Structural Analysis of the DAP5 MIF4G Domain and Its Interaction with eIF4A

Geneviève Virgili Department of Biochemistry Groupe de Recherche Axé sur la Structure des Protéines (GRASP) McGill University Montreal, QC April 2013

______A thesis submitted to McGill University in partial fulfillment of the requirements of the degree of Master of Science under the Faculty of Medicine

© GENEVIÈVE VIRGILI, 2013

Never trust an atom. They make up everything. - Unknown author

ii List of abbreviations 4E-BP eIF4E binding- ATP adenosine triphosphate ATPase adenosine triphosphatase Bcl B cell lymphoma CBP80 cap-binding protein 80 CHESS Cornell high energy synchrotron source Csk casein kinase CDK1 cyclin-dependent kinase 1 DAP5M MIF4G domain of DAP5 DNA deoxyribonucleic acid DTT dithiothreitol eIF eukaryotic initiation factor EMCV encephalomyocarditis virus EMSA electrophoretic mobility shift assay Erk extracellular signal-regulated kinases GTP guanosine triphosphate HCV Hepatitis C virus HEAT Huntingtin, Elongation factor 3, A subunit of protein phosphatase 2A [PP2A], and Target of rapamycin IAP inhibitor of apoptosis protein IPTG isopropyl-β-D-thiogalactoside ITC isothermal titration calorimetry Met methionine (or methionyl) MIF4G middle domain of eIF4G mRNA messenger RNA mRNP mRNA-protein complex Mnk MAPK-interacting kinase mTOR mammalian target of rapamycin NOM1 nucleolar MIF4G domain-containing protein 1 NOT1 negative regulator of transcription subunit 1

iii ORF open reading frame PABP poly(A) binding protein Paip1 PABP-interacting protein 1 PDB Pdcd4 programmed cell death protein 4 PI3K phosphatidylinositol 3-OH kinase PIC pre-initiation complex PKC protein kinase C PKR protein kinase double-stranded RNA-dependent PMSF phenylmethylsulfonyl fluoride RENT2 regulator of nonsense transcripts 2, also known as Upf2 RNA ribonucleic acid RRM RNA-recognition motif S6K ribosomal protein S6 kinase SAXS small-angle X-ray scattering SDS-PAGE sodium dodecyl sulfate-polyacrylamide gel electrophoresis Ser serine TC ternary complex TEV Tobacco Etch virus Thr threonine tRNA transfer RNA Met tRNAi initiator tRNA UTR untranslated region Val valine

iv Abstract Death-associated protein 5 (DAP5/p97) is a homolog of the eukaryotic initiation factor 4G (eIF4G) that promotes the IRES-driven translation of multiple cellular mRNAs. Central to its function is the middle domain (MIF4G), which recruits the RNA helicase eIF4A. The middle domain of eIF4G consists of tandem HEAT repeats that coalesce to form a solenoid-type structure. Here, we report the crystal structure of the DAP5 MIF4G domain. Its overall fold is very similar to that of eIF4G, however, significant conformational variations impart distinct surface properties that could explain the observed differences in IRES binding between the two . Interestingly, quantitative analysis of the DAP5-eIF4A interaction using isothermal titration calorimetry reveals a 10-fold lower affinity than with the eIF4G-eIF4A interaction that appears to affect their ability to stimulate eIF4A RNA unwinding activity in vitro. This difference in stability of the complex may have functional implications in selecting the mode of translation initiation.

v Résumé Death-associated protein 5 (DAP5/p97) est une protéine homologue au facteur d'initiation eIF4G qui soutient la traduction d'ARN cellulaires contenant des éléments IRES chez les eukaryotes. Au centre de sa fonction est le domaine MIF4G, qui recrute l'ARN hélicase eIF4A. Ce domaine est constitué de motifs HEAT qui se répètent en tandem pour former une structure en forme de solénoïde. Nous rapportons ici la structure cristalline du domaine MIF4G de DAP5. Sa structure est très similaire à celle d’eIF4G. Toutefois, d'importantes variations de conformation lui confèrent des propriétés de surface distinctes qui pourraient expliquer les différences de liaison d'IRES observées entre les deux protéines. Notamment, l'analyse quantitative de l'interaction DAP5-eIF4A par calorimétrie de titration isotherme révèle une affinité 10 fois plus faible que l'interaction eIF4G-eIF4A, ce qui semble affecter la capacité de DAP5 à stimuler l'activité hélicasique d'eIF4A in vitro. Cette différence dans la stabilité du complexe pourrait être impliquée dans la sélection du mode d'initiation de la traduction.

vi Table of contents List of abbreviations...... iii Abstract...... v Résumé ...... vi Table of contents...... vii List of tables and figures...... ix Preface...... x Acknowledgements...... xi

Prologue: Regulation of expression from DNA to protein...... 1

Chapter 1: Basic concepts in the study of translation initiation 1 Translation...... 4 1.1 Eukaryotic translation initiation...... 6 1.1.1 Anatomy of the mRNA...... 6 1.1.2 General mechanism of eukaryotic translation initiation...... 7 1.1.3 Components of eIF4F...... 9 1.1.3.1 Regulation of eIF4F...... 9 1.1.3.2 eIF4E...... 10 1.1.3.3 eIF4G...... 10 1.1.3.4 eIF4A...... 14 1.2 The versatility of eIF4G family and eIF4G-like proteins...... 18 1.2.1 Occurrence of eIF4G domains in other proteins...... 18 1.2.2 Role of the MA3 domain of eIF4G...... 19 1.2.3 Alternative modes of ribosome recruitment...... 23 1.2.3.1 Viral infection ...... 23 1.2.3.2 Cap-independent translation of cellular mRNAs...... 24 1.3 Death-Associated-Protein 5 (DAP5)...... 27 1.3.1 Topology of DAP5...... 27 1.3.2 Function in cap-independent translation initiation...... 28 1.3.3 Three-dimensional structure...... 30

vii 1.4 Rationale of the study...... 32

Chapter 2: Structural analysis of the DAP5 MIF4G domain and its interaction with eIF4A 2.1 Introduction...... 34 2.2 Results...... 38 2.2.1 Overall structure of the DAP5 MIF4G domain, or DAP5M...... 38 2.2.2 Comparison of DAP5M and eIF4G MIF4G domains...... 42 2.2.3 Identification of a potential IRES binding site in DAP5M...... 43 2.2.4 The sites of interaction with eIF4A are structurally conserved in the middle domains of eIF4G and DAP5 but have different binding affinities...... 47 2.2.5 Mutational analysis of Site 1 and Site 2 residues in the DAP5-eIF4A complex...... 50 2.2.6 Effect of DAP5M on the helicase activity...... 53 2.3 Discussion...... 55 2.4 Materials and methods...... 56 2.4.1 Expression and purification of recombinant proteins...... 56 2.4.2 Mutagenesis of DAP5M...... 58 2.4.3 Crystallization and data collection...... 58 2.4.4 Structure determination...... 58 2.4.5 Gel filtration chromatography...... 59 2.4.6 In vitro pull-down assays...... 59 2.4.7 Isothermal titration calorimetry (ITC)...... 60 2.4.8 Helicase assays...... 60 2.5 Acknowledgements...... 60 2.6 Accession number...... 61

Epilogue...... 62 References...... 63

viii List of tables and figures Table 1-1: Quantitative analysis of the interaction of eIF4A with the MIF4G and MA3 domains of eIF4G...... 14 Table 1-2: Occurrence of eIF4G domains in other proteins...... 18 Table 1-3: Occurrence of eIF4G domains in eIF4G proteins...... 21 Table 1-4: Experimental phosphorylation sites on DAP5...... 28 Figure 1-1: Peptide transfer during translation elongation...... 5 Figure 1-2: General mechanism of translation initiation in eukaryotes...... 8 Figure 1-3: Domain architecture of eIF4G and associated binding partners...... 11 Figure 1-4: Crystal structures of the MIF4G domain in yeast and human eIF4G...... 11 Figure 1-5: Interactions that guide the conformational changes of eIF4A...... 16 Figure 1-6: Can eIF4A interact simultaneously with the MIF4G and MA3 domains?.....19 Figure 1-7: Simplified representation of internal entry on EMCV IRES...... 25 Figure 1-8: eIF4G proteins come in a variety of flavors...... 26 Figure 1-9: Properties of DAP5 and eIF4G C-termini...... 31 Table 2-1: Diffraction data collection and refinement statistics for DAP5M...... 39 Figure 2-1: Domains Structure and sequence alignment of DAP5 and eIF4G...... 35 Figure 2-2: Structural overview of DAP5M...... 40 Figure 2-3: There are two chains in the asymmetric unit of DAP5M...... 41 Figure 2-4: Surface properties of DAP5M and MIF4G domain from eIF4GII...... 45 Figure 2-5: Scoring based on sequence conservation...... 46 Figure 2-6: Model of the DAP5M-eIF4A complex...... 48 Figure 2-7: ITC titration binding curves...... 51 Figure 2-8: In vitro pull-down experiments...... 51 Figure 2-9: Gel filtration profiles of MIF4G-eIF4AI complexes...... 52 Figure 2-10: In vitro helicase assay...... 54

ix Preface This thesis is based on a manuscript to which I contributed as a first author. The contribution of the co-authors is detailed below. I am lucky to have collaborated with dedicated scientists whose expertise raised my research to a greater level.

Chapter 2 Virgili, G.*, Frank, F.*, Feoktistova, K., Sawicki, M., Sonenberg, N., Fraser, C. S., and Nagar, B. (2013). Structural analysis of the DAP5 MIF4G domain and its interaction with eIF4A. Structure 10.1016/j.str.2013.01.015. *These authors contributed equally to this work.

Author contributions Kateryna Feoktistova performed the RNA unwinding experiments under the supervision of Christopher S. Fraser. Maxime Sawicki helped preparing the recombinant proteins for the RNA unwinding experiments. Nahum Sonenberg contributed to supervision of this work and to the revision of the manuscript. Bhushan Nagar supervised this project. Bhushan Nagar and Filipp Frank solved the crystal structure. Filipp Frank and I conceived the experiments with the help of Bhushan Nagar. Bhushan Nagar, Filipp Frank and I wrote the manuscript. I performed the majority of the experiments and instigated the collaboration with the Fraser laboratory.

x Acknowledgements There are many persons to thank for making my pursuit of greater knowledge a very pleasant journey. I would first like to thank my supervisor Dr. Bhushan Nagar for welcoming me in his laboratory. He gave me the opportunity to work on interesting projects and taught me many things along the way. He is an outstanding supervisor and he greatly influenced my development as a scientist and as a writer. I would like to thank Dr. Nahum Sonenberg, for his continued interest in my research and for insightful advices, and our collaborators, Dr. Christopher Fraser and Kateryna Feoktistova, for making things happen. I also thank my Research Advisory Commitee, Drs. Jason Young, Jerry Pelletier and Masad Damha, for interesting discussions. The members of Nagar lab are exceptional people who are always there to support each other. I am especially grateful to Dr. Filipp Frank, who began the DAP5 project with me back in my first moments in the lab. I have to thank him for training me but also for being a friend and a role model to me. I had the chance to supervise talented and dedicated undergraduate students: Ioana Varlan, Samantha Wala and Maxime Sawicki. I am very grateful for their contribution to my research. I thank Yazan Abbas for taking such good care of the lab equipment, for proofreading my thesis and most importantly for the many good times (and still counting!). I am grateful to Dr. Guennadi Kozlov for his technical guidance and for fun times around the world. Props to Yogita Patel and Kristjan Bloudoff, my cohort mates who are going for the long run to the Ph.D. Finally, I want to thank everybody on Bellini 4th floor, past and present, for creating such a great work atmosphere. I received generous support from the Chemical Biology training initiative and from GRASP, notably for presenting my research to the 2011 IUCR meeting in Madrid, Spain. I have also greatly benefited from the equipment facilities that were provided to us by GRASP. À ma famille, merci pour votre affection et vos encouragements incessants. À mes amis de toujours, merci d'être ce que j'ai de plus précieux. À Mathew, merci d'être tel que tu es. Je suis heureuse de t'avoir à mes côtés.

xi Prologue: Regulation of gene expression from DNA to protein Since the description of the structure of DNA by Watson, Crick and Franklin in the 1950s, it became apparent that the sequence of DNA contains the hereditary information essential for life. This led to the formulation of the central dogma by Francis Crick in 1957, stating that in every cell the genetic information contained in DNA can be replicated or transcribed into RNA and subsequently translated into protein (Crick, 1970). The progress in understanding the genetic flow has not slowed down since. Control mechanisms at the transcriptional, post-transcriptional and the post-translational levels now complement the portrait of gene expression, which has been refined with a better understanding of the mechanisms of execution of the genetic code. As a whole, the genetic information necessary for sustaining the life of an organism is known as the genome, and the sequence of entire genomes is available. In eukaryotic cells, DNA is stored in the nucleus in association with histone proteins in a very compact structure named chromatin. Portions of DNA sequences are transcribed into RNA molecules that are complementary to one strand of DNA, and this process of transcription is mediated and regulated by proteins that remodel the chromatin structure to exclude or allow the RNA polymerase to undergo transcription. There are several types of RNA: messenger RNAs (mRNAs), which code for proteins, and other non-coding RNAs, which have diverse functions. For example, two types of non-coding RNA are integral to the translation machinery: ribosomal RNAs (rRNAs) form the core of the translation machinery and polymerize amino acids into proteins, and transfer RNAs (tRNAs) adapt the mRNA code to the cognate amino acid. Once synthesized and properly folded, proteins can effect specific cellular processes. In charge of the synthesis of the effector molecules in the cell, translation is a crucial, energy-consuming process that is highly regulated and that has been extensively described. Eukaryotic translation is affected by cellular stress, viral infection, cancer and aging, and its control mechanisms are important modulators of development, differentiation, learning and memory. Sonenberg and Hinnebusch (2009) published an excellent review providing a comprehensive overview of the mechanisms of translation and their biological role.

1 Death-associated protein 5 (DAP5) is an RNA-binding protein involved in the early stages of mRNA translation and is the focus of the research presented in this dissertation. X-ray crystallography and other biophysical methods were used to study key players involved in this process. The first chapter of this thesis summarizes major concepts in the field of eukaryotic translation initiation, highlighting the notions related to the fundamental questions that are addressed in the study featured in the second chapter of this document.

2

Chapter 1:

Basic concepts in the study of translation initiation

3 1 Translation Ribosomes are the enzymatic effectors of translation. These megadalton-scale ribonucleoproteins are formed of two major subunits, a large and a small subunit, the sizes of which are commonly described in units of velocity of sedimentation (S for Svedberg units). The eukaryotic ribosome is larger than its prokaryotic counterpart: the 60S and 40S subunits compose the 80S eukaryotic ribosome in comparison to the 50S and the 30S subunits of the 70S prokaryotic ribosome. In 2009, the Nobel Prize in Chemistry was awarded jointly to Venkatraman Ramakrishnan, Thomas A. Steiz and Ada E. Yonath for describing the first crystal structures of ribosomal subunits around the year 2000. Those structures of prokaryotic ribosomes have facilitated the determination of hundreds of structures describing different stages of translation (Schmeing and Ramakrishnan, 2009), which can be divided into three phases: initiation, elongation and termination. The many protein factors assisting each of these stages are omitted in the following description for simplicity purposes. The proteins factors taking part to the initiation stage will be the focus of the remainder of this dissertation. Translation begins by recruitment of the mRNA to the Met small ribosomal subunit bound with the initiator methionyl-tRNA (Met-tRNAi ). The small ribosomal subunit then scans the mRNA for the start site of peptide synthesis, defined by the Shine-Dalgarno sequence in bacteria and by the Kozak consensus Met sequence in eukaryotes, which direct the Met-tRNAi contained in the P site of the ribosome to the first protein-coding codon, termed the start codon (Kozak, 1989; Shine and Dalgarno, 1974). The large subunit then joins this assembly and the second aminoacyl-tRNA is recruited to the A site on the ribosome. The amino acids bound to the tRNAs located in the A and P sites become covalently linked to one another and the formation of the first peptide bond ends the initiation stage of translation. The progression of the ribosome on the mRNA causes the tRNA carrying the peptide to be transferred to the P site and the deacylated tRNA to be transferred from the P site to the E site, where it is then released from the ribosome. The ribosome undergoes multiple cycles of peptide chain elongation and transfer during the elongation stage of translation, as shown in Figure 1-1. The recognition of the stop codon by release factors causes the release of the polypeptide and the dissociation of the translation machinery.

4

Figure 1-1. Peptide transfer during translation elongation This figure is adapted from Protein Data Bank (PDB)`s Molecule of the Month (January 2010) by David Goodsell (doi 10.2210/rcsb_pdb/mom_2010_1), which is itself based on PDB entries 2WDK and 2WDL. The 50S subunit is shown in blue and the 30S in green. A minimal mRNA strand is shown in red and the tRNAs in yellow. During elongation, the ribosome can interact with tRNAs at three different sites. The incoming tRNA binds to the A site. After the transfer of the growing peptide from the P-site tRNA to the A-site tRNA at the peptidyl transferase center (circled), the tRNA holding the elongated peptide gets transferred to the P site by the progression of the ribosome on the mRNA. In parallel, the deacetylated tRNA is transferred from the P site to the E site, from where it will be released.

5 1.1 Eukaryotic translation initiation Translation initiation is the rate-limiting step in protein synthesis and is highly regulated. The recruitment of the ribosome and the initiator tRNA to the mRNA is orchestrated by multiple factors, referred to as eukaryotic initiation factors (eIFs).

1.1.1 Anatomy of the mRNA In the eukaryotic nucleus, DNA coding regions are transcribed into mRNA molecules, which are processed before their export to the cytoplasm. The mature mRNA serves as the template for protein synthesis by the ribosome. During transcription, mRNAs originating from the eukaryotic nucleus are modified at the 5' and 3' ends (Maniatis and Reed, 2002). Untranslated regions (UTRs) flank the coding sequence and modulate the efficiency of its translation by different mechanisms (Sonenberg, 1994). The m7GpppN cap (where N represents any nucleotide) is added to the 5' end of the mRNA. This structure is important for the stability of the mRNA and its recruitment of the translation machinery (Furuichi et al., 1977; Shatkin, 1976). There is usually little secondary structure upstream of the coding sequence to ease its scanning by the small ribosomal subunit and ensure efficient translation (Pelletier and Sonenberg, 1985). However, in the 5' UTR of some RNAs there are specialized structural elements that can recruit the translation machinery directly. These are termed internal ribosome entry sites (IRESs) and were originally discovered in picornaviral RNA (Jang et al., 1988; Pelletier and Sonenberg, 1988). A subset of cellular mRNAs is also thought to contain IRES elements, and this will be discussed in section 1.2.3. The open reading frames (ORFs) of mature mRNAs contain the protein-coding region as a continuous array of nucleotide triplets, or codons. The protein-coding sequence usually starts with the AUG codon as defined by the Kozak consensus sequence in such a way that the ribosome begins the peptide synthesis with methionine (Kozak, 1989). Except for the stop codons (UAA, UAG and UGA), each triplet is recognized by tRNAs via their anticodon to allow the recruitment of the corresponding amino acid that is covalently attached to the other extremity of this L-shaped molecule. The 3' end of most eukaryotic mRNAs is polyadenylated (Dávila López and Samuelsson, 2008) and, together with the 5' cap, it enhances translation initiation

6 synergistically (Michel et al., 2000). Due to the activity of 3'-5' exonucleases, the poly(A) tail gradually becomes shorter, leading to a decrease in its translation and eventually its degradation (Åström et al., 1991; Meijer et al., 2007). Elements that regulate mRNA translation and deadenylation are found in the 3' UTR (Fabian et al., 2009). Initiation factors recognize specific structural features of the mRNA to coordinate its recruitment to the translation machinery.

1.1.2 General mechanism of eukaryotic translation initiation To initiate translation, the 40S subunit first associates with eIFs 1, 1A, 3 and 5 and the Met ternary complex (TC), composed of eIF2-GTP and Met-tRNAi , to form the 43S pre- initiation complex (PIC). To join the 40S subunit, the mRNA needs to be activated by the assembly of the eIF4F heterotrimer at its 5' end. eIF4F is composed of the mRNA cap-binding protein eIF4E and the ATP-dependent RNA helicase eIF4A, connected together by eIF4G. This assembly is the focus the research presented in Chapter 2 and will first be described extensively in the section 1.1.3. In addition, the scaffold protein eIF4G recruits the multisubunit complex eIF3, which in turn interacts directly with the 40S subunit. The junction of eIF3 subunits contained in the 43S PIC and the activated mRNA yields the 48S PIC. Within the 48S PIC, the 40S subunit binds and scans the mRNA in the 5'-3' direction until it locates the start codon with the assistance of eIFs 1 and 1A (Pestova et al., 1998a). eIF4F facilitates the binding of the 40S subunit by remodeling the 5' UTR of mRNAs in a ATP-dependent fashion and is required for scanning mRNAs containing secondary structure elements. In the presence of ATP and its cofactor eIF4B, the eIF4A subunit of eIF4F can unwind the structural elements that would otherwise block the progression of the 40S subunit along the 5' UTR (Pestova and Kolupaeva, 2002). Upon recognition of the start codon, eIF5 stimulates the hydrolysis of the eIF2- bound GTP, triggering the exchange of initiation factors that culminates into the recruitment of the 60S subunit. The 80S ribosome can then proceed with translation elongation until the stop codon is reached (Chakrabarti and Maitra, 1991). Those events are summarized in Figure 1-2.

7

Figure 1-2: General mechanism of translation initiation in eukaryotes (Sonenberg & Hinnebusch, 2009) The formation of the 48S PIC requires mRNA activation and 43S PIC formation. PI3K and Ras signaling as well as eIF2α kinases integrate the cellular requirements for protein synthesis by impinging on the two arms of translation initiation.

8 The rate-limiting step of translation initiation is the recruitment of the 43S PIC to the mRNA. Regulation can operate at two fronts along the 48S formation. Cellular stress leads to a decrease in 43S PIC formation due to the phosphorylation of eIF2α, which inhibits the guanine nucleotide exchange factor eIF2B responsible for replenishing eIF2- GTP (Harding et al., 2000; Gomez et al., 2002). The mRNA activating factors are also subjected to regulation and this will be addressed in the next sections.

1.1.3 Components of eIF4F Cancer, viral infection and aging have been associated with altered levels of eIF4F (Sonenberg and Hinnebusch, 2009). Hence, this protein complex has emerged as a logical pharmacological target to block translation initiation (Malina et al., 2012). The structural dissection of the protein-protein and protein-RNA interactions implicated in the regulation and the activity of the heterotrimer is necessary to improve our fundamental knowledge of translation initiation and may open the way to novel strategies for manipulating cellular translation, particularly for the benefit of human health. This section describes the interactions of eIF4F components and the regulating mechanisms.

1.1.3.1 Regulation of eIF4F The exhaustive regulation of eIF4F components reflects its importance for initiating translation. Mitogens and growth factors stimulate translation by activating receptor tyrosine kinases, which in turn activate phosphatidylinositol 3-OH kinase (PI3K) and Ras signaling, two pathways that converge to the regulation of eIF4F components (Figure 1- 2). Mammalian target of rapamycin (mTOR) is a serine/threonine kinase that integrates the status of the cell as implemented by PI3K to control cell growth and proliferation. The activation of mTOR is implicated in cancer biogenesis and synaptic plasticity (Hay and Sonenberg, 2004). Upon activation by growth factors, mitogens and/or hormones, mTOR promotes translation initiation by phosphorylating eIF4E-binding proteins (4E- BPs) (as described in 1.1.3.2) and ribosomal protein S6 kinase (S6K), an eIF3-associated serine/threonine kinase, which in turn phosphorylates eIF4B and the programmed cell death protein 4 (Pdcd4). The phosphorylation of eIF4B enables its interaction with eIF3,

9 thereby promoting the recruitment of the cofactor to eIF4A (Holz et al., 2005). The phosphorylation of Pdcd4 targets it for degradation so that it cannot interfere with translation by binding eIF4A (Dorrello et al., 2006; Suzuki et al., 2008). Similarly, other serine/threonine kinases controlled by PI3K and Ras target eIF4G to modulate its interactions with initiation factors (described in 1.2.2).

1.1.3.2 eIF4E eIF4E is the least abundant initiation factor and hence is a limiting factor for translation initiation (Duncan et al., 1987). The 26-kDa protein anchors the translation machinery to the mRNA by interacting simultaneously with the mRNA 5' cap and a conserved peptide sequence found in eIF4G (Marcotrigiano et al., 1997; 1999). The same peptide is contained in eIF4E-binding proteins (4E-BPs), which interact with eIF4E through the same surface as eIF4G and thereby prevent the assembly of the translation machinery to the mRNA 5' cap (Marcotrigiano et al., 1999). Under favorable conditions for general translation, the phosphorylation of 4E-BPs by mTOR causes their dissociation from eIF4E (Burnett et al., 1998). The relative low abundance of eIF4E makes it an ideal target for translation control (Duncan et al., 1987). Inversely, the deregulation of eIF4E is strongly associated with metastasis and its overexpression is observed in a number of cancers (Mamane et al., 2004). The oncogenic properties of eIF4E depend on the phosphorylation of serine 209 by the MAPK-interacting kinase (Mnk), another binding partner of eIF4G (Topisirovic et al., 2004; Wendel et al., 2007). When bridged together, their proximity facilitates eIF4E phosphorylation, which decreases its affinity for the mRNA 5' cap, thereby enhancing translation efficiency by easing the progression of the translation machinery toward the coding region (Furic et al., 2010).

1.1.3.3 eIF4G The eIF4G family of proteins contains three members in humans: eIF4GI, eIF4GII and the less conserved DAP5. The mass of bona fide eIF4G is approximately 175 kDa and its overall protein topology can be divided into three portions: the N-terminal third, the

10 middle third and the C-terminal third (Figure 1-3). DAP5 is equivalent to the C-terminal two thirds of other human eIF4Gs and will be thoroughly described in section 1.3. The N-terminal third of eIF4G contains binding sites for the poly(A)-binding protein (PABP) and eIF4E, which respectively engage the poly(A) tail of the mRNA and its 5’ cap. These simultaneous interactions lead to the circularization of the mRNA and enhance translation efficiency by promoting ribosome recycling and by increasing the affinity of eIF4G for the cap-bound eIF4E (Kahvejian et al., 2005). The N-terminus of eIF4G is mostly unstructured and so are its binding sites for eIF4E and PABP. The flexible peptide sequence Tyr-X4-Leu-phi (where X4 represents four variable residues and phi, a hydrophobic residue) contained in the N-terminus of eIF4G undergoes a disorder- to-order transition when interacting with eIF4E (Marcotrigiano et al., 1999). Similarly, the residues 161-216 of eIF4G fold upon binding to PABP (Safaee et al., 2012). It was suggested that the structural transitions could increase eIF4E and PABP-binding specificities or couple their binding to other regulatory events. The middle third of eIF4G (MIF4G) is the minimal core for translation initiation (Morino et al., 2000). It can interact with eIF4A and eIF3, and has RNA-binding properties (De Gregorio et al., 1998; Imataka and Sonenberg, 1997; Lamphear et al., 1995). The binding sites for eIF4A and eIF3 are distinct so that the two initiation factors can bind MIF4G separately (Imataka and Sonenberg, 1997; Morino et al., 2000). Nonetheless, the binding of both proteins appears stronger in the presence of the other, and therefore their interactions with MIF4G are said to be mutually cooperative (Korneeva et al., 2000). In general, the interaction of eIF4G with mRNAs is unspecific, but for alternative mechanisms of translation initiation MIF4G can recognize specific structural elements in the mRNA to bridge it to the small ribosomal subunit (Prévôt et al., 2003). The structure of MIF4G was described for human eIF4GII and for yeast eIF4G (TIF4631) in complex with eIF4A (Marcotrigiano et al., 2001; Schütz et al., 2008) (Figure 1-4). The domain is made of 5 tandem HEAT repeats (for Huntingtin, Elongation factor 3, A subunit of protein phosphatase 2A, and Target of rapamycin), which are helix- loop-helix motifs that are often implicated in protein-protein interactions (Andrade and Bork, 1995). The repeats are stacked on top of one another into a superhelical arrangement, with a convex and concave surfaces, of which the latter binds eIF4A to

11

Figure 1-3: Domain architecture of eIF4G and associated binding partners

Figure 1-4: Crystal structures of the MIF4G domain in yeast and human eIF4G The two structures, respectively described in PDB entries 2VSO and 1HU3, share 36% sequence identity and 2.7 Å root-mean-square deviation (r.m.s.d.) over 196 aligned residues (Holm et al., 2010). The convex and concave surfaces that result from the superhelical fold of the HEAT domains are noticeable on the representation on the left side. From the yeast crystal structure, we know that eIF4A binds the concave surface.

12 enhance its RNA-dependent ATPase activity (Schütz et al., 2008). The C-terminal third of eIF4G contains its second and third HEAT domains and their structure were described by Bellsolell et al. (2006). The second HEAT domain is termed MA3 due to its similarity to the eIF4A-binding domains of the translation inhibitor Pdcd4. This second eIF4A-binding site on eIF4G is thought to have a modulatory role (Imataka and Sonenberg, 1997). The third HEAT domain is called W2 (for two invariant tryptophans, also referred to as eIF5C), and contains two aromatic/acidic box (AA-box) motifs (Bellsolell et al., 2006; Morino et al., 2000). The acidic residues contained in the first motif interact with a stretch of eight basic residues found in Mnk to bring the kinase in proximity of eIF4E for increased translation efficiency. The presence of a second eIF4A-binding site raised questions about the eIF4G- eIF4A interaction and its stoichiometry. Several groups compared the ability of MIF4G and MA3 to bind eIF4A by testing the domains in isolation or within the same protein construct. The dissociation constants (KD) that are summarized in Table 1-1 were obtained using surface plasmon resonance (SPR) or isothermal titration calorimety (ITC). Two groups observed a twenty-fold stronger affinity of eIF4A for MIF4G compared to MA3 (Korneeva et al., 2001; Nielsen et al., 2011). Other groups demonstrated that the difference in affinity is more moderate (Fujita et al., 2009; Li, 2001), in line with the results obtained in our laboratory (G. Virgili, unpublished data). Taken together, those results agree in that the functional interaction of MIF4G with eIF4A is stronger to some extent. However, the eIF4G-eIF4A stoichiometry remains controversial, with some groups observing a second binding event presumably via MA3 while others have not (Fujita et al., 2009; Korneeva et al., 2001; Li, 2001; Nielsen et al., 2011). The binding of eIF4A to MIF4G is the functional interaction, the second binding site likely has an accessory role such as promoting eIF4A recycling (Morino et al., 2000; Nielsen et al., 2011). Still, the limited difference in affinity for eIF4A between the two domains does not explain by itself how the functional interaction dominates over the MA3-eIF4A interaction during translation initiation.

13 Table 1-1: Quantitative analysis of the interaction of eIF4A with the MIF4G and MA3 domains of eIF4G

eIF4GI residues KD (nM) Fold Method Reference domains residues* MIF4G MA3 difference in 711-1450 67 1500 22 SPR Nielsen et al. (2011) tandem 711-1450 50 775 16 ITC MIF4G 653-1117 17 SPR 19 Korneeva et al. (2001) MA3 1117-1599 330 SPR MIF4G 671-1038 412 SPR 1 MA3 1039-1599 429 SPR Fujita et al. (2009) MIF4G 670-1038 77 ITC 11 MA3 1039-1599 862 ITC MIF4G 732-1003 136 ITC G. Virgili, 6 MA3 1234-1571 869 ITC unpublished data * Sequence boundaries adapted to Uniprot entry Q04637-1

1.1.3.4 eIF4A The eIF4A subunit of eIF4F is the enzymatic effector of RNA remodeling, allowing the small ribosomal subunit to bind and scan the 5'UTR for the start codon. eIF4A is the most abundant initiation factor with three-fold excess over the number of ribosomes (Duncan and Hershey, 1983). The study of eIF4A pioneered the description of DEAD-box proteins among a broader class of DNA and RNA helicases dependent on ATP consumption (Linder et al., 1989). DEAD-box helicases contain conserved motifs that mediate ATP binding and hydrolysis, RNA binding and duplex unwinding or both, and that are distributed over two RecA-like domains (Schütz et al., 2010). Pdcd4 inhibits eIF4A by blocking the RNA-binding surface on its C-terminal lobe (Chang et al., 2009). Mammals have two eIF4A isoforms, eIF4AI and eIF4AII, that share 91% sequence identity and that are interchangeably incorporated into eIF4F (Nielsen and Trachsel, 1988). Their relative expression differs among tissues and eIF4AI knockdown elevates the expression of the eIF4AII gene. However, eIF4AII cannot rescue translation upon eIF4AI suppression, suggesting that the two isoforms are not functionally equivalent (Galicia-Vázquez et al., 2012). There is a third isoform of eIF4A that is more distantly related to the others (65% identity with eIF4AI) (Li et al., 1999). eIF4AIII functions as an ATP-dependent RNA helicase involved in the catabolism of aberrant

14 mRNAs (Shibuya et al., 2004) although this document refers to the isoforms of eIF4A that partake in translation initiation simply as eIF4A. eIF4A is a 46 kDa protein made of two globular RecA-like domains with a flexible attachment (Story and Steitz, 1992). The flexibility of the helicase has complicated the efforts of crystallographers who were interested in the structure of the protein and hence its first descriptions were of its domains in isolation (Benz et al., 1999; Caruthers et al., 2000; Johnson and McKay, 1999). The structure of the full-length protein was first captured in the 'open' conformation, where the extended conformation of the linker provides eIF4A with a 'dumbbell' shape, and eventually in the 'closed' conformation in complex with yeast eIF4G, in which the two RecA-like domains come closer to bind the extremities of MIF4G and form an ATP-binding cleft (Caruthers et al., 2000; Schütz et al., 2008). MIF4G promotes the 'closed' conformation by acting like a 'soft clamp' that guides eIF4A`s conformational changes by stabilizing the C-terminus of eIF4A and accommodating or releasing the N-terminal RecA-like domain (Oberer et al., 2005). DEAD-box helicases are directly recruited to their duplex substrates, aided by single-stranded regions in their vicinity (Özeş et al., 2011; Yang and Jankowsky, 2006). The general mechanism of DEAD-box helicases begins by the 'open' conformation, which can bind double-stranded RNA (Bono et al., 2006; Lorsch and Herschlag, 1998; Mallam et al., 2012) (Figure 1-4A). Upon ATP binding, the helicase adopts the 'closed' conformation that is associated with strand displacement so that it becomes bound to single-stranded RNA (Oberer et al., 2005; Schütz et al., 2008). In the case of eIF4A and other DEAD-box helicases, it is the binding and not the hydrolysis of ATP that causes its conformational change (Del Campo et al., 2009) (Figure 1-4B). Instead, ATP hydrolysis modulates the substrate turnover. Through their bipartite interaction, eIF4G guides the conformational changes of eIF4A and facilitates the release of the phosphate, which limits the ATP hydrolysis cycle (Hilbert et al., 2011) (Figure 1-4C). Duplex unwinding by eIF4A is non-processive and bidirectional, since it occurs only a few base pairs at a time both in the 5'-3' and 3'-5' directions (Rogers et al., 1999; Rozen et al., 1990). Compared to other DEAD-box proteins, eIF4A by itself has a relatively weak affinity for RNA and this is reflected in its poor intrinsic ATPase and

15

erfaces erfaces

east eIF4A east eIF4A

re of y

eIF4A eIF4A (PDB entry 1FUU) with the RNA

'open' 'open'

terminal terminal domain of eIF4A (shown in teal). The

-

box protein Mss116p (PDB entry 4DB2). (B) The binding of ATP (shown in orange) entry (B) in(shown binding The orange) (PDB 4DB2). of box ATP Mss116p protein

-

conformational changes of eIF4A of changes conformational

ns ns the unwinding cycle in the 'open' conformation. The interaction of

5: Interactions that guide the Interactions 5:

-

e e transition between the 'open' and 'closed' conformations upon binding the N

Figure1 duplex was modelled based on of the based the duplex DEAD modelled was structure drives th representation shown in is transparency the same as in (A) and the 'closed' conformation was taken from the cocrystal structu and eIF4G (PDB entry 2VSO). (C) The MIF4G (shown in purple) stabilizes the 'closed' conformation of eIF4A 2VSO). entry (PDB via two binding int (A) The helicase begi

16 helicase activities (Abramson et al., 1988; Rogers et al., 1999). The activity of eIF4A increases when associated to RNA-binding proteins eIF4G, and eIF4B or eIF4H, partly by promoting its recruitment to RNA substrates (Méthot et al., 1994; Pestova et al., 1996b; Richter-Cook et al., 1998). When eIF4A or eIF4F are supplemented with eIF4B and/or eIF4H, there is an increase in rate and amplitude of duplex unwinding in vitro that depends on local duplex stability (Özeş et al., 2011; Rogers et al., 2001). eIF4B and eIF4H are related proteins that bind eIF4A and RNA (Rozovsky et al., 2008). The functions of eIF4B and eIF4H seem complementary in vitro but expression studies suggest that the two cofactors have tissue-specific functions (Richter-Cook et al., 1998; Richter et al., 1999; Rogers et al., 2001). eIF4B has one additional RNA-binding site and a stronger ability to stimulate eIF4A activity compared to eIF4H, at least in vitro (Özeş et al., 2011). The stimulation of eIF4A by eIF4B results not only from its capacity to bind RNA, but also from the improvement of the coupling of ATP hydrolysis with strand separation. There is more data available about the function of eIF4B and this is why the rest of this dissertation will focus on eIF4B rather than eIF4H. eIF4F is essential for scanning structured mRNA 5' UTRs and cannot be substituted by eIF4A and eIF4B (Pestova and Kolupaeva, 2002). The association of eIF4A, eIF4B and RNA is very dynamic and is promoted by eIF4G in such a way that the optimal unwinding efficiency is reached in the presence of eIF4G (Nielsen et al., 2011; Özeş et al., 2011). In addition to the stimulation of its enzymatic activity, the association of eIF4G with eIF4A seems to provide the enzyme with distinct properties. While eIF4A alone and together with its cofactors are tolerant for DNA-RNA hybrid substrates, eIF4F shows a greater substrate-specificity by preferring to unwind RNA duplexes with a slight bias for the 5'-3' direction, which is coherent with the direction of mRNA scanning by the small ribosomal subunit (Rogers et al., 2001). eIF4A contains only the minimal helicase core whereas DEAD-box helicases usually contain appendages governing their activity (Andreou and Klostermeier, 2012). eIF4B and eIF4G have a strong influence on its enzymatic activity. Rather than eIF4A alone, the complex of eIF4A with eIF4B and eIF4G is the true effector of RNA unwinding by eIF4F. An advantage of having separate rather than integral modulatory

17 elements is that their association is reversible and subjected to the regulation of each eIF4F components.

1.2 Versatility of eIF4G family and eIF4G-like proteins The properties of eIF4G manifest themselves in many ways along translation initiation. The domains of eIF4G are found in different combinations in eIF4G and other non-eIF4G proteins, thereby diversifying their functions in translation initiation and other cellular processes.

1.2.1 Occurrence of eIF4G domains in other proteins MIF4G, MA3 and W2 are subtypes of HEAT domains that are found in different combinations in proteins other than eIF4G, as summarized in Table 1-2 (Craig et al., 1998; Mendell et al., 2000; Petit et al., 2012; Marintchev and Wagner, 2005; Gunawardena et al., 2008, Asano et al., 2009).

Table 1-2: Occurrence of eIF4G domains in other proteins

Protein Sequence identity with eIF4GI (%) (human) MIF4G (732-1003) MA3 (1234-1427) W2 (1427-1599) Paip1 27 RENT2 20 NOT1 18 CBP80 16 7 15 NOM1 16 10 eIF5 21 eIF2Bε 21 Pdcd4 22, 25

As mentioned earlier, Pdcd4 inhibits translation by binding eIF4A and blocking its interaction with RNA. The MA3 domains of Pdcd4 and eIF4G are structurally similar and were suggested to have overlapping binding sites on eIF4A (Bellsolell et al., 2006; Chang et al., 2009; LaRonde-LeBlanc et al., 2007). A crystal structure of Pdcd4 show that the two MA3 domains can each accommodate one molecule of eIF4A, in competition with the eIF4G/MA3-eIF4A interaction (Chang et al., 2009; Suzuki et al.,

18 2008). In contrast, the binding sites of eIF4G/MIF4G (PDB entry 2VSO) and Pdcd4/MA3 (from Pdcd4, PDB entry 2ZU6) on eIF4A do not overlap. A model based on the structure of the two individual interactions reveals that eIF4G/MIF4G and Pdcd4/MA3 could interact simultaneously with the C-terminal domain of eIF4A and exchange the N-terminal domain of eIF4A (Chang et al., 2009; Schütz et al., 2008) (Figure 1-6A). There is biochemical evidence for the formation of such a Pdcd4/MA3- eIF4A-eIF4G/MIF4G complex (Yang et al., 2003). This suggests that Pdcd4 can capture eIF4F in an unwinding-incompetent state by precluding the interaction of eIF4A with RNA. Therefore, Pdcd4 could target eIF4A in the free and bound states (Figure 1-6B). In another line of thought, it has been suggested that eIF4A could interact simultaneously with its two binding sites on eIF4G, similarly to the Pdcd4/MA3-eIF4A- eIF4G/MIF4G complex that was just described (Morino et al., 2000) (Figure 1-6C). So far, there is no experimental evidence that such a "sandwich" is formed between the MIF4G and MA3 domains of eIF4G (G. Virgili, unpublished results). This putative eIF4G conformation would not be competent for RNA unwinding since eIF4G/MA3 would block eIF4A from binding RNA similarly to Pdcd4/MA3. Thus, based on our model in Figure 1-6C, it can be speculated that the MA3 domain of eIF4G could modulate the interaction of eIF4F with RNA (Figure 1-6D).

1.2.2 Role of the MA3 domain of eIF4G The MA3 domain of eIF4G constitutes a second binding site for eIF4A that is not universal (Imataka and Sonenberg, 1997) (Table 1-3). The yeast homologs TIF 4631/2 are shorter than human eIF4G and they lack the region encompassing the MA3 and W2 domains whereas the wheat homologs only lack the W2 domain. Interestingly, the MA3 domain of DAP5 does not bind eIF4A. Therefore, we can conclude that MIF4G but not MA3 and W2 is essential for the function of eIF4G. Until recently, the mechanistic details of the interplay of eIF4A with its two binding sites were ill-defined. Morino et al. (2000) have suggested previously that MA3 could hinder the eIF4A-binding site on the MIF4G domain of eIF4G so that eIF4A would be first recruited to eIF4G`s MA3 domain and then transferred to its MIF4G domain

19 20 Table 1-3: Occurrence of eIF4G domains in eIF4G proteins

eIF4G Sequence identity with eIF4GI (%) Organism Protein MIF4G (732-1003) MA3 (1234-1427) W2 (1427-1599) eIF4GII 81 64 64 human DAP5 42 26 32 eIF4G 45 31 wheat eIFiso4G 36 25 TIF 3641 34 yeast TIF 3642 33

(Figure 1-6D). This hypothesis was challenged by the small-angle X-ray scattering (SAXS) study of a protein construct encompassing the two eIF4A-binding sites of eIF4G. The structural analysis revealed that the MA3 domain does not hinder the eIF4A-binding site of the MIF4G domain and that when eIF4A is added, it locates to MIF4G

Figure 1-6: Can eIF4A interact simultaneously with the MIF4G and MA3 domains? The N-terminal domain of Pdcd4 is shown in pink and the other molecules follow the same color code as in Figures 1-3 and 1-5. (A) This structural model was constructed from the cocrystal structures of Pdcd4/nMA3-eIF4A and eIF4G/MIF4G-eIF4A (PDB entries 2ZU6 and 2VSO). Both the MIF4G and MA3 domains form more contacts with the C-terminal lobe than the N- terminal lobe of eIF4A. Thus, the structural model was generated by aligning the eIF4A C- termini from the two crystal structures. The binding modes of Pdcd4`s N and C-terminal MA3 domains to eIF4A are equivalent so that its N-terminal MA3 was used to represent both. (B) Cartoon representation of the possible modes of eIF4A-inhibition by Pdcd4. (C) eIF4G/MA3 (PDB entry 1UG3) substitutes the Pdcd4/nMA3 region of the structural model shown in (A) to model the interaction between eIF4A and eIF4G. If the two interactions occur simultaneously, eIF4A would be "sandwiched" between the MIF4G and MA3 domains of eIF4G. (D) The association of eIF4G/MIF4G with eIF4A mediates RNA unwinding and Morino et al. (2000) have suggested that eIF4G/MA3 must play a role in the formation of this functional complex. Does eIF4A interact simultaneously with its two binding sites on eIF4G? In that case, we can speculate that eIF4G/MA3 domain would inhibit RNA unwinding by blocking eIF4A from binding RNA. Another possibility is that eIF4G/MIF4G is sterically hindered so that eIF4A is first recruited to eIF4G/MA3 before its transfer to the MIF4G.

21 (Nielsen et al., 2011). A major inconvenience associated with the study of eIF4G is its instability. To circumvent that problem, recombinant eIF4G is used for in vitro experiments. The subcloning of eIF4G usually excludes its N-terminal third since its flexibility increases its susceptibility to degradation. In the case of the abovementioned SAXS study, the construct of eIF4G encompassed only its two first HEAT domains since the aim was to study the interaction of eIF4A with MIF4G and MA3. Structured (e.g. HEAT domains) as well as unstructured regions (e.g. N-terminal third) contained in eIF4G form an extended network of interactions with initiation factors that are instrumental to translation initiation. Protein kinase C (PKC), casein kinase (Csk) 2 and extracellular signal-regulated kinases (Erk) modulate the access of initiation factors to eIF4G by subjecting the ~200 residue-long linker connecting the MIF4G and MA3 domains to post-translational modifications. In the native protein, the linker binds the W2 domain and blocks its access to Mnk while the MA3 domain remains accessible to eIF4A. Upon PI3K activation, PKC phosphorylates the linker at Ser 1185 to release the W2 domain and allow the recruitment of Mnk (Dobrikov et al., 2011). Then, Csk2 and Erk sequentially phosphorylate the linker at its junction with the MA3 domain (Ser 1238 and 1231), causing its destabilization and leading to the displacement of eIF4A from the MA3 domain potentially in favor of its interaction with the MIF4G domain (Dobrikov et al., 2013). The rearrangement of those intramolecular interactions promotes translation initiation and supports the hypothesis postulated by Morino et al. (2000), by which eIF4A would first be recruited to the MA3 domain. Dobrikov et al. (2013) described elegantly that post-translational modifications regulate the interaction of eIF4A with its two binding sites on eIF4G (Figure 1-8A). They have provided an interesting mechanism by which the native eIF4G is self-inhibitory but can deploy an array of functional interactions when translation initiation is activated by mitogens and growth factors. The presence of the W2 domain enabled the description of this complex network of regulatory interactions that was not observable in the experimental setup used in the SAXS study by Nielsen et al. (2011). Yet, the absence of eIF4G N-terminus excludes its interplay with the rest of the protein so that there might be additional regulatory interactions influencing eIF4G function.

22 1.2.3 Alternative modes of ribosome recruitment The recruitment of the ribosome to mRNA is a critical step in translation initiation. Beforehand, the assembly of the translation machinery on the mRNA generally depends on the interaction of eIF4E with the mRNA 5' cap (Sonenberg, 1981). Translation can also occur on uncapped mRNA as long as the mRNA contains a 5' UTR to accommodate eIF4G, the middle region of which is sufficient to coordinate the recruitment of the small ribosomal subunit (De Gregorio et al., 1998). This type of interaction of eIF4G with RNA is unspecific, but eIF4G is also known to interact specifically with mRNAs that contain IRES elements in their 5' UTR (Prévôt et al., 2003). Those specific interactions have best been described with viral IRES but were also described in a growing number of cellular mRNAs (Lomakin et al., 2000; Baird et al., 2006). An extended network of protein-protein interactions synchronizes translation with biological processes. Cap-dependent translation is downregulated during viral infection, mitosis and apoptosis (Bonneau and Sonenberg, 1987a; 1987b; Marissen et al., 2000). During those processes, alternative mechanisms of ribosome recruitment are utilized to maintain the translation of specific mRNAs in order to fulfill the needs of the cell or to support viral replication.

1.2.3.1 Viral infection Viruses are made up of a protein capsid that surrounds their DNA or RNA-based genome. The viral genome contains some of the components required for viral replication but the synthesis of viral proteins depends on the host translation machinery. Viruses have evolved different strategies to compete with endogenous mRNAs for the synthesis of their own proteins. The translation of some viral RNAs occurs independently from initiation factors that are limiting cap-dependent translation. IRES elements can substitute for the interaction of eIF4E with the mRNA cap, notably in picornavirus, hepatitis C virus (HCV) and dicistrovirus mRNAs (Pestova et al., 1996a). The latter two have the least initiation factor requirements and can bind the small ribosomal subunit directly (Pestova et al., 1998; Wilson et al., 2000). In contrast, the initiation factor requirements for translation initiation on picornaviral IRES are similar to those of cellular initiation.

23 The tactics employed by picornaviruses have been extensively studied. Strategies to shutoff host protein synthesis include the cleavage of eIF4GI and II by picornaviral proteases upon infection with poliovirus, rhinovirus and foot-and-mouth disease virus. The viral proteolysis of the two eIF4G proteins causes the separation of their PABP and eIF4E-binding sites from the rest of the protein so that their C-terminal product can only support cap-independent translation (Borman et al., 1997; Lamphear et al., 1993) (Figure 1-8B). eIF4GI is completely cleaved as of 4 hours post-infection. The inhibition of host translation correlates with the cleavage of eIF4GII, which in turn begins 5 hours after infection (Gradi et al., 1998; Svitkin et al., 1999). In contrast, cardioviruses (such as encephalomyocarditis virus, or EMCV) inhibit cellular translation by causing the dephosphorylation of 4E-BP1 (Gingras et al., 1996). The consequential decrease in eIF4E availability favors internal entry over cap-dependent initiation (Svitkin et al., 2005). As mentioned earlier, the presence of structural elements in the 5'-UTR limits cap-dependent translation efficiency because their unwinding is necessary for the small ribosomal subunit to reach the start codon. In contrast, the structural elements forming IRESs allow the recruitment of the small ribosomal subunit in the vicinity of the start codon. The direct association of the central third of eIF4G to the 5'-UTR of an mRNA is sufficient to recruit the 43S PIC (Pestova et al., 1996b; De Gregorio et al., 1999). For example, the J and K domains of EMCV IRES form the minimal core to bind MIF4G (Figure 1-7). eIF4A synergizes with MIF4G to bind the J-K domains and is instrumental for remodeling their conformation to allow the recruitment of the 43S PIC (Lomakin et al., 2000; Kolupaeva et al., 2003). The interest in picornaviral IRES translation originates from its similarity to cellular IRES translation in terms of initiation factor requirements so that both mechanisms are often described in parallel.

1.2.3.2 Cap-independent translation of cellular mRNAs Cap-dependent translation is also inhibited during mitosis and apoptosis although the translation of certain mRNAs is thought to be selectively maintained due to IRES elements under those stress conditions.

24

Figure 1-7: Simplified representation of internal entry on EMCV IRES (adapted from Hellen & Sarnow, 2001)

Cellular division requires the disruption of the nuclear membrane to split its released content between the two daughter cells. The inhibition of cap-dependent translation during mitosis prevents the translation of immature mRNAs found in the nucleus. The phosphorylation of eIF4GII interferes with eIF4E binding and the hypophosphorylated 4E-BPs sequesters eIF4E, leading to the decrease in eIF4F formation that is observable during mitosis (Olsen et al., 2010; Pyronnet et al., 2001) (Figure 1-8C). The synthesis of proteins that may have mitotic functions is ensured by IRES elements located in their mRNAs (Qin and Sarnow, 2004). Apoptosis, or programmed cell death, is a phenomenon that exists in multicellular organisms to ensure their wellness by the self-elimination of unnecessary or potentially harmful cell. This mechanism is essential for growth, differentiation and development and can be triggered by events inside or at the surface of the cell such as starvation, DNA damage and the activation of extracellular receptor for Fas or tumor necrosis factor-α (TNFα) (Clemens et al., 2000). Signaling cascades lead to the inhibition of translation initiation by the phosphorylation of eIF2α (Holcik and Sonenberg, 2005). In parallel, the formation of a multi-protein complex (termed apoptosomes) triggers a proteolytic cascade that culminates in the activation of serine proteases known as caspases. Similar to the picornaviral proteases, caspase-3 degrades eIF4GI and II distinctively (Marissen et al., 2000). eIF4GI is cleaved between its binding sites for PABP and eIF4E and between

25

Figure 1-8: eIF4G proteins come in a variety of flavors

26 its two eIF4A-binding sites so that its central product can robustly support cap-dependent and independent translation initiation without the regulatory domains MA3 and W2 (Bushell et al., 2006) (Figure 1-8D). This is important as some types of apoptosis require de novo protein synthesis. The cleavage sites are more numerous on eIF4GII, yielding smaller fragments that do not retain any activity so that the degradation of eIF4GII correlates with the inhibition of translation during apoptosis.

1.3 Death-Associated-Protein 5 (DAP5) The cell-cycle progression as well as the stress conditions underlying apoptosis cause a transition from cap-dependent to the cap-independent translation of specific mRNAs (Qin and Sarnow, 2004; Holcik et al., 2000). Death-associated protein 5 (DAP5) is an eIF4G protein that mediates the cap-independent translation of mRNAs whose protein products play a role in the balance between cell survival and death (Marash and Kimchi, 2005). The eIF4G paralog was first described in 1997 and can also be referred to as p97, eIF4G2 and NAT1 (Imataka et al., 1997; Levy-Strumpf et al., 1997; Shaughnessy et al., 1997; Yamanaka et al., 1997). It supports the translation of a specific subset of cellular IRES although the molecular determinants of its specificity remain to be elucidated.

1.3.1 Topology of DAP5 DAP5 lacks the binding sites for eIF4E and PABP and therefore cannot partake in cap- dependent translation initiation. The protein has 907 residues and its synthesis is initiated at a GUG codon, coding for Val (Imataka et al., 1997). Triplets that differ from AUG at Met one position are also recognized by Met-tRNAi and initiate translation with Met albeit less efficiently (Peabody, 1989). This does not seem to affect DAP5 synthesis since the protein is found at similar levels as eIF4GI (Lee and McCormick, 2006). The structure of DAP5 contains the same three HEAT domains as eIF4G: MIF4G (termed DAP5M hereafter), MA3 and W2 (Figure 1-8E). DAP5M is essential for the colocalization of DAP5/p97 with the ribosome and probably plays a role similar to eIF4G/MIF4G (Lee and McCormick, 2006). DAP5 contains a single binding site to eIF4A since its MA3 domain lacks the ability to bind eIF4A (Imataka and Sonenberg, 1997). Similarly to eIF4G, the W2 domain

27 of DAP5 interacts with Mnk. The W2 domain of DAP5 also resembles those found in eIF5 and eIF2Bε and thus can interact with the beta subunit of eIF2 (Lee and McCormick, 2006; Pyronnet et al., 1999). Large-scale mass spectrometry analyses lead to the identification of phosphorylation sites on DAP5 and those are listed in Table 1-4. The NetPhosK 1.0 server (http://www.cbs.dtu.dk/services/NetPhosK) was used to predict the kinases responsible for those modification based on the analysis of the target sequences. It is not surprising that most DAP5 phosphorylation sites are not conserved in eIF4G (data not shown) since the two proteins have different properties and are implicated in different mechanisms of translation. Nonetheless, this suggests that the activity of DAP5 is also influenced by post-translational modifications.

Table 1-4: Experimental phosphorylation sites on DAP5

DAP5 Mitosis Differentiation Putative Kinases References

S395  p38 MAPK, CDK5 Olsen et al. (2010)

T508  CDK5 Mayya et al. (2009)

T514  p38 MAPK, CDK5 Dephoure et al. (2008)

S902   Csk1/2 Rigbolt et al. (2011)

1.3.2 Function in Cap-Independent Translation Initiation The expression of DAP5 is ubiquitous and this initiation factor is essential for mitosis, cell differentiation, proliferation and apoptosis (Henis-Korenblit et al., 2002; Lee and McCormick, 2006; Marash et al., 2008; Yamanaka et al., 2000). Unlike eIF4G, DAP5 does not participate in the translation of viral IRESs (Imataka et al., 1997; Yamanaka et al., 1997). Originally, DAP5 was thought to inhibit cap-dependent translation by titrating initiation factors in a complex excluding eIF4E (Imataka et al., 1997; Pyronnet et al., 1999; Yamanaka et al., 1997). The overexpression of DAP5 artificially induces a decrease in translation that is probably attributable to the out- titration of Mnk and/or eIF2β since translation levels are not affected by the overexpression of an isoform of DAP5 that cannot bind the two initiation factors (Henis- Korenblit et al., 2002).

28 DAP5 is a translation activator that is controlled by mTOR and that colocalizes with polysomes in proliferating cells (Gao et al., 2012; Lee and McCormick, 2006; Nousch et al., 2007). Knockdown and knockout studies have illustrated the role of DAP5 in the total protein synthesis and more particularly to the translation of cyclin-dependent kinase 1 (CDK1) and members of the B cell lymphoma (Bcl) family of proteins (Marash et al., 2008; Nousch et al., 2007). CDK1 activates the anti-apoptotic proteins Bcl-xL and Bcl-2 by phosphorylation. The activation of CDK1 is normally transient during mitosis but its persistence during mitotic arrest causes the hyperphosphorylation of the Bcl proteins and the inactivation of their anti-apoptotic function (Terrano et al., 2010). Hence, DAP5 maintains the synthesis of key proteins that determine the cell fate during mitosis (Liberman et al., 2009). DAP5 is involved in the cellular response to apoptotic stress (Marash and Kimchi, 2005). Caspase activation leads to the cleavage of DAP5 between its MA3 and W2 domains (Liberman et al., 2008). The longer of the two products of caspase-3 proteolysis is termed DAP5/p86 and lacks the binding sites for Mnk and eIF2β. Caspase-activated DAP5 can support the translation of the IRES found in its own mRNA, creating a positive feedback loop for the synthesis of the protein under apoptotic stress (Henis- Korenblit et al., 2000). DAP5/p86 also activates the translation of the IRESs found in the mRNAs of Apaf-1, c-myc and inhibitor of apoptosis proteins (IAPs) (Henis-Korenblit et al., 2002; Nevins et al., 2003; Warnakulasuriyarachchi et al., 2004). As mentioned earlier, eIF4GI is subjected to caspase proteolysis, leading to the isolation of its central functional core that can also support the internal initiation to the IRESs of DAP5, c-myc and IAPs (Hundsdoerfer et al., 2005). Upon caspase activation, both eIF4G proteins are involved in the translation of pro and anti-apoptotic mRNAs and thus contribute to the survival and the death signals that determine the fate of stressed cells. Another target of DAP5 was recently described. p53 acts as a tumor suppressor and is implicated in an astonishing number of cellular processes, notably in mitosis and apoptosis (Levine and Oren, 2009). The mRNA of p53 contains two IRES elements that direct the translation of its two isoforms, p53 and the shorter Δ40p53, which bhave distinct biological activities (Yang et al., 2006). Earlier this year, Weingarten-Gabbay et al. (2013) were the first to report the direct interaction of DAP5 with the IRESs of p53.

29 The weak interaction was captured by RNA-protein UV crosslinking and analyzed by electrophoretic mobility shift assay (EMSA). The direct interaction of DAP5 with IRESs as it was observed for eIF4G suggests that the two proteins may share a similar mechanism for IRES-mediated translation (Lomakin et al., 2000).

1.3.3 Three-Dimensional Structure The structure of the isolated W2 domain (PDB entry 3D3M) of DAP5 uncovered the molecular details of DAP5 cleavage by caspase-3 (Liberman et al., 2008). The W2 domain is very similar to its homolog in eIF4G (PDB entry 1UG3) although the loop containing the caspase cleavage site is longer in DAP5 and its electron density is missing from the crystal structure (Bellsolell et al., 2006). The local flexibility facilitates the recognition of the peptide sequence DETD792 by caspase-3, in contrast to another putative cleavage site (DHVD825) that is confined in a rigid region. Proteolysis leads to the separation of two subdomains, exposing a hydrophobic surface on the C-terminus of DAP5/p86 and potentially causing the rearrangement of intra and extramolecular interactions. Another crystal structure of the C-terminal region of DAP5 shows that MA3 and

W2 form a continuous platform due to their short and rigid attachment by a 310-helix (PDB entry 3L6A) (Fan et al., 2010) (Figure 1-9A). In contrast, the second and third HEAT domains of eIF4G are connected by a flexible linker that allows the domains to move with respect to each other (Bellsolell et al., 2006). The flexibility of eIF4G is presumably important for the rearrangement of its intramolecular interactions as described by Dobrikov et al. (2013). In turn, the rigidity of DAP5 might limit the dynamics of its intramolecular interactions so that the protein escapes the C-terminus- mediated autoinhibition described for eIF4GI. Another interesting dissimilarity between the C-termini of the two proteins is that DAP5/MA3 contains multiple positively-charged residues. The same surface on eIF4G/MA3 is thought to bind positively-charged residues on eIF4A that would otherwise interact with RNA, which is negatively charged (Figure 1-9B). In contrast, the positively-charged surface on DAP5/MA3 is not complementary to the RNA-binding residues of eIF4A, explaining the incompatibility between DAP5/MA3 and eIF4A (Figure 1-9C). The particular charge distribution on DAP5/MA3 favors the

30

Figure 1-9: Properties of DAP5 and eIF4G C-termini (A) The two C-terminal HEAT domains of DAP5 (PDB entry 3L6A) form a continuous platform whereas they have a flexible attachment in eIF4GI (PDB entry 1UG3). Electrostatic potential surface representation of DAP5 and eIF4GI C-termini. Positive charges are shown in blue and negative changes in red. (B) In analogy to Pdcd4, the MA3 domain of eIF4G would bind the RNA-binding surface of eIF4A C-terminal lobe. (C) The positively charged residues found at the same position on DAP5 could explain the lack of binding to the positively-charged region of eIF4A.

31 functional interaction of eIF4A with DAP5M over the autoinhibitory interaction with MA3. This further supports that DAP5 is not subjected to the same type of autoinhibition as observed for eIF4G.

1.4 Rationale of the study eIF4G interacts directly with IRES elements via MIF4G, and eIF4A strongly enhances this interaction (Lomakin et al., 2000). Knowing the similarities between DAP5 and eIF4G-mediated IRES translation, we can predict that DAP5M might also interact directly with IRES elements and that eIF4A might also stimulate that interaction. An obvious question that arises from the crystal structure illustrating the interaction of yeast eIF4G with eIF4A is whether DAP5 binds eIF4A in the same manner. We addressed this issue by solving the crystal structure of DAP5M and by using different biophysical techniques to study its interaction with eIF4A. Our findings are described in the next chapter.

32

Chapter 2: Structural analysis of the DAP5 MIF4G domain and its interaction with eIF4A

Virgili, G.*, Frank, F.*, Feoktistova, K., Sawicki, M., Sonenberg, N., Fraser, C. S., and Nagar, B. (2013). Structure 10.1016/j.str.2013.01.015. *These authors contributed equally to this work.

33 2.1 Introduction Translation initiation is the rate-limiting step of protein synthesis, and involves assembly of the ribosome on the mRNA followed by recognition of the start codon. Initiation can be carried out in two different ways: the canonical mode of translation initiation is cap- dependent and proceeds by assembly of the eIF4F complex on the mRNA 5' m7GpppX cap (where X is any nucleotide) structure and subsequent formation of the pre-initiation complex containing the 40S ribosomal subunit (Sonenberg and Hinnebusch, 2009). An alternative mode of translation initiation is cap-independent and involves access of the ribosome to the mRNA via an internal ribosomal entry site (IRES) typically found in the 5' untranslated region (5' UTR) (Holcik and Sonenberg, 2005). The IRES recruits the ribosome directly without the need for the mRNA cap or eIF4E. The eIF4F complex consists of the cap-binding protein eIF4E, the scaffolding protein eIF4G, and an ATP-dependent RNA helicase eIF4A whose RNA duplex unwinding and ATP hydrolysis activities are coupled and stimulated by eIF4B and eIF4G (Özeş et al., 2011). eIF4G is a large 175 kDa protein with interaction sites for its binding partners spread over multiple domains (Figure 2-1A). There are two isoforms of eIF4G, eIF4GI and eIF4GII, that share 46% sequence identity; in this paper, we refer to both isoforms as 'eIF4G' unless indicated otherwise. In addition to recruiting eIF4E and eIF4A, eIF4G interacts with a number of other factors required for efficient translation, including the 40S ribosome associated eIF3, the poly(A)-binding protein (PABP), and the Ser/Thr kinase Mnk-1 (Prévôt et al., 2003) (Figure 2-1A). PABP connects eIF4F with the poly(A) tail and thus circularizes the mRNA for increased translational efficiency. Mnk-1 phosphorylates eIF4E, which stimulates translation (Furic et al., 2010). The middle domain of eIF4G, termed the MIF4G domain, carries out a number of important functions. This segment of approximately 30 kDa mediates protein-protein interactions with eIF4A and eIF3, and also displays RNA and DNA binding capabilities (Ponting, 2000). It has been shown to interact directly with the IRES element of the encephalomyocarditis virus (EMCV) RNA (Pestova et al., 1996; Lomakin et al., 2000) and allows eIF4G to recruit the ribosome to the EMCV RNA in a cap-independent manner by interacting with eIF3 and the RNA at the same time. EMCV is a member of the picornavirus family of viruses, which have the ability to shutdown cap-dependent

34

Figure 2-1. Domain structure and sequence alignment of DAP5 and eIF4G (A) DAP5 and eIF4GI domain organization highlighting their similarities. Approximate protease cleavage locations are indicated with arrows. (B) Structure-based sequence alignment of MIF4G domains of human DAP5, human eIF4GII, and TIF4631 (S. cerevisiae eIF4G). Residues interacting with eIF4A are highlighted in red. The remainder of the sequence is colored based on conservation using BLOSUM62 scores (Eddy, 2004). Secondary structure elements are shown over the sequence aligned and identified using the same color-code as in Figure 2-2A.

35 translation initiation in the host. One example of such a mechanism commonly employed by picornaviruses is proteolytic cleavage of eIF4G resulting in the separation of the eIF4E and PABP binding sites from the rest of the protein and concomitant blockage of cap-dependent translation (Figure 2-1A). This permits the virus to hijack the translational machinery to efficiently translate its own RNA via IRES interacting elements in the MIF4G domain. Cleavage of eIF4G can also occur in non-infected cells during the process of apoptosis where two caspase cleavage sites in eIF4GI break it into three fragments, thereby impairing cap-dependent translation (Bushell et al., 2000) (Figure 2-1A). Despite this, it has been observed that the translation of a few specific mRNAs is maintained during apoptosis. The translation of these mRNAs occurs via cap-independent mechanisms that is at least to some extent, modulated by the protein DAP5/p97/NAT1/eIF4G2 (Henis-Korenblit et al., 2000; Lewis et al., 2008; Nevins et al., 2003; Warnakulasuriyarachchi et al., 2004) [DAP5 (Levy-Strumpf et al., 1997); p97 (Imataka et al., 1997); NAT1 (Yamanaka et al., 1997), eIF4G2 (Shaughnessy et al., 1997)]. In this paper, we refer to this protein using the ‘DAP5’ nomenclature. DAP5 is homologous to the C-terminal two-thirds of eIF4G and is similar in length to the fragment generated by picornaviral proteolysis (Figure 2-1A). It is abundantly expressed in proliferating cells (Lee and McCormick, 2006) and contains the important MIF4G domain as well as the MA3 domain. Consistent with its homology to eIF4G, DAP5 interacts with eIF4A and eIF3, but not eIF4E (Imataka et al., 1997; Imataka and Sonenberg, 1997; Yamanaka et al., 1997). However, in contrast to eIF4G, the MA3 domain of DAP5 does not support eIF4A binding (Imataka and Sonenberg, 1997), leaving DAP5 with a single interaction domain for eIF4A as compared to eIF4G's two binding domains. Due to the lack of an eIF4E binding site, DAP5 is not involved in cap-dependent translation and was reported to be an inhibitor of translation based on overexpression studies (Imataka et al., 1997; Yamanaka et al., 1997). Later it was demonstrated that DAP5 mediates the IRES-driven translation of a number of cellular mRNAs. For example, during apoptosis and other stress conditions DAP5 specifically enhances the IRES-driven translation of several mRNAs including those coding for the proteins c-

36 IAP1/HIAP2, XIAP, Apaf-1, c-myc and DAP5 itself (Henis-Korenblit et al., 2002; Henis-Korenblit et al., 2000; Hundsdoerfer et al., 2005; Lewis et al., 2008; Nevins et al., 2003; Warnakulasuriyarachchi et al., 2004). During apoptosis DAP5 is cleaved close to the C-terminus near its Mnk-1 binding site generating an 86 kDa fragment that is more active (Figure 2-1A) (Henis-Korenblit et al., 2000). In addition to its role in apoptosis and stress, DAP5 has also been shown to promote cap-independent translation of cell survival factors during mitosis in unstressed cells (Liberman et al., 2009; Marash et al., 2008). Its reported targets include CDK1 and members of the Bcl family of proteins. The involvement of DAP5 in processes aiding apoptosis, and in cell survival during mitosis, renders it an interesting target for therapy of aberrant cellular states characterized by the dysregulation of apoptosis such as cancer and autoimmune diseases. The MIF4G domain of eIF4G and DAP5 is central to their function since it provides a platform for the interaction with eIF4A and eIF3, which are critical for translation initiation. The MIF4G domain of eIF4G was reported to be sufficient to mediate IRES-driven translation initiation (De Gregorio et al., 1998, 1999; Hundsdoerfer et al., 2005; Lomakin et al., 2000) suggesting that the IRES binding determinants of DAP5 may also lie within its MIF4G domain. Surprisingly however, in spite of their homology and shared functionality, eIF4G and DAP5 can act upon different sets of IRES elements, as in the case of EMCV (Nevins et al., 2003). The MIF4G domain of DAP5 share 39% and 43% sequence identity with eIF4GI and eIF4GII, respectively (Figure 2- 1A). The crystal structure of the MIF4G domain from human eIF4GII revealed a HEAT domain consisting of 5 pairs of HEAT repeats of antiparallel α-helices forming a crescent-shaped right-handed solenoid (Marcotrigiano et al., 2001). Mutational analyses mapped the sites of interaction with eIF4A and potential contacts with IRES RNA to adjacent surfaces on the molecule (Marcotrigiano et al., 2001). There is as yet no structural information on MIF4G's binding to IRESs. The crystal structure of the yeast eIF4G-eIF4A complex showed that the convex surface of MIF4G makes contact with eIF4A, at both its N-terminal and C-terminal regions (Schütz et al., 2008). As in human, yeast also possesses two isoforms of eIF4G. The crystal structure contains isoform I, but is referred to as eIF4G throughout the text for simplicity. This structure also revealed a

37 third site of interaction: a conserved tryptophan residue (Trp 579 in yeast eIF4GI also known as TIF4631; Trp 734 in human eIF4GI; Trp 733 in human eIF4GII) at the N- terminus of the MIF4G domain anchored to the eIF4A C-terminal domain. Mutation of this residue in yeast eIF4G weakens its interaction with eIF4A and results in the loss of the ability to stimulate eIF4A ATPase activity in vitro and a temperature-sensitive phenotype in vivo (Schütz et al., 2008). This tryptophan residue and most of the residues in the MIF4G domain that make direct contact with eIF4A are conserved in DAP5. To elucidate the similarities and differences responsible for the crucial functional interactions of the MIF4G domains of DAP5 and eIF4G we solved the crystal structure of the DAP5 MIF4G domain (hereafter referred to as DAP5M) (Frank et al., 2010). DAP5M adopts the same overall fold as eIF4G, but with significant structural differences in some of the helices and their connecting loops that have potential implications for the distinct IRES binding properties of the two proteins. Conserved residues expected to interact with eIF4A are for the most part in the same conformation as seen for the yeast eIF4G-eIF4A complex and the binding properties of the complex it forms with eIF4AI was investigated by mutational analysis. Additionally, quantitative analysis of the affinity of DAP5M to eIF4A indicates that it is one order of magnitude weaker than that of eIF4GI to eIF4A, which likely underlies DAP5’s weaker stimulation of the RNA unwinding activity of eIF4A compared to eIF4GI.

2.2 Results 2.2.1 Overall structure of the DAP5 MIF4G domain, or DAP5M Based on the crystal structure of the middle domain of eIF4GII, we crystallized and determined the structure of a construct encompassing the middle domain of DAP5 (DAP5M; residues 61 to 323) at 2.3 Å resolution using molecular replacement. Subsequent model building, simulated annealing, energy minimization and individual B- factor refinement led to final Rfree and R values of 25.6% and 22.2%. Statistics of data collection and refinement are summarized in Table 2-1. DAP5M belongs to the family of HEAT (Huntingtin, Elongation factor 3, PR65/A, and TOR) domains, which are characterized by repeated pairs of anti-parallel -helices connected by turns/loops arranged about a common axis (Figure 2-2A). Each pair of helices (labeled a and b)

38

Table 2-1: Diffraction data collection and refinement statistics for DAP5M Data collection X-ray source CHESS Beamline A1 Wavelength (Å) 0.9785 Resolution (Å) 30-2.3(2.38-2.3) Space group C2 Cell parameters a=167.630, b=56.573, c=74.339, (Å, ) α=90.00, β=112.05, γ=90.00 Molecules per ASU 2 Mosaicity (°) 0.9 Unique reflections 28017 Redundancy 13.2(10.5) I/σ(I) 23.4(4.8) Completeness (%) 99.8(87.5)

Rsym 0.107(0.35) Refinement

Rwork/Rfree 22.2/25.6 Number of atoms 4038 Protein 3837 Ligand/ion 25 Water 176 R.m.s deviations Bond lengths (Å) 0.002 Bond angles () 0.620 Ramachandran, favored 99.1 % Ramachandran, outliers 0.2 %

Rmsd, root-mean-square deviations.

39

, ,

elical elical axes. The

eIF4G eIF4G (TIF4631, PDB

erevisiae

S. S. c

is is colored differently. Helix numbering is given on the right. In the left orientation,

.

3

-

Figure 2 Figure

d d by 30°, highlights the long loop between helices 2b and 3a (colored orange), which causes the molecule to be divided

Structural overview of overview Structural DAP5M

: :

2

-

2

Figure (A) Ribbon representation of DAP5M. Each HEAT repeat the concave surface of the protein is indicated with a curved line and colored cylinders within the ribbons denote the superh right orientation, rotate See subdomains. two into (B) Structural ofsuperposition DAP5M with the MIF4G domains of human eIF4GII (PDB ID 1HU3) and ID 2VSO). All molecular figures were generated using the program PyMol LLC). Schrödinger, (The PyMOL Molecular Graphics System, Version 1.2r3pre

40

Figure 2-3: There are two chains in the asymmetric unit of DAP5M The arrangement of the two DAP5M molecules in the unit cell gives rise to a noncrystallographic (NCS) 2-fold symmetry axis; A shows top and side views. Ribbon representation B of the overlap of DAP5M chains A and B; views were rotated 30° apart. Chains A and B are respectively coloured burgundy and blue.

41 constitutes one HEAT repeat and in DAP5M, 10 helices form 5 HEAT repeats (labeled 1 to 5), which are stacked on top of each other and stabilized by intervening hydrophobic interactions. Small rotations between the packing of successive HEAT repeats impart a twist to the overall structure, giving rise to a right-handed superhelical axis perpendicular to the repeat helical axes. The MIF4G domain can be subdivided into two smaller subdomains because of the presence of a long 21-residue loop connecting helices 2b and 3a. Thus, subdomain 1 encompasses HEAT repeats 1 and 2 and subdomain 2 includes HEAT repeats 3 to 5. A tilt of ~48° between these subdomains gives the molecule an overall crescent-shaped appearance with two distinctly shaped surfaces, a concave surface and a convex surface (Figure 2-2A). The asymmetric unit of the crystal contains two independent copies of DAP5M (chains A and B) related by a two-fold non-crystallographic symmetry axis (Figure 2-3). Superposition of the two copies using the Dali server indicates a root-mean-squared- deviation (r.s.m.d.) of 1.0 Å for 223 corresponding C atoms (Holm and Rosenström, 2010). The largest deviations occur at the N-terminus where molecule A is longer by 12 residues in the electron density, and at the C-terminus of helix 3a where the paths of the helices diverge considerably between the two molecules. At approximately residue Val 182, helix 3a of molecule B bends by a few degrees leading into the connecting loop to helix 3b. In molecule B, this connecting loop is well ordered and modeled in the structure, whereas in molecule A it is completely disordered. Molecule B is better defined in the electron density with regards to the longer loops in the structure, although the average overall backbone B-factors for both molecules are comparable (molecule A: 42.9 Å2, molecule B: 45.2 Å2). The two molecules bury a significant amount of surface area (~2500 Å2) at their interface, however, the residues in the interface are for the most part polar in nature and gel filtration analysis of DAP5M indicates a monomeric species in solution (data not shown). Thus, the dimer observed here is likely an artifact of crystallization and probably not physiologically relevant.

2.2.2 Comparison of DAP5M and eIF4G MIF4G domains

Two crystal structures of the MIF4G domain from eIF4G have been determined

42 previously: that from human eIF4GII and the structure of the S. cerevisiae eIF4G middle domain in complex with eIF4A (Marcotrigiano et al., 2001; Schütz et al., 2008). Human DAP5 shares 43% and 32% sequence identity (based on structure-based sequence alignments) with human eIF4GII and S. cerevisiae eIF4G, respectively, in their MIF4G domains and all of them adopt the same overall fold (Figure 2-2B). Superposition of DAP5M on human and yeast eIF4G using the Dali server indicates r.m.s.d. values of 1.7 Å and 2.6 Å based on 190 and 212 corresponding C atoms, respectively (Holm and Rosenström, 2010). However, there are significant differences observed in the length and orientation of a number of helices. Additionally, the loops connecting the helices differ considerably in length and conformation. In particular, the concave side of the molecule in the N-terminal region opposite the eIF4A binding site encompassing the helices of HEAT repeats 1, 2, and 3 and the loop connecting repeats 2 and 3 display very different conformations (Figure 2-2B). The loop connecting repeats 2 and 3 (residues 142 to 161) is 18 residues in length and extends outward from the otherwise very compact structure of the HEAT domain. In the eIF4GII structure, this loop is largely disordered and shorter by 6 residues. Other notable differences occur in the loop connecting helices 3a and 3b (residues 185 to 200), which is well ordered in DAP5 and disordered in eIF4GII, where it is longer by 12 residues; and the loop connecting helices 4a and 4b (residues 236 to 249) which is longer in DAP5 by 7 residues. Large structural differences such as these impart significant differences in shape and chemical attributes to their surfaces, and likely contribute to the functional differences observed between these proteins such as IRES binding.

2.2.3 Identification of a potential IRES binding site in DAP5M Although eIF4G and DAP5 have common protein binding partners in eIF4A and eIF3, their interactions with nucleic acids are distinct. In vitro studies of human eIF4GI have demonstrated that its MIF4G domain is able to interact with different RNAs, namely the EMCV IRES RNA and -globin mRNA (Pestova et al., 1996; Lomakin et al., 2000) and one study even reported interaction with DNA (Kim et al., 1999). The DAP5 MIF4G domain, on the other hand, does not interact with the EMCV IRES RNA or -globin mRNA in vitro (Lomakin et al., 2000). Several studies have suggested that DAP5 can

43 initiate translation via IRES elements found in a number of cellular mRNAs (Henis- Korenblit et al., 2002; Henis-Korenblit et al., 2000; Hundsdoerfer et al., 2005; Lewis et al., 2008; Nevins et al., 2003; Warnakulasuriyarachchi et al., 2004). This raises the question as to whether DAP5M interacts with these IRES elements in a manner similar to how eIF4G associates with the EMCV IRES for cap-independent translational initiation. To identify potential sites of IRES interaction, we show the solvent-accessible surface of DAP5M colored according to its electrostatic potential in Figure 2-4A. In the N-terminal region of the molecule, there exists a large area of positively charged surface, which could potentially interact with nucleic acid. Interestingly, we found a sulfate ion from the crystallization buffer bound to a cluster of positively charged residues (Lys 108, Lys 112 and Arg 165) in this region (Figure 2-4A). This interaction could potentially mimic electrostatic interactions of the phosphodiester backbone of IRES RNA with DAP5M. The positively charged area corresponds to a highly conserved region among various orthologues of DAP5 (Figure 2-4B and Figure 2-5). Comparison of the equivalent surface in human eIF4GII also reveals large patches of positive potential, but with significantly different distribution of the charges, perhaps owing to the distinct IRES targeted by these two proteins (Figure 2-4C). The positively charged region in DAP5M is adjacent to the eIF4A-binding site (see below), suggesting that IRES binding may be coupled to the interaction with eIF4A. In support of this, a previous analysis of the MIF4G domain from eIF4GII identified mutants which affect binding of both eIF4A and the EMCV IRES (Marcotrigiano et al., 2001). Furthermore, the IRES interaction is enhanced in the presence of eIF4A (Lomakin et al., 2000). These observations indicate that there is cooperativity in binding of eIF4A and RNA to eIF4G, and that the sites of interaction are probably in close proximity to one another. Considering the functional and structural homology between eIF4G and DAP5 it is likely that DAP5M utilizes a similar mode of cooperative binding to eIF4A and IRES RNA. Based on our crystal structure of DAP5M it will be possible to carry out a mutational analysis of the residues in this region to confirm its role in DAP5 mediated IRES-driven translation initiation.

44

T/e and

B

3k

-

: red <

(Baker (Baker et al., 2001)

). ).

5

-

Figure 2 Figure

Enlargement Enlargement of electrostatic surface surrounding the

Inset Inset

. .

surface surface was calculated using APBS

and and T denotes temperature

denotes denotes Boltzmann constant

B

c potential surface representation of the MIF4G domain from eIF4GII (PDB ID: 1HU3). ID:1HU3). (PDB eIF4GII from domain MIF4G the of representation surface potential c

. Surface properties of DAP5M and MIF4G domain from eIF4GII from domainMIF4G propertiesand of Surface DAP5M .

T/e, T/e, where k

B

4

-

trostatic trostatic potential surface representation of DAP5M. The

2

Electrostati

Figure blue blue > +3k of DAP5M. B molecule on found ion sulfate observed (see orthologues DAP5 eukaryotic selected using conservation surface the of (B)Mapping (C) (A) Elec

45

(Livingstone (Livingstone

ence ence alignment of

B B were generated by processing the sequ

4

-

2

Analysis of Multiply Aligned Sequences (AMAS) server server (AMAS) Aligned ofMultiply Sequences Analysis

code code DAP5M surface based on conservation on Figure

-

Individual scores are found at the bottom of each row of the alignment, as found in this figure adapted from the output file. output the from adapted figure inthis found as alignment, the of row each of thebottom at found are scores Individual

. .

5: Scoring based on sequence on conservation based Scoring 5:

-

Figure2 The The scores used to color tunicate and using DAP5 frog, urchin human, zebrafish, turkey, chicken, 1993) Barton, and

46 2.2.4 The sites of interaction with eIF4A are structurally conserved in the middle domains of eIF4G and DAP5 but have different binding affinities It was previously shown that DAP5M, like eIF4G, recruits eIF4A, a member of the DEAD box RNA helicase family (Imataka et al., 1997). eIF4A is a 46 kDa bilobal protein with its N- and C-terminal domains connected by a flexible linker. Given the conservation of the overall folds of the MIF4G domains of DAP5 and eIF4G, it is likely that they interact with eIF4A in the same manner. To analyze the potential site of eIF4A interaction on DAP5M, we constructed a model of the complex by superposition of the DAP5M structure onto the crystal structure of the S. cerevisiae eIF4G-eIF4A complex (Schütz et al., 2008) (Figure 2-6). As with the yeast eIF4G-eIF4A complex, the DAP5M- eIF4A model reveals that there are two main sites of interaction between the two proteins. The C-terminal domain of eIF4A interacts with the N-terminal region of DAP5M encompassing heat repeats 1 and 2 (Site 1), whereas the N-terminal domain of eIF4A makes contact with the C-terminal part of DAP5M on helix 5b (Site 2). Together, the two sites of interaction bury ~2100 Å2 of surface area at the interface. Additionally, a third site of interaction in the yeast eIF4G-eIF4A complex was identified, which is contributed by a tryptophan residue connected to the MIF4G domain through a flexible, N-terminal linker that buries it in the C-terminal domain of eIF4A (Schütz et al., 2008). This conserved tryptophan residue, although present in DAP5 (Trp 50), could not be modelled as it was not included in our crystallization construct due to its likely flexible attachment. However, the contribution of Trp 50 to the interaction of DAP5 with eIF4A in humans is marginal, as observed in our binding and activity assays in vitro (data not shown). This suggests that, unlike what is observed in yeast, Trp 50 may not be as essential for DAP5's interaction with eIF4A or its effect may be more subtle in humans and will require other assays to fully characterize its role. Yeast eIF4GI middle domain and DAP5M have only 32% identity, yet, the majority of eIF4A-interacting residues in eIF4G (19 of 24 total interface residues), as identified in the crystal structure of the complex, are conserved in DAP5 (Figure 2-1B). Importantly, the model of the DAP5M-eIF4A complex indicates that, in addition to the

47

48 sequence, the conformations of most of the residues at the interface are also conserved, even in the absence of eIF4A. Of the 24 residues that make contact at Site 1 and Site 2, approximately three-quarters are in very similar conformations (Figure 2-6). This suggests that the binding mode between DAP5 and eIF4A mirrors that of eIF4G and that the interaction occurs through, for the most part, a preformed binding site on DAP5. Thus, we expected the binding affinities of eIF4A for the MIF4G domains of eIF4G and DAP5 to be comparable. To ascertain whether this is indeed the case, we carried out ITC experiments to compare dissociation constants for the interaction of human eIF4A (isoform I) with the MIF4G domains of human eIF4G (isoform I: 81% identical to eIF4GII in their MIF4G domains) and DAP5. Both constructs included the conserved tryptophan residue (Trp 50 in DAP5; Trp 734 in eIF4GI) as described in yeast. We measured a dissociation constant of 0.136 μM for the binding of eIF4A to eIF4G with a stoichiometry of 1:1 (Figure 2- 7A). This is in agreement with the crystal structure of the yeast eIF4G-eIF4A complex and our gel filtration analysis (Figure 2-9), which both show one molecule of eIF4A bound to a single MIF4G domain. Surprisingly, we found that the affinity of DAP5M for eIF4A is only 1.1 μM – about ten-fold lower than that of eIF4G, also with a stoichiometry of 1:1 (Figure 2-7B). This suggests that despite the conservation of the

Figure 2-6. Model of the DAP5M-eIF4A complex Center: Shown is the crystal structure of the yeast eIF4G-eIF4A complex (purple-grey) with DAP5M (red) superimposed on eIF4G. W579 (W50 in DAP5M), which comprises the third interaction site in the eIF4G-eIF4A complex, is represented as orange sticks. Left: Detailed view of Site 1 interaction residues. Top: Residues on the surface of eIF4G and DAPM that interact with eIF4A are shown as sticks. The view is rotated relative to the orientation in the center panel. Nitrogen, oxygen, sulfur and carbon atoms are colored blue, red, yellow and pink (DAP5M)/light blue (eIF4G), respectively. The majority of residues are conserved in sequence and conformation. Labels for residue numbers are shown for DAP5M only. Bottom, selected residues from the yeast eIF4G structure (light blue carbon atoms) that interact with yeast eIF4A (white carbon atoms) for which the corresponding residues in DAP5M were mutated. Labels for residue numbers from both yeast eIF4G/eIF4A and human DAP5/eIF4AI are indicated. Right: Detailed view of Site 2 interaction residues with similar coloring and labeling scheme as above.

49 binding site in both sequence and structure, other residues outside of the binding site modulate the affinity of MIF4G domains for eIF4A binding. Alternatively, it may be possible that the few non-conserved residues in the interface reduce the binding affinity or that the model of DAP5M-eIF4A based on the yeast complex does not accurately reflect the actual mode of binding altogether.

2.2.5 Mutational analysis of Site 1 and Site 2 residues in the DAP5- eIF4A complex

To examine whether residues predicted by the model of the DAP5M-eIF4A complex are indeed important in the interaction we mutated conserved residues in DAP5M at the Site 1 and Site 2 interfaces and carried out in vitro binding studies. At Site 1 we chose to mutate two key interactions: Asn 86 to alanine (N86A) and Glu 125 to lysine (E125K). Asn 86 forms several hydrogen bonds with backbone atoms from a loop in the C-terminal lobe of eIF4A and Glu 125 makes a salt-bridge with Arg 312 of yeast eIF4A (Arg 324 in human eIFAI), also located on a loop in the C-terminal lobe (Figure 2-6, lower left panel). At Site 2, Phe 296, which makes interactions with a hydrophobic pocket on the N- terminal lobe of eIF4A, was mutated to alanine (F296A) (Figure 2-6, lower right panel). These mutant proteins were soluble as exhibited by their well-behaved gel filtration profiles (Figure 2-9). Mutation of single residues at either the Site 1 or Site 2 interfaces resulted in complete loss of binding in in vitro pull-down assays (Figure 2-8). The same results were obtained using gel filtration chromatography (Figure 2-9). Our results agree with those obtained by Morino et al. (2000), who showed that the mutation of Phe 977 (equivalent to Phe 296 in DAP5) to Ala (F977A) in eIF4GI resulted in loss of binding to eIF4A in cell extracts. This confirms that DAP5M interacts with eIF4A in the same manner as was observed for eIF4G, albeit with a lower affinity which may impact its ability to stimulate eIF4A activity.

50

Figure 2-7. ITC titration binding curves

(A) eIF4AI with eIF4GI(MIF4G) (residues 732-1003), where KD is the dissociation constant and N is the number of binding sites. (B) eIF4AI with DAP5M (residues 48-323).

Figure 2-8. In vitro pull-down experiments Hexahistidine-tagged constructs of purified DAP5M and its mutants were used as bait and purified eIF4AI as prey. Shown are Coomassie Brilliant Blue stained SDS-PAGE gels. Indicated on the left are protein marker sizes in kDa. Pulldowns of eIF4A with DAP5M containing mutations at Site 1 or Site 2. Lane 1 is a control in which eIF4A alone is incubated with empty Ni-NTA beads to account for non-specific binding of eIF4A to the resin. Lane 2 is a positive control with wild-type DAP5M, which shows strong binding to eIF4A. Lanes 3 to 5 are pulldowns with the indicated mutants of DAP5M, all three of which fail to pull down eIF4A. See Figure 2-9.

51

Figure 2-9: Gel filtration profiles of MIF4G-eIF4AI complexes Mixtures containing a slight molar excess of middle domain were subjected to chromatographic separation on a Superdex 200 10/300 GL column (GE Healthcare) eluted with a buffer containing 25 mM Tris pH 8.0, 150 mM NaCl, 5% glycerol. 0.5-mL fractions of peaks fractions were analyzed by SDS-PAGE analysis on a 12% polyacrylamide gel stained with Coomassie Brilliant Blue. The top protein band corresponds to eIF4AI (46 kDa); the bottom one, to the middle domain (~30 kDa). A-E shows the elution profiles of MIF4G (of either DAP5 or eIF4GI, pink) and eIF4AI (purple), alone or together (ochre). The effects of DAP5M (B-D) mutations are illustrated. (A) DAP5M (48-323) and eIF4AI form a complex. (B) DAP5M N86A, (C) DAP5M E125K and (D) DAP5M F296A do not form a stable complex with eIF4AI. We observe a slight peak shift toward smaller retention volumes that can be attributed to the transient formation of a higher molecular weight species between DAP5M mutants and eIF4AI. (E) MIF4GI forms a complex with eIF4AI.

52 2.2.6 Effect of DAP5M on the helicase activity To assess the effect of the lower affinity DAP5M-eIF4A interaction on helicase activity, we used a fluorescence-based activity assay to monitor RNA unwinding by full-length human eIF4A in vitro (Özeş et al., 2011). The stimulation of eIF4A helicase activity through interaction with eIF4G is well established in vitro using either protein purified from cell extracts or recombinant protein, although the extent of stimulation depends on the particular construct used (Abramson et al., 1988; Rogers et al., 2001). Full-length eIF4G, however, is intrinsically unstable and thus for our activity assays we used a construct encompassing approximately the C-terminal two-thirds of eIF4G, lacking regions upstream of the MIF4G domain (residues 732-1571), similar to one previously demonstrated to robustly stimulate eIF4A ATPase activity in vitro (Korneeva et al., 2005). Analogously for DAP5, we used a near full-length construct beginning just N- terminal to the middle domain and extending to the C-terminus (residues 48-907). The reaction mix also includes the accessory factor eIF4B, which enhances eIF4A processivity (Rogers et al., 2001). eIF4A and eIF4B display relatively low unwinding activity on their own (Figure 2-10, green). Addition of wild-type DAP548-907 to the reaction almost doubled the basal unwinding rate to 4.5% per minute (Figure 2-10, red). The N86A mutant of DAP548-907, which as demonstrated above abrogates binding of eIF4A to DAP5M, returned the activity back to basal levels (Figure 2-10, lavender), confirming the importance of DAP5- eIF4A binding for stimulation of RNA unwinding activity. Thus, as with eIF4G, DAP5 can stimulate eIF4A activity via interaction with its middle domain. However, carrying out the unwinding reaction with eIF4G732-1571 in our assay indicates that it is twice as potent as DAP548-907 in stimulating helicase activity (Figure 2-10, blue). For the abovementioned experiments, all of the proteins in the reaction mixtures were maintained at concentrations of 1 μM. According to the dissociation constants determined above for the middle domains, at 1 uM concentration DAP548-907 would be less occupied by eIF4A than eIF4G732-1571, which would be close to saturation, although the precise situation is more complex since eIF4G732-1571 contains a second binding site for eIF4A in its C- terminal region that is absent in DAP548-907. Nonetheless, the 10-fold difference in binding affinity of eIF4A with the middle domains of the two proteins likely contributes

53

Figure 2-10. In vitro helicase assay The percent unwinding of the RNA substrate over time was monitored by measuring the increase in fluorescence caused by eIF4A and eIF4B in the absence (green) or in the presence of wild-type DAP5 (red and grey) or its N86A mutant (lavender), or in presence of the C-terminal two-thirds of eIF4G (blue and gold). Each protein was present at 1μM in the unwinding reaction unless specified otherwise. The initial unwinding rates were extracted from the initial linear portion of the unwinding time course and expressed as percent unwinding per minute. Error bars represent the standard deviation of three replicates. Wild-type DAP5 and its eIF4G equivalent stimulate the activity of the helicase whereas unwinding in the reaction containing DAP5 N86A is comparable to that of eIF4A and eIF4B alone. There is a ~2-fold difference in unwinding aided by DAP5 and eIF4G at 1 μM, but their eIF4A-stimulation capacities are more similar at 2 μM.

54 to the observed differences in their respective stimulation of unwinding activities. Indeed, increasing the DAP548-907 and eIF4G732-1571 concentrations to 2 μM in the assay (Figure 2-10, grey and gold) substantially reduces the observed difference in unwinding activities between the two proteins.

2.3 Discussion We have determined the crystal structure of the MIF4G domain from DAP5. Our structure reveals that the overall fold of DAP5M is a helical HEAT domain that is very similar to that found in crystal structures of the MIF4G domain from eIF4G. However, significant conformational differences in the connecting loop structures of DAP5M impart distinct shape and surface characteristics to it. We also showed that DAP5M is functionally homologous to eIF4G in its ability to interact with and stimulate eIF4A activity. A precise molecular role for DAP5 is still lacking. The earliest reports, via overexpression in vitro and in transfected cells, ascribed an inhibitory role to DAP5 on both cap-dependent and EMCV-IRES driven translation. Based on this it was postulated that DAP5 could function as a general translation inhibitor by sequestering eIF4A and eIF3, but not eIF4E, into inactive complexes. Later studies often produced many conflicting results with functions of DAP5 ranging from inhibitor of cap-dependent translation (Henis-Korenblit et al., 2002), stimulator of DAP5-IRES dependent translation (Nevins et al., 2003), inert in DAP5-IRES dependent translation (Henis- Korenblit et al., 2002) and even a stimulator of cap-dependent translation (Lee and McCormick, 2006). All of these studies were based on overexpression of DAP5 beyond physiological levels, which could account for the varying results observed depending on the precise conditions and cell types used. Much evidence now points to DAP5 being a scaffolding protein in IRES mediated translation. Knockout and knockdown studies have established that DAP5 is an essential factor for the translation of specific cellular mRNAs containing IRES elements in their 5'- UTRs such as c-myc, CDK1 and Bcl family members and DAP5 itself (Yamanaka et al., 2000; Marash et al., 2008). However, the molecular details of how DAP5 partakes in translation initiation and its relationship to eIF4G have remained elusive. Our structure of

55 DAP5M supports its role in IRES mediated translation in that it possesses a positive surface charge distribution distinct from that found in MIF4G and is well suited for interaction with nucleic acids, although direct interaction between DAP5M and IRES RNA remains to be demonstrated. Our finding of a 10-fold affinity difference between eIF4A and the middle domains of eIF4G and DAP5 resulting in variable stimulation eIF4A helicase activity further supports the notion that DAP5 does not simply function as a general translational inhibitor. eI4FA is the most abundant initiation factor present in the cell, present at concentrations of up to ~30 M whereas DAP5/eIF4G levels can be in the range of ~2 to 6 M (Duncan and Hershey, 1983; Lee and McCormick, 2006). Thus, based on the dissociation constants determined here, both DAP5 and eIF4G would for the most part be in complex with eIF4A, although DAP5 is likely more dynamic in this regard owing to its lower affinity for eIF4A. Indeed, the levels of DAP5 are significantly modulated under cellular stress conditions, particularly during caspase activation, which causes a shift from cap-dependent to cap-independent translation by the degradation of eIF4G (Bushell et al., 2000; Svitkin et al., 1999; Henis-Korenblit et al., 2000), thereby favoring the association of DAP5 with eIF4A. Finally, the importance of the DAP5-eIF4A interaction was further validated in recent translation rescue experiments from rabbit reticulocyte lysate, where the ability of recombinant DAP5 to rescue the translation of cellular IRESs depended specifically on the integrity of interaction between its middle domain and eIF4A (V. Gandin, personal communication). Future studies aimed at deciphering the molecular contribution of DAP5 to IRES driven translation will be required to unravel its true biological role.

2.4 Materials and methods 2.4.1 Expression and purification of recombinant proteins The MIF4G domain of human DAP5 (DAP5M; residues 48 to 323 and 61 to 323) was subcloned into the BamHI and EcoRI restriction sites of the bacterial expression vector pProEx HTb (Invitrogen, Carlsbad). DAP5M was expressed in Escherichia coli strain

BL21 (Rosetta 2) by induction at OD600 = 0.6 with 1 mM isopropyl-β-D-thiogalactoside (IPTG) for 4 hours at 30°C. The protein was purified by nickel affinity chromatography

56 on a Sepharose 6 Fast Flow nickel affinity resin column (HisTrap FF, GE Healthcare, Little Chalfont). Following overnight Tobacco Etch virus (TEV) protease cleavage and dialysis against nickel binding buffer, TEV protease (hexahistidine-tagged) was removed by applying samples onto a HisTrap FF column and collecting the flowthrough. Tagged and cleaved protein was further purified by gel filtration chromatography on a HiLoad 16/60 Superdex 75 prep grade column (GE Healthcare, Little Chalfont) eluted with a buffer containing 25 mM Tris pH 8.0, 150 mM NaCl, 5% glycerol, 1 mM phenylmethylsulfonyl fluoride (PMSF), and 1 mM dithiothreitol (DTT). Human full-length eIF4AI was cloned into the BamHI and EcoRI restriction sites of the bacterial expression vector pProEx Htb (Invitrogen, Carlsbad). The protein was expressed in Escherichia coli strain BL21 (DE3) by auto-induction overnight at 20°C (Studier, 2005). The protein was purified by nickel affinity chromatography and cleaved as described for DAP5M. The cleaved protein was further purified by anion exchange chromatography on a Q Sepharose Fast Flow resin column (HiTrap Q FF, GE Healthcare) before gel filtration, as described for DAP5M. The middle domain of human eIF4GI (MIF4G, residues 732 to 1003) was cloned into pET-28b-SMT3 vector (Mossessova and Lima, 2000) in which we mutated the EcoRI restriction site originally contained in the SUMO coding region, enabling the use of EcoRI as forward restriction site together with XhoI. A longer eIF4G construct encompassing its three HEAT-repeats domains (732-1571) was cloned into the BamHI and NotI sites of the same vector. The purification steps for MIF4G and eIF4G732-1571 are the same as for DAP5M and eIF4A, respectively. Ulp-1 was used to desumoylate the fusion proteins (Mossessova and Lima, 2000). Human DAP5 (residues 48 to 907) was cloned into pProEx Htb (Invitrogen) and expressed as a N-terminal hexahistidine fusion protein in Escherichia coli strain BL21 (Rosetta 2) by auto-induction overnight at 20°C (Studier, 2005). The protein was purified by nickel affinity chromatography on Ni-NTA Superflow resin (Qiagen) followed by anion exchange chromatography on a Q Sepharose Fast Flow resin column (HiTrap Q FF, GE Healthcare). The protein was further purified by gel filtration chromatography on a Superdex 200 HR 10/300 column (GE Healthcare) eluted with a buffer containing 25

57 mM Tris pH 8.0, 150 mM NaCl, 5% glycerol, 1 mM phenylmethylsulfonyl fluoride (PMSF), and 3 mM dithiothreitol (DTT). All DNA sequences were confirmed by sequencing.

2.4.2 Mutagenesis of DAP5M Mutants of DAP5M (48-323) and DAP5 were generated using the QuikChange mutagenesis kit (Stratagene, La Jolla). Sequences of the mutants were confirmed by plasmid DNA sequencing. The mutant proteins were expressed and purified as described above for the wild-type protein.

2.4.3 Crystallization and data collection Tris(2-carboxyethyl)phosphine (TCEP; 1 mM final concentration) was added to DAP5M (61-323) samples before use in crystallization trials. Crystals suitable for structure determination were grown at 291 K using protein concentration of 15–20 mg/ml in drops of 2–4 μL volume (1–2 μL protein solution mixed with 1–2 μL reservoir solution). Crystals were grown in 0.1 M HEPES pH 7.5, 0.2 M ammonium sulfate and 18–20% (w/v) polyethylene glycol 5000 monomethyl ether (PEG 5000 MME). Crystals were flash-cooled in a liquid-nitrogen cryostream at 100 K and data were collected in-house on a Rigaku MicroMax-007 HF microfocus X-ray generator fitted with Varimax X-ray optics and a Saturn 944+ CCD detector.

2.4.4 Structure determination Crystallization, data collection and initial structure solution of DAP5M were carried out as described previously (Frank et al., 2010). In short, DAP5M shares ~43 % sequence identity with its homologous region in eIF4GII. Crystals belong to the space group C2 with two molecules of DAP5M per asymmetric unit and diffracted to 2.4 Å (Table 2-1). Maximum likelihood molecular replacement using the program Phaser (Read, 2001) and the structure of the eIF4GII MIF4G domain (PDB ID 1HU3, Marcotrigiano et al., 2001) resulted in a solution with acceptable Z-scores and log-likelihood gain (RFZ = 5.3; TFZ = 11.3; LLG = 169). However, initial attempts at refinement did not reduce the R factors. Therefore, we broke up the Phaser solution into 20 separate rigid bodies, each

58 corresponding to one helix and carried out rigid body refinement. This reduced the R factors to 47% and subsequent simulated annealing, energy minimization and B-factor refinement with the program CNS (Brünger et al., 1998), as well as density modification including the application of two-fold NCS symmetry averaging using the program DM (Cowtan and Main, 1993), further reduced the R factors and resulted in reasonable electron density for majority of the structure allowing us to build a model. During initial refinement against the home source data set described before (Frank et al., 2010), CNS (Brünger et al., 1998) was used and NCS restraints were applied. This refinement resulted in a final model with an Rfree and R factor of 27.8 % and 24.4 %, respectively. The resulting model was then further refined against a slightly higher resolution data set (2.3 Å) from CHESS beamline A1 (Table 2-1) using Phenix (Adams et al., 2002). NCS restraints were removed for refinement using Phenix, but TLS refinement was employed.

The final model yielded an R factor of 22.2 % and Rfree of 25.6 %.

2.4.5 Gel filtration chromatography Half-milliliter samples containing 0.5 mg His-DAP5M (or MIF4G, or eIF4A alone) with or without 0.5 mg eIF4A was injected over a Superdex 200 10/300 GL column (GE Healthcare, Little Chalfont) eluted with a buffer containing 25 mM Tris pH 8.0, 150 mM NaCl and 5% glycerol.

2.4.6 In vitro pull-down assays Thirty micrograms of His-DAP5M were incubated in binding buffer (25 mM Tris pH 8.0, 150 mM NaCl, 60 mM imidazole pH 8.0, 5% glycerol), together with 25 L Ni-NTA Superflow resin (Qiagen, Hilden) and 100 μg of eIF4AI (~3-fold molar excess) for 30 minutes on ice in 100 μL reaction volume. After washing with 3 x 700 μL of binding buffer, proteins were eluted with 50 μL of elution buffer (25 mM Tris pH 8.0, 500 mM NaCl, 500 mM imidazole pH 8.0). SDS-PAGE was carried out on a 12% polyacrylamide gel and eluted proteins were visualized with Coomassie Brilliant Blue staining.

59 2.4.7 Isothermal titration calorimetry (ITC) Experiments were performed with a VP-ITC instrument (MicroCal, GE Healthcare) at 20°C. His-DAP5M (or desumoylated MIF4G from eIF4GI) and eIF4AI samples were first dialyzed against a buffer containing 25 mM Tris pH 8.0, 150 mM NaCl, 1 mM DTT and then diluted to 0.5-1.0 mM and 0.05-0.1 mM, respectively. MIF4G domain proteins were loaded into the syringe while eIF4AI was loaded into the calorimetric cell. The heat of binding was measured over the injection of 37 μl of MIF4G in 2 μl increments into the cell. Data were fitted to a one binding site model using the Origin software package (MicroCal, GE Healthcare).

2.4.8 Helicase assays Experiments were performed as described by Özeş et al., 2011. eIF4A, eIF4B, DAP5 and eIF4G were used at a concentration of 1 μM unless specified otherwise. The substrate is a double-reporter RNA construct with a 20-nt 5’ overhang and a 24-bp duplex region. It was used at a concentration of 50 nM. Sequences are as follows, with underlined regions corresponding to the duplex regions: template: 5’GAACAACAACAACAACAACAGAAAAAAUUAAAAAAUUAAAAAA CUCGGAGGGGCCGGUGGGGCC - 3’; Cy3 strand sequence: 5’-Cy3- GUUUUUUAAUUUUUUAAUUUUUUC – 3’; BHQ strand sequence: 5’- GGCCCCACCGGCCCCUCCG – BHQ – 3’; 24-nt DNA competitor: 5’- GAAAAAATTAAAAAATTAAAAAAC – 3’. BHQ and Cy3 labeled RNA oligonucleotides were annealed side by side (one nucleotide apart) to the template RNA. Competitor DNA was present in 10x excess to capture Cy3 RNA upon unwinding.

2.5 Acknowledgements

We thank R.Szittner and K. Illes for technical support. B.N. is supported by a Canada Research Chair, a Career Development Award from the Human Frontiers Science Program (CDA 0018/2006-C/1) and an operating grant from the Canadian Institutes of Health Research (CIHR grant MOP-82929). N.S. is funded by a CIHR grant. CF and KF

60 are supported by grant R01GM092927 from the National Institute of General Medical Sciences and a NIH training grant T32 GM-007377-29 to KF. G.V. is supported by the CIHR Strategic Training Initiative in Chemical Biology and by the Groupe de Recherche Axé sur la Structure des Protéines (FRSQ). F.F. is supported by a Boehringer Ingelheim Fonds PhD Fellowship. M.S. is supported by an NSERC-CREATE Training Program in Bionanomachines Undergraduate Award.

2.6 Accession number

The atomic coordinates for the MIF4G domain of DAP5 have been deposited in the Protein Data Bank under ID code 4IUL.

61 Epilogue We have compared the activity of two different eIF4G proteins in vitro in RNA unwinding assays and concluded that DAP5 and eIF4G have comparable abilities to stimulate eIF4A. According to the model of C-terminal autoinhibition proposed by Dobrikov et al. (2013), it is expected that DAP5 and eIF4G would have different activities since native eIF4G and not DAP5 would be in the autoinhibited conformation (as described in section 1.3.3). We can try to reconcile our results with our expectations by taking a closer look to the protein constructs that were used in this experiment. The construct of DAP5 lacked 47 residues at the N-terminus and its function was previously validated in cap-independent translation rescue experiments in rabbit reticulocyte lysate (V. Gandin, personal communication). The construct of eIF4G was designed based on the two crystal structures of eIF4G (PBD entry 1HU3 and 1UG3) to encompass its three HEAT domains so that it lacks its N-terminal third and 28 residues at its C-terminus (Bellsolell et al., 2006; Marcotrigiano et al., 2001). The C-terminal deletion is detrimental to the interaction of eIF4G with Mnk. Those 28 missing residues might also be important for the interaction of W2 with the MIF4G-MA3 linker at the origin of the autoinhibitory conformation so that its suppression might activate eIF4G. Also, the supposition that native DAP5 is constitutively active might be erroneous in such a way that a yet- undescribed autoinhibitory conformation might also exist in DAP5. The study of large proteins like DAP5 and eIF4G is challenging. In those cases, the use of recombinant proteins is not flawless but it allows the use of a specific combination of domains adapted for the study of a particular process. The work that was presented in Chapter 2 is a good example of the potential of recombinant proteins since the hypotheses derived from the structural study of DAP5M were proved correct by functional assays in the context of the (almost) full-length protein.

62 References

Abramson, R.D., Dever, T.E., and Merrick, W.C. (1988). Biochemical evidence supporting a mechanism for cap-independent and internal initiation of eukaryotic mRNA. J Biol Chem 263, 6016–6019.

Adams, P.D., Grosse-Kunstleve, R.W., Hung, L.W., Ioerger, T.R., McCoy, A.J., Moriarty, N.W., Read, R.J., Sacchettini, J.C., Sauter, N.K., and Terwilliger, T.C. (2002). PHENIX: building new software for automated crystallographic structure determination. Acta Crystallogr. D Biol. Crystallogr. 58, 1948–1954.

Andrade, M.A., and Bork, P. (1995). HEAT repeats in the Huntington's disease protein. Nat. Genet. 11, 115–116.

Andreou, A.Z., and Klostermeier, D. (2012). The DEAD-box helicase eIF4A Paradigm or the odd one out? RNA Biol 10, 19–32.

Asano, K., Krishnamoorthy, T., Phan, L., Pavitt, G.D., and Hinnebusch, A.G. (1999). Conserved bipartite motifs in yeast eIF5 and eIF2Bepsilon, GTPase-activating and GDP- GTP exchange factors in translation initiation, mediate binding to their common substrate eIF2. Embo J 18, 1673–1688.

Aström, J., Aström, A., and Virtanen, A. (1991). In vitro deadenylation of mammalian mRNA by a HeLa cell 3' exonuclease. Embo J 10, 3067–3071.

Baird, S.D., Turcotte, M., Korneluk, R.G., and Holcík, M. (2006). Searching for IRES. Rna 12, 1755–1785.

Baker, N. A., Sept, D., Joseph, S., Holst, M. J., and McCammon, J. A. (2001). Electrostatics of nanosystems: application to microtubules and the ribosome. Proc Natl Acad Sci USA 98, 10037–10041.

Bellsolell, L., Cho-Park, P.F., Poulin, F., Sonenberg, N., and Burley, S.K. (2006). Two structurally atypical HEAT domains in the C-terminal portion of human eIF4G support binding to eIF4A and Mnk1. Structure 14, 913–923.

Benz, J., Trachsel, H., and Baumann, U. (1999). Crystal structure of the ATPase domain of translation initiation factor 4A from Saccharomyces cerevisiae--the prototype of the DEAD box protein family. Structure 7, 671–679.

Bonneau, A.M., and Sonenberg, N. (1987a). Proteolysis of the p220 component of the cap-binding protein complex is not sufficient for complete inhibition of host cell protein synthesis after poliovirus infection. J. Virol. 61, 986–991.

Bonneau, A.M., and Sonenberg, N. (1987b). Involvement of the 24-kDa cap-binding protein in regulation of protein synthesis in mitosis. J Biol Chem 262, 11134–11139.

Bono, F., Ebert, J., Lorentzen, E., and Conti, E. (2006). The crystal structure of the exon

63 junction complex reveals how it maintains a stable grip on mRNA. Cell 126, 713–725.

Borman, A.M., Kirchweger, R., Ziegler, E., Rhoads, R.E., Skern, T., and Kean, K.M. (1997). elF4G and its proteolytic cleavage products: effect on initiation of protein synthesis from capped, uncapped, and IRES-containing mRNAs. Rna 3, 186–196.

Brünger, A. T., Adams, P. D., Clore, G. M., DeLano, W. L., Gros, P., Grosse-Kunstleve, R. W., Jiang, J. S., Kuszewski, J., Nilges, M., Pannu, N. S., et al. (1998). Crystallography & NMR system: A new software suite for macromolecular structure determination. Acta Crystallogr. D Biol. Crystallogr. 54, 905–921.

Burnett, P.E., Barrow, R.K., Cohen, N.A., Snyder, S.H., and Sabatini, D.M. (1998). RAFT1 phosphorylation of the translational regulators p70 S6 kinase and 4E-BP1. Proc Natl Acad Sci USA 95, 1432–1437.

Bushell, M., Poncet, D., Marissen, W.E., Flotow, H., Lloyd, R.E., Clemens, M.J., and Morley, S.J. (2000). Cleavage of polypeptide chain initiation factor eIF4GI during apoptosis in lymphoma cells: characterisation of an internal fragment generated by caspase-3-mediated cleavage. Cell Death Differ 7, 628-636.

Bushell, M., Stoneley, M., Kong, Y.W., Hamilton, T.L., Spriggs, K.A., Dobbyn, H.C., Qin, X., Sarnow, P., and Willis, A.E. (2006). Polypyrimidine tract binding protein regulates IRES-mediated gene expression during apoptosis. Molecular Cell 23, 401–412.

Caruthers, J.M., Johnson, E.R., and McKay, D.B. (2000). Crystal structure of yeast initiation factor 4A, a DEAD-box RNA helicase. Proc Natl Acad Sci USA 97, 13080– 13085.

Chang, J.H., Cho, Y.H., Sohn, S.Y., Choi, J.M., Kim, A., Kim, Y.C., Jang, S.K., and Cho, Y. (2009). Crystal structure of the eIF4A-PDCD4 complex. Proc Natl Acad Sci USA 106, 3148–3153.

Clemens, M.J., Bushell, M., Jeffrey, I.W., Pain, V.M., and Morley, S.J. (2000). Translation initiation factor modifications and the regulation of protein synthesis in apoptotic cells. Cell Death Differ 7, 603–615.

Cowtan, K.D., and Main, P. (1993). Improvement of macromolecular electron-density maps by the simultaneous application of real and reciprocal space constraints. Acta Cryst.D 49, 148-157.

Craig, A.W., Haghighat, A., Yu, A.T., and Sonenberg, N. (1998). Interaction of polyadenylate-binding protein with the eIF4G homologue PAIP enhances translation. Nature 392, 520–523.

De Gregorio, E., Preiss, T., and Hentze, M.W. (1998). Translational activation of uncapped mRNAs by the central part of human eIF4G is 5' end-dependent. Rna 4, 828– 836.

64 De Gregorio, E., Preiss, T., and Hentze, M.W. (1999). Translation driven by an eIF4G core domain in vivo. Embo J 18, 4865–4874.

Del Campo, M., Mohr, S., Jiang, Y., Jia, H., Jankowsky, E., and Lambowitz, A.M. (2009). Unwinding by local strand separation is critical for the function of DEAD-box proteins as RNA chaperones. J Mol Biol 389, 674–693.

Dobrikov, M.I., Dobrikova, E.Y., and Gromeier, M. (2013). Dynamic Regulation of the Translation Initiation Helicase Complex by Mitogenic Signal Transduction to Eukaryotic Translation Initiation Factor 4G. Mol Cell Biol 33, 937–946.

Dobrikov, M., Dobrikova, E., Shveygert, M., and Gromeier, M. (2011). Phosphorylation of eukaryotic translation initiation factor 4G1 (eIF4G1) by protein kinase C{alpha} regulates eIF4G1 binding to Mnk1. Mol Cell Biol 31, 2947–2959.

Dorrello, N.V., Peschiaroli, A., Guardavaccaro, D., Colburn, N.H., Sherman, N.E., and Pagano, M. (2006). S6K1- and betaTRCP-mediated degradation of PDCD4 promotes protein translation and cell growth. Science 314, 467–471.

Duncan, R., and Hershey, J.W. (1983). Identification and quantitation of levels of protein synthesis initiation factors in crude HeLa cell lysates by two-dimensional polyacrylamide gel electrophoresis. J Biol Chem 258, 7228–7235.

Duncan, R., Milburn, S.C., and Hershey, J.W. (1987). Regulated phosphorylation and low abundance of HeLa cell initiation factor eIF-4F suggest a role in translational control. Heat shock effects on eIF-4F. J Biol Chem 262, 380–388.

Eddy, S. R. (2004). Where did the BLOSUM62 alignment score matrix come from? Nat Biotechnol 22, 1035–1036.

Fan, S., Jia, M.-Z., and Gong, W. (2010). Crystal structure of the C-terminal region of human p97/DAP5. Proteins 78, 2385–2390.

Frank, F., Virgili, G., Sonenberg, N., and Nagar, B. (2010). Crystallization and preliminary X-ray diffraction analysis of the MIF4G domain of DAP5. Acta Crystallogr. Sect. F Struct. Biol. Cryst. Commun. 66, 15–19.

Fujita, Y., Oe, M., Tutsumino, T., Morino, S., Imataka, H., Tomoo, K., and Ishida, T. (2009). Domain-dependent interaction of eukaryotic initiation factor eIF4A for binding to middle and C-terminal domains of eIF4G. J. Biochem. 146, 359–368.

Furic, L., Rong, L., Larsson, O., Koumakpayi, I.H., Yoshida, K., Brueschke, A., Petroulakis, E., Robichaud, N., Pollak, M., Gaboury, L.A., et al. (2010). eIF4E phosphorylation promotes tumorigenesis and is associated with prostate cancer progression. Proc Natl Acad Sci USA 107, 14134–14139.

Galicia-Vázquez, G., Cencic, R., Robert, F., Agenor, A.Q., and Pelletier, J. (2012). A cellular response linking eIF4AI activity to eIF4AII transcription. Rna.

65 Gao, J.-J., Cai, G.-Y., Ning, Y.-C., Liu, L., Yang, J.-R., Dong, D., Fu, B., Lu, Y., Cui, S.- Y., and Chen, X.-M. (2012). DAP5 ameliorates cisplatin-induced apoptosis of renal tubular cells. Am. J. Nephrol. 35, 456–465.

Gingras, A.C., Svitkin, Y., Belsham, G.J., Pause, A., and Sonenberg, N. (1996). Activation of the translational suppressor 4E-BP1 following infection with encephalomyocarditis virus and poliovirus. Proc Natl Acad Sci USA 93, 5578–5583.

Gradi, A., Svitkin, Y.V., Imataka, H., and Sonenberg, N. (1998). Proteolysis of human eukaryotic translation initiation factor eIF4GII, but not eIF4GI, coincides with the shutoff of host protein synthesis after poliovirus infection. Proc Natl Acad Sci USA 95, 11089– 11094.

Gunawardena, S.R., Ruis, B.L., Meyer, J.A., Kapoor, M., and Conklin, K.F. (2008). NOM1 targets protein phosphatase I to the nucleolus. J Biol Chem 283, 398–404.

Hay, N., and Sonenberg, N. (2004). Upstream and downstream of mTOR. Dev 18, 1926–1945.

Hellen, C.U., and Sarnow, P. (2001). Internal ribosome entry sites in eukaryotic mRNA molecules. Genes Dev 15, 1593–1612.

Henis-Korenblit, S., Strumpf, N.L., Goldstaub, D., and Kimchi, A. (2000). A novel form of DAP5 protein accumulates in apoptotic cells as a result of caspase cleavage and internal ribosome entry site-mediated translation. Mol Cell Biol 20, 496–506.

Henis-Korenblit, S., Shani, G., Sines, T., Marash, L., Shohat, G., and Kimchi, A. (2002). The caspase-cleaved DAP5 protein supports internal ribosome entry site-mediated translation of death proteins. Proc Natl Acad Sci USA 99, 5400–5405.

Hilbert, M., Kebbel, F., Gubaev, A., and Klostermeier, D. (2011). eIF4G stimulates the activity of the DEAD box protein eIF4A by a conformational guidance mechanism. Nucleic Acids Res. 39, 2260–2270.

Holcik, M., Sonenberg, N., and Korneluk, R.G. (2000). Internal ribosome initiation of translation and the control of cell death. Trends Genet. 16, 469–473.

Holcik, M., and Sonenberg, N. (2005). Translational control in stress and apoptosis. Nat Rev Mol Cell Biol 6, 318-327.

Holm, L., and Rosenström, P. (2010). Dali server: conservation mapping in 3D. Nucleic Acids Res. 38, W545–W549.

Holz, M.K., Ballif, B.A., Gygi, S.P., and Blenis, J. (2005). mTOR and S6K1 mediate assembly of the translation preinitiation complex through dynamic protein interchange and ordered phosphorylation events. Cell 123, 569–580.

Hundsdoerfer, P., Thoma, C., and Hentze, M.W. (2005). Eukaryotic translation initiation

66 factor 4GI and p97 promote cellular internal ribosome entry sequence-driven translation. Proc Natl Acad Sci USA 102, 13421–13426.

Imataka, H., and Sonenberg, N. (1997). Human eukaryotic translation initiation factor 4G (eIF4G) possesses two separate and independent binding sites for eIF4A. Mol Cell Biol 17, 6940–6947.

Imataka, H., Olsen, H.S., and Sonenberg, N. (1997). A new translational regulator with homology to eukaryotic translation initiation factor 4G. Embo J 16, 817–825.

Johnson, E.R., and McKay, D.B. (1999). Crystallographic structure of the amino terminal domain of yeast initiation factor 4A, a representative DEAD-box RNA helicase. Rna 5, 1526–1534.

Kahvejian, A., Svitkin, Y.V., Sukarieh, R., M'Boutchou, M.-N., and Sonenberg, N. (2005). Mammalian poly(A)-binding protein is a eukaryotic translation initiation factor, which acts via multiple mechanisms. Genes Dev 19, 104–113.

Kim, C.Y., Takahashi, K., Nguyen, T.B., Roberts, J.K., and Webster, C. (1999). Identification of a nucleic acid binding domain in eukaryotic initiation factor eIFiso4G from wheat. J Biol Chem 274, 10603-10608.

Kolupaeva, V.G., Lomakin, I.B., Pestova, T.V., and Hellen, C.U.T. (2003). Eukaryotic initiation factors 4G and 4A mediate conformational changes downstream of the initiation codon of the encephalomyocarditis virus internal ribosomal entry site. Mol Cell Biol 23, 687–698.

Korneeva, N.L., Lamphear, B.J., Hennigan, F.L., and Rhoads, R.E. (2000). Mutually cooperative binding of eukaryotic translation initiation factor (eIF) 3 and eIF4A to human eIF4G-1. J Biol Chem 275, 41369–41376.

Korneeva, N.L., Lamphear, B.J., Hennigan, F.L., Merrick, W.C., and Rhoads, R.E. (2001). Characterization of the two eIF4A-binding sites on human eIF4G-1. J Biol Chem 276, 2872–2879.

Korneeva, N.L., First, E.A., Benoit, C.A., and Rhoads, R.E. (2005). Interaction between the NH2-terminal domain of eIF4A and the central domain of eIF4G modulates RNA- stimulated ATPase activity. J Biol Chem 280, 1872–1881.

Lamphear, B.J., Kirchweger, R., Skern, T., and Rhoads, R.E. (1995). Mapping of functional domains in eukaryotic protein synthesis initiation factor 4G (eIF4G) with picornaviral proteases. Implications for cap-dependent and cap-independent translational initiation. J Biol Chem 270, 21975–21983.

Lamphear, B.J., Yan, R., Yang, F., Waters, D., Liebig, H.D., Klump, H., Kuechler, E., Skern, T., and Rhoads, R.E. (1993). Mapping the cleavage site in protein synthesis initiation factor eIF-4 gamma of the 2A proteases from human Coxsackievirus and rhinovirus. J Biol Chem 268, 19200–19203.

67 LaRonde-LeBlanc, N., Santhanam, A.N., Baker, A.R., Wlodawer, A., and Colburn, N.H. (2007). Structural basis for inhibition of translation by the tumor suppressor Pdcd4. Mol Cell Biol 27, 147–156.

Lee, S.H., and McCormick, F. (2006). p97/DAP5 is a ribosome-associated factor that facilitates protein synthesis and cell proliferation by modulating the synthesis of cell cycle proteins. Embo J 25, 4008–4019.

Levine, A.J., and Oren, M. (2009). The first 30 years of p53: growing ever more complex. Nat. Rev. Cancer 9, 749–758.

Levy-Strumpf, N., Deiss, L.P., Berissi, H., and Kimchi, A. (1997). DAP-5, a novel homolog of eukaryotic translation initiation factor 4G isolated as a putative modulator of gamma interferon-induced programmed cell death. Mol Cell Biol 17, 1615–1625.

Lewis, S.M., Cerquozzi, S., Graber, T.E., Ungureanu, N.H., Andrews, M., and Holcik, M. (2008). The eIF4G homolog DAP5 supports the translation of select mRNAs during endoplasmic reticulum stress. Nucleic Acids Res 36, 168-178.

Li, Q., Imataka, H., Morino, S., Rogers, G.W., Richter-Cook, N.J., Merrick, W.C., and Sonenberg, N. (1999). Eukaryotic translation initiation factor 4AIII (eIF4AIII) is functionally distinct from eIF4AI and eIF4AII. Mol Cell Biol 19, 7336–7346.

Li, W. (2001). Eukaryotic Initiation Factors 4A (eIF4A) and 4G (eIF4G) Mutually Interact in a 1:1 Ratio in Vivo. Journal of Biological Chemistry 276, 29111–29115.

Liberman, N., Dym, O., Unger, T., Albeck, S., Peleg, Y., Jacobovitch, Y., Branzburg, A., Eisenstein, M., Marash, L., and Kimchi, A. (2008). The crystal structure of the C- terminal DAP5/p97 domain sheds light on the molecular basis for its processing by caspase cleavage. J Mol Biol 383, 539–548.

Liberman, N., Marash, L., and Kimchi, A. (2009). The translation initiation factor DAP5 is a regulator of cell survival during mitosis. Cell Cycle 8, 204–209.

Linder, P., Lasko, P.F., Ashburner, M., Leroy, P., Nielsen, P.J., Nishi, K., Schnier, J., and Slonimski, P.P. (1989). Birth of the D-E-A-D box. Nature 337, 121–122.

Livingstone, C. D., and Barton, G. J. (1993). Protein sequence alignments: a strategy for the hierarchical analysis of residue conservation. Comput. Appl. Biosci. 9, 745–756.

Lomakin, I.B., Hellen, C.U., and Pestova, T.V. (2000). Physical association of eukaryotic initiation factor 4G (eIF4G) with eIF4A strongly enhances binding of eIF4G to the internal ribosomal entry site of encephalomyocarditis virus and is required for internal initiation of translation. Mol Cell Biol 20, 6019–6029.

Lorsch, J.R., and Herschlag, D. (1998). The DEAD box protein eIF4A. 2. A cycle of nucleotide and RNA-dependent conformational changes. Biochemistry 37, 2194–2206.

68 Malina, A., Mills, J.R., and Pelletier, J. (2012). Emerging therapeutics targeting mRNA translation. Cold Spring Harb Perspect Biol 4, a012377.

Mallam, A.L., Del Campo, M., Gilman, B., Sidote, D.J., and Lambowitz, A.M. (2012). Structural basis for RNA-duplex recognition and unwinding by the DEAD-box helicase Mss116p. Nature.

Mamane, Y., Petroulakis, E., Rong, L., Yoshida, K., Ler, L.W., and Sonenberg, N. (2004). eIF4E--from translation to transformation. Oncogene 23, 3172–3179.

Marash, L., and Kimchi, A. (2005). DAP5 and IRES-mediated translation during programmed cell death. Cell Death Differ 12, 554–562.

Marash, L., Liberman, N., Henis-Korenblit, S., Sivan, G., Reem, E., Elroy-Stein, O., and Kimchi, A. (2008). DAP5 promotes cap-independent translation of Bcl-2 and CDK1 to facilitate cell survival during mitosis. Molecular Cell 30, 447–459.

Marcotrigiano, J., Gingras, A.C., Sonenberg, N., and Burley, S.K. (1997). Cocrystal structure of the messenger RNA 5' cap-binding protein (eIF4E) bound to 7-methyl-GDP. Cell 89, 951–961.

Marcotrigiano, J., Gingras, A.C., Sonenberg, N., and Burley, S.K. (1999). Cap-dependent translation initiation in eukaryotes is regulated by a molecular mimic of eIF4G. Molecular Cell 3, 707–716.

Marcotrigiano, J., Lomakin, I.B., Sonenberg, N., Pestova, T.V., Hellen, C.U., and Burley, S.K. (2001). A conserved HEAT domain within eIF4G directs assembly of the translation initiation machinery. Molecular Cell 7, 193–203.

Marintchev, A., and Wagner, G. (2005). eIF4G and CBP80 share a common origin and similar domain organization: implications for the structure and function of eIF4G. Biochemistry 44, 12265–12272.

Marissen, W.E., Gradi, A., Sonenberg, N., and Lloyd, R.E. (2000). Cleavage of eukaryotic translation initiation factor 4GII correlates with translation inhibition during apoptosis. Cell Death Differ 7, 1234–1243.

Mendell, J.T., Medghalchi, S.M., Lake, R.G., Noensie, E.N., and Dietz, H.C. (2000). Novel Upf2p orthologues suggest a functional link between translation initiation and nonsense surveillance complexes. Mol Cell Biol 20, 8944–8957.

Méthot, N., Pause, A., Hershey, J.W., and Sonenberg, N. (1994). The translation initiation factor eIF-4B contains an RNA-binding region that is distinct and independent from its ribonucleoprotein consensus sequence. Mol Cell Biol 14, 2307–2316.

Morino, S., Imataka, H., Svitkin, Y.V., Pestova, T.V., and Sonenberg, N. (2000). Eukaryotic translation initiation factor 4E (eIF4E) binding site and the middle one-third of eIF4GI constitute the core domain for cap-dependent translation, and the C-terminal

69 one-third functions as a modulatory region. Mol Cell Biol 20, 468–477.

Mossessova, E., and Lima, C.D. (2000). Ulp1-SUMO crystal structure and genetic analysis reveal conserved interactions and a regulatory element essential for cell growth in yeast. Molecular Cell 5, 865-876.

Nevins, T.A., Harder, Z.M., Korneluk, R.G., and Holcík, M. (2003). Distinct regulation of internal ribosome entry site-mediated translation following cellular stress is mediated by apoptotic fragments of eIF4G translation initiation factor family members eIF4GI and p97/DAP5/NAT1. J Biol Chem 278, 3572–3579.

Nielsen, K.H., Behrens, M.A., He, Y., Oliveira, C.L.P., Jensen, L.S., Hoffmann, S.V., Pedersen, J.S., and Andersen, G.R. (2011). Synergistic activation of eIF4A by eIF4B and eIF4G. Nucleic Acids Res. 39, 2678–2689.

Nielsen, P.J., and Trachsel, H. (1988). The mouse protein synthesis initiation factor 4A gene family includes two related functional genes which are differentially expressed. Embo J 7, 2097–2105.

Nousch, M., Reed, V., Bryson-Richardson, R.J., Currie, P.D., and Preiss, T. (2007). The eIF4G-homolog p97 can activate translation independent of caspase cleavage. Rna 13, 374–384.

Oberer, M., Marintchev, A., and Wagner, G. (2005). Structural basis for the enhancement of eIF4A helicase activity by eIF4G. Genes Dev 19, 2212–2223.

Olsen, J.V., Vermeulen, M., Santamaria, A., Kumar, C., Miller, M.L., Jensen, L.J., Gnad, F., Cox, J., Jensen, T.S., Nigg, E.A., et al. (2010). Quantitative phosphoproteomics reveals widespread full phosphorylation site occupancy during mitosis. Sci Signal 3, ra3.

Özeş, A.R., Feoktistova, K., Avanzino, B.C., and Fraser, C.S. (2011). Duplex unwinding and ATPase activities of the DEAD-box helicase eIF4A are coupled by eIF4G and eIF4B. J Mol Biol 412, 674–687.

Peabody, D.S. (1989). Translation initiation at non-AUG triplets in mammalian cells. J Biol Chem 264, 5031–5035.

Pestova, T.V., Hellen, C.U., and Shatsky, I.N. (1996a). Canonical eukaryotic initiation factors determine initiation of translation by internal ribosomal entry. Mol Cell Biol 16, 6859–6869.

Pestova, T.V., Shatsky, I.N., and Hellen, C.U. (1996b). Functional dissection of eukaryotic initiation factor 4F: the 4A subunit and the central domain of the 4G subunit are sufficient to mediate internal entry of 43S preinitiation complexes. Mol Cell Biol 16, 6870–6878.

Pestova, T.V., Shatsky, I.N., Fletcher, S.P., Jackson, R.J., and Hellen, C.U. (1998). A prokaryotic-like mode of cytoplasmic eukaryotic ribosome binding to the initiation codon

70 during internal translation initiation of hepatitis C and classical swine fever virus RNAs. Genes Dev 12, 67–83.

Pestova, T.V., and Kolupaeva, V.G. (2002). The roles of individual eukaryotic translation initiation factors in ribosomal scanning and initiation codon selection. Genes Dev 16, 2906–2922.

Petit, A.-P., Wohlbold, L., Bawankar, P., Huntzinger, E., Schmidt, S., Izaurralde, E., and Weichenrieder, O. (2012). The structural basis for the interaction between the CAF1 nuclease and the NOT1 scaffold of the human CCR4-NOT deadenylase complex. Nucleic Acids Res. 40, 11058–11072.

Ponting, C.P. (2000). Novel eIF4G domain homologues linking mRNA translation with nonsense-mediated mRNA decay. Trends Biochem Sci 25, 423-426.

Prévôt, D., Décimo, D., Herbreteau, C.H., Roux, F., Garin, J., Darlix, J.-L., and Ohlmann, T. (2003). Characterization of a novel RNA-binding region of eIF4GI critical for ribosomal scanning. Embo J 22, 1909–1921.

Prévôt, D., Darlix, J.-L., and Ohlmann, T. (2003). Conducting the initiation of protein synthesis: the role of eIF4G. Biol Cell 95, 141-156.

Pyronnet, S., Dostie, J., and Sonenberg, N. (2001). Suppression of cap-dependent translation in mitosis. Genes Dev 15, 2083–2093.

Pyronnet, S., Imataka, H., Gingras, A.C., Fukunaga, R., Hunter, T., and Sonenberg, N. (1999). Human eukaryotic translation initiation factor 4G (eIF4G) recruits mnk1 to phosphorylate eIF4E. Embo J 18, 270–279.

Qin, X., and Sarnow, P. (2004). Preferential translation of internal ribosome entry site- containing mRNAs during the mitotic cycle in mammalian cells. J Biol Chem 279, 13721–13728.

Read, R.J. (2001). Pushing the boundaries of molecular replacement with maximum likelihood. Acta Cryst.D 57, 1373-1382.

Richter-Cook, N.J., Dever, T.E., Hensold, J.O., and Merrick, W.C. (1998). Purification and characterization of a new eukaryotic protein translation factor. Eukaryotic initiation factor 4H. J Biol Chem 273, 7579–7587.

Richter, N.J., Rogers, G.W., Hensold, J.O., and Merrick, W.C. (1999). Further biochemical and kinetic characterization of human eukaryotic initiation factor 4H. J Biol Chem 274, 35415–35424.

Rogers, G.W., Richter, N.J., and Merrick, W.C. (1999). Biochemical and kinetic characterization of the RNA helicase activity of eukaryotic initiation factor 4A. J Biol Chem 274, 12236–12244.

71 Rogers, G.W., Richter, N.J., Lima, W.F., and Merrick, W.C. (2001). Modulation of the helicase activity of eIF4A by eIF4B, eIF4H, and eIF4F. J Biol Chem 276, 30914–30922.

Rozen, F., Edery, I., Meerovitch, K., Dever, T.E., Merrick, W.C., and Sonenberg, N. (1990). Bidirectional RNA helicase activity of eucaryotic translation initiation factors 4A and 4F. Mol Cell Biol 10, 1134–1144.

Rozovsky, N., Butterworth, A.C., and Moore, M.J. (2008). Interactions between eIF4AI and its accessory factors eIF4B and eIF4H. Rna 14, 2136–2148.

Safaee, N., Kozlov, G., Noronha, A.M., Xie, J., Wilds, C.J., and Gehring, K. (2012). Interdomain allostery promotes assembly of the poly(A) mRNA complex with PABP and eIF4G. Molecular Cell 48, 375–386.

Schütz, P., Bumann, M., Oberholzer, A.E., Bieniossek, C., Trachsel, H., Altmann, M., and Baumann, U. (2008). Crystal structure of the yeast eIF4A-eIF4G complex: an RNA- helicase controlled by protein-protein interactions. Proc Natl Acad Sci USA 105, 9564– 9569.

Schütz, P., Karlberg, T., van den Berg, S., Collins, R., Lehtiö, L., Högbom, M., Holmberg-Schiavone, L., Tempel, W., Park, H.-W., Hammarström, M., et al. (2010). Comparative structural analysis of human DEAD-box RNA helicases. PLoS ONE 5.

Shaughnessy, J.D., Jenkins, N.A., and Copeland, N.G. (1997). cDNA cloning, expression analysis, and chromosomal localization of a gene with high homology to wheat eIF- (iso)4F and mammalian eIF-4G. Genomics 39, 192–197.

Shibuya, T., Tange, T.Ø., Sonenberg, N., and Moore, M.J. (2004). eIF4AIII binds spliced mRNA in the exon junction complex and is essential for nonsense-mediated decay. Nat Struct Mol Biol 11, 346–351.

Sonenberg, N. (1981). Atp-Mg++-Dependent Cross-Linking of Cap Binding-Proteins to the 5' End of Eukaryotic Messenger-Rna. Nucleic Acids Res. 9, 1643–1656.

Sonenberg, N., and Hinnebusch, A.G. (2009). Regulation of translation initiation in eukaryotes: mechanisms and biological targets. Cell 136, 731–745.

Story, R.M., and Steitz, T.A. (1992). Structure of the recA protein–ADP complex. Nature 355, 374–376.

Studier, F.W. (2005). Protein production by auto-induction in high density shaking cultures. Protein Expression and Purification 41, 207-234.

Suzuki, C., Garces, R.G., Edmonds, K.A., Hiller, S., Hyberts, S.G., Marintchev, A., and Wagner, G. (2008). PDCD4 inhibits translation initiation by binding to eIF4A using both its MA3 domains. Proc Natl Acad Sci USA 105, 3274–3279.

Svitkin, Y.V., Gradi, A., Imataka, H., Morino, S., and Sonenberg, N. (1999). Eukaryotic

72 initiation factor 4GII (eIF4GII), but not eIF4GI, cleavage correlates with inhibition of host cell protein synthesis after human rhinovirus infection. J. Virol. 73, 3467–3472.

Svitkin, Y.V., Herdy, B., Costa-Mattioli, M., Gingras, A.-C., Raught, B., and Sonenberg, N. (2005). Eukaryotic translation initiation factor 4E availability controls the switch between cap-dependent and internal ribosomal entry site-mediated translation. Mol Cell Biol 25, 10556–10565.

Terrano, D.T., Upreti, M., and Chambers, T.C. (2010). Cyclin-dependent kinase 1- mediated Bcl-xL/Bcl-2 phosphorylation acts as a functional link coupling mitotic arrest and apoptosis. Mol Cell Biol 30, 640–656.

Topisirovic, I., Ruiz-Gutierrez, M., and Borden, K.L.B. (2004). Phosphorylation of the eukaryotic translation initiation factor eIF4E contributes to its transformation and mRNA transport activities. Cancer Res. 64, 8639–8642.

Warnakulasuriyarachchi, D., Cerquozzi, S., Cheung, H.H., and Holcík, M. (2004). Translational induction of the inhibitor of apoptosis protein HIAP2 during endoplasmic reticulum stress attenuates cell death and is mediated via an inducible internal ribosome entry site element. J Biol Chem 279, 17148–17157.

Wendel, H.-G., Silva, R.L.A., Malina, A., Mills, J.R., Zhu, H., Ueda, T., Watanabe- Fukunaga, R., Fukunaga, R., Teruya-Feldstein, J., Pelletier, J., et al. (2007). Dissecting eIF4E action in tumorigenesis. Genes Dev 21, 3232–3237.

Wilson, J.E., Pestova, T.V., Hellen, C.U., and Sarnow, P. (2000). Initiation of protein synthesis from the A site of the ribosome. Cell 102, 511–520.

Yamanaka, S., Poksay, K.S., Arnold, K.S., and Innerarity, T.L. (1997). A novel translational repressor mRNA is edited extensively in livers containing tumors caused by the transgene expression of the apoB mRNA-editing enzyme. Genes Dev 11, 321–333.

Yamanaka, S., Zhang, X.Y., Maeda, M., Miura, K., Wang, S., Farese, R.V., Iwao, H., and Innerarity, T.L. (2000). Essential role of NAT1/p97/DAP5 in embryonic differentiation and the retinoic acid pathway. Embo J 19, 5533–5541.

Yang, H.-S., Jansen, A.P., Komar, A.A., Zheng, X., Merrick, W.C., Costes, S., Lockett, S.J., Sonenberg, N., and Colburn, N.H. (2003). The transformation suppressor Pdcd4 is a novel eukaryotic translation initiation factor 4A binding protein that inhibits translation. Mol Cell Biol 23, 26–37.

Yang, D.-Q., Halaby, M.-J., and Zhang, Y. (2006). The identification of an internal ribosomal entry site in the 5'-untranslated region of p53 mRNA provides a novel mechanism for the regulation of its translation following DNA damage. Oncogene 25, 4613–4619.

Yang, Q., and Jankowsky, E. (2006). The DEAD-box protein Ded1 unwinds RNA duplexes by a mode distinct from translocating helicases. Nat Struct Mol Biol 13, 981–

73 986.

74