5 6 4
7 8 3
9 2
10 1
Research Activities

Research Activities

Programmed Cell Death: from single genes and molecular pathways towards systems level studies

Our lab studies programmed cell death, a fundamental process in cell biology, by proceeding from single gene studies towards global network analysis.  Every cell has a built-in molecular network of 150-200 proteins, which once activated by the appropriate input signals, leads to cell death.  It is now recognized that the cell has several routes by which to die, including apoptosis, autophagic cell death and programmed necrosis, and that different death stimuli can activate different combinations of these death programs.  Each of these cell death modules is characterized by a set of morphological features (Fig. 1), and is regulated and executed by individual signaling pathways.  Certain protein nodes in these pathways are shared by one or more module, representing points of integration or inter-modular cross-talk.  The sum total of these individual pathways and the interactions among them is referred to as the PCD network (Fig. 2).  By performing large scale anti-sense RNA screens in mammalian cells, we previously identified several new components of the PCD network.  These genes, the DAPs (Death Associated Proteins), were further characterized at the structural and functional levels.  Although DAP genes were originally selected from the same cellular stress settings, they actually differ substantially in their biochemical properties, intracellular localization and the molecular pathways in which they take part.  We currently employ multiple approaches, including biochemistry, cell biology, proteomics-based strategies, 3D protein structure analysis, advanced microscopy and systems biology to elucidate the functions of the individual DAPs, to map their positions within the cell death network, and to analyze network functionality and integration as a whole.  Our ultimate goal is to integrate our findings with previously known data about cell death, in order to define a global cell death network, in which the functional significance of each death pathway, and the inter- and intra-modular interactions among individual proteins or protein complexes within specific pathways, can be predicted for any given death stimulus (e.g., DNA damaging drugs, cytokines, ER stress).

Fig. 1 Electron micrographs of cells undergoing various types of programmed cell death. A. Apoptosis. Note the membrane blebs, fragmented cytoplasm, and condensed chromatin (arrows). B. Autophagy. Note the double-membrane enclosed autophagosomes (arrows). Boxed regions in b1 are shown at higher magnification in b2 and b3. C. Programmed necrosis. Note the extensive vacuolization and swollen organelles.

Fig. 2 The PCD network.  A map of the molecular components of the apoptosis (green), autophagy (blue) and programmed necrosis death (yellow) pathways, which together comprise the three known modules of the PCD network.  The edges possess directionality, indicating either activation of inhibition of the target, and are color coded as indicated in the key.  Gray
boxes define multi-protein complexes.  For more detailed maps, see also
www.cs.tau.ac.il/spike/listOfMaps_files/Apoptosis.html and www.cs.tau.ac.il/spike/listOfMaps_files/autophagy.html.


DAP-kinase (DAPk), a Ca2+/calmodulin (CaM) regulated, Ser/Thr kinase, is one of the most extensively studied proteins in our laboratory.  We have shown that, depending on the cell type and death trigger, DAPk can regulate various forms of programmed cell death, including apoptosis, autophagic cell death and programmed necrosis, as well as several non-death related signaling events (Fig. 3).  These pathways comprise the multiple functional arms of DAPk.  In order to characterize these functional arms, we have focused on identifying DAP-kinase substrates and interacting proteins.  Over the years we have identified several important substrates and interactors, including the myosin II regulatory light chain, phosphorylation of which mediates the process of membrane blebbing, a phenotype common to both apoptotic and autophagic cell deaths; Mcm3, a DNA replication initiation factor, a novel substrate without any previously known link to cell death which may define a new functional arm of DAPk; and Protein Kinase D (PKD), whose phosphorylation and activation by DAPk leads to both autophagy and caspase-independent necrotic cell death. PKD activation in turn leads to activation of JNK, which we have shown to be necessary for both autophagy and programmed necrosis during oxidative stress.  Furthermore, PKD can directly phosphorylate the class III PI3K Vps34, a central regulator of autophagy, providing a direct pathway by which DAPk/PKD can induce autophagy.  In addition, we identified Beclin 1, an essential component of the Vps34 complex that is necessary for autophagosome formation, as a direct DAPk substrate.  DAPk phosphorylates Beclin 1 in its BH3 domain, thereby preventing its interaction with its negative regulators Bcl-2 and Bcl-XL.  Thus, Beclin 1 is another critical molecular link between DAPk and the central autophagy machinery.  An additional DAPk substrate that we have identified is smARF, a small isoform of p19ARF that is produced by internal initiation of translation.  smARF localizes to the mitochondria, and disrupts mitochondrial membrane potential, ultimately leading to caspase-independent autophagic cell death.  We have further shown that smARF is a short-lived protein that undergoes ubiquitin-dependent proteasome-mediated degradation, despite its localization to the mitochondrial matrix. Its stability is enhanced by interaction with the mitochondrial matrix protein p32.
Interestingly, we have also recently identified a novel functional connection between DAPk and metabolism.  A yeast-2-hybrid screen in our lab identified Pyruvate Kinase isoform M2 (PKM2), the enzyme that controls the final rate-limiting step of glycolysis in cancer cells, as a DAPk-interacting protein, an interaction that was further confirmed in vivo.  DAPk enhances PKM catalytic activity in vitro and in vivo, resulting in enhanced glylcolytic flux in cells.  Surprisingly, these effects require direct physical interaction between the proteins but occur independently of DAPk catalytic activity.

Fig. 3 The multiple functional arms of DAPk.  Direct targets are depicted by polygons, indirect targets by ovals.  Double red lines refer to pathways identified in our lab, dashed red lines represent unknown indirect mechanisms.  Known substrates are indicated with a P for phosphorylation.  Note that DAPk physically interacts with and activates PKM and MARK1/2 independently of phosphorylation.  The DAPk substrate Pin1, a phosphoSer/Thr directed peptidy-prolyl isomerase, is itself a master regulator of numerous cell signaling events

We have also generated a DAPk knock-out mouse, in order to enable an understanding of how DAPk regulates physiologically relevant death programs.  For example, we have shown that during ER stress, cell death results from the simultaneous activation of apoptosis and autophagy, both in cultured MEFs and in the intact kidney in vivo.  Significantly, deletion of DAPk inhibits both apoptotic and autophagic death processes and rescues overall cell viability, implying that it lies at a critical integrating junction of these cell death modules.  Several successful collaborations with outside researchers have utilized the DAPk -/- mice to understand additional functions of DAPk in other model systems, including its function in inflammatory responses.
An important direction in the DAPk project is the elucidation of regulatory mechanisms that control its activity.  We have identified an autophosphorylation event on a critical Ser residue within its CaM binding domain, which suppresses kinase function.  DAPk is dephosphorylated and activated by numerous cell-death stimuli.  Work done in our lab and also in collaboration with others (P. Mehlen, Universite de Lyon, France) has indicated that the PP2A phosphatase mediates Ser308 dephosphorylation, and is thus a critical upstream activator of DAPk.  We have also identified an additional level of intra-molecular regulation that occurs via a recently recognized structural feature known as the ROC-COR domains, which it shares with other members of the ROCO family of proteins.  The ROC domain is highly homologous to small G-proteins such as Ras, and in fact, DAPk’s ROC domain can bind GTP and possesses intrinsic GTPase activity.  Significantly, loss of GTP binding promotes DAPk activity, through a mechanism that involves decreased Ser308 phosphorylation.  Based on this, we have proposed a model in which GTP binding to the ROC domain suppresses DAPk function.  Activation of DAPk thus involves three inter-related events: CaM binding to the CaM regulatory domain, dephosphorylation of Ser308 by PP2A, and GTP hydrolysis by the ROC domain.

The DAPk Family

DAPk has two closely related family members, ZIP-kinase (ZIPk, also known as DAPK3) and DRP-1 (also known as DAPK2) (Fig 4).  They share with DAPk several common functional and structural features, including active participation in the autophagic process.  In addition, they all possess a highly ordered basic loop located at the surface of the upper lobe of the catalytic domain, named the ‘fingerprint’ of the DAPk family.  This basic loop mediates heterodimerization between DAPk and ZIPk, which results in trans-phosphorylation and subsequent functional activation of ZIPk, as well as homodimerization of DRP-1.  Currently, we are studying the hierarchy among these kinases, with particular emphasis on elucidating the specific roles that each serves in the various death modules.  In addition, bioinformatics analysis has uncovered several novel features of this very interesting gene family. First, we found that the murine ZIPk underwent unusual and specific sequence divergence from a conserved consensus found in all vertebrates.  This divergence led to changes in protein properties concomitant with adaptive compensatory mechanisms, which developed to preserve the ultimate function of murine ZIPk in cell death.  In addition, we identified a novel isoform of DRP-1, which we call DRP-1b, which shares structural features of both DRP-1 and ZIPk.  A previously unidentified exon in DRP-1 produces a splice variant that replaces the CaM regulatory and dimerization domains of DRP-1 with a ZIPk-like extra catalytic domain.  This results in a CaM-independent kinase that shares interacting partners with ZIPK via the C-terminal Leu Zipper domain, but is exclusively cytoplasmic like DRP-1, in contrast to ZIPK, which localizes to the nucleus and cytoplasm. 
DRP-1β is expressed in early developmental stages of rat brain.


Fig. 4 The members of the DAPK family.  The numbers within the kinase domain represent % identity to DAPk.

DAP1- a Stress Response Modulator

DAP1 is a small (~15Kd), ubiquitously expressed phosphoprotein rich in prolines and lacking known functional motifs.  The gene is highly conserved through evolution in all multicellular organisms.  A bioinformatics search revealed a human homolog of DAP1, named DAP1-Like (DAP1-L).  DAP1 and DAP1-L segregate into completely separate branches in a phylogenetic tree.  Our recent data has shown that DAP1 serves as a negative regulator of starvation-induced autophagy.  In growing cells, the energy sensor and growth regulator mTOR phosphorylates DAP1 on Ser3 and Ser51, thereby inactivating the protein.  Upon amino acid withdrawal, which leads to inhibition of mTOR, DAP1 protein is upregulated and rapidly dephosphorylated, resulting in increased active, suppressive function.  In contrast, depletion of DAP1 by siRNA results in enhanced autophagic flux in starved cells.  Thus DAP1 acts as the “brake” in a “gas and brake” model that is induced upon inactivation of mTORC1 following starvation (Fig. 5).  We are currently investigating the mechanism by which DAP1 suppresses autophagy and have identified several interesting protein interacting partners.

Fig. 5. Gas and Brake model of DAP1 function during autophagy.  Upon starvation, mTORC1 is inactivated, leading to dephosphorylation and activation of the Ulk1/2 complex and DAP1.  While the former induces autophagy, the latter inhibits it.

DAP5 and Cap-Independent Protein Translation

DAP5 is a translation initiation factor of the eIF4G family, which directs IRES-dependent translation under stress conditions when cap-dependent translation is compromised.  As such, it provides a mechanism to maintain translation of itself and specific target mRNAs under circumstances when overall protein translation is suppressed, such as during apoptosis.  Several novel DAP5 mRNA targets were recently identified in our laboratory by screening cDNA arrays with mRNAs that interact with the DAP5 protein or by examining the outcome of knock-down of DAP5 on the expression levels of specific proteins.  Analysis of the profile of DAP5 targets indicates that by driving IRES-mediated translation of target mRNAs, and as a consequence, changing the relative steady state levels of critical proteins, DAP5 can either promote or inhibit the process of cell death.  For example, depletion of DAP5 induces M-phase specific cell death.  During a normal mitotic cycle, DAP5 prevents cell death by facilitating cap-independent translation of several pro-survival proteins: the anti-apoptotic Bcl-2 and Bcl-XL and M-phase kinase CDK1.  In addition to these target screens, analysis of the resolved 3D structure of DAP5 has shed light on its activation and function during cell death.


The Global Cell Death Network

The complicated non-linearity of protein connectivity within each functional cell death module in the PCD network, and the potential inter-modular interactions, further emphasize the need for new strategies capable of dissecting the architecture of the PCD network, and converting it from a static to a dynamic map.  To this end, we developed a new platform for dissecting the network’s complexity based on single and double sets of RNAi-mediated perturbations of apoptotic and autophagic genes.  By applying this strategy to cells exposed to the DNA damaging drug etoposide, we discovered several new principles delineating the structure/function organization of the protein network underlying cell death.  First, our data provided an explanation for the robustness of the network; system performance (i.e. cell death output) can be maintained through activation of dormant death modules upon knock down of the alternate pathways. For example, knockdown of the apoptosis gene caspase-3 attenuated its own apoptotic pathway, and surprisingly also programmed necrosis, but instead triggered the induction of autophagy, resulting in negligible effects on the overall cell death performance.  Significantly, cell death was attenuated upon perturbation of both apoptotic and autophagic modules, indicating that in this setting, autophagy provides an alternative backup mechanism of cell death.  We also discovered an unexpected high degree of positive and negative inter-modular connectivity occurring through several apoptotic and autophagic genes. For example, although the autophagy module is dormant when apoptosis is intact, depletion of the autophagy gene Atg5 partially reduced apoptosis, implying that Atg5 is connected to the apoptotic module independently of its canonical role in autophagy.  Using computational tools to mine the protein-protein interaction databases, we identified a potential molecular pathway linking Atg5 to caspase-3, which was experimentally validated.  The systems-biology approach can be used to address additional issues, such as identification of the minimal number of perturbations that will lead to the collapse of the network (i.e.,~100% reduction in cell death), elucidation of the factors that determine why one cell death pathway predominates in a specific circumstance/cell type, and understanding how a cell establishes the final outcome when multiple death and/or survival pathways are activated simultaneously. 
We are currently employing similar high-throughput platforms to identify additional points of interface and cross-talk between the apoptosis and autophagy modules.  In a high-throughput siRNA screen of autophagic proteins that modulate apoptosis, we discovered the dual nature of the ubiquitin-like protein Atg12, an essential autophagy regulator that emerged as a necessary factor for caspase activation during apoptosis.  Atg12 binds and inhibits anti-apoptotic proteins of the Bcl-2 family through a novel BH3-like motif, and is required for Bax activation upstream of the mitochondrial pathway (Fig. 6).  Interestingly, this function is independent and separable from its autophagic function, in which it is conjugated to Atg5.  This type of crosstalk between regulators of autophagic and apoptotic pathways, particularly proteins that interact with the Bcl-2 family, is emerging as a recurring theme, and underscores the importance of interconnectivity within the PCD network to ensure a coordinated response of the various modules to individual death triggers.

Fig. 6. Dual Function of Atg12 in autophagy and apoptosis.  As a regulator of autophagy, Atg12 is covalently linked to Atg5 in one of two ubiquitin-like conjugation reactions that are necessary for expansion of the autophagosome membrane.  In apoptosis, Atg12 interacts with and inhibits anti-apoptotic Bcl-2 family members, leading to induction of the mitochondrial apoptotic pathway, i.e. activation of Bax, cytochrome C release and subsequent caspase activation.