5 2 62 PALLD Palladin 151 2.4 14 12 DNA-dependent protein 63 PRKDC 469 2.3 24 6 kinase catalytic subunit 64 SCRIB Protein scribble homolog 178 2.3 10 8 65 TNS3 Tensin-3 155 2.3 3 3 Coatomer subunit gamma- 66 COPG1 98 2.3 5 8 1 67 RPL31 60S ribosomal protein L31 14 2.3 2 18 ATP-dependent RNA 68 DDX39A 49 2.3 5 14 helicase DDX39A 69 TUBB2B Tubulin beta-2B chain 50 2.2 14 43 60S acidic ribosomal 70 RPLP2 12 2.2 2 27 protein P2 RPS27A; Ubiquitin-40S ribosomal 71 18 2.2 4 28 protein S27a;

E3 SUMO-protein ligase 72 RANBP2 RanBP2;Putative peptidyl- 358 2.2 76 32 prolyl cis-trans isomerase DNA replication licensing 73 MCM7 81 2.2 9 18 factor MCM7 Interleukin-1 receptor- 74 IRAK1 77 2.1 2 3 associated kinase 1

219

Gene MW. Ratio Pep. Seq. No Protein names names kD H/L No (%) mRNA-decapping enzyme 75 DCP1B 68 2.1 6 13 1B LIM domain-containing 76 LIMD1 72 2.1 4 8 protein 1 77 RPS19 40S ribosomal protein S19 16 2.1 5 34 Delta-1-pyrroline-5- 78 ALDH18A1 87 2.1 3 5 carboxylate synthase 79 TUBB4A Tubulin beta-4A chain 50 2.1 16 54 80 RPS14 40S ribosomal protein S14 16 2.1 3 28 81 TUBB6 Tubulin beta-6 chain 50 2.0 9 28 Eukaryotic translation 82 EIF5 49 2.0 8 23 initiation factor 5 83 RPS16 40S ribosomal protein S16 16 2.0 3 20 Elongation factor Tu, 84 TUFM 50 2.0 2 6 mitochondrial 85 TUBB Tubulin beta chain 50 2.0 19 61 86 TUBB4B Tubulin beta-4B chain 50 2.0 19 61

220

Table S2. List of the Potential PAK5 interacting partners identified by GFP-PAK5 pull down

Gene MW. Ratio Pep. Seq. No. Protein names names kD H/L No (%) Serine/threonine-protein 1 PAK7 81 64.9 37 57 kinase PAK 7 Prolyl 4-hydroxylase subunit 2 P4HA2 61 29.6 16 39 alpha-2 Prolyl 4-hydroxylase subunit 3 P4HA1 61 15.3 16 38 alpha-1 BAG family molecular 4 BAG2 24 11.5 12 60 chaperone regulator 2 CAD protein;Glutamine- dependent carbamoyl- 5 CAD phosphate synthase;Aspartate 243 11.5 68 41 carbamoyltransferase;Dihydro orotase 6 SEC13 Protein SEC13 homolog 41 11.3 5 17 Protein transport protein 7 SEC16A 252 11.1 22 12 Sec16A 8 P4HB Protein disulfide-isomerase 57 10.9 18 38 9 AKAP8L A-kinase anchor protein 8-like 72 10.8 3 5 10 FAM83H Protein FAM83H 127 10.0 19 23 Cell division control protein 42 11 CDC42 21 10.0 4 23 homolog ATPase family AAA domain- 12 ATAD3A 66 9.0 7 12 containing protein 3A Apoptosis-inducing factor 1, 13 AIFM1 67 8.9 10 19 mitochondrial KN motif and ankyrin repeat 14 KANK2 92 8.3 4 5 domain-containing protein 2 15 LGALS3BP Galectin-3-binding protein 65 8.2 3 7 T-complex protein 1 subunit 16 TCP1 60 7.4 21 44 alpha Prolow-density lipoprotein 17 LRP1 505 7.4 16 4 receptor-related protein 1; T-complex protein 1 subunit 18 CCT7 59 7.4 10 23 eta T-complex protein 1 subunit 19 CCT3 61 7.1 16 32 gamma T-complex protein 1 subunit 20 CCT4 58 7.1 17 40 delta Microtubule-associated 21 MAPRE1 protein RP/EB family member 30 7.0 2 14 1 T-complex protein 1 subunit 22 CCT5 60 6.9 16 39 epsilon

221

Gene MW. Ratio Pep. Seq. No. Protein names names kD H/L No (%) 23 PKP2 Plakophilin-2 97 6.9 6 8 T-complex protein 1 subunit 24 CCT6A 58 6.8 12 30 zeta T-complex protein 1 subunit 25 CCT2 57 6.8 11 29 beta 26 YWHAH 14-3-3 protein eta 28 6.6 6 26 14-3-3 protein gamma;14-3-3 27 YWHAG protein gamma, N-terminally 28 5.9 10 37 processed T-complex protein 1 subunit 28 CCT8 60 5.9 19 39 theta 29 RUVBL1 RuvB-like 1 50 5.8 3 8 Ras GTPase-activating-like 30 IQGAP3 185 5.4 2 1 protein IQGAP3 E3 ubiquitin-protein ligase 31 UBR5 309 5.2 13 7 UBR5 Serine/threonine-protein 32 PGAM5 phosphatase PGAM5, 32 5.2 12 40 mitochondrial DNA-directed RNA polymerase II subunit 33 POLR2B 134 4.9 5 4 RPB2;DNA-directed RNA polymerase 34 EMD Emerin 29 4.9 10 46 35 YWHAE 14-3-3 protein epsilon 29 4.8 12 48 36 FLNC Filamin-C 291 4.8 53 27 Melanoma-associated antigen 37 MAGED1 92 4.6 5 7 D1 38 LEPRE1 Prolyl 3-hydroxylase 1 91 4.5 5 8 Probable ubiquitin carboxyl- USP9X; terminal hydrolase FAF- 39 292 4.5 8 4 USP9Y X;Probable ubiquitin carboxyl- terminal hydrolase FAF-Y 40 LOXL4 Lysyl oxidase homolog 4 84 4.3 5 8 41 RUVBL2 RuvB-like 2 51 4.3 5 12 Heat shock cognate 71 kDa 42 HSPA8 71 4.1 34 60 protein 43 FLNB Filamin-B 282 3.6 5 2 Rapamycin-insensitive 44 RICTOR 195 3.4 2 1 companion of mTOR Procollagen-lysine,2- 45 PLOD1 88 3.4 5 8 oxoglutarate 5-dioxygenase 1 Coiled-coil-helix-coiled-coil- 46 CHCHD3 helix domain-containing 27 3.4 5 24 protein 3, mitochondrial Heat shock 70 kDa protein 47 HSPA1A 70 3.3 21 37 1A/1B Mitochondrial inner membrane 48 IMMT 84 3.3 11 17 protein

222

Gene MW. Ratio Pep. Seq. No. Protein names names kD H/L No (%) NADH dehydrogenase 49 NDUFAF4 [ubiquinone] 1 alpha 20 3.2 2 10 subcomplex assembly factor 4 50 HAX1 HCLS1-associated protein X-1 32 3.2 3 12 51 MYO1C Unconventional myosin-Ic 120 3.1 12 14 52 PHB2 Prohibitin-2 33 3.1 9 35 53 HSPA6 Heat shock 70 kDa protein 6 71 3.1 11 17 14-3-3 protein beta/alpha;14- 54 YWHAB 3-3 protein beta/alpha, N- 28 3.1 9 46 terminally processed Cytochrome c1, heme protein, 55 CYC1 35 3.0 5 15 mitochondrial Heterogeneous nuclear 56 HNRNPM 78 3.0 8 15 ribonucleoprotein M Serine/threonine-protein 57 PPP6C phosphatase 6 catalytic 39 2.9 4 14 subunit 58 PHB Prohibitin 30 2.8 8 32 DnaJ homolog subfamily B 59 DNAJB6 36 2.8 5 15 member 6 DNA mismatch repair protein 60 MSH6 153 2.7 14 13 Msh6 61 YWHAQ 14-3-3 protein theta 28 2.6 12 42 Mitochondrial import inner 62 TIMM21 membrane translocase 28 2.6 4 17 subunit Tim21 63 MYO1B Unconventional myosin-Ib 132 2.6 4 5 64 FUS RNA-binding protein FUS 53 2.5 2 6 DNA-dependent protein 65 PRKDC 469 2.5 70 20 kinase catalytic subunit DnaJ homolog subfamily A 66 DNAJA1 45 2.5 2 7 member 1 Calcium-binding mitochondrial 67 SLC25A13 74 2.3 11 19 carrier protein Aralar2 Dolichyl- diphosphooligosaccharide-- 68 RPN2 69 2.3 3 7 protein glycosyltransferase subunit 2 69 YWHAZ 14-3-3 protein zeta/delta 28 2.3 14 61 70 CRYAB Alpha-crystallin B chain 20 2.2 4 23 Mitochondrial import inner 71 TIMM50 membrane translocase 50 2.2 4 12 subunit TIM50 Cytochrome b-c1 complex 72 UQCRQ 10 2.1 3 28 subunit 8 Mitochondrial dicarboxylate 73 SLC25A10 32 2.1 9 31 carrier

223