דוקטור לפילוסופיה Doctor of Philosophy מוגשת למועצה המדעית של Submitted to the Scientific Council of the מכון ויצמן למדע Weizmann Institute of Science

עבודת גמר (תזה) לתואר Thesis for the degree דוקטור לפילוסופיה Doctor of Philosophy מוגשת למועצה המדעית של Submitted to the Scientific Council of the מכון ויצמן למדע Weizmann Institute of Science

מאת By יובל גלעד Yuval Gilad

ניטור אינטראקציות חלבון-חלבון ברשת המוות התאי המתוכנת

Monitoring Protein-Protein Interactions within the Programmed Cell Death Network

מנחה: :Advisor פרופ' עדי קמחי Prof. Adi Kimchi

תשרי תשע"ו September 2015

Table of Contents

Abstract ...... 3 4 ...... תקציר Introduction ...... 5 Results ...... 7

Global Screen of the PCD Network ...... 7 Validation of GLuc PCA Reporters for Known Protein-Protein Interactions: Follow-up of Specificity, Reversibility, and the Dynamics of the Luminescent Signal16 Previously Unidentified Interactions Detected in the GLuc PCA Screen ...... 19 Focusing on the DAPK2/14-3-3τ Interaction: DAPK2 Interacts with 14-3-3τ via a Ser/Thr-Rich Stretch Mapped to the Former’s C Terminus ...... 21

Measuring the Functional Outcome of the Interaction: 14-3-3τ Inhibits DAPK2’s Activity both in Cells and in Cell-free Systems ...... 26 High-throughput siRNA kinome screen to identify modifiers of the DAPK2- 14-3-3 interaction ...... 28

ULK1/WIPI2b interaction ...... 34

ULK1 phosphorylates WIPI2b in vitro and in cells...... 35 Mapping the phosphorylation sites on WIPI2b ...... 37 Deciphering the functional role of the phosphorylation ...... 38

Utilizing the PCA reporters as a screening system for drug discovery ...... 44

Autophagy inhibition for cancer therapy ...... 45 Atg5/Atg16 interaction ...... 46

Discussion ...... 48 Materials and Methods ...... 52

Bioluminescence Assay and Analysis ...... 52 ELISA Kinase Assay ...... 52 Blebbing Assay ...... 53 DNA Constructs ...... 53

Cell Culture and Induction of Cell Death ...... 53 Protein Analysis ...... 54 Immunoprecipitation ...... 54 siRNA Screen...... 55 PIP Binding Assay ...... 55 Statistical Analysis ...... 55 Protein Interaction Maps ...... 55

References ...... 56

Abstract

Apoptosis and autophagy are distinct biological processes, each driven by a different set of protein-protein interactions, with significant cross-talk via direct interactions among apoptotic and autophagic proteins. To measure the global profile of these interactions, we adapted the Gaussia luciferase Protein-fragment Complementation Assay (GLuc PCA), which monitors binding between proteins fused to complementary fragments of a luciferase reporter. A library encompassing 63 apoptotic and autophagic proteins was constructed for the analysis of ~3600 protein-pair combinations. This generated a detailed landscape of the apoptotic and autophagic modules and points-of- interface between them, identifying 46 previously unknown interactions. One of these interactions, between DAPK2, a Ser/Thr kinase that promotes autophagy, and 14-3-3, was further investigated. We mapped the region responsible for 14-3-3binding and proved that this interaction inhibits DAPK2 dimerization and its biological and biochemical activities. This proof-of-concept underscores the power of the GLuc PCA platform for the discovery of biochemical pathways within the cell death network. In addition, we discovered by this screen a novel interaction between two core machinery proteins in autophagy; the Ser/Thr kinase ULK1 and WIPI2. We discovered that ULK1 phosphorylates WIPI2 both in vitro and in cells and mapped several putative phosphorylation sites. Functional analysis of the phosphorylation on Ser185 revealed that it abrogates the ability of WIPI2 to bind PI(3)P, an essential function for its role in autophagy. This documents for the first time a direct interaction between an upstream kinase essential for autophagy and a scaffold protein that recruits some of the critical Atg proteins at the growing autophagic membranes. We believe that the understanding of the functional role of WIPI2 phosphorylation will uncover a key step in the first stages of autophagosome biogenesis.

תקציר אפופטוזיס ואוטופאגיה הם שני תהליכים ביולוגיים נפרדים, אשר כל אחד מהם מבוצע ומבוקר על ידי מערכת מוגדרת של אינטראקציות חלבון-חלבון, עם רמה משמעותית של קישוריות ביניהם. תהליכים אלו פועלים במסגרת רשת העברת אותות מולקולרית הנקראת "רשת המוות התאי המתוכנת" ושולטת בבקרה וביצוע של תהליכי מוות תאי. על מנת למדוד את אינטראקציות החלבון -חלבון ברשת המוות התאי המתוכנת, אימצנו טכנולוגיה בשם (Protein-Fragment Complementation Assay (PCA המשמשת למדידת אינטראקציות בין חלבונים המאוחים לחצאים משלימים של אנזים מדווח. על מנת למפות את רשת האינטראקציות בתוך תהליכי האוטופאגיה והאפופטוזיס וביניהם, יצרנו ספריה של 63 חלבונים המעורבים בתהליכים אלו וסרקנו כ3600- קומבינציות של אינטראקציות ביניהם. כתוצאה מכך, זיהינו 46 אינטראקציות חדשות שלא תוארו עד כה. מתוכן, האינטראקציה בין הקינאז DAPK2 לבין 14-3-3, נחקרה לעומק. מיפינו את אזור הקישור של 14-3-3 בDAPK2- והראינו שהאינטראקציה בין חלבונים אלו מעכבת את הפעילות הביולוגית והביוכימית של DAPK2. הוכחת היתכנות זאת ממחישה את הפוטנציאל של מערכת ה- PCA שיצרנו באיתור מסלולים ביוכימיים חדשים ברשת המוות התאי המתוכנת. בנוסף, בסריקה שביצענו גילינו אינטראקציה חדשה בין שני חלבונים מרכזיים באוטופאגיה – הקינאז ULK1 והחלבון WIPI2. הראינו ש - ULK1 מזרחן את WIPI2 בתנאי מבחנה ובתאים ומיפינו את אתרי הפולספורילציה. כמו כן, הצלחנו להראות שהזרחון באתר S185 בחלבון WIPI2, מבקר את יכולתו להיקשר לממברנות, תכונה החשובה לפעילותו באוטופאגיה. תיעוד ראשוני זה, של קשר בין קינאז רגולטורי באוטופאגיה וחלבון המתווך את יצירת האוטופאגוזום יכול לתרום להרחבת הבנתינו לגבי המנגנון המולקולרי של השלבים הראשונים ביצירת האוטופאגוזום.

Introduction

The life-or-death decision of a cell is a highly complex process, by which multiple pro-survival and pro-death cues are integrated into a single output that determines the cell’s fate. The main cellular pathways that mediate these decisions are apoptosis and autophagy (Rubinstein and Kimchi, 2012). Apoptosis is a programmed cell death mechanism in which activation of the caspase family of cysteine proteases leads to breakdown of the cell via the specific cleavage of substrate proteins. Autophagy is a highly conserved process in which double-membrane vesicles, known as autophagosomes, form to engulf and consume intracellular content upon fusion with the lysosomes. By recycling macromolecular building blocks and removing damaged proteins and organelles, autophagy functions as a survival mechanism, especially during cell stress. While many of the proteins driving each of these two basic biological processes have been discovered over the last few decades, some gaps may still exist, especially in autophagy, which is a younger field. Moreover, autophagy and apoptosis are interconnected processes. In most cases, autophagy blocks apoptosis progression by mitigating the cellular damage that triggers cell death (Suzuki et al., 2011, Ravikumar et al., 2006, Kaminskyy et al., 2012, Bhogal et al., 2012), or by directly targeting pro-death proteins for degradation (Hou et al., 2010, Sandilands et al., 2012b, Sandilands et al., 2012a). Conversely, the apoptotic process can actively suppress autophagy by caspase- dependent degradation of major autophagic proteins, to ensure a death response when necessary (Cho et al., 2009, Luo and Rubinsztein, 2010, Oral et al., 2012, Pagliarini et al., 2012). Interestingly, in some cellular settings, autophagy can contribute to cell death, either by promoting other death pathways (Basit et al., 2013, Laussmann et al., 2011, Qu et al., 2007, Young et al., 2012, Rubinstein and Kimchi, 2012), or potentially, through over-consumption of critical cellular survival factors (Yu et al., 2006, Nezis et al., 2010). The latter type of autophagic cell death can act as a backup mechanism to ensure cell death when caspase dependent pathways are blocked (Shimizu et al., 2004, Denton et al., 2012). Based on this high degree of connectivity, it has been previously suggested that a combined map should be constructed wherein apoptosis and autophagy constitute two modules within the overall programmed cell death (PCD) network (Bialik et al., 2010).

The incomplete proteomics landscapes of autophagy and apoptosis, and the growing evidence as to the mechanistic and therapeutic importance of understanding the cross-talk between these two distinct PCD modules, motivated us to develop new tools for systematic network analysis. Since most of the connectivity between the network’s components involves protein-protein interactions, we utilized a technology called Protein-fragment Complementation Assay (PCA), which enables detection and analysis of such interactions within cells. The PCA strategy is based upon the fusion of separate inactive and structurally unfolded fragments of a reporter protein to two proteins of interest. When these proteins interact, the fragments are brought into close proximity and the reporter is folded into its functional conformation, regaining its activity (Figure 1). The PCA strategy was developed and utilized for a variety of applications using different reporter genes, including the Gaussia princeps luciferase gene (GLuc) (reviewed in (Michnick et al., 2007)). Examples include pathway discovery using cell perturbations and drug actions (Remy and Michnick, 2001, Remy et al., 2004, MacDonald et al., 2006), library screening using GFP PCA in mammalian cells (Remy and Michnick, 2004, Ding et al., 2006), systematic, large scale screening of the yeast interactome (Tarassov et al., 2008) and systematic measurement of protein concentrations (Levy et al., 2014). It was also implemented in the context of cell death for measuring the dynamics of specific interactions, such as caspase-2 dimerization (Bouchier-Hayes et al., 2009) and Bax homodimerization/activation (Yivgi-Ohana et al., 2011). Here we utilized the GLuc PCA strategy for the first time for a global unbiased analysis of a large number of protein-protein interactions within the PCD network in mammalian cells. A library comprising both apoptosis and autophagy proteins was fused to fragments of GLuc, enabling detection of reversible interactions and quantification of temporal alterations in protein complexes within the cell (Remy and Michnick, 2006). Its application resulted in the analysis of the global landscape of protein connectivity and also in specific pathway discovery, as reported here.

Results

Global Screen of the PCD Network

In order to detect protein-protein interactions within the cell death network, we generated the programmed cell death GLuc PCA library. The library encompasses 63 proteins, including apoptosis proteins, autophagy proteins, the Death Associated Proteins (DAPs), and proteins previously reported to associate with and/or regulate apoptotic or autophagic pathways (Table 1). Since the orientation of the fused luciferase fragments can have a significant effect on the interaction between some pairs of proteins and/or on the reconstitution of the luminescence signal, the majority of the proteins in the library were cloned in more than one orientation (i.e., GLuc(1)-X, X-GLuc(1), and X-GLuc(2), where GLuc(1) and GLuc(2) correspond to the N-terminal and C-terminal fragments of the luciferase gene; see Figure 1 and Table 1).

Fusion to luciferase fragment No. Gene Accession Number X-GLuc1 X-GLuc2 GLuc1-X GLuc2-X 1 Bcl-2 NM_000633 + + + + 2 Bcl-XL NM_138578 + 3 Bcl-W NM_004050 + + + 4 Mcl-1 NM_021960 + + 5 Mcl-1(R263A) NM_021960 + 6 BCL2A1 NM_001037100 + 7 BIK NM_001197 + + 8 BAD NM_004322 + + 9 BIM NM_138621 + + 10 BID NM_197966 + + 11 BMF NM_001003940 + 12 Bax NM_138761 + + + + 13 Bak NM_001188 + + 14 Bnip3L NM_004331 + 15 XIAP NM_001167 + + + 16 Survivin NM_001168 + + + 17 Livin NM_139317 + + 18 SMAC NM_019887 + + 19 CytC NM_018947 + + + 20 Casp6 NM_001226 + + + 21 Casp9 NM_001229 + + + 22 Casp3 NM_004346 + + 23 Casp8 NM_001080125 + + 24 Casp2 NM_032982 + 25 Casp10 NM_032977 + 26 ICAD NM_004401 + + + 27 CAD NM_004402 + + + 28 FasR NM_000043 + 29 FADD NM_003824 + + + 30 TRADD NM_003789 + 31 cFLIP NM_001127183 + + 32 RAIDD NM_003805 + +

33 DR5 NM_003842 + + 34 RIPK3 NM_006871 + + 35 PGAM5 NM_001170543 + 36 PIDD NM_145886 + 37 DAPK1 NM_004938 + + 38 DAPK2 NM_014326 + + + 39 DAPK3 NM_001348 + 40 P62 NM_003900 + + 41 Atg3 NM_022488 + 42 Atg4A NM_001261491 + + 43 Atg4B NM_013325 + + 44 Atg5 NM_004849 + 45 Atg7 NM_006395 + + + 46 Atg10 NM_001131028 + 47 Atg12 NM_004707 + + + 48 Atg13 NM_001205119 + + 49 Atg14 NM_014924 + + 50 Atg16 NM_030803 + + 51 mULK1 NM_009469 + 52 mAtg9 NM_001003917 + 53 WIPI2 NM_016003 + 54 UVRAG NM_003369 + + 55 Bif-1 NM_016009 + 56 Beclin-1 NM_003766 + + + 57 LC3 NM_025735 + 58 GABARAPL2 NM_007285 + 59 Ambra1 NM_017749 + + 60 DAP1 NM_004394 + + + 61 BAG3 NM_004281 + 62 14-3-3 NM_006826 + + + 63 LC8 NM_001037494 + + +

Table 1. The gene composition of the programmed cell death GLuc PCA library. Gene names, accession numbers and orientation of their fusion to the different luciferase fragments are documented.

Figure 1. The Gaussia luciferase PCA reporter. A. The PCA strategy is based on the fusion of separately unfolded and inactive fragments of a reporter protein to two proteins of interest. Upon binding of these proteins the reporter is refolded into its active conformation and reconstitutes its activity. B. Schematic representation of the Gaussia luciferase fragments and different fusion orientations.

The resulting PCD library was screened for all possible complementation pairs (3604 combinations using 63 proteins) in HEK293T cells. For each data point, a normalized interaction score (NIS) was calculated (See Materials and Methods for details). The screen was performed three times and the repeats were highly reproducible (r=0.93, 0.86, 0.87, p<0.0001 for repeats 2 and 3, 1 and 2, 1 and 3 respectively; see Figure 2). The mean NIS values calculated from the three experiments, covering the 3604 interactions, are shown in the attached Excel document. Interactions were considered positive for mean NIS>3, and if they were detected in at least two out of the three repeats. The results of the screen are displayed in Figure 3 as a heat map, in which NIS values are represented by a color gradient. The detected interactions were compared to a manually curated list of previously reported interactions among these 63 proteins (see Table 2 for a list of 136 published interactions). Out of a total 131 protein-protein interactions detected in the screen, 85 were previously reported (representing 62.5% of the 136 known interactions) and 46 were not yet identified (Fig. 3B and Table 3). As shown in Fig. 4, the GLuc PCA system enabled the detection of most of the known “core machinery” interactions of the apoptotic and autophagic networks. This includes interactions between proteins from the death-inducing signaling complex (DISC), the Inhibitors of Apoptosis Proteins (IAPs), and the Bcl-2 family, in the apoptotic module. In the autophagic module the interactions within the ULK1 complex, the Beclin-1/Vps34 complex, and the ubiquitin-like conjugation systems were detected. Also several known interactions between autophagic and apoptotic proteins were detected, such as the Beclin-1-Bcl-2 interaction. The 46 previously unknown interactions resulting from the screen are summarized in Table 3. They include two pairs of autophagic proteins, 5 pairs of apoptotic proteins, 9 cross- interactions between apoptotic and autophagic proteins, and 13 interactions between the DAP proteins and components of the autophagic or apoptotic machinery. Additionally, the screen extends the list of proteins interacting with the two scaffold proteins, p62 and FADD, beyond the current published information (11 and 5 new interactions, respectively). Detailed examination of the NIS values within the group of known interactions revealed that although some of the proteins in the library are components of larger multi-protein complexes, the system is specific for direct interactions between

9 pairs of proteins, and does not detect indirect interactions between proteins from the same complex. For example, UVRAG and Atg14 are both part of the autophagic Beclin- 1/Vps34 complex, each interacting directly with Beclin-1 (Itakura et al., 2008). Both proteins exhibited high NIS values for complementation with Beclin-1 (379.36 and 402.61, respectively. see attached Excel file for details), but not with each other. Another example relates to the interaction of Atg16 with the Atg12-Atg5 conjugate. Atg12 and Atg5 are covalently linked in cells by a highly conserved ubiquitin-like conjugation system. The Atg12-Atg5 conjugate associates with Atg16 and this interaction occurs via direct binding to Atg5 (Matsushita et al., 2007). While the Atg5/Atg16 interaction was detected in our screen, the indirect interaction between Atg12 and Atg16 yielded a below- threshold complementation value. This means that the newly discovered positive hits may reflect direct protein-protein interactions, although we cannot exclude the possibility that in some protein complexes the luciferase fragments may be oriented close enough in space to fold even if the interaction is indirect, as previously reported (Tarassov, K., et al. Science, 2008). Altogether, this unbiased GLuc PCA screen provided for the first time a global view of the landscape of protein interactions between and within the apoptotic and autophagic modules.

Figure 2. Comparison among the three screen repeats. A, B and C represent correlation plots of Log2 of the NIS values of repeats 2 and 3, 1 and 3, and 1 and 2, respectively. r= Pearson product-moment correlation coefficient. p<0.0001. D. Venn diagram representing the overlap of the new interactions (NIS>3) detected in the three screen repeats.

Figure 3. PCA screen in HEK293T cells. A. Heat-map showing mean NIS values of the three screen repeats between individual pairs of GLuc(1) (rows) and GLuc(2) (columns)-fusion proteins within the library. B. Venn diagram of the interactions detected in the PCA screen vs. previously known interactions. C. Histogram representing the distribution of NIS values across the library. The number of samples was plotted as a function of Log2 of NIS values.

Figure 4. Network view of all the detected protein-protein interactions in the GLuc PCA screen. The nodes are colored according to their pathway association: apoptotic genes (red), autophagic genes (blue), DAP genes (purple) and proteins that were reported to associate and regulate the apoptotic and autophagic modules (gray). Edges represent positive interactions detected in the GLuc PCA screen (NIS>3) in at least two out of the three repeats. Previously reported interactions are represented by black edges and newly identified interactions detected in the screen are represented by red edges.

No. protein A protein B Reference (PMID) No. protein A protein B Reference (PMID) 17899380, 8900201, 1 14-3-3 14-3-3 9428519 69 Casp3 Casp6 9922454 2 Ambra1 LC8 20921139 70 Casp6 Casp6 19694615 3 Ambra1 Beclin1 17589504 71 Casp8 Casp8 12399450, 12620239 9765224, 8755496, 4 Atg12 Atg10 10508157, 12482611 72 Casp8 Casp3 8962078 5 Atg12 Atg5 9759731, 9852036 73 Casp8 BID 9727491, 9727492 8900201, 9922454, 6 Atg16 Atg16 12665549, 10406794 74 Casp9 Casp3 9390557 7 Atg3 Atg12 20723759 75 Casp9 Survivin 11069302 9880531, 9228018, 9217161, 10200473, 8 Atg4A LC3 21177865 76 cFLIP Casp8 9289491 GABARAP 9 Atg4A L2 21177865 77 cFLIP Casp10 9289491, 9228018 11038174, 15355958, 10 Atg4B LC3 17102583, 15187094 78 cFLIP Atg3 19838173 GABARAP 11 Atg4B L2 14530254, 20404487 79 DAPK1 Beclin1 19180116, 19395874 17192262, 12665549, 12 Atg5 Atg16 10406794 80 DAPk1 DAPK1 21738225, 19712061 13 Atg7 Atg7 11139573, 22170151 81 DAPK1 DAPK3 15367680, 18995835 14 Atg7 Atg12 9759731, 11890701, 82 DAPK2 DAPK2 11230133, 21497605

11096062 15 Atg7 Atg3 11825910 83 DR5 DR5 15538968, 10549288 16 Atg7 Atg10 23388412 84 DR5 FADD 9430228 17 BAD 14-3-3 9369453, 11410287 85 DR5 TRADD 9430228 7536190, 7536190, 18 BAK BAK 17157251 86 FADD FADD 16710361 8681376, 8681377, 19 BAX BAX 8631771 87 FADD Casp8 9184224 12198137, 15721256, 9217161, 9228018, 20 BAX BIM 12242151 88 FADD cFLIP 9289491 21 BAX BID 15574335, 8918887 89 FADD TRADD 8565075 22 BAX BAK 11571294 90 FADD Casp10 11717445 21622563, 14634622, 23 Bcl-2 BIK 9305912 91 FADD PIDD 10825539 7644501, 9111042 24 Bcl-2 BAX ,8358790 92 FADD Atg5 15778222 8967952 ,7538907 25 Bcl-2 BAK 94633819 93 Fas FADD ,7536190 GABARA 26 Bcl-2 BAD 7834748, 15694340 94 PL2 Atg3 20562859 9731710, 15694340, GABARA 27 Bcl-2 BIM 9430630 95 PL2 Atg7 20562859 GABARA 28 Bcl-2 BID 8918887 96 PL2 Atg13 23043107 GABARA 29 Bcl-2 BMF 11546872 97 PL2 Atg16 20562859 9422506, 9422513, 30 Bcl-2 Bnip3L 19273585 98 ICAD CAD 9108473, 9480834 31 Bcl-2 Bcl-2 7744846, 9111042 99 ICAD ICAD 15909126, 16204257 32 Bcl-2 Beclin1 9765397, 16179260 100 LC3 Atg7 11100732, 11096062 33 Bcl-2 Atg12 22152474 101 LC3 Atg3 11100732, 11825910 34 BCL2A1 BID 11929871 102 LC3 P62 20562859 35 BCL2A1 BIM 15694340 103 LC8 BIM 10198631, 14561217 36 BCL2A1 BAK 10381646 104 LC8 LC8 18084006, 10426949 37 BCL2A1 BIK 10381646 105 Livin Casp3 11024045, 14559822 38 BCL2A1 BAX 10753914 106 Livin Livin 16729033 39 Bcl-W BIK 10381646 107 Livin Casp9 11024045 40 Bcl-W BAX 10381646 108 Livin Smac 16729033 41 Bcl-W BAK 10381646 109 Mcl-1 BAX 7644501 42 Bcl-W BAD 10381646, 15694340 110 Mcl-1 BAK 9356461 9430630, 9731710, 43 Bcl-W BIM 15694340 111 Mcl-1 BIM 15694340, 9731710 44 Bcl-W BID 15694340 112 Mcl-1 BID 16380381 45 Bcl-W BMF 11546872 113 Mcl-1 BMF 11546872 46 Bcl-W Beclin1 17643073 114 Mcl-1 Atg12 22152474 GABARAP 47 Bcl-XL BIK 9305912 115 P62 L2 20562859 48 Bcl-XL BAX 9111042, 8521816 116 P62 P62 19931284 15901672, 7644501, 49 Bcl-XL BAK 8521816 117 P62 BAG3 19229298 9372935, 7834748, 50 Bcl-XL BAD 15694340 118 P62 Casp8 19427028, 21628531 9430630, 9731710, 51 Bcl-XL BIM 15694340 119 PGAM5 BCL-XL 22072718 15073321, 17637755, 52 Bcl-XL BID 9727492, 15694340 120 RAIDD PIDD 16183742 8985253 ,9695946, 53 Bcl-XL BMF 11546872 121 RAIDD Casp2 9044836, 15073321 54 Bcl-XL Bnip3L 19273585 122 RIPK3 RIPK3 18533105 17337444, 17446862, 55 Bcl-XL Beclin1 17643073 123 RIPK3 PGAM5 22265414 18843052, 19050071, 19270696, 19270693, 56 Beclin1 Atg14 19223761 124 Survivin Survivin 10876248 57 Beclin1 BIM 22742832 125 Survivin Smac 12660240 58 BID BID 11805084 126 Survivin Casp3 9850056, 11170436 59 Bif1 BAX 11259440, 16227588 127 ULK1 Atg14 23685627

60 Bif1 UVRAG 17891140 128 ULK1 UVRAG 23685627 61 BMF LC8 14561217 129 ULK1 Ambra1 20921139 19258318, 19225151, 62 Casp10 BID 16186808 130 ULK1 Atg13 19211835 16799551, 18843052, 63 Casp2 Casp8 16193064 131 UVRAG Beclin1 19223761 64 Casp2 BID 11832478 132 UVRAG BAX 21597469 65 Casp3 Casp2 9922454 133 XIAP Casp9 11390657, 11927604 9230442, 11359776, 66 Casp3 Casp3 11714276 134 XIAP Casp3 15650747, 11257232 9422506 ,9108473, 22194841, 17698078, 67 Casp3 ICAD 9422513, 9480834 135 XIAP XIAP 23259674 68 Casp3 Bcl-2 9395403 136 XIAP Smac 10929712

Table 2. Previously reported interactions. The list of previously reported interactions was generated by a critical curation of validated interactions from the literature. This list is restricted to the group of genes available in our PCA library (see Table 1).

No. Protein A Protein B Repeats No. Protein A Protein B Repeats 1 Ambra1 Casp8 3 24 DAPK3 FADD 3 2 Ambra1 P62 3 25 DAPK3 DAPK2 3 3 Ambra1 BAG3 3 26 DAPK3 cFLIP 3 4 Ambra1 PIDD 2 27 FADD Bcl-Xl 3 5 Atg14 DAPK2 3 28 FADD CytC 3 6 Atg14 P62 2 29 FADD LC8 3 7 Atg4B DAPK2 2 30 LIVIN DAP1 2 8 Atg4B LC8 2 31 Mcl-1 FADD 3 9 Atg9 WIPI2 2 32 P62 FADD 3 10 Bcl-2 P62 2 33 P62 LC8 3 11 Beclin-1 P62 3 34 P62 CAD 2 12 BID BAD 2 35 P62 DAPK2 2 13 BID FADD 3 36 PGAM5 Bcl-2 2 14 Bnip3L Bax 3 37 Smac CytC 2 15 Casp8 PIDD 2 38 Smac DAP1 2 16 cFLIP P62 3 39 Survivin Atg12 2 17 CytC Casp8 3 40 TRADD P62 3 18 DAP1 BID 3 41 WIPI2 XIAP 3 19 DAP1 BAD 2 42 WIPI2 Casp8 3 20 DAP1 FADD 2 43 WIPI2 ULK1 3 21 DAPK2 14-3-3 3 44 WIPI2 P62 3 22 DAPK2 CytC 2 45 WIPI2 LIVIN 2 23 DAPK3 BAD 3 46 WIPI2 PIDD 2

Table 3. Previously unknown interactions detected in the programmed cell death GLuc PCA screen. Interactions were considered positive for mean NIS>3, and if they were detected in at least two out of the three repeats.

Validation of GLuc PCA Reporters for Known Protein‐ Protein Interactions: Follow‐up of Specificity, Reversibility, and the Dynamics of the Luminescent Signal

In order to demonstrate that the luciferase signal truly represents the state of the detected interactions, and that the positive GLuc PCA pairs can be used as reporters for specific protein-protein interactions, we performed several validation experiments by manipulating known protein interactions and testing their effect on the luminescence signal, either by applying cellular stress or introducing inhibitory point mutations. The first to be tested was the GLuc PCA reporter of a well-defined pair of interacting proteins from the apoptotic module, the endonuclease CAD and its inhibitor ICAD. Apoptosis induction results in caspase-3-mediated cleavage of ICAD and the release of CAD, leading to the characteristic inter-nucleosomal DNA cleavage (Nagata, 2000). Co- expression of ICAD and CAD fused to the Gluc(1) and GLuc(2) fragments resulted in high complementation signals. Upon treatment with the apoptotic trigger staurosporine (STS), a gradual decrease in the luciferase signal was observed over time (Figure 5A), which inversely correlated with increased ICAD cleavage as detected by western blotting of the cell lysates (Figure 5B), indicating the reversibility of the ICAD-CAD GLuc PCA. Notably, addition of the pan-caspase inhibitor QVD-OPH prevented caspase-mediated cleavage of ICAD, and attenuated the decrease in luminescence (Figure 5A,B). Importantly, both endogenous and luciferase-fused ICAD responded to STS with similar temporal kinetics (Figure 5B), indicating that the addition of the luciferase fragment did not change the sensitivity of the ICAD protein to caspase cleavage. This line of experiments confirm that changes in luminescent signals can faithfully reflect the dynamics of interaction between pairs of proteins which change in cells after their exposure to triggers. It also documents the reversibility of the luminescent signal. Assessment of the Atg12/Atg5 interaction provided an additional validation strategy, based on the knowledge that the conjugation between these two proteins depends on the C-terminal glycine of Atg12, and therefore might be inhibited by fusing

16 the luciferase fragment to the C-terminus. Indeed, while co-expression of GLuc(1)- Atg12 with Atg5-GLuc(2) resulted in a positive NIS value (5.75), co-expression of Atg12-GLuc(1) and Atg5-GLuc(2) resulted in a signal below the threshold. To further verify this result, GLuc(1)-Atg12 and Atg12-GLuc(1) were each co-expressed with Atg5- GLuc(2) and the cell lysates were analyzed for both luminescence and western blotting (Figure 5C, D). As expected, the exogenous Atg5-Atg12 conjugate was detected by western blot only in GLuc(1)-Atg12+Atg5-GLuc(2) sample, correlating with the PCA signal, which was 12-fold higher. Thus, the GLuc(1)-Atg12-Atg5-GLuc(2) PCA reporter truly represents the state of the conjugate in cells. Another validation was performed with the anti-apoptotic protein Mcl-1. Expression of wild-type Mcl-1 resulted in high NIS values with the pro-apoptotic proteins Bax and Bak. In contrast, insertion of a single mutation within the BH3 binding pocket of Mcl-1 (R263A), previously shown to ablate Mcl-1 binding to pro-apoptotic proteins (Meng et al., 2007), strongly decreased the luminescence signals, further confirming the specificity of the system (Figure 5E). Taken together, these validations demonstrate that the GLuc PCA pairs detected in the screen truly represent the state of different protein-protein interactions, and can be further utilized as sensitive and quantitative probes for these interactions.

Figure 5. Validation of GLuc PCA reporters using cellular stress and mutagenesis. A. ICAD-GLuc(1) and CAD-GLuc(2) were expressed in HeLa cells in the presence or absence (NT) of staurosporine (2µM, STS) at the indicated time points, with or without QVD-OPH (50µM), a pan-caspase inhibitor, and the GLuc PCA signal was measured in cell lysates. B. Western blotting of the corresponding cell lysates shown in B for the indicated proteins using anti-Gaussia luciferase, anti-ICAD, and anti-tubulin Abs. The upper and lower ICAD fragments result from cleavage of ICAD-GLuc(1) and endogenous ICAD, respectively. C. Relative luminescence signal (RLU) as measured by co-expression of the different Atg12 and Atg5 GLuc PCA plasmids. D. HEK293T cells were transfected with different combinations of GLuc(1)-Atg12, Atg12- GLuc(1) and Atg5-GLuc(2) and western blotting was performed using anti-Atg5, Atg12 and Gaussia antibodies, showing the expression levels of each of the proteins and the generation of the GLuc(1)-Atg12- Atg5-GLuc(2) conjugate (marked with *). E. Wild type MCL1-GLuc(1) and MCL1(R263A)-GLuc(1) were co-expressed with Bax-GLuc(2) or Bak-GLuc(2) and the luminescence signal was measured from cell lysates after 24 h. The results are displayed as meanSD of RLU from three independent repeats. *p<0.05, **p<0.01, ***p<0.001. Student’s t-test.

Previously Unidentified Interactions Detected in the GLuc PCA Screen

Among the interactions detected in the screen novel interacting partners of Ambra1 were discovered. Ambra1 is a member of the Beclin1/Vps34 complex. It was shown to control the localization of the Beclin complex to the cytoskeleton by interacting with the dynein light chain (LC8). Upon autophagy induction, Ambra1 is phosphorylated by ULK1 and the phosphorylation induces its release from LC8, an essential step for autophagy induction (Di Bartolomeo et al., 2010). Ambra1 GLuc PCA resulted in positive complementation values with Beclin-1, LC8 and ULK1 (Figure 6A) confirming these three previously reported interactions (Di Bartolomeo et al., 2010, Fimia et al., 2007). In addition, novel interactions were detected between Ambra1 and the pro- apoptotic proteins caspase-8 and PIDD, providing interesting points of interface between apoptotic and autophagic proteins. Notably, the molecular chaperone BAG3, an anti- apoptotic protein that interacts with Bcl-2 and also mediates selective autophagy of misfolded proteins by binding to p62 (Gamerdinger et al., 2009), was also identified as an Ambra-1 partner (Table 3). These interactions reveal new links between Ambra1 and different components of the cell death network (Figure 6A), and may provide clues for its molecular function. Interestingly, among the different proteins in the library, the Bcl-2 family was found to be highly compatible with the GLuc PCA system, as the screen detected most of the interactions among the family members with relatively high NIS values (Figure 6B,C). Moreover, the different anti-apoptotic Bcl-2 members showed differential BH3- only binding profiles that correspond to previous analysis of this protein family (Chen et al., 2005). For example, while Bcl-2, Bcl-XL and Bcl-W had high complementation signals with the BH3-only protein BAD, Mcl-1 complementation with BAD was below background. Interestingly, we discovered an interaction between Bax and Bnip3L (Figure 6B.C). Bnip3L (Bnip3-like, also known as NIX), and the closely related Bnip3, are BH3-only like proteins that play similar roles in both autophagy and cell death

(Yasuda et al., 1998, Zhang and Ney, 2009). Genetically, Bax/Bak were shown to be required for Bnip3-induced mitochondrial dysfunction and cell death in MEFs (Kubli et al., 2007). The GLuc PCA screen now shows for the first time a direct physical interaction between Bnip3L and Bax, thus providing a new direction for investigation into the mechanism of Bnip3L-mediated cell death.

Figure 6. Ambra1 and the Bcl-2 family sub-networks. A. Ambra1 interacting proteins as detected in the screen. B. Bcl-2 family sub-network. C. Heat-map representing NIS values of various combinations of Bcl- 2 family members.

Focusing on the DAPK2/14‐3‐3τ Interaction: DAPK2 Interacts with 14‐3‐3τ via a Ser/Thr‐Rich Stretch Mapped to the Former’s C Terminus

One of the benefits of the current GLuc PCA library was that it enabled unbiased analysis of the direct interacting proteins of little known proteins previously shown to regulate apoptosis and/or autophagy. One of these candidates is DAPK2 (Death Associated Protein Kinase 2, also named DRP-1), a Ca2+/calmodulin regulated serine/threonine kinase (Figure 7B) that belongs to the DAP-kinase family of proteins, which function as positive mediators of apoptosis and autophagy (Inbal et al., 2002, Inbal et al., 2000, Bialik and Kimchi, 2006, Zalckvar et al., 2009, Eisenberg-Lerner and Kimchi, 2011, Shiloh et al., 2014). Very little is known on DAPK2’s mode of activation and biochemical function. The GLuc PCA screen identified six DAPK2 interacting proteins (Figure 7A), five of which were also validated by co-immunoprecipitation experiments (Figure 8 for Atg4B, Atg14, DAPK3, and p62; and Figure 7 D,E for 14-3- 3). The interaction between DAPK2 and 14-3-3 was further studied here in detail. 14- 3-3 is a member of the highly conserved 14-3-3 family of proteins, which, by preferentially binding specific consensus motifs within phospho Ser-containing proteins (Muslin et al., 1996, Yaffe et al., 1997, Liu et al., 1997), play important roles in the regulation of diverse biological processes, including cell growth control, neuronal development, apoptosis and autophagy (Fu et al., 2000, Pozuelo-Rubio, 2011). In order to validate the PCA interaction, co-immunoprecipitation experiments were performed, in which either 14-3-3-FLAG or FLAG-DAPK2 pulled down HA- DAPK2 or 14-3-3-HA, respectively (Figure 7 D,E). FLAG-DAPK2 also interacted with endogenous 14-3-3 (Figure 9). To check whether other 14-3-3 isoforms bind DAPK2, FLAG-DAPK2 was immunoprecipitated from 293T cells and resolved by SDS-PAGE. Two bands at the predicted size of 14-3-3 (28 kDa) were analyzed by mass spectrometry and identified at high confidence as the theta and epsilon isoforms of 14-3-3 (Figure 7F), suggesting that either DAPK2 binding is not restricted to a specific 14-3-3 isoform, or that a heterodimer of the two isoforms binds DAPK2 within cells.

Examination of DAPK2’s amino acid sequence identified a serine-rich motif at its

C-terminus that potentially matches a 14-3-3 binding site (RX1–2SX2–3S) (Figure 7C) (Liu et al., 1997). Notably, the 14-3-3-GLuc(2) interaction with GLuc(1)-DAPK2 resulted in a much higher complementation signal than with DAPK2-GLuc(1), consistent with the C-terminal fusion blocking the binding site (Figure 10). Since 14-3-3 proteins usually bind phosphorylated proteins, we searched the PhosphoSitePlus (http://www.phosphosite.org) database for detected phosphorylation of the DAPK2 protein, and found that C-terminal S367, S368, T369 and S370 are phosphorylated in both human and mouse orthologues. A mutant lacking the last 5 amino acids of the protein (DAPK2Δ5) was generated, and tested using the GLuc PCA system (Figure 7G). Significantly, while co-expression of GLuc(1)-DAPK2 and 14-3-3-GLuc(2) resulted in a high complementation of the luciferase fragments, the Δ5 mutation almost completely abolished the signal. Loss of 14-3-3 binding was also observed upon co- immunoprecipitation of FLAG-DAPK2Δ5 (Figure 7F), suggesting that the C-terminal tail of DAPK2 is indeed essential for its interaction with 14-3-3. To examine the phosphorylation status of these Ser/Thr residues in HEK293 cells at basal growth conditions, an antibody targeted against the RXRXXpS/T consensus sequence was used. This antibody recognized immunoprecipitated WT FLAG-DAPK2 but not FLAG- DAPK2 Δ5 (Figure 7I). Moreover, pre-treatment of cell extracts with λ phosphatase eliminated the signal, confirming that the Abs indeed recognize phosphorylated residues in DAPK2’s C-terminus, and that these residues are phosphorylated at basal growth conditions (Figure 7J). Additional point-mutations of the suspected Ser residues indicated a reduced complementation signal between 14-3-3-GLuc(2) and either GLuc(1)-DAPK2(S367A), GLuc(1)-DAPK2(S368A) or GLuc(1)-DAPK2(T369A), suggesting a partial role for each of these residues in 14-3-3 binding (Figure 7G). In contrast, the GLuc(1)-DAPK2(S370A) mutation had no effect on the interaction. Importantly, the WT and mutant proteins were expressed at comparable levels in these experiments (Figure 7H).

Figure 7. DAPK2 interacts with 14-3-3. A. DAPK2 interacting proteins as detected in the PCA screen. B. Schematic of domain organization of DAPK2. C. Amino acid sequence of DAPK2. The 14-3-3 binding sequence is colored in red. D. Co-immunoprecipitation of FLAG-tagged DAPK2 with HA-tagged 14-3-3 from HEK293T cells. E. Reciprocal co-immunoprecipitation of FLAG-tagged 14-3-3 with HA-tagged DAPK2 from HEK293T cells. F. Immunoprecipitation of FLAG-DAPK2 and FLAG-DAPK2Δ5 from HEK293T cells. Samples were subjected to SDS-PAGE and gel stained with Gelcode reagent. Arrows indicate endogenous 14-3-3 isoforms  and , as detected by mass spectrometry. G. GLuc(1)-DAPK2 and 14-3-3-GLuc(2) PCA. The different DAPK2 mutants were co-transfected into HEK293T cells with 14-3- 3-GLuc(2). Relative luminescence was measured after 24h in the cell lysates as indicated, ***p<0.001, student’s t-test. H. Western blot of samples from E. were blotted using anti-Gaussia luciferase antibody, which recognizes both luciferase fragments. I. FLAG tagged DAPK2 or FLAG-DAPK2Δ5 were immunoprecipitated from HEK293T cells and blotted with anti-RXRXXpS/T and anti-DAPK2 Abs. J. immunoprecipitated FLAG-DAPK2 was incubated with or without λ phosphatase for 30 min at 370c and blotted with anti-RXRXXpS/T and anti-DAPK2 Abs.

Figure 8. Validation of DAPK2’s sub-network by co-immunoprecipitation. A. Co-immunoprecipitation of FLAG-tagged DAPK2 with HA-tagged Atg4B from HEK293T cells. B. Co-immunoprecipitation of FLAG-tagged Atg14 with HA-tagged DAPK2. C. Co-immunoprecipitation of FLAG-tagged DAPK2 with HA-tagged DAPK3. D. Co-immunoprecipitation of FLAG-tagged p62 with HA-tagged DAPK2.

Figure 9. DAPK2 interacts with endogenous 14-3-3τ. The membrane from the co-immunoprecipitation experiment presented in Figure 7D was stripped and re-blotted with anti-14-3-3τ antibody (Santa Cruz). The overexpressed FLAG-DAPK2 pulled down the endogenous 14-3-3τ as well as the overexpressed 14-3- 3τ -HA.

8.010 6

6.010 6

4.010 6 RLU

2.010 6

(2) c uc(2) u uc(2) L L L 3-G 2-G 2-G K K P P A A 2+14-3- )+D K 3+D 1 P ( A 4-3- uc D 1 L 3-G uc(1)- L GLuc(1)- DAPK2-GLuc(1)+14-3-3-GLuc(2)G 14-3-

Figure 10. 14-3-3 and DAPK2 GLuc PCA. Relative luminescence signal (RLU) as measured by co- expression of the different 14-3-3 and DAPK2 GLuc PCA plasmids. Co-expression of GLuc(1)-DAPK2 and 14-3-3-GLuc(2) resulted in a much higher complementation signal, suggesting a functional role for DAPK2’s C-terminus that is disrupted by the luciferase fusion to its C-terminus. The results are displayed as meanSD of RLU from six replicates.

Measuring the Functional Outcome of the Interaction: 14‐3‐3τ Inhibits DAPK2’s Activity both in Cells and in Cell‐free Systems

Overexpression of DAPK2 was previously shown to cause extensive membrane blebbing, leading to detachment from extracellular matrix and cell death (Shani et al., 2001). FLAG-DAPK25 was slightly more active than FLAG-DAPK2 in inducing membrane blebbing upon over-expression in HEK293T cells. Significantly, co- expression of 14-3-3 strongly inhibited the blebbing phenotype induced by FLAG- DAPK2, but not FLAG-DAPK2Δ5, which cannot bind 14-3-3 (Figure 11A,B). Western blot analysis verified that the differences were not due to differential expression of the DAPK2 constructs (Figure 11C). In order to test the effect of 14-3-3 on DAPK2’s kinase activity in vitro, an ELISA kinase assay using myosin II regulatory light chain (MLC) as an exogenous substrate was performed. FLAG-DAPK2 was incubated with recombinant MLC in the presence of increasing amounts of 14-3-3, and phosphorylation levels quantified using an anti-phosphoSer19-MLC antibody. Consistent with the in vivo data, the presence of 14- 3-3 significantly reduced DAPK2 activity in vitro in a dose-dependent manner (Figure 11D,E). The 40 amino acid tail of DAPK2 was previously shown to be required for its homodimerization (Shani et al., 2001). To determine whether 14-3-3 inhibits DAPK2 activity by preventing its dimerization, a competition assay was conducted, by overexpressing GLuc(1)-DAPK2 and DAPK2-GLuc(2) with increasing levels of 14-3-3- FLAG. The ICAD-GLuc(1) and CAD-GLuc(2) PCA pair was used as a negative control. The presence of 14-3-3 decreased the dimerization signal of DAPK2 by more than 50%, but had no effect on the ICAD-CAD interaction (Figure 11F), suggesting that 14-3-3 binding to the DAPK2 C-terminus prevents its homodimerization, thereby reducing its kinase activity (Figure 12).

Figure 11. 14-3-3τ inhibits DAPK2’s activity. A. 293T cells were transfected with GFP and either 14-3-3- FLAG, FLAG-DAPK2, or FLAG-DAPK2Δ5 in various combinations and imaged after 24h. Arrows indicate examples of blebbed cells. B. Quantification of the extent of blebbing among GFP-positive cells, as meanSD of three independent repeats. C. Western blot of a representative experiment from A. D. Purified DAPK2 was incubated with MLC in an in vitro kinase assay in the presence of purified 14-3-3τ, in various molar ratios, and phosphorylated MLC was quantitated by ELISA assay with anti-phospho MLC antibodies. E. Mean kinase activity ±SD (n=6) of purified DAPK2 with or without 14-3-3 (at the 1:2 molar concentrations) towards non-limiting amounts of MLC (1ug) as measured by the quantitative ELISA assay described in D. F. GLuc(1)-DAPK2 and DAPK2-GLuc(2) or ICAD-GLuc(1) and CAD-GLuc(2) were expressed alone or in the presence of 2 or 4 μg 14-3-3-Flag plasmid. 24h later, the cells were lysed and luminescence was monitored. Complementation levels of each pair in the absence of 14-3-3τ were set as 100%. **p<0.01, ***p<0.001, student’s t-test.

Figure 12. Integrated model of regulation of DAPK2. In the basal state, DAPK2 is autophosphorylated on Ser308, which keeps it inactive. Binding to Ca2+/CaM and dephosphorylation of Ser308 turn on the catalytic activity. In addition, dimerization is necessary for activation. Phosphorylation on Ser367, Ser368 and T369 by a yet unknown kinase (“X”) enables binding of 14-3-3, which sequesters DAPK2 and prevents dimerization. Note that the sequence of the regulatory processes is not yet known, and it is not known if the tail phosphorylation occurs on the active or inactive DAPK2; as such CaM and phospho-Ser308 are faded in the tail-phosphorylated version.

High‐throughput siRNA kinome screen to identify modifiers of the DAPK2‐14‐ 3‐3 interaction

In order to identify the kinase responsible for the phosphorylation of DPAK2’s C- terminus, we performed an unbiased high-throughput siRNA screen of the entire Human kinome using the DAPK2-14-3-3 PCA reporter as readout for the screen. The rational of the screen was based on the notion that 14-3-3 binding to its target proteins is dependent on phosphorylation and that the phosphorylation of DAPK2’s C-terminus is occurring under basal growth conditions in HEK293T cells. Therefore, siRNA silencing of genes which mediate the DAPK2’s tail phosphorylation would result in decreased binding and 28 thus decreased luminescence of the GLuc(1)-DAPK2-14-3-3-GLuc(2) PCA reporter. To this end, we used the siGENOME kinome siRNA library from Dharmacon which include siRNA SMARTpools targeting over 700 genes including all the known human kinases and additional kinase-related genes (e.g. phosphatases, scaffold proteins). siRNAs were reversed transfected into HEK293T cells in 96-well plates and incubated for 48 h to allow effective silencing of the target genes. The cells were then co-transfected with the GLuc(1)-DAPK2 and 14-3-3-GLuc(2) plasmids and incubated for another 24 h. The cells were then lysed and luminescence was read using our standard protocol (Figure 13A). The screen was repeated three times for DAPK2 and 14-3-3 and the average z-score of the two most similar repeats was used for the analysis. In order to avoid false-positive hits of siRNA which cause reduction of the signal due to indirect effects on cell viability or proliferation, we performed the screen another time with a control PCA reporter, expressing Leucine zipper domain fused with the complementary GLuc(1) and GLuc(2) fragments. Therefore, siRNAs which effect viability/proliferation are expected to decrease both the DAPK2-14-3-3 and the ZIP-ZIP signal and would be excluded. Only siRNAs which lead to decreased DAPK2-14-3-3 signal but not decreased ZIP-ZIP signal would be considered as positive hits (Figure 13B). We used a threshold of z-score<-0.5 for DAPK2-14-3-3 PCA and z-score>-0.2 for ZIP-ZIP PCA. DAPK2 and 14-3-3 siRNAs were also included in the library and were used as positive control for the screen sensitivity. Both siRNAs resulted within the lowest 10 z- scores, suggesting that our screening platform is highly sensitive (Figure 14A). Among the other genes which had a strong effect on the signal was PLK which is an essential gene used as positive control in the library since its knock down using siRNA leads to cell death. Overall the screen resulted in 41 positive hits (Figure 14B and Table 4), of which 31 are Ser/Thr kinases, three are Ser/Thr and Tyrosine kinases, 6 Tyrosine kinases and 11 are kinase related proteins such as phosphatase subunits or scaffold proteins. Therefore, while the genes encoding Ser/Thr kinase activity may include some of the direct effectors of the DAPK2-14-3-3 interaction, the rest are probably upstream/indirect regulators. Interestingly, while we were performing the siRNA screen, a paper was published by (Yuasa et al., 2015) which repeated our published results in the context of DAPK2 and

14-3-3 interaction and suggested Akt1 as the kinase responsible for DAPK2’s C-terminus phosphorylation on Thr369. The authors of this paper based their conclusions on co- expression experiments of inactive DAPK2 (K52A) with activated Akt, using an antibody that recognizes p-Akt substrates (RXRXXpS/pT) as readout from western blots of the cell lysates. In addition, they used the Akt inhibitor MK2206 to inhibit the DAPK2-14-3-3 interaction. It should be noted that the inhibitor was used in relatively high concentrations (1-10µM) which may pose a non-specific effect on other kinases. Furthermore, according to our observations, the phosphorylation of the DAPK2 c-terminal tail occurs under basal growth condition and doesn’t require the overexpression of Akt. This further raises doubts about their conclusions. In addition, and the other Ser/Thr residues in the tail may be phosphorylated by different kinases. Our screen identified Akt2 and Akt3 as positive hits but not Akt1, which had no effect on the interaction in all three repeats. In addition, we have performed a bioinformatics analysis of the screen hits for pathway enrichment using the GeneAnalytics tool (http://geneanalytics.genecards.org/) and found that our gene set was highly enriched with genes related to Akt signaling (11 matched genes, score – 22.82). While the Akt family members are thought to have redundant roles and share the same substrates, phenotypic analysis of Akt isoforms knockout mice suggests that there also have some differential roles in metabolism and cancer (Gonzalez and McGraw, 2009). Therefore, it would be interesting to further investigate the involvement of the different Akt isoforms in the regulation of the DAPK2-14-3-3 interaction and use additional biochemical methods such as in-vitro kinase assay to test whether DAPK2 is a direct target of some of the Akt isoforms. Since the C-terminal tail of DAPK2 includes three Ser and one Thr residues, it is possible that it is regulated by more than a single kinase. Another interesting hit from our screen is MAPKAPK2, a stress kinase activated by p38- alpha. Ser368 of DAPK2 fits the MAPKAPK2 consensus site (Hyd-X-R-X(2)-S). The AMPK-related kinase AR5/Nuak1 and its paralog TRIB1 were also detected in the screen as possible regulators of the DAPK2-14-3-3 interaction. AR5/Nuak1 acts downstream of Akt and is involved in various processes including cell adhesion, cell proliferation and tumor progression.

Overall, while we still don’t know which kinase is responsible for DAPK2’s C- terminus phosphorylation, the siRNA kinome screen helped us to narrow down the list of suspects. This in turn would enable us to proceed towards small-scale validation experiments of specific candidates. Currently, there are two main possible directions for the continuing of this work:

1. Educated guess – picking of 2-5 genes of the hit list, based on the existing literature and bioinformatic analysis and examining their ability to phosphorylate DAPK2 by common biochemical assays (e.g. in-vitro kinase assays, measuring the effect on the interaction by siRNA or targeted inhibitors). 2. Unbiased – performing a secondary validation screen on all hits using ON- TARGETplus siRNA, scoring their effect on the interaction and proceeding to biochemical analysis with the top hits. Before this second validation we can refine the list by removing the kinases that are not expressed in the HEK293T cells at basal growth conditions and therefore have no relevance in our screen.

Reverse transfection of siRNA with Dharmafect1in 293T cells ZIP-ZIP

48 hours

Co-transfection of GLuc1-DAPK2 and 14-3-3-GLuc2 with JetPEI DAPK2-14-3-3

24 hours

Read luminescence

Figure 13. Outline of the kinase siRNA screen. A. illustration of the experimental setting of the screen. 293T cells are transfected with individual siRNA in 96-well format. 48 h after transfection, the cells are transfected with the DAPK2-14-3-3 PCA reporter. 24 h later, the cells are lysed and luminescence is measured as readout for the DAPK2-14-3-3 interaction. B. in order to eliminate false-positive hits from siRNAs which effect the viability/proliferation of the cells without having a direct effect on the interaction, the results are plotted against a control reporter. Blue – represents positive hits which lower the DAPK2- 14-3-3 signal only. Red – false-positive hits caused by siRNA which effect viability. Green – negative results.

DAPK2

YWHAQ

6 Z-score of DAPK2-14-3-3 Vs. ZIP-ZIP PCA

2 ZIP-ZIP

0 -2-1.5-1-0.500.511.522.533.544.5

-1

-2 DAPK2-14-3-3

Figure 14. Summary of the screen results. A. Plot of z-score values for the entire siRNA library. DAPK2 and 14-3-3 (YWHAQ) siRNAs are marked by arrows and resulted within the top 10 lowest z-scores, supporting the sensitivity of the screen. B. Scatter plot of DAPK2-14-3-3 PCA and ZIP-ZIP PCA Z-scores obtained from the siRNA screen. Positive hits are colored in red, representing siRNA with Z-score lower than -0.5 for DAPK2-14-3-3 PCA and higher than -0.2 for ZIP-ZIP PCA.

32 genes Z‐score z‐score ZIP‐ZIP genes Z‐score Z‐score ZIP‐ZIP DAPK2‐14‐3‐3 DAPK2‐14‐3‐3 'CLK3' ‐0.66 0.03 'GRK5' ‐0.60 0.59 'CIT' ‐0.60 2.35 FLJ10761' ‐0.75 0.89 'ARK5' ‐0.67 0.19 'HRI' ‐0.88 0.06 'CERK' ‐0.52 0.02 'GK' ‐0.66 ‐0.09 'AK1' ‐0.72 ‐0.03 'MAP3K1' ‐0.68 0.29 'CKMT1' ‐0.60 0.10 'IRAK3' ‐0.83 ‐0.14 'TNK2' ‐0.51 ‐0.04 'ITK' ‐0.87 0.05 'AKT2' ‐0.96 0.69 'MARK3' ‐0.64 0.59 'ACVRL1' ‐0.51 0.03 ‘MATK' ‐0.66 ‐0.18 'AKAP11' ‐0.75 0.11 ‘MAPKAPK2' ‐0.59 0.63 'AKT3' ‐0.65 0.24 'MPP2' ‐0.61 0.49 'BMPR2' ‐0.56 0.84 'MYLK2' ‐0.55 0.49 'TRIB1' ‐0.55 ‐0.01 'NEK7' ‐0.54 ‐0.17 'CDK7' ‐0.59 0.06 'NEK9' ‐0.93 0.44 'DUSP22' ‐0.86 0.44 ‘PPP2CB' ‐0.57 0.10 EPHA4' ‐0.50 0.74 'PLK4' ‐0.76 ‐0.14 'DDR2' ‐0.63 ‐0.14 'RFP' ‐0.91 ‐0.09 ''ERBB4 ‐1.05 0.20 'TRPM7' ‐0.59 0.11 FN3K' ‐0.51 0.93 'TTBK1' ‐0.79 0.50 'GALK2' ‐0.63 0.45 'TTK' ‐0.64 ‐0.09 'TXK' ‐0.79 ‐0.07

Table 4. Positive hits from the siRNA kinome screen for regulators of the DAPK2-14-3-3 interactions. Black-Ser/Thr kinases, Red – Ser/Thr and Tyrosine kinases, Orange – Tyrosine kinases, Green – other regulatory genes (i.e. scaffold proteins, phosphoatases).

ULK1/WIPI2b interaction

Among the interactions detected in the screen, a new interacting partner for the autophagy regulating kinase ULK1 was identified (Table 3). The ULK1 complex (which also includes the regulatory proteins Atg13 and RB1CC1/FIP200) plays an essential role in the regulation of autophagy. Although it is well established that the activity of the complex is dependent on ULK1 kinase activity, the mechanism by which it controls the downstream autophagic machinery is poorly understood (Wong et al., 2013). As shown in Figure 15A, the GLuc PCA screen detected the known interaction between ULK1 and Atg13, a member of the ULK complex that interacts directly with ULK1 and is essential for autophagosome formation (Hosokawa et al., 2009). In addition, interactions between ULK1 and two members of the Beclin1/Vps34 complex, UVRAG and Atg14, were detected. These interactions have recently been described as essential for ULK1- mediated phosphorylation of Beclin-1, as Atg14 and UVRAG serve as adaptors that enable ULK1 association with the Beclin-1/Vps34 complex (Russell et al., 2013). Consistent with these results, the GLuc PCA screen did not detect any direct binding between ULK1 and Beclin-1. Interestingly, a novel interaction between ULK1 and WIPI2 was also detected (Figure 15A and Table 3). WIPI2, an effector of Ptdins3P, is required for LC3 lipidation and for the maturation of omegasome structures, and was shown by co-immunostaining to be recruited to the pre-autophagosomal structure which also contained Atg16 (Polson et al., 2010). Recently, WIPI2 was shown to directly bind Atg16, regulating the localization of the Atg5-Atg12-Atg16 complex and controlling LC3 lipidation at the PAS (Dooley et al., 2014). We hypothesized that the novel interaction between ULK1 and WIPI2b may suggest a functional relationship between these two proteins. Therefore, we decided to zoom into this interaction and understand its role in autophagy regulation. Our first step was to validate the interaction by another independent assay. Co- immunoprecipitation experiments, in which ectopically expressed FLAG-ULK1 pulled down WIPI2b-HA from cells (Figure 15B) confirmed the interaction detected by the GLuc PCA screen. The fact that this was the only novel interaction of ULK1 in our

34 screen, in addition to the detection of four previously described interactions, also encouraged us to further investigate it.

Figure 15. ULK1 sub-network as detected in the PCA screen. A. ULK1 interacting proteins as detected in the screen. Known interactions are colored in blue and previously unknown interaction in red. B. Co- immunoprecipitation of FLAG-tagged ULK1 with HA-tagged WIPI2b from HEK293T cells.

ULK1 phosphorylates WIPI2b in vitro and in cells After validating the interaction, we decided to check whether ULK1 phosphorylates WIPI2b. When WIPI2b-HA was immunoprecipitated from 293T cells, co-expressing FLAG-ULK1, a clear migration shift was observed in WIPI2b protein which migrated slower in SDS-PAGE than when expressed alone (Figure 16A). The migration shift was also observed in western blotting of total cell extracts, revealing a clear transition to the slowly migrating forms of WIPI2b-HA, in response to FLAG- ULK1 overexpression (see Figure 16B). In order to test whether the migration shift is dependent on the kinase activity of ULK1, we co-expressed WIPI2b-HA with wt or kinase-dead mutant (K46I) of ULK1. While co-expression of WIPI2b-HA with wt FLAG-ULK1 resulted in a clear migration shift, co-expression with FLAG-ULK1(K46I) failed to induce the shift, suggesting that the shift is dependent on ULK1 kinase activity (Figure 16C). In order to validate that the migration shift is due to a phosphorylation of WIPI2b, immunoprecipitated WIPI2b-HA was treated in vitro with -phosphatase for 30 min at 370C. It was found that the -phosphatase treatment abolished the migration shift, suggesting that it results of WIPI2b phosphorylation (Figure 16D). We next tested the

35 phosphorylation by a radioactive in vitro kinase assay. Recombinant GST-WIPI2a, which was produced in wheat germ cells, was incubated with or without recombinant GST- ULK1(1-649) in the presence of 32P labeled ATP for 30min at 300C. A clear band was observed in the autoradiogram at the expected size of GST-WIPI2. Additional band was observed at the expected size of GST-ULK1, indicating autophosphorylation. Equal loading of GST-WIPI2 was verified by Gelcode staining (Figure 16E).

Figure 16. ULK1 phosphorylates WIPI2b. A. Evaluation of WIPI2b migration shift in response to ULK1 co-expression by SDS-PAGE and Gelcode staining. B. Evaluation of the migration shift by western- blotting. C. Co-expression of FLAG-ULK1 induced the migration shift of WIPI2b-HA but the expression of a kinase-dead mutant of ULK1 (K46I) failed to do induce the shift. D. phosphatase treatment of cell lysates from 293T cells co-expressing WIPI2b-HA and FLAG-ULK1. Cell lysates were treated with - phosphatase for 30nim at 370C. E. Recombinant GST-ULK1a (1-649) was incubated with recombinant GST-WIPI2 in the presence of radioactive ATP32 for 30min at 300C. Equal quantities of substrate assayed were verified by Gelcode staining.

Mapping the phosphorylation sites on WIPI2b In order to identify the ULK1 phosphorylation sites on WIPI2b, WIPI2b-HA was expressed in 293T cells, with or without FLAG-ULK1, immunoprecipitated and sent for LC/MS analysis. 6 potential phosphorylation sites were identified – S39, S68, S96, S185, S284 and S360. We next wanted to see whether we could detect the phosphorylation of WIPI2b in response to autophagy induction without ULK1 overexpression. For that purpose, WIPI2b-HA was expressed in 293A cells which were treated with DMSO or with the potent mTOR inhibitor Torin1. Immunoprecipitated WIPI2 was analyzed by SDS-PAGE and the migration shift was clearly detected (Figure 17B). Torin1 strongly induces autophagy in 293A cells via inhibition of the mTOR pathway, as measured by enhanced p62 degradation and LC3 lipidation (Figure 17A).

Figure 17. HEK293A cells in response to Torin1 treatment. A. Western blot analysis of 293A cells treated with 200nM Torin1 for 3h. autophagy was measured by enhanced degradation of p62 and LC3-II/LC3-I ratio. B. WIPI2b-HA was expressed in 293A cells treated with DMSO or 200nM Torin1 for 3h, immunoprecipitated and analyzed by SDS-PAGE and Gelcode staining.

Deciphering the functional role of the phosphorylation WIPI2 is a member of the “PROPPIN” family, which is characterized by a - propeller fold and the presence of a FRRG motif required for phosphoinositide binding (Proikas-Cezanne et al., 2015). Several members of the PROPPIN family were shown to bind PI(3)P and PI(3,5)P2. While recombinant expression of most of the members of the family resulted in large soluble aggregates which are not suitable for crystallization, the K. lactis Hsv2 was successfully crystallized (Baskaran et al., 2012). The structural analysis revealed that the conserved FRRG motif participates in two phosphoinositide binding pockets, both required for the proper generation of autophagosomes. These binding sites are located in the C-terminal part of the protein, in blades 5 and 6. Furthermore, the structure analysis identified a hydrophobic loop in blade 6, which is thought to mediate docking of Hsv2 to membranes. The N-terminal part of the protein (mostly blades 1-2) is thought to mediate protein-protein interactions and was shown to be essential for Atg16L1 binding (Dooley et al., 2014). We used the structural and functional insights regarding the different PROPPIN family members in order to hypothesize the possible effects of the recognized ULK1 phosphorylation sites on WIPI2b. Together with Miriam Eisenstein (from the Department of Structural Biology) we have performed a homology modeling of WIPI2b sequence on the Hsv2 structure and tried to estimate the effect of the different phosphorylations on WIPI2 function. According to Figure 18, S68 is located very close to R108, which was found to make direct contact with Atg16. Phosphorylation of S68 is therefore likely to affect binding of WIPI2b to Atg16. The insert including S284 is located in the unstructured loop which is considered to mediate membrane docking; therefore it may affect this function of WIPI2 in cells. In addition, S185 is located next to H183, which is a highly conserved PI(3)P binding residue (Wilson et al., 2014). Therefore, phosphorylation in this site may also have a functional effect on PI(3)P binding.

Figure 18. Localization of the detected phosphorylation sites on WIPI2 sequence and structure. A. WIPI2b amino acid sequence. Phosphorylation sites are marked in yellow, FRRG motif in red, residues which bind PI(3)P in green, residues which bind Atg16 in blue and the unstructured stretch in blade 6 is marked in pink. B. Ribbon diagram of the homology modeling template based on Hsv2 structure. The six phosphorylated sites are indicated in green. Residues implicated in Atg16 binding are indicated by blue ribbon segments and those implicated in PI(3)P binding are implicated by dark red ribbon segments.

While WIPI2b is considered to bind PI(3)P and PI(3,5)P2 based on its structural similarity to other PROPPIN family members, its direct binding and specificity to key cellular phosphoinositides wasn’t clearly demonstrated. To this end, we have performed a lipid binding assay using PIP stripTM (Echelon). This hydrophobic membrane is spotted with various phospholipids and enables to determine protein-lipid interactions using a simple far-western protocol. Overexpressed WIPI2-HA was immunoprecipitated from 293T cells, incubated with the PIP strip array and blotted with anti-HA antibodies (Figure 19). As expected, WIPI2 bound specifically to PI(3)P, and to a less extent to PI(3,5)P2. Surprisingly, additional binding was detected with phosphatidic acid (PA) and phosphatidylserine (PS).

Phosphatidic acid is a precursor for biosynthesis of acylglycerol lipids in the cell including phosphatidylethanolamine (PE) which is covalently conjugated to LC3 during autophagy, a central and indispensable event in autophagy. Interestingly, PA was also found to activate mTOR, a master regulator of autophagy (Yoon et al., 2011). Phosphatidylserine (PS), which also binds WIPI2b, is mainly located on the inner side of the plasma membrane and is flipped to the outer side during apoptosis and thus signal for cell engulfment by macrophages. The affinity for PS may explain the partial localization of WIPI2 to the plasma membrane, as detected by Freeze-fracture replica immune-labeling (Proikas-Cezanne and Robenek, 2011). After establishing and calibrating the PIP binding assay for WIPI2, we generated phospho-mimicking and phospho-silencing mutations of all suspected sites (hereafter referred to as 6D and 6A mutants respectively) and tested their ability to bind PIPs. While the 6A mutant had no effect on the binding compared to the wild-type WIPI2b, the 6D mutations completely abolished its ability to bind phospholipids (Figure 19B). Since Ser185 is located in close proximity to an essential residue for PI(3)P binding (H183), we hypothesized that the phosphorylation of this site is responsible to the observed phenotype of the 6D mutant. Therefore, we generated a single phospho-mimicking mutant (S185D) and compared its PIP binding to the wild-type WIPI2b. As expected, the S185D mutant failed to bind PI(3)P and PI(3,5)2P (Figure 19B), suggesting that the phosphorylation of WIPI2b in this sites by ULK1 regulates its membrane association. Interestingly, the ULK1 complex subunits ULK1, Atg13 and FIP200 all contain a LIR motif which is associated with their binding to Atg8/LC3 family proteins. The LIR motif of ULK1 was found to be dispensable for ULK1 role in early phagophore nucleation but essential for its recruitment into autophagosomes and co-localization with WIPI2 and LC3 positive structures (Carlsson and Simonsen, 2015). Thus, the phosphorylation of WIPI2 by ULK1 may be spatio-temporal regulated by accumulation of LC3 on the growing autophagosome membrane.

Figure 19. WIPI2b phospholipid binding assay. Wild-type or mutant WIPI2-HA was transiently expressed in 293T cells, immunoprecipitated and incubated with PIP StripTM in equal concentrations. Phospholipid bound proteins were detected by far-western blot analysis using anti-HA antibody.A. Outline of the PIP strip membrane. B. Comparison of Wt, 6A and 6D mutants. The presented blots are representative images of three independent repeats. C. Comparison of Wt WIPI2b to the S185D mutant.

In order to test the functional effect of the phosphorylation of WIPI2 on autophagy, we have utilized a rescue system for WIPI2b which was established in Sharon Tooze lab. To this end, we have inserted silence mutations to the WIPI2b-HA plasmids in a region which is targeted by a specific siRNA against all WIPI2 isoforms. Silencing of WIPI2 in HeLa cells stably expressing GFP-LC3 dramatically decreased autophagy in these cells, as was measured by image analysis of GFP-LC3 puncta (Figure 20A and B). In order to check the effect of the 6A and 6D mutants, cells were transfected with siWIPI2 or siCtrl for 48 h and then WIPI2b levels were reconstituted by transfection of siRNA resistance plasmids. While the Wt and 6A mutant rescued the WIPI2 knockdown phenotype, the 6D mutant failed to do so (Figure 20C), corresponding with its inability to bind PI(3)P, an essential function for autophagy. As expected, the S185D mutant also

41 failed to rescue autophagy in WIPI2 depleted cells (Figure 20D), confirming our hypothesis that this is the site responsible for the 6D phenotype. Interestingly, the reconstitution of WIPI2 depleted cells with the WIPI2b-6A mutant resulted in an increase in the GFP-LC3 puncta size compared to the Wt and siCtrl autophagosomes suggesting an increase in the size of autophagosomes or aggregates between several autophagosomes (Figure 21). Although this is currently only a qualitative observation and needs to be quantified and analyzed by more sensitive means (i.e. electron microscopy), it may suggest a possible role of WIPI2b phosphorylation in the regulation of autophagosome size. This implies a second role of ULK1 in autophagy, in addition to the activation of nucleation through Beclin-1 phosphorylation. This second role may block further supply of WIPI2b capable of interacting with PI(3)P through its phosphorylation by ULK1 which is recruited to the growing autophagosome once a critical threshold of lipidated LC3 is reached. This further controls optimal autophagosome size and/or prevent fusion with other autophagosomes. Overall, our results suggest that WIPI2b is regulated by the ULK1 kinase by phosphorylation on six putative sites. The phosphorylation on Ser185 inhibits its PI(3)P binding ability and as a consequence phospho-mimicking mutation in this site prevents its ability to reconstitute autophagy in a WIPI2 knockdown system. Assuming that this regulation is a dynamic process which occurs in a time and localization specific context, the inhibitory effect of the phospho-mimicking mutant in the reconstitution assays probably reflect the second step in the process (after WIPI2b has recruited the Atg16/Atg5/Atg12 to the PI3P to drive LC3 lipidation), to further limit the LC3 accumulation on the growing autophagosomes. The observation from the phospho- silencing mutation, in which large autophagosomes were detected, may suggest that the phosphorylation is involved in the regulation of autophagosome size. It should be noted that although the 6D phenotype was repeated in the S185D mutant, it doesn’t apply about the importance of the other phosphorylation sites. It would be interesting to further study the role of these phosphorylations, especially in the context of other essential functions of WIPI2b, such as Atg16 binding.

Figure 20. Measuring the effect of WIPI2b mutants on autophagy using WIPI2b rescue settings in GFP- LC3 HeLa cells. A. Impact of siWIPI2 on autophagy levels. B. Validation of the WIPI2 knockdown in western blot. C. Comparison of the ability of the 6A and 6D mutants to rescue WIPI2 knockdown. D. repeat of C with S185D single mutant.

Figure 21. Rescue of WIPI2 knockdown with the 6A mutants results in enlarged GFP-LC3 puncta. HeLa cells with stable GFP-LC3 expression were treated with the indicated siRNA for 48 h followed by expression of siRNA resistance WIPI2 plasmids. The shown cells are from representative images used for the analysis presented in Figure 20.

Utilizing the PCA reporters as a screening system for drug discovery

The expansion of our understanding of the cell death network and the interconnectivity of the different cell death pathways enables the identification of specific protein-protein interactions as potential targets for drug development. Some examples such as BH3 mimetic and SMAC mimetic compounds already show promising results in triggering apoptosis in cancer cells (Bai and Wang, 2014). Specific autophagy inhibitors are also of great interest, as inhibition of autophagy was shown to enhance the cellular response to chemotheraphy (Amaravadi et al., 2011). The GLuc PCA reporters detected in our screen can be easily used as quantitative and cost-effective readouts for novel drug discovery. To test the feasibility and sensitivity of the GLuc PCA system for drug screening, we measured the effect of ABT-737, a well characterized protein-protein interaction inhibitor which targets the BH3 binding pocket of anti-apoptotic members of the BCL2 family. Cell lysates, separately expressing Bax-GLuc(2), BCLXL-GLuc(1) or MCL1- GLuc(1) were incubated in-vitro in the presence of increasing levels of ABT-737 and the PCA signal was measured. While the MCL1-Bax interaction was not affected by the presence of ABT-737, the BCLXL-Bax PCA signal decreased in response to elevated levels of the compound (Figure 22). ABT-737 is known to bind with high affinity to BCL2, BCLXL and BCL-W, but only weakly to MCL1. Thus, the GLuc PCA system can be used for quantitative evaluation of BH3 mimetic drugs with high specificity.

Figure 22. GLuc PCA reporter for the activity of BH3-mimetic drugs. A. Schematic representation of the mode of action of BH3 mimetic drugs. B. Bax-GLuc(2), MCL1-GLuc(1) or BCLXL-GLuc(1) were expressed in 293T cells and cell lysates were mixed in the presence of increasing concentrations of ABT- 737. PCA signal is normalized to the control sample.

Autophagy inhibition for cancer therapy The role of autophagy in cancer is context dependent. On one hand, autophagy can suppress tumor development. Defective autophagy leads to the accumulation of damaged organelles and protein aggregates in the cytoplasm, which is linked to increase production of reactive oxygen species (ROS). On the other hand, autophagy can promote cancer cell survival by helping the cell to cope with stressful environment, lack of essential nutrient and metabolic stress (Amaravadi et al., 2011). Therefore, autophagy is induced in response to a variety of anti-cancer drugs, and combinational therapy with autophagy inhibitors leads to synergistic effects. To date, the most widely used autophagy inhibitor is hydroxychloroquine (HCQ), which inhibits the final step of autophagy – fusion of autophagosomes with the lysosome. The antitumor effect of HCQ is currently investigated under several clinical trials. The development of more specific autophagy inhibitors which target different steps in the process will enable better understanding of the effect of autophagy inhibition in cancer therapy.

Atg5/Atg16 interaction We decided to focus on the interaction between Atg16 and Atg5 and utilize its reporter for targeted screening of protein-protein interaction inhibitors. Atg16 creates a complex with the Atg5-12 conjugate, and this complex plays an essential role in autophagy. Atg16 was shown to be essential for the proper localization of the complex to the pre-autophagosomal structure (mostly by interaction with WIPI2b), enabling membrane elongation and acting as an E3 ligase for LC3 lipidation (Dooley et al., 2014). The structure of the Atg16-Atg5 interaction was solved in high resolution (Matsushita et al., 2007), which enables to evaluate its potential for targeting with small molecular inhibitors. Atg5 comprises two ubiquitin-like and a helix-rich domain (see Figure 23A) which together form a groove on the protein surface. This groove is bound by a helical region on the N-terminal part of Atg16. In order to utilize the Atg16-Atg5 PCA reporter for drug screening, we decided to check its complementation under the same setting of the BCLXL-GLuc(1) and Bax- GLuc(2) experiment with ABT-737. Atg5-GLuc(2) and Atg16-L1 were separately or co- expressed in HEK293T cells, cell lysates were prepared and mixed and the complementation signal was measured (Figure 23B). Surprisingly, as the co-expression of both plasmids resulted in a high complementation signal, the lysate mix of the separately expressed plasmids failed to generate a signal higher than the background. While the reason for the inability to generate a complementation signal from the Atg5-GLuc(2) and Atg16-GLuc(1) lysates is currently unclear, we are motivated to investigate it in the future in order to understand the reason for the difference between the complementation capability under separately Vs. co-expression of the plasmids. Since the Gaussia luciferase PCA reporter is reversible, it would also be possible to use the co- expressing lysates for the small molecule inhibitor screening. In order to optimize the reporter for the screen, we have cloned both Atg16 and Atg5 in various fusion conformations to the luciferase fragments and compared their PCA signal under co- expression in HEK293T cells. In addition, the N-terminal fragment of Atg16 which is responsible for Atg5 binding was also cloned (i.e. Atg16(1-57)). Using the N-terminal part alone prevents homodimerization of Atg16 and thus was expected to have a higher

46 complementation signal with Atg5. Indeed, the Atg(16)-GLuc(1) complementation with Atg5-GLuc(2) resulted in the highest signal among all the examined combinations and thus may be the most suitable construct for the future drug screening.

8.010 6

6.010 6

4.010 6 RLU

2.010 6

0 2 -L2 L -L2 5 6 g5-L2 Atg16-L2 6+Atg 1+At +Atg55- +Atg1 1 -L 1 tg 6 -L1 tg5+ -L ) 5 7 -A tg1 -5 1 L1-A A (1 L Atg 6

Atg1

Figure 23. Utilizing the Atg5-GLuc(2) and Atg16-GLuc(1) PCA reporter for drug screening. A. a view of the interaction between Atg5 and Atg16 (adopted from Matsushita et al., 2006). The two ubiquitin-like domains (UblA and UblB) and the helix rich domain (HR) of Atg5 are colored in yellow, pink and green, respectively. Atg16 is shown as a cyan ribbon diagram. B. Atg5-GLuc(2) and Atg16-GLuc(1) were separately or Co-expressed in HEK293T cells. Lysates were mixed and incubated and the interaction was assayed using PCA. C. Comparison of the various PCA reporters of the Atg5-Atg16 interaction. Atg5 and Atg16 fused to the GLuc(1) and GLuc(2) fragments in various orientations were co-expressed in HEK293T cells and PCA signal was compared. Atg16(1-57) represents the N-terminal fragment of Atg16 responsible for Atg5 binding.

Discussion

In the past, our conception about programmed cell death was mostly limited to apoptosis. However, it is now becoming clear that other pathways such as programmed necrosis and autophagy are involved in the cell’s life-or-death decisions. Although many points of interface between these pathways have been uncovered in recent years (Marino et al., 2014), the crosstalk between the different cell death pathways and the backup relationships between them is not completely understood. Protein-protein interactions play a major role in cell signaling pathways. Environmental and intra-cellular signals are constantly collected and converted to a variety of post-translational modifications, such as phosphorylations, protein cleavage, ubiquitinations and other protein modifications. These in turn converge to spatial or temporal changes in protein-protein interactions. In the context of programmed cell death, key steps in apoptosis and autophagy and in the crosstalk between these pathways are regulated by protein-protein interactions and are dependent on protein complexes (Bialik et al., 2010). In this work, we have generated the programmed cell death GLuc PCA library, which enables the discovery and quantitative measurements of protein-protein interactions within the cell death network. By performing an unbiased screen on all possible protein-protein interactions in the library, we managed to detect most of the core-machinery interactions of both autophagy and apoptosis and in addition, to uncover 46 previously unknown interactions. Among those some reflected novel points of interface between apoptosis and autophagy. One of the novel interactions identified in this work, between DAPK2 and 14-3-3т was further functionally investigated both at the biochemical and cellular levels. We showed that 14-3-3 specifically binds to a Ser-rich region in the C-terminal tail of DAPK2 and thus inhibits its dimerization and activation. Furthermore, we have identified five more interacting partners for DAPK2 which may provide clues on its role in autophagy regulation. This proof-of-concept emphasizes the power of the GLuc PCA cell death library for the discovery of and characterization of novel protein-protein interactions within the cell death network. Notably, this work was done under basal growth

48 conditions. One of the future challenges will be to apply different external triggers to this system and test to what extent and how these triggers change the proteomic landscape. By following the novel interaction between ULK1 and WIPI2, we managed to show that ULK1 phosphorylates WIPI2 in-vitro and in cells. Functional and biochemical analysis revealed that phosphorylation of Ser185 abrogated the PI(3)P binding ability of WIPI2b and thus failed to rescue autophagy in WIPI2 knockdown system. While the elucidation of the functional role of this interaction is still ongoing, we believe that it may shed a light on a critical step in autophagy regulation. That is, ULK1 may reach the PAS membranes, through its LIR domain, when the lipidated LC3 reach critical levels, functioning as a gauge that stops further binding of WIPI2b to the autophagosome membranes by its phosphorylation on S185. This provides novel evidence on a molecular connection between ULK1 and the downstream autophagic machinery. Currently we forecast several exciting future applications for the GLuc PCA cell death library (Figure 24). Due to the rapid development of high-throughput functional screening, the field faces an increasing number of genes which are positively or negatively linked to programmed cell death, especially to autophagy, which is a younger field (Orvedahl et al., 2011, Lipinski et al., 2010). The bottleneck however, remains in the annotation of these genes to the pathway’s core machinery and the identification of their mode of action. Our GLuc PCA library can be used as a complementary tool for functional screening as it enables high throughput screening of protein-protein interactions. As such, functionally associated genes can be screened against the library in order to identify their potential protein partners which mediate their function. A second application relates to drug screening. The expansion of our understanding of the cell death network and the interconnectivity of the different cell death pathways enables the identification of specific protein-protein interactions as potential targets for drug development. Some examples such as BH3 mimetic and SMAC mimetic compounds already show promising results in triggering apoptosis in cancer cells (Bai and Wang, 2014). Specific autophagy inhibitors are also of great interest, as inhibition of autophagy was shown to enhance the cellular response to chemotherapy (Amaravadi et al., 2011). The GLuc PCA reporters detected in our screen can be easily used as quantitative and cost-effective readouts for novel drug discovery. This approach

49 has several advantages over using a general pathway marker as a readout (e,g, caspase activation in apoptosis or LC3 lipidation in autophagy). First, since a specific pair of protein interactions can be chosen as a target, it makes it easier to understand the drug’s mode of action and minimize the chance for hidden off-target effects. Second, it enhances the screening sensitivity and enables the detection of drugs with milder effect that may not influence the overall process alone but can be used in combinations with other drugs to achieve a synergistic effect. The GLuc PCA library can help to identify subtle differences between family members. The mammalian apoptosis and autophagy pathways include several protein families whose members display high degree of similarity in protein structures and functional roles. The Bcl-2 family proteins, the inhibitor of apoptosis proteins (IAPs) and the LC3 paralogues are examples of such protein groups. Understanding the difference/similarity among those family members and quantitative mapping of their interacting partners is highly important for our understanding of whether they are fully redundant or display complementary functions in these biological processes. The quantitative information which may result from these experiments can also be used as a powerful resource for computational modeling of the cell death network. Thus, it should be possible to use the GLuc PCA reporters in order to improve our understanding of the specified roles of each member in these protein families.

Unbiased screening Annotation of of pathway-based functionally related GLuc PCA libraries genes to the PPI network

GLuc PCA

Discovering novel interactions Quantitative reporters for specific PPIs

Drug screening Mapping the interaction landscape

Figure 24. Possible applications of the programmed cell death GLuc PCA library. Functionally related genes or pathway-based libraries can be screen against the library for protein-protein interaction mapping. Detected interactions can be further developed as quantitative reporters for specific PPIs or can be used as readouts for drug screening.

Materials and Methods

Bioluminescence Assay and Analysis HeLa or HEK293T cells were reverse-transfected in white 96-well tissue culture plates (BD Falcon) coated with 25ng of each PCA plasmid in the presence of 150mM NaCl containing 0.5µl JetPEI reagent. 24 hours later, the cells were lysed in luciferase lysis buffer (25 mM Tris pH 8.5, 150 mM NaBr, 5 mM EDTA, 0.1% NP40, 5% glycerol, 65 µM sodium oxalate, 0.5 mM reduced glutathione, 0.5 mM oxidized glutathione). Native coelenterazine (Nanolight) was diluted in luciferase assay buffer (25 mM Tris, pH 7.75, 1 mM EDTA, 0.5 mM reduced glutathione, 0.5 mM oxidized glutathione, 75 mM urea) to a final concentration of 20 µM. Luminescence signal (integrated over 10 seconds) was read using a Veritas microplate luminometer (Turner BioSystems). Variations in background signal of the different proteins in the library were detected, which may result from differences in expression levels, higher tendency of some proteins to interact non-specifically with other proteins (“stickiness level”), basal affinity to the luciferase fragments, and allosteric constrains due to variance in protein size or structure. In order to avoid such false readouts, each data point was given a normalized value (normalized interaction score, NIS) calculated by dividing the luminescence readout for each pair of proteins (A+B) by the sum of the median complementation values of each of A  B complementation the proteins, over the entire set: NIS  [Median A][Median B]

ELISA Kinase Assay 96 well ELISA maxiSorp plates (Nunc) were coated with various amounts of recombinant hMLC (purified from E.coli) and incubated at 4C for 72h. The plate wells were washed and blocked with 100µl blocking solution (4% BSA, 0.02% azide in PBS) for 1h at 37C. FLAG-DAPK2 and 14-3-3-FLAG, which were immunopurified from HEK293T cells, were pre-incubated in various amounts in protein kinase buffer (NEB) with 10 mM ATP and added to the ELISA plate for 20 min. The reaction was stopped

52 with 28mM EDTA. The wells were then incubated with anti-pSer19 MLC primary antibody (Cell Signaling), followed by HRP-conjugated goat anti-rabbit secondary antibody, and fresh ABTS substrate solution (28mM mM citric acid, 22mM Na2HPO4

0.006% H2O2 and 1.8mM ABTS), and the color reaction was quantified using an ELISA plate reader.

Blebbing Assay HEK293T cells were transfected with GFP, FLAG-DAPK2 or FLAG-DAPK2Δ5 either with or without 14-3-3-FLAG. 24h later, the cells were imaged by fluorescence microscopy (Olympus BX41) and the blebbed cells were counted. The cells were then harvested and lysed for western blot analysis in order to verify equal expressions of the DNA constructs.

DNA Constructs

A cDNA library was generated from HEK293 cells using RNAeasy MiniElute Cleanup kit (QIAGEN) for mRNA production and SuperScript First-Strand kit (Invitrogen) for RT-PCR. The different genes were amplified with specific primers containing additional restriction sites for BSPEI and XbaI, or NotI and ClaI for the insertion into GLuc(1)-X or X-GLuc(1)/X-GLuc(2) plasmids, respectively (The GLuc(2)- X orientation is less effective and therefore it was not used in most cases). After a standard cloning procedure, the final product was confirmed by direct sequencing of the entire ORF. All mutant plasmids were generated by PCR mediated site directed mutagenesis. The ULK1, Atg9 and LC3 genes in the library are of Mus musculus origin, and were cloned from source DNA plasmids in the same procedure.

Cell Culture and Induction of Cell Death

HeLa and HEK293T were grown in Dulbecco’s Modified Eagle’s medium (DMEM, Biological Industries, Beit Haemek, Israel), supplemented with 10% fetal bovine serum (FBS, GibcoBRL), 4mM glutamine (GibcoBRL) and combined antibiotics

(100µg/ml penicillin and 0.1mg/ml streptomycin). Plasmids were transfected using JetPEI transfection reagent (PolyPlus), according to the manufacturer’s protocol. To induce apoptosis, cells were washed with PBS and then incubated in DMEM containing 2µM Staurosporine (Sigma) for various time-points, as indicated in the text. Caspase inhibition was performed using 50µM QVD-OPH (BioVision).

Protein Analysis

Cells were lysed in B buffer (20 mM HEPES, pH 7.6, 100 mM KCl, 0.5 mM

EDTA, 0.4% NP-40, 20% glycerol) or in PLB (10 mM NaPO4, pH 7.5, 5 mM EDTA, 100 mM NaCl, 1% Triton X-100, 1% Na deoxycholate, 0.1% SDS). The buffers were supplemented with 10l/ml 0.1M PMSF and 1% protease and phosphatase inhibitor cocktails (Roche). Proteins were separated by SDS-PAGE and transferred to nitrocellulose membranes, which were incubated with antibodies to FLAG, HA, - tubulin, Atg5 and actin (Sigma), PARP-1 (Biomol), 14-3-3 and ICAD (Santa Cruz), Atg12 and RXRXXpS/T (Cell Signaling), or Gaussia luciferase (NanoLight). Detection was done with either HRP-conjugated goat anti-mouse or anti-rabbit secondary antibodies (Jackson ImmunoResearch), followed by enhanced chemiluminescence (SuperSignal, Pierce).

Immunoprecipitation

Cells were washed with PBS and extracted in ice-cold lysis buffer as described above. Protein extracts were pre-cleared with protein G beads (Santa Cruz) and then incubated with anti-Flag M2 beads (Sigma). Immunoprecipitates were washed and the protein eluted from the beads with an excess of FLAG peptide.

54 siRNA Screen HEK293T cells were reverse transfected in 96 well plates with the kinome siRNA library (Dharmacon). siRNA transfection was performed at concentration of 50nM using Dharmafect reagent according to the manufacture protocol. 48 h later, cells were transfected with 25ng of L1-DAPK2 and 25ng of 14-3-3-L2 plasmids using JetPEI reagent (0.5µl JetPEI per well). 24 h later, the cells were lysed and luminescence was read according to the bioluminescence assay described above. Z-scores were calculated using MATLAB.

PIP Binding Assay WIPI2b-HA plasmids of the Wt or mutants forms were expressed in HEK293T cells and immunoprecipitated using HA beads (Sigma). The eluted proteins were quantified using SDS-PAGE and gelcode staining to make sure that they were pulled down at equal concentrations. PIP strip membranes (Echelon) were blocked using 3% BSA in PBST (0.05% tween 20) for 1h, followed by incubation with the immunoprecipitated proteins for 1 h. membranes were then incubated for 1 h with anti HA antibody (Covance Ltd.) 1:2000. Finally, detection was done with either HRP- conjugated goat anti-mouse secondary antibodies (Jackson ImmunoResearch, 1:5000), followed by enhanced chemiluminescence (SuperSignal, Pierce).

Statistical Analysis

The statistical significance of differences between means was assessed by two- tailed Student’s T-test. Values of p<0.05 were considered significant.

Protein Interaction Maps

All Protein interaction maps were generated using Cytoscape (http://www.cytoscape.org/).

References

1. AMARAVADI, R. K., LIPPINCOTT-SCHWARTZ, J., YIN, X. M., WEISS, W. A., TAKEBE, N., TIMMER, W., DIPAOLA, R. S., LOTZE, M. T. & WHITE, E. 2011. Principles and current strategies for targeting autophagy for cancer treatment. Clin Cancer Res, 17, 654-66. 2. BAI, L. & WANG, S. 2014. Targeting apoptosis pathways for new cancer therapeutics. Annu Rev Med, 65, 139-55. 3. BASIT, F., CRISTOFANON, S. & FULDA, S. 2013. Obatoclax (GX15-070) triggers necroptosis by promoting the assembly of the necrosome on autophagosomal membranes. Cell Death Differ, 20, 1161-73. 4. BASKARAN, S., RAGUSA, M. J., BOURA, E. & HURLEY, J. H. 2012. Two-site recognition of phosphatidylinositol 3-phosphate by PROPPINs in autophagy. Mol Cell, 47, 339-48. 5. BHOGAL, R. H., WESTON, C. J., CURBISHLEY, S. M., ADAMS, D. H. & AFFORD, S. C. 2012. Autophagy: a cyto-protective mechanism which prevents primary human hepatocyte apoptosis during oxidative stress. Autophagy, 8, 545-58. 6. BIALIK, S. & KIMCHI, A. 2006. The death-associated protein kinases: structure, function, and beyond. Annu Rev Biochem, 75, 189-210. 7. BIALIK, S., ZALCKVAR, E., BER, Y., RUBINSTEIN, A. D. & KIMCHI, A. 2010. Systems biology analysis of programmed cell death. Trends Biochem Sci, 35, 556-64. 8. BOUCHIER-HAYES, L., OBERST, A., MCSTAY, G. P., CONNELL, S., TAIT, S. W., DILLON, C. P., FLANAGAN, J. M., BEERE, H. M. & GREEN, D. R. 2009. Characterization of cytoplasmic caspase-2 activation by induced proximity. Mol Cell, 35, 830-40. 9. CARLSSON, S. R. & SIMONSEN, A. 2015. Membrane dynamics in autophagosome biogenesis. J Cell Sci, 128, 193-205. 10. CHEN, L., WILLIS, S. N., WEI, A., SMITH, B. J., FLETCHER, J. I., HINDS, M. G., COLMAN, P. M., DAY, C. L., ADAMS, J. M. & HUANG, D. C. 2005. Differential targeting of prosurvival Bcl-2 proteins by their BH3-only ligands allows complementary apoptotic function. Mol Cell, 17, 393-403. 11. CHO, D. H., JO, Y. K., HWANG, J. J., LEE, Y. M., ROH, S. A. & KIM, J. C. 2009. Caspase- mediated cleavage of ATG6/Beclin-1 links apoptosis to autophagy in HeLa cells. Cancer Lett, 274, 95-100. 12. DENTON, D., NICOLSON, S. & KUMAR, S. 2012. Cell death by autophagy: facts and apparent artefacts. Cell Death Differ, 19, 87-95. 13. DI BARTOLOMEO, S., CORAZZARI, M., NAZIO, F., OLIVERIO, S., LISI, G., ANTONIOLI, M., PAGLIARINI, V., MATTEONI, S., FUOCO, C., GIUNTA, L., D'AMELIO, M., NARDACCI, R., ROMAGNOLI, A., PIACENTINI, M., CECCONI, F. & FIMIA, G. M. 2010. The dynamic interaction of AMBRA1 with the dynein motor complex regulates mammalian autophagy. J Cell Biol, 191, 155-68. 14. DING, Z., LIANG, J., LU, Y., YU, Q., SONGYANG, Z., LIN, S. Y. & MILLS, G. B. 2006. A retrovirus-based protein complementation assay screen reveals functional AKT1-binding partners. Proc Natl Acad Sci U S A, 103, 15014-9. 15. DOOLEY, H. C., RAZI, M., POLSON, H. E., GIRARDIN, S. E., WILSON, M. I. & TOOZE, S. A. 2014. WIPI2 links LC3 conjugation with PI3P, autophagosome formation, and pathogen clearance by recruiting Atg12-5-16L1. Mol Cell, 55, 238-52. 16. EISENBERG-LERNER, A. & KIMCHI, A. 2011. PKD is a kinase of Vps34 that mediates ROS- induced autophagy downstream of DAPk. Cell Death Differ., 19, 788-97. 17. FIMIA, G. M., STOYKOVA, A., ROMAGNOLI, A., GIUNTA, L., DI BARTOLOMEO, S., NARDACCI, R., CORAZZARI, M., FUOCO, C., UCAR, A., SCHWARTZ, P., GRUSS, P., PIACENTINI, M., CHOWDHURY, K. & CECCONI, F. 2007. Ambra1 regulates autophagy and development of the nervous system. Nature, 447, 1121-5. 18. FU, H., SUBRAMANIAN, R. R. & MASTERS, S. C. 2000. 14-3-3 proteins: structure, function, and regulation. Annu Rev Pharmacol Toxicol, 40, 617-47. 19. GAMERDINGER, M., HAJIEVA, P., KAYA, A. M., WOLFRUM, U., HARTL, F. U. & BEHL, C. 2009. Protein quality control during aging involves recruitment of the macroautophagy pathway by BAG3. EMBO J, 28, 889-901.

20. GONZALEZ, E. & MCGRAW, T. E. 2009. The Akt kinases: isoform specificity in metabolism and cancer. Cell Cycle, 8, 2502-8. 21. HOSOKAWA, N., HARA, T., KAIZUKA, T., KISHI, C., TAKAMURA, A., MIURA, Y., IEMURA, S., NATSUME, T., TAKEHANA, K., YAMADA, N., GUAN, J. L., OSHIRO, N. & MIZUSHIMA, N. 2009. Nutrient-dependent mTORC1 association with the ULK1-Atg13-FIP200 complex required for autophagy. Mol Biol Cell, 20, 1981-91. 22. HOU, W., HAN, J., LU, C., GOLDSTEIN, L. A. & RABINOWICH, H. 2010. Autophagic degradation of active caspase-8: a crosstalk mechanism between autophagy and apoptosis. Autophagy, 6, 891-900. 23. INBAL, B., BIALIK, S., SABANAY, I., SHANI, G. & KIMCHI, A. 2002. DAP kinase and DRP- 1 mediate membrane blebbing and the formation of autophagic vesicles during programmed cell death. J Cell Biol, 157, 455-68. 24. INBAL, B., SHANI, G., COHEN, O., KISSIL, J. L. & KIMCHI, A. 2000. Death-associated protein kinase-related protein 1, a novel serine/threonine kinase involved in apoptosis. Mol Cell Biol, 20, 1044-54. 25. ITAKURA, E., KISHI, C., INOUE, K. & MIZUSHIMA, N. 2008. Beclin 1 forms two distinct phosphatidylinositol 3-kinase complexes with mammalian Atg14 and UVRAG. Mol Biol Cell, 19, 5360-72. 26. KAMINSKYY, V. O., PISKUNOVA, T., ZBOROVSKAYA, I. B., TCHEVKINA, E. M. & ZHIVOTOVSKY, B. 2012. Suppression of basal autophagy reduces lung cancer cell proliferation and enhances caspase-dependent and -independent apoptosis by stimulating ROS formation. Autophagy, 8, 1032-44. 27. KUBLI, D. A., YCAZA, J. E. & GUSTAFSSON, A. B. 2007. Bnip3 mediates mitochondrial dysfunction and cell death through Bax and Bak. Biochem J, 405, 407-15. 28. LAUSSMANN, M. A., PASSANTE, E., DUSSMANN, H., RAUEN, J. A., WURSTLE, M. L., DELGADO, M. E., DEVOCELLE, M., PREHN, J. H. & REHM, M. 2011. Proteasome inhibition can induce an autophagy-dependent apical activation of caspase-8. Cell Death Differ, 18, 1584-97. 29. LEVY, E. D., KOWARZYK, J. & MICHNICK, S. W. 2014. High-resolution mapping of protein concentration reveals principles of proteome architecture and adaptation. Cell Rep, 7, 1333-40. 30. LIPINSKI, M. M., HOFFMAN, G., NG, A., ZHOU, W., PY, B. F., HSU, E., LIU, X., EISENBERG, J., LIU, J., BLENIS, J., XAVIER, R. J. & YUAN, J. 2010. A genome-wide siRNA screen reveals multiple mTORC1 independent signaling pathways regulating autophagy under normal nutritional conditions. Dev Cell, 18, 1041-52. 31. LIU, Y. C., LIU, Y., ELLY, C., YOSHIDA, H., LIPKOWITZ, S. & ALTMAN, A. 1997. Serine phosphorylation of Cbl induced by phorbol ester enhances its association with 14-3-3 proteins in T cells via a novel serine-rich 14-3-3-binding motif. J Biol Chem, 272, 9979-85. 32. LUO, S. & RUBINSZTEIN, D. C. 2010. Apoptosis blocks Beclin 1-dependent autophagosome synthesis: an effect rescued by Bcl-xL. Cell Death Differ, 17, 268-77. 33. MACDONALD, M. L., LAMERDIN, J., OWENS, S., KEON, B. H., BILTER, G. K., SHANG, Z., HUANG, Z., YU, H., DIAS, J., MINAMI, T., MICHNICK, S. W. & WESTWICK, J. K. 2006. Identifying off-target effects and hidden phenotypes of drugs in human cells. Nat Chem Biol, 2, 329-37. 34. MARINO, G., NISO-SANTANO, M., BAEHRECKE, E. H. & KROEMER, G. 2014. Self- consumption: the interplay of autophagy and apoptosis. Nat Rev Mol Cell Biol, 15, 81-94. 35. MATSUSHITA, M., SUZUKI, N. N., OBARA, K., FUJIOKA, Y., OHSUMI, Y. & INAGAKI, F. 2007. Structure of Atg5.Atg16, a complex essential for autophagy. J Biol Chem, 282, 6763-72. 36. MENG, X. W., LEE, S. H., DAI, H., LOEGERING, D., YU, C., FLATTEN, K., SCHNEIDER, P., DAI, N. T., KUMAR, S. K., SMITH, B. D., KARP, J. E., ADJEI, A. A. & KAUFMANN, S. H. 2007. Mcl-1 as a buffer for proapoptotic Bcl-2 family members during TRAIL-induced apoptosis: a mechanistic basis for sorafenib (Bay 43-9006)-induced TRAIL sensitization. J Biol Chem, 282, 29831-46. 37. MICHNICK, S. W., EAR, P. H., MANDERSON, E. N., REMY, I. & STEFAN, E. 2007. Universal strategies in research and drug discovery based on protein-fragment complementation assays. Nat Rev Drug Discov, 6, 569-82. 38. MUSLIN, A. J., TANNER, J. W., ALLEN, P. M. & SHAW, A. S. 1996. Interaction of 14-3-3 with signaling proteins is mediated by the recognition of phosphoserine. Cell, 84, 889-97.

39. NAGATA, S. 2000. Apoptotic DNA fragmentation. Exp Cell Res, 256, 12-8. 40. NEZIS, I. P., SHRAVAGE, B. V., SAGONA, A. P., LAMARK, T., BJORKOY, G., JOHANSEN, T., RUSTEN, T. E., BRECH, A., BAEHRECKE, E. H. & STENMARK, H. 2010. Autophagic degradation of dBruce controls DNA fragmentation in nurse cells during late Drosophila melanogaster oogenesis. J Cell Biol, 190, 523-31. 41. ORAL, O., OZ-ARSLAN, D., ITAH, Z., NAGHAVI, A., DEVECI, R., KARACALI, S. & GOZUACIK, D. 2012. Cleavage of Atg3 protein by caspase-8 regulates autophagy during receptor-activated cell death. Apoptosis, 17, 810-20. 42. ORVEDAHL, A., SUMPTER, R., JR., XIAO, G., NG, A., ZOU, Z., TANG, Y., NARIMATSU, M., GILPIN, C., SUN, Q., ROTH, M., FORST, C. V., WRANA, J. L., ZHANG, Y. E., LUBY- PHELPS, K., XAVIER, R. J., XIE, Y. & LEVINE, B. 2011. Image-based genome-wide siRNA screen identifies selective autophagy factors. Nature, 480, 113-7. 43. PAGLIARINI, V., WIRAWAN, E., ROMAGNOLI, A., CICCOSANTI, F., LISI, G., LIPPENS, S., CECCONI, F., FIMIA, G. M., VANDENABEELE, P., CORAZZARI, M. & PIACENTINI, M. 2012. Proteolysis of Ambra1 during apoptosis has a role in the inhibition of the autophagic prosurvival response. Cell Death Differ, 19, 1495-504. 44. POLSON, H. E., DE LARTIGUE, J., RIGDEN, D. J., REEDIJK, M., URBE, S., CLAGUE, M. J. & TOOZE, S. A. 2010. Mammalian Atg18 (WIPI2) localizes to omegasome-anchored phagophores and positively regulates LC3 lipidation. Autophagy, 6, 506-22. 45. POZUELO-RUBIO, M. 2011. Regulation of autophagic activity by 14-3-3zeta proteins associated with class III phosphatidylinositol-3-kinase. Cell Death Differ, 18, 479-92. 46. PROIKAS-CEZANNE, T. & ROBENEK, H. 2011. Freeze-fracture replica immunolabelling reveals human WIPI-1 and WIPI-2 as membrane proteins of autophagosomes. J Cell Mol Med, 15, 2007-10. 47. PROIKAS-CEZANNE, T., TAKACS, Z., DONNES, P. & KOHLBACHER, O. 2015. WIPI proteins: essential PtdIns3P effectors at the nascent autophagosome. J Cell Sci, 128, 207-217. 48. QU, X., ZOU, Z., SUN, Q., LUBY-PHELPS, K., CHENG, P., HOGAN, R., GILPIN, C. & LEVINE, B. 2007. Autophagy gene-dependent clearance of apoptotic cells during embryonic development. Cell, 128, 931-46. 49. RAVIKUMAR, B., BERGER, Z., VACHER, C., O'KANE, C. J. & RUBINSZTEIN, D. C. 2006. Rapamycin pre-treatment protects against apoptosis. Hum Mol Genet, 15, 1209-16. 50. REMY, I. & MICHNICK, S. W. 2001. Visualization of biochemical networks in living cells. Proc Natl Acad Sci U S A, 98, 7678-83. 51. REMY, I. & MICHNICK, S. W. 2004. Regulation of apoptosis by the Ft1 protein, a new modulator of protein kinase B/Akt. Mol Cell Biol, 24, 1493-504. 52. REMY, I. & MICHNICK, S. W. 2006. A highly sensitive protein-protein interaction assay based on Gaussia luciferase. Nat Methods, 3, 977-9. 53. REMY, I., MONTMARQUETTE, A. & MICHNICK, S. W. 2004. PKB/Akt modulates TGF-beta signalling through a direct interaction with Smad3. Nat Cell Biol, 6, 358-65. 54. RUBINSTEIN, A. D. & KIMCHI, A. 2012. Life in the balance - a mechanistic view of the crosstalk between autophagy and apoptosis. J Cell Sci, 125, 5259-68. 55. RUSSELL, R. C., TIAN, Y., YUAN, H., PARK, H. W., CHANG, Y. Y., KIM, J., KIM, H., NEUFELD, T. P., DILLIN, A. & GUAN, K. L. 2013. ULK1 induces autophagy by phosphorylating Beclin-1 and activating VPS34 lipid kinase. Nat Cell Biol. 56. SANDILANDS, E., SERRELS, B., MCEWAN, D. G., MORTON, J. P., MACAGNO, J. P., MCLEOD, K., STEVENS, C., BRUNTON, V. G., LANGDON, W. Y., VIDAL, M., SANSOM, O. J., DIKIC, I., WILKINSON, S. & FRAME, M. C. 2012a. Autophagic targeting of Src promotes cancer cell survival following reduced FAK signalling. Nat Cell Biol, 14, 51-60. 57. SANDILANDS, E., SERRELS, B., WILKINSON, S. & FRAME, M. C. 2012b. Src-dependent autophagic degradation of Ret in FAK-signalling-defective cancer cells. EMBO Rep, 13, 733-40. 58. SHANI, G., HENIS-KORENBLIT, S., JONA, G., GILEADI, O., EISENSTEIN, M., ZIV, T., ADMON, A. & KIMCHI, A. 2001. Autophosphorylation restrains the apoptotic activity of DRP-1 kinase by controlling dimerization and calmodulin binding. EMBO J, 20, 1099-113. 59. SHILOH, R., BIALIK, S. & KIMCHI, A. 2014. The DAPK family: a structure-function analysis. Apoptosis, 19, 286-97.

60. SHIMIZU, S., KANASEKI, T., MIZUSHIMA, N., MIZUTA, T., ARAKAWA-KOBAYASHI, S., THOMPSON, C. B. & TSUJIMOTO, Y. 2004. Role of Bcl-2 family proteins in a non-apoptotic programmed cell death dependent on autophagy genes. Nat Cell Biol, 6, 1221-8. 61. SUZUKI, S. W., ONODERA, J. & OHSUMI, Y. 2011. Starvation induced cell death in autophagy-defective yeast mutants is caused by mitochondria dysfunction. PLoS One, 6, e17412. 62. TARASSOV, K., MESSIER, V., LANDRY, C. R., RADINOVIC, S., SERNA MOLINA, M. M., SHAMES, I., MALITSKAYA, Y., VOGEL, J., BUSSEY, H. & MICHNICK, S. W. 2008. An in vivo map of the yeast protein interactome. Science, 320, 1465-70. 63. WILSON, M. I., DOOLEY, H. C. & TOOZE, S. A. 2014. WIPI2b and Atg16L1: setting the stage for autophagosome formation. Biochem Soc Trans, 42, 1327-34. 64. WONG, P. M., PUENTE, C., GANLEY, I. G. & JIANG, X. 2013. The ULK1 complex: sensing nutrient signals for autophagy activation. Autophagy, 9, 124-37. 65. YAFFE, M. B., RITTINGER, K., VOLINIA, S., CARON, P. R., AITKEN, A., LEFFERS, H., GAMBLIN, S. J., SMERDON, S. J. & CANTLEY, L. C. 1997. The structural basis for 14-3- 3:phosphopeptide binding specificity. Cell, 91, 961-71. 66. YASUDA, M., THEODORAKIS, P., SUBRAMANIAN, T. & CHINNADURAI, G. 1998. Adenovirus E1B-19K/BCL-2 interacting protein BNIP3 contains a BH3 domain and a mitochondrial targeting sequence. J Biol Chem, 273, 12415-21. 67. YIVGI-OHANA, N., EIFER, M., ADDADI, Y., NEEMAN, M. & GROSS, A. 2011. Utilizing mitochondrial events as biomarkers for imaging apoptosis. Cell Death Dis, 2, e166. 68. YOON, M. S., SUN, Y., ARAUZ, E., JIANG, Y. & CHEN, J. 2011. Phosphatidic acid activates mammalian target of rapamycin complex 1 (mTORC1) kinase by displacing FK506 binding protein 38 (FKBP38) and exerting an allosteric effect. J Biol Chem, 286, 29568-74. 69. YOUNG, M. M., TAKAHASHI, Y., KHAN, O., PARK, S., HORI, T., YUN, J., SHARMA, A. K., AMIN, S., HU, C. D., ZHANG, J., KESTER, M. & WANG, H. G. 2012. Autophagosomal membrane serves as platform for intracellular death-inducing signaling complex (iDISC)-mediated caspase-8 activation and apoptosis. J Biol Chem, 287, 12455-68. 70. YU, L., WAN, F., DUTTA, S., WELSH, S., LIU, Z., FREUNDT, E., BAEHRECKE, E. H. & LENARDO, M. 2006. Autophagic programmed cell death by selective catalase degradation. Proc Natl Acad Sci U S A, 103, 4952-7. 71. YUASA, K., OTA, R., MATSUDA, S., ISSHIKI, K., INOUE, M. & TSUJI, A. 2015. Suppression of death-associated protein kinase 2 by interaction with 14-3-3 proteins. Biochem Biophys Res Commun, 464, 70-5. 72. ZALCKVAR, E., BERISSI, H., MIZRACHY, L., IDELCHUK, Y., KOREN, I., EISENSTEIN, M., SABANAY, H., PINKAS-KRAMARSKI, R. & KIMCHI, A. 2009. DAP-kinase-mediated phosphorylation on the BH3 domain of beclin 1 promotes dissociation of beclin 1 from Bcl-XL and induction of autophagy. EMBO Rep, 10, 285-92. 73. ZHANG, J. & NEY, P. A. 2009. Role of BNIP3 and NIX in cell death, autophagy, and mitophagy. Cell Death Differ, 16, 939-46.