University Pierre and Marie Curie Identification of New Regulators For

University Pierre and Marie Curie

Doctoral school : Complexité Du Vivant - ED515 Nuclear Organization and Oncogenesis Unit (INSERM U993)

Identification of new regulators for PML Nuclear Bodies

By Thibaut Snollaerts

Doctoral thesis of BIOCHEMISTRY AND MOLECULAR BIOLOGY OF THE CELL

Supervised by Anne Dejean

Presented and defended publicly on September 28th 2016

In front of a jury composed of:

Dr. Laurent LE CAM INSERM research director Reviewer Dr. Dimitris XIRODIMAS CNRS research director Reviewer Dr. Mounira CHELBI-ALIX INSERM research director Examinator Dr. Robert WEIL CNRS research director Examinator Dr. Frederic DEVAUX UPMC university professor President Prof. Anne DEJEAN INSERM research director Thesis Director (Invited member)

La persévérance, c'est ce qui rend l'impossible possible, le possible probable et le probable réalisé.

Perseverance is what makes the impossible possible, the possible probable and the probable realized

Robert Half

Acknowledgements

First, I would like to thank all the members of the jury: Laurent Le Cam, Dimitris Xirodimas, Mounira Chelbi-Alix, Robert Weil and Frederic Devaux who, despite a very busy schedule, took the time to take part in this thesis defense jury.

I would like to thank Anne Dejean for giving me the opportunity to be part of her research team since my master internship, for giving me the chance to meet very interesting people, and for providing me with many opportunities. Thank you for the trust you put in me to carry out this complicated project as well as for providing the best working conditions, and collaborations to realize it. Thank you for all the things I was able to learn from you, your team and the world of science in which we are involved.

I would like to thank the entire ONO team for their support during these past few, short, years: Thank you to the past members: Adrien, the basketball player from room 306, and Pierre for your help, advices and good sense of humor; Yoon Ra for your kindness and zeness; Ricardo for bringing a little bit of Mexican spirit to the lab; Louise-Mary for all your help with the administrative paperwork and Sophie aka Fofi, aka “la personne la plus désagréable du labo, mais qui en vrai, est plus ou moins relativement super sympa”. A big thank for your help and support to you and all the people I might have met during these years.

I would like to give a special thank you to Hélène-Neyret Khan for giving me the opportunity to take over one of her projects. And also for her patience, kindness and understanding while teaching most of the techniques used for this project. Thank you for keeping contact, for supporting and helping me, even though, you moved on to another lab.

Thank you to the present members of the team : Alexandra, merci pour ton aide précieuse sur ce projet notamment à compter des points verts sans devenir folle. Merci aussi pour ton soutien et ton franc parler tout au long de ces années. Thank you Elefteria (Elma) for your support, for your friendship and all your advices over the years. Thank you for listening, understanding, supporting and for being one hell of a crazy friend. Pablo, muchas gracias for

your kindness, enthusiasm, optimism and for our scientific discussions on the project; Juan Pablo, my statistics genius, mi amigo, for all your help inside and out of the lab; Ying for sharing crazy Chinese pastries; Jack for your help, good sense of humor and support; Ilan for bringing some craziness in the lab; Oliver for your insights in the project and your good spirit; Lucas for bringing a Canadian vibe to the lab; Pierre Francois and Greg, your communicative sense of humor, for all the laughs and good times; Jacob, king of the cloning, for your help and knowledge on the project; Pierre Tiollais, for all of your stories, Friday seminar organization and of course the design of this amazing fridge, I mean building! Thank you to all the interns for making this lab so lively.

Thank you people from room 307; Agnes, qPCR master, for being there to listen and help me with my various problems. Thank you Pascal for sharing your passion of science and politics, for your availability under any circumstances, for your help, for listening and for everything else.

Thank you Florence for helping us with all of the administrative paperwork. Thank you for your good sense of humor and energy.

Merci à Chaty et Monique, qui parfois dans l’ombre, nous facilitent tant le travail.

I would also like to thank all of my floor mates and in particular people from the Hepacivirus team: Barbara, the one and only unicorn from Britany; Pierick, aka the seagull, for the famous “Soirée Piscine!TM”; Dona, thank you for your support (long live ferrets and penguins!). Thank you to all the other members of the team: Christine; Stephanie; Mila; Aurora and Patrick. I would also like to extend that thanks to the fabulous people of our building: Claudia; Iratxe; Florence; Yu; but also from Pasteur: Mathilde; Lilliana; Emmanuelle and all the people I might have missed, thank you all.

I would like to thank all my friends outside the Institute who have also supported me. Thank you Paul, for always having my back for all these years; Jessica for your support, help and Tanguy, for your good sense of humor and for making teaching so interesting; Dr David Moffet for introducing me to this crazy world of science. Thank you Raphael; Lola; Joel; Estelle; Leo; Elelta and of course all the friends that are not mentioned here.

Thank you Camille for your support, patience and understanding over the years. Thank you to your family for their support.

I would like to thank my family in the broad sense for believing in me and supporting me, especially my parents for allowing me to pursue my dreams. Thank you for your patience and unwavering support. I would also like to give a special thanks to my brothers and sister who were always supporting me no matter what. A big extra thank you for Audrey who helped me a lot with corrections. Thank you all for being here, family rocks!

I would like -to once again- thank all the people who helped me shape this thesis and make it what it is. Thank you all, family/friends for your support, it means a lot for me. Over the past years, you were always there for me, no matter the circumstances, no matter how hard or difficult the situation might have been. I am actually very proud to be one of your friend/family and I hope that I would never let you down. Thank you for the smiles; the good and bad jokes; for coffee breaks (thank you coffee machine), for reminding me to eat and sleep (thank you to the night guard for checking that I was still alive) and for everything. I would like to thank the -80 freezer for breaking on a Sunday night at 11:30pm; penguins because they are funny, and because I have one yelling at me every time I open my fridge; and otters, because I can and I like them.

Finally, thank you reader, for taking the time to look at all the people that helped me in this project. I hope you will enjoy reading this work and find it as interesting as I did.

Collaborations

Because this work is also the result of a collaborative effort, I would like to thank our collaborators for their help by providing their expertise and tools to this project.

- Michael Howell from the “London Research Institute” and the members of the high throughput-screening platform, who helped realize the primary and secondary siRNA screening.

- Michele Pagano from the “Howard Hughes Medical Institute” of New York, who shared knowledge on the SCF complex systems and F-Box proteins, as well as, implementing the first co-immunoprecipitating screening for the F-Box.

- Florian Bassermann and his team, and in particular Vanessa Fernandez, from the School of Medicine at the Technical University of Munich who helped us with the in- vitro experiments and provided us with FBXO9 antibody.

- All the people and scientists who I had the chance to encounter at the Pasteur Institute and in conferences, which led to discussion allowing to further this project.

Table of Contents

Acknowledgements ...... 3 Collaborations ...... 6 Table of Contents ...... 7 Abbreviations ...... 10 Introduction ...... 11 I) Discovery of ProMyelocytic Leukemia protein (PML) ...... 11 1) Acute Promyelocytic Leukemia (APL) ...... 11 2) The PML protein ...... 15 3) PML Nuclear Bodies ...... 19 II) Post-translational modifications ...... 23 1) Diversity of post-translational modifications ...... 24 2) Phosphorylation ...... 26 3) Acetylation ...... 29 4) SUMOylation ...... 30 5) The SUMO/Ubiquitin coupled pathway ...... 38 6) Ubiquitination ...... 41 III) The SKP-Cullin-F-Box containing Complex (SCF) ...... 49 1) General structure of Cullin-RING Ligases complex ...... 49 2) The SCF complex and the F-Box proteins ...... 50 3) Regulation of the SCF complex ...... 53 4) Substrate recognition ...... 57 5) SCFs and diseases ...... 60 6) PML and SCF ...... 64 IV) PML, a tumor suppressor...... 65 1) PML physiological functions ...... 65 2) PML and diseases ...... 66 3) PML and the apoptotic pathway ...... 68 4) PML and the P53 pathway ...... 68 5) PML and transcriptional regulation ...... 69 6) Role of PML in DNA damage repair ...... 70 7) PML and the Akt pathway ...... 71 8) Cytoplasmic PML in tumorigenesis...... 72 9) PML in cancers ...... 73 Thesis Objective ...... 74 Results ...... 76 I) Initial Data ...... 77 1) Screen Design ...... 77

2) Primary screen results ...... 79 3) Validation screen ...... 80 II) Functional study of selected candidates ...... 82 1) Identification of SKP1a and RBX1 ...... 82 2) Manual validation of screen results ...... 83 3) Depletion of SKP1a and RBX1 stabilizes PML ...... 84 4) The overexpression of SKP1a and RBX1 destabilizes PML ...... 86 5) RBX1 and SKP1a are both interacting with PML ...... 88 6) Identification of the specific F-Box protein for PML ...... 89 7) Validation of the interaction between PML and FBXO9 ...... 90 8) FBXO9 interacts with all PML isoforms ...... 92 9) Localization of FBXO9-PML interaction ...... 93 10) SUMOylation, arsenic and PML-FBXO9 interaction ...... 95 11) FBXO9 degrades PML under arsenic trioxide treatment ...... 97 12) Impact of FBXO9 on the half-life of PML ...... 106 13) SCFFBXO9 ubiquitinates PML in vitro ...... 107 14) An attempt to localize the FBXO9 degron on PML ...... 110 15) Kinase mini-screen to localize the FBXO9 degron on PML ...... 112 16) Search for physiological stimuli leading to SCFFBXO9-induced PML degradation . 114 17) Possible links to diseases ...... 116 Discussion and perspectives ...... 120 1) SKP1a and RBX1 are members of an ubiquitination complex involved in the degradation of PML ...... 120 2) The Cullin-1 is involved in SCF complexes ...... 122 3) FBXO9 specifically interacts with PML ...... 123 4) FBXO9 is involved in PML stability ...... 123 5) Is PML degradation dependent on SUMOylation? ...... 125 6) SCFFBXO9 specifically ubiquitinates PML ...... 125 7) PML’s degron ...... 126 8) Kinases phosphorylating PML’s degron ...... 127 9) PML and FBXO9 in diseases ...... 131 10) Mouse model Fbxo9 KO ...... 133 11) PML and cellular differentiation ...... 133 12) PML and innate immunity ...... 134 13) The SCF complex: a druggable target ...... 135 14) Other potential candidates to be studied ...... 138 Conclusion ...... 141 Material and Methods ...... 142 Annexes ...... 150 Annex 1: The 69 Mammalian F-Box proteins ...... 150 Annex 2: F-Box protein and E3 ubiquitin ligase implication in cellular pathways...... 152

Annex 3: Validated candidates inducing a morphological change of PML Nuclear Bodies...... 153 Annex 4: Co-immunoprecipitation screen to identify PML interacting F-Box protein. .... 165 Annex 5: Cell lines used in mRNA mini screen...... 166 Annex 6: FBXO9 is overexpressed in some types of breast cancers...... 169 Annex 7: Cancer tissue PML antibody staining of Breast and lung cancers...... 170 Bibliographical References ...... 171 Table and illustration table ...... 194 Abstract: ...... 197

Abbreviations

Phosphorylation-Dependent ALT: Alternative Lenghtening of Telomere PDSM: SUMOylation Motif APL: Acute Promelocytic Leukemia PIAS: Protein Inhibitor of Activated STAT ATR: Ataxia Telangiectasia and Rad3-related protein PML: ProMelocytic Leukemia protein

ATO: Arsenic TriOxide (AS2O3, Ars) PML-NBs: PML Nuclear Bodies ATRA: All-Trans Retinoic Acid PTMs: Post-Translational Modifications CAND1: Cullin-Associated NEDD8-Dissociated protein 1 RARα: Retinoic Acid Receptor α CDK1/2: Cyclin-Dependent Kinase 1/2 Rb: Retinoblastma protein RBCC CK2: Casein Kinase 2 (α1, α2, α3, β) RING B-Box Coiled-Coil domain domain: CRL: Cullin-Ring Ligases RBX1: Ring-BoX protein 1 CUL1: CULlin-1 RING: Really Interesting New Gene DUBs: DeUBiquitinating enzyme RNF4: RING Finger protein 4 E1: Activation Enzyme SAE1: SUMO Activating Enzyme 1 E2: Conjugation Enzyme SAE2: SUMO Activating Enzyme 2 SKP-Cullin-F-Box containing E3: Ligation Enzyme SCF: complex ERK: Extracellular signal-Regulated Kinase SENP: SENtrin specific Proteases FBXL: F-BoX Leucine rich repeats protein shRNA: small hairpin RNA FBXO9: F-BoX Only protein 9 SIM: SUMO Interacting Motif FBXO: F-BoX Only protein siRNA: small interfering RNA FBXW: F-BoX WD40 repeats protein SKP1a: S-phase Kinase-associated Protein 1 GFP: Green Fluorescent Protein STUbL: SUMO Targeted Ubiquitin Ligases HAT: Histone Acetyl Transferase SUMO: Small Ubiquitin like MOdifier HDAC: Histone DeACetylase TNF: Tumor Necrosis Factor HECT: Homologous to the E6-AP Carboxyl Terminus TRIM: TRIpartite Motif-containing protein HIPK2: Homeodomain-Interacting Protein Kinase 2 Ub: Ubiquitin IQR Inter Quartile Range UBA1: UBiquitin Activating enzyme 1 KO: Knock-Out UBC: UBiquitin Conjugating enzyme MAPK: Mitogen-Activated Protein Kinase UBC9: UBiquitin Conjugating enzyme 9 MDM2: Murine Double Minute 2 UBD: Ubiquitin Binding Domain MEFs: Murine Embryonic Fibroblasts UPS: Ubiquitin Proteasome System mRNA: messenger RNA WT: Wild Type Negatively charged amino-acid-Dependent NDSM: SUMOylation Motif NEDD8: Neural precursor cell Expressed Developmentally Down-regulated protein 8 NES: Nuclear Export Signal NLS: Nuclear Localization Signal

Introduction

Statistics available from the World Health Organization state that around 14 million new cases of cancer were diagnosed in 2012. They caused 8.2 million deaths worldwide. As such, tumoral diseases are recognized as being among the leading causes of morbidity and mortality worldwide and in addition are predicted to increase in number (Stewart & Wild 2014). Cancer is linked to the accumulation of genetic or epigenetic events that enable uncontrolled proliferation of cells. One type of cancer, leukemia, is caused by the malignant proliferation of cells derived from bone marrow. Leukemic cells disrupt the process called hematopoiesis and invade distant organs as well as the bloodstream. There are different forms of leukemia, which range from relatively un-impactful conditions -and therefore rarely shorten life expectancy- to highly malignant cases, for which very few therapeutic options are available to date. Leukemias are usually classified as myeloid, lymphoid, chronic or acute, depending on the phenotype of the malignant cells. However, even though each class has specific clinic-biological features, they often share common traits like anemia, hemorrhages mainly caused by the loss of platelets (thrombopenia), as well as infections related to myeloid and lymphoid deficiencies. Thanks to treatments involving inhibitors of nucleotide synthesis or DNA replication, such as DNA cross- linkers and topoisomerase inhibitors, some Acute Leukemias can be cured definitively in a great majority of patients. Most of who are children suffering from lymphoblastic leukemia. Such favorable prognosis is far less frequent in other types of acute or chronic Leukemias (de Thé et al. 2012). Acute Promyelocytic Leukemia is a well-studied disease cured through arsenic based treatment and used as a model to study PML.

I) Discovery of ProMyelocytic Leukemia protein (PML)

1) Acute Promyelocytic Leukemia (APL)

a) The disease

The M3 subtype of Acute Myeloid Leukemia (AML-M3), also known as Acute Promyelocytic Leukemia (APL) is a rare condition, with around 100 new cases per year in France (de Thé et al. 2012), around 10% of all AML cases. It is one of the most malignant conditions due to its rapid and spontaneous evolution, as well as its sudden hemorrhages. It was first described by

the Swedish hematologist Hillestad in 1957 (Hillestad 1957). Hemorrhages come from a coagulation disorder, due to the release of coagulation cascade activating agents from APL cells, and a low amount of platelets in the blood (de Thé et al. 2012). Normally, blood cells are produced through a process called hematopoiesis, from myeloid progenitor cells (Figure 1). In APL, however, myeloid differentiation is blocked at the promyelocyte stage causing a heavy burden of leukemia blasts (Mi et al. 2015).

Figure 1 : Hematopoietic differentiation from hematopoietic stem cells to mature cells. (Wikipedia n.d.)

b) The genetic defect

APL is the result of specific chromosomal translocation always involving Retinoic Acid (RA) Receptor α (RARα), present on chromosome 17. In more than 98% of cases, a gene called ProMyelocytic Leukemia (PML, also known as MYL, RNF71, TRIM19 or PP8675) present on chromosome 15, can also be translocated (Kakizuka et al. 1991; de Thé et al. 1991; de Thé et al. 2012; Mi et al. 2015) (Figure 2). The second most common translocation t(11.17) is encoding for PLZF/RARα, and is clinically associated with Retinoic Acid resistant APL and a poorer prognosis (Licht et al. 1995). Promyelocytic Leukemia Zinc Finger protein (PLZF) is a transcriptional repressor and epigenetic regulator involved in hematopoietic stem cell

quiescence or natural Killer T cells formation (McConnell et al. 2015). Here however, we are interested in the translocation t(15;17) (q22;q21) resulting in the production of a fusion oncoprotein called ProMyelocytic Leukemia-Retinoic Acid Receptor α (PML/RARα) that is capable of blocking cell differentiation at the promyelocitic stage (Wang et al. 2010).

Figure 2 : PML-RARα fusion protein comes from the t(15;17) (q22;q21) translocation. (Lo-Coco & Hasan 2014)

c) PML-RARα dominant negative effect on Retinoic Acid Receptor α

RARα is a Retinoic Acid (RA) nuclear receptor acting as a hormone dependent transcriptional switch. Retinoids are derivatives from vitamin A, and have very important effects on development, differentiation and cell proliferation, by regulating specific gene expression. RA is implicated in many types of cellular response, such as differentiation from multiple progenitors cell, as well as myeloid cells (de Thé et al. 2012). Retinoids can interact with two classes of nuclear receptor proteins: the steroid and thyroid hormone superfamily receptors, the RARs (which include RARα) and the Retinoid X Receptors (RXR). Both receptors can be activated by 9-cis-RA, but All-Trans Retinoic Acid (ATRA) only activates RARα (Lo-Coco & Hasan 2014). Under normal conditions, RARα interacts with RXR to form the RARα/RXR heterodimer that will bind typical Retinoic Acid Response Elements (RAREs) generally located in target genes promoters. In absence of RA, the heterodimer can recruit a corepressor complex (CoR) notably composed of histone deacetylases (HDACs) and will repress transcription of the target genes. However, in presence of physiological concentrations of RA (1x10-9M) (or in presence of ATRA), a conformational change occurs and dissociation of the corepressors ensues promoting the recruitment of coactivators (CoA) with histone acetyltransferase activity leading to a chromatin remodeling and subsequent transcriptional activation (Figure 3A).

However, in APL, PML-RARα behaves as an altered RARα that can oligomerise with or without RXR. PML-RARα acts as a constitutive repressor, capable of recruiting a CoR that is not sensitive to physiological concentration of ATRA, finally leading to the characteristic block in differentiation observed in APL (Martens et al. 2010; Mi et al. 2015; Lo-Coco & Hasan 2014). In addition, PML-RARα also binds the hematopoietic transcription factor PU.1 leading to repression of genes depending on it, like glycolytic enzyme HK3 needed for hematopoietic differentiation (Wang et al. 2010; Martens et al. 2010) (Figure 3B). The PML moiety of the oncoprotein also plays a role in APL self-renewal acquisition and differentiation block, by weakening through deacetylation the P53 response to DNA damage (Mi et al. 2015), and by blocking senescence (Korf et al. 2014). However, although both the function of RARα and the effects of the PML-RARα fusion have been clearly characterized, the data available from the PML counterpart does not lead to a clear model of its function and regulation. So what do we know so far about PML?

Figure 3 : Retinoic Acid Receptor α and PML-RAR α function in normal and APL cells. A. RAR/RXR heterodimer function in presence or not of ATRA. B. PML-RAR interaction leading to differentiation block in APL. (Mi et al. 2015)

2) The PML protein

a) PML isoforms and nomenclature

ProMelocytic Leukemia protein (PML) is a multi-faceted protein that plays key roles in cellular events under physiological and pathological conditions. PML is a strongly conserved protein expressed in all mammals, testifying of its relevant role in cell function (Cheng & Kao 2012) (Figure 4).

Figure 4 : PML protein is conserved in mammals. PML protein phylogenetic tree based on maximum likelihood method with amino acid substitution (Jones-Taylor-Thornton model). Bootstrap values are displayed on branches. (Cheng & Kao 2012)

Seven PML isoforms are generated through alternative splicing of a single PML gene in the 3̍ exons. This gene is composed of a total of nine exons of which exons 7 and 8 can be divided in two (a and b) and can sometimes contain intronic sequences (Figure 5A). Because of the very high complexity of the gene alternative splicing to form PML isoforms, difficulties rose in the attempt to reach a unifying nomenclature between data banks (GeneBank, NCBI and UniProt), although finally a unified name nomenclature was adopted (Jensen et al. 2001) (Figure 5B). The result of this complex splicing generated six nuclear and one cytoplasmic isoform of PML, which could be experimentally validated. However, other isoforms can be expressed missing exons 4, 5 or 6. In these cases, a letter is added after their name depending on which exons are missing: “a” for isoforms missing exon 5, “b” for missing exons 5 and 6 and “c” for missing exons 4, 5 and 6. For example, PML IVa corresponds to PML isoform IV

missing exon 5. The longest isoform, PML I, is 882 amino acids whereas the shortest is about half of the size at 435 amino acids (PML VII) (Figure 5C) (Nisole et al. 2013).

Figure 5 : PML protein isoforms and nomenclature generated from PML gene alternative splicing. A. PML gene contains nine exons, exons 7 and 8 can be separated into “a” and “b”. Some intronic sequences can be included. B. PML isoforms nomenclature. C. PML protein isoforms encoded by mRNA variants. Asterisks indicate the presence of an incomplete exon or intron caused by an in-frame STOP codon (Nisole et al. 2013).

b) Domains and structure of PML

All isoforms have a common sequence, encoded in exons 1 through 4, which contains the TRIpartite Motif (TRIM). TRIM proteins form a wide family of proteins involved in a variety of cellular processes such as cell differentiation, cell growth, development, apoptosis, and morphogenesis (Munir et al. 2010). More recently, an increasing number of studies are also looking into the role of TRIM proteins in immune signaling. PML is part of the TRIM family (TRIM 19) and possesses, like all TRIMs, an RBCC domain composed of a Really Interesting New Gene (RING) finger domain (R), two cysteine/histidine-rich B-box domains (B1 and B2) and an α-helical coiled-coil domain (CC) (Figure 6). TRIM proteins possess homo and hetero dimerization properties through their α-helical coiled-coil domain. However, PML seems to only homodimerize (Guan & Kao 2015; Reymond et al. 2001). Both B-boxes and RING domains (involving Cysteine and Histidine residues) require zinc ions to stabilize their structure. These structures could be used for protein-protein, protein-DNA or protein-RNA interactions. In addition, PML isoforms I to VII hold a Nuclear Localization Signal (NLS) and isoforms I to V have a SUMO interacting Motif (SIM). Only PML isoform I contains a Nuclear Export Signal (NES) (Figure 6). There are relatively few information on PML structure except from the well-studied TRIM motif meaning that structural information for more than two thirds of the protein are currently unknown.

Figure 6 : PML protein domain and structure. PML functional domains diagram with RING (R); B-Box1 (B1); B-Box2 (B2) and coiled coil (CC) domains present in all isoforms. Nuclear Export Signal (NES) is only present in PML I while Nuclear Localization Signal (NLS) and SUMO Interacting Motif (SIM) are only present in PML I to V (Guan & Kao 2015).

c) Differential role of PML isoforms

The existence of that many different PML isoforms with such variability in their C-terminus suggests probable specific functions. Here are a few examples of specific functions discovered for PML isoforms:

PML I is the most abundant isoform (Condemine et al. 2006), interacts directly with Herpes Simplex Virus 1 Ubiquitin ligase ICP0 (Cuchet-Lourenco et al. 2012); PML-I can also form a complex with AML1 (Acute myeloid leukemia 1 protein) leading to the transcriptional activation of genes driving myeloid cell differentiation (Nguyen et al. 2005). PML I is also the only isoform to possess a Nuclear Export Signal (NES).

PML II is involved in innate immune response as it is specifically required for the induction of IFN-stimulated gene transcription in response to type I interferon (IFN) like NF- κB, STAT1 and CBP (Y. Chen et al. 2015). PML II is also involved in preventing Human Adenovirus type 5 infection through interferon and HSP70 (Atwan et al. 2016).

PML III is interacting with the centrosome, an organelle regulating cell cycle progression, and controls its duplication through its interaction with Aurora B kinase (Xu et al. 2005). It is also suggested to interact and work with the tumor suppressor TIP60 (also known as Histone acetyltransferase KAT5) (Wu et al. 2009).

PML IV is probably the most studied isoform of PML, it is involved in apoptosis regulation, DNA damage and senescence (Bischof et al. 2002; Guo et al. 2000). It was also thoroughly studied for its interaction or indirect effect on P53 as well as for its implication in immune response against viruses, such as rabies virus (Blondel et al. 2010; El Asmi et al. 2014; Ivanschitz et al. 2015).

PML V possess an α-helix in its C-terminal domain that allows for strong homodimerization and seems to be important for recruitment of key partner proteins such as Death Domain-Associated protein 6 (DAXX) or SP100 to particular nuclear structures (Weidtkamp-Peters et al. 2008; Geng et al. 2012).

PML VI is not very well studied and the few papers available have conflicting results. A study states that PML VI was resistant to arsenic trioxide induced degradation (Maroui et al. 2012) while another one showed that it was not the case (Hands et al. 2014).

PML VII also known as PMLc (cytoplasmic), was found to be essential for the activation of Transforming Growth Factor Beta (TGFβ) signaling through its interaction with SMAD2/3 and Smad Anchor for Receptor Activation (SARA) (Carracedo et al. 2011; Lin et al. 2004). A recent study underlined the importance of studying this pathways as it plays a key role in some diseases like prostate cancer (Buczek et al. 2015). Cytoplasmic PML was also shown to have protective effects against viral infections through cytoplasmic sequestration of key proteins (McNally et al. 2008; Nisole et al. 2013).

Despite the high diversity of isoforms, PML function appears to be mediated mostly through the formation of complex nuclear structures involving several isoforms, called the PML Nuclear Bodies.

3) PML Nuclear Bodies

Nuclear punctate structures enriched in PML were first described to be localized around the chromatin. These spherical nuclear speckles structures were observed and given different names over the years: Kremer bodies, ND10 (Nuclear domain 10), POD (PML oncogenic domains) or PML Nuclear Bodies (PML-NBs) (Hodges et al. 1998). These structures are heterogeneous at the protein level and very dynamic as many proteins uses PML-NBs as temporary storage or as a platform to get modified and/or interact with other proteins. If PML is not present in the cell, PML-NBs do not form indicating that PML is the scaffold protein required for the formation of these structures (Lallemand-Breitenbach et al. 2010; Batty et al. 2012; Seeler & Dejean 1999). To that end the RBCC domain of PML is also essential (Shen et al. 2006; Jensen et al. 2001; Guan & Kao 2015). Another major component of PML-NBs is the protein SP100, a transcriptional regulator that also happened to be the first and major Nuclear Body associated protein described (Szostecki et al. 1990). Another common component of these bodies is DAXX which is involved in apoptosis regulation (Lallemand-Breitenbach et al. 2010). The size of these structures ranges from 0.1 to 1 µm in diameter. From 5 to 30 PML NBs can be found per nucleus depending on cell types, phase of the cell cycle, stress and nutritional conditions (Figure 7). For example, the number and size of PML NBs will increase under interferon response (Salomoni & Pandolfi 2002; Guan & Kao 2015; Everett et al. 1999).

Figure 7 : PML Nuclear Bodies. HT1080 cell stably expressing GFP-PML IV (fluorescence microscopy). PML Nuclear Bodies appear in green while, the nucleus is stained using DAPI (blue).

The question is how do these structures form. So far, two models have been proposed to explain the genesis of PML NBs. Based on the identification of Small Ubiquitin like Modifier (SUMO) Interacting Motif (SIM) in the Carboxyl terminus of the protein and the sequence requirements in PML to observe co-localization with GFP-SUMO1, the first model proposes that the nucleation of PML-NBs depend on PML SUMOylation and non-covalent interaction between SUMOylated PML and its SIM (Shen et al. 2006). SUMOylation is a type of post- translational modification that will be discussed later. In addition, this model is comforted by the fact that PML mutants, no longer able to be modified by SUMO, do not form structured PML-NBs but rather aggregates in nucleus thus emphasizing the essential role of SUMO modification in the formation of these nuclear structures (Zhong et al. 2000) (Figure 8).

Figure 8 : PML Nuclear Body formation requires SUMO modification. Mouse Embryonic Fibroblasts (MEFs) PML-/- immunofluorescence expressing either WT or a non-SUMOylable mutant of (PML-ΔSUMO) (Zhong et al. 2000).

However, PML VI, which does not have the SIM is still capable of forming PML NBs and polymers most likely through its RBCC domain (Shen et al. 2006) leading us to the second

model. This new model suggests a two-step process leading to PML NB formation. Firstly, Reactive Oxygen Species (ROS) such as peroxides, superoxides, hydroxyl radicals or singlet oxygen induce PML oxidation resulting in the formation of disulfide covalent bonds between PML monomers. At this point, PML NB shell becomes associated with the nuclear matrix. Moreover, this seeding step is an accurate reflection of the redox status of the cell. A second step involves UBC9 -the SUMO conjugating enzyme- recruitment causing an increase of PML SUMOylation. In the third and fourth step, a SIM-SUMO-dependent mechanism recruits SUMOylated or SIM-containing partners such as DAXX or SP100, leading to an increased interaction and mature PML NB formation (Sahin Umut et al. 2014). This model also helps to give an explanation regarding the nuclear aggregates formation by non-SUMOylable mutant or SIM deficient isoforms of PML (Figure 9).

Figure 9 : PML Nuclear Body biogenesis model. Reactive Oxygen Species (ROS) induce PML oligomerization through disulphide bonds. The E2 SUMO conjugating enzyme UBC9 is recruited. Finally, partners are mobilized and form the mature PML NBs (Sahin et al. 2015).

An important and increasing amount of proteins are associated to PML NBs and it has been estimated that more than 160 proteins functionally interact with PML directly or indirectly (Van Damme et al. 2010). Around 120 of these proteins were reported to physically interact with PML as shown through data obtained and analyzed from BIOGRID (http://www.thebiogrid.org/). The data suggest a potential co-regulation of some of its partners and explain PML NBs involvement in different essential cellular functions such as transcription regulation, apoptosis or stress response (Guan & Kao 2015) (Figure 10).

Figure 10 : PML interactome based on data from BIOGRID. Data obtained from BIOGRID (http://www.thebiogrid.org/) of 120 proteins interacting with PML through affinity interactions (immunoprecipitation) followed by Western blotting experiments. The more publication referencing the interaction, the thicker the line is (Guan & Kao 2015).

Most of these interactions are made possible thanks to post-translational modifications. PML is a heavily modified protein: these modifications regulate the ability of PML to interact

with various partners and allow for stress- and signal-dependent regulation of PML or its interacting partners.

II) Post-translational modifications

As previously described the PML gene is associated with a great diversity of proteins that can be created from a few number of exons thanks to the process of alternative splicing. It is estimated that genes from the entire genome can produce in average around five to six different messenger RNAs (mRNAs). Altogether, the mRNAs comprise the transcriptome, which will encode for as many slightly different proteins. Moreover, another layer of diversification is available at the post-translational level usually by adding simple or multiple modifications through covalent bonding of short amino acids chains or molecules. This process called post- translational modifications can produce up to 10 or more different forms of the same protein. All together, they form the proteome of the cell (Figure 11). In addition to simply increasing the variety of proteins, these modifications provide and regulate protein function and/or cellular life. They also allow the cell to adapt in a very dynamic and fined tuned way to internal and external stimulus by directly acting on existing proteins rather than producing new ones (Jensen 2004).

Figure 11 : Protein diversity explained, from Genome to Proteome. Protein diversity obtained from a single gene thanks to alternative splicing and post- translational events. The genome is composed of around 30000 genes and could generate around 1.8 million different proteins thanks to these processes (Jensen 2004).

1) Diversity of post-translational modifications

There is are wide range of post-translational modifications available to the cell to diversify its proteome. These modifications can be classified into two main groups. First, a chemical group, such as phosphate groups, or amino acid chains can be added on a target protein. This study will focus on the latter. On the contrary, the second group removes parts of a protein through hydrolase activity that cuts peptidic bonds found on proteins to be secreted by the cell or addressed to the membrane

These types of modifications can be found in prokaryotes but are mainly present in eukaryotes. On a biochemical level, there are five types of covalent bonds typically found which are oxidation, glycosylation, alkylation, acylation and the most common one: phosphorylation (Walsh et al. 2005) (Figure 12).

Figure 12 : Five major types of covalent modification. Oxidation, glycosylation, alkylation, acylation and phosphorylation are the main covalent modifications (Walsh et al. 2005).

These modifications can be very diverse and can also target 15 of the 21 amino acids typically found in eukaryotes. Knowing that around 5% of the human genome codes for proteins involved in those modifications, it is not surprising to see hundreds of described covalent

modifications. Targeted proteins can gain new functions, localization or general regulation depending on the nature of the modification (Walsh et al. 2005) (Table 1).

Table 1 : Some examples of post-translational modifications. (Walsh et al. 2005).

However, another layer of complexity is added since proteins can be modified multiple times at different sites with different modifications, each of them having a possible impact on future or existing modifications. Moreover, some modifications are targeting the same residue on the same protein leading to post-translational modification competition. In addition, these modifications can also, for the most part, be removed by specialized enzymes. This multitude of modification with infinite combination and reversibility is a major key to understand cellular life as they affect protein function. Therefore any conditions impairing this key ecosystem could lead to diseases. PML being a heavily post-translationally modified protein, it is the perfect example to look at some of these modifications in more details.

2) Phosphorylation

Phosphorylation is the most predominant type of modification in eukaryotes for transducing signals. Phosphates have unique properties; they are chemically versatile, are able to form mono, di or tri-esters and are very abundant on earth. Moreover, the unique size of its ionic shell and charge characteristics allows for specific and inducible protein-protein interactions (Hunter 2012). In mammals, five residues are principally targeted for phosphorylation by specific enzymes called kinases: Serine (Ser, S), Threonine (Thr, T), Tyrosine (Tyr, Y), Histidine (His, H) and Aspartic acid (Asp, D) (Walsh et al. 2005) (Figure 13).

Figure 13 : Five residues phosphorylated in mammals. (Walsh et al. 2005)

Phosphorylation is a major regulatory mechanism for proteins and PML does not escape that observation since PML protein abundance as well as the number and size of PML NBs are depending on its phosphorylation. PML is a very heavily modified protein with around eighty potential sites for phosphorylation with forty experimentally confirmed sites (Phosphonet 2016) (Figure 14).

Figure 14 : Known PML phosphorylation sites. Data from PhosphoSitePlus website (www.phosphosite.org/); red residues correspond to data from publications and black residues from expected sites. Zf-B_box corresponds to B-box zinc finger domain and DUF3583 stands for Domain of Unknown Function (PhosphoSitePlus 2016).

Cells use phosphorylation of PML to respond to various stimuli. PML protein possess an N-terminal region stretch enriched in prolines residues (36% of prolines between amino acids 3 and 46) usually exposed at the protein surface which participate in protein-protein interactions like signal transduction and post-translational modification (Kay et al. 2000). Within this region, a lot of residues were identified as phosphorylated in response to Epidermal Growth Factor (EGF) treatment at S8, S36, S38, S40 and T42 (Olsen et al. 2006). Extracellular signal Regulated Kinase (ERK1/2) -which is linked to EGF signaling- is probably phosphorylating site that also promotes SUMOylation at T28, S36, S38 and S40 (Hayakawa & Privalsky 2004). On the other hand, S8, S36 and S38 are phosphorylated by Homeodomain- Interacting Protein Kinase 2 (HIPK2) following DNA damage leading to the accumulation of PML protein and its SUMOylation. HIPK2 activity on PML is also required for effective pro- apoptotic activity of PML after DNA damage (Cheng & Kao 2012; Gresko et al. 2009).

In response to DNA damage like double strand breaks (DSBs), the number of PML NBs increases and multiple sites on PML protein are phosphorylated. However, this can be inhibited by caffeine or wortmannin, which are both Serine-protein Kinase ATM inhibitors. A study suggests that ATM is regulating PML NBs by phosphorylating PML directly or some of it components. (Dellaire et al. 2006). In response to gamma irradiation, DNA damage Check point Kinase 2 (CHK2) phosphorylates S117, which suggests a linked to PML-mediated apoptosis after DNA damage (Yang et al. 2002).

Also, PML nuclear localization requires phosphorylation at an unknown site by Ataxia Telangiectasia and Rad-3-related Kinase (ATR) (Bernardi et al. 2004). Several groups reported that the Nuclear Localization Signal (NLS) region was phosphorylated as well as the SUMO Interacting Motif (SIM) under different stimuli (Cheng & Kao 2012). The Peptidyl-prolyl cis- trans Isomerase NIMA-interacting 1 (PIN-1) mediated degradation of PML is also observed following phosphorylation at S403 and S505 by ERK2 (Lim et al. 2011). Similarly, a study showed that Cyclin-Dependent Kinase 1/2 (CDK1/2) phosphorylation of PML at S518, followed by an isomerization event mediated by PIN-1, in a prostate cancer model, was triggering PML degradation under hypoxia conditions through Cullin3-KLHL20 ubiquitin ligase as part of Hypoxia-Inducible Factor 1-alpha (HIF1α) tumor hypoxia response (Yuan et al. 2011). Under arsenic stress, PML is phosphorylated by ERK1/2 at S527 and S530 leading to PML mediated apoptosis (Hayakawa & Privalsky 2004). PML degradation is also promoted by its phosphorylation by Casein Kinase 2 at S565. This phosphorylation site (560-566,

SSSEDSE) is adjacent to a SIM (VVVI, 556-559) and is influencing interactions with SUMO (Scaglioni et al. 2006; Stehmeier & Muller 2009).

PML is also a key regulator of the cell cycle since PML overexpression in Hela cells leads to cell cycle arrest in G1/S phase whereas cell cycle progression is promoted by loss of PML (Mu et al. 1997; Wang, Delva, et al. 1998). PML phosphorylation appears to be subjected to cell cycle regulation as it is directly interacting with Aurora Kinase A (AURKA) during M phase and during G1 phase of the cell cycle. However, it is unknown if PML phosphorylation plays a direct role in cell cycle control (Cheng & Kao 2012). PML is also phosphorylated at S403 and T409 by Mitogen-Activated Kinase (MAPK) BMK1/ERK5 hereby inhibiting the activation of Cyclin-Dependent Kinase Inhibitor 1 (CDKN1A also known as p21), a key modulator of cell proliferation, through PML (Yang et al. 2010).

Phosphorylation of regions next to NLS, SIM and ubiquitination sites are very important to PML regulation and function, involving many different kinases. However, very little is known about the required stimuli leading to PML phosphorylation and coordination of other post-translational modifications inducing new specific functions (Figure 15).

Figure 15 : Known site-specific kinases for PML phosphorylation and associated function. Diagram showing known phosphorylated residues by corresponding kinases. Arrows indicate the targeted sites and functional consequences are annotated in red next to the corresponding kinase (Guan & Kao 2015).

There is also a great quantity of proteins that are phosphorylated in PML NBs. For example, Cellular tumor antigen p53 (or p53) a famous tumor suppressor protein involved in cell cycle regulation, is phosphorylated in PML NBs by HIPK2 following UV radiation which in turn will promote p53 acetylation by CREB-binding protein (CBP) (Hofmann et al. 2002; D’Orazi et al. 2002). These modifications enhance its pro-apoptotic and transactivation activities, as well as its ability to arrest the cell cycle. Moreover, P53 is enriched in PML NBs after DNA damage where it is stabilized by phosphorylation by CHK2, whose autophosphorylation is enhanced in PML NBs (Cheng & Kao 2012; Louria-Hayon et al. 2003; S. Yang et al. 2006).

3) Acetylation

Acetylation is a type of acylation characterized by the transfer of an acetyl group onto a substrate, mainly on Lysine (K) residues and is involved, for example, in gene expression

regulation through histone modification (Walsh et al. 2005). PML is also acetylated by Histone acetyltransferase p300 at K487 and K515 promoting its SUMOylation and leading to apoptosis upon Trichostatin A (TSA) treatment, an inhibitor of Histone deacetylases (Hayakawa et al. 2008). PML protein abundance is increased under NAD-dependent protein deacetylase Sirtuin- 1 (SIRT1) overexpression whereas its loss causes a decrease in PML protein accumulation. Even though SIRT1 is deacetylating PML, this effect does not appear to be linked to its deacetylating properties (Campagna et al. 2011; Miki et al. 2012). PML modification at K487 is essential for its nuclear localization, as shown through K487R mutants, but the role of acetylation at this site is not well understood yet. On the other hand, K515 acetylation site does not seem to have any effects on PML or PML NBs (Duprez et al. 1999). It is also interesting to note that some post-translational modifications, such as SUMO, can also be acetylated and may play an inhibitory role on PML NB assembly by preventing interactions with its partners such as DAXX, through their SUMO-SIM interphase thus introducing the idea of modified modifiers (Cheng & Kao 2012; Ullmann et al. 2012).

As described, phosphorylation sites for PML are numerous and important, both for its stability and function. Acetylation is important as well but to a minor extend. These modifications are often linked to another very important modification of PML that is SUMOylation.

4) SUMOylation

a) The SUMO protein

Human Small Ubiquitin-like Modifiers (SUMOs) are ~10kDa proteins that have a three- dimensional structure close to Ubiquitin, a protein used in one of the most studied protein modification system that will be described later (Figure 16). Studies found different isoforms for SUMO: SUMO1 (also known as Smt3c, PIC1, GMP1, Sentrin or Ubl1), SUMO2 (also known as Smt3a or Sentrin3), SUMO3 (also known as Smt3b or Sentrin2) and SUMO4. Ubiquitin and SUMO only share around 20% sequence identity at the protein level and display different charge distribution. All SUMO isoforms have an unstructured stretch of 10-25 amino acids at their N-terminus that is not found in any other ubiquitin related protein. It is probably used for the formation of SUMO chains (Tatham et al. 2001) (Figure 16). SUMO proteins are expressed in all eukaryotes and, in vertebrates, all SUMO isoforms are expressed in all tissues except for SUMO4 who is mainly expressed in the kidney, lymph node and spleen (Guo et al.

2004). However, it is not clear whether SUMO4 is present at the active protein level in-vivo especially since it can not be conjugated (Sinha et al. 2016; Geiss-Friedlander & Melchior 2007; Owerbach et al. 2005). Very recently, this protein family got a little bit bigger with the discovery of SUMO5, an isoform highly homologous to SUMO1, essential for PML NB formation and stability through its conjugation on PML at K160 (Liang et al. 2016).

Figure 16 : Ubiquitin and SUMO three dimensional structure comparison. Structures obtained from crystallography (ubiquitin) and Nuclear Magnetic Resonance Spectroscopy (SUMO). Both structures share the tightly packed secondary structure made of α-helices and β-sheets along with the di-glycine motif at the C-terminus. A long flexible chain at the N-terminus is only observed in SUMO (Dohmen 2004).

All SUMO proteins need to be maturated: the immature forms carry a C-terminal stretch of amino acids (2-11) after an invariant di-glycine (Gly-Gly) motif marking the C-terminus of the mature protein. The mature form of SUMO2 is 95% identical in structure to SUMO1, but shares only 50% sequence identity with it while SUMO2 and 3 only differ from one another by three amino acids on the N-terminus. Given this similarity, they cannot be distinguished by using antibodies so they are usually referred to as SUMO2/3. Despite some functional

redundancy, SUMO1 and SUMO2/3 also display distinct functions as they are conjugated to different proteins in-vivo (Saitoh & Hinchey 2000; Vertegaal et al. 2006; Hay 2005). SUMOylation is an essential process in most eukaryotic organisms like S. cerevisiae, C. elegans, Arabidopsis thaliana and mice but not for fission yeast (Geiss-Friedlander & Melchior 2007). Absence of SUMOylation in mice causes embryonic lethality while SUMO1 haploinsufficiency or SUMO2 deficiency induces development defect in mice (Geiss-Friedlander & Melchior 2007; Alkuraya et al. 2006; L. Wang et al. 2014). Moreover, it has been shown during the recent years that SUMO was playing an important role in many key cellular processes, such as cardiac development and function (Lee et al. 2015), but also notably in the cellular stress response (Enserink 2015).

b) Enzymatic cascade involved in SUMOylation

SUMOylation results in the formation of an isopeptide bond linking the C-terminal glycine (Gly, G) residue of the modifier protein and the lysine (Lys, K) residue of the acceptor protein. This process involves an enzymatic cascade comprised of three classes of enzymes that are very well conserved and unique to this pathway (Geiss-Friedlander & Melchior 2007). In the first step, SUMO precursor protein is processed by cysteine-specific SUMO proteases, ULPs in yeast but called SENPs (SENtrin specific Proteases) in mammals. This exposes the di-glycine motif mentioned earlier which will then be linked to the unique E1 activating heterodimer enzyme AOS1–UBA2, also known as SAE1-SAE2 dimer. The E1 will catalyze the covalent attachment of SUMO to a reactive cysteine (Cys, C) residue in SAE2 through an ATP- dependent thioesterification reaction. Through a thioester linkage, SUMO is then transferred to the cysteine residue of the unique SUMO E2 conjugating enzyme: UBC9. In vitro, the E2 enzyme is sufficient for conjugating SUMO to a lysine residue on the substrate. However, it is likely receiving help from E3 SUMO ligases in vivo. E3 ligases can serve as scaffold proteins that will bring the SUMO-charged UBC9 and the substrate in close proximity providing in the process the efficiency and specificity to the SUMOylation reaction or just stimulating the E2 enzyme (Enserink 2015; Geiss-Friedlander & Melchior 2007) (Figure 17) . There are relatively few E3 ligases identified, nine to date, the best-known class being the Protein Inhibitor of Activated STAT protein (PIAS) family. For example, PIASy was reported to interact with p53 protein and to be involved in the regulation of cellular senescence and apoptosis (Bischof et al. 2006; Nelson et al. 2001). RANBP2 is also a very well-known E3 ligase targeting SP100 and involved in nuclear import (Pichler et al. 2002). E3 ligases are very interesting to study since

they provide substrate specificity to the pathway making them good candidates for drug development.

An important aspect of SUMOylation is that it is a very dynamic and reversible process. SUMOylated proteins can be deSUMOylated through SENPs activity, the same enzymes used for SUMO maturation (Figure 17). These enzymes have an important functional role in the turnover and spatial regulation of SUMO (Mukhopadhyay & Dasso 2007) and are essential for many cellular processes like chromosome cohesion, mitosis or transcription (Enserink 2015).

Figure 17 : Mechanism of reversible SUMOylation. Diagram representing the activation (E1: SAE1-SAE2), conjugation (E2: UBC9) and ligation steps leading to substrate SUMOylation. SENPs activity allows for SUMO maturation and deSUMOylation of targeted proteins. It is also important to note than the E1 and E2 are unique unlike E3 ligases.

c) Consensus motifs for SUMOylation

SUMOylation of substrates usually occurs on lysine residues in a canonical SUMO consensus motif ΨKx(D/E), in which Ψ is a large hydrophobic residue and x any amino acid followed by an acidic residue (Rodriguez et al. 2001). The hydrophobic and acidic residues promote stability of the interaction between the substrate and the E2 conjugating enzyme (Lin et al. 2002; Enserink 2015). Different variations of this motif have been identified including Negatively

charged amino acid-Dependent SUMO Motifs (NDSMs) and Phosphorylation Dependent SUMO Motif (PDSMs). This last motif is an extended canonical motif in which phosphorylation, by proline-directed kinases, increase SUMOylation efficiency. PDSM and NDSM probably increase SUMOylation efficiency by increasing the stability of the interaction between UBC9 and the substrate because of the negatively charged amino acid (NDSM) or phosphate (PDSM) which will interact with charged amino acids of UBC9 (S.-H. Yang et al. 2006; Yang & Grégoire 2006). It is also important to note that although there is a canonical consensus motif, non-consensus SUMOylation sites are a relatively common event (Enserink 2015).

d) Chain formation and SIM

SUMO is capable to form polymeric chains (only with SUMO isoforms 2 and 3) through its K11 site found in a canonical SUMO consensus motif. However SUMO1 does not contain K11 and is conjugated to its substrate once or at the end of a poly-SUMO chain (Hay 2005; Tatham et al. 2001). These chains were mostly studied and characterized because of their role as an indirect degradation signal. SUMO chains can recruit conserved enzymes known as SUMO Targeted Ubiquitin ligases (STUbls). Theses E3 ubiquitin ligases can then ubiquitinate polySUMOylated substrates leading them to proteasomal degradation as it is the case for PML under arsenic trioxide induced stress (Enserink 2015; Lallemand-Breitenbach et al. 2008).

As shown, SUMO can be used as signal relay for protein degradation but most importantly, it is creating a new interface for protein-protein interactions through a particular motif called the SUMO Interacting Motif (SIM). NMR spectroscopic characterization of the interaction of SUMO and peptides derived from known substrates identified a hydrophobic core with the following consensus to identify SIM: [V/I]x[V/I] [V/I] (Song et al. 2004). Following studies confirmed the essential role of this hydrophobic core and showed that a hydrophobic pocket on SUMO was interacting with hydrophobic side chains of the SIM (Song et al. 2005). This site is also often flanked by acidic amino acid or in some cases phosphorylated residues that will interact with the lysine residue of the SUMO protein to stabilize the interaction (Stehmeier & Muller 2009). This also gives a way for the cell to control protein SUMOylation spatially and temporally as phosphorylated SIM might add specificity for appropriate substrates (Enserink 2015). SUMO can also serve as glue in a complex, stabilizing it through a cooperation of multiple weak SUMO-SIM interactions that would significantly increasing stability (Enserink 2015; Psakhye & Jentsch 2012).

SUMO plays an important role under normal conditions by maintaining cell homeostasis, promoting cell growth and proliferation whereas under stress conditions such as heat shock, DNA damage, oxidative stress but also viral infections, it is a key component in the cellular response by activating pro-survival pathways (Saitoh & Hinchey 2000; Enserink 2015).

e) PML SUMOylation and SIMs

Ran GTPase-activating protein 1 (RanGAP1) was the first SUMO target identified (Matunis et al. 1996) and was closely followed by another heavily SUMOylated protein that is PML. PML possess one SUMO-Interacting Motif (SIM) at the C-terminus of its sequence (Shen et al. 2006) and is also heavily SUMOylated by SUMO1 and SUMO2/3 (Cheng & Kao 2012). Three main SUMOylation sites on PML were identified at K65, K160 and K490 (Kamitani et al. 1998) (Table 2). Three more sites were identified for poly-SUMOylation under arsenic trioxide treatment at K380, K400 and K497 (Galisson et al. 2011). Two more potential polySUMO sites were discovered through quantitated proteomics at K226 and K616 but further confirmation is needed (Vertegaal et al. 2006) (Table2). Table 2 : Known PML SUMOylation sites

Known and verified SUMOylation sites using Consensus Motif (CM); Inverted Consensus Motif (ICM); Hydrophobic Cluster SUMO Motif (HCSM) or Negatively charged amino-acid Dependent SUMO Motif (NDSM). Targeted lysine are shown in red (Nisole et al. 2013).

Fluorescence microscopy data show that PML NBs and SUMO1 co-localize and that SUMOylation is essential to the maintenance of their structure and function (Müller, Matunis, et al. 1998; Gao, Cheng, et al. 2008). PML dimerization is a prerequisite for de novo assembly of PML NBs along with PML SUMOylation for recruitment of components such as SP100 or DAXX but also for the turnover and retention of PML in NBs and their integrity (Weidtkamp- Peters et al. 2008; Shen et al. 2006; Lallemand-Breitenbach et al. 2010). The importance of SUMO1 conjugation to PML for the maintenance of the integrity of PML NBs was also shown in Mouse Embryonic Fibroblasts (MEFs). In those cells, SUMO1 was knocked out and a fewer

number of PML NBs was observed as well as a decrease in the amount of SUMO2/3 conjugated to PML compared to wild type cells (Evdokimov et al. 2008). SUMO3 conjugation also appears to regulate nuclear localization and formation of PML NBs when conjugated to K160 (Fu et al. 2005). Also, the SUMOylation of PML is suggested to play a role on the localization of other components of the NBs probably through SUMO-SIM interaction (Cheng & Kao 2012).

f) Regulation of PML through the SUMO pathway

PML SUMOylation seems to be dependent on the cell cycle: its SUMOylation is elevated during interphase and declines during mitosis (Everett et al. 1999). SUMOylation of PML can also be triggered through DNA damage induced chemically, for example with Adriamycin, a chemotherapeutic agent (Gresko et al. 2009). PML NBs regulate transcription through sequestration or dissociation of transcription factors. Therefore, PML SUMOylation might have both direct and indirect effects on transcriptional regulation. One example of such indirect regulation is the release from PML NBs of Signal Transducer and Activator of Transcription 3 (STAT3), due to PML deSUMOylation by SENP1 previously activated through Interleukin-6 treatment (Kawasaki et al. 2003; Ohbayashi et al. 2008). PML SUMOylation also plays a role in apoptosis regulation through PML NBs interaction with DAXX or P53 proteins (Meinecke et al. 2007; Cheng & Kao 2012).

SUMOylation of PML is controlled in part by the E3 SUMO ligases targeting it. The first identified PML E3 SUMO ligase was RAN Binding Protein 2 (RanBP2) which mediated the K490 SUMOylation, essential for PML NBs maintenance (Tatham et al. 2005; Satow et al. 2012; Saitoh et al. 2006). Another E3 ligase recently discovered for PML is Protein Inhibitor of Activated STAT1 (PIAS1) that would be responsible for the SUMOylation of K65 and K160 residues. These two sites also appear to influence Casein Kinase 2 activity on PML leading to S565 phosphorylation and subsequent degradation of PML (Rabellino et al. 2012). Histone Deacetylate 7 (HDAC7) was proposed as an E3 ligase for PML since its presence is required to keep PML SUMOylated, but there is no evidence so far that it is acting directly as a SUMO E3 ligase (Gao, Ho, et al. 2008). Recently a new SUMO2/3 E3 ligase for PML was discovered, ZNF451-1, modifying PML at established SUMOylation sites (K65/K160/K490) and involved in its stability (Koidl et al. 2016). A more indirect approach can also alter PML SUMOylation status: for example Beta-catenin was shown to prevent RanBP2 from interacting with PML thus preventing its SUMOylation (Satow et al. 2012). Finally another way to prevent PML from

interacting with its E3 ligases would be by restricting the localization of the substrate or the enzymes involved (Cheng & Kao 2012).

Another way to regulate the SUMOylation of PML is through the involvement of specific deSUMOylases (SENPs). In Humans, there are six SENPs, SENP-1, -2,-3,-5 and -6. The nuclear SENP1 was shown to specifically remove SUMO1 from PML (Gong et al. 2000). SENP2 isoform (SuPr-1) has also been involved on PML deSUMOylation, causing c-Jun, transactivation (Best et al. 2002). PML PolySUMO chains formed by SUMO2/3 are removed by SENP3 under mild oxidative stress causing PML NBs disruption and stimulating cell proliferation (Han et al. 2010). Just as SENP3, SENP5 deconjugates SUMO2/3 at the SUMOylation sites K160 and K490 as well as all SUMO isoforms present on K65 (Gong & Yeh 2006). SENP6 was reported to specifically remove SUMO2/3 and a loss of this SENP causes an increase of cell death as well as an increase in PML NBs (Hattersley et al. 2011; Mukhopadhyay et al. 2006). In line with MEFs experiment, removing SENP increases PML NBs, whereas knocking out SUMO1 in MEFs causes a decrease. It is rather obvious that SENPs, just as E3 ligases, are involved in PML SUMOylation dynamics however, it is still poorly understood under which conditions each component play a role and how they are coordinated (Cheng & Kao 2012) (Figure 18). Besides SENPs, other deSUMOylases exist such as Ubiquitin-specific peptidase-like protein 1 (USPL1) (Schulz et al. 2012).

Figure 18 : Known SUMOylation sites and E3 ligases of PML. Diagram showing residues being modified. SUMO E3 ligases target is indicated with an arrow and functional consequences are indicated in red. SUMOylation resulting from arsenic trioxide stress are indicated in green. Both Mono-SUMOylation by SUMO1 and PolySUMOylation are observed at K65 while poly-SUMO chains were described at K160, K380, K400, K490 and K497. Some reports also indicate polySUMOylation at K226 and K616 but this needs to be verified (Guan & Kao 2015).

5) The SUMO/Ubiquitin coupled pathway

SUMOylation of PML is a key factor that controls PML stability in response to extracellular or intracellular stimuli. Arsenic Trioxide (AS2O3 or ATO) is a well-known drug currently used as a therapeutic agent for Acute Promyelocytic Leukemia (APL) treatment. It induces an increase of PML SUMOylation about an hour after treatment, which later leads to its degradation by the proteasome. PML-RARα is also degraded in the same way (Müller, Matunis, et al. 1998) (Figure 19).

Figure 19 : PML is hyperSUMOylated and degraded under arsenic stress. Western blot from Acute Promyelocytic Leukemia (APL) cells after treatment with 1µM Arsenic Trioxide (AS2O3) for indicated times.

Both All Trans-Retinoic Acid (ATRA) and Arsenic Trioxide (ATO) are used to treat APL as they both target Leukemic Blasts. ATRA will lead to PML-RARα, and RARα degradation and patient remission by primarily forcing differentiation by activating transcription. However, this treatment is unable to efficiently destroy Leukemia Initiating Cells (LICs) ultimately leading to disease relapse (Figure 20, right panel). On the other hand, ATO leads to PML-RARα and PML degradation and ultimately to patient remission by inducing both differentiation and apoptosis of leukemic blasts and eradication of LICs thus preventing the disease relapse (Figure 20, left panel). The vast majority of patients are now cured with a combination of both ATRA and ATO (Dos Santos et al. 2013). A study also showed that trivalent antimonials could be used instead of ATO to maybe reduce side effects such as mitochondrial toxicity or ROS (Müller, Miller, et al. 1998).

Figure 20 : Arsenic Trioxide (ATO) and All-Trans Retinoic Acid (ATRA) effects in the cure of APL. ATRA forces differentiation but does not destroy Leukemia Initiating Cell (LICs) causing disease relapse. ATO causes apoptosis and differentiation and is able to efficiently destroy LICs. A combination of both treatments (dashed arrow) is currently used and cures most cases as they both work synergistically (Dos Santos et al. 2013).

Arsenic trioxide (As2O3) directly binds to cysteine-rich zinc finger in the RBCC domains of PML (Figure 21, top part). Arsenic trioxide binding causes a conformational change in PML that will in turn promote the interaction between PML and UBC9, the unique SUMO2 E2 conjugation enzyme (Zhang et al. 2010). This SUMOylation event can however be inhibited by calyculin, a serine/threonine phosphatase inhibitor, suggesting that some dephosphorylation events are needed prior to PML SUMOylation, either on PML or on one of its interacting partners (Müller, Miller, et al. 1998). HyperSUMOylated PML is targeted for ubiquitination by the E3 ubiquitin ligase RNF4 (also known as SNURF) to be degraded by the proteasome. RNF4 protein contains multiple SIMs in its N-terminus and a C-terminal RING- type E3 ligase domain. These SIMs interact with SUMO2 chains of PML and allow RNF4 to ubiquitinate PML and its chains leading to its degradation (Lallemand-Breitenbach et al. 2008; Tatham et al. 2008) (Figure, 21 bottom part). SUMOylated PML also primes PML phosphorylation through Casein Kinase 2 (CK2), which contribute as well to the ubiquitination

of PML and leads to its degradation as shown in APL cells, in Non-Small Cell Lung carcinoma cells (NSCL) as well as in human primary tumor specimens (Rabellino et al. 2012).

Figure 21 : PML degradation key-step events under Arsenic trioxide (AS2O3) induced stress. Diagram representing Arsenic Trioxide induced SUMOylation of PML and subsequent recognition of SUMO2/3 chain by RNF4 SIMs leading to Ubiquitination and degradation of PML by the proteasome (Nisole et al. 2013).

PML degradation under Arsenic Trioxide induced stress allowed the discovery of a coupled SUMO-Ubiquitin pathway in which RNF4 interacts with PML through SUMO chains leading to its ubiquitination and proteasomal degradation. Thus, ubiquitination also plays an important role on PML stability.

6) Ubiquitination

Ubiquitination is a multifaceted post-translational modification that is highly dynamic and involved in all aspects of the cell biology. Ubiquitin is a 76 amino acid protein that can be modified and is involved in many types of signal transduction leading to various cellular outcomes. The most common and remarkable one is the targeting for proteasome dependent degradation but Ubiquitination is not only about degradation. For example, P53 function can be regulated through its ubiquitination by the ubiquitin ligase E4F1 leading to cell cycle arrest specific transcriptional program (Le Cam et al. 2006).

a) Enzymatic cascade

The post-genomic era provided insight into the complexity of the ubiquitin system with more than 1000 proteins regulating ubiquitination in Humans. Ubiquitin is attached to substrates by

a sophisticated three-step enzymatic cascade very similar to SUMOylation. However, the two pathways are completely independent and do not share any enzymes. First, there is the priming of ubiquitination through an ATP-dependent covalent attachment of the Ubiquitin protein to one of the two ubiquitin activating enzymes E1. Spontaneous transfer of ubiquitin to an ubiquitin conjugation enzyme E2 is facilitated by the E1-Ubiquitin intermediate. At this point, two different paths could be taken, both resulting in substrate ubiquitination. In the first option, Ubiquitin (Ub) is directly transferred to the substrate through an E3 ubiquitin ligase containing a RING domain capable of directing the E2-Ub species to its substrate. In the second option, Ubiquitin is passed on to a HECT (Homologous to the E6-AP Carboxyl Terminus) domain of an E3 ubiquitin ligase to which the E2-Ub enzymes would have been associated. Ubiquitin is then directly transferred from the E3 ligase to the substrate through a thioester bond (Lee & Diehl 2013). Ubiquitinated proteins can be recognized by Ubiquitin Binding Domains (UBDs) on receptors for example (Husnjak & Dikic 2012). Like SUMOylation, this process is reversible thanks to a specialized family of enzymes called DeUbiquitinases (DUBs) (Komander et al. 2009) (Figure 22).

Figure 22 : Ubiquitination enzymatic cascade leading to substrate degradation. Diagram representing the consecutive activities of the three enzymes types involved in Ubiquitination: activation (E1s), conjugation (E2s) and ligation (E3s). This process is reversible thanks to deubiquitinases (DUBs). Many enzymes are involved in this process; approximate number are shown next to the type of enzyme. Note that this diagram takes degradation as an example, not all ubiquitination event lead to degradation (Skaar et al. 2014).

b) The ubiquitin code

Comprehensive proteomics studies revealed that most proteins will experience ubiquitination at some point in their cellular lifetime (Swatek & Komander 2016). Ubiquitination starts by the attachment of a single ubiquitin protein to a lysine residue on a substrate. These mono- ubiquitination events have many roles in the cell such as signal transduction (Komander & Rape 2012). However, the key feature of ubiquitin is that it has seven lysine residues that can also be ubiquitinated to form chains of different types (Figure 23).

Figure 23 : Structure of Ubiquitin. Ubiquitin structure showing the seven lysine residues. Blue spheres show amino acids involved in chain formation (Komander & Rape 2012).

The diversity in the way chains are built gave rise to a complex “ubiquitin code” which got even more complex over the years with the discovery that these modifications by the ubiquitin protein could also be modified by SUMO, NEDD (another ubiquitin like modifier) or even phosphorylated or acetylated. This allows cells to store and transfer signal information through the complex ubiquitin code (Swatek & Komander 2016) (Figure 24).

Figure 24 : The Ubiquitin code. A. Diagram of the different possible modifications of Ubiquitin by SUMO2/3, NEDD8 or by chemical modifications. B. Secondary messengers in cells include modified or un-modified single ubiquitin or chains (Swatek & Komander 2016).

Even though it is important to appreciate the wide range of signal transduction, which can be provided by this ubiquitin code, this study will focus on the degradation signaling since it is the most studied for PML.

c) Proteasomal degradation signal

New studies and insights into individual chain types, new rules for proteasome degradation and the rising of branched and mixed linkage chains could support the ubiquitination threshold

model for proteasomal degradation instead of the more common single Lys48-linked tetra ubiquitin. It appeared quite clear over the years that the main task of the ubiquitin system was the degradation of targeted proteins through the proteasome. This complex can degrade a wide range of substrates in various contexts that do not necessarily require poly-ubiquitination (Finley 2009). Latest data suggest that the most efficient signal seems to be multiple modifications with short (four) Lys48-linked chains, or branched structures with Lys11- or Lys- 48 linkages. This model proposes that a protein could be targeted for non-degradative signaling up to a certain point. The branching possibility enables any non-degradative chain types to become a degradation signal. In short, the amount of poly-ubiquitin is more important than the type of modification (Swatek & Komander 2016) (Figure 25). For example, the most abundant E2 enzyme (the UB2D family) seems to add many types of short chains on substrates at random sites and as a result cyclin B degradation is facilitated when modified in this way (Kirkpatrick et al. 2006; Swatek & Komander 2016). In contrast, specialized E3 ubiquitin ligase systems, such as SKP-Cullin-F-Box containing complex (SCF) E3 ligases, use Lys-48-specific E2 enzyme UBE2R1. This E2 enzyme assembles medium sized chains (3-6 ubiquitin molecules) on substrates, that are used as canonical degradation signals (Pierce et al. 2009). Once the signal is recognized by the 26S subunit of the proteasome, the substrate is unfolded in an ATP- dependent manner and allows the unfolded polypeptide chain into its catalytic lumen. Once there, a host of proteases degrades the protein into short peptides very rapidly (Baumeister et al. 1998).

Figure 25 : Proteasomal degradation based on the Ubiquitin threshold model. Ubiquitination can result in either cellular signaling or degradation. Multiple short chains or branched ubiquitin will be recognized as degradation signals but not others. Spheres correspond to Ubiquitin (Light Grey) or to Lys63 (bleu), Met1 (green), Lys29 (orange) and Lys48 (dark grey) linkage types (Swatek & Komander 2016).

It is also important to keep in mind that ubiquitination is a very dynamic process especially since many E3 ligases work in conjunction with DUBs. They can regulate how chains are assembled through chain editing functions hereby shaping the ubiquitination status of the substrate. The length and stability of the chains also define their recognition by DUBs and therefore the stability of the signal (Swatek & Komander 2016).

d) PML ubiquitination

We saw earlier that PML could be ubiquitinated in a SUMO dependent manner under Arsenic Trioxide induced stress by RNF4 at K380, K400 (or K401) and K476 (Tatham et al. 2008). In a similar way, the protein ARKADIA also ubiquitinates PML under Arsenic Trioxide induced stress (Erker et al. 2013).

Ubiquitin-like-containing PHD and RING finger domains protein 1 (UHRF1), an epigenetic regulator, was identified as a new ubiquitin E3 ligase for PML but the process by which it leads to its degradation is still unclear (Guan et al. 2013). Some viral proteins such as Herpes Simplex Virus (HSV-1) viral protein ICP0 are E3 ligases that target PML for

degradation through ubiquination and in a SUMOylation dependent manner (Gu et al. 2005; Boutell et al. 2003).

Mammalian homologues of Drosophila Seven in Absentia (SIAHs) are involved in the proteasomal degradation of several factors implicated in cell growth and tumorigenesis. SIAH1/2 are involved in PML degradation through interaction with its Coiled-Coil region (Fanelli et al. 2004). In Burkitt’s lymphoma, HPV E6-associated protein (E6AP) was shown to degrade PML (Wolyniec et al. 2012). The Herpes virus Associated Ubiquitin Specific Protease, USP7 was also described as regulating PML stability in nasopharyngeal carcinoma cells but this regulation did not seem to involve its catalytic activity (Sarkari et al. 2011; Sivachandran et al. 2008). In prostate cancer cells and under hypoxia, PML is degraded by the a Cullin3 based complex involving KLH20 mediated ubiquitination of PML (Yuan et al. 2011). More recently, RNF8 and RNF168 which are two proteins involved in the DNA damage response, were described as regulating PML and PML NBs in a SUMO dependent manner (Shire et al. 2016).

So far, there is only one known deubiquitinating enzyme for PML that is USP11. In Human glioma cells, USP11 is repressed by the Notch/Hey1 pathway leading to PML degradation and to more aggressive gliomas (H.-C. Wu et al. 2014).

On a more general note, as mentioned previously, PML is a very heavily modified protein under many different stimuli (Figure 26). However, ubiquitination, with its impact on PML stability, is rather important and it would be interesting to know whether Ubiquitin would take an active part in signal transduction in PML NBs other than degradation. As shown, there are many different types of E3 Ubiquitin ligases however, this study focused on a particular type of ligase called SKP-Cullin-F-Box containing Complex (SCF).

Figure 26 : Summary of human PML post-translational modifications. Pro, proline-rich region; R, RING domain; B1 and B2, B-box domains; CC, coiled-coil region; (p), phosphorylation sites; (s), SUMOylation sites; (u), Ubiquitination sites; (a), acetylation sites. Ex1-9 and grey lines show exon boundaries whereas dotted lines show common sequences for all PML isoforms. NCBI nomenclature is shown in red and “Jensen2001” corresponds to nomenclature proposed by Jensen et al. (2001). Modification sites are indicated by one letter code amino acid for each PML isoform (Cheng & Kao 2012).

III) The SKP-Cullin-F-Box containing Complex (SCF)

Substrate specific degradation is a key component of the Ubiquitin Proteasome System (UPS), it governs very diverse cellular processes such as cell cycle progression, apoptosis, transcription or cell proliferation. Therefore, it is not surprising that the human genome encodes for more than 300 proteins containing a RING domain, each targeting proteins more or less specifically. In order to add another layer of selectivity in this process, multi-unit complexes can be formed in which RING ubiquitin ligase associates with other components that will bring substrate specificity to the newly formed complex.

1) General structure of Cullin-RING Ligases complex

The best studied example of Ubiquitin ligase complex is Cullin-Ring Ligases (CRLs). These complexes are composed of a Cullin-family protein that serves as a scaffold by interacting directly with the RING domain of the enzyme, through a protein-protein interaction domain on the C-terminus of the protein. In mammals, there are eight different Cullins able to make complexes; some of these can be interchangeable leading to the formation of many different complexes (Petroski & Deshaies 2005; Lee & Diehl 2013). Almost all Cullin proteins including CUL-1, -2, -3, -4a, -4b, -7 and -9, bind the E2 ubiquitin conjugation enzyme (UBER1) through the small RING-box protein 1 (RBX1 also known as ROC1) except for CUL5 which is only recruiting RBX2 (Lydeard et al. 2013) (Figure 27). Cullin-repeat motifs situated at the N- terminus of the Cullin protein allow for a large number of adaptor proteins to interact causing the assembly of more than 200 CRL complexes using the eight Cullin scaffold proteins (Skaar et al. 2014). These adaptor proteins include S phase Kinase-associated Protein 1 (SKP1) for CRLs 1 and 7, Elongin B and C for CRLs 2 and 5, DNA Damage-Binding protein 1 (DDB1) for CRLs 4A and 4B and finally, CRLs 3 and 9 do not use adaptor proteins or are unknown (Figure 27). The purpose of adaptors is to recruit proteins that will be able to specifically recognize a substrate and efficiently recruit it to the complex for its ubiquitination. These substrates vary widely from one complex to the other, and depending on the adaptor protein recruiting it (Figure 27). The best prototype for these CRLs is the Cul1-containing complex also known as the SCF ligase (Lee & Diehl 2013).

Figure 27 : The Cullin RING Ligase (CRL) family ubiquitination complex. Diagram showing the variety of CRL complexes. Cullin proteins are shown in green, the RING- box proteins are shown in red and the E2 ubiquitin conjugating enzyme recruited by it in orange. Adaptor proteins linking the Cullin scaffold to the substrate binding protein are shown in bleu (these include SKP1, DDB1, Elongin B and C). Substrates (in yellow) binds to specific protein capable of recognizing them (in violet). These include the F-Box family, SOCS box, DCAF and BTB proteins. APC2 is the scaffold protein of the APC/C (anaphase promoting complex or cyclosome) and is related to CRLs in its architecture and function even though it comprises many more proteins (in blue) (Skaar et al. 2013).

2) The SCF complex and the F-Box proteins

In the SCF complex, the RING-type Zinc finger containing protein recruiting the E2 conjugating enzyme is RBX1. The scaffold protein is represented by CUL1 and SKP1 serves as the adaptor protein that will recruit an interchangeable F-Box protein which will confer the substrate specificity to the ligase (Figure 27, CRL1/SCF). The name, F-Box, comes from a conserved motif of around 40 amino acid that was identified in all proteins binding to SKP1. One of them called cyclin F (CCNF, also known as FBXO1), gave the name to the F-Box motif (Bai et al. 1996). There are 69 genes encoding for F-Box proteins identified in humans, and are classified into three different families based on their protein interaction domain (Annex 1, Mammalian F-Boxes). Ten proteins containing the WD40 repeat domain were identified and named FBXW. WD40 repeat domain are usually composed of 4 to 16 repeating units forming

a circularized beta-propeller. 21 other proteins containing Leucine-Rich Repeats (LRR) domains were called FBXL. LRR domains are composed of stretches of repeating 20 to 30 amino acids rich in Leucine (a hydrophobic amino acid) forming an α/β horseshoe fold. Finally, the most important class, with 38 members, are proteins that have, in addition to their F-Box domain used to interact with the adaptor protein, other more diverse domains to recognize their substrates and are called FBXO (Jin et al. 2004; Skaar et al. 2013) (Figure 28).

Figure 28 : Mammalian F-Box protein structural domains. Diagram showing known structural motifs for the 69 F-Box protein. The most common one include the F-Box motif (F), WD40 repeat motif (WD) and the Leucine-rich repeat motif (L). Other domains include transmembrane domain (T), F-Box-associated domain (FBA), between-ring domain (IBR), domain in carbohydrate binding proteins and sugar hydrolases (CASH), kelch repeat (K), calponin homology domain (CH), domain found in cupin metalloenzyme family (Jmjc), domain present in PSD-95, Dlg, and ZO-1 (PDZ), zinc-binding domain (Lim), HNH nuclease family (HNHc), novel eukaryotic zinc- binding domain (CHORD), and tetratrico peptide repeat (TPR), ApaG-like motif (ApaG); the apolipophorin-III-like fold (Apolipophorin), the ubiquitin-like fold (Ubl), Traf- domain like (TDL), placental RNase inhibitor like (RNI-like), regulator of chromatin condensation-1 fold (RCC1) (Jin et al. 2004).

Most of the well-studied F-Box proteins were discovered to have multiple substrates. SCF targets have broad biological functions in very diverse pathways making the substrate network of SCF rather complex, very broad and difficult to study (Skaar et al. 2014; Skaar et al. 2009; Randle & Laman 2015) (Figure 29 and Annex 2).

Figure 29 : Known F-Box substrates and biological implications. Examples of known F-Box substrates depending on their class (FBXL, FBXW or FBXO) with a description of biological substrate function (Skaar et al. 2009).

3) Regulation of the SCF complex

SCF plays a very important role as a regulator of key signaling pathways whose dysregulation causes major oncogenic consequences (Z. Wang et al. 2014). Therefore, it is important to understand how these complexes are regulated.

a) NEDDylation

CUL1 association with RING E3 ubiquitin ligases and therefore SCF activity is increased by a post-translational modification of CUL1 called NEDDylation. This modification consists in the addition of a small ubiquitin-like modifier called NEDD8/RUB1, on the C-terminus of CUL1. NEDD8 facilitates the recruitment of the ubiquitin E2 to the complex therefore enhancing ubiquitin chain elongation as well as helping in bringing the E2 closer to the substrate (Saha & Deshaies 2008).

Like other post-translational modifications such as SUMOylation or Ubiquitination, NEDD8 is transferred to its substrate following a three-step enzymatic cascade involving an E1 activating enzyme (NAE) and one of the two E2 conjugation enzymes (UBC12 or UBE2F) (Tanaka et al. 2012) (Figure 30). The E3 ligase for this system that is specific to CUL1 is the DCN1-like protein 1 (DCUN1D or DCN1). It enhances NEDDylation through the recruitment of the E2 (UBC12) to the Cul1-CAND1-RBX1 complex. The activity of this E3 is important, as described through knockout experiments, for nuclear localization and its impact on cellular proliferation (Kim et al. 2008; Lee & Diehl 2013). Moreover, just like SUMOylation and Ubiquitination, NEDDylation is a reversible process thanks to an octameric complex called COP9 signalosome (CSN) that contains the metalloenzyme CSN5 subunit that will catalyze the deNEDDylation of the substrate (Lyapina et al. 2001; Lee & Diehl 2013) (Figure 30). Recent studies suggest that NEDDylation is promoted by the ejection of the CSN complex from SCF by the substrate binding (Lee & Diehl 2013).

Figure 30 : Regulation of Cullin Ring Ligases (CRLs) through NEDDylation. Diagram showing structural variety of CRLs as well as NEDDylation processes regulating their formation. The number of substrate receptors for each CRL is written on the far left. NEDDylation system is composed of one activating E1 enzyme (NAE) and two E2 conjugating enzymes. NEDDylation E2 UBC12 operates with RBX1 based CRLs while UBE2F operates with RBX2 based CRL5. RBX1/2 and DCN1 are E3 NEDDylation ligases. DeNEDDylation is mediated through the CSN complex (asterisk indicates catalytic subunit). CAND1 is a CRL exchange factor. CSN stands for COP9 signalosome, N for NEDD8, S for Substrate, SR for Substrate Receptor and U for Ubiquitin (Lydeard et al. 2013).

b) Cullin-associated NEDD8-dissociated protein 1 (CAND1)

NEDDylation is not the only way SCF complexes are regulated. Another protein called the Cullin-associated NEDD8-dissociated protein 1 (CAND1) was proposed to play the role of substrate receptor exchange to regulate these complexes (Pierce et al. 2013; Wu et al. 2013) (Figure 30 and 31A). An empty Cullin protein can either interact with a new substrate receptor or be used in the exchange cycle. Basic CRLs, formed by the substrate receptor, RBX1 and Cullin can either be NEDDylated and stabilized by the NEDDylation machinery or enter CAND1-dependent exchange cycle. In the case of its NEDDylation, the complex can start interacting with its substrates leading to their ubiquitination and subsequent degradation. Eventually, substrate abundance will diminish and CRLs will associate with CSN to get deNEDDylated leading to the dissociation of the complex that can now re-enter the cycle (Figure 31A). In the CAND1 exchange pathway, substrate receptors are exchanged with either a pre-existing or newly synthesized one. This is achieved through the formation of an intermediate complex where CAND1 and a substrate receptor both bind Cullin. The resulting outcome is a new complex with a new receptor targeting new substrates (Figure 31A). SKP1 association with CUL1 is prevented by the interaction of CAND1 with CUL1 through its β- hairpin (Figure 31B). In addition, CAND1 dissociation is favored by NEDDylation of Cullin (Liu et al. 2002). Crystallographic studies showed that CAND1 adopts a conformation within the CAND1-CUL1-Roc1 complex that causes the SKP1 binding site on CUL1 to be hidden. In addition to this, HEAT repeats domain from CAND1, forming rod-like helical structures, seems to slightly mask the lysine residues of CUL1, which are sites of NEDDylation. This is showing a form of completion between NEDDylation and CAND1 interaction with CUL1 further increasing SCF regulation (Goldenberg et al. 2004). Finally, just as the assembly of SCF is tightly regulated, its disassembly also appears to be tightly regulated as well, through active processes. For example, ATPase Cdc48/p97 recruitment is required to remove F-Box proteins in yeast under stress (Yen et al. 2012).

Figure 31 : Regulation of Cullin Ring Ligases (CRLs) through CAND1 substrate receptor exchange. A. Diagram showing pathways leading CRL activation, deNEDDylation and substrate receptor exchange. The main steps of the different pathways are labelled 1 through 6. SR stands for Substrate Receptor (asterisk shows newly synthesized SR), S1 and S2 represent Substrates, N stands for NEDD8, U for Ubiquitin and CSN for COP9 signalosome. B. Superimposing structures of CUL1-RBX1-CAND1 and CUL1-RBX1-SKP1-SKP2F-Box reveal how the β- hairpin of CAND1 (in red) bounds to CUL1 (in green) and prevents SKP1 (in blue) from interacting with CUL1 (Lydeard et al. 2013).

c) F-Box binding

Another layer of regulation occurs through the F-Box protein association to the complex. The first mechanism uses F-Box dimerization domains, which are usually situated at the N-terminus of the F-Box, to interact with the entire complex. For example, FBX4 is phosphorylated by a cell cycle dependent GSK2-β kinase in this dimerization domain which is then used by 14-3- 3β as docking site to facilitate the dimerization (Barbash et al. 2011). Another F-Box, FBW7 shows similar behavior by dimerizing (Zhang & Koepp 2006). Another mechanism is the self- regulation of FBXWs. F-Box proteins that have WD40 repeats domains can bind Ubiquitin

directly therefore impacting their turnover by auto-ubiquitination (Lee & Diehl 2013). To summarize, SCF substrates are not only regulated by SCF activity but also by the SCF own regulatory system.

4) Substrate recognition

SCF substrates are usually recognized by F-Box proteins through a domain that enables them to recognize on their substrate, highly specific short peptide sequences that allows protein- protein interaction. This motif located in the substrate is called a degron. In most of the well- studied F-Box systems, the substrate interaction is dependent on a post-translational modification, mainly phosphorylation of the substrate’s degron. Even though some degrons do not require to be modified in order to be recognized by their F-Box, access to these sequences is regulated in most cases, by post-translational modifications (Skaar et al. 2013) (Figure 32). For example, p27 (also known as CDK1B) is phosphorylated on T187 by a Cyclin Dependent Kinase (CDK) which allows its specific recognition by SKP2 (also known as FBXL1) and Cofactor cyclin-dependent kinase subunit 1 (CKS1 or CKS1B) leading to its degradation (Spruck et al. 2001). This type of degron recognition using phosphorylation is known as the canonical phosphodegron (Figure 32). However, recent studies showed that F-Box proteins were able to recognize very different degrons with different modifications like glycosylation (Mello et al. 2002) or non-modified degrons which allow different alternative mechanisms for substrate recognition by SCFs (Skaar et al. 2013) (Figure 32). In some cases, phosphorylation of the substrate can even prevent the interaction with its F-Box and therefore preventing its degradation. It is the case with Chromatin licensing and DNA replication factor 2 (CDT2), degraded through SCFFBXO11 and p85β (also known as PIK3R2) degraded via SCFFBXL2 (Kuchay et al. 2013; Rossi et al. 2013; Skaar et al. 2014). In some other cases where a phosphorylation is needed to induce a conformational change on the substrate to reveal the degron, a priming phosphorylation might be required. In some of these cases, a first kinase would phosphorylate the substrate so that a second kinase can recognize it and in turn phosphorylate its degron. Finally, F-Box recognition of the substrate’s degron might require the implication of a cofactor or might be regulated by competition with another binding partner (Skaar et al. 2013) (Figure 32). It is important to note that different recognition mechanisms can be combined thus increasing specificity and regulation of SCF activities. An additional layer of regulation is added by the fact that F-Box proteins themselves are regulated by post- translational modifications such as phosphorylation and proteolytic turnover.

Figure 32 : Various mechanisms used to regulate F-Box protein substrate recognition. Diagram representing eight different ways F-Box proteins can interact with their substrate degron. Note that multiple mechanisms can be combined to increase specific regulation of the substrate degradation (Skaar et al. 2013).

Even though many different ways are available to recognize a degron and interact with the substrate, most of F-Box substrates identified so far present phosphodegrons (Table 3). In addition, the study of these interactions showed once more that SCF complexes are involved in many essential cellular pathways such as cell differentiation and proliferation but also cell signaling (Nakayama & Nakayama 2006) (Table 3). It is therefore not surprising that any dysregulation in this system might lead to cellular miss function and ultimately could be linked to diseases.

Table 3 : Example of F-Box substrate and known degron regulation. (Skaar et al. 2013)

5) SCFs and diseases

Since SCF complexes are able to recognize so many different targets involved in so many cellular regulatory pathways, they are also involved in oncogenic pathways. In diseases, there are many ways in which SCF complex activity can be altered, one of them would be through its F-Box. F-Box proteins can be overexpressed or mutated causing a deletion or point mutation thus modulating SCF activity. Moreover, a more indirect modulation can also occur through regulatory pathways leading to degron modification. This dysregulation is however not necessarily bound to cancer: microbial infections could also cause such an imbalance by degrading the F-Box or blocking its action. In some cases SCF can even get hijacked to target specific protein of the host cell (Skaar et al. 2013; Skaar et al. 2014) (Figure 33).

Figure 33 : Protein degradation can be dysregulated in diseases due to F-Box activity alterations. Diagram representing the different ways in which F-Box protein-mediated degradation can be altered in disease. These include overexpression or deletion of the F-Box protein, mutation of the F-Box or of its degron and dysregulation caused by microbial proteins. Ub stands for Ubiquitin (Skaar et al. 2013).

Targeting E3 ligases has been considered as an easy way to impact a specific step in protein regulation. One such example is the E3 Ubiquitin ligase MDM2 which is influencing the tumor suppressor protein P53 (or TP53) stability by ubiquitinating it resulting in its subsequent degradation. P53 blocks cell proliferation and induces apoptosis of tumor cells (Tchelebi et al. 2014). A new call of small molecules, based on cis-imidazoline backbone, already passed phase I clinical trials. These small molecules, called Nutlins, are competitive

inhibitors of MDM2 thus stabilizing P53 (Skaar et al. 2014). Targeting E3 ligases can be an effective and specific way to control turnover of proteins. However, it is also possible to create chemotherapy targeting multiple targets -like the formation of SCF complexes- through drugs such as MLN4924, which prevents Cullin NEDDylation. MLN4924 binds to NEDD8 to form NEDD8-MLN4924 adducts catalyzed by NAE1 activity. These adducts will block NEDDylation of Cullin and will ultimately cause the disassembly of CRLs complexes effectively inhibiting E3 ubiquitin ligase activity (Zhang et al. 2016). This compound managed to prevent growth of lung, ovarian, breast, myeloma, leukemia, melanomas and Ewing’s sarcoma cells in vitro and has proven its efficacy in rodent models (Skaar et al. 2014). But inhibition of NEDDylation can also affect other targets that are not CRL related, such as P53 whose nuclear localization is dependent on its modification by NEDD8 (Abidi & Xirodimas 2015). Therefore, in some cases, it seems more preferable to look for a more specific drug with a better therapeutic index. It means that a drug needs low amounts of a therapeutic agent to cause maximum therapeutic effects while causing the least amount of toxicity. Therefore, the use of specific inhibitors seems a better option, especially for SCF ligases involved in cancers, rather than broad drugs such as MLN4624. Various disorders such as sleep or mood disorder, inflammation and acquired infections might obtain better therapies by developing more studies focused on the function of the SCF family E3 ligases (Skaar et al. 2014). Currently, the main problem is that relatively little is known about the biological characterization of F-Box proteins. However, a few laboratories are starting to pay some attention to them as they are being described as tumor suppressors (like FBXW7) or oncogenes (like SKP2) based on experimental observations and analysis of their domains (Z. Wang et al. 2014) (Figure 34).

Figure 34 : Examples of F-Box proteins involvement in cancer. Illustration based on data obtained from mice models, biochemical substrates identified for each F-Box and pathological profiles. F-Box proteins were grouped in two categories: tumor suppressor (a) or oncogene (b). β-helix (BH); F-Box motif (F); F-Box and leucine-rich repeat protein (FBXL); F-box only (FBXO); F-Box/WD repeat-containing protein (FBXW); in between ring fingers domain (IBR); leucine-rich repeat (L); NOP14-like family domain (Nop14); periplasmic copper-binding protein (NosD); S-phase kinase-associated protein 2 (SKP2); transmembrane region (T); tetratricopeptide repeat (TPR); UvrD/REP helicase N- terminal domain (UH); UvrD-like helicase C-terminal domain (UvrD); WD40 repeat (W); putative zinc finger in N-recognin (Zu) (Z. Wang et al. 2014).

Taking this into consideration, less than 20 F-Box proteins have well-characterized substrates. The identification of multiple substrates for many F-Box proteins, which is key to understand basic biochemical principles for substrate targeting, still needs to be done. Some F- Box proteins have essential cell functions such as driving or stopping proliferation through degradation of targets with common biological functions. However, other F-Box proteins can target substrates with opposing or disparate biological functions such as FBXO9, involved both in adipocyte differentiation through the degradation of Peroxisome Proliferator-Activated Receptor gamma (PPARγ) and in cell survival through the mTOR pathway (Fernández-Sáiz et al. 2013; Lee et al. 2013; Skaar et al. 2014; Lee et al. 2016). Understanding the role of each F- Box protein is important not just to get a deeper insight into the regulation of normal and oncogenic cellular events, but also to determine the best targeting strategies for other pathological processes (Frescas & Pagano 2008; Welcker & Clurman 2008; Z. Wang et al. 2014). Still, some topics on F-Box protein biology and biochemistry have been deeply

investigated and their link to human diseases has been clarify, allowing the development of new concepts and targeting therapies (Skaar et al. 2014) (Table 4).

Table 4 : Drug therapy development targeting E3 Ubiquitin Ligases. (Skaar et al. 2014)

6) PML and SCF

Regarding PML, there is only one case reported of a CRL interacting with PML. KLHL20- CUL3-ROC1 complex was reported to ubiquitinate PML under hypoxia conditions in prostate cancer cells model. In these cells, under hypoxic stress, Hypoxia-Inducible Factor 1 (HIF-1α) causes an increase in KLHL20 protein production leading to an increased amount of KLHL20- CUL3-ROC1 complex formation. Meanwhile, CDK1/2 phosphorylates PML at S518-519 followed by a prolyl cis/trans isomerization by PIN-1. Phosphorylated PML is then recognized by KLHL20, ubiquitinated and subsequently degraded by the proteasome. PML degradation and mTOR derepression reinforces HIF-1α activity and participates in patient’s poor clinical prognosis (Yuan et al. 2011) (Figure 35). There was also evidence of FBXO3 interacting with PML (BIOgrid data) but it was reported to target HIPK2 rather than PML (Shima et al. 2008).

Figure 35 : KLHL20-CUL3-ROC1 is targeting PML for degradation under hypoxic stress. Diagram showing CUL3-KLHL20 complex activity under normoxia and hypoxia in prostate cancer. Under hypoxia, the transcription factor HIF-1α is activated causing KLHL20 protein abundance to increase leading to more CRL complex formed and ultimately to increased PML degradation. A double-negative feedback loop to amplify HIF-1 signaling through PML degradation and mTOR derepresion is also shown (Yuan et al. 2011).

IV) PML, a tumor suppressor

1) PML physiological functions

Over the past 25 years, PML has been intensively studied, and its implication in key cellular events under physiological and pathological conditions gave rise to a model of PML depicting a functional multi-faceted protein (Lallemand-Breitenbach et al. 2010; Bernardi & Pandolfi 2007; Sahin et al. 2014). PML was reported as being implicated under physiological conditions in a number of cellular pathways, such as stress response mediated through PML NBs. Besides, PML is involved in metabolic pathways through its involvement in AKT/mTOR pathway and in the hematopoietic stem cell maintenance by the activation of Fatty Acid Oxidation (FAO) pathway (Sahin et al. 2014; Nakahara et al. 2014). PML plays a role in circadian rhythm by controlling the nuclear localization of Period circadian protein homolog 2 (PER2), a circadian clock regulator (Miki et al. 2012; Miki et al. 2016). Furthermore, PML is also involved in aging; PML NBs decrease with age with a concomitant decrease in response abilities of cells to stress (Wenger et al. 2014). It is also involved in the unfolded protein response by causing misfolded protein SUMOylation and degradation through RNF4 mediated Ubiquitination (Guo et al. 2014). Finally, PML is implicated in many other processes where, under pathological conditions, it displays a tumor suppressor role (Figure 36).

Figure 36 : PML functions in diseases. PML plays an important role in indicated pathways under physiological (yellow) and pathological conditions (green) (Guan & Kao 2015).

Studies based on a mouse model knocked out for Pml (Pml-/-) also gave interesting insights on PML functions. First, Pml knockout mice are viable meaning that even though PML is involved in many key cellular processes, its absence is not lethal. However, these mice are

more susceptible to develop tumors when exposed to carcinogenic chemicals (Bernardi & Pandolfi 2003). Moreover, Pml-/- mice are more prone to viral infections and immunopathologies, thus supporting a role for PML in innate immunity (Bonilla et al. 2002). Pml-/- mice and derivated cells are resistant to the lethal effects caused by apoptotic stimuli thus proving the importance of PML in apoptotic pathways (Bernardi & Pandolfi 2003). Another study showed that Pml-/- Mouse Embryonic Fibroblasts (MEFs) displayed a decreased cellular adhesion with slower migration, but increased proliferation compared to Pml+/+ MEFs (Tang et al. 2013). Finally, a behavioral study performed on Pml-/- mice showed impaired spatial memory and conditional learning. This study suggests that PML also plays an important role in synaptic plasticity and hippocampus-dependent learning (Butler et al. 2013). Interestingly, it has been also shown that -in the absence of Pml- the transition between radial glial cells and basal progenitors during the brain development is impaired. As a result, there is an overall decrease in the thickness of the cortex wall (Regad et al. 2009).

2) PML and diseases

After the discovery of the PML gene, studies were able to show that the encoded protein was in fact a tumor suppressor in several types of cancer such as gastric (Lee et al. 2007), lung (Zhang et al. 2000), bladder, prostate, colorectal and even breast cancer in which loss of PML leads to more aggressive or invasive tumors (Gurrieri, Capodieci, et al. 2004; Koken et al. 1995; Guan & Kao 2015; Gambacorta et al. 1996). Inhibition of cell proliferation, cell cycle arrest, apoptosis or senesce have been observed in cells overexpressing PML, whereas cells that do not express it show increased cytokine-induced apoptosis, resistance to UV and increased proliferation. In addition, elevated spontaneous and chemically induced tumorigenesis was observed in PML knockout mice (Mu et al. 1997; Wang, Delva, et al. 1998; Rego et al. 2001; Guan & Kao 2015). Besides, Pml-/- mice are viable but do not react well to stress such as infections (Rego et al. 2001). These studies collectively suggest a tumor suppressor role for PML, dependent of the cellular context. In addition, PML NBs are also believed to be storage structures that allow accumulation or sequestration of specific proteins that can be released when needed. As shown previously, these PML NBs could also serve as platforms for protein- protein interactions and promote post-translational modifications like SUMOylation, acetylation, ubiquitination or even phosphorylation of important proteins like the tumor suppressor protein P53 (Mao et al. 2011; Reineke & Kao 2009).

PML tumor suppressor activity is mediated through different mechanisms (Guan & Kao 2015) (Figure 37). PML NBs can sequester proteins in order to repress their function. PML NBs-associated proteins are recruited by PML, which mediates protein interaction for their activation. In fact, PML NBs are used by the cell as post-translational platforms or hubs that allow protein activity and functional regulation. PML is also involved in gene expression control by facilitating interaction of transcription factors and co-regulators with specific regions of the genome. DNA damage repair is regulated by complexes involving PML and PML NBs in order to contribute to the maintenance of the genome integrity through Alternative Lengthening of Telomeres (ALT), a process that involves homologous recombination of DNA (Cesare & Reddel 2013; Guan & Kao 2015). PML being involved in so many pathways, it is bound to be part of essential and key cellular pathways like apoptosis, AKT signaling, P53 stability or even gene regulation (Guan & Kao 2015) (Figure 37).

Figure 37 : Tumor suppressor pathways involving PML Nuclear Bodies. Diagram showing the different pathways on which PML Nuclear Bodies are acting through protein-protein interactions, sequestration or through post-translational modifications (Guan & Kao 2015).

3) PML and the apoptotic pathway

PML appears to be involved in apoptotic pathways directly through the activation of Caspase- 3 (CASP-3), in response to various stimuli such as γ-irradiation in which the lethal effect is attenuated in PML knockout mice and cells; Tumor necrosis factor receptor superfamily member 6 (FAS); Tumor Necrosis Factor-α (TNFα); type I and II interferon (INFs) and ceramide (a molecule composed found in membranes). However, the apoptosis regulator BAX and Cyclin-dependent kinase inhibitor 1B (p27K1P1) can also be recruited by PML to PML NBs to trigger apoptosis indirectly thus bypassing caspase-3 pathway (Huang et al. 2011; Bernardi & Pandolfi 2003; Wang, Ruggero, et al. 1998; Quignon et al. 1998; Guan & Kao 2015) (Figure 37, Caspase pathway).

4) PML and the P53 pathway

PML is involved in the regulation of P53, an extensively studied protein in tumor biology. PML is able to regulate its activity and cellular function such as cell cycle arrest, senescence, DNA repair or apoptosis. In fact, PML is able to sequester MDM2, the E3 ubiquitin ligase of P53, in NBs upon cellular stress or DNA damage (Lane & Levine 2010; Bernardi et al. 2004; Louria- Hayon et al. 2003). MDM2 stability and therefore P53’s, is also depending on Ubiquitin carboxyl-terminal hydrolase 7 (USP7 also known as HAUSP) deubiquitinase activity on MDM2. However, the association of PML and MDM2 is disrupted by the Big Map Kinase 1 (BMK1). MDM2 is then able to induce P53 destabilization through its ubiquitination (Yang et al. 2013). Under DNA damage conditions, P53 is phosphorylated by Chk2 (on S20 residue) preventing P53 degradation by blocking its interaction with MDM2. Furthermore, P53 is acetylated by the acetyltransferase CBP/p300 on K382 and phosphorylated by the Homeodomain-Interacting Protein Kinase 2 (HIPK2) on S46. Both of these proteins are recruited to PML NBs, along with tumor suppressor AXIN, under DNA damage or ultraviolet stress. These modifications induce cellular apoptosis or senescence through P53 transcriptional activity (Li et al. 2011; Hofmann et al. 2002; Guo et al. 2000). This transcriptional activity of P53 can be repressed through its deacetylation by the deacetylase SIRT1, also recruited to PML NBs, upon PML overexpression or through activation of oncogenic RAS (Ha-RAS V12) (Langley et al. 2002). To summarize, P53 regulation is dependent on PML and more specifically on the composition of PML NBs that will allow to control the abundance and

activity of the tumor suppressor P53 by sequestering MDM2 or SIRT1 for example (Guan & Kao 2015) (Figure 37, P53 pathway).

PML NBs also have an effect on another potent tumor suppressor protein called Retinoblastoma protein (RB) whose phosphorylation prevents its interaction with the transcription factor E2F (a transcription activator) hereby promoting cell cycle progression. This process can be reversed through dephosphorylation of RB by the phosphatase1α (PP1a). E2Fs can then bind to RB preventing E2F dependent transcription and cell cycle progression. In MEFs, PML protein expression is induced by oncogenic RAS causing an hypo- phosphorylation of RB and its co-localization in PML NBs to finally lead to cellular senescence (Regad et al. 2009; Vernier et al. 2011; Ferbeyre et al. 2000).

5) PML and transcriptional regulation

PML is able to repress gene transcription through sequestration of transcription factors in PML NBs or by association to complexes that repress transcription. For example, in TNFα induced apoptosis, PML NBs can sequester the RelA/p65 subunit of NF-κB to inhibit its transcriptional abilities. NF-κB can no longer bind to the promoter of the A20 gene that inhibits apoptosis from TNFα signaling using negative feedback (Wu, Xu & Chang 2002; Wu et al. 2003). Also sequestered in PML NBs are the transcription factors Sp1 and Nur77 but also Signal Transducer and Activator of Transcription 3 (STAT3), efficiently preventing them from binding to their target promoters (Wu, Xu, Ran, et al. 2002; Vallian, Chin, et al. 1998; Kawasaki et al. 2003). PML can also associate with co-repressors such as c-ski, Nuclear receptor CoRepressor 1 (N- CoR), mSin3A and Histone Deacetylase 1 (HDAC1) that can for example mediate transcriptional repression through the tumor suppressor Mad (Khan et al. 2001; Guan & Kao 2015) (Figure 37, transcriptional repression).

However, PML can also positively regulate transcription. DAXX, for example, which represses the expression of transcription factor Pax3 and Glucocorticoid Receptor alpha (GRα) genes, is recruited in PML NBs leading to derepression of these genes (Chang et al. 2011; Lin et al. 2006; Lin et al. 2003; Li et al. 2000). This interaction is SUMO dependent since it involves both SUMO1 and a SIM in DAXX. It was also reported that HDAC7 is sequestered in PML NBs upon TNFα stimulation leaving MMP-10 promoter, leading to increased MMP-10 expression (Gao, Cheng, et al. 2008). MMP-10 is also known as Stromelysin-2 and can degrade fibronectin and gelatins. Class II major histocompatibility complex expression is also promoted

through blocked degradation of class II transactivator (CIITA) by PML (Ulbricht et al. 2012). Interferon-responsive gene (IRG) expression and Interferon beta (IFNβ) activation is facilitated by transcription factors association (NF-κB, CREB-binding protein (CBP) and STAT1) with PML isoform II even though there is currently no proof of PMLII being present at these promoters (Y. Chen et al. 2015). ADP-ribosylation factor (ARF) and Siamois, which are β- catenin responsive genes, are activated by complexes based on P300, β-catenin and PML (Shtutman et al. 2002). The oncoprotein c-Fos was also reported to collaborate with PML and enhance AP-1 transcriptional activity which is essential for c-Jun DNA binding and activity under UV stress (Salomoni et al. 2005; Vallian, Gäken, et al. 1998). Another example where PML is essential, is the trans-activation of the p21WAF1/C1P1 gene through all-trans retinoic acid (AT-RA) which requires PML (Wang, Delva, et al. 1998). Furthermore, master transcription factors for hematopoietic stem cells, GATA1 and GATA2, interact with PML hereby facilitating their transcriptional activities (J. Wu et al. 2014; Tsuzuki et al. 2000) (Figure 37, transcriptional activation). In summary, PML is able to activate the transcription of a gene by favoring post-translational modifications of transcriptional factors or by just stabilizing them or even by sequestering transcriptional co-repressors in PML NBs.

6) Role of PML in DNA damage repair

PML and PML NBs are believed to be playing a crucial role in DNA damage repair and in an alternative mechanism of telomere maintenance called ALT, which is not dependent on telomerase, in cancer and immortalized cells. In these cells, telomeric DNA colocalizes with PML as well as with telomere-binding proteins TRF1 and TRF2 but also with NBS1, Mre11, Rad51 and 52, which are proteins, involved in DNA synthesis and recombination. These proteins are important to guaranty the integrity and genomic stability of ALT cells, therefore PML and PML NBs play an important role in DNA damage by binding to these proteins (Dellaire & Bazett-Jones 2004; Grobelny et al. 2000; Chung et al. 2012; Flynn et al. 2015).

After exposure to ionizing radiation (IR), the DNA damage protein DNA topoisomerase 2-binding protein 1 (TopBP1) is co-localized and stabilized by PML (Xu et al. 2003). PML NB formation is facilitated by the activities of Nibrin (NBS1) and ATM, ATR and CHK2 kinases upon DNA Double Stranded Breaks (DSBs) (Dellaire et al. 2006). Also co-localizing in PML NBs is the DNA helicase BLM (Bloom syndrome protein) which is an important player in genomic stability regulation. After DNA damage or during late S/G2 phase in normal cells,

Replication Protein-A (RPA), BLM and RAD51 assemble in PML NBs (Bischof et al. 2001; Zhong et al. 1999). The RAD50-Mre11-NBS1 complex plays a role in the maintenance of telomere length when telomerase is absent and it is also implicated in Double Stranded Breaks repair through homologous or non-homologous recombination repair (Le et al. 1999). This complex also co-localizes with PML under Ionizing Radiation, suggesting a role for PML in the repair of DNA damage and genomic stability (Guan & Kao 2015) (Figure 37, DNA damage repair pathway).

7) PML and the Akt pathway

Metabolism, migration, survival and cell cycle progression are regulated by substrates of RAC- alpha serine/threonine-protein kinase (also known as Akt). PML recruits protein phosphatase 2a (PP2a) to PML NBs as well as Akt, which gets dephosphorylated by PP2a leading to inhibition of its kinase activity. Thus, PML sequesters Akt in PML NBs and favors its inactivation (Carnero & Paramio 2014; Trotman et al. 2006). Akt activity can also be inhibited by PML through the eIF4E-NBS1-PI3K-Akt axis. Eukaryotic translation initiation factor 4E (eIF4E) can be found in PML NBs where PML negatively regulates its activity by directly interacting with it. This, in turn, causes a reduction in the activation of the Akt pathway through a decrease of NSB1, an upstream activator of the PI3K/Akt pathway and mRNA level that is dependent on eIF4E activity (Culjkovic et al. 2008). Another phosphoinositide-3 kinase-Akt (PI3K/Akt) activity suppressor is also positively regulated by PML: PTEN (phosphatase and tensin homolog), another tumor suppressor. In order for PTEN to exercise its tumor suppressor activity, it needs to be mono ubiquitinated on K289 by NEDD4-1 E3 Ub ligase, which will allow for its nuclear localization and activity. This post-transitional modification can be removed by the action of HAUSP thus preventing PTEN nuclear localization (Trotman et al. 2007). If PML is lost or inactivated, a decrease in PTEN nuclear localization is observed while overexpression of PML causes a decrease of PTEN deubiquitination by HAUSP in PC3 prostate cancer cell model. This is caused by the overexpressed PML binding DAXX protein which normally stabilizes HAUSP (Song et al. 2008). It also seems that cytoplasmic PML (cPML) plays a role in calcium release-associated apoptosis. In fact, cPML allows for 1.4.5-triphosphate receptor phosphorylation that is dependent on PP2a and Akt activation, which will in turn triggers the release of calcium from the Endoplasmic Reticulum ultimately leading to apoptosis (Giorgi et al. 2010). To summarize, PML regulates cell proliferation and survival by inhibiting

Akt kinase activity through PP2a, eIF4E and HAUSP (Guan & Kao 2015) (Figure 37, Akt pathway).

8) Cytoplasmic PML in tumorigenesis

Depending on the biological context, cytoplasmic PML (cPML) have been reported to display both tumor suppressive and oncogenic functions. The cytoplasmic isoform of PML, PMLVII, was identified in plasmacytoma J558 cells and it is the smallest of the PML isoforms containing exons 1-4, 6, 7 and part of exon 9. This isoform helps tumors to evade the host immune defenses by contributing to Major Histocompatibility Complex (MHC) class I antigen presentation. This antigen presentation at the cell surface normally alerts the immune system of compromised cells (Liu et al. 1998). Apart from normal genetic splicing, cytoplasmic PML can be the result of neutrophil elastase activity that truncates PML-RARα protein in APL (upstream of the Nuclear Localization Signal (NLS) at V420 or 432). Rare events producing a small mutation in an exon (1272delAG) or at a splice site (IVS3-1G to A) causing a premature STOP codon, will also lead to PML cytoplasmic isoforms lacking an NLS and identified in aggressive APLs. These cells from APL patients display an increased proliferation, reduced levels of apoptosis and are resistant to Retinoic Acid treatments (Lane & Ley 2003; Gurrieri, Nafa, et al. 2004; Gao et al. 2013; C. Bellodi et al. 2006). In addition, truncated PML mutant appear to be able to sequester wild type cytoplasmic PML by dimerizing with it hereby preventing P53 tumor suppressive functions (Cristian Bellodi et al. 2006). In hepatocellular carcinoma, an increased expression of PML and an increased cytoplasmic localization was observed but there is no proof so far of PML mutations in these tissues (Terris et al. 1995; Chan et al. 1998). Even though mutated forms of cytoplasmic PML are suggested to have an oncogenic effect, there are currently no data suggesting that wild type cytoplasmic PML does.

On the other hand, there are no evidence pointing out a tumor suppressive function of cytoplasmic PML. The activity of M2 type pyruvate kinase (PKM2), that is overexpressed in many cancers, is inhibited by a endogenous cytoplasmic PML mutant lacking the NLS (Chen et al. 2014; Shimada et al. 2008). After 24 hours of treatment with Transforming Growth Factor beta (TGFβ) -which can both promote or inhibit tumorigenesis depending on the biological context- cytoplasmic PML expression is induced. This isoform lacking the NLS and containing exons 1-3, 7a, 8a and 8b, is required for the Smad2/3-dependent transcription through the formation of a complex in endosomes (TβRI/ TβRII/SARA/Smad2/3). This complex is required

for cell senescence, apoptosis and inhibition of proliferation under TGFβ treatment (Lin et al. 2004; Ikushima & Miyazono 2010). This signaling can be blocked through the overexpression of the homeoprotein TGIF that will cause the retention of this isoform in the nucleus by interacting with TGIF (Seo et al. 2006). These studies collectively suggest that TGFβ tumor suppressor activity is promoted by cytoplasmic PML regulation.

Inositol 1,4,5-trisphosphate receptor (IP(3)R), Akt and PP2a form complexes with PML at the Endoplasmic Reticulum (ER) and Mitochondria Associated Membranes (MAM). Apoptosis by calcium release is promoted when overexpressed PML is targeted to the outer -/- surface of the ER in MEFs. In Pml MEFs, apoptosis response to H2O2 or menadione is impaired due to a decreased release of calcium from the ER caused by an enhanced phosphorylation of IP(3)R by Akt thus further supporting the tumor suppressive function of cytoplasmic PML (Guan & Kao 2015; Giorgi et al. 2010).

9) PML in cancers

Inactivation of PML has been observed in many different cancers and occurs through multiple mechanisms other than somatic mutation, since only very few of these mutations have been reported to date (Gurrieri, Capodieci, et al. 2004; Salomoni et al. 2012). Studies suggest that PML decreases and/or inactivation might be caused by events at the transcriptional and/or post- translational level. However, alternative splicing of PML mRNA and epigenetic regulation of the gene have not been extensively investigated, unlike PML post-translational modifications (Reineke & Kao 2009). In addition, PML protein down-regulation is observed in many types of cancers but no concomitant decrease of its mRNA is observed in those samples (Reineke & Kao 2009), which makes it difficult to associate PML levels with cancer, since most gene expression databases today are established on mRNA levels. This suggests that post- translational mechanisms could play an important role in these diseases, through the regulation of PML stability (Guan & Kao 2015). Accumulating evidence suggests that Post-Translational Modifications (PTMs) crosstalk regulates PML function. PML and PML NBs are involved in many key cellular processes, the alteration of some of which has been linked to cancer. It is relevant to better understand how PML and PML NBs, as central players in the regulation of cellular processes, are regulated in both normal and altered conditions. This knowledge might allow for the development of new therapeutic strategies.

Thesis Objective

The Nuclear Organization and Oncogenesis unit directed by Anne DEJEAN, is interested in studying the molecular and cellular mechanisms underlying the development of human cancers, with a particular emphasis on the role of the post-translational modifications.

In 1990, this laboratory identified the chromosomal translocation associated with Acute Promyelocytic Leukemia (APL) (de Thé et al. 1991). This translocation generates the oncogenic fusion PML-RARα protein where the RARα moiety can be targeted by Retinoic Acid (RA) and the PML (ProMyelocytic Leukemia) moiety by arsenic trioxide treatment. The combined treatment of RA and arsenic trioxide leads to cure nearly 100% of APL patients and represent the prime example of oncoprotein targeted therapy (Mi et al. 2015). Since then, the lab kept a strong implication in subsequent discoveries concerning PML, including its post- translational modification by SUMO. Logically, the team started the characterization of this pathway through functional studies of its components and identification of new enzymes and targets. Thereafter, the aims of the team widened and multiple thematics are now investigated in the lab such as the genetic/epigenetics aspect of liver tumors, cellular senescence mechanisms and in depth study of the SUMO pathway to expand our knowledge about its involvement in cancer or even cell reprogramming at the cellular level as well as at the organismal level. This was made possible through the initial study of PML, a well-established tumor suppressor that is both a key player and a target in APL. In spite of its tumor suppressor properties, little is known about the normal function and regulation of PML. Therefore, PML is more than ever an interesting target of study that will hopefully lead to better understanding of the normal or pathological function of the cell.

This thesis is part of the general thematic studied in the lab. It aims at providing a better comprehension of PML and PML Nuclear Bodies (PML NBs) regulation and by extension of SUMOylation and their implication in oncogenesis through the identification of new regulators of PML Nuclear Bodies. Our knowledge regarding PML and the SUMO pathway components is expanding, however new enzymes regulating the PML NBs still remain to be discovered. A number of deSUMOylases and SUMO-dependent ubiquitin ligases such as RNF4 and ARKADIA, have been discovered in the last decade (Erker et al. 2013; Lallemand- Breitenbach et al. 2008). Hence, it was demonstrated that arsenic could induce the

polySUMOylation of PML and PML-RARα leading to the degradation of the fusion oncoprotein (Müller, Matunis, et al. 1998; Tatham et al. 2001; Lallemand-Breitenbach et al. 2008). However, few actors of the enzymatic machinery involved in this process have been yet identified, even though they are probably implicated in the SUMOylation of many other substrates when submitted to variable triggers. In order to identify new regulators PML NBs and then possibly of the SUMO pathway, we used PML unique properties notably its dependence on SUMOylation to form PML NBs to elaborate a screen looking for modulators of PML NB structure, used here as a phenotypic read-out. The method consisted in a high throughput genome wide screen with small interfering RNAs (siRNAs) coupled to high throughput imaging, allowing for the analysis of effects caused by the inactivation of each gene on the morphological integrity of PML NBs. Candidates identified in this screen that reproduce significant PML NB phenotypes were then subjected to in-depth biochemical and functional characterizations.

Results

The goal of this project was to uncover and characterize new molecular players regulating PML, directly or through SUMOylation, using changes in PML NBs structure as an amenable and quantifiable proxy. In fact, our team, as well as others, was able to show in the past that SUMOylation of PML was conditioning the formation of PML NBs in cells (Zhong et al. 2000;

Müller, Matunis, et al. 1998). Moreover, one hour of arsenic trioxide (As2O3) treatment on HT1080 cells induces the polySUMOylation of PML, which leads to an increase in PML NB number and size (Müller, Matunis, et al. 1998). After several hours (24 to 48 hours), the ubiquitination by RING Finger protein 4 (RNF4) and subsequent degradation of PML triggered by its polySUMOylation leads to the disappearance of PML Nuclear Bodies (Zhang et al. 2010; Tatham et al. 2008; Lallemand-Breitenbach et al. 2008) (Figure 38A). This phenomenon can also be observed through Western blot of our stable clone expressing GFP-PML IV, with the presence of multiple high molecular weight bands (hyperSUMOylated forms) after one hour which slowly disappear within 24 hours of arsenic trioxide treatment (Figure 38B). Hence, PML NBs number and morphology characterize a reliable mark of PML integrity and/or of its level of modification by SUMO.

Figure 38 : Effect of Arsenic trioxide treatment on PML and PML Nuclear Bodies. A. Fluorescence microscopy of HT1080 GFP-PML IV cell line after indicated time of arsenic trioxide treatment (1µM). Scale bars represent 10.0µm. B. Effect of arsenic trioxide on PML observed by Western blot using indicated antibodies.

I) Initial Data

1) Screen Design

The screen was designed and carried out by Hélène Neyret-Khan, in collaboration with the team of Michael Howell in London (High throughput screening platform, London Research Institute, Cancer Research UK). This method involved the transfection of cells by a bank of 21000 siRNAs targeting the human genome, and the analysis of PML NBs morphology by fluorescence imaging. Cells in 800 x 96-well plates were transfected before result visualization under the microscope. It was necessary to use an easily transfectable cell line in which the number of PML NBs could be readily counted by an automated microscope. This cell line was obtained from HT1080 (human fibrosarcoma) cells in which GFP-PML IV was stably expressed. PML IV isoform was chosen because it has been extensively studied in the laboratory for its implication in cellular senescence (Bischof et al. 2002; Bischof et al. 2005). Preliminary control experiments were carried out to make sure that this cell line was correctly responding to arsenic trioxide treatment (Figure 38A and B) and that the level of GFP-PML IV was as close as possible to the endogenous level. Once these features were validated, siRNA controls were chosen. To show that PML NB morphology was indeed modulated by known SUMO pathway regulators, siUBC9, the only E2 enzyme for the SUMO pathway, was chosen as a positive control. Inactivation of UBC9 or of all SUMO isoforms leads to the disappearance of PML NBs and to the formation of one to two big protein aggregates into the cell nucleus. In addition, another control (siPML) shows that inhibition of PML correlated with the complete disappearance of PML NBs.

After the adequate cell line was validated and siRNA controls were properly identified, the first genome wide screen was subsequently carried out. HT1080 cells expressing GFP-PML IV were transfected with a bank of siRNAs covering 21360 human genes (Dharmacon), each gene being targeted by a mix of four siRNAs and treated to minimize off targets effects (SMARTpool series) (Figure 39A). These transfections were performed in triplicates and included four positive controls (siPML, siUBC9, siSUMO2/3, siRNF4) and a negative control (siSCRAMBLE). 72 hours post-transfection, plates were fixed, stained with DAPI and read on the Cellomics ArrayScan (Figure 39A). The automated acquisition and analysis of immunofluorescence images allowed for the phenotypic study of PML NBs morphology based on approximatively 20 different parameters including NB number, size, intensity and

localization (nucleus vs cytoplasm) (Figure 39A). At the same time, nuclei measurements were taken, allowing estimation of both cell viability and proliferation, therefore eliminating all candidates that would create a cell cycle arrest or a global destructuration of the cell. The analysis of these results showed that the control siRNAs present on each plate is as expected at the external boundaries of distributions for the different chosen parameters. For example, siPML exhibits a low Nuclear Body count along with a low intensity (Figure 39B) whereas cells treated with siUBC9 show less NBs but a stronger fluorescence intensity (Figure 39B and C).

Figure 39 : Methodology and controls for the genome wide siRNA screen. A. Genome wide siRNA screen methodology and analysis. B. Primary screen distribution of Z scores for two indicated parameters. Controls siRNA are indicated in colors. C. Plot showing primary screen combination of two parameters and highlighting siUBC9 control wells (Neyret-Kahn 2012).

2) Primary screen results

After this primary genome wide screen, the number of candidates altering PML NBs morphology in a significant manner was still exceeding the thousand candidates. Therefore, in order to classify the candidates into three major phenotypic classes, different parameters, mainly number, localization and intensity of bodies, were combined (Figure 40):

- Class I (“hypoSUMOylation like” phenotype) is characterized by the formation of big PML aggregates in the nuclei, as observed in cells treated with siUBC9 or siSUMO2/3. - Class II (absence of PML NBs) is characterized by the complete absence of PML NBs as observed in cells treated with siPML. - Class III (“hyperSUMOylation like” phenotype) is characterized by an increase of PML NBs number and intensity compared to the control. This phenotype can be obtained after arsenic trioxide treatment and under siRNF4 or siARKADIA treatment, two SUMOdependent E3 ubiquitin ligases, causing the stabilization of PML (Erker et al. 2013; Yin et al. 2012; Tatham et al. 2008; Lallemand-Breitenbach et al. 2008); or by an overproduction of PML as observed for example upon interferon stimulation (Zhou & Bao 2014).

Figure 40 : Phenotypic classes used to regroup candidates in the screen. (Neyret-Kahn 2012)

3) Validation screen

Out of the 1000 primary hits exhibiting changes in PML NBs morphology, 380 of our best candidates were chosen for a second validation screen using a deconvolution siRNA library. Four independent siRNAs for each target gene were transfected individually (1well = 1 siRNA for 1 gene). This allowed to control for off-targets or randoms, false positive effects. 124 candidates for which at least two independent siRNAs were able to induce a change in the number or intensity of PML NBs were validated as potential PML and/or SUMO regulators (Annex 2). This list contains UBC9 and SAE1/SAE2, the E2 and E1 enzymes of the SUMO pathway, which are well known to regulate PML and PML NB integrity (Nacerddine et al. 2005), thus validating the approach (Figures 41A and 41B). In order to identify proteins involved in the negative regulation of PML stability, we looked for siRNAs increasing both the number and intensity of PML NBs (class III). Seven of our validated candidates met these criteria, among them SKP1a and RBX1. Three out of four siRNAs targeting SKP1a or RBX1, were able to induce a change in count and two out of four in PML NBs intensity compared to the control (Figure 41C).

Figure 41 : Results obtained from the second validation screen. A. Examples of immunofluorescence data obtained from the second screen showing morphological changes under indicated siRNA treatment. B. Network obtained using STRING (Szklarczyk et al. 2015) with the list of 124 validated hits. C. List of validated target proteins whose depletion significantly increases both PML NBs number and intensity (Class III) (Neyret-Kahn 2012).

II) Functional study of selected candidates

1) Identification of SKP1a and RBX1

Among the 124 validated candidates, SKP1a and RBX1 whose inactivation mimics the “hyperSUMOylation like” phenotype (Class III), were selected from the best hits (Figure 41C). In fact, these two proteins are well characterized since they are both part of an E3 Ubiquitin ligase complex called the SKP1-Cul1-F-Box-containing complex or SCF (also known as CRL1). This complex is composed of an adaptor protein, S-phase kinase-associated protein 1 (SKP1, also known as SKP1a), capable of recognizing an F-Box protein which provides the substrate specificity to the complex. In addition, this complex also includes an E2 conjugating enzyme, UBC3, which allows the transfer of ubiquitin onto the target protein. This E2 is recruited by Ring Box protein 1 (RBX1 also known as ROC1) through its RING finger domain (Wei & Sun 2010). Finally, SKP1a and RBX1 are linked to one another through a scaffold protein called Cullin 1 (Cul1) (Bond & Wu 2011; Skaar et al. 2014) (Figure 42). This complex notably regulates cell cycle by degrading cyclin-dependent kinases inhibitors (Nakayama & Nakayama 2006).

Figure 42 : SKP1-CUL1-F-Box containing complex (SCF). P stands for Phosphorylation and Ub for Ubiquitin (Bond & Wu 2011).

2) Manual validation of screen results

Results from the screen were manually validated by transfecting siRNAs targeting SKP1a and RBX1 genes in HT1080 GFP-PML IV cell line. Cells transfected with siSKP1a or siRBX1 display a higher number of PML NBs, with a median of 24 bodies per nuclei for SKP1a (IQR 18-35) and RBX1 (IQR 18-32) depleted cells, compared to a median of 19 bodies observed in control cells (IQR 14-22) (Figure 43A and 43B). We can also observe an increase in surface and intensity of PML NBs under these conditions compared to the control (Figure 43A). Around 300 cells were counted for each conditions in this experiment and the difference of PML NB count between the control and the depleted cells is statistically significant (p-values in both cases (P<0.0001)) (Figure 43B). This result was also reproduced in another clonal cell line HT1080 GFP-PML IV, excluding clone-specific artefacts (data not shown). In summary, knock down of SKP1a and RBX1, as verified by Western blot, induces a significant increase in the number of PML NBs.

Figure 43 : SKP1a and RBX1 depletion effect on PML Nuclear Bodies morphology in HT1080 GFP-PML IV cells. A. Fluorescence data from cells transfected with indicated siRNAs. B. Box and whiskers plot of PML NBs count per nuclei in cells treated with indicated siRNAs. Statistical calculations were made using Mann-Whitney test, **** reports p-values of p<0.0001.

3) Depletion of SKP1a and RBX1 stabilizes PML

The simplest hypothesis to explain the increase in the number, size and intensity of PML NBs upon depletion of SKP1a and RBX1 would be a stabilization of the PML protein. To test this assumption, cells were transfected with the corresponding siRNAs, lysed and the proteins extracted analyzed by Western blot. Depletion of SKP1a or RBX1 caused an increase of all PML forms: the unmodified form with the lower molecular weight band (around 110kD), as well as the SUMOylated ones with higher molecular weights (between 120 and 200kD) (Figure 44A, left blot).

In order to determine if the increase in PML observed in cells depleted for SKP1a and RBX1 was dependent or not on its modification by SUMO, a treatment with arsenic trioxide was coupled to the siRNAs transfections targeting SKP1a, RBX1 and RNF4. RNF4 was used here as a positive control since this protein is known to specifically target SUMOylated forms of PML for ubiquitination and further degradation (Tatham et al. 2008; Lallemand-Breitenbach et al. 2008). As a reminder, under arsenic trioxide treatment, PML is hyperSUMOylated (one hour post-treatment) and subsequently slowly degraded 24 to 48h post-treatment. As expected, RNF4-depleted cells treated with arsenic trioxide display a stabilization of hyperSUMOylated forms of PML. In these cells, hyperSUMOylated PML is no longer degraded compared to the control where most of it was degraded after 24 hours treatment (Figure 44A, right blot). After 24 hours of arsenic trioxide treatment, depletion of SKP1a, RBX1 or both induced a strong increase in the amount of the PML protein, at levels close to, or higher than our positive control siRNF4 (Figure 44A, right blot). Knockdown efficiency was also checked by Western blot (Figure 44C). We also verified through RT-qPCR experiments that this stabilization was not due to transcriptional up-regulation (Figure 44B). A minor increase in PML mRNA levels is however observed, but to levels comparable to those obtained with our positive control siRNF4. The remarkable additive effect seen when using siRNAs against SKP1a together with RBX1 (Figure 44A, right blot), in addition to the current knowledge on these two candidates, indicates that these two proteins could modulate PML stability through a single complex, the SCF complex.

Figure 44 : SKP1a and RBX1 are involved in PML stability in presence or not of arsenic trioxide treatment. A. Western blot of HT1080 GFP-PML IV cells transfected with indicated siRNAs coupled to arsenic trioxide treatment (1µM) for indicated time. B. RT-qPCR to determine PML mRNA variation under indicated siRNA treatments. Expression normalized to housekeeping gene β- Actin (n=3). C. Western blots validating knockdown induced by indicated siRNAs.

Since these experiments were done using SMARTpools combining four siRNAs targeting the same gene, individual siRNAs were transfected to verify the reproducibility of the observed effects. The same experiment was carried out by transfecting single siRNAs followed by arsenic trioxide treatment and subsequent Western blot analysis. Individual siRNA for SKP1a or RBX1 led to the same stabilization of the PML protein as observed with the mixed siRNAs. Whereas the effect of individual siRNAs is hardly visible for the lowest-migrating PML bands, or at the beginning of the treatment, a strong stabilizing effect was observed after 24 hours arsenic trioxide treatment, and this was particularly clear for the most heavily SUMOylated forms of PML (Figure 45). Together these data indicate that depletion of SKP1a and RBX1 stabilize PML most likely through an alteration of the post-translational regulation of the protein stability.

Figure 45 : Validation of SKP1a and RBX1 involvement in PML stability through single siRNA experiment. Western blot of HT1080 GFP-PML IV cells treated with indicated siRNAs (one example for each gene) coupled to arsenic trioxide (Ars) treatment (1µM) for indicated time.

4) The overexpression of SKP1a and RBX1 destabilizes PML

Loss of function experiments point toward a role of SKP1a and RBX1 in the regulation of PML protein. The role of these two candidates was also tested in HeLa cells under overexpression conditions. Cells were transfected with plasmids coding for GFP-PML IV and/or one of the candidates fused to a FLAG tag. The overexpression of SKP1a or RBX1 display a drastic effect on PML stability since all PML bands detected by Western blot were reduced in intensity when SKP1a or RBX1 were overexpressed (Figure 46A, left blot). The profile for PML on Western blot is slightly different in Hela cells compared to HT1080 cells, since, depending on the cell line, PML is expressed and modified differently (Lallemand-Breitenbach et al. 2010). These results correlate with the disappearance of PML NBs observed in the immunofluorescence experiment (Figure 46B). In a similar manner, we observed that HT1080 stably expressing GFP-PML IV display a slightly lower number of NBs and a clear decrease in their intensity when transfected with vectors overexpressing SKP1a or RBX1 (here in red) when compared to control or neighboring untransfected cells (Figure 46B). Taking into consideration that cells overexpressing SKP1a and RBX1 are quite low in number due to poor transfection efficiency, it could be difficult to properly interpret this result. However, the observation that the decrease in PML observed by Western blot was inhibited by MG132 treatment (a proteasome inhibitor) strongly suggests that SKP1a and/or RBX1 overexpression induces PML degradation in a proteasome-dependent manner (Figure 46A, right blot).

Figure 46 : SKP1a and RBX1 are involved in PML stability: gain of function experiment. A. Western blot from transfected Hela cells with indicated plasmids. Cells were treated with MG132 (25 µM for 8 to 12 hours) when indicated. B. Immunofluorescence data from HT1080 GFP-PML IV cell line transfected with indicated plasmid. FLAG tagged proteins appear in red while GFP-PML IV is in green and nucleus are stained with DAPI (blue).

In summary, SKP1a and RBX1 are regulating PML stability in a proteasome-dependent manner both in normal conditions and under arsenic trioxide-induced stress as shown through gain and loss of function experiments.

5) RBX1 and SKP1a are both interacting with PML

To look for possible interactions between PML and SKP1a and/or RBX1, co- immunoprecipitated experiments were performed. First, 293T cells (Human Embryonic Kidney cells), chosen for their high transfection efficiency, were transfected with GFP-PML IV and FLAG-SKP1a or FLAG-RBX1 expression vectors. SKP1a and RBX1 were immunoprecipitated using anti-FLAG magnetic beads, and the presence of co- immunoprecipitated PML was assessed by Western blot. This assay showed that PML was co- immunoprecipitating with SKP1a and even more efficiently with RBX1 (Figure 47A). The reciprocal experiment was performed with GFP-PML IV, using anti GFP-magnetic beads, and the presence of FLAG-SKP1a or FLAG-RBX1 assessed by Western blot. Here again, a strong interaction between RBX1 and PML was observed and a weaker one with SKP1a (Figure 47B). Also in this experiment, RBX1 seemed to weakly bind to the beads in a nonspecific manner but it is still very clear that PML and RBX1 interact with one another. In addition, CUL1 was also co-immunoprecipitating with RBX1 as expected from the literature (data not shown).

Figure 47 : SKP1a and RBX1 interaction with PML. A. Western blot from 293T cells overexpressing the indicated proteins followed by immunoprecipitation of FLAG proteins (IP). Whole cell extract is shown on the right side (WCE). B. Western blot from 293T cells overexpressing indicated proteins followed by immunoprecipitation of GFP-PML IV (IP). Whole cell extract is shown on the right side (WCE).

In summary, PML interacts with RBX1 and SKP1a as part of a SCF complex that regulates its stability. This raises the essential question of the identity of the F-box protein that will bring within the SCF complex, the substrate specificity for PML.

6) Identification of the specific F-Box protein for PML

F-Box proteins are essential components of SCF complexes. Indeed, in order to target in a specific manner the multiple protein substrates for ubiquitination and subsequent degradation, different F-Box proteins can be loaded on the SCF complex. There are around 70 different F- Box proteins categorized into three classes: FBXL, FBXW and FBXO. In order to identify which F-Box protein is interacting with PML, a co-immunoprecipitation screen was designed and carried out in collaboration with F. Bassermann’s team, from the Technical University of Munich. Plasmids expressing FLAG-tagged F-Box proteins were transfected in 293T cells along with GFP-PML IV. F-Box proteins were then immunoprecipitated using their FLAG Tag. Co-immunoprecipitation of SKP1 was used as an internal control since, in the majority of the cases, SKP1 and F-Box proteins are interacting and forming dimers (Annex 4). Co- immunoprecipitation of PML was then assessed by Western blot. F-Box proteins of each of the three categories were tested and only one F-Box, FBXO9, was able to efficiently co- immunoprecipitate PML (Figure 48).

Figure 48 : Immunoprecipitation screen for F-Box proteins specifically recognizing PML. Western blot from 293T cells transfected with FLAG-F-Box and GFP-PML expression vectors after immunoprecipitation of FLAG-tagged proteins (IP). Whole Cell Extract is shown on the right side (WCE). Identified F-Box candidate is circled in red.

So far, very little is known about this protein. One publication showed FBXO9 implication in multiple myeloma through the degradation of TEL2/TTI1. This situation enables cell survival under growth factor withdrawal conditions (Fernández-Sáiz et al. 2013). The possible implication of FBXO9 in adipocyte differentiation through the degradation of Peroxisome Proliferator-Activated Receptor gamma (PPARγ) was also shown in mice (Lee et

al. 2013) and humans (Lee et al. 2016). Three isoforms of FBXO9 were described, which share an F-Box domain for their interaction with SKP1a and a Tetratrico Peptide Repeat (TPR) domain, a stretch of 34 amino acids probably involved in substrate recognition. There is one reported site for phosphorylation at S136 but no associated function has been yet identified. In addition, the protein also contains predicted sites for SUMOylation and two SUMO Interacting Motif (SIM) identified using GPS-SUMO software (Zhao et al. 2014). But, none of them have been experimentally confirmed yet (Figure 49).

Figure 49 : Knowledge summary on FBXO9 protein post-translational modifications and domains. F-Box domain allows for the interaction with SKP1 and TPR repeat domain is involved in protein-protein interaction.

7) Validation of the interaction between PML and FBXO9

To confirm the interaction between PML and FBXO9, 293T cells were transfected with plasmids coding for FLAG-tagged FBXO9 and GFP-PML IV. PML was immunoprecipitated using GFP magnetic beads and the presence of FBXO9 assessed by Western blot using anti- FLAG antibodies (Figure 50A). The reciprocal experiment has also been done in parallel. The result shows that PML was co-immunoprecipitated by FBXO9 to the extent as SKP1a and RBX1 used here as positive controls (Figure 50B). To confirm that this result was not an artefact caused by the overexpression of one or both of the proteins, an immunoprecipitation on endogenous proteins was performed. In order to increase the detection of the interaction, 293T cells were treated with Leptomycin B (a nuclear export inhibitor) to concentrate FBXO9

in the cell nuclei. The main technical challenge with this experiment was the detection of endogenous FBXO9 by Western blotting since FBXO9 protein migrates at the same molecular weight than light chains of immunoglobulin, thus masking the signal. However, a faint but specific band corresponding to the immunoprecipitated endogenous PML protein was detected indicating that endogenous FBXO9 co-immunoprecipitated with endogenous PML (Figure 50C).

In summary, the finding that PML interacts with RBX1, SKP1a and FBXO9, is in strong support for PML’s interaction with the SCFFBXO9 complex.

Figure 50 : FBXO9 interacts with PML. A. Western blot from 293T cells transfected with indicated plasmids and subsequent immunoprecipitation of GFP-tagged proteins (IP). Whole cell extract is shown on the right side (WCE). B. Western blot from 293T cells transfected with indicated plasmids and subsequent immunoprecipitation of FLAG-tagged proteins (IP). Whole cell extract is shown on the right side (WCE). C. Western blot from 293T treated with 20nM Leptomycin B and subsequent immunoprecipitation of PML proteins with a PML antibody (IP PML). IgGs serve as negative control (IP IgG). Whole cell extract is shown on the left side (WCE). Suspected FBXO9 band is shown with an arrow and vinculin was used here as loading control.

8) FBXO9 interacts with all PML isoforms

The next question that needed to be addressed was the specificity of the interaction for the different isoforms of PML, knowing that seven different isoforms of PML have been described so far, six of which are located in the nucleus and one is cytosolic (Nisole et al. 2013). To address this point, we performed co-immunoprecipitation experiments using plasmids expressing non-tagged FBXO9 along with plasmids expressing FLAG-PML isoforms. Three nuclear isoforms that are available in the lab (IV, V and VI) were tested as well as the cytoplasmic isoform of PML (VII). All isoforms were found to co-immunoprecipitate FBXO9 and whereas PML IV displayed the lowest efficiency of binding to FBXO9, PML VI and VII exhibited the highest level of interaction (Figure 51). It was difficult to raise any conclusion for PML V since it was reproducibly less expressed than the three other isoforms. From this experiment, it appears that FBXO9 is capable of interacting with all PML isoforms tested.

Figure 51 : FBXO9 interacts with different PML isoforms. Western blot from 293T cells transfected with indicated plasmids, lysed and subjected to immunoprecipitation of FLAG-tagged PML isoforms (IP). Whole cell extract is shown on the right side (WCE).

9) Localization of FBXO9-PML interaction

FBXO9 has been reported to be mostly cytoplasmic therefore we focused more precisely on the interaction with PML VII, the cytoplasmic isoform of PML that lacks a Nuclear Localization Signal (NLS). In order to characterize the location of this interaction in the cell, HT1080 cells were transfected with FLAG-PML VII along with GFP-FBXO9 and their interaction localized by immunofluorescence. Cells mounted on slides, were observed using an Apotome fluorescence microscope to look at slices (four z-stack slices in this case) of the cell (Figure 52A). GFP-FBXO9 (in green) co-localized with FLAG-PML VII (in red) in cytoplasmic foci and their co-localization appeared in yellow (Figure 52A).

However, one question remains: is the interaction between PML and FBXO9 only occurring in the cytoplasm? To answer this question, a co-immunoprecipitation experiment was designed in which FLAG-PML IV is overexpressed alone or together with untagged FBXO9. Cells were then lysed and the nuclear and cytoplasmic fractions collected and analyzed by Western blot. The cytoplasmic localization of GAPDH (Glyceraldehyde-3-phosphate dehydrogenase) was used here as an indicator of the quality of the fractionation (Figure 52B). Whereas PML IV was detected, as expected, mostly in the nucleus, a significant amount of the protein was also present in the cytoplasmic fraction. In agreement with the literature (Fernández-Sáiz et al. 2013), FBXO9 was mostly detected in the cytoplasmic faction, although a minor amount could be detected in nucleus (Figure 52B). Moreover, co-immunoprecipitation of FBXO9 with PML IV was observed in both nuclear and cytoplasmic fractions. It is important to keep in mind that, although nuclear or cytoplasmic fractions are clearly enriched in nuclear or cytoplasmic proteins respectively, this fractionation does not guaranty a 100% purity. A co- localization experiment by immunofluorescence was also carried out with ectopically expressed FLAG-PML IV and FBXO9 in order to observe a potential nuclear interaction. However, it was difficult to make a clear conclusion since PML IV was mainly detected in NBs while FBXO9 was mainly detected in the cytoplasm. These results, in addition to previous data suggest that FBXO9 is capable to interact with all PML isoforms and that it does so, both in the nucleus and/or the cytoplasm.

Figure 52 : FBXO9 is co-localized with PML VII and interacts with PML IV both in the nucleus and in the cytoplasm. A. 293T cells transfected with GFP-FBXO9 (green) and FLAG-PMLVII (red). Co-localization can be observed in yellow, nuclei were stained with DAPI (bleu). Immunofluorescence of four slices (z- stack) acquired using Apotome fluorescence microscope. B. Western blot from 293T cells transfected with indicated plasmids and subsequent immunoprecipitation of FLAG-tagged PML isoforms (IP). Whole cell extract is shown on the right side (WCE).

10) SUMOylation, arsenic and PML-FBXO9 interaction

Since hyperSUMOylation by arsenic trioxide regulates PML stability, we next tested whether arsenic trioxide and hence SUMOylation, would affect the interaction between FBXO9 and its PML target. To answer this question, we focused on PML V since it is known to be the isoform most susceptible to degradation upon arsenic treatment (Hands et al. 2014) and we compared cells treated with arsenic trioxide for one hour to untreated cells. The rationale for this approach was based on the observation that after one hour arsenic trioxide treatment PML becomes hyperSUMOylated, concomitant to the onset of ubiquitination and subsequent degradation. If FBXO9 requires SUMOylation to interact with PML, an increase in interaction would be expected in the presence of arsenic trioxide. However, this experiment failed to detect any increase in FBXO9 interaction with PML under arsenic trioxide treatment (Figure 53A).

PML uses SUMOylation and SIMs to interact with many partners in PML NBs and FBXO9 have predicted SUMO sites and SIMs that could be used for protein interaction as well. Therefore, to address a possible dependence of FBXO9 activity on the SUMOylation status of its substrates, an experiment was designed using GFP-PML3KR, a non-SUMOylable mutant for PML (Zhong et al. 2000). In this experiment, non-tagged FBXO9 was overexpressed in 293T cells along with GFP-PML IV or GFP-PML3KR. PML or the PML mutant were then immunoprecipitated using their GFP-tag and FBXO9 was detected by Western blot. FBXO9 co-immunoprecipitated efficiently both wild-type and SUMO-deficient PML mutant (Figure 53B). In conclusion, these data suggest that SUMOylation of PML is not required for interaction with FBXO9, although it is possible that binding of PML3KR to FBXO9 is indirect and mediated by the multimerization of the mutant with endogenous wild-type PML. Indeed PML shows a remarkable propensity to multimerize to form the PML NBs (Shen et al. 2006). Restoration experiments in PML-depleted (shRNA or PML-/-) cells should help clarifying this issue.

Figure 53 : FBXO9 interaction with PML is not dependent on arsenic trioxide treatment or SUMOylation. A. Western blot from 293T cells transfected with indicated plasmids, treated or not with arsenic trioxide for one hour followed by immunoprecipitation of FLAG-tagged PML (IP). Whole cell extract is shown on the right side (WCE). B. Western blot from 293T cells transfected with either Wild Type PML or SUMO deficient mutant PML3KR followed by immunoprecipitation of GFP-tagged PML constructions (IP). Whole cell extract is shown on the right side (WCE)

Having shown that FBXO9 is part of a SCF complex interacting with PML, the rest of the study aimed at investigating the role of FBXO9 in PML stability. To increase the level of sensitivity, id est to perform the experiments in cells harboring very low and limiting levels of PML, most of the studies below have been performed in the presence of arsenic trioxide, as used in published work on RNF4.

11) FBXO9 degrades PML under arsenic trioxide treatment

a) Transient depletion of FBXO9

For this part of the study, we intended to assess the role of FBXO9 on PML stability therefore Western blots were performed targeting FBXO9 only. HT1080 cells stably expressing GFP- PML IV were transfected with SMARTpool siRNAs targeting FBOX9 to analyze the effect of its knockdown on PML stability. siRNF4, siSKP1a and siRBX1 were used here as positive controls and siCrtl as a negative control. As previously described, in cells transfected with siCrtl, several PML bands were detected (Figure 54A, siCrtl lanes). In HT1080 cells, these can be clearly discriminated: the unmodified forms migrate at the lowest molecular weights (around 100kDa), followed by a higher molecular weight species corresponding to the monoSUMOylated form of PML (around 115kDa) and by a ladder representing the multiSUMOylated forms of PML (over 150kDa). After one hour arsenic trioxide treatment, PML is hyperSUMOylated and a shift in upper molecular weight bands could be observed. Finally, after 24 hours, PML is degraded and only the unmodified PML remains along with a small fraction of the hyperSUMOlated forms of PML (Figure 54A, siCrtl lanes).

However, when cells are depleted in FBXO9, an increase or a stabilization of all forms of PML is observed, especially for the unmodified and hyperSUMOylated forms after 24 hours of arsenic trioxide treatment (Figure 54A). The effects seen in these cases were comparable to the one obtained with the positive control siRNF4 (Figure 17A). RNF4, as previously described, is one of the few ubiquitin E3 ligases (along with ARKADIA) known to target hyperSUMOylated PML for ubiquitination and subsequent proteasomal degradation (Lallemand-Breitenbach et al. 2008). The effect after 24 hours of arsenic trioxide treatment is particularly strong for cells depleted in FBXO9 whereas this effect is barely detectable when cells are left untreated A similar situation is observed in RNF4-depleted cells. (Figure 54A). In the current settings, arsenic trioxide is used to force PML degradation since the physiological stimulus triggering its degradation through the SCF complex is unknown. Though it is still

possible that, like RNF4 and ARKADIA, FBXO9 needs to interact with SUMO chains of PML to trigger its ubiquitination. We observed that siSKP1a, in addition to depriving cells of SKP1a protein, also knocked down FBXO9 protein (Figure 54A, FBXO9 blot). This is consistent with the fact that SKP1a and FBXO9 usually work as dimers, therefore, under certain circumstances, removing one partner probably destabilizes the other as described in the literature (Yoshida et al. 2011). Interestingly, this effect is not reciprocal as knocking down FBXO9 did not have an effect on the quantity of SKP1a in the cell (data not shown). This result was rather expected since whereas FBXO9 protein is expected to dimerize with the sole SKP1a protein, SKP1a protein dimerizes with multiple F-Box proteins.

The effect of FBXO9 depletion on PML stability was also verified using individual siRNAs. Here again, depletion of FBXO9 induces an increase or stabilization of PML after 24 hours arsenic trioxide treatment, to the same extent as the stabilization induced by RNF4 depletion (Figure 54B). The consequences of knocking down FBXO9 on the endogenous PML protein were then investigated in HT1080 cells, using RNF4 knockdown as a control. HT1080 cells depleted in FBXO9 showed a small increase or stabilization of endogenous PML after 24 hours arsenic trioxide treatment whereas a very strong stabilization of endogenous PML in cells depleted in RNF4 was observed (Figure 54C). Of note, simultaneous suppression of RNF4 and FBXO9 did not display any additive effects when compared to suppression of each targeted protein alone.

RT-qPCRs were also carried out to check any transcriptional effects on PML. While no significant effect of siSKP1a, siRBX1 or siRNF4 on PML mRNA levels was observed, siFBXO9 led to a significant increase in PML transcripts (Figure 54D). However, even though PML mRNA levels were higher than in the control, no significant increase in protein amount was detected by Western blots under siFBXO9 in untreated cells (Figure 54A). These data support the idea that the stabilization observed is mainly based on a post-translational regulation process and not on a transcriptional level. In addition, some studies showed that mRNA levels alone could not explain the variation in PML protein abundance (Vogel & Marcotte 2012).

Figure 54 : FBXO9 depletion has an effect on PML stability: siRNA approach. A. Western blot from HT1080 cells stably expressing GFP-PML IV, transfected with indicated SMARTpool siRNAs and treated for indicated times with 1µM arsenic trioxide (Ars). B. Western blot from HT1080 cells stably expressing GFP-PML IV, transfected with indicated single siRNAs and treated for indicated times with arsenic trioxide (Ars). Asterisk indicates a non-specific band while the arrow shows the band of interest for FBXO9. C. Western blot looking at endogenous PML from HT1080 cells, transfected with indicated SMARTpool siRNAs and treated for indicated times with arsenic trioxide (Ars). D. RT-qPCR from total RNA extracts of untreated HT1080 cells stably expressing GFP-PML IV and transfected with indicated SMARTpool siRNAs, done on three biological replicates and normalized on POLRIIA gene expression (n=3).

b) Constitutive depletion of FBXO9

In order to reproduce the observed effects with a different approach, a short hairpin RNA (shRNA) was used to knockdown FBXO9. shRNAs have the advantage of making a more stable knockdown with, in theory, less off-target effects than siRNAs. One inconvenient though, is that the delivery has to be through a viral infection, plasmidic DNAs being poorly transfected in HT1080 GFP-PML IV cell line. Infection of HT1080 cells stably expressing GFP-PMLIV with a lentiviral transduced shRNA led to the stabilization or an increase in PML under arsenic trioxide treatment, an effect visible with at least two different shRNAs (Figure 55A). The most significant effect was again observed after 24 hours of arsenic trioxide treatment where a

stabilization of up to 60% can be observed with shFBXO9-2 (Figure 55B) while FBXO9 knockdown is around 80% (quantification data not shown). These results thus confirm that FBXO9 is down-regulating PML stability under arsenic trioxide treatment with an efficiency similar to that observed for RNF4.

Figure 55 : FBXO9 depletion has an effect on PML stability: shRNA approach. A. Western blot from HT1080 cells stably expressing GFP-PML IV, infected with indicated single shRNAs and treated for indicated times with 1µM arsenic trioxide (Ars). B. Quantification of PML profile from experiment described in A, based on pixel count and normalized to GAPDH.

c) Effect of FBXO9 depletion on PML, an indirect approach through CAND1

Cullin-associated NEDD8-dissociated protein 1 (CAND1) is an important protein acting as an exchange factor for F-Box proteins on SCF complexes (Figure 56A). This protein binds to CUL1 and RBX1 preventing the interaction between CUL1 and SKP1a. This allows to increase the turnover of SKP1a-F-Box dimers on the complex making the cell capable to react faster to various stimuli. Knocking down CAND1 should therefore stabilize already formed complexes and prevent the formation of new ones that would be, eventually, involved in arsenic trioxide stress induced response. Knocking down CAND1 in arsenic trioxide-treated HT1080 cells stably expressing GFP-PML IV induced a stabilization or an increase in hyperSUMOylated forms of PML after 24 hours treatment (Figure 56B). The effect of CAND1 depletion on endogenous PML was also checked in HT1080 cells (Figure 56C). An increase or stabilization of non-modified and hyperSUMOylated forms of endogenous PML was clearly observed in cells depleted in CAND1, at levels comparable to those obtained upon depletion of FBXO9 or

100

RNF4. Knockdown of PML with siPML was used here to confirm that the bands analyzed on Western blot are indeed specific for PML (Figure 56C).

Taken together, these data indicate that the degradation of PML, triggered by FBXO9 is mediated by a bona fide CAND1-regulated SCF complex.

Figure 56 : FBXO9 and CAND1 depletion have an effect on PML stability. A. Diagram summarizing the implication of CAND1 in SCF regulation as an F-BOX exchange factor. B. Western blot from HT1080 cells stably expressing GFP-PML IV, transfected with indicated SMARTpool siRNAs and treated for indicated times with 1µM arsenic trioxide (Ars). Knockdown of RNF4 and CAND1 were also verified through Western blot C. Western blot of endogenous PML from HT1080, transfected with indicated SMARTpool siRNAs and treated for indicated times with 1µM arsenic trioxide (Ars).

101

d) FBXO9 depletion effect on PML NBs using fluorescence microscopy

We next assessed the consequences of depleting FBXO9 on PML NB morphology using fluorescence imaging. To this aim, HT1080 cells stably expressing GFP-PML IV were transfected with siRNAs targeting FBXO9 or RNF4 (positive control) and were left untreated or treated with arsenic trioxide (1µM) for 1 hour and 24 hours. The cells were subsequently mounted on slides and observed under a fluorescence microscope. Morphology of PML NBs was then analyzed for three different parameters: the number per nucleus, the size per NB and the intensity per NB using ICY Bioimaging software (de Chaumont et al. 2012). In the absence of any treatment, depletion of RNF4 led to an increase in both the number of PML NBs per nucleus up to 16 (median, +14%, IQR 12-21, p-value<0.0001) and their intensity (median, +26%, IQR 1045-1948, p<0.0001) whereas their size remained unaffected (Figure 57). Suppression of FBXO9 induced similar changes in PML NB structure though in this case it increased their size (median, +25%, p<0.0001) and their intensity (median, +17%, IQR 909- 2006, p<0.0001) but it did not affect their number (Figure 57).

102

Figure 57 : FBXO9 and RNF4 depletion effect on PML Nuclear Bodies without arsenic trioxide treatment: fluorescence microscopy approach. Fluorescence data from HT1080 cells stably expressing GFP-PML IV and treated with indicated SMARTpool siRNAs. Graphs showing data of PML Nuclear Body morphology depending on three criteria. Box and wishers plots were used for Nuclear Body count per Nucleus (n=around 900 nuclei per condition) and Nuclear Body intensity (n=around 14000 bodies per condition). Nuclear Body surface median data is represented by a bar graph (n=around 14000 bodies per condition). P-values were calculated using Mann-Whitney test (****=p-value<0.0001).

After one hour arsenic trioxide treatment, as expected, the number of PML NBs in control cells strongly increased, from a median of 14 to 18 bodies per nuclei (+28%, IQR 12- 24) a similar increase in their intensity (+25%, IQR 1232-2136) and size (+20%) was observed (Figure 58). This phenomenon is due to the hyperSUMOylation induced by arsenic trioxide. Suppression of FBXO9 or RNF4 did not affect the number and the size of the PML NBs when compared to the control. On the contrary, cells depleted for FBXO9 exhibit a decrease in their

103

intensity (-15%, IQR 704-2609, p<0.0001) whereas suppression of RNF4 leads to more intense PML NBs (+9%, IQR 1232-2136, p<0.0001) (Figure 58).

Figure 58 : FBXO9 and RNF4 depletion effect on PML Nuclear Bodies after one hour arsenic trioxide treatment: fluorescence microscopy approach. Fluorescence data from HT1080 cells stably expressing GFP-PML IV, transfected with indicated SMARTpool siRNAs and treated for an hour with 1µM arsenic trioxide. Graphs showing data of PML Nuclear Body morphology depending on three criteria. Box and wishers plots were used for Nuclear Body count per Nucleus (n=around 900 nuclei per condition) and Nuclear Body intensity (n=around 17000 bodies per condition). Nuclear Body surface median data is represented by a bar graph (n= around 17000 bodies per condition). P-values were calculated using Mann-Whitney test (****=p-value<0.0001).

Finally, after 24 hours arsenic trioxide treatment, control cells exhibited a strong decrease in the median PML NBs intensity (-27%, IQR 712-1703) and count, from a median of 18 to 7 bodies per nucleus (-62%, IQR 5-10), a phenomenon due to PML degradation (Figure 59). Cells depleted in FBXO9 or RNF4 displayed a slightly lower PML NBs count than control cells with a median of 6 bodies per nuclei compared to 7 for control cells (-15%, IQR siFBXO9

104

5-8, IQR siRNF4 5-9, p<0.0001). However, both the size and the intensity of PML NBs were significantly higher: +37% increase in size for both RNF4 and FBXO9-depleted cells (p<0.0001) and +104% (IQR 1464-3217) increase in intensity for FBXO9 and +44% (IQR 989- 2386) increased intensity for RNF4-depleted cells compared to control cells (IQR 712-1703, p<0.0001) (Figure 59).

Figure 59 : FBXO9 and RNF4 depletion effect on PML Nuclear Bodies after 24 hours arsenic trioxide treatment: fluorescence microscopy approach. Fluorescence data from HT1080 cells stably expressing GFP-PML IV, transfected with indicated SMARTpool siRNAs and treated 24 hours with 1µM arsenic trioxide. Graphs showing data of PML Nuclear Body morphology depending on three criteria. Box and wishers plots were used for Nuclear Body count per Nucleus (n=around 830 nuclei per condition) and Nuclear Body intensity (n=around 5900 bodies per condition). Nuclear Body surface median data is represented by a bar graph (n=around 5900 bodies per condition). P-values were calculated using Mann-Whitney test (****=p-value<0.0001).

105

These experiments have been done in biological triplicates and data were pooled for the statistical analysis, meaning that between 750 and 1000 cells were analyzed in each conditions. Given that a large number of cells were analyzed, although some of the numerical differences observed were not outstanding, a highly significant statistical difference was observed with very low p-values.

To summarize, FBXO9 is controlling, like SKP1a and RBX1, PML stability under arsenic trioxide treatment as observed both by Western blot as well as by immunofluorescence and the effect is similar to that observed for RNF4. Gain of function experiments were attempted but results were inconclusive due to a very low plasmid transfection efficiency on HT1080 expressing GFP-PML IV cell line.

12) Impact of FBXO9 on the half-life of PML

We then analyzed the effect of FBXO9 overexpression on the half-life of PML. 293T cells were transfected with GFP-PML IV with or without FLAG-FBXO9 and then treated with cycloheximide, an inhibitor of protein biosynthesis, for different time intervals (Figure 60A). The quantification of PML on Western blot profiles was normalized to GAPDH at a given cycloheximide treatment time point. Under these conditions, GFP-PML IV was degraded faster when co-expressed with FBXO9 (Figure 60A). As a consequence, only 20% of PML remained after 8h of treatment, whereas 41% remained when only GFP-PML IV is overexpressed (Figure 60B).

These data indicate that FBXO9 significantly shortens the half-life of PML. Surprisingly, a reciprocal effect could be observed in that overexpression of GFP-PML IV leads to the faster degradation of FLAG-FBXO9 (Figure 60A). Would this effect be non- transcriptional (to be tested), this could suggest that the presence of PML stabilizes FBXO9 as shown previously for SKP1a (Figure 54A).

106

Figure 60 : FBXO9 shortens the half-life of PML. A. Western blot from 293T cells, transfected with indicated plasmids and treated for 2 to 8 hours with 100µg/mL cycloheximide (CHX). B. Quantification of PML profile from experiment described in A, based on pixel count and normalized to GAPDH for each time point.

13) SCFFBXO9 ubiquitinates PML in vitro

As a logical step in the functional analysis of an E3 ubiquitin ligase, we determined whether the SCFFBXO9 complex was able to specifically ubiquitinate PML. To answer this question, we initiated a collaboration with F. Bassermann’s team at the Technical University of Munich who had already studied the activity of the SCFFBXO9 complex on the other substrate TEL2 and TTI1 (Fernández-Sáiz et al. 2013). We overexpressed FLAG-PML IV in 293T cells treated the cells for one hour with arsenic trioxide or vehicle, and purified PML using its FLAG tag. The complex SCFFBXO9 was produced in 5B insect cells and purified in a similar way. All these purified components were then used in an in vitro ubiquitination experiment along with commercial E1, E2 enzymes and ubiquitin. The reaction was allowed to run for the indicated time, and then stopped by adding Laemmli to the reaction and the results were visualized by Western blot (Figure 61A and 61B). In the vehicle treated cells, PML was found to be ubiquitinated rapidly, within 20 minutes, while the non-ubiquitinated form of PML was decreasing (Figure 61A). When PML had been treated for one hour with arsenic trioxide prior to in vitro ubiquitination assay, we observed it became ubiquitinated significantly faster than in control cells and concomitantly, that the non-ubiquitinated form of PML was decreasing rapidly (Figure 61B). As a control, incubation of E1, E2 enzymes and ubiquitin, failed to lead to PML ubiquitination unless FBXO9 was added. Moreover, no ubiquitination of PML could be detected substituting FBXO9 for FBXO25, another F-Box protein known to work efficiently in this system on its specific substrate (Baumann et al. 2014), thus emphasizing FBXO9 specificity for PML ubiquitination (Figure 61A).

107

This experiment was reproduced using an in-vivo approach. To this aim, FLAG-PML V was transfected along with non-tagged FBXO9 and/or HA-Ubiquitin in 293T cells. Immunoprecipitated PML was then observed by Western blot. Even though many technical difficulties arose while performing this type of experiment, such as vector expression efficiency, lysing buffer optimization or aggregating beads, a faint smear was detected when PML V and Ubiquitin were overexpressed that became more pronounced upon co-expression of FBXO9 (Figure 61C). However, these results are to be taken with caution since the in vivo ubiquitination assays are known to be technically challenging for certain substrates and the signal obtained is faint and not clearly defined. Therefore, it is difficult to completely rule out the possibility of an artefact for this in vivo experiment.

108

Figure 61 : SCFFBXO9 is able to specifically ubiquitinate PML: ubiquitination assays approach. A. Western blot from in vitro ubiquitination experiment using purified FLAG-PML from 293T cells and purified SCF complexes (FBXO9 or FBXO25) from insect cells. Reactions were carried out in the presence of the indicated proteins and a reaction mix containing E1, E2s, ubiquitin as well as necessary catalyzers. B. Same experiment as described in A but PML was either treated with 1µM arsenic trioxide (Ars) or with vehicle for 1 hour prior to in vitro testing. C. Western blot from 293T cells, transfected with indicated plasmids followed by PML V immunoprecipitation (IP-FLAG). Whole cell extract can be found on the right blot (WCE).

109

In summary, these data indicate that the SCFFBXO9 complex efficiently and specifically ubiquitinates PML in vitro.

14) An attempt to localize the FBXO9 degron on PML

It is well established that F-Box proteins recognize their substrates through a small stretch of amino acids, usually modified post-translationally mostly through phosphorylation. This small stretch of residues is called a “degron” (Skaar et al. 2013). In the case of SCFFBXO9, the F-Box recognizes on TEL2 and TTI1, a degron phosphorylated by Casein Kinase 2 (CK2) (Fernández- Sáiz et al. 2013). Since PML is a big and structurally complex protein, a map of all known post- translational modifications and kinases was put together to clarify the following approach (Figure 62).

Figure 62 : PML post-translational modification map. Diagram representing all known, published post-translational modification on PML isoforms I to VI. Highlighted SUMO sites correspond to mutated sites in PML3KR mutant. RING, B-Box and Coiled-Coil domains form together the RBCC domain found in TRIM family proteins. NLS= Nuclear Localization Signal.

110

In order to identify the degron, a small immunoprecipitation screen of different PML mutants, which we designed according to previously accumulated knowledge, was initiated. These mutants are lacking either internal domains, like part of the RBCC domain, or parts of N or C-terminus (Figure 63A). Eight different FLAG-tagged PML mutants were tested and only seven are shown here. These constructs were transfected in 293T cells along with non-tagged FBXO9 followed by PML mutant immunoprecipitation via FLAG tags and detection of FBXO9 by Western blot (Figure 63B). All our mutants were correctly expressed however, unexpectedly, all of them were interacting to some extent with FBXO9. The strongest mutant interaction was observed with the ΔRBCC mutant, followed by RBCC, ΔN+C, Δ320-371 and PML VII. The weakest interaction was detected with PML IV, ΔRWGD, ΔC-terminus and ΔRING mutants (Figure 63B). Strong interactions with mutants that only shared 30 amino acids (ΔRBCC and RBCC mutants) was detected suggesting a possible interaction site on this stretch. However, a mutant that lacked this small sequence also interacted with FBXO9 rather strongly (Δ320-371). We concluded that the PML-FBXO9 interaction might occur through several different regions on PML. Alternatively, these experiments will have to be performed in cells depleted for endogenous PML using shRNA or in Mouse Embryonic Fibroblasts (MEFs) KO for PML (PML-/-) (Wang, Delva, et al. 1998), to prevent possible multimerization of the different PML mutants with the endogenous PML protein.

111

Figure 63 : FBXO9 degron identification on PML: immunoprecipitation screen approach. A. PML mutant constructs used in the immunoprecipitation screen. B. Western blot from 293T cells, transfected with indicated constructs described in A followed by mutant immunoprecipitation (IP-FLAG). Whole cell extract can be found on the right blot (WCE).

15) Kinase mini-screen to localize the FBXO9 degron on PML

It has been previously shown that SCF substrate-specific interactions and specifically interactions implicating, FBXO9 with other substrates, are dependent on phosphorylation (Fernández-Sáiz et al. 2013). In order to check if the interaction between FBXO9 and PML requires a prior phosphorylation event, a small siRNA screen of all the kinases known to phosphorylate PML was performed (Figure 64). HT1080 cells overexpressing GFP-PML IV were transfected with siRNAs targeting the indicated kinase and then treated with arsenic trioxide (Figure 64). This screen led to the identification of three candidate kinases: Cyclin- Dependent Kinase 1 (CDK1), Mitogen-activated Protein Kinase 1 (MAPK1, also known as ERK2) and CK2, whose depletion induced a significant increase or stabilization of PML compared to the control siRNA (Figure 64). MAPK1, also known as ERK2 phosphorylates PML allowing PIN-1 to target PML for phosphorylation dependent degradation (Lim et al. 2011). CDK1 is already known to phosphorylate PML under hypoxia conditions leading to its degradation by a Cullin-3 based complex involving the E3 ligase KLHL20 and the isomerase

112

PIN-1 (Yuan et al. 2011). CK2 is known to phosphorylate PML under different types of stimulus, like UV, oxidative stress or osmotic stress, leading to its proteasomal degradation through an unknown pathway (Scaglioni et al. 2006).

In summary, whereas the degron within PML that mediates the recruitment of FBXO9 remains unknown, we have here identified three promising candidate kinases, CDK1, CK2 and MAPK1, and future effort will be dedicated to the clarification of their role in the SCFFBXO9 mediated degradation of PML.

Figure 64 : Mini kinase screen to identify FBXO9 degron on PML. Western blot from HT1080 cells stably expressing GFP-PML IV, transfected with indicated SMARTpool siRNAs and later treated for indicated time with 1µM arsenic trioxide (Ars).

113

16) Search for physiological stimuli leading to SCFFBXO9-induced PML degradation

Whereas the SCFFBXO9 complex induces PML degradation in response to arsenic trioxide, other biological stimuli remain to be identified. We thus tested different stimuli that are known to activating CDK1 activity such as cobalt to induce hypoxia or CK2 activity such as UV irradiation or oxidative stress. A drug preventing CK2 activity, 4,5,6,7-Tetrabromo-2- azabenzimidazole (TBB) (Sarno et al. 2001), was used in cells treated with H2O2 or arsenic trioxide. Under these conditions, a transient stabilization of PML, after one hour H2O2 treatment was observed when compared to untreated HT1080 cells (Figure 65A) but not under arsenic trioxide-induced stress (Figure 65B). No significant results were obtained using other stimuli such as hypoxia or UV radiation (data not shown).

Figure 65 : Test of different types of stress on PML stability in the presence of CK2 inhibitor. A. HT1080 cells were treated with 100mM H2O2 to induce oxidative stress and 50µM CK2 inhibitor TBB, for indicated times. Cells were subsequently lysed and PML stability observed through Western blot. B. Same experiment as described in A except that 1µM arsenic trioxide was used instead of H2O2.

Another reported external stimulus which causes CK2 activation is osmotic shock (Scaglioni et al. 2006). To test this condition, 293T cells were treated with 0.2M NaCl for a given time and PML stability followed by Western blot. Phosphorylated P38 (p-P38), which activates CK2, was used as an indicator for CK2 activation in response to osmotic stress (Figure 66). As expected, we observed an increase in phosphorylated P38 (p-P38) over time along with a slow decrease of endogenous PML in HT1080 cells (Figure 66A). In order to determine if

114

this degradation was linked to SCFFBXO9 activity, the same experiment was performed in HT1080 cells stably expressing GFP-PML IV previously infected with shRNAs targeting FBXO9. In cells depleted in FBXO9, a stabilization of all PML bands was observed compared to control cells. Osmotic shock could thus act as an additional stimulus for PML degradation mediated by SCFFBXO9 and CK2 activity (Figure 66B). These preliminary results, yet to be confirmed and should be confirmed, might open new interesting avenues toward the elucidation of regulatory mechanisms underlying PML stability in response to stress.

Figure 66 : Osmotic shock causes PML degradation that is dependent at least in part on SCFFBXO9. A. HT1080 cells were treated with 0.2M NaCl to induce osmotic shock stress for indicated times. Cells were subsequently lysed and PML stability observed through Western blot. CK2 activation was monitored through p-P38 accumulation. B. HT1080 GFP-PML IV cells infected with indicated shRNAs for 72 hours and treated with 0.2M NaCl to induce osmotic shock stress for indicated times. Cells were subsequently lysed and PML stability as well as FBXO9 knockdown observed by Western blot.

115

17) Possible links to diseases

A large number of studies have established the tumor suppressor function of PML both in mouse models and human cancers (Salomoni & Pandolfi 2002; Guan & Kao 2015). Notably down-regulating of the PML protein is frequently found in cancer cells. Deciphering the mechanisms that control PML stability are thus of key interest for mechanistic studies and translational purposes.

To unravel possible alteration in PML’s expression level and the different members of the SCFFBXO9 complex in cancer, we performed a pilot study aimed at quantifying the expression of the corresponding mRNAs by RT-qPCR in 72 cancer cell lines derived from various tissues including, liver, breast, lung and prostate (Annex 5). A total of 21 genes were analyzed, including members of the SCFFBXO9 complex along with related kinases. Were also included in the study SP100 known to interact with PML at the protein level, or known targets of SCFFBXO9 such as TEL2 and TTI1. Results were normalized with respect to the expression of the three housekeeping genes TRIM 44, β-Actin and RPLP0 (de Jonge et al. 2007) and the two PLC/PRF5 and HT1080 cell lines as an internal reference. Normalization was done using 2- ΔΔCT method (Livak & Schmittgen 2001). These data allowed to generate a heat map displaying mRNA abundance (log2) for each of the targets described (Figure 67). In this analysis, HUH13 cell line, which is a mouse cell line, was used as a voluntary outlier (control) to verify specificity. Whereas the level of PML transcripts was expectedly low in a significant (54%) number of cell lines, FBXO9 transcripts showed high level of expression in more than 85% of cases.

Thus, it is tempting to speculate that, in tumor cells, the PML protein could be down regulated in two ways, either directly at the transcriptional level or via FBXO9-mediated proteasomal degradation, in the subset of cases expressing high levels of PML transcripts.

116

Figure 67 : Heat map of mRNA expression of 21 genes in 72 different cell lines. Heat map based on total mRNA extract, all samples were validated through Bioanalyser. Normalization based on TRIM 44, β-Actin and RPLP0. CSNK2-A1, A2, A3, B corresponds to subunits forming CK2.

We also looked for correlation between gene expression levels using Pearson's coefficient (Figure 68). As expected, the expression of the members of the SCFFBXO9 complex are positively correlating together, while PML expression correlates with SP100 and the SUMO E2 enzyme UBE2I (UBC9). CSNK2A1 that encodes for the catalytic subunit of CK2, also displays a strong positive correlation with SCFFBXO9 targets, TELO2 and TTI1. Interestingly, PML display a rather strong correlation with the other catalytic CSNK2A3 subunit of CK2. There were not any negative correlations between the expressions of the genes analyzed in our series (Figure 68).

117

Figure 68 : Correlation matrix of target gene expression. Correlation matrix based on Pearson's correlation coefficient. Analysis of the matrix and rendering based on literature (Friendly & Friendly 2002).

Finally, clusters were generated integrating the levels of the 19 out of 21 transcripts using a Multiple Factor Analysis (MFA); this will give us the data variability dispersion based on PML and FBXO9 expression. Two significant hierarchical clusters were obtained, where 99% of the data variability was explained (Figure 69). The first cluster (in black) is characterized by a lower PML and UBE2I expression and the second one (in red) exhibits opposite characteristics. Interestingly, the second cluster was significantly enriched for Non- Small Lung Cancer cell line (CALU1, GILI5C, SK-MES1), 3/6 out of 7/72 included in the study, and for head and neck carcinoma cell lines (BB-48 and SCC15), 2/6 out of 6/72 (Figure 69). Remarkably, it was reported that PML is often completely or partially lost in Non-Small Cell Lung Cancer at the protein level (Gurrieri, Nafa, et al. 2004; Koken et al. 1995). A study was even able to show, in a murine model of lung cancer, that CK2 pharmacological inhibition was increasing PML activity as a tumor suppressor (Scaglioni et al. 2006). Moreover, this study also shows an inverse correlation between PML protein abundance and CK2 activity in human normal lung and lung-cancer-derived cell lines. Thus, Non-Small Lung Cancer cell lines showing high or normal levels of PML transcripts but reduced amount of PML protein may provide an amenable system to investigate possible involvement of the SCFFBXO9 complex in regulating PML stability.

118

Figure 69 : Hierarchical clustering of studied cell lines based on gene expression. Multiple Factor Analysis (MFA) based on PML and FBXO9 normalized expression followed by hierarchical clustering.

119

Discussion and perspectives

The initial project was relying on a genome wide screening of PML Nuclear Body (PML-NB) morphology, in order to identify new regulators of these nuclear structures. The list of candidates obtained by the two sequential screenings contains enzymes of the SUMO pathway already known to have an effect on PML-NB. Target validation, on different cell lines, helped us to confirm that the screening results identified some bona fide PML NB modifiers.

1) SKP1a and RBX1 are members of an ubiquitination complex involved in the degradation of PML

This study focused on the function of different candidates whose depletion led to a “hyperSUMOylation like” phenotype (Class III) (Figure 40). In fact, this phenotype exhibiting an increase in PML-NB count and size, implies that these candidates could be part of a SUMO dependent Ubiquitin ligase class -like RNF4- or more largely as described in this study, players of the ubiquitination pathway controlling PML stability. Moreover, the two candidates with the strongest phenotypes, SKP1a and RBX1, are proteins biologically known for being part of an ubiquitination complex called SKP-Cullin-F-Box containing complex (SCF) (Skaar et al. 2014).

Our experiments showed that targeted depletion of SKP1a and RBX1 induced an increase of PML NB count and fluorescent intensity, as well as a stabilization of PML that is not caused by transcriptional upregulation (Figures 43 and 44). Under arsenic trioxide induced stress -used in the treatment of APL- a strong stabilization is observed of all modified PML forms in cells depleted in SKP1a and RBX1 (Figure 44). The cooperative effect of SKP1a and RBX1, coupled to co-immunoprecipitation data (Figure 47), suggests that they are likely involved in PML regulation via a single complex.

In accordance with the hypothesis that SKP1a and RBX1 are involved in PML degradation, overexpression of these two proteins caused the disappearance of PML and PML NBs (Figure 46). Under MG132 treatment, a proteasome inhibitor, there are no more significant differences on PML protein abundance on Western blot between cells overexpressing GFP-PML IV and one of the two candidates compared to control cells expressing only GFP-PML IV (Figure 46). To strengthen this point, experiments could be

120

performed by infecting the cell line HT1080 GFP-PML IV, with plasmids expressing SKP1a and RBX1 to decrease artefactual risks involved in co-overexpression. Finally, results obtained through immunofluorescence microscopy also confirmed the effect of SKP1a and RBX1 overexpression on PML NBs. Indeed, the immunofluorescence approach using the stable cell line GFP-PML IV allowed us to obtain quantitative results. Results that confirmed what was observed by Western Blot, even though the transfection efficiency was quite low.

Co-immunoprecipitation experiments suggest that SKP1a and RBX1 both interact with PML (Figure 47). However, throughout the study, the quantity of RBX1 co- immunoprecipitating with PML appeared consistently higher than of SKP1a. This might be due to the fact that there are proportionally more Cullin Ring Ligases (CRLs) complexes involving RBX1 (almost all of CRLs), than SKP1a (only CRL1 and CRL7) (Skaar et al. 2013; Yoshida et al. 2011) (Figure 27). There is indeed at least one published CRL complex interacting with PML: The Cullin3-KLHL20 complex (Yuan et al. 2011). Overall, our data support the interaction of both proteins with PML, probably through an SCF complex.

The involvement of other CRLs was also addressed. The primary screen identified members of the CRL2 and CRL5. Cells depleted in Elongin B and C as well as Cullin-2 and SOCS-Box proteins (all members of CRL2 and CRL5) were exhibiting lower PML-NB count. The depletion of these proteins causing a downregulation of PML-NB, this complex could be involved in an indirect effect on PML stability by affecting the stability of PML’s regulator. CRL2 and 5 being involved in occurrence and progression of tumors, it would be interesting to study their implication in PML NB maintenance (Siwei Wang et al. 2016). However, our focus remained on SKP1a and RBX1, since they are more likely to have a direct effect on PML stability. Both candidates are part of a known Cullin Ring Ligase complex: the SCF complex family (or CRL1 on SCF complex). These complexes share a common structure composed of a RING containing protein (RBX1) recruiting the E2 ubiquitin conjugating enzyme, an adaptor protein (SKP1a), a scaffold protein from the Cullin family, and an F-Box protein which will bring the substrate specificity to the complex (Figure 70 ). This project led us to identify the other putative components of the complex involved on PML degradation as well as the “physiological” context in which this complex is activated.

121

Figure 70 : Diagram of the SKP1-Cullin-F-Box (SCF) containing complex (Morgan 2007).

2) The Cullin-1 is involved in SCF complexes

Out of the eight Cullin proteins engaged in CRLs formation, Cullin-1 (CUL1) is clearly described in the literature as the only one involved in SCF formation (Skaar et al. 2013; Lee & Diehl 2013; Willems et al. 2004). Experiments of inactivation -through siRNAs or dominant negative mutants- or overexpression of this protein did not give any conclusive results so far, regarding PML NBs stability (data not shown). It is possible that other isoforms of Cullin -like Cullin-7 (CUL7)- could be used as a scaffold to replace the missing CUL1. CUL7 is already described as an atypical CRL involved in growth control, binding both SKP1 and RBX1 (Sarikas et al. 2008). CUL7 could therefore have been a candidate. However, results from our primary screen did not show any variations in PML NBs morphology in cells depleted of CUL7. However, CUL3, CUL4B and CUL5 depletions showed moderate effects in the first screen. The study focused on some of these candidates since one of them, CUL3, was recently implicated in PML degradation under hypoxic conditions found in prostate cancer progression (Yuan et al. 2011). So far, there is no evidence that CUL1 is involved in PML stability notwithstanding that CUL1 is considered as the main Cullin protein interacting with both RBX1 and SKP1a. In addition, FBXO9 was shown to interact with a CRL complex involving CUL1 as the scaffold protein to degrade TEL2 and TTI1 (Fernández-Sáiz et al. 2013). Taking this into consideration, CUL1 is a reasonable candidate for the complex that recognizes PML.

122

3) FBXO9 specifically interacts with PML

Substrate specificity is brought to the SCF complex through the F-Box protein. As previously mentioned in the introduction, there are 69 different F-Box proteins encoded in the human genome (Jin et al. 2004) and at least one of them recognizes PML as its substrate. Only FBXO9 was identified, thanks to a co-immunoprecipitation screen of PML with tagged F-Box proteins (Figure 48). This interaction was verified through co-immunoprecipitation of PML or FBXO9 and even if it is weaker than PML-RBX1 interaction (likely due to the same reasons as suggested for SKP1a), it is highly reproducible. Moreover, different protein tags, GFP or FLAG -or none at all- were used to make sure that the interaction observed was not due to tags. To further confirm that this interaction was genuine, it was also showed with endogenous proteins (Figure 50). In addition, a small part of FBXO9 was bound to PML probably because no appropriate stimuli were used to induce SCF complex formation and activation. This situation would also explain why FBXO9 did not show up in the first siRNA screen. Other F-Box proteins were detected in the siRNA genome wide screen, causing mild effects on PML-NB morphology. However, none of them co-immunoprecipitated PML indicating that, under our experimental conditions, there is no detectable interaction. In literature, another F-Box protein, FBXO3, was described as potentially interacting with PML because of its localization in PML NBs. However, this F-Box protein known to interact with HIPK2 (Shima et al. 2008) was not detected in the screening.

4) FBXO9 is involved in PML stability

Both fluorescence microscopy and Western blotting showed that depletion of FBXO9 in cells treated with arsenic trioxide caused an increase or stabilization of PML and PML NBs (number and size after 24 hours treatment) (Figures 54-59). All the effects observed with FBXO9 depletion were not very intense compared to RNF4 knockdown, but very reproducible. Even though RTqPCR method showed an increase in PML mRNA when cells were treated with siFBXO9, no significant increase in the amount of PML protein before arsenic trioxide treatment was observed. Keeping in mind that an increase in mRNA levels does not necessarily translate into an increase in protein abundance, and the implication of FBXO9 in regulating protein stability, it is likely that the effect observed at 24 hours is mostly due to post- translational regulation (Vogel & Marcotte 2012). To exacerbate the effects observed with PML, the oncoprotein PML-RARα -which is extremely responsive to arsenic trioxide- could be

123

used to show if a depletion of FBXO9 would play a role on its stability (Zhang et al. 2010). Western blots also revealed that cells depleted in SKP1 displayed a decrease in FBXO9 protein level, thus confirming SKP1 stabilization function of F-Box proteins previously described (Yoshida et al. 2011). An indirect approach to study FBXO9 depletion was through CAND1 (Figure 56). Cells depleted in this protein displayed a stabilization of PML under arsenic stress. CAND1 is stimulating F-Box turnover on SCF complexes, and helping to maintain the “SCF landscape” in the cell (Pierce et al. 2013; Wu et al. 2013). This “landscape” describes the types of active SCF complex in a cell under a specific condition. The knockdown of CAND1 can have two opposite effects: (i) it can either stabilize existing complexes or (ii) prevent the formation of new ones. In addition, it leads to deNEDDylation of CUL1 (Lydeard et al. 2013). In cells, F-Box proteins detected as heterodimers (with SKP1) are considered in a “free” state, in contrast to the “engaged” state translating their assembly in an SCF complex. The effect of CAND1 will depend on the ratio of already formed SCFFBXO9 complexes (Pierce et al. 2013; Wu et al. 2013). However, the quantity of these complexes and their location are still poorly understood except for specific cases. Moreover, depending on stimuli received by the cells, the SCF landscape will change. Some research teams have already started to study the abundance of some SCF complexes under specific stresses, such as arsenic trioxide. The initial screen did not show any variation in PML NBs in cells depleted for members of the NEDDylation pathway or CAND1. Further analysis of the screen revealed that CAND2 depleted cells exhibited an increased PML NBs count and fluorescent intensity. Information on CAND2 are scarce, but it is known to interact with CUL1 and to display the same function as CAND1 (Pierce et al. 2013). This protein is a regulator of myogenesis, for example by causing the dissociation of the SCF- myogenin complex (Shiraishi et al. 2007).

Arsenic trioxide (ATO, AS2O3) is used to force the degradation of PML, although it is not the main physiological stimulus activating the SCF complex. We have shown that FBXO9 is interacting with PML, and influences its stability under arsenic trioxide induced stress. PML binds arsenic through a dicysteine motif (C212/C213) causing a conformational change leading to HyperSUMOylation of PML (Jeanne et al. 2010). PML is not the only protein exhibiting such a motif. Therefore, it would be interesting to look at the potential side effects caused by targeting other arsenic binding proteins. There are around seventy proteins on the human genome with at least one cysteine residue available for arsenic binding raising the question of their impact on cell function (B. Chen et al. 2015).

124

5) Is PML degradation dependent on SUMOylation?

PML being one of the major substrates of SUMOylation, along with RANGAP1, we wondered if the interaction between PML and FBXO9 was dependent on SUMOylation. To answer this, PML3KR, the triple mutant defective for SUMOylation, was used (Zhong et al. 2000). No significant differences in the interaction of FBXO9 with PML3KR and the wild type were observed (Figure 53). We hypothesized that PML mutants might have dimerized with the wild type PML, biasing final outcome. In fact, PML being a TRIM protein, it can form homo and heterodimers through their Coiled-Coil domain. These proteins can also form higher order interactions such as trimers of hexamers (Jensen et al. 2001). To further test the implication of SUMOylation in PML-FBXO9 interaction, PML3KR mutant is currently being used for in- vitro ubiquitination experiment. The fact that FBXO9 presents two predicted SUMOylation sites and one SUMO Interacting Motif (Figure 48) -found using GPS-SUMO software (Zhao et al. 2014)- suggests that it might be capable to efficiently interact with PML. However, FBXO9 was not identified in recent SUMO proteomics studies as a SUMO target (Hendriks & Vertegaal 2016). To test FBXO9 capacity to interact with SUMO, a GST-SUMO pull down will be performed as it was described for ARKADIA (Erker et al. 2013).

6) SCFFBXO9 specifically ubiquitinates PML

The specific role played by the SCFFBXO9 complex on PML stability was a major question raised through our research. GFP-PML IV and FBXO9 were overexpressed in 293T cells, treated with cycloheximide to prevent protein biosynthesis, and PML stability was assessed through Western blotting. Our work clearly showed that, GFP-PML IV in these circumstances was degraded faster when co-expressed with FBXO9 (Figure 60). A half-life measurement experiment using an HT1080 cell line Knocked Out (KO) for FBXO9, created using the CRISPR/Cas9 technology (Ceasar et al. 2016) was performed under arsenic trioxide conditions. Unfortunately, no significant outcomes were observed for this experiment.

The in vitro ubiquitination experiment showed a very strong and clear specific effect of SCFFBXO9 and no ubiquitination was observed when using another functional F-Box protein (FBXO25) (Figure 61A). SCFFBXO25 degrades the prosurvival protein HCLS1-associated protein X-1 (HAX-1) after its phosphorylation by PRKCD under apoptotic stress. As a result, SCFFBXO25 is involved in apoptotic regulation in mantle cell lymphoma (Baumann et al. 2014).

125

Regarding the in-vitro experiments with PML treated for one hour with arsenic trioxide, a strong ubiquitination effect was unexpectedly seen, even after a few seconds (data not shown). PML is hyperSUMOylated within an hour, and RNF4 along with ARKADIA are already having an effect on PML stability (Tatham et al. 2008; Erker et al. 2013). The rapidity of the ubiquitination effect observed leads us to think that RNF4 and/or ARKADIA protein could have been co-immunoprecipitated along with SUMOylated PML causing a faster response. However, this would have implied that some PML would already have been ubiquitinated when it was purified from 293T cells, which was not detected in the experiments (Figure 61B).

7) PML’s degron

The degron is a small motif responsible for the specific interaction of the F-Box with its substrate, in this case PML. PML mutants were created to investigate the possible involvement of structural domains, and to locate the degron on PML. Unfortunately, all mutants were interacting to some degree with FBXO9 (Figure 63). One hypothesis could be that mutants were forming dimers or tetramers through their RBCC domain with endogenous wild type PML since PML wild type PML is known to form such structures (Lallemand-Breitenbach et al. 2010; Jensen et al. 2001). Different strategies were carried out without success using Pml-/- MEF or 3T3 cells. An alternative approach would consist in using an shRNA against PML targeting the 3’UTR of PML mRNA, and overexpressing the mutants in 293T cells -as it was described in a paper studying PML isoforms (Hands et al. 2014). PML mutants would later be immunoprecipitated, and their possible interaction with FBXO9 analyzed.

Another hypothesis infers that PML mutant structural alteration might be exposing hidden interaction and/or modification sites or that PML contains multiples degrons recognized by FBXO9. This last hypothesis could be supported by the fact that PML contains a large number of post-translational modification sites such as phosphorylation, SUMOylation or acetylation sites, that could be used to interact with FBXO9 (Guan & Kao 2015). Many examples of F-Box proteins capable of recognizing different substrates with similar degron sequence are described in literature, but not multiple degron sequences within the same substrate (Skaar et al. 2013). However, a recent study showed that FBXO9 was capable of interacting with two distinct domains of Peroxisome Proliferator-Activated Receptor gamma (PPARγ) (Lee et al. 2016).

126

Over the years, the definition of a degron evolved with now the emerging concept of tripartite degron motifs. This model includes (i) the primary degron motif (peptide sequence) recognized by the E3 ubiquitin ligase, (ii) single or multiple neighboring lysine residues for poly-ubiquitination forming the secondary degron site and (iii) to initiate unfolding at the 26S proteasome of the substrate, an intrinsically disordered regions forming the tertiary degron is preferable (Guharoy et al. 2016). According to this model, it would be unlikely that the degron is contained in the RBCC domain of PML, since it is a highly structured area without any nearby known ubiquitination sites.

Another approach to identify the degron region would be to use the fusion oncoprotein PML-RARα, lacking a parts of PML’s C-terminus in an in vitro ubiquitination experiment. The PML moiety of PML-RARα is truncated at residue K394 in type A, and A552 in type B APL. Therefore, knowing if SCFFBXO9 complex is capable of ubiquitinating this oncoprotein would give some clues about the possible location of PML’s degron.

F-Box proteins can recognize different substrates by recognizing a degron with similar characteristics. For example, β-TrCP can bind β-catenin (DpSGIHpS), IκBα (DpSGLDpS), p105 (DpSGVETpS), and CDC25B (DpSGFCLDpS) (Kanarek et al. 2010). FBXO9 already has two known substrates, TEL2 and TTI1, with known phosphodegron. However, sequence alignment did not reveal significant clues as to where the degron could be located on PML. Therefore, the identification of the kinases relevant for FBXO9-PML interaction might offer a better clue.

8) Kinases phosphorylating PML’s degron

Information available in literature on FBXO9 and SCF complex activity report that FBXO9 recognizes a canonical phosphodegron on its substrate, suggesting that a kinase might need to phosphorylate PML beforehand (Fernández-Sáiz et al. 2013). Since the degron was unsuccessfully revealed using PML mutants, the next step was to find kinases capable of phosphorylating it. However, there are more than 80 potential phosphorylation sites identified in silico on PML (Phosphonet 2016), of which about 40 were confirmed experimentally and published (Guan & Kao 2015). We checked the effect on PML stability of the knockdown of kinases commonly associated with this protein. CDK1, MAPK1, and CK2 were the best candidates with strong stabilization of PML bands observed through Western blot, and under arsenic trioxide stress. These kinases might not have necessarily came up on our initial screen,

127

since we did not use arsenic trioxide or any stimulus. However, cells depleted in CSNK2A1 - the catalytic subunit of CK2- displayed in our initial screening a tendency toward an increase of PML NB count and fluorescent intensity. We also tested the serine/threonine-protein kinase ICK, since it maps in close proximity to FBXO9 gene on chromosome 11 (Kent et al. 2002), suggesting that they might be co-regulated and/or acting in the same pathway. However, no change was observed when cells were depleted in ICK (data not shown). MAPK1, CDK1 and CK2 were shown to have the most significant effect on PML stability with and without arsenic trioxide treatment (Figure 64).

PML phosphorylation by MAPK1, also known as Extracellular signal Regulated Kinase 2 (ERK2) and its implications are not yet fully understood. The ERK pathway is usually involved in cell survival, proliferation and differentiation but in the case of ERK2, it is also involved in apoptotic signaling (Cobb 1999). Under Epithelial Growth Factor (EGF) stimulation PML is phosphorylated at S403, S505 and S527 by ERK2, which trigger the interaction of PIN-1 with PML leading to its degradation by an unknown E3 ligase (Lim et al. 2011; Reineke et al. 2008). In vitro, ERK2 can phosphorylate PML at multiple sites, both in the N and C-terminal domains (T28, S36, S38, S40, S527 and S530). Interestingly, Arsenic trioxide treatment induces PML phosphorylation in an ERK2-dependent manner and is associated to enhanced SUMOylation (Reineke & Kao 2009). PML constructs mutated for residues T28, S36, S38 and S40 (construct one) and for residues S527 and S530 (construct two) displayed an impaired ability to mediate cell death under arsenic trioxide treatment (Hayakawa & Privalsky 2004). Because of its implication in arsenic trioxide response, MAPK1 is proving to be an interesting candidate to study further. The number of phosphorylation sites on PML recognized by MAPK1 -and therefore the great number of possible degron sites for FBXO9 interaction- would explain in part why we were unable to locate PML’s degron.

CDK1 is known to phosphorylate PML under hypoxia, at S518, leading to its ubiquitination by the CRL CUL3-KLHL20, followed by its degradation (Yuan et al. 2011). Interestingly, PML NB display a change in number, size and intensity during cell cycle with only a few PML NBs remaining in mitosis, when CDK1 is mostly active, making it an interesting target to study (Everett et al. 1999; Malumbres & Barbacid 2009) (Figure 71).

128

Figure 71 : Localization of PML through cell cycle. Immunofluorescence of PML (green) and SP100 (red) in Hep2 cells synchronized and fixed at intervals, scale bars represent 10µm. A. S-phase, PML and SP100 co-localized (yellow) making clear PML NBs. B. C. D. E. shows various stages of mitosis and early G1 phase where PML NBs number increases and slowly disappear as the cell goes through mitosis. F. most of the cells are in G1 phase. Based on Everett et al. 1999.

129

Both the literature about CK2, and the preliminary data obtained with different stimuli (Figure 66) suggest that CK2 might trigger PML degradation through its phosphorylation upon specific CK2 activating stimuli. CK2 is known to phosphorylate PML at S565, leading to its proteasomal degradation, but the exact mechanism by which PML is ubiquitinated and degraded is still unknown (Nisole et al. 2013; Scaglioni et al. 2006) (Figure 72). Therefore, it is very tempting to propose SCFFBXO9 as the ubiquitin ligase recognizing CK2’s phosphorylation on PML, causing its ubiquitination and subsequent degradation. This hypothesis however needs further confirmation.

Figure 72 : CK2 phosphorylation of PML leads to ubiquitination and subsequent degradation through an unknown process. Diagram showing two paths leading to PML degradation. One though arsenic trioxide induced stress, the other through PML phosphorylation by CK2 in response to anisomycin (an antibiotic preventing DNA and protein synthesis), UV radiation or osmotic shock. CK2 phosphorylation sites are indicated in red next to the SIM hydrophobic core in blue (Nisole et al. 2013).

Identifying the kinase involved in PML stability regulation will hopefully provide some hints about the physiological stimulus causing PML’s kinase activation, phosphorylation and subsequent degradation by the proteasome.

130

9) PML and FBXO9 in diseases

It has been demonstrated that PML plays a relevant role in some key cellular processes in both normal and pathological conditions, like transcription regulation, cell differentiation or oncogenesis (Sahin et al. 2014; Guan & Kao 2015). Taking this into consideration, we wondered if any diseases showed signs of increased SCF activity, or PML stability dysregulation. In this context, Acute Promyelocytic Leukemia (APL) might be an interesting case, since it directly involves PML through the fusion oncoprotein PML-RARα, and is treated using Arsenic trioxide (ATO), which triggers the degradation of the oncoprotein. It would be interesting to test whether the SCFFBXO9 complex could be playing a role in APL remission. To test this hypothesis, APL cells (NB4) will be depleted in FBXO9 and treated with All Trans- Retinoic Acid (ATRA) or ATO, in order to check if their differentiation or apoptosis is affected.

Data obtained from our mini screen revealed that Non-Small Lung cancer (NSCLC) could be another interesting model to investigate, as PML appears upregulated at the mRNA level in our study (Figure 69). However, PML was reported to be completely or partially lost in NSCLC (Gurrieri, Nafa, et al. 2004; Koken et al. 1995). We also wondered if FBXO9 might be overexpressed in some diseases leading to SCF unbalanced activity on PML, and this might also stand true for the kinase modifying PML’s degron. In order to observe potential unbalances in the SCF-PML axis, databases analysis software were used such as genevestigator (Hruz et al. 2008), which allowed to see where FBXO9 was mostly expressed in normal tissues and in some neoplastic samples.

FBXO9 is highly expressed under normal conditions in bones. However, some solid tumor neoplasm samples -such as invasive breast cancer- were exhibiting a high level of FBXO9 mRNA, and an almost unchanged level of PML mRNA level (Annex 6). In this study, seeking to identify human Triple-Negative Breast Cancer (TNBC) subtypes (Lehmann et al. 2011), 24 breast cancer samples were selected, all of them showing an overexpression of FBXO9 and a slight decrease in PML mRNA. Moreover, in this study, two patients received and responded to a taxanes based treatment (docetaxel), an anti-mitotic agent, causing a significant increase of PML mRNA levels (Annex 6). However, the number of samples included in this study was relatively small, and this avenue needs to be further investigated and a deeper insight should be gained regarding the molecular consequences of the treatment. In addition, it is important to keep in mind that these studies are based on mRNA levels and as

131

shown earlier, mRNA and protein quantities are not necessarily directly linked (Vogel & Marcotte 2012). In addition, we propose that the effect of FBXO9 on PML level is mainly due to a post-translational mechanism, instead of a direct transcriptional effect. However, this pilot study based on mRNA levels could provide some useful information regarding the possible correlation between PML levels and some specific types of cancer. In addition, besides a transcriptional regulation of the PML gene, we can not rule out the possible contribution of post-translational mechanisms as well. Since we proposed a new post-translational mechanism of regulation for PML stability, analysis of studies like this and of available databases might help to orientate future research, to clarify its possible involvement in the development of these pathologies, and to explore the possibility of influencing cancer progression by targeting it. Moreover, multiple studies reported a decrease in PML protein in some types of breast cancer (Mazza & Pelicci 2013) while others mention CK2 involvement in that disease as well (Filhol et al. 2015). Since there are more studies implicating PML and breast cancer than PML and lung cancer, we decided to go deeper in the search of possible links according to available datasets. To have an idea of the protein levels detected in lung and breast cancers we consulted the Human Protein Atlas (Uhlén et al. 2015; Pontén et al. 2008).

This database was created using antibody staining to quantify protein abundance in cells and tissues. PML staining in lung and breast cancer samples revealed that there was a strong decrease in PML at the protein level especially for breast cancers. In ten of the patients, PML was no longer detected while it is at medium abundance in control tissues (glandular cells) (Annex 7). This decrease in PML protein level is also observed in lung cancer but to a minor extend (Annex 7). As discussed earlier, mRNA levels do not necessarily reflects protein levels and the results observed through the Human protein Atlas confirmed previous studies. In some breast cancers, PML mRNA levels show minor variations -unlike in lung cancer- but the PML protein level is strongly reduced, pointing toward a post-translational modification mechanism of regulation of PML stability (Gurrieri, Capodieci, et al. 2004; Mazza & Pelicci 2013). Moreover, a recent study showed that CUL1, the scaffold protein of SCFFBXO9, was overexpressed in some breast cancers with bad prognosis due to its impact on cell proliferation, invasion and migration (Bai et al. 2013). In addition, CK2 expression is also playing a role in epithelial-to-mesenchymal transition and a dysregulation of its regulatory subunit (CK2β) leads to a poor prognostic in breast cancer (Filhol et al. 2015). Taking all this into consideration, it would be of great interest to go deeper in the possible relation between PML stability misregulation and the initiation and/or the development of specific types of cancers. In time, it

132

would be relevant to explore the therapeutic possibilities of targeting this new PML regulatory system.

Future work based on results obtained from our mini screen include the verification of PML protein levels in cell lines exhibiting high or normal PML mRNA levels. Mainly cell lines exhibiting PML downregulation at the protein level, but not at the mRNA level will be selected for future functional studies, with a main focus on cell lines also exhibiting high mRNA and protein levels of FBXO9. These cell lines would then be treated with shRNA targeting FBXO9 to observe the effects on PML stability and PML NB morphology.

10) Mouse model Fbxo9 KO

The mouse model KO for Fbxo9 is viable, but exhibits adipocyte differentiation problems (Lee et al. 2013) and abnormal locomotor behavior (unpublished data). Obese mice displayed high expression of FBXO9 and an increased expression during adipogenesis (Lee et al. 2013). The role of FBXO9 in adipocyte differentiation was recently confirmed in humans. Peroxisome proliferator-activated receptor gamma (PPARγ) is a master regulator of adipocyte differentiation and lipogenesis by binding to PPAR response elements leading to transcription activation (Shuibang Wang et al. 2016). FBXO9 interacts with PPARγ through its N-Terminal activation function-1 and Ligand Binding Domain (LBD) leading to its ubiquitination and subsequent degradation (Lee et al. 2016). PPARγ is also involved in Fatty-Acid Oxidation (FAO) pathway by controlling for example acyl-CoA oxidase transcription. Interestingly, PML is also involved in the FAO pathway but through PPARδ allowing asymmetric cell division leading to the maintenance of hematopoietic stem cells (Ito et al. 2012). Both cellular models (cancer cell lines where we can correlate low levels of PML protein with high activity of SCFFBXO9 and/or CK2) and mice models such as Fbxo9 -/- mice could be used to investigate SCFFBXO9 effect on PML stability, and the consequences of the misregulation of the PML- SCFFBXO9 axis.

11) PML and cellular differentiation

PML was also reported to play a role in cellular differentiation, in particular by controlling Oct4 gene activity, critical for the maintenance of Embryonic Stem Cells (ESC). PML maintains an open chromatin conformation of the Oct4 promoter, causing its constant expression in stem

133

cells. However, upon Retinoic Acid treatment, PML is no longer located at the promoter and Oct4 is no longer expressed leading to cell differentiation (Chuang et al. 2011). Moreover, Pml was found to also be important in mammary gland development in mice, as its depletion disturbs the balance of the luminal progenitor populations (Li et al. 2009). Another study showed that Pml was also involved in the development of the mice neocortex by controlling the function of Neural Progenitor Cells (NPCs) (Regad et al. 2009). PML expression in brain became recently a source of interest, since it appears to be actively involved in brain development and synaptic plasticity (Butler et al. 2013; Korb & Finkbeiner 2013). Interestingly, the PML protein has been detected in neuronal precursors in the brain, but it decreases to undetectable levels just at the onset of the migration and differentiation of these precursors to neurons. This decrease in endogenous PML protein was further and independently confirmed in the induced neural differentiation on a mouse cellular model (Korb & Finkbeiner 2013). Together with the poorly characterized change on PML protein level during the progression of the cell cycle, this is the only example to date under physiological conditions where PML protein level is endogenously regulated in a natural process different from induced stress. It would be very interesting to investigate if the SCFFBXO9 is involved in regulating PML stability in this context -which may also contribute to explain the abnormal locomotor behavior observed in the Fbxo9 KO mice.

12) PML and innate immunity

Finally, PML role in innate immunity is also increasingly studied, and in particular for its effects on defenses against viruses. PML and PML NBs prevent replication of the dengue virus (Giovannoni et al. 2015) but also of rabies virus (Blondel et al. 2010) and HIV (Kahle et al. 2015) through an unknown mechanism. Upon infection, PML also activates IRF3 allowing to enhance the production of IFN-β, that will warn neighboring uninfected cells leading to activation of anti-viral defenses (El Asmi et al. 2014; Y. Chen et al. 2015). Even though this type of response is very interesting as it implicates PML directly, it might be challenging to use it as a new model of stress inducing SCFFBXO9 activity. Indeed, viruses usually code for proteins directly interfering with innate immunity, such as HSV1 E3 ubiquitin ligase ICP0 which targets directly PML for degradation (Cuchet-Lourenco et al. 2012; McNally et al. 2008). Some studies showed that viruses were able to hijack the host cell post-translational modifications such as SUMOylation (Chang et al. 2016). PML being an important player in cell innate immunity, it would not be surprising if some viruses would hijack the SCF complex to trigger its degradation.

134

Our project allowed the characterization of a new Ubiquitin ligase complex, SCFFBXO9, as a new regulator of PML stability. A longer perspective would be to investigate the possible effects of this complex on other proteins present in NBs and/or SUMOylated like SP100, another major constituent of PML NBs. Insights into the possible influence of SCFFBXO9 on these PML partners might shed some light on the functional outcome of PML stability misregulation, and its effects on innate immunity, cell differentiation or carcinogenesis.

13) The SCF complex: a druggable target

Besides its implication in many different diseases and cellular processes, PML stability could be pharmacologically modulated either by targeting SUMOylation and the formation of NBs or through the SCF or (more generally) Cullin RING Ligases (CRLs) complexes responsible for its ubiquitination. Few SUMOylation modulators have been described so far, such as Ginkgolic Acid (Fukuda, Ito, Hirai, et al. 2009) and Kerriamicin B (Fukuda, Ito, Uramoto, et al. 2009), which are used to inhibit the SUMO E1 activating enzyme SAE1/SAE2. Two other inhibitors targeting UBC9 were also recently discovered: an oxygenated flavonoid derivative (Kim et al. 2013) and a small molecule binding the active site of UBC9 (Kumar et al. 2014). These are potent substances, but inhibiting SUMOylation altogether is predicted to cause serious side effects precluding their use in clinical practice. Ubc9 deficient mice embryos are not viable (Nacerddine et al. 2005) and in adult mice, Ubc9 depletion affects the small intestine ultimately leading to death (Demarque et al. 2011).

Another approach will consist in acting on the CRLs, which could in turn be even less specific by affecting different CRLs involved on several pathways, but there are many ways to target them, especially for the case of SCF complexes. SCF complexes can be inhibited by preventing their formation using NEDDylation inhibitors such as MLN4924 (Zhang et al. 2016; Tanaka et al. 2012). Other approaches consist in inhibiting directly the ubiquitin E2 conjugation enzyme, blocking the interaction between SKP1 and the F-Box protein (inhibitor: 6-OAP) (Liu et al. 2015; Gorelik et al. 2016), or finally preventing the F-Box protein from interacting with its substrate (Skaar et al. 2014; Hussain et al. 2016) (Figure 74). A study looked for genetic profile of tumors (of cervical origin) in order to develop personalized patient care that will detect sensitivity or resistance to certain therapies (Muller et al. 2015). Since in some cancers PML is repressed or its stability altered (Gurrieri, Nafa, et al. 2004; Koken et al. 1995), it could

135

be a therapeutic advantage to be able to specifically target PML stability in patients that present such dysregulation.

Figure 73 : SCF complexes can be targeted by drugs to treat diseases. Diagram showing a brief overview of SCF complex ligase activity under normal and oncogenic conditions and the different ways this complex activity can be inhibited. β-TRCP, β-transducin repeat-containing protein; CRLs, Cullin RING Ligases; Cul1/7, Cullin-1 or -7; Fbxw7, F-box and WD repeat domain containing 7; Fbxo, F-Box protein; NEDD8, Neural precursor cell expressed developmentally down-regulated protein 8; Rbx1, Ring-Box protein 1; SCF, Skp1- Cullin-F-Box protein; Skp1, S phase kinase-associated protein 1; Skp2, S phase kinase- associated protein 2; SRD, Substrate Recognition Domain; Ub, Ubiquitin (Hussain et al. 2016).

F-Box proteins are the most specific part of the complex, the one that is direct and specifically recognizing substrates. Therefore, they might represent the best targets for specific drug design. There are two ways to manipulate the SCF ubiquitin ligase activity: it can either be restored or inhibited. Restoration of function can be achieved using Proteolysis Targeting Chimaeras (PROTACs). PROTACs are chimeric molecules composed at one end by a degron- mimicking compound capable of recruiting an E3 ligase, while the other end is capable of

136

recruiting substrate resulting in its degradation. One example is the degradation of methionine aminopeptidase 2 (METAP2) by SCFβTrCP where the nuclear factor-κB inhibitor-α (IκBα) degron was used to recruit SCFβTrCP and the drug ovalicin was used to recruit METAP2 (Sakamoto et al. 2003; Sakamoto et al. 2001). Other methods to restore SCF function involve the use of small molecules to help interaction, and/or increase affinity with the substrate, or redirect degradation using small molecules.

The inhibition of SCF function including the use of competitive inhibitors binding the same F-Box, as well as allosteric inhibitors physically preventing interaction by causing conformational changes at the substrate recognition site. For example, βTrCP inhibitors efficiently blocked prostate and breast cancer proliferation (Skaar et al. 2014). Finally, as previously seen, inhibition of the SCF assembly could also prevent SCF activity (Figure 75) (Skaar et al. 2014). In addition, since FBXO9 can recognize -like most F-Box proteins- a phosphorylated degron, it can also be targeted with small molecules to modulate its interaction (Watanabe & Osada 2016). Moreover, the responsible kinase could also be a target for inhibition, even though it will not be as specific as targeting an F-Box protein. SCFFBXO9 activity can be modulated in a number of ways, allowing a pharmacological regulation of PML stability.

137

Figure 74 : Manipulation of SCF ubiquitin ligase activity trough different strategies. Diagram showing the normal function of SCF, recognizing and ubiquitnating its substrate (a). The first type of strategy consists in restoring SCF function through the use of Proteolysis Targeting Chimaeras (PROTACs) (b); small molecules acting as a sort of molecular glue (c); small molecules used to allow the recognition of a different substrate (d). The second type of strategy consist in inhibition of SCF activity through the use of competitive inhibitors for substrate binding (e), allosteric inhibitor whose binding will cause a change in the substrate recognition site (f) and finally through the inhibition of SCF complex assembly (g). Ub, ubiquitin (Skaar et al. 2014).

14) Other potential candidates to be studied

The screen performed at the beginning of the development of this project, allowed the identification of many new proteins potentially implicated in PML regulation, and/or the SUMO pathway (Table 5).

138

Table 5 : Table summarizing results from the validating screen. Enzymes involved in the SUMO pathway are in bold and underscored. The two candidates used in this study are in bold.

Professor Anne Dejean’s research group will extend the analysis to other candidates presenting hyperSUMOylation like phenotypes. In order to identify new regulators of the SUMO pathway under arsenic treatment, a new small scale screen under arsenic trioxide treatment will be designed on a selection of candidates. Arsenic trioxide effects combined to selected siRNAs will be observed, and candidates exhibiting a prevention or exacerbation of phenotypes will be further investigated.

Other candidates in the remaining two classes, “hypoSUMOylation like” and absence of PML NBs, could also be studied as they might contain essential enzymes for the SUMO

139

pathway or PML stability, as well as regulators of PML expression. Among these candidates, we could find enzymes involved in chromatin remodeling and transcription regulation such as CHD4, whose structure includes a great number of SIMs, and BRD4, recently identified in the laboratory as a SUMO substrate (data not published) (Annex 2). Interestingly, as seen through the STRING network (Figure 41B, ) some of the candidates are part of the proteasome (such as PSMB5, PSMA3, PSMD10). Other proteins (CHD4, RBBP4, HDAC1) are part of the nucleosome remodeling and histone deacetylase complex (the NuRD complex) involved in transcriptional repression through histone deacetylation and nucleosome remodeling. These are only examples, as many other candidates show heterogeneous functions such as cold stress response or calcium homeostasis.

SUMOylation affects many different subtracts and is implicated in diverse pathologies, such as breast cancer (Zhang 2014), suggesting that its manipulation could be used as a therapeutic approach, as it is the case with arsenic trioxide treatment for APL. In that line, proteasome inhibitors targeting the ubiquitin pathway have already proved their effectiveness in the treatment for some cancers (Burger & Seth 2004).

This study allowed us to identify new key regulatory enzymes for PML, and might also provide new insights on PML protein and PML NBs regulation under normal and stress conditions. It will also shed some light on the mechanisms and relevance of PML misregulation in disease. Besides the fundamental aspects of this research, the characterization of these enzymes as new PML regulators could lead to the development and use of new therapeutic strategies aiming to directly -or indirectly- modulate PML stability, or to affect the SUMOylation regulatory pathways in different pathologies like cancer.

140

Conclusion

We identified a new complex regulating PML stability, and its principal cellular reservoirs, PML NBs. This complex, SCFFBXO9, is composed of at least four proteins: an adaptor protein (SKP1), a RING protein recruiting the ubiquitin E2 conjugation enzyme (RBX1), a scaffold protein (CUL1) and an F-Box protein recognizing specifically PML (FBXO9). At least three of its components were proven to interact with PML, and to modulate its stability under arsenic trioxide induced stress, both at the protein level and of PML NB morphology. This complex was also shown to specifically ubiquitinate PML in vitro. However, we were unable to identify PML’s degron putatively recognized by the F-Box protein or by a hypothetical kinase. Preliminary work identified three interesting candidates: CDK1, MAPK1 and CK2 for the role of kinase phosphorylating PML prior to SCFFBXO9 complex interaction. Moreover, tumoral mRNA expression coupled with an in silico study analysis revealed that lung and breast cancers might be interesting disease models to study SCFFBXO9 activity, and its effect on PML stability.

On a fundamental point of view, we hope that this work will lead to a better understanding of PML stability regulation, which could lead to the development of new therapeutic applications. PML is a tumor suppressor implicated in numerous oncogenic processes and downregulated in a number of human cancers. SCF complex can be targeted using many approaches, and with high specificity as well as some of the analyzed kinases that might be involved in the regulatory circuit proposed here. Ubiquitination and related post- translational modifications already constitute a great field of study and development for a growing number of laboratories. The proposed new model for PML regulation links these post- translational modifications with the specific SCFFBXO9 complex, thus providing new fronts to understand and target PML stability regulation.

141

Material and Methods

Cell culture HT-1080, HEK293T, Hela, 3T3 cells were cultured in Dulbecco's Modified Eagle's Medium (DMEM) + 10% Fetal Bovine Serum (FBS) along with 10µg/mL streptomycin and 10UI/µL penicillin at 37C° and 5% CO2. Primary Murine Embryonic Fibroblasts (MEFs) were cultivated at 37C° and 3% CO2 in the same medium. All cells tested negative for mycoplasma using a PCR-based detection method. siRNA screen HT1080 cells stably expressing GFP-PML IV were screened in triplicate with the Dharmacon Genome-wide siRNA SMARTpool library by reverse transfection in 96-well plate format. 0.3µl per well Lullaby transfection reagent (Oz Bioscience) and a final concentration of 37.5nM siRNA SMARTpool were added to each well containing 4000 cells. Control siRNA treatments were included in all plates (neutral controls Scrambled, RISC-free and On-Target Non-targeting siRNAs; positive controls Sumo2/3 as previously described (Lallemand-Breitenbach et al. 2008) and UBC9, RNF4 and PML siRNAs are from Dharmacon). Transfections were incubated for 3 days at 37˚C and 10% CO2 at which time the media were removed and cells fixed with 4% formaldehyde in Phosphate Buffer Saline (PBS) for 1 hour at room temperature. After washing with PBS, plates were stored or stained with 1µg/ml DAPI reagent for 1 hour, washed with PBS and imaged on the Cellomics Arrayscan VTi. Captured images were analyzed using the Spot detector (V3) Bioapplication and the number, area and intensity of PML foci in both the nucleus and cytoplasm were quantified. Data analysis were performed using bespoke pipelines and Bioconductor packages. Raw replicate data showed good reproducibility (correlation coefficient of between 0.76 and 0.88) and control data was well-separated (Z’ factor of 0.4). Raw data were normalized to the plate median and scored against the Median absolute deviation of the screen, replicates were summarized as the mean of the final scored values

Drug treatment Where indicated, the following drugs were used: Arsenic Trioxide (As2O3) (1µM final concentration prepared from 0.1M stock diluted in NaOH, or the same volume of NaOH as control); MG132 (25µM for 8 to 12h, Calbiochem 474790); Cycloheximide (100µg/mL, Sigma

142

C4859); NaCl (0.2M to induce osmotic shock); TBB (50µM, T0826 sigma); Leptomycin B, (20nM, Sigma L2913).

Transfections siRNA transfections were performed as described in siRNA screen method (Lullaby transfection reagent standard protocol (Oz Bioscience)) using concentrations ranging from 37.5 to 100nM.

Cells to be transfected with plasmids were washed with PBS, tripsinized (0.05% trypsin/EDTA, Thermo Fischer Scientific) resuspended, counted (TC10 BioRad counter) and plated the day before transfection. The next day, plasmid DNA (2µg for six-well plates, 10µg for 10cm plates and 20µg for 150cm plates) and X-tremeGENE™ HP DNA Transfection Reagent (Roche) were added as specified by Roche and left to incubate at room temperature for 20 min. The mix was later added to the cells drop by drop. The next day, the medium was changed and cells were finally collected 48 hours after transfection.

Stable cell lines, vectors, plasmids, shRNAs, and siRNAs HT1080 GFP-PML IV stable cell line was generated by retroviral infection with pBabe-GFP- PML retrovirus as previously described (Bischof et al. 2006) of the ecotrope HT-1080-E14 cell line (a gift from Eugene Kandel, Roswall Park Cancer Institute, NY, USA)(Kandel & Nudler 2002). Clonal selection was performed by serial dilution of the polyclonal population.

HT1080 Knock-Out (KO) FBXO9 cells were generated by CRISPR/Cas9. Plasmid design to target and generate selectable KO cells (with puromycin) were designed by TEBUBio and transfected as previously described. Cells were selected after 48h under puromycin selection and Clonal selection was performed by serial dilution of the polyclonal population. Selected clones were later controlled for FBXO9 deletion at the protein level by Western Blot.

Synthesized plasmids from the laboratory were pCMV-Flag-PML mutants (ΔRing; ΔRWGB. ΔRBCC, RBCC; ΔC-term; ΔN+Cterm; Δ320-371) with pCMV backbone from OriGene; pLenti-Flag-PML mutants and pLenti-GFP-PML mutants; pGFP-PML3KR; pBABE- GFP-PML IV; pCDNA-His-UB, pCDNA-Ha-Ub.

The other plasmids (pCDNA 3.1 backbone) were a generous gift from Dr Bassermann’s team: pCDNA-FBXO9, pCDNA-Flag-FBXO9 and pCDNA-GFP-FBXO9.

143

pCMV-Flag-SKP1a; pCMV-Flag-RBX1; pCMV TEL2 and pCMV-Flag-PML IV, V, VI, VII were bought from OriGene.

FBXO9 shRNAs were purchased from SIGMA (Mission shRNA plasmid DNA (pLKO.1)): shCRTL; shFBXO9-2 3’- CCGGCCAGAGGTTCAACAAACTCATCTCGAGAT GAGTTTGTTGAACCTCTGGTTTTTG-5’; shFBXO9-3 3’-CCGGTGGAATATTACAGGT ACATAACTCGAGTTATGTACCTGTAATATTCCATTTTTG-5’.

The siRNAs were purchased from Dharmacon (GE Healthcare): On-Target plus SMARTpool: siPML (L-006574); siRBX1 (L-004087); siSKP1 (L-003323); siCUL1 (L- 004086); siRNF4 (L-006557); siRNF111 (Arkadia) (L-007002); siCAND1 (L-015562); siCSNK2A1 (L-003475); siCSNK2A2 (L-004752); siCSNK2B (L-007679); siICK (L- 004811); siHIPK2 (L-003266); siCDK2 (L-003236); siCDK1 (L-003224); siATR (L-003202); siCHEK2 (L-003256); siMAPK7 (L-003513); siMAPK1 (L-003555); siKLHL20 (L-004893) siNon-targeting (D-001810-10). siGENOME: FBXO9 (M-012469). Single siRNA target sequences: siNon-targeting (D-001810-01); siRNF4 (J-006557-08) 3’- GAAUGGACGUCUCAUCGUU-5’; siSKP1 (J-003323-14) 3’-CGCAAGACCUUCAA UAUCA-5’; siRBX1 (J-004087-09) 3’-GGAACCACAUUAUGGAUCU-5’; siFBXO9-1 (D- 012469-04) 3’-GCAACUUGUACC UGAUAUA-5’; siFBXO9-2 (D-012469-01) 3’- GGUGUAAGCUCUAGCAAUU-5’.

Generation of viral particles and infections On the first day, HEK293T cells were transfected using calcium phosphate transfection protocol with a mix of three plasmids: pVSV, pRSV8.71 (both used for the production of viruses), and a plasmid to express the desired protein or shRNA (30µg of total DNA. The DNA was mixed with 93µL CaCl2 (2M) diluted in H2O to reach a total volume of 750µL. Meanwhile, 750µL of HEPES2X (Sigma, 51558) was placed in a 15mL Falcon tube. HEK293T cells (80% confluence) were washed with PBS followed by 5 minutes tripsinization and collected. After ^6 2+ cell counting, a master mix containing 8x10 cells per condition was made. DNA/Ca /H2O mix was added dropwise to HEPES2X while gently vortexing to make precipitates. 9mL of cells were immediately added to the precipitate and transferred into a T75 flask. Medium was changed the following morning. Supernatants were collected and filtered (0.45µm, Millipore) in a P3 laboratory for three consecutive days. All collections from the same condition were pulled together and viral particles were concentrated using a concentration kit, overnight at 4°C

144

(Lenti-X™ Concentrator, 631231, CloneTech). The next morning, tubes were spun down at 4000rpm for 45 min, the supernatant removed and the pellet re-suspended in 500µL DMEM. The re-suspended viruses were either used right away for an infection or were aliquoted and stored at -80C°.

To perform the infection, target cells (MEFs, 3T3 or HT1080) were washed with PBS, tripsinized for 5 minutes and resuspended in DMEM. After cell counting, 5x10^6 cells were plated on 150mm plates or 4x10^5 cells in six-well plates. The supernatant containing viruses was directly added to cells after addition of polybrene® (8mg/mL, Sigma). The following morning, medium was replaced and a new batch of viral supernatant added. The medium is changed again 24 hours later and cells were left for an additional 24 hours before being treated and/or collected.

Immunofluorescence Cells were previously plated on round coverslips in six-well plates and treated (transfections and/or arsenic trioxide treatment). Cells were washed twice in cold PBS, fixed in Formalin 10% (Sigma-Aldrich) for 15 minutes at room temperature and finally permeabilized using a PBS solution with 0.5% of Triton X-100 (final concentration) for 15 min at room temperature. Cells were then washed three times in PBS for 5 minutes. Coverslips were washed with PBS-Tween (PBST) and primary antibodies, diluted in wash buffer according to manufacturer’s instruction, were added and incubated for 1 hour at room temperature. Coverslips were then washed three times with PBS and once with PBST. Secondary antibodies, coupled to Alexa® fluorophores (1/400), were diluted in PBS and were incubated for 45 minutes in the dark. A wash with PBST followed, along with three PBS washes. One of these washes was supplemented with 4',6- diamidino-2-phenylindole (DAPI) to stain the cell nuclei, and incubated for 1 minute. Next, coverslips were mounted using Vectashield mounting medium (VectorLabs). Then, immunofluorescence was observed under a Fluorescence Microscope (ZEISS, axiocam MRM) and or Apotome (ApoTome.2) using ZEISS software (ZEN). Further analysis were carried on ICY Bioimage analysis software (de Chaumont et al. 2012).

Antibodies The following primary antibodies were purchased from Abcam: anti-CK2β (ab76025, rabbit monoclonal); anti-CUL1 C-ter (ab75817, rabbit polyclonal); anti-CUL1 full length ([AS97.1] ab11047, mouse monoclonal); anti-GAPDH (ab8245, Mouse monoclonal); anti-ICK

145

(ab194411, mouse polyclonal); anti-RBX1(ab86862, rabbit polyclonal); anti-SKP1a (ab10546, rabbit polyclonal and ab124473 for mouse monoclonal); Vinculin (SP3227) (ab18058, mouse monoclonal). Others were bought from SIGMA: anti-ACTINβ (A2066, Rabbit polyclonal); anti-DKK (or FLAG tag) (F7425, Rabbit polyclonal); anti-HA tag (H3663, mouse monoclonal); anti-TUBULIN (T5168, Mouse monoclonal); or from Cell Signaling (anti-His- tag (2365P, Rabbit polyclonal); anti-p-AKT (587f11, mouse monoclonal)); Santa Cruz biotechnologies (anti-GFP (SC-9996, mouse monoclonal); anti-HA tag (sc805, rabbit polyclonal), or other companies (anti-c-Myc (Calbiochem, OP10, mouse monoclonal); anti-DKK (OriGene, TA50011-1, mouse monoclonal); anti-RNF4 (Novus, H00006047-A01, mouse polyclonal); anti-TEL2 (Proteintech,15975-1-AP, Rabbit polyclonal); anti-Ub (Thermo scientific, 1859660, rabbit polyclonal)).

Multiple antibodies were used against PML: anti-PML (IF, IP; Abcam, ab50637, Mouse monoclonal); anti-PML (IF, WB; Homemade antibody, pan PML '42', rabbit polyclonal); anti- PML (IF, WB; Novus, NB100-59787, pan PML, Rabbit polyclonal).

Anti-FBXO9, recognizing amino acids 390 to 404, rabbit polyclonal antibody was generously provided by Dr. Bassermann’s team (Department of Medicine III, Technische Universitat Munchen, Munich, Germany) (Fernández-Sáiz et al. 2013). Anti-FBXO25 was also sent from Dr. Bassermann’s team (Baumann et al. 2014). Anti-RNF4, rabbit monoclonal antibody was a generous gift from Dr. Palvimo’s team (University of Eastern Finland, Kuopio, Finland) (Häkli et al. 2001).

Secondary antibodies for immunofluorescence were purchased at Invitrogen (Thermo Fisher Scientific): anti-Mouse IgG (H+L) (A-11001, goat, Alexa® fluor 488nm and A-11032, goat, Alexa® fluor 594nm); anti-Rabbit IgG (H+L) (A-11034, goat, Alexa® fluor 488nm and A-11012, goat, Alexa® fluor 594nm, goat, Alexa® fluor 488nm). The following antibodies were used for Western blots: anti-Rabbit IgG (H+L) (Thermo Fisher Scientific, A-21109, goat, Alexa® fluor 680nm and Cell Signaling Technologies, 5151, goat, DyLight™ 800nm conjugate) and anti-Mouse IgG (H+L) (Cell Signaling Technologies, 5257, goat, DyLight™ 800nm conjugate and 5470, goat, DyLight™ 680nm conjugate).

146

Co-immunoprecipitation Cells were washed with cold PBS, scrapped and transferred into a 2mL tube then lysed at 4°C for 15 minutes using lysing buffer: Tris buffer 50mM pH 8.0; EDTA 0.1mM; NaCl 200mM; 0.5% of NP-40; 10% glycerol and 1 tab of EDTA-free protease inhibitor (Roche). When protein SUMOylation had to be preserved, lysis buffer was supplemented with N-Ethylmaleimide (NEM), an inhibitor of SUMO proteases. Samples were later quantified and equilibrated using PierceTm 660nm protein assay (Thermo Fisher Scientific). Next, samples were immunoprecipitated using anti-FLAG® M2 coated magnetic beads (Sigma, M8823) or anti GFP coated magnetic beads GFP-Trap®_M (ChromoTek, gtm-20) previously equilibrated in Chris or TRIS buffer. A pre-clear step was made using Trap®_M control beads for GFP based immunoprecipitations. Then, samples were incubated with beads overnight at 4°C on a wheel. The next morning, FLAG beads were washed in Tris Buffer Saline (TBS: 50mM Tris HCl, 150 mM NaCl, pH 7.4) whereas GFP beads were washed in GFP beads wash buffer (10mM Tris HCl, 150 mM NaCl, 0.01% NP-40, 0.5mM EDTA, pH 7.5). One of these washes was supplemented with salt (500mM NaCl, final concentration) to reduce non-specific binding to beads. Elution from FLAG-beads was done using 3X FLAG® peptide (200ng/µL, Sigma), twice at 30°C for 2h. Laemmlli 2x was used for GFP beads, twice at 60°C for 10 minutes and once at 95°C for 5 minutes. Samples were later analyzed by Western blots.

Nucleo-cytoplasmic fractionation Cells were washed with PBS and collected as previously described. Then, the fractionation was done with nuclear extraction kit (Abcam, ab113474) according to the manufacturer instructions. The immunoprecipitation of the cytosolic and nuclear fractions was done as described above.

Western blotting Cells were washed with cold PBS and lysed in 1.25x Laemmli buffer. Samples were then heated at 95°C for 5 minutes and sonicated (2x10seconds). Proteins were separated using SDS-PAGE electrophoresis, 3-8% or 4-12% gels (Criterion XT, Bio-Rad). Once proteins were transferred to a nitrocellulose membrane, it was blocked using I-blockTm (Thermo Fisher Scientific) and then incubated with primary antibodies, diluted as specified by the antibody manufacturer, for one hour at room temperature or overnight at 4°C. Next, membranes were washed five times for five minutes in PBST and diluted secondary antibodies coupled to Alexa® fluorophore (1/10000) were incubated for up to 45 minutes. Fluorescence signals from the blots were monitored and analyzed using Odyssey instrument (Li-Cor).

147

Ubiquitination assays In-Vitro ubiquitination of PML IV was performed in a volume of 15µL containing 50mM Tris -1 buffer at pH 7.6, 5mM MgCl2, 0.6 mM dithiothreitol, 2mM ATP, 1.5 ngµL E1 (Boston Biochem), 10 ng µL-1 Ubc3, 10 ngµL-1 UBC5, 2.5 µg µL-1 ubiquitin (Sigma), 1µM ubiquitin aldehyde, 4µL of purified FLAG-PML IV or PML mutant protein from HEK293T cells treated in absence or presence of Arsenic via FLAG immunoprecipitation followed by FLAG-peptide elution and dialysis against PBS with 5% glycerol. Approximately 1µg of each purified Skp1Fbxo9 or Skp1Fbxo25 together with recombinant Cul-1Roc1 complexes purified from 5B insect cells, were added into the ubiquitin reactions (Baumann et al. 2014; Fernández-Sáiz et al. 2013).

In-Vivo ubiquitination assay were performed on HEK293T cells transfected with HA- Ub along with FLAG-PMLV and/or un-tagged FBXO9. PML was subsequently immunoprecipitated following the immunoprecipitation protocols described above. mRNA expression in cell lines Cells were washed with cold PBS and lysed using Qiagen’s RTL buffer (supplemented with β- mercaptoethanol). Samples were later applied to QiAshredder spin column (Qiagen) and RNA purification was carried out following RNeasy RNA isolation kit procedure (Qiagen). Purified RNA samples were later quantified using QubitTM RNA BR assay kit (Invitrogen) and, if quantities were sufficient, submitted to the Bioanalyser (RNA 6000 Nano chip, Agilent technologies) to check for RNA quality. Samples were kept in the study using a RNA Integrity Number (RIN) cut off value of five as recommended for qPCR studies (Fleige & Pfaffl 2006). Complementary DNA (cDNA) was synthesized using random primer kit according to manufacturer’s instruction (High-Capacity cDNA Reverse Transcription Kit, Applied BiosystemsTM). Quantitative-PCRs (qPCR) were performed using the SYBR green PCR kit from Applied BiosystemsTM, on the CFX-96 cycler (Bio-Rad).

To obtain relative quantities of gene expression level, some or all of the following housekeeping genes ACTINβ, GAPDH, POLR2A, RPLPO and TRIM 44 were used; these were chosen based on published data (de Jonge et al. 2007) and availability of primers in the laboratory. HT1080 cells (connective tissue) and ALEX cells (liver tissue) were chosen as

148

calibrators for their low housekeeping expression variation. The gene expression raw data were normalized using 2-ΔΔCT method (Livak & Schmittgen 2001).

Primers used for the qPCRs were purchased from Qiagen (Quantitect primers): SP100 (QT00056343); PML (QT00090447); CAND1 (QT00093310); RBX1 (QT00074214); CDK1 (QT00042672); FBXO9 (QT00034426); CSNK2B (QT00012446); CSNK2A2 (QT00014385); CSNK2A1 (QT00064148); CSNK2A3 (QT02503963); SKP1 (QT00040320); TELO2 (QT00046114); CUL1 (QT00024591); UBE2I (QT00046424); TTI1 (QT00060921); MYC (QT01663361); ACTB (QT01680476); GAPDH (QT00079247); TRIM44 (QT00042994); POLR2A (QT00033264); RPLP0 (QT00075012).

Statistical analysis/bioinformatics Data mining of mRNAs levels for PML and FBXO9 genes was performed on Genevestigator software (Hruz et al. 2008). All statistical analysis for PML Nuclear Body morphology study were performed using GraphPad Prism version 6.07 for Windows, GraphPad Software, San Diego California USA, www.graphpad.com (Mann-Whitney test). R software was used to analyse data from the mRNA cell line screen (ANOVA tests) (Wickham 2007; R Core Team 2015). Data obtained from the study of the 72 cell lines allowed to generate a heat map displaying mRNA abundance (log2) for each of the targets described (Kolde 2015; Neuwirth 2014), a correlation matrix based on Pearson's coefficient (Frank E Harrell Jr 2016; Wright 2015) and finally, Multiple Factor Analysis (MFA) was used to define clusters based on PML and FBXO9 mRNA levels (Lê et al. 2008; Josse & Husson 2016).

Mouse models Mouse Embryonic Fibroblasts (MEFs) Pml-/- were generated from Pml Knock Out mice acquired through the generous collaboration with Dr. Pandolfi’s team (Beth Israel Deaconess Medical Center, Boston, USA) (Wang, Delva, et al. 1998).

MEF FBXO9 -/- were extracted from the mouse line Fbxo9_tm1b_DO5 bought from the Toronto Center for phenogenomics, Canada.

149

Annexes Annex 1: The 69 Mammalian F-Box proteins (Jin et al. 2004)

150

151

Annex 2: F-Box protein and E3 ubiquitin ligase implication in cellular pathways.

F-box proteins are shown in red, other E3 ligases in green and key processes in bleu. Interactions leading to activations are shown with an arrow, inhibitions with a bar and indirect effects with dashed lines, direct binding or complex formation is shown by a double blue line (Randle & Laman 2015).

152

Annex 3: Validated candidates inducing a morphological change of PML Nuclear Bodies.

Annex key: Green box: significative increase of measured values Red box: significative decrease of measured values Italic characters: siRNA lethal in other screens

Validated Validated Oligos Potential SUMO1 or Oligos Potential type Potential PML Gene modifying consensus SUMO 2 Gene Name modifying II SUMO sites SIMs interactant symbol PML NBs SUMO sites interactant PML NBs (SUMOplot) (SUMOplot) (Pubmed) count per (SUMOplot) (Pubmed) intensity nuclei

S-phase kinase- SKP1A 3 2 0 0 1 associated protein 1 chromosome X open CXORF1 2 4 0 0 2 reading frame 1 RBX1 ring-box 1 3 2 0 0 0 vacuolar protein VPS45A sorting 45 homolog 2 2 0 3 13 (S. cerevisiae) shisa homolog 5 SCOTIN 2 2 0 0 15 (Xenopus laevis) RCOR3 REST corepressor 3 2 3 2 1 4 SUMO2 DKFZP54 zinc finger protein 4 3 2 1 7 7K1113 710

NADH dehydrogenase NDUFB9 (ubiquinone) 1 beta 2 3 1 3 1 subcomplex, 9, 22kDa mitochondrial MRPS11 ribosomal protein 2 3 0 2 5 S11 ITGB8 integrin, beta 8 3 2 0 2 18 homer homolog 2 HOMER2 3 2 2 4 2 (Drosophila) HNRNPA heterogeneous 3 2 0 1 2 0 nuclear

153

ribonucleoprotein A0 bombesin-like BRS3 2 2 0 0 19 receptor 3 ENDOG endonuclease G 2 2 0 0 7 PNOC prepronociceptin 2 2 0 0 4 proline PRODH dehydrogenase 2 2 0 0 7 (oxidase) 1 spondin 1, SPON1 extracellular matrix 2 2 3 0 4 protein solute carrier family SLC39A1 39 (metal ion 2 2 0 0 6 1 transporter), member 11 zinc finger protein ZNF460 3 1 0 0 1 460 similar to zinc finger ZNF532 protein 347; zinc 3 1 1 7 5 finger protein 532 zinc finger protein ZNF77 3 1 0 0 0 77 actin binding LIM ABLIM3 protein family, 2 1 2 1 4 member 3 SRY (sex SUMO1/ SOX6 determining region 2 1 3 2 5 2 Y)-box 6 chromosome 19 CATSPER open reading frame 2 1 2 0 25 G 15 stimulated by STRA6 retinoic acid gene 6 2 1 1 1 30 homolog (mouse) signal peptide, CUB SCUBE1 2 1 0 1 7 domain, EGF-like 1 delta/notch-like EGF DNER 2 1 0 0 27 repeat containing zinc finger protein ZNF684 2 1 1 0 3 684 U2 small nuclear U2AF1L4 RNA auxiliary factor 2 1 0 0 3 1-like 4 WD repeat domain WDR27 2 1 1 0 15 27 SNX33 sorting nexin 33 2 1 0 3 5 proteasome PSMA3 3 0 2 0 2 (prosome,

154

macropain) subunit, alpha type, 3 3-hydroxy-3- methylglutaryl- HMGCR 2 0 1 1 24 Coenzyme A reductase interferon-induced protein with IFIT1 2 0 1 1 6 tetratricopeptide repeats 1 proteasome (prosome, PSMB5 2 0 0 0 2 macropain) subunit, beta type, 5 proteasome (prosome, PSMD10 macropain) 26S 2 0 0 1 4 subunit, non- ATPase, 10 PDPN podoplanin 2 0 0 0 13 tudor and KH TDRKH 2 0 0 1 6 domain containing limb region 1 LMBR1L homolog (mouse)- 2 0 0 1 41 like GSG1L GSG1-like 2 0 0 0 14 WD repeat domain WDR53 2 0 1 0 5 53

ubiquitin- conjugating enzyme SUMO1/ UBE2I 4 4 0 0 2 OUI E2I (UBC9 homolog, 2 yeast) ubiquitin-like SUMO1/ UBA2 modifier activating 4 4 0 1 7 2 enzyme 2 SUMO1 activating SUMO1/ SAE1 4 4 0 1 7 enzyme subunit 1 2 IQ motif and Sec7 IQSEC1 2 4 3 4 10 domain 1 bromodomain BRD4 2 4 1 15 12 containing 4 chromosome 16 C16ORF5 open reading frame 2 4 1 1 5 7 57 phospholipase A2, PLA2G2E 4 3 1 1 4 group IIE zinc finger, DHHC- ZDHHC4 3 3 0 2 20 type containing 4

155

chymotrypsin-like CELA2A elastase family, 3 3 0 0 4 member 2A chromodomain CHD4 helicase DNA 2 3 9 19 10 OUI SUMO2 binding protein 4 2',3'-cyclic CNP nucleotide 3' 2 3 1 1 9 phosphodiesterase solute carrier family 37 (glucose-6- SLC37A4 phosphate 2 3 1 1 14 transporter), member 4 MATN2 matrilin 2 2 3 1 0 16 RNA binding motif RBM3 (RNP1, RRM) 2 3 0 0 2 protein 3 cold shock domain CSDC2 containing C2, RNA 2 3 1 binding lens epithelial LENEP 2 3 0 0 1 protein chromosome 11 C11ORF4 open reading frame 2 3 0 0 2 8 48 TMEM17 transmembrane 2 3 0 0 2 7 protein 177 ARCN1 archain 1 4 2 3 0 5 histone deacetylase HDAC1 3 2 2 5 5 OUI SUMO2 1 calcium homeostasis CALHM2 3 2 0 1 12 modulator 2 STE20-related STRADA kinase adaptor 3 2 0 1 5 alpha FERM domain FRMD6 3 2 1 0 10 containing 6 family with FAM50B sequence similarity 2 2 1 4 3 50, member B A kinase (PRKA) AKAP8L anchor protein 8- 2 2 1 3 1 like DnaJ (Hsp40) DNAJC27 homolog, subfamily 2 2 1 1 8 C, member 27 mesoderm posterior MESP1 2 2 0 1 4 1 homolog (mouse)

156

cytidine monophosphate N- CMAS 2 2 0 1 7 acetylneuraminic acid synthetase mitochondrial MRPL40 ribosomal protein 2 2 0 5 1 L40 tetratricopeptide TTC25 2 2 2 4 0 repeat domain 25 HUS1 checkpoint HUS1B homolog b (S. 2 2 0 0 9 pombe) KRT72 keratin 72 2 2 1 4 2

ACTN3 actinin, alpha 3 4 1 1 7 CRYGA crystallin, gamma A 3 0 0 1 FGL1 fibrinogen-like 1 3 0 2 2 MT2A metallothionein 2A 3 0 0 0 MYO1C myosin IC 3 2 1 28 nuclear cap binding NCBP1 protein subunit 1, 3 1 1 18 80kDa hypothetical LOC642954; RBBP4 3 0 0 1 SUMO2 retinoblastoma binding protein 4 mitochondrial MRPS12 ribosomal protein 3 0 2 3 S12 cytokine receptor- CRLF1 4 0 0 12 like factor 1 RAB11 family RAB11FI interacting protein 3 3 1 3 13 P3 (class II) RNA binding motif RBM12 2 1 3 12 protein 12; copine I cleavage and polyadenylation CPSF6 4 1 0 3 specific factor 6, 68kDa DIP2 disco- interacting protein 2 DIP2A 3 2 2 41 homolog A (Drosophila) family with FAM53C sequence similarity 3 0 0 2 53, member C

157

sushi domain SUSD4 4 1 0 16 containing 4 zinc finger family ZNF673 2 1 1 1 member 673 NSFL1 (p97) NSFLl1C 3 0 1 4 cofactor (p47) JPH1 junctophilin 1 3 1 5 9 vacuolar protein VPS33A sorting 33 homolog 3 1 1 16 A (S. cerevisiae) SH3 domain and SH3TC2 tetratricopeptide 3 2 1 22 repeats 2 Rab interacting RILP 4 1 0 6 lysosomal protein ARHGAP Rho GTPase 2 0 5 18 18 activating protein 18 cytochrome P450, CYP2U1 family 2, subfamily 2 3 1 16 U, polypeptide 1 spermatogenesis SPATA2L 2 0 0 6 associated 2-like ankyrin repeat and ASB7 SOCS box-containing 4 1 0 4 7 adenomatosis APCDD1 polyposis coli down- 4 1 0 7 regulated 1 SET and MYND SMYD1 4 0 1 6 domain containing 1 BEN domain BEND7 2 1 1 7 containing 7 CCR4-NOT transcription CNOT6L 4 1 1 15 complex, subunit 6- like prolyl 4- P4HA3 hydroxylase, alpha 2 2 0 10 polypeptide III Na+/K+ transporting NKAIN3 3 0 0 17 ATPase interacting 3 KRT26 keratin 26 3 1 1 3 kelch-like 30 KLHL30 3 0 1 17 (Drosophila)

polymerase (RNA) II (DNA directed) POLR2B 4 4 2 0 polypeptide B, 140kDa

158

polymerase (RNA) II POLR2G (DNA directed) 4 4 1 0 polypeptide G polymerase (RNA) II (DNA directed) POLR2I 4 4 0 0 polypeptide I, 14.5kDa RAN, member RAS RAN 4 4 0 0 oncogene family suppressor of Ty 6 SUPT6H homolog (S. 4 4 1 4 cerevisiae) THOC1 THO complex 1 4 4 1 2 AF4/FMR2 family, AFF4 3 4 6 8 member 4 DEAD (Asp-Glu-Ala- DDX19A As) box polypeptide 3 4 2 2 19A mediator complex MED9 3 4 0 1 subunit 9 polymerase (RNA) II POLR2D (DNA directed) 3 4 0 0 polypeptide D tumor protein p53 TP53I11 3 4 0 3 inducible protein 11 CD33 CD33 molecule 2 4 0 2 corticotropin CRHBP releasing hormone 2 4 2 0 binding protein glucosidase, beta GBA2 2 4 1 1 (bile acid) 2 mediator complex MED28 2 4 0 2 subunit 28 metallophosphoeste MPPED2 rase domain 2 4 1 0 containing 2 tetratricopeptide TTC23 2 4 2 2 repeat domain 23 CD160 CD160 molecule 4 3 0 0 G protein-coupled GPR39 4 3 0 2 receptor 39 dihydrodiol DHDH dehydrogenase 3 3 1 1 (dimeric) GLI family zinc finger GLI3 3 3 5 4 3 guanine nucleotide GNG8 binding protein (G 3 3 1 0 protein), gamma 8

159

NIN1/RPN12 binding protein 1 homolog NOB1 (S. cerevisiae); 3 3 1 2 hypothetical LOC100132364 hypothetical gene supported by NM_014886; TGF beta-inducible NSA2 3 3 1 2 nuclear protein 1; similar to TGF beta- inducible nuclear protein 1 myosin, heavy chain MYH4 3 3 11 19 4, skeletal muscle Purkinje cell protein PCP4 3 3 0 0 4 post-GPI PGAP1 attachment to 3 3 3 2 proteins 1 phosphatidylinositol glycan anchor PIGM 3 3 0 1 biosynthesis, class M phospholipase C, gamma 2 PLCG2 3 3 0 4 (phosphatidylinosito l-specific) secreted LY6/PLAUR SLURP1 3 3 0 0 domain containing 1 zinc finger protein ZNF142 3 3 1 3 142 ABI1 abl-interactor 1 2 3 0 0 aldehyde ALDH18 dehydrogenase 18 2 3 0 1 A1 family, member A1 adaptor protein, phosphotyrosine APPL2 interaction, PH 2 3 1 3 domain and leucine zipper containing 2 B double prime 1, subunit of RNA BDP1 polymerase III 2 3 6 15 transcription initiation factor IIIB coiled-coil domain CCDC12 2 3 1 2 containing 12

160

CDC42 effector DCD42E protein (Rho GTPase 2 3 0 1 P1 binding) 1 DIM1 dimethyladenosine DIMT1L 2 3 1 0 transferase 1-like (S. cerevisiae) G patch domain and GPKOW 2 3 2 0 KOW motifs GPR109 niacin receptor 2; 2 3 B niacin receptor 1 histone cluster 1, H3j; histone cluster 1, H3i; histone cluster 1, H3h; histone cluster 1, H3g; histone cluster 1, H3f; histone cluster 1, H3e; histone cluster 1, HIST1H2 H3d; histone cluster 2 3 ? ? AD 1, H3c; histone cluster 1, H3b; histone cluster 1, H3a; histone cluster 1, H2ad; histone cluster 2, H3a; histone cluster 2, H3c; histone cluster 2, H3d interleukin ILF3 enhancer binding 2 3 2 7 factor 3, 90kDa late cornified LCE3A 2 3 0 0 envelope 3A OSTN osteocrin 2 3 0 0 polycomb group PCGF6 2 3 1 1 ring finger 6 transcription TCEAL7 elongation factor A 2 3 2 0 (SII)-like 7 tripartite motif- TRIM31 2 3 0 2 containing 31 zinc finger protein ZNF593 2 3 0 0 593 alkB, alkylation ALKBH3 repair homolog 3 (E. 3 2 0 0 coli) cytoplasmic CPEB1 3 2 0 0 polyadenylation

161

element binding protein 1 immunoglobulin IGDCC3 superfamily, DCC 3 2 0 1 subclass, member 3 polymerase (DNA- POLD4 3 2 0 0 directed), delta 4 polypyrimidine tract PTBP1 3 2 0 1 SUMO1 binding protein 1 replication factor C RFC5 (activator 1) 5, 3 2 0 1 36.5kDa serpin peptidase inhibitor, clade A SERPINA (alpha-1 3 2 1 4 antiproteinase, antitrypsin), member 4 2 SH3GLB SH3-domain GRB2- 3 2 0 2 2 like endophilin B2 small proline-rich SPRR1B 3 2 0 2 protein 1B (cornifin) adrenergic, alpha- ADRA2C 2 2 0 0 2C-, receptor CD79a molecule, CD79A immunoglobulin- 2 2 0 1 associated alpha G protein-coupled GPR132 2 2 0 1 receptor 132 programmed cell PDCD2 2 2 1 1 death 2 ring finger protein RNF122 2 2 0 0 122 thymosin-like 2 (pseudogene); thymosin-like 1 TMSL1 2 2 0 3 (pseudogene); thymosin beta 4, X- linked WNT1 inducible WISP2 signaling pathway 2 2 0 0 protein 2 apolipoprotein L APOLD1 3 0 0 domain containing 1 eukaryotic EIF3B translation initiation 3 0 3 factor 3, subunit B DIRAS family, GTP- DIRAS2 2 0 2 binding RAS-like 2

162

MYST histone Myst1 2 1 3 acetyltransferase 1 phosphoribosyl pyrophosphate PRPSAP2 2 0 0 synthetase- associated protein 2 transmembrane TMEM14 protein 14D; 2 0 0 D transmembrane protein 14B sterile alpha motif SAMD8 4 1 0 domain containing 8 SIX6 SIX homeobox 6 4 0 2 SRY (sex SOX2 determining region 4 1 1 Y)-box 2 TDP-glucose 4,6- TGDS 4 1 0 dehydratase V-set and transmembrane VSTM2L 4 0 1 domain containing 2 like amyloid beta (A4) precursor protein- APBA1 3 0 3 binding, family A, member 1 serologically defined CWC27 colon cancer 3 3 5 antigen 10 deoxyhypusine DOHH hydroxylase/monoo 3 1 0 xygenase gap junction GJC3 protein, gamma 3, 3 0 0 30.2kDa HAUS augmin-like HAUS6 3 4 6 complex, subunit 6 ISL2 ISL LIM homeobox 2 3 0 1 katanin p80 (WD KATNB1 repeat containing) 3 1 2 subunit B 1 LPA lipoprotein, Lp(a) 3 NADH dehydrogenase NDUFAF (ubiquinone) 1 3 0 0 3 alpha subcomplex, assembly factor 3 WAP four-disulfide WFDC2 3 0 0 core domain 2

163

zinc finger CCCH- ZC3H12A 3 0 0 type containing 12A zinc finger, CCHC ZCCHC3 3 0 2 domain containing 3 FBN2 fibrillin 2 2 2 2 FAD1 flavin adenine dinucleotide FLAD1 2 1 0 synthetase homolog (S. cerevisiae) ladybird homeobox LBX1 2 0 3 1 ribosomal protein L29 pseudogene 9; ribosomal protein L29 pseudogene 12; ribosomal protein RPL29 2 0 2 L29 pseudogene 11; ribosomal protein L29; ribosomal protein L29 pseudogene 26 synovial sarcoma, X SSX3 2 0 0 breakpoint 3 tumor necrosis TNFAIP8 factor, alpha- 2 0 0 L2 induced protein 8- like 2 UBQLN2 ubiquilin 2 2 0 1

164

Annex 4: Co-immunoprecipitation screen to identify PML interacting F- Box protein.

165

Annex 5: Cell lines used in mRNA mini screen.

Samples Origin Tissue organ/tissue Diseased

Immortalized AKN1 Liver No Hepatocytes Hepatocellular ALEX Liver Yes carcinoma BJAB lymphocytic leukemia Blood Yes Head and Neck Head and BB48 Yes carcinoma Neck BG1 Breast cancer breast Yes Non-small cell Lung Calu1 Lung Yes cancer Non-small cell Lung Calu6 Lung Yes cancer Caski Cervical cancer Cervix Yes T cells T cells Blood No CEM lymphocytic leukemia Blood Yes Connective Clone IV2 Fibrosarcoma Yes tissue Biliary EGl1 Cholangiocarcinoma Yes track ES2 ovarian cancer Ovary Yes immortalized FA2N Liver No Hepatocytes Fadu squamous cell cancer Skin Yes Hepatocellular FOCUS Liver Yes carcinoma FS Liver Liver No Biliary GBC-SD Cholangiocarcinoma Yes track Non-small cell Lung GILI5C Lung Yes cancer NCI-H1299 small-cell lung cancer Lung Yes Immortalized HaCaT skin Yes keratinocytes Hepatocellular HBG Liver Yes carcinoma HCT116 Colorectal cancer Intestine Yes HEC1B gastric cancer stomach Yes human embryonic HEK293 Kidney No kidney HELA Cervical cancer Cervix Yes Hepatocellular HepaRG Liver Yes carcinoma HepG2 Hepatoblastoma Liver Yes HEY ovarian cancer Ovary Yes Human primary HH Liver No hepatocytes HL60 lymphocytic leukemia Blood Yes Connective HT1080 Fibrosarcoma Yes tissue

166

HT29 Colorectal cancer Intestine Yes HUH13 (M. Hepatocellular Liver Yes musculus) carcinoma Hepatocellular HuH2.2 Liver Yes carcinoma HuH6 Hepatoblastoma Liver Yes Hepatocellular HuH7 Liver Yes carcinoma Hepatocellular HuSp Liver Yes carcinoma Non-small cell Lung ISUL Lung Yes cancer JURKAT lymphocytic leukemia Blood Yes K562 lymphocytic leukemia Blood Yes immortalized Hs399Li Liver Yes Hepatocytes Hepatocellular Li7A Liver Yes carcinoma LnCaP prostate cancer Prostate Yes Hepatocellular Malhavu Liver Yes carcinoma MCF7 Breast cancer breast Yes MDAMB231 Breast cancer breast Yes Connective MRC5 Fibroblast No tissue cholangiocellular MZ-ChA1 Liver Yes carcinoma Non-small cell Lung NIIH441 Lung Yes cancer Hepatocellular NUK1 Liver Yes carcinoma Biliary OCUCG1 Cholangiocarcinoma Yes track OVCAR3 ovarian cancer Ovary Yes cholangiocellular OZ Liver Yes carcinoma PC3 prostate cancer Prostate Yes PEO14 ovarian cancer Ovary Yes PEO16 ovarian cancer Ovary Yes RAJI lymphocytic leukemia Blood Yes Head and Neck Head and RPMI2650 Yes carcinoma Neck SAOS2 Osteosarcoma Bone Yes Head and Neck Head and SCC15 Yes Squamous cell cancer Neck Head and Neck Head and SCC9 Yes Squamous cell cancer Neck SK-BN2 Neuroblastoma Brain Yes SKNSH Neuroblastoma Brain Yes Non-small cell Lung SK-MES1 Lung Yes cancer SKMEL28 Melanoma Skin Yes SKV-TU Vulvar Cancer Vulva Yes

167

Hepatocellular SuHC1 Liver Yes carcinoma SW480 Colorectal cancer Intestine Yes T47D Breast cancer Breast Yes U2OS Osteosarcoma Bone Yes Connective WI38 Fibroblast No tissue

168

Annex 6: FBXO9 is overexpressed in some types of breast cancers.

Data mining results from Genevestigator (Hruz et al. 2008). Upper figure: anatomy tissue samples are used as control (n=46) to be compared to breast cancer tissues (n=24). Lower figure details mRNA levels obtained from the 24 patients among which two were responsive to treatment.

169

Annex 7: Cancer tissue PML antibody staining of Breast and lung cancers. (Human Protein Alas)

For each cancer, the fraction of samples with antibody staining/protein expression level is indicated by the blue-scale color. The length of the bar represents the number of patient samples analyzed (max=12 patients). The number of patient in each category is written within boxes. Staining made using two antibodies (CAB010194 and CAB16304). Immunofluorescence validation of the antibody in A-431 cells is provided to the left of the figure. Next to the cancer staining data, the protein expression data of normal tissues or specific cell types corresponding to each cancer are shown and protein expression levels are indicated by the blue-scale color coding. For breast cancer, glandular cells were used while respiratory epithelial cells and pneumocytes were used for lung cancer.

170

Bibliographical References

Abidi, N. & Xirodimas, D.P., 2015. Regulation of cancer-related pathways by protein NEDDylation and strategies for the use of NEDD8 inhibitors in the clinic. Endocrine- Related Cancer, 22(1), pp.T55–T70. Alkuraya, F.S. et al., 2006. SUMO1 haploinsufficiency leads to cleft lip and palate. Science (New York, N.Y.), 313(5794), p.1751. Available at: http://www.ncbi.nlm.nih.gov/pubmed/16990542 [Accessed July 11, 2016]. El Asmi, F. et al., 2014. Implication of PMLIV in Both Intrinsic and Innate Immunity. PLoS Pathogens, 10(2). Atwan, Z. et al., 2016. Promyelocytic leukemia protein isoform II inhibits infection by human adenovirus type 5 through effects on HSP70 and the interferon response. The Journal of general virology. Available at: http://www.microbiologyresearch.org/content/journal/jgv/10.1099/jgv.0.000510.v1 [Accessed June 9, 2016]. Bai, C. et al., 1996. SKP1 connects cell cycle regulators to the ubiquitin proteolysis machinery through a novel motif, the F-box. Cell, 86(2), pp.263–74. Available at: http://www.ncbi.nlm.nih.gov/pubmed/8706131 [Accessed April 19, 2016]. Bai, J. et al., 2013. Cullin1 is a novel marker of poor prognosis and a potential therapeutic target in human breast cancer. Annals of Oncology, 24(8), pp.2016–2022. Barbash, O., Lee, E.K. & Diehl, J.A., 2011. Phosphorylation-dependent regulation of SCF(Fbx4) dimerization and activity involves a novel component, 14-3-3ɛ. Oncogene, 30(17), pp.1995–2002. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3084329&tool=pmcentrez&r endertype=abstract [Accessed June 4, 2016]. Batty, E.C., Jensen, K. & Freemont, P.S., 2012. PML nuclear bodies and other TRIM-defined subcellular compartments. Advances in experimental medicine and biology, 770, pp.39– 58. Available at: http://www.ncbi.nlm.nih.gov/pubmed/23630999 [Accessed May 31, 2016]. Baumann, U. et al., 2014. Disruption of the PRKCD–FBXO25–HAX-1 axis attenuates the apoptotic response and drives lymphomagenesis. Nature Medicine, 20(12), pp.1401– 1409. Available at: http://www.nature.com/doifinder/10.1038/nm.3740 [Accessed June 17, 2016]. Baumeister, W. et al., 1998. The proteasome: paradigm of a self-compartmentalizing protease. Cell, 92(3), pp.367–80. Available at: http://www.ncbi.nlm.nih.gov/pubmed/9476896 [Accessed June 2, 2016]. Bellodi, C. et al., 2006. A Cytoplasmic PML Mutant Inhibits p53 Function. Cell Cycle, 5(22), pp.2688–2692. Available at: http://www.tandfonline.com/doi/abs/10.4161/cc.5.22.3504 [Accessed June 8, 2016]. Bellodi, C. et al., 2006. Cytoplasmic Function of Mutant Promyelocytic Leukemia (PML) and PML-Retinoic Acid Receptor-. Journal of Biological Chemistry, 281(20), pp.14465– 14473. Available at: http://www.jbc.org/cgi/doi/10.1074/jbc.M600457200 [Accessed June 8, 2016]. Bernardi, R. et al., 2004. PML regulates p53 stability by sequestering Mdm2 to the nucleolus. Nature cell biology, 6(7), pp.665–72. Available at: http://www.nature.com/doifinder/10.1038/ncb1147 [Accessed June 1, 2016]. Bernardi, R. & Pandolfi, P.P., 2003. Role of PML and the PML-nuclear body in the control of programmed cell death. Oncogene, 22(56), pp.9048–57. Available at:

171

http://www.ncbi.nlm.nih.gov/pubmed/14663483 [Accessed June 6, 2016]. Bernardi, R. & Pandolfi, P.P., 2007. Structure, dynamics and functions of promyelocytic leukaemia nuclear bodies. Nature reviews. Molecular cell biology, 8(12), pp.1006–16. Available at: http://www.ncbi.nlm.nih.gov/pubmed/17928811 [Accessed March 3, 2012]. Best, J.L. et al., 2002. SUMO-1 protease-1 regulates gene transcription through PML. Molecular Cell, 10(4), pp.843–855. Available at: http://linkinghub.elsevier.com/retrieve/pii/S1097276502006998 [Accessed June 24, 2016]. Bischof, O. et al., 2002. Deconstructing PML-induced premature senescence. The EMBO journal, 21(13), pp.3358–69. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=126090&tool=pmcentrez&re ndertype=abstract. Bischof, O. et al., 2001. Regulation and localization of the Bloom syndrome protein in response to DNA damage. The Journal of cell biology, 153(2), pp.367–80. Available at: http://www.ncbi.nlm.nih.gov/pubmed/11309417 [Accessed June 27, 2016]. Bischof, O. et al., 2006. The E3 SUMO Ligase PIASy Is a Regulator of Cellular Senescence and Apoptosis. Molecular Cell, 22(6), pp.783–794. Bischof, O., Nacerddine, K. & Dejean, A., 2005. Human papillomavirus oncoprotein E7 targets the promyelocytic leukemia protein and circumvents cellular senescence via the Rb and p53 tumor suppressor pathways. Molecular and cellular biology, 25(3), pp.1013– 24. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=543993&tool=pmcentrez&re ndertype=abstract [Accessed June 6, 2016]. Blondel, D. et al., 2010. Resistance to Rabies Virus Infection Conferred by the PMLIV Isoform. Journal of Virology, 84(20), pp.10719–10726. Available at: http://jvi.asm.org/lookup/doi/10.1128/JVI.01286-10 [Accessed June 21, 2016]. Bond, M. & Wu, Y.-J., 2011. Proliferation unleashed: the role of Skp2 in vascular smooth muscle cell proliferation. Frontiers in bioscience (Landmark edition), 16, pp.1517–35. Available at: http://www.ncbi.nlm.nih.gov/pubmed/21196245 [Accessed January 19, 2015]. Bonilla, W. V et al., 2002. Effects of promyelocytic leukemia protein on virus-host balance. Journal of virology, 76(8), pp.3810–8. Available at: http://www.ncbi.nlm.nih.gov/pubmed/11907221 [Accessed July 12, 2016]. Boutell, C., Orr, A. & Everett, R.D., 2003. PML residue lysine 160 is required for the degradation of PML induced by herpes simplex virus type 1 regulatory protein ICP0. Journal of virology, 77(16), pp.8686–94. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=167235&tool=pmcentrez&re ndertype=abstract [Accessed June 2, 2016]. Brand, P., Lenser, T. & Hemmerich, P., 2010. Assembly dynamics of PML nuclear bodies in living cells. PMC biophysics, 3(1), p.3. Buczek, M.E. et al., 2015. Cytoplasmic PML promotes TGF-β-associated epithelial– mesenchymal transition and invasion in prostate cancer. Oncogene. Available at: http://www.nature.com/doifinder/10.1038/onc.2015.409 [Accessed June 21, 2016]. Burger, A.M. & Seth, A.K., 2004. The ubiquitin-mediated protein degradation pathway in cancer: therapeutic implications. European journal of cancer (Oxford, England : 1990), 40(15), pp.2217–29. Available at: http://www.ncbi.nlm.nih.gov/pubmed/15454246 [Accessed March 28, 2012]. Butler, K., Martinez, L.A. & Tejada-Simon, M. V., 2013. Impaired cognitive function and reduced anxiety-related behavior in a promyelocytic leukemia (PML) tumor suppressor

172

protein-deficient mouse. Genes, Brain and Behavior, 12(2), pp.189–202. Available at: http://doi.wiley.com/10.1111/gbb.12014 [Accessed July 12, 2016]. Le Cam, L. et al., 2006. E4F1 Is an Atypical Ubiquitin Ligase that Modulates p53 Effector Functions Independently of Degradation. Cell, 127(4), pp.775–788. Campagna, M. et al., 2011. SIRT1 stabilizes PML promoting its sumoylation. Cell death and differentiation, 18(1), pp.72–9. Available at: http://www.ncbi.nlm.nih.gov/pubmed/20577263 [Accessed June 24, 2016]. Carnero, A. & Paramio, J.M., 2014. The PTEN/PI3K/AKT Pathway in vivo, Cancer Mouse Models. Frontiers in oncology, 4, p.252. Available at: http://www.ncbi.nlm.nih.gov/pubmed/25295225 [Accessed June 7, 2016]. Carracedo, A., Ito, K. & Pandolfi, P.P., 2011. The nuclear bodies inside out: PML conquers the cytoplasm. Current opinion in cell biology, 23(3), pp.360–6. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3109131&tool=pmcentrez&r endertype=abstract [Accessed March 26, 2012]. Ceasar, S.A. et al., 2016. Insert, remove or replace: A highly advanced genome editing system using CRISPR/Cas9. Biochimica et biophysica acta. Available at: http://www.ncbi.nlm.nih.gov/pubmed/27350235. Cesare, A.J. & Reddel, R.R., 2013. Alternative Lengthening of Telomeres in Mammalian Cells. Chan, J.Y.. et al., 1998. Altered expression of the growth and transformation suppressor PML gene in human hepatocellular carcinomas and in hepatitis tissues. European Journal of Cancer, 34(7), pp.1015–1022. Chang, C.-C. et al., 2011. Structural and Functional Roles of Daxx SIM Phosphorylation in SUMO Paralog-Selective Binding and Apoptosis Modulation. Molecular Cell, 42(1), pp.62–74. Chang, P.-C., Campbell, M. & Robertson, E.S., 2016. Human Oncogenic Herpesvirus and Post-translational Modifications - Phosphorylation and SUMOylation. Frontiers in microbiology, 7, p.962. Available at: http://www.ncbi.nlm.nih.gov/pubmed/27379086 [Accessed July 8, 2016]. de Chaumont, F. et al., 2012. Icy: an open bioimage informatics platform for extended reproducible research. Nature Methods, 9(7), pp.690–696. Available at: http://www.nature.com/doifinder/10.1038/nmeth.2075 [Accessed June 17, 2016]. Chen, B. et al., 2015. Therapeutic and analytical applications of arsenic binding to proteins. Metallomics : integrated biometal science, 7(1), pp.39–55. Available at: http://pubs.rsc.org.login.ezproxy.library.ualberta.ca/en/content/articlehtml/2015/mt/c4mt 00222a. Chen, Y. et al., 2015. Promyelocytic Leukemia Protein Isoform II Promotes Transcription Factor Recruitment To Activate Interferon Beta and Interferon-Responsive Gene Expression. Molecular and cellular biology, 35(10), pp.1660–72. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=4405644&tool=pmcentrez&r endertype=abstract [Accessed May 31, 2016]. Chen, Z. et al., 2014. Co-expression of PKM2 and TRIM35 predicts survival and recurrence in hepatocellular carcinoma. Oncotarget, 6(4), pp.2539–2548. Cheng, X. & Kao, H.-Y., 2012. Post-translational modifications of PML: consequences and implications. Frontiers in oncology, 2(January), p.210. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3539660&tool=pmcentrez&r endertype=abstract [Accessed November 14, 2013]. Chuang, Y.S. et al., 2011. Promyelocytic leukemia protein in retinoic acid-induced chromatin remodeling of Oct4 gene promoter. Stem Cells, 29(4), pp.660–669. Chung, I. et al., 2012. PML body meets telomere: the beginning of an ALTernate ending?

173

Nucleus (Austin, Tex.), 3(3), pp.263–75. Available at: http://www.ncbi.nlm.nih.gov/pubmed/22572954 [Accessed June 6, 2016]. Cobb, M.H., 1999. MAP kinase pathways. Progress in Biophysics and Molecular Biology, 71(3), pp.479–500. Condemine, W. et al., 2006. Characterization of endogenous human promyelocytic leukemia isoforms. Cancer research, 66(12), pp.6192–8. Available at: http://www.ncbi.nlm.nih.gov/pubmed/16778193 [Accessed July 19, 2016]. Cuchet-Lourenco, D. et al., 2012. Herpes Simplex Virus 1 Ubiquitin Ligase ICP0 Interacts with PML Isoform I and Induces Its SUMO-Independent Degradation. Journal of Virology, 86(20), pp.11209–11222. Available at: http://jvi.asm.org/cgi/doi/10.1128/JVI.01145-12 [Accessed June 21, 2016]. Culjkovic, B. et al., 2008. The eIF4E RNA regulon promotes the Akt signaling pathway. Journal of Cell Biology, 181(1), pp.51–63. D’Orazi, G. et al., 2002. Homeodomain-interacting protein kinase-2 phosphorylates p53 at Ser 46 and mediates apoptosis. Nature cell biology, 4(1), pp.11–9. Available at: http://www.ncbi.nlm.nih.gov/pubmed/11780126 [Accessed June 23, 2016]. Van Damme, E. et al., 2010. A manually curated network of the PML nuclear body interactome reveals an important role for PML-NBs in SUMOylation dynamics. International journal of biological sciences, 6(1), pp.51–67. Available at: http://www.ijbs.com/v06p0051.htm [Accessed June 24, 2016]. Dellaire, G. et al., 2006. Promyelocytic leukemia nuclear bodies behave as DNA damage sensors whose response to DNA double-strand breaks is regulated by NBS1 and the kinases ATM, Chk2, and ATR. The Journal of cell biology, 175(1), pp.55–66. Available at: http://www.ncbi.nlm.nih.gov/pubmed/17030982 [Accessed May 9, 2016]. Dellaire, G. & Bazett-Jones, D.P., 2004. PML nuclear bodies: dynamic sensors of DNA damage and cellular stress. BioEssays : news and reviews in molecular, cellular and developmental biology, 26(9), pp.963–77. Available at: http://www.ncbi.nlm.nih.gov/pubmed/15351967 [Accessed June 6, 2016]. Demarque, M.D. et al., 2011. Sumoylation by Ubc9 regulates the stem cell compartment and structure and function of the intestinal epithelium in mice. Gastroenterology, 140(1), pp.286–96. Available at: http://dx.doi.org/10.1053/j.gastro.2010.10.002. Dohmen, R.J., 2004. SUMO protein modification. Biochimica et biophysica acta, 1695(1-3), pp.113–31. Available at: http://www.ncbi.nlm.nih.gov/pubmed/15571812. Duprez, E. et al., 1999. SUMO-1 modification of the acute promyelocytic leukaemia protein PML: implications for nuclear localisation. Journal of cell science, 112 ( Pt 3, pp.381– 93. Available at: http://www.ncbi.nlm.nih.gov/pubmed/9885291. Enserink, J.M., 2015. Sumo and the cellular stress response. Cell division, 10(1), p.4. Available at: http://dx.doi.org/10.1186/s13008-015-0010- 1\nhttp://www.celldiv.com/content/10/1/4. Erker, Y. et al., 2013. Arkadia, a novel SUMO-targeted ubiquitin ligase involved in PML degradation. Molecular and cellular biology, 33(11), pp.2163–77. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3648077&tool=pmcentrez&r endertype=abstract [Accessed December 13, 2013]. Evdokimov, E. et al., 2008. Loss of SUMO1 in mice affects RanGAP1 localization and formation of PML nuclear bodies, but is not lethal as it can be compensated by SUMO2 or SUMO3. Journal of cell science, 121(Pt 24), pp.4106–13. Available at: http://www.ncbi.nlm.nih.gov/pubmed/19033381 [Accessed June 24, 2016]. Everett, R.D. et al., 1999. Cell cycle regulation of PML modification and ND10 composition. Journal of cell science, 112 ( Pt 2, pp.4581–8. Available at: http://www.ncbi.nlm.nih.gov/pubmed/10574707 [Accessed June 1, 2016].

174

Fanelli, M. et al., 2004. The Coiled-coil Domain Is the Structural Determinant for Mammalian Homologues of Drosophila Sina-mediated Degradation of Promyelocytic Leukemia Protein and Other Tripartite Motif Proteins by the Proteasome. Journal of Biological Chemistry, 279(7), pp.5374–5379. Available at: http://www.jbc.org/cgi/doi/10.1074/jbc.M306407200 [Accessed June 25, 2016]. Ferbeyre, G. et al., 2000. PML is induced by oncogenic ras and promotes premature senescence. Genes and Development, 14(16), pp.2015–2027. Available at: http://www.ncbi.nlm.nih.gov/pubmed/10950866 [Accessed June 7, 2016]. Fernández-Sáiz, V. et al., 2013. SCFFbxo9 and CK2 direct the cellular response to growth factor withdrawal via Tel2/Tti1 degradation and promote survival in multiple myeloma. Nature cell biology, 15(1), pp.72–81. Available at: http://www.ncbi.nlm.nih.gov/pubmed/23263282 [Accessed November 11, 2013]. Filhol, O. et al., 2015. Protein kinase CK2 in breast cancer: The CK2?? regulatory subunit takes center stage in epithelial plasticity. Cellular and Molecular Life Sciences, 72(17), pp.3305–3322. Finley, D., 2009. Recognition and processing of ubiquitin-protein conjugates by the proteasome. Annual review of biochemistry, 78, pp.477–513. Available at: http://www.ncbi.nlm.nih.gov/pubmed/19489727 [Accessed June 25, 2016]. Fleige, S. & Pfaffl, M.W., 2006. RNA integrity and the effect on the real-time qRT-PCR performance. Molecular Aspects of Medicine, 27(2-3), pp.126–139. Flynn, R.L. et al., 2015. Alternative lengthening of telomeres renders cancer cells hypersensitive to ATR inhibitors. Science (New York, N.Y.), 347(6219), pp.273–7. Available at: http://www.ncbi.nlm.nih.gov/pubmed/25593184 [Accessed June 6, 2016]. Frank E Harrell Jr, with contributions from C.D. and many others., 2016. Hmisc: Harrell Miscellaneous. R package version 3.17-4. Available at: https://cran.r- project.org/package=Hmisc. Frescas, D. & Pagano, M., 2008. Deregulated proteolysis by the F-box proteins SKP2 and beta-TrCP: tipping the scales of cancer. Nature reviews. Cancer, 8(6), pp.438–49. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=2711846&tool=pmcentrez&r endertype=abstract [Accessed May 10, 2016]. Friendly, M. & Friendly, M., 2002. Corrgrams: Exploratory displays for correlation matrices. The American Statistician, 34(9), pp.1447–9. Available at: http://www.ncbi.nlm.nih.gov/pubmed/19412301. Fu, C. et al., 2005. Stabilization of PML nuclear localization by conjugation and oligomerization of SUMO-3. Oncogene, 24(35), pp.5401–13. Available at: http://www.ncbi.nlm.nih.gov/pubmed/15940266 [Accessed June 24, 2016]. Fukuda, I., Ito, A., Hirai, G., et al., 2009. Ginkgolic acid inhibits protein SUMOylation by blocking formation of the E1-SUMO intermediate. Chemistry & biology, 16(2), pp.133– 40. Available at: http://www.ncbi.nlm.nih.gov/pubmed/19246003 [Accessed April 15, 2012]. Fukuda, I., Ito, A., Uramoto, M., et al., 2009. Kerriamycin B inhibits protein SUMOylation. The Journal of antibiotics, 62(4), pp.221–4. Available at: http://www.ncbi.nlm.nih.gov/pubmed/19265871 [Accessed April 27, 2012]. Galisson, F. et al., 2011. A novel proteomics approach to identify SUMOylated proteins and their modification sites in human cells. Molecular & cellular proteomics : MCP, 10(2), p.M110.004796. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3033685&tool=pmcentrez&r endertype=abstract [Accessed June 1, 2016]. Gambacorta, M. et al., 1996. Heterogeneous nuclear expression of the promyelocytic

175

leukemia (PML) protein in normal and neoplastic human tissues. The American journal of pathology, 149(6), pp.2023–35. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=1865355&tool=pmcentrez&r endertype=abstract [Accessed June 6, 2016]. Gao, C., Ho, C.-C., et al., 2008. Histone Deacetylase 7 Promotes PML Sumoylation and Is Essential for PML Nuclear Body Formation. Molecular and Cellular Biology, 28(18), pp.5658–5667. Available at: http://mcb.asm.org/cgi/doi/10.1128/MCB.00874-08 [Accessed June 24, 2016]. Gao, C., Cheng, X., et al., 2008. Signal-dependent Regulation of Transcription by Histone Deacetylase 7 Involves Recruitment to Promyelocytic Leukemia Protein Nuclear Bodies. Molecular Biology of the Cell, 19(7), pp.3020–3027. Available at: http://www.molbiolcell.org/cgi/doi/10.1091/mbc.E07-11-1203 [Accessed June 8, 2016]. Gao, Y.-M. et al., 2013. PML(NLS - ) Inhibits Cell Apoptosis and Promotes Proliferation in HL-60 Cells. International Journal of Medical Sciences, 10(5), pp.498–507. Available at: http://www.medsci.org/v10p0498.htm [Accessed June 8, 2016]. Geiss-Friedlander, R. & Melchior, F., 2007. Concepts in sumoylation: a decade on. Nature reviews. Molecular cell biology, 8(12), pp.947–56. Available at: http://www.ncbi.nlm.nih.gov/pubmed/18000527 [Accessed March 1, 2012]. Geng, Y. et al., 2012. Contribution of the C-terminal regions of promyelocytic leukemia protein (PML) isoforms II and V to PML nuclear body formation. The Journal of biological chemistry, 287(36), pp.30729–42. Available at: http://www.jbc.org/cgi/doi/10.1074/jbc.M112.374769 [Accessed May 3, 2016]. Giorgi, C. et al., 2010. PML regulates apoptosis at endoplasmic reticulum by modulating calcium release. Science (New York, N.Y.), 330(6008), pp.1247–51. Available at: http://www.ncbi.nlm.nih.gov/pubmed/21030605 [Accessed June 7, 2016]. Giovannoni, F., Damonte, E.B. & García, C.C., 2015. Cellular promyelocytic leukemia protein is an important dengue virus restriction factor. PLoS ONE, 10(5), pp.1–17. Available at: http://dx.doi.org/10.1371/journal.pone.0125690. Goldenberg, S.J. et al., 2004. Structure of the Cand1-Cul1-Roc1 complex reveals regulatory mechanisms for the assembly of the multisubunit cullin-dependent ubiquitin ligases. Cell, 119(4), pp.517–28. Available at: http://www.ncbi.nlm.nih.gov/pubmed/15537541 [Accessed May 13, 2016]. Gong, L. et al., 2000. Differential regulation of sentrinized proteins by a novel sentrin- specific protease. Journal of Biological Chemistry, 275(5), pp.3355–3359. Available at: http://www.jbc.org/cgi/doi/10.1074/jbc.275.5.3355 [Accessed June 24, 2016]. Gong, L. & Yeh, E.T.H., 2006. Characterization of a Family of Nucleolar SUMO-specific Proteases with Preference for SUMO-2 or SUMO-3. Journal of Biological Chemistry, 281(23), pp.15869–15877. Available at: http://www.jbc.org/cgi/doi/10.1074/jbc.M511658200 [Accessed June 24, 2016]. Gorelik, M. et al., 2016. Inhibition of SCF ubiquitin ligases by engineered ubiquitin variants that target the Cul1 binding site on the Skp1–F-box interface. Proceedings of the National Academy of Sciences, (6), p.201519389. Available at: http://www.pnas.org/lookup/doi/10.1073/pnas.1519389113. Gresko, E. et al., 2009. PML tumor suppressor is regulated by HIPK2-mediated phosphorylation in response to DNA damage. Oncogene, 28(5), pp.698–708. Available at: http://www.ncbi.nlm.nih.gov/pubmed/19015637 [Accessed June 23, 2016]. Grobelny, J. V, Godwin, A.K. & Broccoli, D., 2000. ALT-associated PML bodies are present in viable cells and are enriched in cells in the G(2)/M phase of the cell cycle. Journal of cell science, 113 Pt 24, pp.4577–85. Available at: http://www.ncbi.nlm.nih.gov/pubmed/11082050 [Accessed June 6, 2016].

176

Gu, H. et al., 2005. Components of the REST/CoREST/histone deacetylase repressor complex are disrupted, modified, and translocated in HSV-1-infected cells. Proceedings of the National Academy of Sciences of the United States of America, 102(21), pp.7571–6. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=1140450&tool=pmcentrez&r endertype=abstract. Guan, D. et al., 2013. The epigenetic regulator UHRF1 promotes ubiquitination-mediated degradation of the tumor-suppressor protein promyelocytic leukemia protein. Oncogene, 32(33), pp.3819–28. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3578017&tool=pmcentrez&r endertype=abstract [Accessed June 2, 2016]. Guan, D. & Kao, H.-Y., 2015. The function, regulation and therapeutic implications of the tumor suppressor protein, PML. Cell & bioscience, 5, p.60. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=4632682&tool=pmcentrez&r endertype=abstract. Guharoy, M. et al., 2016. Tripartite degrons confer diversity and specificity on regulated protein degradation in the ubiquitin-proteasome system. Nature Communications, 7, p.10239. Available at: http://www.nature.com/doifinder/10.1038/ncomms10239 [Accessed July 25, 2016]. Guo, A. et al., 2000. The function of PML in p53-dependent apoptosis. Nature cell biology, 2(10), pp.730–6. Available at: http://www.ncbi.nlm.nih.gov/pubmed/11025664 [Accessed June 6, 2016]. Guo, D. et al., 2004. A functional variant of SUMO4, a new I kappa B alpha modifier, is associated with type 1 diabetes. Nature genetics, 36(8), pp.837–41. Available at: http://www.ncbi.nlm.nih.gov/pubmed/15247916 [Accessed May 31, 2016]. Guo, L. et al., 2014. A cellular system that degrades misfolded proteins and protects against neurodegeneration. Molecular cell, 55(1), pp.15–30. Available at: http://www.ncbi.nlm.nih.gov/pubmed/24882209 [Accessed June 27, 2016]. Gurrieri, C., Capodieci, P., et al., 2004. Loss of the tumor suppressor PML in human cancers of multiple histologic origins. Journal of the National Cancer Institute, 96(4), pp.269– 79. Available at: http://www.ncbi.nlm.nih.gov/pubmed/14970276 [Accessed May 26, 2016]. Gurrieri, C., Nafa, K., et al., 2004. Mutations of the PML tumor suppressor gene in acute promyelocytic leukemia. Blood, 103(6), pp.2358–62. Available at: http://www.ncbi.nlm.nih.gov/pubmed/14630830 [Accessed June 8, 2016]. Häkli, M. et al., 2001. The RING Finger Protein SNURF Is a Bifunctional Protein Possessing DNA Binding Activity. Journal of Biological Chemistry, 276(26), pp.23653–23660. Han, Y. et al., 2010. SENP3-mediated De-conjugation of SUMO2/3 from Promyelocytic Leukemia Is Correlated with Accelerated Cell Proliferation under Mild Oxidative Stress. Journal of Biological Chemistry, 285(17), pp.12906–12915. Available at: http://www.jbc.org/cgi/doi/10.1074/jbc.M109.071431 [Accessed June 24, 2016]. Hands, K.J. et al., 2014. PML isoforms in response to arsenic: high-resolution analysis of PML body structure and degradation. Journal of cell science, 127(Pt 2), pp.365–75. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3889398&tool=pmcentrez&r endertype=abstract [Accessed November 14, 2013]. Hattersley, N. et al., 2011. The SUMO protease SENP6 is a direct regulator of PML nuclear bodies. Molecular biology of the cell, 22(1), pp.78–90. Available at: http://www.molbiolcell.org/cgi/doi/10.1091/mbc.E10-06-0504 [Accessed June 24, 2016].

177

Hay, R.T., 2005. SUMO: a history of modification. Molecular cell, 18(1), pp.1–12. Available at: http://www.ncbi.nlm.nih.gov/pubmed/15808504 [Accessed March 10, 2012]. Hayakawa, F. et al., 2008. Acetylation of PML is involved in histone deacetylase inhibitor- mediated apoptosis. Journal of Biological Chemistry, 283(36), pp.24420–24425. Available at: http://www.jbc.org/cgi/doi/10.1074/jbc.M802217200 [Accessed June 24, 2016]. Hayakawa, F. & Privalsky, M.L., 2004. Phosphorylation of PML by mitogen-activated protein kinases plays a key role in arsenic trioxide-mediated apoptosis. Cancer cell, 5(4), pp.389–401. Available at: http://www.ncbi.nlm.nih.gov/pubmed/15093545 [Accessed June 2, 2016]. Hendriks, I.A. & Vertegaal, A.C.O., 2016. A comprehensive compilation of SUMO proteomics. Nature Reviews Molecular Cell Biology. Available at: http://www.nature.com/doifinder/10.1038/nrm.2016.81. Hillestad, L.K., 1957. Acute promyelocytic leukemia. Acta medica Scandinavica, 159(3), pp.189–94. Available at: http://www.ncbi.nlm.nih.gov/pubmed/13508085 [Accessed June 20, 2016]. Hodges, M. et al., 1998. Structure, organization, and dynamics of promyelocytic leukemia protein nuclear bodies. American journal of human genetics, 63(2), pp.297–304. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=1377331&tool=pmcentrez&r endertype=abstract [Accessed May 31, 2016]. Hofmann, T.G. et al., 2002. Regulation of p53 activity by its interaction with homeodomain- interacting protein kinase-2. Nature cell biology, 4(1), pp.1–10. Available at: http://www.nature.com/doifinder/10.1038/ncb715 [Accessed June 5, 2016]. Hruz, T. et al., 2008. Genevestigator V3: A Reference Expression Database for the Meta- Analysis of Transcriptomes. Advances in Bioinformatics, 2008, pp.1–5. Available at: http://www.hindawi.com/journals/abi/2008/420747/ [Accessed June 17, 2016]. Huang, Q. et al., 2011. Caspase 3-mediated stimulation of tumor cell repopulation during cancer radiotherapy. Nature medicine, 17(7), pp.860–6. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3132290&tool=pmcentrez&r endertype=abstract [Accessed June 6, 2016]. Husnjak, K. & Dikic, I., 2012. Ubiquitin-Binding Proteins: Decoders of Ubiquitin-Mediated Cellular Functions. http://dx.doi.org/10.1146/annurev-biochem-051810-094654. Hussain, M. et al., 2016. Skp1: Implications in cancer and SCF-oriented anti-cancer drug discovery. Pharmacological research, 111, pp.34–42. Available at: http://www.ncbi.nlm.nih.gov/pubmed/27238229. Ikushima, H. & Miyazono, K., 2010. TGFβ signalling: a complex web in cancer progression. Nature Reviews Cancer, 10(6), pp.415–424. Available at: http://www.nature.com/doifinder/10.1038/nrc2853 [Accessed June 8, 2016]. Ito, K. et al., 2012. A PML–PPAR-δ pathway for fatty acid oxidation regulates hematopoietic stem cell maintenance. Nature medicine, 18(9), pp.1350–8. Available at: http://www.nature.com/doifinder/10.1038/nm.2882 [Accessed June 24, 2016]. Ivanschitz, L. et al., 2015. PML IV/ARF interaction enhances p53 SUMO-1 conjugation, activation, and senescence. Proceedings of the National Academy of Sciences, 112(46), pp.14278–14283. Available at: http://www.pnas.org/lookup/doi/10.1073/pnas.1507540112 [Accessed June 21, 2016]. Jeanne, M. et al., 2010. PML/RARA oxidation and arsenic binding initiate the antileukemia response of As2O3. Cancer cell, 18(1), pp.88–98. Available at: http://www.ncbi.nlm.nih.gov/pubmed/20609355 [Accessed May 16, 2012]. Jensen, K., Shiels, C. & Freemont, P.S., 2001. PML protein isoforms and the RBCC/TRIM

178

motif. Oncogene, 20(49), pp.7223–33. Available at: http://www.nature.com/doifinder/10.1038/sj.onc.1204765 [Accessed June 24, 2016]. Jensen, O.N., 2004. Modification-specific proteomics: Characterization of post-translational modifications by mass spectrometry. Current Opinion in Chemical Biology, 8(1), pp.33– 41. Jin, J. et al., 2004. Systematic analysis and nomenclature of mammalian F-box proteins. Genes & development, 18(21), pp.2573–80. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=525538&tool=pmcentrez&re ndertype=abstract [Accessed June 2, 2016]. de Jonge, H.J.M. et al., 2007. Evidence based selection of housekeeping genes. PLoS ONE, 2(9), pp.1–5. Josse, J. & Husson, F., 2016. missMDA: A Package for Handling Missing Values in Multivariate Data Analysis. Journal of Statistical Software, 70(1). Available at: http://www.jstatsoft.org/v70/i01/. Kahle, T. et al., 2015. TRIM19/PML Restricts HIV infection in a cell type-dependent manner. Viruses, 8(1), pp.1–18. Kakizuka, A. et al., 1991. Chromosomal translocation t(15;17) in human acute promyelocytic leukemia fuses RAR alpha with a novel putative transcription factor, PML. Cell, 66(4), pp.663–74. Available at: http://linkinghub.elsevier.com/retrieve/pii/009286749190112C [Accessed January 13, 2015]. Kamitani, T. et al., 1998. CELL BIOLOGY AND METABOLISM : Identification of Three Major Sentrinization Sites in PML Identification of Three Major Sentrinization Sites in PML *. , 273(41), pp.26675–26682. Kanarek, N. et al., 2010. Ubiquitination and Degradation of the Inhibitors of NF- B. Cold Spring Harbor Perspectives in Biology, 2(2), pp.a000166–a000166. Available at: http://cshperspectives.cshlp.org/lookup/doi/10.1101/cshperspect.a000166 [Accessed July 25, 2016]. Kandel, E.S. & Nudler, E., 2002. Template switching by RNA polymerase II in vivo. Evidence and implications from a retroviral system. Mol Cell, 10(6), pp.1495–1502. Kawasaki, A. et al., 2003. Opposing effects of PML and PML/RAR alpha on STAT3 activity. Blood, 101(9), pp.3668–73. Available at: http://www.ncbi.nlm.nih.gov/pubmed/12506013 [Accessed June 8, 2016]. Kay, B.K., Williamson, M.P. & Sudol, M., 2000. The importance of being proline: the interaction of proline-rich motifs in signaling proteins with their cognate domains. FASEB journal : official publication of the Federation of American Societies for Experimental Biology, 14(2), pp.231–41. Available at: http://www.ncbi.nlm.nih.gov/pubmed/10657980 [Accessed June 23, 2016]. Kent, W.J. et al., 2002. The Human Genome Browser at UCSC. Genome Research, 12(6), pp.996–1006. Available at: http://www.genome.org/cgi/doi/10.1101/gr.229102 [Accessed July 4, 2016]. Khan, M.M. et al., 2001. Role of PML and PML-RARα in Mad-Mediated Transcriptional Repression. Molecular Cell, 7(6), pp.1233–1243. Kim, A.Y. et al., 2008. SCCRO (DCUN1D1) is an essential component of the E3 complex for neddylation. The Journal of biological chemistry, 283(48), pp.33211–20. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=2586271&tool=pmcentrez&r endertype=abstract [Accessed June 3, 2016]. Kim, Y.S. et al., 2013. An Electrophoretic Mobility Shift Assay Identifies a Mechanistically Unique Inhibitor of Protein Sumoylation. Chemistry & Biology, 20(4), pp.604–613. Kirkpatrick, D.S. et al., 2006. Quantitative analysis of in vitro ubiquitinated cyclin B1 reveals complex chain topology. Nature cell biology, 8(7), pp.700–10. Available at:

179

http://www.ncbi.nlm.nih.gov/pubmed/16799550 [Accessed June 2, 2016]. Koidl, S. et al., 2016. The SUMO2/3 specific E3 ligase ZNF451-1 regulates PML stability. The International Journal of Biochemistry & Cell Biology, pp.1–10. Available at: http://linkinghub.elsevier.com/retrieve/pii/S1357272516301510. Koken, M.H. et al., 1995. The PML growth-suppressor has an altered expression in human oncogenesis. Oncogene, 10(7), pp.1315–24. Available at: http://www.ncbi.nlm.nih.gov/pubmed/7731682 [Accessed June 6, 2016]. Kolde, R., 2015. pheatmap: Pretty Heatmaps. R package version 1.0.8. Available at: https://cran.r-project.org/package=pheatmap. Komander, D., Clague, M.J. & Urbé, S., 2009. Breaking the chains: structure and function of the deubiquitinases. Nature reviews. Molecular cell biology, 10(8), pp.550–63. Available at: http://www.ncbi.nlm.nih.gov/pubmed/19626045 [Accessed June 25, 2016]. Komander, D. & Rape, M., 2012. The Ubiquitin Code. http://dx.doi.org/10.1146/annurev- biochem-060310-170328. Korb, E. & Finkbeiner, S., 2013. PML in the Brain: From Development to Degeneration. Frontiers in oncology, 3(September), p.242. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3775456&tool=pmcentrez&r endertype=abstract [Accessed June 7, 2016]. Korf, K. et al., 2014. The PML domain of PML-RARα blocks senescence to promote leukemia. Proceedings of the National Academy of Sciences of the United States of America, 111(33), pp.12133–12138. Available at: http://www.ncbi.nlm.nih.gov/pubmed/25092303. Kuchay, S. et al., 2013. FBXL2- and PTPL1-mediated degradation of p110-free p85β regulatory subunit controls the PI(3)K signalling cascade. Nature cell biology, 15(5), pp.472–80. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3865866&tool=pmcentrez&r endertype=abstract [Accessed June 5, 2016]. Kumar, A. et al., 2014. Identification of sumoylation inhibitors targeting a predicted pocket in Ubc9. Journal of chemical information and modeling, 54(10), pp.2784–93. Available at: http://www.ncbi.nlm.nih.gov/pubmed/25191977 [Accessed July 5, 2016]. Lallemand-Breitenbach, V. et al., 2008. Arsenic degrades PML or PML-RARalpha through a SUMO-triggered RNF4/ubiquitin-mediated pathway. Nature cell biology, 10(5), pp.547– 555. Available at: http://www.ncbi.nlm.nih.gov/pubmed/18408733 [Accessed March 23, 2012]. Lallemand-Breitenbach, V., de The, H. & de Thé, H., 2010. PML nuclear bodies. Cold Spring Harbor perspectives in biology, 2(5), p.a000661. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=2857171&tool=pmcentrez&r endertype=abstract [Accessed March 12, 2012]. Lane, A.A. & Ley, T.J., 2003. Neutrophil Elastase Cleaves PML-RARα and Is Important for the Development of Acute Promyelocytic Leukemia in Mice. Cell, 115(3), pp.305–318. Lane, D. & Levine, A., 2010. p53 Research: the past thirty years and the next thirty years. Cold Spring Harbor perspectives in biology, 2(12), p.a000893. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=2982174&tool=pmcentrez&r endertype=abstract [Accessed June 6, 2016]. Langley, E. et al., 2002. Human SIR2 deacetylates p53 and antagonizes PML/p53-induced cellular senescence. The EMBO journal, 21(10), pp.2383–96. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=126010&tool=pmcentrez&re ndertype=abstract [Accessed June 6, 2016]. Le, S. et al., 1999. RAD50 and RAD51 define two pathways that collaborate to maintain telomeres in the absence of telomerase. Genetics, 152(1), pp.143–52. Available at:

180

http://www.ncbi.nlm.nih.gov/pubmed/10224249 [Accessed June 27, 2016]. Lê, S., Josse, J. & Husson, F., 2008. FactoMineR: An R Package for Multivariate Analysis. J. of Statistical Software, 25(1), pp.1–18. Available at: http://www.jstatsoft.org/v25/i01. Lee, A. et al., 2015. Post-translational Modifications in Heart Failure : Small Changes , Big Impact. , pp.1–6. Lee, E.K. & Diehl, J. a, 2013. SCFs in the new millennium. Oncogene, (January), pp.1–8. Available at: http://www.ncbi.nlm.nih.gov/pubmed/23624913 [Accessed November 14, 2013]. Lee, H.E. et al., 2007. Loss of promyelocytic leukemia protein in human gastric cancers. Cancer letters, 247(1), pp.103–9. Available at: http://www.ncbi.nlm.nih.gov/pubmed/16713073 [Accessed June 6, 2016]. Lee, K.W. et al., 2016. F-box only protein 9 is an E3 ubiquitin ligase of PPARγ. Experimental & molecular medicine, 48, p.e234. Available at: http://www.ncbi.nlm.nih.gov/pubmed/27197753 [Accessed June 27, 2016]. Lee, K.W. et al., 2013. F-box only protein 9 is required for adipocyte differentiation. Biochemical and biophysical research communications, 435(2), pp.239–43. Available at: http://www.ncbi.nlm.nih.gov/pubmed/23643813 [Accessed November 14, 2013]. Lehmann, B.D.B. et al., 2011. Identification of human triple-negative breast cancer subtypes and preclinical models for selection of targeted therapies. Journal of Clinical Investigation, 121(7), pp.2750–2767. Available at: http://www.ncbi.nlm.nih.gov/pubmed/21633166\nhttp://www.pubmedcentral.nih.gov/art iclerender.fcgi?artid=3127435&tool=pmcentrez&rendertype=abstract\nhttp://www.ncbi. nlm.nih.gov/pmc/articles/pmc3127435/. Li, H. et al., 2000. Sequestration and Inhibition of Daxx-Mediated Transcriptional Repression by PML. Molecular and Cellular Biology, 20(5), pp.1784–1796. Available at: http://mcb.asm.org/cgi/doi/10.1128/MCB.20.5.1784-1796.2000 [Accessed June 8, 2016]. Li, Q. et al., 2011. AXIN is an essential co-activator for the promyelocytic leukemia protein in p53 activation. Oncogene, 30(10), pp.1194–204. Available at: http://www.ncbi.nlm.nih.gov/pubmed/21057547 [Accessed June 6, 2016]. Li, W. et al., 2009. PML depletion disrupts normal mammary gland development and skews the composition of the mammary luminal cell progenitor pool. Proceedings of the National Academy of Sciences of the United States of America, 106(12), pp.4725–30. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=2660741&tool=pmcentrez&r endertype=abstract. Liang, Y.-C. et al., 2016. SUMO5, a Novel Poly-SUMO Isoform, Regulates PML Nuclear Bodies. Scientific Reports, 6, p.26509. Available at: http://www.nature.com/articles/srep26509 [Accessed June 9, 2016]. Licht, B.J.D. et al., 1995. Clinical and Molecular Characterization. Blood, 85(4), pp.1083– 1095. Available at: http://www.ncbi.nlm.nih.gov/pubmed/7849296. Lim, J.H. et al., 2011. Mitogen-activated protein kinase extracellular signal-regulated kinase 2 phosphorylates and promotes Pin1 protein-dependent promyelocytic leukemia protein turnover. The Journal of biological chemistry, 286(52), pp.44403–11. Available at: http://www.jbc.org/cgi/doi/10.1074/jbc.M111.289512 [Accessed June 23, 2016]. Lin, D. et al., 2002. Identification of a substrate recognition site on Ubc9. The Journal of biological chemistry, 277(24), pp.21740–8. Available at: http://www.ncbi.nlm.nih.gov/pubmed/11877416 [Accessed May 31, 2016]. Lin, D.-Y. et al., 2003. Promyelocytic Leukemia Protein (PML) Functions as a Glucocorticoid Receptor Co-activator by Sequestering Daxx to the PML Oncogenic Domains (PODs) to Enhance Its Transactivation Potential. Journal of Biological Chemistry, 278(18),

181

pp.15958–15965. Available at: http://www.jbc.org/cgi/doi/10.1074/jbc.M300387200 [Accessed June 8, 2016]. Lin, D.-Y. et al., 2006. Role of SUMO-Interacting Motif in Daxx SUMO Modification, Subnuclear Localization, and Repression of Sumoylated Transcription Factors. Molecular Cell, 24(3), pp.341–354. Available at: http://linkinghub.elsevier.com/retrieve/pii/S109727650600709X [Accessed June 8, 2016]. Lin, H.-K., Bergmann, S. & Pandolfi, P.P., 2004. Cytoplasmic PML function in TGF-β signalling. Nature, 431(7005), pp.205–211. Available at: http://www.nature.com/doifinder/10.1038/nature02783 [Accessed June 8, 2016]. Liu, J. et al., 2002. NEDD8 modification of CUL1 dissociates p120(CAND1), an inhibitor of CUL1-SKP1 binding and SCF ligases. Molecular cell, 10(6), pp.1511–8. Available at: http://www.ncbi.nlm.nih.gov/pubmed/12504025 [Accessed June 3, 2016]. Liu, Y. et al., 1998. Proto-oncogene PML controls genes devoted to MHC class I antigen presentation. Nature, 396(6709), pp.373–376. Available at: http://www.nature.com/doifinder/10.1038/24628 [Accessed June 8, 2016]. Liu, Y. et al., 2015. Skp1 in lung cancer : Clinical significance and therapeutic efficacy of its small molecule inhibitors. Oncotarget, 6(33), pp.34953–67. Livak, K.J. & Schmittgen, T.D., 2001. Analysis of relative gene expression data using real- time quantitative PCR and the 2(-Delta Delta C(T)) Method. Methods (San Diego, Calif.), 25(4), pp.402–8. Available at: http://www.ncbi.nlm.nih.gov/pubmed/11846609. Lo-Coco, F. & Hasan, S.K., 2014. Understanding the molecular pathogenesis of acute promyelocytic leukemia. Best practice & research. Clinical haematology, 27(1), pp.3–9. Available at: http://www.sciencedirect.com/science/article/pii/S1521692614000243. Louria-Hayon, I. et al., 2003. The promyelocytic leukemia protein protects p53 from Mdm2- mediated inhibition and degradation. The Journal of biological chemistry, 278(35), pp.33134–41. Available at: http://www.jbc.org/cgi/doi/10.1074/jbc.M301264200 [Accessed June 6, 2016]. Lyapina, S. et al., 2001. Promotion of NEDD-CUL1 conjugate cleavage by COP9 signalosome. Science (New York, N.Y.), 292(5520), pp.1382–5. Available at: http://www.ncbi.nlm.nih.gov/pubmed/11337588 [Accessed June 3, 2016]. Lydeard, J.R., Schulman, B.A. & Harper, J.W., 2013. Building and remodelling Cullin-RING E3 ubiquitin ligases. EMBO reports, 14(12), pp.1050–61. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3849489&tool=pmcentrez&r endertype=abstract [Accessed June 2, 2016]. Malumbres, M. & Barbacid, M., 2009. Cell cycle, CDKs and cancer: a changing paradigm. Nature reviews. Cancer, 9(3), pp.153–166. Mao, Y.S., Zhang, B. & Spector, D.L., 2011. Biogenesis and function of nuclear bodies. Trends in genetics : TIG, 27(8), pp.295–306. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3144265&tool=pmcentrez&r endertype=abstract [Accessed May 23, 2016]. Maroui, M.A. et al., 2012. Requirement of PML SUMO interacting motif for RNF4- or arsenic trioxide-induced degradation of nuclear PML isoforms. PloS one, 7(9), p.e44949. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3445614&tool=pmcentrez&r endertype=abstract [Accessed November 14, 2013]. Martens, J.H.A. et al., 2010. PML-RAR??/RXR Alters the Epigenetic Landscape in Acute Promyelocytic Leukemia. Cancer Cell, 17(2), pp.173–185. Available at: http://dx.doi.org/10.1016/j.ccr.2009.12.042. Matunis, M.J., Coutavas, E. & Blobel, G., 1996. A novel ubiquitin-like modification

182

modulates the partitioning of the Ran-GTPase-activating protein RanGAP1 between the cytosol and the nuclear pore complex. The Journal of cell biology, 135(6 Pt 1), pp.1457– 70. Available at: http://www.ncbi.nlm.nih.gov/pubmed/8978815 [Accessed June 24, 2016]. Mazza, M. & Pelicci, P.G., 2013. Is PML a Tumor Suppressor? Frontiers in oncology, 3(July), p.174. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3705425&tool=pmcentrez&r endertype=abstract [Accessed January 6, 2015]. McConnell, M.J. et al., 2015. Post transcriptional control of the epigenetic stem cell regulator PLZF by sirtuin and HDAC deacetylases. Epigenetics & chromatin, 8, p.38. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=4581162&tool=pmcentrez&r endertype=abstract. McNally, B.A. et al., 2008. A role for cytoplasmic PML in cellular resistance to viral infection. PloS one, 3(5), p.e2277. Available at: http://www.ncbi.nlm.nih.gov/pubmed/18509536 [Accessed June 21, 2016]. Meinecke, I. et al., 2007. Modification of nuclear PML protein by SUMO-1 regulates Fas- induced apoptosis in rheumatoid arthritis synovial fibroblasts. Proceedings of the National Academy of Sciences, 104(12), pp.5073–5078. Available at: http://www.pnas.org/cgi/doi/10.1073/pnas.0608773104 [Accessed June 24, 2016]. Mello, C. et al., 2002. E3 ubiquitin ligase that recognizes sugar chains. Nature., 418(July), pp.438–442. Mi, J.-Q. et al., 2015. Synergistic targeted therapy for acute promyelocytic leukaemia: a model of translational research in human cancer. Journal of Internal Medicine, p.n/a–n/a. Available at: http://doi.wiley.com/10.1111/joim.12376. Miki, T. et al., 2012. PML regulates PER2 nuclear localization and circadian function. The EMBO journal, 31(6), pp.1427–39. Available at: http://www.ncbi.nlm.nih.gov/pubmed/22274616 [Accessed June 24, 2016]. Miki, T., Zhao, Z. & Lee, C.C., 2016. Interactive Organization of the Circadian Core Regulators PER2, BMAL1, CLOCK and PML. Scientific reports, 6, p.29174. Available at: http://www.ncbi.nlm.nih.gov/pubmed/27383066 [Accessed July 8, 2016]. Morgan, D.O., 2007. The Cell Cycle: Principles of control, Corby, United Kingdom: Oxford University Press. Mu, Z.M. et al., 1997. Stable overexpression of PML alters regulation of cell cycle progression in HeLa cells. Carcinogenesis, 18(11), pp.2063–9. Available at: http://www.carcin.oupjournals.org/cgi/doi/10.1093/carcin/18.11.2063 [Accessed June 2, 2016]. Mukhopadhyay, D. et al., 2006. SUSP1 antagonizes formation of highly SUMO2/3- conjugated species. The Journal of Cell Biology, 174(7), pp.939–949. Available at: http://www.jcb.org/lookup/doi/10.1083/jcb.200510103 [Accessed June 24, 2016]. Mukhopadhyay, D. & Dasso, M., 2007. Modification in reverse: the SUMO proteases. Trends in biochemical sciences, 32(6), pp.286–95. Available at: http://www.ncbi.nlm.nih.gov/pubmed/17499995 [Accessed March 30, 2012]. Muller, E. et al., 2015. Genetic profiles of cervical tumors by high-throughput sequencing for personalized medical care. Cancer Medicine, 4(10), pp.1484–1493. Available at: http://doi.wiley.com/10.1002/cam4.492 [Accessed July 24, 2016]. Müller, S., Matunis, M.J. & Dejean, A., 1998. Conjugation with the ubiquitin-related modifier SUMO-1 regulates the partitioning of PML within the nucleus. The EMBO journal, 17(1), pp.61–70. Available at: http://emboj.embopress.org/cgi/doi/10.1093/emboj/17.1.61 [Accessed June 5, 2012].

183

Müller, S., Miller, W.H. & Dejean, a, 1998. Trivalent antimonials induce degradation of the PML-RAR oncoprotein and reorganization of the promyelocytic leukemia nuclear bodies in acute promyelocytic leukemia NB4 cells. Blood, 92(11), pp.4308–16. Available at: http://www.ncbi.nlm.nih.gov/pubmed/9834237. Munir, M. et al., 2010. TRIM proteins: another class of viral victims. Science signaling, 3(118), p.jc2. Available at: http://www.ncbi.nlm.nih.gov/pubmed/20407122 [Accessed July 9, 2016]. Nacerddine, K. et al., 2005. The SUMO pathway is essential for nuclear integrity and chromosome segregation in mice. Developmental cell, 9(6), pp.769–79. Available at: http://www.ncbi.nlm.nih.gov/pubmed/16326389 [Accessed June 6, 2016]. Nakahara, F., Weiss, C.N. & Ito, K., 2014. The role of PML in hematopoietic and leukemic stem cell maintenance. International Journal of Hematology, 100(1), pp.18–26. Nakayama, K.I. & Nakayama, K., 2006. Ubiquitin ligases: cell-cycle control and cancer. Nature reviews. Cancer, 6(5), pp.369–81. Available at: http://www.nature.com/doifinder/10.1038/nrc1881 [Accessed June 9, 2016]. Nelson, V., Davis, G.E. & Maxwell, S.A., 2001. A putative protein inhibitor of activated STAT (PIASy) interacts with p53 and inhibits p53-mediated transactivation but not apoptosis. Apoptosis : an international journal on programmed cell death, 6(3), pp.221– 34. Available at: http://www.ncbi.nlm.nih.gov/pubmed/11388671 [Accessed June 24, 2016]. Neuwirth, E., 2014. RColorBrewer: ColorBrewer Palettes. R package version 1.1-2. Available at: https://cran.r-project.org/package=RColorBrewer. Neyret-Kahn, H., 2012. Etude du role de la voie SUMO sur la chromatine et dans l’oncogenese. (Unpublished doctoral thesis). Univerité Pierre et Marie Curie, Paris, France. Nguyen, L.A. et al., 2005. Physical and functional link of the leukemia-associated factors AML1 and PML. Blood, 105(1), pp.292–300. Available at: http://www.ncbi.nlm.nih.gov/pubmed/15331439 [Accessed June 21, 2016]. Nisole, S. et al., 2013. Differential Roles of PML Isoforms. Frontiers in oncology, 3(May), p.125. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3660695&tool=pmcentrez&r endertype=abstract [Accessed January 20, 2014]. Ohbayashi, N. et al., 2008. The IL-6 family of cytokines modulates STAT3 activation by desumoylation of PML through SENP1 induction. Biochemical and Biophysical Research Communications, 371(4), pp.823–828. Available at: http://linkinghub.elsevier.com/retrieve/pii/S0006291X08008644 [Accessed June 24, 2016]. Olsen, J. V et al., 2006. Global, in vivo, and site-specific phosphorylation dynamics in signaling networks. Cell, 127(3), pp.635–48. Available at: http://www.ncbi.nlm.nih.gov/pubmed/17081983 [Accessed June 23, 2016]. Owerbach, D. et al., 2005. A proline-90 residue unique to SUMO-4 prevents maturation and sumoylation. Biochemical and biophysical research communications, 337(2), pp.517–20. Available at: http://www.ncbi.nlm.nih.gov/pubmed/16198310 [Accessed April 16, 2016]. Petroski, M.D. & Deshaies, R.J., 2005. Function and regulation of cullin-RING ubiquitin ligases. Nature reviews. Molecular cell biology, 6(1), pp.9–20. Available at: http://www.ncbi.nlm.nih.gov/pubmed/15688063 [Accessed July 21, 2015]. Phosphonet, 2016. Human Phosphosite KnowledgeBase. Kinexus Bioinformatics Corporation. Available at: http://www.phosphonet.ca/. PhosphoSitePlus, 2016. PML known phosphorylation sites. Cell Signaling Technology, Inc.

184

Available at: http://www.phosphosite.org/proteinAction.action?id=5135&showAllSites=true. Pichler, A. et al., 2002. The nucleoporin RanBP2 has SUMO1 E3 ligase activity. Cell, 108(1), pp.109–120. Available at: http://linkinghub.elsevier.com/retrieve/pii/S009286740100633X [Accessed June 24, 2016]. Pierce, N.W. et al., 2013. Cand1 promotes assembly of new SCF complexes through dynamic exchange of F box proteins. Cell, 153(1), pp.206–215. Available at: http://dx.doi.org/10.1016/j.cell.2013.02.024 [Accessed November 14, 2013]. Pierce, N.W. et al., 2009. Detection of sequential polyubiquitylation on a millisecond timescale. Nature, 462(7273), pp.615–9. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=2791906&tool=pmcentrez&r endertype=abstract [Accessed June 2, 2016]. Pontén, F., Jirström, K. & Uhlen, M., 2008. The Human Protein Atlas--a tool for pathology. The Journal of pathology, 216(4), pp.387–93. Available at: http://www.ncbi.nlm.nih.gov/pubmed/18853439 [Accessed July 26, 2016]. Psakhye, I. & Jentsch, S., 2012. Protein group modification and synergy in the SUMO pathway as exemplified in DNA repair. Cell, 151(4), pp.807–20. Available at: http://www.ncbi.nlm.nih.gov/pubmed/23122649 [Accessed June 24, 2016]. Quignon, F. et al., 1998. PML induces a novel caspase-independent death process. Nature genetics, 20(3), pp.259–65. Available at: http://www.ncbi.nlm.nih.gov/pubmed/9806544 [Accessed June 6, 2016]. R Core Team, 2015. R: A Language and Environment for Statistical Computing. Available at: http://www.r-project.org/. Rabellino, A. et al., 2012. The SUMO E3-ligase PIAS1 regulates the tumor suppressor PML and its oncogenic counterpart PML-RARA. Cancer research, 72(9), pp.2275–84. Available at: http://cancerres.aacrjournals.org/cgi/doi/10.1158/0008-5472.CAN-11-3159 [Accessed June 1, 2016]. Randle, S.J. & Laman, H., 2015. F-box protein interactions with the hallmark pathways in cancer. Seminars in Cancer Biology, 36, pp.3–17. Available at: http://dx.doi.org/10.1016/j.semcancer.2015.09.013. Regad, T. et al., 2009. The tumor suppressor Pml regulates cell fate in the developing neocortex. Nature neuroscience, 12(2), pp.132–40. Available at: http://www.ncbi.nlm.nih.gov/pubmed/19136970 [Accessed March 23, 2012]. Rego, E.M. et al., 2001. Role of promyelocytic leukemia (PML) protein in tumor suppression. The Journal of experimental medicine, 193(4), pp.521–29. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=2195907&tool=pmcentrez&r endertype=abstract [Accessed June 6, 2016]. Reineke, E.L. et al., 2008. Degradation of the Tumor Suppressor PML by Pin1 Contributes to the Cancer Phenotype of Breast Cancer MDA-MB-231 Cells. Molecular and Cellular Biology, 28(3), pp.997–1006. Available at: http://mcb.asm.org/cgi/doi/10.1128/MCB.01848-07 [Accessed June 24, 2016]. Reineke, E.L. & Kao, H.-Y., 2009. Targeting promyelocytic leukemia protein: a means to regulating PML nuclear bodies. International journal of biological sciences, 5(4), pp.366–76. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=2686094&tool=pmcentrez&r endertype=abstract. Reymond, A. et al., 2001. The tripartite motif family identifies cell compartments. EMBO Journal, 20(9), pp.2140–2151. Available at: http://emboj.embopress.org/cgi/doi/10.1093/emboj/20.9.2140 [Accessed June 21, 2016].

185

Rodriguez, M.S., Dargemont, C. & Hay, R.T., 2001. SUMO-1 conjugation in vivo requires both a consensus modification motif and nuclear targeting. The Journal of biological chemistry, 276(16), pp.12654–9. Available at: http://www.ncbi.nlm.nih.gov/pubmed/11124955 [Accessed June 24, 2016]. Rossi, M. et al., 2013. Regulation of the CRL4(Cdt2) ubiquitin ligase and cell-cycle exit by the SCF(Fbxo11) ubiquitin ligase. Molecular cell, 49(6), pp.1159–66. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3624904&tool=pmcentrez&r endertype=abstract [Accessed June 5, 2016]. Saha, A. & Deshaies, R.J., 2008. Multimodal activation of the ubiquitin ligase SCF by Nedd8 conjugation. Molecular cell, 32(1), pp.21–31. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=2644375&tool=pmcentrez&r endertype=abstract [Accessed June 3, 2016]. Sahin Umut, U. et al., 2014. Oxidative stress-induced assembly of PML nuclear bodies controls sumoylation of partner proteins. The Journal of cell biology, 204(6), pp.931–45. Available at: http://www.ncbi.nlm.nih.gov/pubmed/24637324 [Accessed March 19, 2014]. Sahin, U. et al., 2014. PML nuclear bodies: regulation, function and therapeutic perspectives. The Journal of pathology, 234(3), pp.289–291. Available at: http://www.ncbi.nlm.nih.gov/pubmed/25138686 [Accessed August 22, 2014]. Sahin, U., de Thé, H. & Lallemand-Breitenbach, V., 2015. PML nuclear bodies: Assembly and oxidative stress-sensitive sumoylation. Nucleus, 5(6), pp.499–507. Available at: http://www.tandfonline.com/doi/abs/10.4161/19491034.2014.970104. Saitoh, H. & Hinchey, J., 2000. Functional heterogeneity of small ubiquitin-related protein modifiers SUMO-1 versus SUMO-2/3. The Journal of biological chemistry, 275(9), pp.6252–8. Available at: http://www.ncbi.nlm.nih.gov/pubmed/10692421. Saitoh, N. et al., 2006. In situ SUMOylation analysis reveals a modulatory role of RanBP2 in the nuclear rim and PML bodies. Experimental Cell Research, 312(8), pp.1418–1430. Available at: http://linkinghub.elsevier.com/retrieve/pii/S0014482706000061 [Accessed June 24, 2016]. Sakamoto, K.M. et al., 2003. Development of Protacs to target cancer-promoting proteins for ubiquitination and degradation. Molecular & cellular proteomics : MCP, 2(12), pp.1350–8. Available at: http://www.ncbi.nlm.nih.gov/pubmed/14525958 [Accessed July 6, 2016]. Sakamoto, K.M. et al., 2001. Protacs: chimeric molecules that target proteins to the Skp1- Cullin-F box complex for ubiquitination and degradation. Proceedings of the National Academy of Sciences of the United States of America, 98(15), pp.8554–9. Available at: http://www.ncbi.nlm.nih.gov/pubmed/11438690 [Accessed July 6, 2016]. Salomoni, P. et al., 2005. The promyelocytic leukemia protein PML regulates c-Jun function in response to DNA damage. Blood, 105(9), pp.3686–90. Available at: http://www.ncbi.nlm.nih.gov/pubmed/15626733 [Accessed June 8, 2016]. Salomoni, P., Dvorkina, M. & Michod, D., 2012. Role of the promyelocytic leukaemia protein in cell death regulation. Cell death & disease, 3, p.e247. Available at: http://www.ncbi.nlm.nih.gov/pubmed/22237204 [Accessed June 7, 2016]. Salomoni, P. & Pandolfi, P.P., 2002. The role of PML in tumor suppression. Cell, 108(2), pp.165–70. Available at: http://www.ncbi.nlm.nih.gov/pubmed/11832207 [Accessed May 31, 2016]. Dos Santos, G.A., Kats, L. & Pandolfi, P.P., 2013. Synergy against PML-RARa: targeting transcription, proteolysis, differentiation, and self-renewal in acute promyelocytic leukemia. The Journal of experimental medicine, 210(13), pp.2793–802. Available at: http://www.ncbi.nlm.nih.gov/pubmed/24344243.

186

Sarikas, A. et al., 2008. The cullin7 E3 ubiquitin ligase: a novel player in growth control. Cell cycle (Georgetown, Tex.), 7(20), pp.3154–61. Available at: http://www.tandfonline.com/doi/abs/10.4161/cc.7.20.6922 [Accessed July 24, 2016]. Sarkari, F. et al., 2011. The Herpesvirus Associated Ubiquitin Specific Protease, USP7, Is a Negative Regulator of PML Proteins and PML Nuclear Bodies D. Harrich, ed. PLoS ONE, 6(1), p.e16598. Available at: http://dx.plos.org/10.1371/journal.pone.0016598 [Accessed June 25, 2016]. Sarno, S. et al., 2001. Selectivity of 4,5,6,7-tetrabromobenzotriazole, an ATP site-directed inhibitor of protein kinase CK2 (“casein kinase-2”). FEBS Letters, 496(1), pp.44–48. Available at: http://doi.wiley.com/10.1016/S0014-5793%2801%2902404-8 [Accessed July 4, 2016]. Satow, R. et al., 2012. β-Catenin Inhibits Promyelocytic Leukemia Protein Tumor Suppressor Function in Colorectal Cancer Cells. Gastroenterology, 142(3), pp.572–581. Available at: http://linkinghub.elsevier.com/retrieve/pii/S0016508511016398 [Accessed June 24, 2016]. Scaglioni, P.P. et al., 2006. A CK2-dependent mechanism for degradation of the PML tumor suppressor. Cell, 126(2), pp.269–83. Available at: http://www.ncbi.nlm.nih.gov/pubmed/16873060 [Accessed December 31, 2013]. Schulz, S. et al., 2012. Ubiquitin-specific protease-like 1 (USPL1) is a SUMO isopeptidase with essential, non-catalytic functions. EMBO reports, 13(10), pp.930–938. Available at: http://dx.doi.org/10.1038/embor.2012.125. Seeler, J.S. & Dejean, A., 1999. The PML nuclear bodies: actors or extras? Current opinion in genetics & development, 9(3), pp.362–7. Available at: http://www.ncbi.nlm.nih.gov/pubmed/10377280 [Accessed May 31, 2016]. Seo, S.R. et al., 2006. Nuclear Retention of the Tumor Suppressor cPML by the Homeodomain Protein TGIF Restricts TGF-?? Signaling. Molecular Cell, 23(4), pp.547– 559. Shen, T.H. et al., 2006. The mechanisms of PML-nuclear body formation. Molecular cell, 24(3), pp.331–9. Available at: http://linkinghub.elsevier.com/retrieve/pii/S1097276506006617 [Accessed May 31, 2016]. Shima, Y. et al., 2008. PML activates transcription by protecting HIPK2 and p300 from SCFFbx3-mediated degradation. Molecular and cellular biology, 28(23), pp.7126–38. Available at: http://www.ncbi.nlm.nih.gov/pubmed/18809579 [Accessed June 26, 2016]. Shimada, N., Shinagawa, T. & Ishii, S., 2008. Modulation of M2-type pyruvate kinase activity by the cytoplasmic PML tumor suppressor protein. Genes to cells : devoted to molecular & cellular mechanisms, 13(3), pp.245–54. Available at: http://doi.wiley.com/10.1111/j.1365-2443.2008.01165.x [Accessed June 8, 2016]. Shiraishi, S. et al., 2007. TBP-interacting Protein 120B (TIP120B)/Cullin-associated and Neddylation-dissociated 2 (CAND2) Inhibits SCF-dependent Ubiquitination of Myogenin and Accelerates Myogenic Differentiation. Journal of Biological Chemistry, 282(12), pp.9017–9028. Available at: http://www.jbc.org/cgi/doi/10.1074/jbc.M611513200 [Accessed July 25, 2016]. Shire, K. et al., 2016. Identification of RNF168 as a PML nuclear body regulator. Journal of Cell Science, 129(3), pp.580–591. Available at: http://jcs.biologists.org/content/129/3/580.long. Shtutman, M. et al., 2002. PML is a target gene of beta-catenin and plakoglobin, and coactivates beta-catenin-mediated transcription. Cancer research, 62(20), pp.5947–54. Available at: http://www.ncbi.nlm.nih.gov/pubmed/12384561 [Accessed June 8, 2016]. Sinha, N. et al., 2016. SUMO4 163 G>A variation is associated with kidney disease in Indian

187

subjects with type 2 diabetes. Molecular biology reports, 43(5), pp.345–8. Available at: http://www.ncbi.nlm.nih.gov/pubmed/27055882 [Accessed May 31, 2016]. Sivachandran, N., Sarkari, F. & Frappier, L., 2008. Epstein-Barr Nuclear Antigen 1 Contributes to Nasopharyngeal Carcinoma through Disruption of PML Nuclear Bodies J. U. Jung, ed. PLoS Pathogens, 4(10), p.e1000170. Available at: http://dx.plos.org/10.1371/journal.ppat.1000170 [Accessed June 25, 2016]. Skaar, J.R. et al., 2009. SnapShot: F Box Proteins II. Cell, 137(7), pp.1358, 1358.e1. Available at: http://www.ncbi.nlm.nih.gov/pubmed/19563764 [Accessed June 2, 2016]. Skaar, J.R., Pagan, J.K. & Pagano, M., 2013. Mechanisms and function of substrate recruitment by F-box proteins. Nature reviews. Molecular cell biology, 14(6), pp.369– 81. Available at: http://www.ncbi.nlm.nih.gov/pubmed/23657496 [Accessed November 12, 2013]. Skaar, J.R., Pagan, J.K. & Pagano, M., 2014. SCF ubiquitin ligase-targeted therapies. Nature reviews. Drug discovery, 13(12), pp.889–903. Available at: http://dx.doi.org/10.1038/nrd4432. Song, J. et al., 2004. Identification of a SUMO-binding motif that recognizes SUMO- modified proteins. Proceedings of the National Academy of Sciences of the United States of America, 101(40), pp.14373–8. Available at: http://www.ncbi.nlm.nih.gov/pubmed/15388847 [Accessed June 24, 2016]. Song, J. et al., 2005. Small ubiquitin-like modifier (SUMO) recognition of a SUMO binding motif: a reversal of the bound orientation. The Journal of biological chemistry, 280(48), pp.40122–9. Available at: http://www.ncbi.nlm.nih.gov/pubmed/16204249 [Accessed June 24, 2016]. Song, M.S. et al., 2008. The deubiquitinylation and localization of PTEN are regulated by a HAUSP-PML network. Nature, 455(7214), pp.813–7. Available at: http://www.ncbi.nlm.nih.gov/pubmed/18716620 [Accessed June 7, 2016]. Spruck, C. et al., 2001. A CDK-independent function of mammalian Cks1: targeting of SCF(Skp2) to the CDK inhibitor p27Kip1. Molecular cell, 7(3), pp.639–50. Available at: http://www.ncbi.nlm.nih.gov/pubmed/11463388 [Accessed June 5, 2016]. Stehmeier, P. & Muller, S., 2009. Phospho-regulated SUMO interaction modules connect the SUMO system to CK2 signaling. Molecular cell, 33(3), pp.400–9. Available at: http://www.ncbi.nlm.nih.gov/pubmed/19217413 [Accessed December 19, 2013]. Stewart, B.W. & Wild, C., 2014. World cancer report 2014, Swatek, K.N. & Komander, D., 2016. Ubiquitin modifications. Cell research, 26(4), pp.399– 422. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=4822133&tool=pmcentrez&r endertype=abstract. Szklarczyk, D. et al., 2015. STRING v10: protein-protein interaction networks, integrated over the tree of life. Nucleic acids research, 43(Database issue), pp.D447–52. Available at: http://www.ncbi.nlm.nih.gov/pubmed/25352553 [Accessed June 14, 2016]. Szostecki, C. et al., 1990. Isolation and characterization of cDNA encoding a human nuclear antigen predominantly recognized by autoantibodies from patients with primary biliary cirrhosis. Journal of immunology (Baltimore, Md. : 1950), 145(12), pp.4338–47. Available at: http://www.ncbi.nlm.nih.gov/pubmed/2258622 [Accessed June 22, 2016]. Tanaka, T., Nakatani, T. & Kamitani, T., 2012. Inhibition of NEDD8-conjugation pathway by novel molecules: potential approaches to anticancer therapy. Molecular oncology, 6(3), pp.267–75. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3826113&tool=pmcentrez&r endertype=abstract [Accessed June 3, 2016]. Tang, M.K. et al., 2013. Promyelocytic Leukemia (PML) Protein Plays Important Roles in

188

Regulating Cell Adhesion, Morphology, Proliferation and Migration N. A. Hotchin, ed. PLoS ONE, 8(3), p.e59477. Available at: http://dx.plos.org/10.1371/journal.pone.0059477 [Accessed July 12, 2016]. Tatham, M.H. et al., 2001. Polymeric chains of SUMO-2 and SUMO-3 are conjugated to protein substrates by SAE1/SAE2 and Ubc9. The Journal of biological chemistry, 276(38), pp.35368–74. Available at: http://www.ncbi.nlm.nih.gov/pubmed/11451954 [Accessed June 24, 2016]. Tatham, M.H. et al., 2008. RNF4 is a poly-SUMO-specific E3 ubiquitin ligase required for arsenic-induced PML degradation. Nature cell biology, 10(5), pp.538–46. Available at: http://www.nature.com/doifinder/10.1038/ncb1716 [Accessed March 23, 2012]. Tatham, M.H. et al., 2005. Unique binding interactions among Ubc9, SUMO and RanBP2 reveal a mechanism for SUMO paralog selection. Nature Structural & Molecular Biology, 12(1), pp.67–74. Available at: http://www.nature.com/doifinder/10.1038/nsmb878 [Accessed June 24, 2016]. Tchelebi, L., Ashamalla, H. & Graves, P.R., 2014. Mutant p53 and MDM2 in Cancer, Available at: http://www.ncbi.nlm.nih.gov/pubmed/25201193. Terris, B. et al., 1995. PML nuclear bodies are general targets for inflammation and cell proliferation. Cancer research, 55(7), pp.1590–7. Available at: http://www.ncbi.nlm.nih.gov/pubmed/7882370 [Accessed June 6, 2016]. de Thé, H. et al., 1991. The PML-RAR alpha fusion mRNA generated by the t(15;17) translocation in acute promyelocytic leukemia encodes a functionally altered RAR. Cell, 66(4), pp.675–84. Available at: http://linkinghub.elsevier.com/retrieve/pii/009286749190113D [Accessed January 31, 2015]. de Thé, H., Le Bras, M. & Lallemand-Breitenbach, V., 2012. The cell biology of disease: Acute promyelocytic leukemia, arsenic, and PML bodies. The Journal of cell biology, 198(1), pp.11–21. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3392943&tool=pmcentrez&r endertype=abstract [Accessed November 13, 2013]. Trotman, L.C. et al., 2006. Identification of a tumour suppressor network opposing nuclear Akt function. Nature, 441(7092), pp.523–7. Available at: http://www.ncbi.nlm.nih.gov/pubmed/16680151 [Accessed June 7, 2016]. Trotman, L.C. et al., 2007. Ubiquitination regulates PTEN nuclear import and tumor suppression. Cell, 128(1), pp.141–56. Available at: http://www.ncbi.nlm.nih.gov/pubmed/17218261 [Accessed June 7, 2016]. Tsuzuki, S. et al., 2000. Potentiation of GATA-2 Activity through Interactions with the Promyelocytic Leukemia Protein (PML) and the t(15;17)-Generated PML-Retinoic Acid Receptor alpha Oncoprotein. Molecular and Cellular Biology, 20(17), pp.6276–6286. Available at: http://mcb.asm.org/cgi/doi/10.1128/MCB.20.17.6276-6286.2000 [Accessed June 8, 2016]. Uhlén, M. et al., 2015. Proteomics. Tissue-based map of the human proteome. Science (New York, N.Y.), 347(6220), p.1260419. Available at: http://www.ncbi.nlm.nih.gov/pubmed/25613900 [Accessed July 26, 2016]. Ulbricht, T. et al., 2012. PML promotes MHC class II gene expression by stabilizing the class II transactivator. The Journal of Cell Biology, 199(1), pp.49–63. Available at: http://www.jcb.org/lookup/doi/10.1083/jcb.201112015 [Accessed June 8, 2016]. Ullmann, R. et al., 2012. An acetylation switch regulates SUMO-dependent protein interaction networks. Molecular cell, 46(6), pp.759–70. Available at: http://linkinghub.elsevier.com/retrieve/pii/S1097276512003024 [Accessed January 2, 2014].

189

Vallian, S., Gäken, J.A., et al., 1998. Modulation of Fos-mediated AP-1 transcription by the promyelocytic leukemia protein. Oncogene, 16(22), pp.2843–2853. Available at: http://www.nature.com/doifinder/10.1038/sj.onc.1201837 [Accessed June 8, 2016]. Vallian, S., Chin, K.-V. & Chang, K.-S., 1998. The Promyelocytic Leukemia Protein Interacts with Sp1 and Inhibits Its Transactivation of the Epidermal Growth Factor Receptor Promoter. Molecular and Cellular Biology, 18(12), pp.7147–7156. Available at: http://mcb.asm.org/lookup/doi/10.1128/MCB.18.12.7147 [Accessed June 8, 2016]. Vernier, M. et al., 2011. Regulation of E2Fs and senescence by PML nuclear bodies. Genes and Development, 25(1), pp.41–50. Available at: http://www.ncbi.nlm.nih.gov/pubmed/21205865 [Accessed June 7, 2016]. Vertegaal, A.C.O. et al., 2006. Distinct and overlapping sets of SUMO-1 and SUMO-2 target proteins revealed by quantitative proteomics. Molecular & cellular proteomics : MCP, 5(12), pp.2298–310. Available at: http://www.ncbi.nlm.nih.gov/pubmed/17000644 [Accessed April 9, 2012]. Vogel, C. & Marcotte, E.M., 2012. Insights into the regulation of protein abundance from proteomic and transcriptomic analyses. Nature Reviews Genetics, 13(4), pp.227–232. Available at: http://dx.doi.org/10.1038/nrg3185 [Accessed July 10, 2014]. Walsh, C.T., Garneau-Tsodikova, S. & Gatto, G.J., 2005. Protein posttranslational modifications: the chemistry of proteome diversifications. Angewandte Chemie (International ed. in English), 44(45), pp.7342–72. Available at: http://www.ncbi.nlm.nih.gov/pubmed/16267872 [Accessed June 23, 2016]. Wang, K. et al., 2010. PML/RAR?? Targets Promoter Regions Containing PU.1 Consensus and RARE Half Sites in Acute Promyelocytic Leukemia. Cancer Cell, 17(2), pp.186– 197. Available at: http://dx.doi.org/10.1016/j.ccr.2009.12.045. Wang, L. et al., 2014. SUMO2 is essential while SUMO3 is dispensable for mouse embryonic development. EMBO reports, 15(8), pp.878–85. Available at: http://www.ncbi.nlm.nih.gov/pubmed/24891386 [Accessed July 19, 2016]. Wang, S. et al., 2016. Atlas on substrate recognition subunits of CRL2 E3 ligases. Oncotarget, 7(29), pp.46707–46716. Wang, S., Dougherty, E.J. & Danner, R.L., 2016. PPARγ signaling and emerging opportunities for improved therapeutics. Pharmacological Research, 111, pp.76–85. Wang, Z. et al., 2014. Roles of F-box proteins in cancer. Nature reviews. Cancer, 14(4), pp.233–47. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=4306233&tool=pmcentrez&r endertype=abstract [Accessed June 5, 2016]. Wang, Z.G., Ruggero, D., et al., 1998. PML is essential for multiple apoptotic pathways. Nature genetics, 20(3), pp.266–72. Available at: http://www.ncbi.nlm.nih.gov/pubmed/9806545 [Accessed June 6, 2016]. Wang, Z.G., Delva, L., et al., 1998. Role of PML in cell growth and the retinoic acid pathway. Science (New York, N.Y.), 279(5356), pp.1547–51. Available at: http://www.sciencemag.org/cgi/doi/10.1126/science.279.5356.1547 [Accessed June 2, 2016]. Watanabe, N. & Osada, H., 2016. Small molecules that target phosphorylation dependent protein-protein interaction. Bioorganic & Medicinal Chemistry. Available at: http://linkinghub.elsevier.com/retrieve/pii/S0968089616301717. Wei, D. & Sun, Y., 2010. Small RING Finger Proteins RBX1 and RBX2 of SCF E3 Ubiquitin Ligases: The Role in Cancer and as Cancer Targets. Genes & cancer, 1(7), pp.700–7. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=2983490&tool=pmcentrez&r endertype=abstract [Accessed May 15, 2012].

190

Weidtkamp-Peters, S. et al., 2008. Dynamics of component exchange at PML nuclear bodies. Journal of cell science, 121(Pt 16), pp.2731–43. Available at: http://www.ncbi.nlm.nih.gov/pubmed/18664490 [Accessed May 4, 2016]. Welcker, M. & Clurman, B.E., 2008. FBW7 ubiquitin ligase: a tumour suppressor at the crossroads of cell division, growth and differentiation. Nature reviews. Cancer, 8(2), pp.83–93. Available at: http://www.ncbi.nlm.nih.gov/pubmed/18094723 [Accessed April 25, 2016]. Wenger, B. et al., 2014. PML-nuclear bodies decrease with age and their stress response is impaired in aged individuals. BMC geriatrics, 14(1), p.42. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3992156&tool=pmcentrez&r endertype=abstract. Wickham, H., 2007. Reshaping Data with the reshape Package. Journal of Statistical Software, 21(12), pp.413–420. Available at: http://www.jstatsoft.org/v21/i12. Wikipedia, hematopoesis fig. Wikipedia: The Free Encyclopedia. Wikimedia Foundation Inc. Available at: https://en.wikipedia.org/wiki/Haematopoiesis [Accessed June 27, 2016]. Willems, A.R., Schwab, M. & Tyers, M., 2004. A hitchhiker’s guide to the cullin ubiquitin ligases: SCF and its kin. Biochimica et biophysica acta, 1695(1-3), pp.133–70. Available at: http://www.ncbi.nlm.nih.gov/pubmed/15571813 [Accessed March 23, 2012]. Wolyniec, K. et al., 2012. E6AP ubiquitin ligase regulates PML-induced senescence in Myc- driven lymphomagenesis. Blood, 120(4), pp.822–32. Available at: http://www.ncbi.nlm.nih.gov/pubmed/22689861 [Accessed June 25, 2016]. Wright, K., 2015. corrgram: Plot a Correlogram. R package version 1.8. Available at: https://cran.r-project.org/package=corrgram. Wu, H.-C. et al., 2014. USP11 regulates PML stability to control Notch-induced malignancy in brain tumours. Nature communications, 5, p.3214. Available at: http://www.ncbi.nlm.nih.gov/pubmed/24487962 [Accessed February 5, 2014]. Wu, J. et al., 2014. PML4 facilitates erythroid differentiation by enhancing the transcriptional activity of GATA-1. Blood, 123(2), pp.261–70. Available at: http://www.ncbi.nlm.nih.gov/pubmed/24255919 [Accessed June 8, 2016]. Wu, Q. et al., 2009. PML3 Orchestrates the Nuclear Dynamics and Function of TIP60. The Journal of biological chemistry, 284(13), pp.8747–59. Available at: http://www.ncbi.nlm.nih.gov/pubmed/19150978 [Accessed June 21, 2016]. Wu, S. et al., 2013. CAND1 controls in vivo dynamics of the cullin 1-RING ubiquitin ligase repertoire. Nature communications, 4, p.1642. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3637025&tool=pmcentrez&r endertype=abstract. Wu, W.-S., Xu, Z.-X., Ran, R., et al., 2002. Promyelocytic leukemia protein PML inhibits Nur77-mediated transcription through specific functional interactions. Oncogene, 21(24), pp.3925–3933. Available at: http://www.nature.com/doifinder/10.1038/sj.onc.1205491 [Accessed June 8, 2016]. Wu, W.-S. et al., 2003. Promyelocytic Leukemia Protein Sensitizes Tumor Necrosis Factor - Induced Apoptosis by Inhibiting the NF- B Survival Pathway. Journal of Biological Chemistry, 278(14), pp.12294–12304. Available at: http://www.jbc.org/cgi/doi/10.1074/jbc.M211849200 [Accessed June 8, 2016]. Wu, W.-S., Xu, Z.-X. & Chang, K.-S., 2002. The promyelocytic leukemia protein represses A20-mediated transcription. The Journal of biological chemistry, 277(35), pp.31734–9. Available at: http://www.ncbi.nlm.nih.gov/pubmed/12080044 [Accessed June 8, 2016]. Xu, Z.-X. et al., 2005. A role for PML3 in centrosome duplication and genome stability. Molecular cell, 17(5), pp.721–32. Available at: http://www.ncbi.nlm.nih.gov/pubmed/15749021 [Accessed June 21, 2016].

191

Xu, Z.-X. et al., 2003. PML colocalizes with and stabilizes the DNA damage response protein TopBP1. Molecular and cellular biology, 23(12), pp.4247–56. Available at: http://www.ncbi.nlm.nih.gov/pubmed/12773567 [Accessed June 27, 2016]. Yang, Q. et al., 2013. BMK1 is involved in the regulation of p53 through disrupting the PML- MDM2 interaction. Oncogene, 32(26), pp.3156–64. Available at: http://www.ncbi.nlm.nih.gov/pubmed/22869143 [Accessed November 14, 2013]. Yang, Q. et al., 2010. Pharmacological inhibition of BMK1 suppresses tumor growth through promyelocytic leukemia protein. Cancer cell, 18(3), pp.258–67. Available at: http://www.ncbi.nlm.nih.gov/pubmed/20832753 [Accessed June 23, 2016]. Yang, S. et al., 2002. PML-dependent apoptosis after DNA damage is regulated by the checkpoint kinase hCds1/Chk2. Nature cell biology, 4(11), pp.865–70. Available at: http://www.ncbi.nlm.nih.gov/pubmed/12402044 [Accessed June 1, 2016]. Yang, S. et al., 2006. Promyelocytic leukemia activates Chk2 by mediating Chk2 autophosphorylation. The Journal of biological chemistry, 281(36), pp.26645–54. Available at: http://www.ncbi.nlm.nih.gov/pubmed/16835227 [Accessed June 23, 2016]. Yang, S.-H. et al., 2006. An extended consensus motif enhances the specificity of substrate modification by SUMO. The EMBO journal, 25(21), pp.5083–93. Available at: http://www.ncbi.nlm.nih.gov/pubmed/17036045 [Accessed June 24, 2016]. Yang, X.-J. & Grégoire, S., 2006. A recurrent phospho-sumoyl switch in transcriptional repression and beyond. Molecular cell, 23(6), pp.779–86. Available at: http://www.ncbi.nlm.nih.gov/pubmed/16973431 [Accessed June 24, 2016]. Yen, J.L. et al., 2012. Signal-induced disassembly of the SCF ubiquitin ligase complex by Cdc48/p97. Molecular cell, 48(2), pp.288–97. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3483439&tool=pmcentrez&r endertype=abstract [Accessed June 3, 2016]. Yin, Y. et al., 2012. SUMO-targeted ubiquitin E3 ligase RNF4 is required for the response of human cells to DNA damage. Genes & development, 26(11), pp.1196–208. Available at: http://www.ncbi.nlm.nih.gov/pubmed/22661230 [Accessed June 5, 2012]. Yoshida, Y., Murakami, A. & Tanaka, K., 2011. Skp1 stabilizes the conformation of F-box proteins. Biochemical and biophysical research communications, 410(1), pp.24–8. Available at: http://www.ncbi.nlm.nih.gov/pubmed/21640084 [Accessed March 15, 2012]. Yuan, W.-C. et al., 2011. A Cullin3-KLHL20 Ubiquitin ligase-dependent pathway targets PML to potentiate HIF-1 signaling and prostate cancer progression. Cancer cell, 20(2), pp.214–28. Available at: http://www.ncbi.nlm.nih.gov/pubmed/21840486 [Accessed March 1, 2012]. Zhang, P. et al., 2000. Lack of expression for the suppressor PML in human small cell lung carcinoma. International journal of cancer, 85(5), pp.599–605. Available at: http://www.ncbi.nlm.nih.gov/pubmed/10699936 [Accessed June 6, 2016]. Zhang, W. & Koepp, D.M., 2006. Fbw7 isoform interaction contributes to cyclin E proteolysis. Molecular cancer research : MCR, 4(12), pp.935–43. Available at: http://www.ncbi.nlm.nih.gov/pubmed/17189384 [Accessed June 26, 2016]. Zhang, X., 2014. Analysis of Changes in SUMO-2/3 Modi fi cation during Breast Cancer Progression and Metastasis. Zhang, X.-W. et al., 2010. Arsenic trioxide controls the fate of the PML-RARalpha oncoprotein by directly binding PML. Science (New York, N.Y.), 328(5975), pp.240–3. Available at: http://www.ncbi.nlm.nih.gov/pubmed/20378816 [Accessed April 12, 2012]. Zhang, Y. et al., 2016. MLN4924 suppresses neddylation and induces cell cycle arrest, senescence, and apoptosis in human osteosarcoma. Oncotarget, 5(0).

192

Zhao, Q. et al., 2014. GPS-SUMO: a tool for the prediction of sumoylation sites and SUMO- interaction motifs. Nucleic acids research, 42(Web Server issue), pp.W325–30. Available at: http://www.ncbi.nlm.nih.gov/pubmed/24880689 [Accessed June 15, 2016]. Zhong, S. et al., 1999. A role for PML and the nuclear body in genomic stability. Oncogene, 18(56), pp.7941–7. Available at: http://www.ncbi.nlm.nih.gov/pubmed/10637504 [Accessed June 27, 2016]. Zhong, S. et al., 2000. Role of SUMO-1-modified PML in nuclear body formation. Blood, 95(9), pp.2748–52. Available at: http://bloodjournal.hematologylibrary.org/content/95/9/2748.short [Accessed May 29, 2012]. Zhou, W. & Bao, S., 2014. PML-mediated signaling and its role in cancer stem cells. Oncogene, 33(12), pp.1475–84. Available at: http://www.ncbi.nlm.nih.gov/pubmed/23563177.

193

Table and illustration table

Figure 1 : Hematopoietic differentiation from hematopoietic stem cells to mature cells...... 12 Figure 2 : PML-RARα fusion protein comes from the t(15;17) (q22;q21) translocation...... 13 Figure 3 : Retinoic Acid Receptor α and PML-RAR α function in normal and APL cells. ... 14 Figure 4 : PML protein is conserved in mammals...... 15 Figure 5 : PML protein isoforms and nomenclature generated from PML gene alternative splicing...... 16 Figure 6 : PML protein domain and structure...... 17 Figure 7 : PML Nuclear Bodies...... 20 Figure 8 : PML Nuclear Body formation requires SUMO modification...... 20 Figure 9 : PML Nuclear Body biogenesis model...... 21 Figure 10 : PML interactome based on data from BIOGRID...... 22 Figure 11 : Protein diversity explained, from Genome to Proteome...... 23 Figure 12 : Five major types of covalent modification...... 24 Figure 13 : Five residues phosphorylated in mammals...... 26 Figure 14 : Known PML phosphorylation sites...... 26 Figure 15 : Known site-specific kinases for PML phosphorylation and associated function. 29 Figure 16 : Ubiquitin and SUMO three dimensional structure comparison...... 31 Figure 17 : Mechanism of reversible SUMOylation...... 33 Figure 18 : Known SUMOylation sites and E3 ligases of PML...... 38 Figure 19 : PML is hyperSUMOylated and degraded under arsenic stress...... 39 Figure 20 : Arsenic Trioxide (ATO) and All-Trans Retinoic Acid (ATRA) effects in the cure of APL...... 40 Figure 21 : PML degradation key steps events under Arsenic trioxide (AS2O3) induced stress...... 41 Figure 22 : Ubiquitination enzymatic cascade leading to substrate degradation...... 42 Figure 23 : Structure of Ubiquitin...... 43 Figure 24 : The Ubiquitin code...... 44 Figure 25 : Proteasomal degradation based on the Ubiquitin threshold model...... 46 Figure 26 : Summary of human PML post-translational modifications...... 48 Figure 27 : The Cullin RING Ligase (CRL) family ubiquitination complex...... 50 Figure 28 : Mammalian F-Box protein structural domains...... 51 Figure 29 : Known F-Box substrates and biological implications...... 52 Figure 30 : Regulation of Cullin Ring Ligases (CRLs) through NEDDylation...... 54 Figure 31 : Regulation of Cullin Ring Ligases (CRLs) through CAND1 substrate receptor exchange...... 56 Figure 32 : Various mechanisms used to regulate F-Box protein substrate recognition...... 58 Figure 33 : Protein degradation can be dysregulated in diseases due to F-Box activity alterations...... 60 Figure 34 : Examples of F-Box proteins involvement in cancer...... 62 Figure 35 : KLHL20-CUL3-ROC1 is targeting PML for degradation under hypoxic stress. 64

194

Figure 36 : PML functions in diseases...... 65 Figure 37 : Tumor suppressor pathways involving PML Nuclear Bodies...... 67 Figure 38 : Effect of Arsenic trioxide treatment on PML and PML Nuclear Bodies...... 76 Figure 39 : Methodology and controls for the genome wide siRNA screen...... 78 Figure 40 : Phenotypic classes used to regroup candidates in the screen...... 79 Figure 41 : Results obtained from the second validation screen...... 81 Figure 42 : SKP1-CUL1-F-Box containing complex (SCF)...... 82 Figure 43 : SKP1a and RBX1 depletion effect on PML Nuclear Bodies morphology in HT1080 GFP-PML IV cells...... 83 Figure 44 : SKP1a and RBX1 are involved in PML stability in presence or not of arsenic trioxide treatment...... 85 Figure 45 : Validation of SKP1a and RBX1 involvement in PML stability through single siRNA experiment...... 86 Figure 46 : SKP1a and RBX1 are involved in PML stability: gain of function experiment. . 87 Figure 47 : SKP1a and RBX1 interaction with PML...... 88 Figure 48 : Immunoprecipitation screen for F-Box proteins specifically recognizing PML. . 89 Figure 49 : Knowledge summary on FBXO9 protein post-translational modifications and domains...... 90 Figure 50 : FBXO9 interacts with PML...... 91 Figure 51 : FBXO9 interacts with different PML isoforms...... 92 Figure 52 : FBXO9 is co-localized with PML VII and interacts with PML IV both in the nucleus and in the cytoplasm...... 94 Figure 53 : FBXO9 interaction with PML is not dependent on arsenic trioxide treatment or SUMOylation...... 96 Figure 54 : FBXO9 depletion has an effect on PML stability: siRNA approach...... 99 Figure 55 : FBXO9 depletion has an effect on PML stability: shRNA approach...... 100 Figure 56 : FBXO9 and CAND1 depletion have an effect on PML stability...... 101 Figure 57 : FBXO9 and RNF4 depletion effect on PML Nuclear Bodies without arsenic trioxide treatment: fluorescence microscopy approach...... 103 Figure 58 : FBXO9 and RNF4 depletion effect on PML Nuclear Bodies after one hour arsenic trioxide treatment: fluorescence microscopy approach...... 104 Figure 59 : FBXO9 and RNF4 depletion effect on PML Nuclear Bodies after 24 hours arsenic trioxide treatment: fluorescence microscopy approach...... 105 Figure 60 : FBXO9 shortens the half-life of PML...... 107 Figure 61 : SCFFBXO9 is able to specifically ubiquitinate PML: ubiquitination assays approach...... 109 Figure 62 : PML post-translational modification map...... 110 Figure 63 : FBXO9 degron identification on PML: immunoprecipitation screen approach. 112 Figure 64 : Mini kinase screen to identify FBXO9 degron on PML...... 113 Figure 65 : Test of different types of stress on PML stability in the presence of CK2 inhibitor...... 114 Figure 66 : Osmotic shock causes PML degradation that is dependent at least in part on SCFFBXO9...... 115 Figure 67 : Heat map of mRNA expression of 21 genes in 72 different cell lines...... 117

195

Figure 68 : Correlation matrix of target gene expression...... 118 Figure 69 : Hierarchical clustering of studied cell lines based on gene expression...... 119 Figure 70 : Diagram of the SKP1-Cullin-F-Box (SCF) containing complex ...... 122 Figure 71 : Localization of PML through cell cycle...... 129 Figure 72 : CK2 phosphorylation of PML leads to ubiquitination and subsequent degradation through an unknown process...... 130 Figure 73 : SCF complexes can be targeted by drugs to treat diseases...... 136 Figure 74 : Manipulation of SCF ubiquitin ligase activity trough different strategies...... 138

Table 1 : Some examples of post-translational modifications...... 25 Table 2 : Known PML SUMOylation sites ...... 35 Table 3 : Example of F-Box substrate and known degron regulation...... 59 Table 4 : Drug therapy development targeting E3 Ubiquitin Ligases...... 63

Annex 1: The 69 Mammalian F-Box proteins………………………………………………150 Annex 2: F-Box protein and E3 ubiquitin ligase implication in cellular pathways………....152 Annex 3: Validated candidates inducing a morphological change of PML Nuclear Bodies..153 Annex 4: Co-immunoprecipitation screen to identify PML interacting F-Box protein…...... 165 Annex 5: Cell lines used in mRNA mini screen……………………………………………..166 Annex 6: FBXO9 is overexpressed in some types of breast cancers…………………………169 Annex 7: Cancer tissue PML antibody staining of Breast and lung cancers…………………170

196 Snollaerts Thibaut – Thèse de doctorat -2016

Abstract: ProMyelocytic Leukemia (PML) protein is implicated in a number of key cellular processes, and was identified as a tumor suppressor. This protein is one of the main structural components forming the PML Nuclear Bodies (PML-NBs) whose integrity -compromised in some leukemias- is strictly dependent on PML SUMOylation. SUMOylation of protein, of which PML is one of the major substrates, is a post-translational modification playing a central role in regulation mechanisms involved in numerous important cellular processes and notably, oncogenesis. The goal of this thesis project was to identify new regulators of PML Nuclear Bodies, and by extension of the SUMO pathway, using PML-NBs, which are extremely sensitive to global cellular SUMOylation level, as a read out. This work is based on a high throughput siRNA screen, which led to the identification of 20 candidates, of which two are part of an Ubiquitin E3 ligase described in details throughout the project. The two proteins, SKP1a and RBX1, are both part of a complex called SKP-Cullin-F-Box containing complex (SCF). SKP1 is the adaptor protein of the complex while RBX1 is the protein recruiting the conjugation enzyme necessary for ubiquitination, another post-translational modification that can lead to protein degradation. We were able to show the involvement of SKP1 and RBX1 in PML protein stability through gain and loss of function experiments. We also identified FBXO9 as the F-Box capable of specifically recognizing PML, causing its ubiquitination and subsequent degradation by the proteasome. The newly discovered SCFFBXO9 complex is capable of specifically ubiquitinate PML in-vitro. However, FBXO9 site of interaction on PML and the identity of the kinase implicated in this recognition processes are yet to be discovered. PML being degraded in numerous cancers such as breast cancer, it is essential to acquire a better understanding of post-translational mechanism leading to the degradation of this tumour suppressor. In the long term, this work should, allow the discovery of new PML Nuclear Body regulators in addition to SCFFBXO9, and potentially allow the development of new strategies aiming to modulate PML Nuclear Bodies in tumoral cells.

Keywords : PML – SUMOylation – Ubiquitin - Post-Translational Modification - Therapeutic targets

Identification de nouveaux régulateurs des Corps Nucléaires PML Résumé : La protéine Promyelocytic Leukemia (PML) est impliquée dans de nombreux processus cellulaires, et identifiée comme un suppresseur de tumeur. Cette protéine est le composant structural des Corps Nucléaires PML (CNs-PML) dont l’intégrité, compromise dans certaines leucémies, dépend strictement de sa SUMOylation. La SUMOylation des protéines dont PML est l’un des substrats majeurs, est un mécanisme central de régulation agissant sur de nombreux processus cellulaires, et joue notamment un rôle important dans certains processus oncogéniques. Ce projet de thèse visait à identifier de nouveaux régulateurs des CNs-PML, et par extension de la voie SUMO, en utilisant comme ‘read-out’ l’anatomie des CNs-PML, laquelle est extrêmement sensible au niveau de SUMOylation cellulaire globale. Ces travaux sont basés sur un criblage siARNs à grande échelle qui a conduit à l'identification de 20 candidats, dont deux font partie intégrante d’un complexe d’ubiquitine E3 ligase décrit dans ces travaux de thèse. Il s’agit de SKP1et RBX1, tous deux faisant partie d’un complexe appelé « SKP1-CUL1-F- Box containing complex » (SCF). SKP1 est la protéine adaptatrice de ce complexe, et RBX1 la protéine recrutant l’enzyme de conjugaison nécessaire à l’ubiquitination, Nous avons pu démontrer l’implication de SKP1 et RBX1 dans la stabilité de la protéine PML avec des expériences de gain et perte de fonction. Nous avons également identifié FBXO9 comme la protéine F-Box capable de reconnaitre spécifiquement PML, causant son ubiquitination suivie de sa dégradation par le protéasome. Le complexe SCFFBXO9, ainsi formé, est capable d’ubiquitiner spécifiquement PML in-vitro. En revanche, le site d’interaction de FBXO9 sur PML -tout comme la kinase impliquée dans ce processus de reconnaissance- restent encore à identifier. PML étant dégradée dans de nombreux cancers, comme le cancer du sein, il apparait essentiel d’avoir une meilleure compréhension des mécanismes post-traductionnels menant à la dégradation de ce suppresseur de tumeur. Ces travaux devraient à long terme permettre de révéler de nouveaux régulateurs des CNs-PML, en plus de SCFFBXO9, et potentiellement permettre le développement de nouvelles stratégies thérapeutiques, visant à moduler les CNs-PML dans la cellule tumorale.

Mots clés : PML – SUMOylation – Ubiquitine - Modification Post-traductionnelles - Cibles thérapeutiques.